_______________________________________________________________

_______________________________________________________________

Report Information from ProQuest08 March 2013 19:36_______________________________________________________________

Table of Contents

1. Abnormal Audit Fee and Audit Quality......................................................................................................... 1

2. AN EMPIRICAL INVESTIGATION OF THE EFFECTS OF MERGER AND ACQUISITION ON FIRMS'PROFITABILITY.............................................................................................................................................. 14

3. Does reduced trade tax revenue affect government spending patterns?..................................................... 27

4. THE DECLINING USE OF INTERNAL SERVICE FUNDS: HOW LOCAL GOVERNMENTS ARECHANGING THE ALLOCATION OF INDIRECT COSTS................................................................................ 43

5. International differences in IFRS policy choice: a research note.................................................................. 49

6. CITIZEN PARTICIPATION IN THE BUDGET PROCESS: THE EFFECT OF CITY MANAGERS............... 62

7. Auditor Disaffiliation Program in China and Auditor Independence............................................................. 75

8. Corporate governance, firm characteristics and risk management committee formation in Australiancompanies....................................................................................................................................................... 94

9. A Methodology for Evaluating the Cost-Effectiveness of Alternative Management Tools in Public-SectorInstitutions: An Application to Public Education............................................................................................... 109

10. The provision of non-audit services and earnings conservatism: Do New Zealand auditors compromisetheir independence?........................................................................................................................................ 130

08 March 2013 ii ProQuest

Document 1 of 10 Abnormal Audit Fee and Audit Quality Author: Asthana, Sharad C; Boone, Jeff P Publication info: Auditing 31. 3 (Aug 2012): 1-22.ProQuest document link Abstract: This study tests the hypotheses that below-normal audit fees signal important nuances in the balanceof bargaining power between the auditor and the client, and that such power may ultimately influence auditquality. We find that audit quality, proxied by absolute discretionary accruals and meeting or beating analysts'earnings forecasts, declines as negative abnormal audit fees increase in magnitude, with the effect amplified asproxies for client bargaining power increase. We find that this effect is dampened in years following theSarbanes-Oxley Act (SOX), suggesting that SOX was effective in enhancing auditor independence.[PUBLICATION ABSTRACT] Full Text: Headnote SUMMARY: This study tests the hypotheses that below-normal audit fees signal importantnuances in the balance of bargaining power between the auditor and the client, and that such power mayultimately influence audit quality. We find that audit quality, proxied by absolute discretionary accruals andmeeting or beating analysts' earnings forecasts, declines as negative abnormal audit fees increase inmagnitude, with the effect amplified as proxies for client bargaining power increase. We find that this effect isdampened in years following the Sarbanes-Oxley Act (SOX), suggesting that SOX was effective in enhancingauditor independence. Keywords: abnormal audit fees; bargaining power; economic bonding. (ProQuest: ...denotes formulae omitted.) INTRODUCTION Understanding the factors that lead auditors to compromise onaudit quality is an important issue of concern to scholars, investors, and regulators. One possible factor that hasreceived significant research attention is economic bonding between the auditor and client. The basic idea inthese studies is that positive abnormal audit fees reflect the extent of economic bonding between the auditorand client, and greater economic bonding degrades audit quality by impairing auditor independence. Based onthis premise, studies have examined the linkage between audit quality and positive abnormal audit fees, withsuch studies documenting a negative association. In this study, we reexamine the association betweenabnormal audit fees and audit quality, and we do so in a way that allows our paper to offer two importantcontributions to the literature. First, we incorporate into our conceptual framework insights about clientbargaining power (Casterella et al. 2004) in addition to the economic bonding story that forms the conceptualframework in prior studies. Expanding the framework to consider both bargaining power and economic bondingallows us to offer a novel prediction that would not be meaningful within a framework based exclusively oneconomic bonding. That prediction is: audit quality will decline as negative abnormal audit fees increase inmagnitude, and the magnitude of the decline will increase as proxies for client bargaining power increase. Theeconomic bonding literature that informs prior studies places the focus on positive abnormal audit fees,predicting no association between negative abnormal audit fees and audit quality, and a negative associationbetween positive abnormal audit fees and audit quality. Second, we include in our analysis data taken from thepost-Sarbanes-Oxley (SOX) period (i.e., years 2004-2009), in addition to data from the pre-SOX period (i.e.,years 2000-2003). The pre-SOX period has been the data source for virtually all of the prior studies. With datataken from both the pre- and post-SOX periods, we are able to probe for a dampened association betweenabnormal audit fees and audit quality that would be manifest if SOX reforms meaningfully increased auditorindependence or strengthened the auditor's bargaining power, leading to higher audit quality. To date, there isonly limited empirical evidence that speaks to the issue of whether SOX reforms increased audit quality.Consistent with prior research (e.g., Choi et al. 2010; Hope et al. 2009; and Higgs and Skantz 2006), wedecompose total audit fees into normal and abnormal components, and test for an association between audit

08 March 2013 Page 1 of 145 ProQuest

quality and abnormal audit fees, conditioning our tests on the sign of the abnormal audit fee metric (i.e., above-normal audit fees and below-normal audit fees). Under the bargaining power story, we expect to find that auditquality declines as negative abnormal fees increase in magnitude, with the effect amplified as proxies for clientbargaining power increase. Under the economic bonding story (and consistent with prior studies), we expect tofind that audit quality declines as positive abnormal fees increase in magnitude. We also partition our analysisby regulatory regime (i.e., a pre-SOX reporting period or a post-SOX reporting period) in order to assesswhether the sensitivity of audit quality to abnormal audit fees differs between these two regimes. Our twoproxies for audit quality are absolute discretionary accruals and meeting or beating analysts' earnings forecasts,and our two proxies for client bargaining power reflect the importance of the client to the local practice office.Our tests produce evidence consistent with both the economic bonding story and the bargaining power story. Aspredicted by the economic bonding story, we find that absolute discretionary accruals and the probability ofmeeting or beating earnings forecasts both increase as positive abnormal audit fees increase. As predicted bythe client bargaining power story, we find that absolute discretionary accruals and the probability of meeting orbeating earnings forecasts increase with the magnitude of negative abnormal audit fees, with the effectamplified as client bargaining power increases. With respect to the effects of SOX, we find that the effects ofeconomic bonding and client bargaining power are both dampened in the post-SOX regime, suggesting thatSOX was effective in enhancing auditor independence. The evidence presented in our paper is important in atleast two respects. First, our results suggest that investors, regulators, and others interested in assessing theeffects of auditor remuneration on audit quality should be concerned with both above-normal and below-normalauditor remuneration, but for different reasons. The potential effect of above-normal audit fees in degradingaudit quality by economically bonding the auditor with the client is well recognized and extensively investigated.Less recognized is the possibility that below-normal audit fees signal important nuances in the balance of powerbetween the auditor and the client, and that such power may ultimately influence audit quality. Thus, our studyhighlights the importance of considering the bargaining power of the client when assessing audit quality.Second, our study presents the first evidence (of which we are aware) to suggest that reforms introduced by theSarbanes-Oxley Act dampened the deleterious effects of economic bonding and client power on audit quality,and hence increased audit quality. Both of these important insights from our study increase understanding ofthe factors that may lead auditors to compromise on audit quality, and hence our paper should be of interest toaccounting scholars, investors, and regulators. The remainder of the study is organized as follows. The secondsection develops our theoretical framework and presents the hypotheses that we test. We present our researchdesign in the third section, followed by a discussion of our sample in the fourth section and our results in the fifthsection. We offer concluding comments in the sixth section. THEORETICAL FRAMEWORK ANDHYPOTHESES An audit firm is not a single person ''auditor''; rather, it is a decentralized organization in whichindividual audit partners act as agents for the audit firm (Liu and Simunic 2005). To the extent the partnershipprofit-sharing plan does not effectively align the interests of the partner with that of the audit firm partners as awhole, an uncontrolled moral hazard problem exists that might result in an individual audit partner succumbingto client pressure for earnings management. This is because the individual partner captures a significant portionof the expected benefit from acquiescing to client demands, while passing to the partnership as a whole theexpected cost (Trompeter 1994). What factors might lead an audit partner to compromise on audit quality? Asdescribed below, engagement profitability and client bargaining power are two possible explanations.Engagement Profitability and Economic Bonding Engagement profitability should influence audit quality for thefollowing reason. Audit startup costs and client switching costs allow the auditor to price audit services at a pricein excess of the avoidable cost of producing the audit, and thus create for the incumbent auditor clientspecificquasi-rents (DeAngelo 1981a, 1981b). The client-specific quasi-rents economically bond the auditor to theclient, reduce auditor independence, and increase the likelihood that the auditor will acquiescence to a client'sdemand for earnings management. However, succumbing to client pressure risks audit firm forfeiture of some or

08 March 2013 Page 2 of 145 ProQuest

all of the quasi-rents from the firm's entire client portfolio (if the earnings management is discovered), andadditional economic loss through litigation and government penalties (DeAngelo 1981a, 1981b). The auditor willcompromise audit integrity only if the expected gain (preserving the client-specific quasi-rent) exceeds theexpected loss (forfeited quasi-rents from the overall client portfolio, litigation costs, and penalties), and thus thequestion of whether economic bonding undermines audit quality depends upon the relative magnitude ofexpected costs and benefits, which is an empirical question.1 Bargaining Power Bargaining power shouldinfluence audit quality for the following reason. Audited financial statements, and hence audit quality, are thejoint effort of the auditor and the client (Antle and Nalebuff1991) that arise from a process of negotiationbetween the two (Gibbins et al. 2001). The negotiation literature shows that when negotiators differ inbargaining power, the more powerful party expects greater concessions (e.g., Pruitt and Carnevale 1993;Hornstein 1965; Michener et al. 1975), and Barnes (2004) shows that audit quality may decrease as clientbargaining power increases. In an experimental auditing setting, Hatfield et al. (2008) show that the effect ofclient bargaining power on the audited financial statements depends upon the negotiation strategy employed bythe auditor, with a reciprocity negotiating strategy leading to more conservative financial statements. Thus, thequestion of whether client bargaining power undermines audit quality depends upon whether the auditor is ableto employ a negotiating strategy that weakens the advantage held by a client with strong bargaining power.Abnormal Audit Fees Simunic (1980) shows that the auditor's expected fee charged to the client is driven by theunits of audit resources expended, the per-unit cost of those resources, and the auditor's expected future lossarising from the engagement (e.g., litigation losses, government penalties). Empirically, extant research modelsthe expected audit fee as a function of observable factors that proxy for the auditor's cost in performing theaudit, including auditor effort (i.e., resources expended and their cost), expected future litigation losses, andnormal profit. If the audit fee model is well specified, the residual audit fee reflects abnormal profits from theaudit engagement. To the extent that some factors are unobservable (and hence omitted from the audit feemodel), the residual audit fee metric measures abnormal audit profitability with error. Abnormal audit profitabilityshould be associated with both client bargaining power and economic bonding. With respect to the former,Casterella et al. (2004) show a negative association between proxies for client bargaining power and audit feesearned by industry specialists. Their research suggests that, ceteris paribus, below-normal audit fees mayreflect billing concessions granted by the auditor due to client bargaining power. With respect to the later,Kinney and Libby (2002) note that ''Unexpected fees may also better capture the profitability of the servicesprovided . . . more insidious effects on economic bond may result from unexpected nonaudit and audit fees thatmay more accurately be likened to attempted bribes.'' Although there is scant evidence on the associationbetween abnormally low audit fees and audit quality, a growing literature, described below, examines theassociation between abnormally high audit fees (as a proxy for economic bond) and audit quality. DeFond et al.(2002), Krishnan et al. (2005), Hoitash et al. (2007), and Hribar et al. (2010) test for a linear associationbetween abnormal audit and/or engagement fees and audit quality (i.e., the curve relating audit quality toabnormal audit fees exhibits the same slope for both positive and negative abnormal audit fees).2 DeFond et al.(2002) find no association between abnormal engagement fees and auditors' going concern opinions during2000-2001, while Krishnan et al. (2005) find that during year 2001, earnings response coefficients (a directindicator of audit quality) decline as abnormal engagement fees increase. Hoitash et al. (2007) find during years2000-2007 a positive association between abnormal engagement fees and two (inverse) audit quality metrics-the Dechow and Dichev (2002) accrual quality metric and the absolute value of performance-adjusteddiscretionary accruals. Hribar et al. (2010) find during years 2000 to 2007 a positive association betweenabnormal audit fees and accounting fraud, restatements, and SEC comment letters.3 Larcker and Richardson(2004), Higgs and Skantz (2006), Hope et al. (2009), Mitra et al. (2009), and Choi et al. (2010) test for anonlinear association between abnormal audit fees and audit quality (i.e., the curve relating audit quality toabnormal audit fees exhibits different slope for positive as compared to negative abnormal audit fees). Larcker

08 March 2013 Page 3 of 145 ProQuest

and Richardson (2004) use data from 2000 and 2001 to examine absolute discretionary accruals (an inverseindicator of audit quality), and find that audit quality increases as abnormal engagement fees increase inabsolute magnitude. Higgs and Skantz (2006) find that during 2000-2002, earnings response coefficients (adirect indicator of audit quality) are greater in firms with positive abnormal audit fees. Hope et al. (2009) find thatduring 2000-2003, equity discount rates (an inverse indicator of audit quality) increase as positive abnormalengagement fees increase, but find no association with negative abnormal engagement fees. Mitra et al. (2009)find during years 2000-2005 a negative association between positive abnormal audit fees, and both absolutediscretionary accruals and incomeincreasing accruals, but find no association between negative abnormal auditfees and discretionary accruals. Choi et al. (2010) find during 2000-2003 a positive association between positiveabnormal audit fees and absolute discretionary accruals, but no association when abnormal audit fees arenegative. The weight of the preceding evidence suggests a negative association between audit quality andpositive abnormal audit fees, and no association between audit quality and negative abnormal audit fees-findings consistent with the economic bonding hypothesis. However, as discussed above, negative abnormalaudit fees may reflect client bargaining power that could degrade audit quality-with the degrading being larger inmagnitude the greater the bargaining power of the client. Also as previously discussed, the question of whetherclient bargaining power undermines audit quality ultimately depends upon whether the auditor is able to employa negotiating strategy that weakens the advantage held by a client with strong bargaining power. Thus, whetherclient bargaining power affects audit quality remains an open empirical question, leading us to specify and testthe following client bargaining power hypotheses. H1a: Audit quality will decline as below-normal audit feeincreases in magnitude. H1b: The association predicted in H1a will be amplified as proxies for client bargainingpower increase. For completeness, we also specify and test the following economic bonding hypothesis: H2:Audit quality will decline as above-normal audit fees increase in magnitude. Post-Sox Abnormal Audit Fees Inaddition to the question of client bargaining power, another as yet unanswered question is whether the auditquality/abnormal audit fee association changed following SOX. Passed in 2002 following discovery of a series ofhigh-profile financial reporting scandals, SOX seeks to improve corporate governance and enhance auditorindependence by mandating federal government oversight of auditors, enhancing audit committee auditoroversight, and limiting the opportunity for auditors to sell nonaudit services to clients (U.S. House ofRepresentatives 2002). If these reforms are sufficiently salient, they should manifest in a reduced associationbetween abnormal audit fees and audit quality post-SOX relative to pre-SOX. This leads to our final hypothesis:H3: The association between audit quality and abnormal audit fees will be attenuated in the post-SOX period ascompared to the pre-SOX period. RESEARCH DESIGN To test our hypotheses, we need to measure abnormalaudit fee and audit quality. We estimate abnormal audit fee (ABNAFEE) as the actual audit fee paid by the clientto its auditor minus the predicted (normal) audit fee, with the difference deflated by the total audit fee revenue ofthe audit office conducting the client's audit. We deflate the abnormal audit fee by total audit fee revenues of thepractice office conducting the audit in order to capture the relative profitability of the engagement to the opiningaudit office. We do so since prior research (e.g., Reynolds and Francis 2001) suggests that economic incentivesimpacting audit quality are best measured at the local office level rather than at the national firm level. Forexample, the Enron audit failure largely stemmed from decisions made in the Houston office of Arthur Andersen(Chaney and Philipich 2002). We define two separate variables from ABNAFEE. If ABNAFEE . 0 thenHIABNAFEE1⁄4ABNAFEE, and 0 otherwise. If ABNAFEE , 0 then LOABNAFEE 1⁄4 jABNAFEEj, and 0otherwise. This allows us to study the relationship of the dependent variables with the positive and negativeabnormal audit fee, separately. The predicted audit fee is estimated from an audit fee model based on extantresearch.5 All the variables used in various tests are summarized in Table 1. Audit quality is unobservable.Consistent with prior research, we define audit quality as the client's earnings quality (Higgs and Skantz 2006;Lim and Tan 2008; Davis et al. 2009; Francis and Yu 2009; Reichelt and Wang 2010; Choi et al. 2010).Following this research, we use two commonly used proxies for earnings quality: absolute discretionary

08 March 2013 Page 4 of 145 ProQuest

accruals and propensity to meet or beat earnings expectations. We also conduct additional tests using anearnings response coefficient. The first two proxies are surrogates for actual earnings management, while thelast one is related to perceived earnings quality. However, for the sake of brevity, we only report results for thefirst two, since the conclusions are similar. The tests for these proxies are discussed in more detail below.Discretionary Accruals Model The level of discretionary accruals has commonly been used as a surrogate formanagers' exercise of discretion provide by GAAP. To the extent the discretionary component of accruals isused by managers to opportunistically manage earnings and auditors allow the manipulation to remainuncorrected, discretionary accruals adversely reflect on the audit and earnings quality (Schipper 1989; Jones1991; Levitt 2000; DeFond and Park 2001). Discretionary accruals can be used for increasing or decreasingearnings depending on the incentives of managers. Since we are not looking at any specific managerialincentives, we have no directional predictions for accruals. We therefore use the absolute value of discretionaryaccruals (jDACCj) as the independent variable in our next test. Discretionary accruals (DACC) are calculatedusing the cross-sectional modified version of the Jones model (Jones 1991; Dechow et al. 1995), deflated bytotal assets, and estimated by year and for each industry. We adjust discretionary accruals for performance assuggested by Kothari et al. (2005). Following Hribar and Collins (2002), we use the difference between netincome and cash from operations, deflated by lagged assets, as our measure of total accruals (TACC). Thus:TACC 1⁄4 ðIBC L OANCFÞ=LagðATÞ; where IBC is the income before extraordinary items (Compustat cashflow item), OANCF is net cash flow from operating activities, and AT is total assets. The model to estimatediscretionary accruals is: ... (1) where Lag(AT) is total assets of prior year; DSALE is change in revenue;RECCH is the decrease in accounts receivables; PPEGT is property plant and equipment (gross total); andROA is return on assets, calculated as IBC deflated by AT. Equation (1) is estimated by year for each industry(twodigit SIC code). Then, TACC minus the predicted value from the above regression is our measure ofdiscretionary accruals. Our test of jDACCj is based on the following model: ... (2) The control variables are fromextant research. Francis and Yu (2009) show that larger offices of Big 4 auditors have higher audit quality. Thelogarithm of total office-specific audit fee of all clients in a given year (LOFFICE) is included to capture thiseffect. Reynolds and Francis (2001) provide evidence that auditors report more conservatively for larger clients.Consistent with this research, the variable INFLUENCE, defined as the ratio of a client's total fee relative to thetotal annual fee of the practice office that audits the client, is included as an independent variable. TENURE (1/0dummy variable for audit tenures of three years or less) controls for potential effect of short auditorclientassociation on audit quality (Johnson et al. 2002; Carey and Simnett 2006). Balsam et al. (2003) argue thatindustry expertise increases audit quality. We include USLEADER and CITYLEADER, consistent with Francisand Yu (2009), to control for national-level and city-level auditor industry expertise. USLEADER is an indicatorvariable that is coded 1 if the\ auditor is the national audit fee leader in that industry. CITYLEADER is similarlydefined at the city level. BUSSEG (GEOSEG) is the number of business (geographic) segments reported in theCompustat segment file. LOGMV is the natural logarithm of market value of equity at the end of the fiscal year.This variable controls for any size-related effects. Prior research (Ashbaugh et al. 2003; Butler et al. 2004;Menon and Williams 2004; Geiger and North 2006) finds LOGMV to be negatively associated to discretionaryaccruals. SGROWTH is the annual growth in sales and has been found to be positively related to discretionaryaccruals (Menon and Williams 2004). CFFO is the cash flow from operations deflated by total assets. Previousresearchers (Frankel et al. 2002; Chung and Kallapur 2003) find a negative association between discretionaryaccruals and CFFO. Hribar and Nichols (2007) find that accrual volatility may be related to firm-specificoperating characteristics as measured by the volatility of the firm's cash flows and sales. Hence, we includeSDSALES and SDCFFO (calculated as the standard deviations over the current and past four years of cashflow from operations and sales, respectively, deflated by the total assets) as control variables. Doyle et al.(2007) suggest that earnings quality may be a function of the quality of the firm's internal control. ICOPINION(number of material internal control weaknesses reported in Audit Analytics in the post-SOX period) is,

08 March 2013 Page 5 of 145 ProQuest

therefore, included as a control variable. Consistent with DeFond and Jiambalvo (1994), Reynolds and Francis(2001), and Francis and Yu (2009), LEVERAGE, LOSS, and DISTRESS are included to control for the effects ofdebt and financial distress. LEVERAGE is total debt deflated by total assets. LOSS is a dichotomous variablewith a value of 1 if client has a negative net income before extraordinary items, and 0 otherwise. DISTRESS isZmijewski's (1984) measure of financial distress. B2M is the book-to-market value and captures the growthopportunities. Ashbaugh et al. (2003), Butler et al. (2004), Menon and Williams (2004), and Geiger and North(2006) suggest B2M and DACC to be negatively associated. Following Matsumoto (2002), Hribar and Nichols(2007), and Francis and Yu (2009), we include stock-return volatility (VOLATILITY) to proxy for capital marketpressures that can result in increased earnings management. FINANCED and ACQUIRED are dichotomousvariables that have values of 1 if the company was involved in significant financing activities or acquisitions,respectively, and 0 otherwise. Ashbaugh et al. (2003) and Chung and Kallapur (2003) find positive coefficientson these two variables. LAGROA is the previous year's return on assets and is included to control for priorperformance. BIG-N (a 0/1 dummy variable) and QUALIFIED (a 0/1 dummy variable) capture the effect ofauditor size and qualified opinions on earnings quality. Natural logarithm of 1 þnumber of calendar days fromfiscal year-end to date of auditor's report (LDELAY) controls for the effect of extra effort by the auditor onearnings quality. Since restatements can influence earnings management, the dichotomous variableRESTATEMENT is added as a control variable. MBEX is a dichotomous variable with a value of 1 if the firmmeets or beats the earnings expectation (proxied by the most recent median consensus analysts forecastavailable on I/B/E/S file) by two cents or less, and 0 otherwise. Since jDACCj and MBEX can be jointlydetermined, we use Maddala's (1988) two-stage-least-squares estimation (2SLS) for Equations (2) and (3) toavoid any endogeneity problems. I_MBEX denotes the instrumented variable for MBEX from the 2SLS.6 Meet-or-Beat Earnings Expectation Model There is evidence in extant literature that managers are rewarded(penalized) for meeting (missing) earnings forecasts (Bartov et al. 2002; Kasznik and McNichols 2002; Lopezand Rees 2002) and this leads to incentives for managers to manage earnings. If auditors' incentives to curtailearnings management vary with abnormal audit fee, the propensity for meeting or beating analysts' earningsforecasts will be a function of the abnormal audit fee. We estimate the following logit model to test thisconjecture: ... (3) In Model (3), F(|) denotes the logistic cumulative probability distribution function. I_jDACCjdenotes the instrumented variable for jDACCj from the 2SLS. Following Reichelt and Wang (2010), we includethe standard deviation of analysts' earnings forecasts (SDFOR) and natural logarithm of number of analysts'forecasts (LNUMFOR) to control for characteristics of the forecasts. Additional Tests To test H1b, we furtherinteract LOABNAFEE and HIABNAFEE with two measures of clients' bargaining powers in Models (2) and (3).LARGEST is a dichotomous variable with value of 1 if the client pays the highest audit fee in the practice officethat audits the client; TOP10%INFL is a dichotomous variable with a value of 1 if INFLUENCE is in the top 10percent, and 0 otherwise. To make the interpretations of the interactions easier, we use the dichotomousversions of abnormal audit fees, DLOABNAFEE and DHIABNAFEE, where DLOABNAFEE (DHIABNAFEE) is adichotomous variable with a value of one if LOABNAFEE (DHIABNAFEE) - median value, and 0 otherwise. Totest our conjecture that SOX has dampened the association between abnormal audit fees and earningsmanagement (H3), we rerun Models (2) and (3) with an additional independent variable DSOX and interactionterms, DLOABNAFEE * DSOX and DHIABNAFEE * DSOX, where DSOX is a dichotomous variable with a valueof 1 for the period 2004-2009, and 0 otherwise.7 Adjustment for Clustering Since our data are pooled over time,the same firm may appear more than once in the sample, resulting in clustering. Clustered samples can lead tounderestimation of standard errors and overestimation of significance levels (Cameron et al. 2011). Wetherefore estimate the t-statistics adjusted for clustering using robust standard errors corrected for firm-levelclustering and heteroscedasticity, consistent with Petersen (2009) and Gow et al. (2010). SAMPLE The sampleselection procedure is outlined in Table 2. We start with 61,953 firm-year observations for the period 2000-2009available in the Audit Analytics database for non-Andersen clients. We exclude Andersen since the auditor-

08 March 2013 Page 6 of 145 ProQuest

client relationship might be atypical, given the woes of Andersen. After merging with Compustat we are leftwith35,081 observations. We then exclude the financial (SIC codes 6000-6900) and utility (SIC codes 4400-4900)industries, since the incentives of managers from these regulated industries may be different. This leaves uswith 28,925 observations. Missing data for estimating the variables in a modified Jones (1991) model andEquation (2) further reduce our sample size to 18,873 observations (Sample 1). Finally, I/B/E/S and CRSP dataare needed for Equation (3), which results in Sample 2 with 14,796 observations. Untabulated statistics showthat the sample industry composition is very close to that of the Compustat population. Across years,untabulated statistics generally show an even distribution of observations. Sample characteristics, includingmeasures of central tendency, are presented in Table 3. Mean (median) jDACCj and MBEX are 0.0655 (0.0418)and 0.2039 (0.0000), respectively. Mean LOABNAFEE (HIABNAFEE) is 0.0259 (0.0267). Both variables havemedian values , 0.0001. A typical client accounts for under 2 percent of the audit office's revenue (median valueof INFLUENCE). A mean value of over 10 percent for this variable suggests some large influential clients. Over23 percent of the clients have been with their auditors for three years or less. On average, firms have betweentwo and three business and geographic segments. Mean (median) log of firm market value (in $ million) is 6.15(6.14). Clients experienced a mean of 13 percent sales growth during the sample period. The median firm hadone internal control weakness reported in Audit Analytics. On average, total debt was 19 percent of firm assetsand 30 percent of the firm-years reported losses. Book value was around 59 percent of market value of the firm.While 31 percent of the firms financed during the year, 19 percent were involved in acquisition activities. Asexpected, a large proportion (80 percent) of the clients are audited by Big 4 auditors. RESULTS Results ofestimation for Models (2) and (3) are presented in Panel A of Table 4. For the sake of comparison, regressionsare reported with and without control variables. The adjusted R2 of models with controls range from 18.39percent to 20.39 percent.8 For the jDACCj regression (Equation (2)), 14 of the 26 control variables aresignificant (at 10 percent level or better). SGROWTH, SDSALES, CFFO, SDCFFO, LOSS, DISTRESS,VOLATILITY, FINANCED, and MBEX are significantly positive; LOGMV, LEVERAGE, B2M, LAGROA, and BIG-N are significantly negative. Thus, firms with more sales growth, sales volatility, cash from operations, cashvolatility, losses, distress, stock price volatility, and financing activities are more likely to have higher jDACCj.There is also a positive association between jDACCj and MBEX. On the other hand, firms with larger size,higher leverage, lower growth potential, profitability, and with BIG-N auditors are more likely to have lowerjDACCj. Both LOABNAFEE and HIABNAFEE are significant and positive. This suggests that as the magnitudeof ABNAFEE increases on either side (positive or negative), the magnitude of jDACCj increases. This supportsH1a and H2. In Model (3), since the dependent variable MBEX is a dichotomous variable, we use logisticregression to estimate the propensity to meet or beat earnings expectations. The results for Model (3) areexhibited in the last two columns of Panel A of Table 4. Twenty-one of the 28 control variables are significant:CFFO, FINANCED, ACQUIRED, SDFOR, and jDACCj are significantly positive; LOFFICE, INFLUENCE,USLEADER, BUSSEG, LOGMV, ICOPINION, LEVERAGE, LOSS, B2M, VOLATILITY, LAGROA, BIG-N,QUALIFIED, LDELAY, RESTATEMENT, and LNUMFOR are significantly negative. LOABNAFEE andHIABNAFEE are significantly positive. Thus, this test also supports H1a and H2. Tests for H1b are reported inPanels B and C of Table 4. DLOABNAFEE and DHIABNAFEE are positive and significant as predicted by H1aand H2. DLOABNAFEE * LARGEST and DLOABNAFEE * TOP10%INFL are both positive and significant at a 5percent level or better for jDACCj and MBEX. This suggests that highly influential firms with more bargainingpower get more freedom to manage their earnings, for a given level of LOABNAFEE. This supports H1b.Moreover, DHIABNAFEE * LARGEST and DHIABNAFEE * TOP10%INFL are insignificant in Panel C, asexpected, since the bargaining effect is expected on the below-normal audit fee side (firms getting discounts)and not on the above-normal audit fee side (firms paying premiums). To test H3, we rerun Equations (2) and(3), with LOABNAFEE * DSOX and HIABNAFEE * DSOX added as additional independent variables (see PanelD of Table 4). Both of these interaction terms are significantly negative for Model (2), suggesting that the

08 March 2013 Page 7 of 145 ProQuest

management of jDACCj postulated in H1a and H2 has dampened post-SOX, supporting H3. However, theterms LOABNAFEE þLOABNAFEE * DSOX and HIABNAFEEþHIABNAFEE * DSOX are both significantlypositive at a 1 percent level, suggesting that the effects have not yet been completely erased as a result ofSOX. On the other hand, estimation of Model (3) with interactions yields insignificant interactions, suggestingthat the association of abnormal fees with MBEX is unaffected by SOX. Additional Analyses with EarningsResponse Coefficient Lim and Tan (2008) use the ERC (earnings response coefficient) as a proxy for investors'perception of earnings quality and, in turn, audit quality. We run the ERC model with interactions of unexpectedearnings (UE) with LOABNAFEE and HIABNAFEE. The independent variables in this model are based onFrancis andKe (2006), Limand Tan (2008), Wilson (2008), and Ghosh et al. (2009). We do not report detailedresults for the sake of brevity. However, coefficients of UE * LOABNAFEE and UE * HIABNAFEE are bothnegative, suggesting that the market perceives that audit quality declines with LOABNAFEE and HIABNAFEE,consistent with results reported for jDACCj and MBEX. Tests of Robustness We conduct several tests ofrobustness discussed below to convince ourselves that our results are not driven by any bias or modelmisspecification. 1. We first test for possible endogeneity bias in our audit fee model that potentially couldintroduce error into our measurement of ABNAFEE. The concern arises because Whisenant et al. (2003)suggest that audit and nonaudit fees are simultaneously determined, and thus not exogenous. To rule out thispossibility, we run Davidson and MacKinnon's (1993) endogeneity test. The null hypothesis of no endogeneity isnot rejected. 2. Unlike prior studies, we focus on audit fee revenue rather than nonaudit fee revenue as a sourceof impaired audit quality. We do so for two reasons. First, audit fee revenue is a recurring annuity whereasnonaudit fee revenue, with the exception of revenues from tax services, is not. Recurring engagements (ascompared to nonrecurring engagements) create greater incentive for the auditor to compromise independence.Second, reforms enacted by the Sarbanes-Oxley Act of 2002 have significantly reduced the scope of nonauditservices that auditors can provide to their clients while also expanding the scope of work required in the audit.Together, these two factors suggest that in the current milieu, audit fees emerge as a potentially importantdeterminant of audit quality.10 Recent research (Choi et al. 2010) has, therefore, focused on audit fees.However, as a robustness check, we repeat all of our tests with total engagement fee (i.e., the sum of audit andnonaudit fees) and arrive at similar conclusions. 3. We try an alternate specification for our models. Instead ofthe split-linear regression, we estimate the arctan regressions suggested by Freeman and Tse (1992) as below:... This specification yields interesting conclusions. The relationship of jDACCj with HIABNAFEE (LOABNAFEE)is S-shaped (inverse S-shaped). The flat curve around HIABNAFEE 1⁄4 LOABNAFEE 1⁄4 0 suggests that theauditors do not impair their independence until positive abnormal fee from a client exceeds what the auditorcould earn by simply resigning from this client and contracting with a new one.11 Similarly, on the negativeabnormal audit fee side, the auditor may have a pecking order. Initially they would grant their influential clients(with bargaining powers) fee discounts, and as the client gains more bargaining power, the auditor may alsoallow earnings management. The flattening of the curve for extreme departures from normal fee suggests thatthe auditors have a threshold or tolerance level for earnings management, and as the firm tries to reportextremely high levels of discretionary, accruals the auditors start resisting. 4. Since audit fee data becameavailable on Audit Analytics in 2000, one could question how the investors derive their ''normal'' audit fees toassess the audit quality in the year 2000. We rerun our analyses without the year 2000 data and ourconclusions do not change qualitatively. For Model (2), the coefficients of LOABNAFEE and HIABNAFEE are0.0251 (p1⁄40.0049) and 0.0316 (p1⁄40.0136), respectively; for Model (3), the coefficients of LOABNAFEE andHIABNAFEE are 1.9900 (p , 0.0001) and 1.2004 (p 1⁄4 0.0004), respectively. 5. Since quality of corporategovernance may also affect earnings quality, as a sensitivity analysis we add the Gompers et al. (2003) index ofcorporate governance to our regressions and obtain similar results. We do not report these findings as our mainresults since the sample size after merging with the Gompers' database was considerably smaller. 6. To theextent that the first few years of the auditor-client relationship might be atypical, we run our tests excluding the

08 March 2013 Page 8 of 145 ProQuest

first three years of new clients. The results are not altered qualitatively. For Model (2), the coefficients ofLOABNAFEE and HIABNAFEE are 0.0153 (p1⁄40.0016) and 0.0233 (p 1⁄4 0.0011), respectively; for Model (3),the coefficients of LOABNAFEE and HIABNAFEE are 1.3200 (p1⁄40.0002) and 0.3991 (p1⁄40.0813),respectively. 7. Firms that have consistent positive or negative abnormal fees across years might have differentmotivations from firms that switch between the two categories. We split the sample into firms that have five ormore years with abnormal audit fees of the same sign and those that do not. Our hypotheses hold for bothsubsets, suggesting that our results are applicable to both categories of firms. 8. We try different cutoffs for thetenure variable to see if the results are sensitive to the definition of tenure. We try five years and also twodummy variables for three years or under and more than nine years. Our main conclusions are not affected. 9.The pricing of audit engagements for accelerated filers may be different from that of nonaccelerated filers, sinceaccelerated filers could have a higher audit fee due to more internal control work. The pricing of materialweaknesses may also be different. If the size variable in the audit fee model is unable to control for theseeffects, the conclusions might be biased. To ascertain that the results are not different for accelerated/non-accelerated filers, we run the tests separately for the two filing categories. We do not find any evidence of anycategory leading the results. For example, in Model (2) for accelerated filers, the coefficients of LOABNAFEEand HIABNAFEE are 0.0095 (significant at 1 percent) and 0.0054 (significant at 10 percent), respectively; fornon-accelerated filers, the coefficients of LOABNAFEE and HIABNAFEE are 0.0147 (significant at 10 percent)and 0.0202 (significant at 5 percent), respectively. 10. As with any ratio, the scaling of abnormal audit fee by thetotal revenue of the audit office might introduce inflation (deflation) of the same numerator for smaller (larger)audit offices. To ascertain if this scaling affects our results, we rerun all the tests with unscaled abnormal auditfees. The coefficients of LOABNAFEE and HIABNAFEE continue to be positive and significant at a 5 percentlevel or better for both Models (2) and (3). 11. Prior research (e.g., Bedard and Johnstone 2004; Johnstone andBedard 2001, 2003) suggests that proxies for audit risk used in extant research may be inadequate. There maybe above-normal effort in cases of earnings manipulation risk and corporate governance risk (Bedard andJohnstone 2004) that are not captured by the audit fee model. This extra effort may result in an abnormally highaudit fee and lead to higher audit quality. Extending this argument, below-normal audit fees may representabbreviated audit efforts, and not clients' bargaining power, leading to poorer quality audits. In the absence ofdata on engagement hours and staffmix, we utilize the number of days between fiscal year-end and release ofthe audit report as a proxy for audit effort. We include this variable, square of the variable, and natural logarithmof the variable in our audit fee model, in an effort to control for auditor effort that is in response to missing auditrisk factors. All of our conclusions are supported as before. Additionally, unless the missing measures for auditrisk are correlated with bargaining power, the differential variation of audit quality with clients' bargaining power(documented in Panel C of Table 4) gives additional credibility to our results. However, to the extent that we areunable to control for audit risks in our model, the factors discussed above continue as limitations of ourresearch. 12. In Panels B, C, and D of Table 4, we use logistic regressions for the dichotomous dependentvariable MBEX, along with interaction terms. Norton et al. (2004) caution that since logistic regressions arenonlinear models, the interaction terms must be interpreted in a different way. They show that the marginaleffect of a change in both interacted variables is not equal to the marginal effect of the change in just theinteraction term. They suggest applying the INTEFF function in Stata for estimating the correct marginal effectof the interaction term. As a robustness check, we run the INTEFF function as detailed in Norton et al. (2004).Our conclusions are unchanged, suggesting that the results reported in Panels B, C, and D are robust. Oneconstraint with the INTEFF procedure is that it works for only one interaction term. In Panel D, for the SOXtests, we have two interaction terms. We therefore run two separate tests with INTEFF for each interaction term.In addition, to convince ourselves that our interpretation of the interaction effects is correct, we run the aboveregressions with MBEX replaced with a continuous variable, excess earnings defined as the reported earningsper share minus the analyst forecast. Since this dependent variable is not a dichotomous variable, we no longer

08 March 2013 Page 9 of 145 ProQuest

have to use the logistic regression. We are able to replicate all the results reported before. This further providesassurance that our interpretation of the interaction terms is reliable. CONCLUSION Auditors are hired andcompensated by their clients, and this creates an economic bond between the two. The question of whether thiseconomic bond ultimately leads to reduced audit quality remains an important public policy issue and has beenthe topic of prior research. We extend this research by examining the role of client bargaining power in additionto economic bonding as a determinant of audit quality. In our study, we examine the association betweenabnormal audit fees and two audit quality proxies: (1) abnormal discretionary accruals and (2) the probability ofmeeting or beating analysts' consensus earnings forecasts. Above-normal audit fees suggest quasi-rentsarising from a highly profitable audit engagement, while below-normal audit fees suggest the client has strongbargaining power (and hence is able to negotiate billing concessions). Both factors-quasi-rents and clientbargaining power-may lead the auditor to succumb to client pressure for earnings management. Thus, absoluteabnormal audit fees reflect the presence of these independence-impairing underlying factors and is our variableof interest. As hypothesized, we find that audit quality declines as actual audit fees depart from ''normal'' feelevels. Our finding that audit quality declines as positive abnormal audit fees increase in magnitude is consistentwith prior research. New to the literature is our finding that audit quality declines as negative abnormal auditfees increase in magnitude, with the decline increasing in magnitude as proxies for client bargaining powerincrease. We also examine whether the audit quality/abnormal audit fee association changed followingimplementation of SOX, and use this examination as a means of testing whether SOX was effective inincreasing auditor independence. We find evidence that this association in the post-SOX era is milder than inthe pre-SOX era. Thus, SOX reforms increase the risk of and reduce the gain from succumbing to clientpressure and compromising audit quality. It is important to note that our results regarding abnormally high or lowaudit fees could be attributable not to economic bonding or client bargaining power, but rather to factors that arenot captured by our audit fee model. Alternative explanations that we cannot investigate due to data limitationsinclude such items as audit team composition (i.e., the relative allocation of audit hours between partners,managers, and staff), the audit work allocation between interim and year-end, the influence of internal auditassistance, the relative quality of client financial reporting systems, and individual differential audit firmproduction functions. Accordingly, the reader should remain mindful of this limitation when interpreting ourevidence. Overall, our study increases understanding of the factors that may lead auditors to compromise onaudit quality, and hence our paper should be of interest to accounting scholars, investors, and regulators.Footnote 1 See Beck et al. (1988), Magee and Tseng (1990), and Zhang (1999) for extensions of DeAngelo's(1981a, 1981b) idea that quasi-rents impair auditor independence. 2 We refer to ''engagement fees'' when thetest variable in the study was total fees (i.e., the sum of audit and nonaudit fees). 3 In contemporaneousunpublished research, Hollingsworth and Li (2011) investigate the pre-SOX (years 2000- 2001) versus post-SOX (years 2003-2004) linear association between audit fee ratios and ex ante cost of capital. They findevidence of a positive association between total audit fees and ex ante cost of capital during 2003- 2004, but nosuch association during 2000-2001. 4 We do not deflate this measure by the total abnormal audit fee for thepractice office since this results in nearzero denominator problem. 5 Our audit fee model is based on modelsused in Ghosh and Lustgarten (2006), Craswell and Francis (1999), Craswell et al. (1995), and Simon andFrancis (1988). The adjusted-R2 of this model is over 81 percent. However, we do not report details for the sakeof brevity. Detailed specifications are available from the authors on request. 6 See Rusticus and Larcker (2010)for a more detailed discussion of instrumental variable estimation. 7 One of the SOX provisions to affect theaudit relationship was the requirement that auditors evaluate and report on management's assessment of theeffectiveness of the firm's internal controls (SOX Section 404(b) [U.S. House of Representatives 2002]). SinceSOX Section 404(b) was implemented for fiscal years ending on or after November 15, 2004 (per SEC ReleaseNo. 33-8392, issued February 24, 2004), we classify 2000-2003 as the pre-SOX period and 2004-2009 as thepost-SOX period. However, using 2002 as the cutoffdoes not alter the conclusions. 8 Tests for outliers are

08 March 2013 Page 10 of 145 ProQuest

conducted on all the regressions using Belsley et al.'s (1980) procedure. Results (not reported) do not changequalitatively when influential outliers are removed. Thus, our conclusions are not driven by outliers. 9 As anextra precaution, we reestimate the audit fee model using two-stage least squares as a control for potentialendogeneity bias. None of our conclusions are affected significantly. For Model (2), the coefficients ofLOABNAFEE and HIABNAFEE are 0.0238 (p , 0.0001) and 0.0154 (p 1⁄4 0.0138), respectively; for Model (3),the coefficients of LOABNAFEE and HIABNAFEE are 1.3262 (p , 0.0001) and 0.4209 (p 1⁄4 0.0522),respectively. 10 Findings of Hope et al. (2009) support this approach. They report that their results of positiveassociation between cost of equity capital and excess auditor remuneration only hold for audit fee and not fornonaudit fee. 11 We thank one of our reviewers for this insight. References REFERENCES Antle, R., and B.Nalebuff. 1991. Conservatism and auditor-client negotiations. Journal of Accounting Research 29 (Supplement):31-54. Ashbaugh, H., R. LaFond, and B. Mayhew. 2003. Do nonaudit services compromise auditorindependence? Further evidence. The Accounting Review 78 (3): 611-639. Balsam, S., J. Krishnan, and J.Yang. 2003. Auditor industry specialization and earnings quality. Auditing: A Journal of Practice &Theory 22 (2):71-97. Barnes, P. 2004. The auditor's going concern decision and Types I and II errors: The Coase theorem,transaction costs, bargaining power and attempts to mislead. Journal of Accounting and Public Policy 23 (6):415-440. Bartov, E., D. Givoly, and C. Hayn. 2002. The rewards to meeting or beating earnings expectations.Journal of Accounting and Economics 33 (2): 173-204. Beck, P., T. Frecka, and I. Solomon. 1988. A model ofthe market for MAS and audit services: Knowledge spillovers and auditor-auditee bonding. Journal ofAccounting Literature 7: 50-64. Bedard, J. C., and K. Johnstone. 2004. Earnings manipulation risk, corporategovernance risk, and auditor's planning and pricing decisions. The Accounting Review 79 (2): 277-304. Belsley,D. A., E. Kuh, and R. E. Welsch. 1980. Regression Diagnostics: Identifying Influential Data and Sources ofCollinearity. New York, NY: John Wiley &Sons. Butler, M., A. J. Leone, and M. Willenborg. 2004. An empiricalanalysis of auditor reporting and its association with abnormal accruals. Journal of Accounting and Economics37 (2): 139-165. Cameron, A. C., J. G. Gelbach, and D. L. Miller. 2011. Robust inference with multi-wayclustering. Journal of Business and Economic Statistics 29 (2): 238-249. Carey, P., and R. Simnett. 2006. Auditpartner tenure and audit quality. The Accounting Review 81 (3): 653- 676. Casterella, J., J. Francis, B. Lewis,and P. Walker. 2004. Auditor industry specialization, client bargaining power, and audit pricing. Auditing: AJournal of Practice &Theory 23 (1): 123-140. Chaney, P. K., and K. L. Philipich. 2002. Shredded reputation: Thecost of audit failure. Journal of Accounting Research 40 (4): 1221-1245. Choi, J., J. Kim, and Y. Zang. 2010.The association between audit quality and abnormal audit fees. Auditing: A Journal of Practice &Theory 29 (2):115-140. Chung, H., and S. Kallapur. 2003. Client importance, nonaudit services, and abnormal accruals. TheAccounting Review 78 (4): 931-956. Craswell, A., and J. Francis. 1999. Pricing initial audits engagements: Atest of competing theories. The Accounting Review 74 (2): 201-215. Craswell, A., J. Francis, and S. Taylor.1995. Auditor brand name reputations and industry specialization. Journal of Accounting and Economics 20 (3):297-322. Davidson, R., and J. G. MacKinnon. 1993. Estimation and Inference in Econometrics. New York, NY:Oxford University Press. Davis, L., B. Soo, and G. Trompeter. 2009. Auditor tenure and the ability to meet orbeat earnings forecasts. Contemporary Accounting Research 26 (2): 517-548. DeAngelo, L. 1981a. Auditorindependence, ''low balling,'' and disclosure regulation. Journal of Accounting and Economics 3 (2): 113-127.DeAngelo, L. 1981b. Auditor size and audit quality. Journal of Accounting and Economics 3 (3): 183-199.Dechow, P., and I. Dichev. 2002. The quality of accruals and earnings: The role of accrual estimation errors.The Accounting Review 77 (Supplement): 35-59. Dechow, P., R. Sloan, and A. Sweeney. 1995. Detectingearnings management. The Accounting Review 70 (2): 193-225. DeFond, M. L., and J. Jiambalvo. 1994. Debtcovenant violation and manipulation of accruals. Journal of Accounting and Economics 17 (1/2): 145-176.DeFond, M. L., and C. W. Park. 2001. The reversal of abnormal accruals and the market valuation of earningssurprises. The Accounting Review 76 (3): 375-404. DeFond, M., K. Raghunandan, and K. Subramanyam. 2002.Do nonaudit service fees impair auditor independence? Evidence from going concern audit opinions. Journal of

08 March 2013 Page 11 of 145 ProQuest

Accounting Research 40 (4): 1247-1274. Doyle, J., G. Weili, and S. McVay. 2007. Accruals quality and internalcontrol over financial reporting. The Accounting Review 82 (5): 1141-1170. Francis, J., and B. Ke. 2006.Disclosure of fees paid to auditors and the market valuation of earnings surprises. Review of AccountingStudies 11 (4): 495-523. Francis, J., and M. Yu. 2009. Big 4 office size and audit quality. The AccountingReview 84 (5): 1521-1552. Frankel, R. M., M. F. Johnson, and K. K. Nelson. 2002. The relation betweenauditors' fees for nonaudit services and earnings quality. The Accounting Review 77 (Supplement): 71-105.Freeman, R., and S. Tse. 1992. A nonlinear model of security price: Responses to unexpected earnings.Journal of Accounting Research 30 (2): 185-209. Geiger, M. A., and D. S. North. 2006. Does hiring a new CFOchange things? An investigation of changes in discretionary accruals. The Accounting Review 81 (4): 781-809.Ghosh, A., and S. Lustgarten. 2006. Pricing of initial audit engagements by large and small audit firms.Contemporary Accounting Research 23 (2): 333-368. Ghosh, A., S. Kallapur, and D. Moon. 2009. Audit andnonaudit fees and capital market perceptions of auditor independence. Journal of Accounting and Public Policy28 (5): 369-385. Gibbins, M., S. Salterio, and A. Webb. 2001. Evidence about auditor-client managementnegotiation concerning client's financial reporting. Journal of Accounting Research 39 (3): 535-563. Gompers,P., J. Ishii, and A. Metrick. 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118(1): 107-155. Gow, I., G. Ormazabal, and D. Taylor. 2010. Correcting for cross-sectional and time-seriesdependence in accounting research. The Accounting Review 85 (2): 483-512. Hatfield, R., C. Agoglia, and M.Sanchez. 2008. Client characteristics and the negotiation tactics of auditors: Implications for financial reporting.Journal of Accounting Research 46 (5): 1183-1207. Higgs, J. L., and T. R. Skantz. 2006. Audit and nonauditfees and the market's reaction to earnings announcements. Auditing: A Journal of Practice &Theory 25 (1): 1-26. Hoitash, R., A. Markelevich, and C. Barragato. 2007. Audit fees and audit quality. Managerial AuditingJournal 22 (8): 761-786. Hollingsworth, C. W., and C. Li. Investors' perceptions of auditors' economicdependence on the client- Post-SOX evidence. Available at: http://ssrn.com/abstract1⁄41800765 Hope, O., T.Kang, W. Thomas, and Y. K. Yoo. 2009. Impact of excess auditor remuneration on cost of equity capital aroundthe world. Journal of Accounting, Auditing, and Finance 24 (2): 177-210. Hornstein, H. 1965. Effects of differentmagnitudes of threat upon interpersonal bargaining. Journal of Experimental Social Psychology 1 (3): 282-293.Hribar, P., and D. W. Collins. 2002. Errors in estimating accruals: Implications for empirical research. Journal ofAccounting Research 40 (1): 105-134. Hribar, P., and D. Nichols. 2007. The use of unsigned earnings qualitymeasures in tests of earnings management. Journal of Accounting Research 45 (5): 1017-1053. Hribar, P., T.Kravet, and R. Wilson. 2010. A new measure of accounting quality. Available at: http://ssrn.com/abstract1⁄41283946 Johnson, V., I. Khurana, and J. Reynolds. 2002. Audit firm tenure and the quality ofaccounting earnings. Contemporary Accounting Research 19 (4): 637-660. Johnstone, K. M., and J. C. Bedard.2001. Engagement planning, bid pricing, and client response in the market for initial attest engagements. TheAccounting Review 76 (2): 199-220. Johnstone, K. M., and J. C. Bedard. 2003. Risk management in clientacceptance decisions. The Accounting Review 78 (4): 1003-1025. Jones, J. 1991. Earnings managementduring import relief investigations. Journal of Accounting Research 29 (2): 193-228. Kasznik, R., and M.McNichols. 2002. Does meeting or beating expectations matter? Evidence from analysts' forecast revisions andshare prices. Journal of Accounting Research 40 (3): 727-759. Kinney, W., and R. Libby. 2002. Discussion ofthe relation between auditors' fees for nonaudit services and earnings management. The Accounting Review 77(Supplement): 107-114. Kothari, K., A. Leone, and C. Wasley. 2005. Performance matched discretionaryaccruals. Journal of Accounting and Economics 39 (1): 163-197. Krishnan, J., H. Sami, and Y. Zhang. 2005.Does the provision of nonaudit services affect investor perceptions of auditor independence? Auditing: AJournal of Practice &Theory 24 (2): 111-135. Larcker, D., and S. Richardson. 2004. Fees paid to audit firms,accrual choices, and corporate governance. Journal of Accounting Research 42 (3): 625-658. Levitt, A. 2000.Reviewing the covenant with investors. Remarks delivered at New York University, Center for Law andBusiness, May 10. Lim, C., and H. Tan. 2008. Nonaudit service fees and audit quality: The impact of auditor

08 March 2013 Page 12 of 145 ProQuest

specialization. Journal of Accounting Research 46 (1): 199-246. Liu, X., and D. Simunic. 2005. Profit sharing inan auditing oligopoly. The Accounting Review 80 (2): 677- 702. Lopez, T., and L. Rees. 2002. The effect ofbeating and missing analysts' forecasts on the information content of unexpected earnings. Journal ofAccounting, Auditing, and Finance 17 (2): 155-184. Maddala, G. S. 1988. Introduction to Econometrics. NewYork: Macmillan Publishing Company. Magee, R., and M. Tseng. 1990. Audit pricing and independence. TheAccounting Review 65 (2): 315-336. Matsumoto, D. 2002. Management's incentives to avoid negative earningssurprises. The Accounting Review 77 (3): 483-514. Menon, K., and D. D. Williams. 2004. Former audit partnersand abnormal accruals. The Accounting Review 79 (4): 1095-1118. Michener, H., J. Vaske, L. Schleifer, J.Plazewski, and J. Chapman. 1975. Factors affecting concession rate and threat usage in bilateral conflict.Sociometry 38 (1): 62-80. Mitra, S., D. Deis, and M. Hossain. 2009. The association between audit fees andreported earnings quality in pre- and post-Sarbanes-Oxley regimes. Review of Accounting and Finance 8 (3):232-253. Norton, E., H. Wang, and C. Ali. 2004. Computing interaction effects and standard errors in logit andprobit models. The Stata Jamal 4 (2): 154-167. Petersen, M. 2009. Estimating standard errors in finance paneldatasets: Comparing approaches. The Review of Financial Studies 22 (1): 435-481. Pruitt, D., and P.Carnevale. 1993. Negotiation in Social Conflict. Pacific Grove, CA: Books/Cole Publishing. Reichelt, K., and D.Wang. 2010. National and office-specific measures of auditor industry expertise and effects on audit quality.Journal of Accounting Research 48 (3): 647-686. Reynolds, J., and J. Francis. 2001. Does size matter? Theinfluence of large clients on office-level auditor reporting decisions. Journal of Accounting and Economics 30(3): 375-400. Rusticus, T., and D. Larcker. 2010. On the use of instrumental variables in accounting research.Journal of Accounting and Economics 49 (3): 186-205. Schipper, K. 1989. Commentary on earningsmanagement. Accounting Horizons 3 (4): 91-102. Simon, D., and J. Francis. 1988. The effects of auditorchange on audit fees: Test of price cutting and price recovery. The Accounting Review 63 (2): 255-269.Simunic, D. 1980. The pricing of audit services: Theory and evidence. Journal of Accounting Research 18 (1):161-190. Trompeter, G. 1994. The effect of partner compensation schemes and generally accepted accountingprinciples on audit partner judgment. Auditing: Journal of Practice &Theory 13 (2): 56-69. U.S. House ofRepresentatives. 2002. The Sarbanes-Oxley Act of 2002. Public Law 107-204 [H.R. 3763]. Washington, D.C.:Government Printing Office. Whisenant, S., S. Sankaraguruswamy, and K. Raghunandan. 2003. Evidence onthe joint determination of audit and nonaudit fees. Journal of Accounting Research 41 (4): 721-744. Wilson, W.2008. The empirical analysis of the decline in information content of earnings following restatements. TheAccounting Review 83 (2): 519-548. Zhang, P. 1999. A bargaining model of auditor reporting. ContemporaryAccounting Research 16 (1): 167- 184. Zmijewski, M. 1984. Methodological issues related to the estimation offinancial distress. Journal of Accounting Research 22 (Supplement): 59-82. AuthorAffiliation Sharad C. Asthanaand JeffP. Boone are both Professors at The University of Texas at San Antonio. Editor's note: Accepted byJean Bedard. Submitted: December 2009 Accepted: March 2012 Published Online: August 2012 Subject: Auditing; Quality; Public Company Accounting Reform&Investor Protection Act 2002-US; Studies;Fees&charges Location: United States--US Classification: 4130: Auditing, 4320: Legislation, 9130: Experiment/theoretical treatment, 9190: United States Publication title: Auditing Volume: 31 Issue: 3 Pages: 1-22

08 March 2013 Page 13 of 145 ProQuest

Number of pages: 22 Publication year: 2012 Publication date: Aug 2012 Year: 2012 Publisher: American Accounting Association Place of publication: Sarasota Country of publication: United States Journal subject: Business And Economics--Accounting ISSN: 02780380 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Equations;Tables;References ProQuest document ID: 1143891718 Document URL: http://search.proquest.com/docview/1143891718?accountid=86413 Copyright: Copyright American Accounting Association Aug 2012 Last updated: 2012-11-09 Database: Accounting&Tax

_______________________________________________________________ Document 2 of 10 AN EMPIRICAL INVESTIGATION OF THE EFFECTS OF MERGER AND ACQUISITION ON FIRMS'PROFITABILITY Author: Ferrer, Rodiel C Publication info: Academy of Accounting and Financial Studies Journal 16. 3 (2012): 31-55.ProQuest document link Abstract: This study is considered a causal and correlational research, which aims to determine the relationshipof the mergers and acquisitions to the firm's profitability. It is a quantitative study that measured the effects ofmergers and acquisitions on return on assets and return on equity of the companies. Besides knowing therelationship, this study also obtained an estimate of the possible impact of the independent variable to thedependent variables. This study covered all the listed companies in the Philippines Stock Exchange for theyears 2006 until 2010. This covered companies from the different sectors of the economy, which comprise of 30companies in the financial sector, 75 firms in the industrial sector, 39 businesses classified as holding firms, 39companies in the property sector, 54 businesses in the service sector and 22 companies in the mining and oilsector. The research made use of two linear regressions to analyze the effect of having a merger or acquisitionon the profitability of the companies. Findings suggest that there is significant negative relation of merger andreturn on equity, having a merger or acquisition to return on equity implies that most mergers and acquisitionsdo harm to the financial well-being of the companies, rather than good. Furthermore, merger and acquisitionprovide an insignificant relation to the return on total assets, as evidenced by the insignificant p-value. As a

08 March 2013 Page 14 of 145 ProQuest

result, the finding of this variable provides empirical evidence that having a merger and acquisition does notaffect the return on assets ratio of companies in the Philippines. Full Text: Headnote ABSTRACT Economic advantage and competitive edge is the name of the game. Businesscombination is one proven and tested method by companies wanting to grow and gobble up a larger marketshare. The emerging business scenario created an additional burden to the already struggling corporations'existence, in almost all types of industry, which is due to the ever increasing demand for innovative strategies.To survive the dog-eats-dog world of competitiveness, a number of these players engage in businesscombination - wherein two or more companies incorporate into a single accounting entity. This study isconsidered a causal and correlational research, which aims to determine the relationship of the mergers andacquisitions to the firm's profitability. It is a quantitative study that measured the effects of mergers andacquisitions on return on assets and return on equity of the companies. Besides knowing the relationship, thisstudy also obtained an estimate of the possible impact of the independent variable to the dependent variables.This study covered all the listed companies in the Philippines Stock Exchange for the years 2006 until 2010.This covered companies from the different sectors of the economy, which comprise of 30 companies in thefinancial sector, 75 firms in the industrial sector, 39 businesses classified as holding firms, 39 companies in theproperty sector, 54 businesses in the service sector and 22 companies in the mining and oil sector. Theresearch made use of two linear regressions to analyze the effect of having a merger or acquisition on theprofitability of the companies. Two separate regressions are needed because profitability would be proxy by twodifferent but widely used variables: the return on equity and the return on assets ratio. Since the study coveredthe entire publicly listed companies in the Philippines for the period 2006 until 2010, this essentially means thatpanel data was used in the study. Hence, the appropriate panel analysis was conducted. Findings suggest thatthere is significant negative relation of merger and return on equity, having a merger or acquisition to return onequity implies that most mergers and acquisitions do harm to the financial well-being of the companies, ratherthan good. Furthermore, merger and acquisition provide an insignificant relation to the return on total assets, asevidenced by the insignificant p-value. As a result, the finding of this variable provides empirical evidence thathaving a merger and acquisition does not affect the return on assets ratio of companies in the Philippines.Keywords: Merger and Acquisition, Firms' Profitability, Panel Analysis, Return on Equity and Return on Asset.(ProQuest: ... denotes formulae omitted.) INTRODUCTION With the rapid advancement in technology, theglobal business industry is also at the forefront of such changes. For the past couple of years, we havewitnessed the introduction of new products in the market. Over-capacity indeed is the glaring issue here for thevery basic of the Law of Demand &Supply seems to have been ignored entirely. In the face of this predicament,firms need to reinvent ways of coping with the harsh reality of the industry. Should production be cut or totallycease from operation and rely solely on robust branding, or be more market pro-active and buy-up fledglingcompetition to emerge as the "last-man-standing"? Economic advantage and competitive edge is the name ofthe game. Business combination is one proven and tested method by companies wanting to grow and gobbleup a larger market share. The emerging business scenario created an additional burden to the alreadystruggling corporations' existence, in almost all types of industry, which is due to the ever increasing demand forinnovative strategies. To survive the dog-eats-dog world of competitiveness, a number of these players engagein business combination - wherein two or more companies incorporate into a single accounting entity. It is acommon practice for companies who underwent such combination to still continue with their product or brandsand distinct identities. However, after the business combination has commenced, the combined companies willnow share a common culture and mission aside from bearing the same corporate name. The major types ofbusiness combination as follows: mergers, consolidations and stock acquisitions. Merger is the combination oftwo or more entities by purchase acquisition whereby the identity of one of the entities remain while the othersare being dissolved. The reasons behind the merger transactions are basically gaining market share,competitive advantage, increasing revenues and risk and product diversifications. With the global financial

08 March 2013 Page 15 of 145 ProQuest

crises, it is noticeable that mergers and acquisitions have considerably increased. Corporations employed suchcombination not only for the sake of competitiveness but to maintain a firm foothold in the industry as well. Thishas lead to the significant transformation in the business landscape. Though one question that hounds theindustry, will the entities be able to handle the ramifications of the merger coupled with the risks involved in suchactivity and will the business combination improve the profitability of the firms. This paper investigates the effectof merger and acquisition on firm's profitability in terms of return on equity and return on assets. REVIEW OFRELATED LITERATURE According to Yurtoglo, there are three usual effects of merger and acquisitions:financial performance, industry and aggregate concentration levels and social welfare. The financialperformance of the company would definitely be affected by the business combination, as a result of synergiesor disruption that may either increase or decrease the company's operating performance. On the other hand,the combination of two or more companies would decrease the number of players in a given industry. Finally,Yurtoglo indicated that as a result of the changes in financial performance and aggregate concentration levels,the social welfare of the people would also be affected by merger or acquisition. For example, the recentacquisition of Equitable PCI by Banko De Oro (BDO) led to the rise of BDO as the number one bank in thePhilippines, in terms of assets and amount of deposits. In addition to this, BDO also received increased marketpresence, as a result of converting the former Equitable PCI banks into BDO banks. This, in effect, made the lifeof BDO depositors and the society at large easier, since they have more alternatives on which branch they optto conduct their transaction. There had been numerous studies from different countries conducted on the effectof merger or acquisition on the financial profitability of the companies. Ollinger, Nguyen, Blayney, Chambersand Nelson (2006) provided empirical evidence that merger and acquisition improved the labor productivity inthe food sector. Holmstrom (2001) examined changes in the merger activity and corporate governancemechanism in the United States. He concluded that there is a rise in merger and acquisition activity for theperiod 1980 until 1999 and that the corporate governance mechanism has evolved from leveraged hostiletakeovers and buyouts in the 1980s to incentive-based compensation in the last portion of the 1990s. Kemal(2011) investigated the effects of merger or acquisition on the different financial ratios of companies in thebanking industry of Pakistan for the years 2006 until 2009. They found evidence that mergers or acquisitionsactually worsen the liquidity, profitability, return on investment and market stock ratios of the banks while thesolvency ratio is the only one that improved. Altunbas and Ibanes (2004) provided evidence that bank mergersin Europe resulted to an improvement on the companies' return on capital, particularly on cross-border mergers,as a result of organizational and strategic fits. They found out that the improved performance can primarily beattributed to the broad similarities between merger participants. Altunbas and Ibanes (2004) separated theiranalysis between domestic and cross-border mergers and acquisitions and found evidence that financialinnovation, capitalization and investment in technology resulted to an enhanced performance for domesticmergers while differences in loan and risks strategies boost the performance of cross-border businesscombinations. On the flip side, they indicated that the disparities in earnings, loans and deposit strategiesbrought about damaging consequences for domestic mergers while inconsistencies in capitalization, technologyand financial innovation promulgated harmful effects on cross-border mergers and acquisitions. Hu (2009)examined the long-term financial performance of Chinese-acquiring companies during the post-acquisitionperiod and found that the buying companies receive no significant positive abnormal returns over one yearfollowing the merger or acquisition. However, he was also able to establish that acquiring companies receivesignificant positive abnormal returns three years after the merger or acquisition. In addition, he was able todetermine that only acquirers using asset acquisitions experience positive abnormal returns aver the three yearperiod. In contrast, even if the overall result over the one year period is insignificant, Hu (2009) found evidencethat acquirers using tender offers receive positive abnormal returns. In summary, Hu (2009) indicated that thetype of transaction, the industry characteristics and the year of acquisition have a significant influence on theacquirer's financial performance over the long term. Mantravadi and Reddy (2008) found empirical evidence

08 March 2013 Page 16 of 145 ProQuest

that, overall, companies in India are experiencing slight increases in their profitability following the merger oracquisition. However, the impact is different when the different industries are considered in isolation. Theyindicated that businesses in the banking and finance industry and pharmaceutical companies receive slightpositive impact in their profitability while companies in the textile and electrical equipment sectors obtained anegative impact on their performance. In contrast, they depicted that firms in the chemicals and agri-productsindustry experience a significant decline in their profitability, as measured by their return on investment andreturn on assets, after the merger or acquisition. Wong, Cheung and Mun (2009) conducted their research usingmarket measures, specifically the relationship between security returns and returns on the market portfolio, onthe Asian markets, particularly Hong Kong, China, Taiwan, Singapore, South Korea and Japan. They providedempirical evidence that stockholders of target firms regard merger or acquisition as bad news while theshareholders in the acquiring company regard it as good news. They found out that there is no abnormal returnto the shareholders of target firms surrounding the announcement period. They attributed this to the low marketvalue of the shares before the announcement period as the results of either poor financial performance of thetarget company or an information leakage with regards to the features of the takeover, such as the acquiringprices. In contrast, they regard the decreasing market value after the announcement as the reaction of themarket to the news. On the other hand, Wong et. al (2009) also furnished proof that the market value of thebuying company's shares receives abnormal returns right after the announcement of the acquisition, dependingon the type of acquisitions. Yen and Andre (2010) cited several evidences that shareholders of acquiring firmseither suffer losses, as a result of merger or acquisition, or, at best, break even. They investigated the effects ofconcentrated ownership, governance mechanisms and legal protection on corporate performance of acquiringfirms. As a result, they found out that buying firms with shareholders owning 25% to 30% of the company tendsto improve their operating performance within the next three years following the business combination. Theyindicated that their result is actually in conformity with the prior literature providing that firm value rises asownership of the largest shareholders. This finding may have been the result of the agency problem. Whenindividuals own a larger portion of the corporation, their financial well-being is linked more closely to theperformance of the said business, which creates an incentive on that individual's part to ensure that theoperations of the firm would be profitable. In contrast, if an individual owns only a small portion of a corporation,their financial well-being may be less affected compared to that of the controlling shareholder. In addition, thatperson would most likely have no capacity to influence corporate decisions, as a result of his small interest.Hence, a corporation with shareholders owning a huge chunk of its corporate ownership tends to have highercorporate performance after merger or acquisition, than businesses whose ownership is levelled out among ahandful of shareholders. Williams (2010) and Lafosse (1999) provided further evidences that may explain theeffects of merger and acquisition on the profitability of the companies. Williams (2010) indicated that companiesoften overlook the marketing synergies that may result from mergers and acquisitions, which actually result toeither its failure or undermines the benefit that acquiring firms get from such business combination. Heexplained that operating synergies on marketing components can also be obtained from mergers andacquisitions, particularly in a horizontal integration. Lafosse (1999) provided empirical evidence that theaccounting method (pooling or purchasing) used to consider the merger activity does not give target firmsabnormal returns. However, upon separating their analysis between firms listed in the NASDAQ and companiestraded in the NYSE, they found out that target companies listed in the NASDAQ tend to have higher premiumspaid by the acquirers for their shares. Singh and Zollo (1999) provided empirical evidence that knowledgecodification has a significant positive impact on the post-acquisition financial performance of companies in thebanking industry while experience accumulation on mergers or acquisitions provides no such impact.Knowledge codification is defined as the process of converting tacit knowledge or the type of knowledge that isdifficult to transmit to another person into explicit knowledge or knowledge that can easily be transmitted(turing.edu). An example of tacit knowledge is the capability of swimming. Any swimmer can easily describe the

08 March 2013 Page 17 of 145 ProQuest

proper way of swimming, yet after hearing the lecture, the person who does not know how to swim would still beunable to swim. On the other hand, explicit knowledge can readily be seen in manuals, lectures and cookinginstructions. Singh and Zollo (1999) also expressed a direct relation between the level of integration andfinancial performance while a negative relation is exhibited by the replacement of top management to corporateperformance. Ismail, Abdou and Annis (2011) pointed out that these studies have conflicting results, primarilybecause of the differences in the scope and measures used by the contrasting studies. Some studies focusedon a particular industry, such as the steel industry, construction sector and railroad industry while others tookinto account all the listed companies in their area (Ismail et. al, 2011). Hence, it is not really surprising thatstudies focusing on the food sector would have different results from studies engaged in the telecommunicationsector or that studies concentrating on American firms would have different findings as a similar study onCanada or Germany. On the other hand, the studies also differ on the measures used to signify profitability.According to Ismail et. al (2011), a handful studies used either market measures, such as market power, bookto market ratio and cumulative abnormal returns, or accounting measures like operating income over sales,sales ratio and solvency to signify the company's operating performance. In contrast, Ismail et. al (2011)indicated that some studies used a combination of both market-measures and accounting-measures whileothers used qualitative-measures, particularly the theoretical researches. Either way, the studies falling in eachof the four categories depicted conflicting results. The inconsistent result can primarily be attributed to thecombination of the differences in both the scope and the measures used to signify operating performance.STATEMENT OF THE PROBLEM Accounting is the tool often used to simplify the complex environment of thebusiness world, where every now and then countless transactions are involved. Over the years, business hasdeveloped and diversified into various forms and methodologies. This has induced the need for a specializedsystem of monitoring and evaluation of its objective, to earn profit, without jeopardizing ethics and the welfare ofits various stakeholders. Audit is one of these resulting systems. The primary objective of audits is to ascertainthe validity and reliability of information and to administer an assessment of a system's internal controls.Classifications of audit include: operational audit, financial audit, compliance audit, information systems audit,and investigative or forensic audit. Financial statements are the primary source of quantitative financialinformation regarding important aspects about a company that is useful to a wide range of users in makingeconomic decisions. In order to ensure the veracity of the reported information, financial statements should beaudited by independent certified public accountants (CPAs). The CPA is guided by generally accepted auditingstandards (GAAS) in conducting the audit examination and in rendering an opinion as to whether such financialstatements were presented fairly and in conformity with the generally accepted accounting principles (GAAP).Nonetheless, it is still the management of the business enterprise that is principally liable for the preparation andpresentation of financial statements that conform to GAAP. Management approval is essential to enact anychanges or adjustments needed to rectify material misstatements discovered in the audit. If such approval is notobtained, the CPA practitioner would be obliged to make the necessary modification in the "IndependentAuditor's Report." Racasa (2003) Companies use the annual reports as the primary mode of communication tocorrespond with stakeholders (Botosan, 1997; and Lang and Lundholm, 1993). It is through these reports,where companies disclose relevant information that plays a crucial role in the decision-making processes.Cooke claims that it is important to assess the extent of disclosures made by a corporation, as stakeholders relyheavily on these pieces of information when making different types of decisions (Cooke, 1989). These pieces ofinformation are crucial in the decision-making processes regarding the allocation of scarce resources forstakeholders. The problem addressed by this paper is: What effect does merger or acquisition have on acompany's profitability? Null Hypotheses Null Hypothesis (Hoi): Merger and acquisition has no significant effecton return on total assets. Null Hypothesis (Ho2): Merger and acquisition has no significant effect on return onequity. THEORETICAL FRAMEWORK The different theories underlying the study includes the transaction costtheory, technological competence theory and internalization theory. Transaction cost theory actually

08 March 2013 Page 18 of 145 ProQuest

encompasses the other two theories and, hence, could be designated as the "parent" theory. Transaction costtheory Transaction cost theory addresses the problem of organizing interdependencies among many individuals(Hennart, 2001). Hennart (2001) explained transaction costs as "the information, enforcement and bargainingcosts incurred by economic agents, as a result of bounded rationality and opportunism." Bounded rationality,also known as cognitive limitation, is the idea that an individual's decision, although rational, is limited by theinformation they have, their capacity to evaluate the information available and the amount of time to makedecisions (Jones, 1999). Opportunism, on the other hand, is discussed by Williamson (1981), as the tendencyof individuals to pursue actions that would maximize their self-interests. Klein (2006) explained thatopportunistic individuals cannot be relied on to retain their promises, to fulfill their obligations and to respect theinterests of their trading partners. Hence, safeguards should be placed. Here is a brief discussion of thetransaction cost framework. The transaction cost framework, which was adapted from the study of Mikkonen,signified that companies incur two types of transaction costs, namely external or market transaction costs andinternal or bureaucratic transaction costs. External transaction costs comprise of expenses that the companywould incur if it chooses to engage in a transaction with another company. On the other hand, internaltransaction costs consist of the expenses that the company would incur, if it chooses to produce the product in-house. This is actually the same expenses that the company will incur, if it chooses to merge with anothercompany. There are two human factors within the framework: bounded rationality and opportunism. In contrast,there are three environmental factors: asset specificity, uncertainty and complexity. The more specialized orrare an asset is; the higher is the transaction costs that the firm would face in order to obtain that asset. Thisactually provides the company with an incentive to simply merge with the supplier ofthat rare asset. The samegoes for the other two environmental factors. The greater the uncertainty in the environment or the morecomplex the environment where the company operates, the higher is the transaction costs that the firms wouldincur. Bounded rationality and opportunism, on the other hand, would also increase transaction cost. Acompany may choose to merge with its supplier when it finds it cheaper to produce a product in-house. If acompany is thinking whether to outsource its production for a given product, it may assess the costs related tosuch a transaction with the environment (businessmate.org). If the company sees it as difficult to formulate acontract that controls the uncertainties related to the exchange, the company may regard it as too costly tooutsource the production (businessmate.org). This is because the transaction costs of monitoring the exchangeare perceived to be higher, than the bureaucratic costs of performing the activity in-house (businessmate.org).Nevertheless, the company may decide to simply acquire the supplier in order to take advantage of its expertisein producing the given product and, hence, also avoiding the high transaction costs associated with the externaltransaction. Transaction cost theory, as stated by Williamson (1981), discusses that "the task of economicorganizations is to coordinate transactions so as to economize on bounded rationality while simultaneouslysafeguarding them against the hazards of opportunism." Hennart (2001) explained that when individuals formgroups with different or similar capabilities, they are able to generate economic rents or excess returns. Thistheory claims that firms or organizations arise when they are the most efficient institutions to organize theinterdependencies among individuals (Hennart, 2001). This means that firms would enable individuals to takeadvantage of each other's skills while minimizing the costs that will be incurred to conduct the transaction.These transaction costs include the legal expenses of forming the contract that will ensure the relation of theparties and the cost of organizing the relation. Applied to the company level, the transaction cost theory explainsthat companies merge in order to take advantage of the skills or expertise of another company, so as tominimize the transaction costs that need to be incurred everytime they will do business. Hence, the objective ofthe transaction cost theory is to be able to form groups that would minimize transaction costs. CONCEPTUALFRAMEWORK The effect of having a merger and acquisition to the company's profitability would be the mainobjective of this study. The company's profitability, which is the dependent variable, would be represented bythe return on total assets and the return on equity ratio. On the other hand, the only independent variable of the

08 March 2013 Page 19 of 145 ProQuest

study is merger and acquisition, which signify whether the company had engaged in the said businesscombination in a given year. SIGNIFICANCE OF THE STUDY This section discusses how the results of thestudy would be of importance to the various sectors of the economy. This study would particularly be helpful tothe following stakeholders: corporate stockholders and managers, investors, academe, government, brokers,consultancy firms, investment banks, business partners, creditors, customers, the Philippine Stock Exchangeand the public in general. Corporate Stockholders and Managers This study would help managers indetermining whether doing business combinations is a viable option in increasing profitability. It would indicatethe effects of merger and acquisition on the company's profitability, which would allow managers to give betterdecisions for the best interest of the company. Stockholders would also profit from this study by gainingessential information on the effects of mergers and acquisitions on the value of their investments. Investors Theresults of this study would be beneficial to prospective investors, by helping them make correct decisions, withregards to investing in companies that would undergo merger or acquisitions. In addition, it would improve theinvestors' prediction of the future cash flows and earnings that would be generated by these companies.Academe The academe would profit from this study by gaining knowledge on the effect of merger or acquisitionon the profitability of publicly listed companies in the Philippines. In addition, this research would also incitefuture studies with regards to similar topics as merger and acquisition. Government With regards to thegovernment, this study would provide them with information about the usual effects of merger and acquisition onthe companies' profitability. This would help the government promulgate relevant policies that would regulatethe practice of merger between companies, as well as the acquisition of one company on another. In addition,the findings of this study would enable the government to effectively control market conditions of the varioussectors of the economy that would be affected by such business combination. Business Partners, Creditors andCustomers The findings of this research would help companies determine the most likely effects of businesscombinations on its profitability ratio. This would signal other firms to take measures that will minimize theharmful effects of merger or acquisition of their business partners on the company. Creditors, on the other hand,would gain more knowledge about business combinations, which will allow them to decide on the appropriateinterest rate to charge firms undergoing such business strategy. Customers, likewise, would benefit by enablingthem to find out the probable effect of such merger and acquisition on the company's ability to continue goingconcern. Hence, this research would provide valuable knowledge to business partners, creditors andcustomers. The Philippine Stock Exchange This study would aid the Philippine Stock Exchange whenconducting seminars with regard to the possible consequences of business combinations here in the country.Specifically, the PSE would be able to extend the knowledge of the impact of merger and acquisitions on thecompany's profitability ratios. Public The public would be given a clearer picture of the market conditions ofindustries where business combinations occur. They would be able to usual impact brought about by businesscombination, particularly merger and acquisition, on the company's profitability. SCOPE AND LIMITATIONSThis section discussed the scope and limitations of the study. This study covered all the listed companies in thePhilippines Stock Exchange for the years 2006 until 2010. It used secondary data, which were obtainedprimarily from the Osiris database. On the other hand, the main limitation of the study is brought about by itsvery low r-squared. This is not really surprising; since there is only one independent variable and r-squarednormally rise as the number of predictor variables increase. OBJECTIVE AND METHODOLOGY OF THESTUDY Research Design This study is considered a causal and correlational research, which aims to determinethe relationship of the mergers and acquisitions to the firm's profitability. It is a quantitative study that measuredthe effects of mergers and acquisitions on return on assets and return on equity of the companies. Being acausal research, it determined whether the change in the company's profitability is caused by businesscombination. On the other hand, as a correlational research, the study seeks to know the relationship ofmergers and acquisitions to the companies' financial performance. Besides knowing the relationship, this studyalso obtained an estimate of the possible impact of the independent variable to the dependent variables.

08 March 2013 Page 20 of 145 ProQuest

Population of the Study The population of the study comprised the entire Philippine publicly listed companiesfrom 2006 until 2010. This covered companies from the different sectors of the economy, which comprise of 30companies in the financial sector, 75 firms in the industrial sector, 39 businesses classified as holding firms, 39companies in the property sector, 54 businesses in the service sector and 22 companies in the mining and oilsector. The financial sector comprise of the banks, other financial institutions, preferred companies and half ofthe small and medium enterprise. On the other hand, the industrial sector includes companies that providechemicals; construction, infrastructure and allied services; utilities; food, beverage and tobacco; as well as thediversified industrials. In contrast, the holding firms sector and the property sector is composed only of holdingfirms and businesses engaged in the buying and selling of property, respectively. The services sector includesfirms that provide diversified services, education, hotel and leisure, information technology, media,telecommunications, transportation services and the other half of the small and medium enterprise. Finally, themining and oil sector is composed of companies engaged in the mining business and firms that sell oils. Overall,there are 259 listed companies in the Philippine Stock Exchange. Research Procedures The study made use ofonly secondary data, which were obtained from the OSIRIS database. The merger and acquisition informationin the OSIRIS database were used to identify the occurrence of the said business combination, which were thenrecorded as dummy variables. As a result, the independent variable is a dichotomous variable that equals 1 ifeither merger or acquisition occurred in a given year for the respective companies and 0 otherwise. On theother hand, return on assets and return on equity for the publicly listed companies are readily available in theOSIRIS database. Hence, the ratios used in the data analysis are the exact values obtained from the OSIRISdatabase. Methods of Data Analysis The research made use of two linear regressions to analyze the effect ofhaving a merger or acquisition on the profitability of the companies. Two separate regressions are neededbecause profitability would be proxy by two different but widely used variables: the return on equity and thereturn on assets ratio. Since the study covered the entire publicly listed companies in the Philippines for theperiod 2006 until 2010, this essentially means that panel data was used in the study. Hence, the appropriatepanel analysis was conducted. Linear Regression Analysis Equation 1 : Return on assets ratio as the predictedvariable roa it = ma it Where roa it = return on assets ratio of company i at period t ma it = merger andacquisition of company i at period t On the above equation, roa it signified the endogenous variable while ma itindicated the exogenous variable, roa itis composed of the return on assets ratio of the respective companiesfor the period 2006 until 2010. On the other hand, ma it is a binary variable that takes the value of 1 granted thatcompany i had a merger or acquisition in period t and 0 if it did not. Equation 2: Return on assets ratio as thepredicted variable roe it = ma it Where roe it = return on equity ratio of company i at period t ma it = merger andacquisition of company i at period t On the above equation, roe it signified the dependent variable while ma itindicated the independent variable, roe itis composed of the return on equity ratio of the respective companiesfor the period 2006 until 2010. On the other hand, maitis a binary variable that takes the value of 1 granted thatcompany i had a merger or acquisition in period t and 0 if it did not. Panel Data Regression Since panel datawas used in the study, the appropriate panel analysis was conducted. Panel analysis is composed of a series ofsteps to determine which among three models namely, the ordinary least squares, fixed effects model and therandom effects model was appropriate for the data used. The ordinary least squares (OLS), also known as thenaive model, does not take into account the variations and possible interactions between cross-sectionalobservations, as well as its potential time effect. On the other hand, the fixed effects model (FEM), which is alsoknown least squares dummy variable (LSDV), takes into account either/both the individuality of each companiesand time. Lastly, the random effects model (REM) assumes that the observations were randomly drawn fromthe population. The Wald test and the Breusch-Pagan test were used to determine which among the threemodels is most appropriate. Panel analysis was done twice because there are two dependent variables thatwere studied. The final model for the return on equity may not essentially be the resulting model for the returnon assets. However, as a matter of coincidence, the panel analysis produces similar results for both dependent

08 March 2013 Page 21 of 145 ProQuest

variables. That is, both models used OL S as the final model. Naïve Panel Data Regression/Ordinary LeastSquares Regression The naive panel data regression treats the panel data as a simple cross-sectional data andruns it as a linear regression. For the naive panel data regression, the panel data was regressed, first, withreturn on equity as the output variable and merger and acquisition as the independent variable then,subsequently, with return on assets as the dependent variable and merger and acquisition as the predictorvariable. The equation for the naive panel data regression is as follows: y = x Where y = dependent variable ? =independent variable Fixed Effects Panel Data Regression After running the naive panel data regression, thethree types of fixed effects models, namely the space-varying fixed effects model, the time-varying fixed effectsmodel and the spaceand time-varying fixed effects model, was tested. Using the fixed model would assume thattimeinvariant characteristics (e.g. error term and constant) are unique to the company and that it must not becorrelated to another company's individual characteristic (Torres-Reyna, 2009). Afterwards, the test of overallsignificance of dummies was conducted to determine which among the three models is most appropriate for thedataset gathered. Space- Varying Fixed Effects Model - Test for Effects of Company-Specific Attributes(LSDV1) The first type of fixed effects panel data regression was the space-varying fixed effects model, which isalso called the within-groups regression or LSDV1. LSDVltakes into account the "individuality" of each companyby enabling the intercept for each cross-sectional data or company to vary but still assume constant slopecoefficients among firms. Although the use of this model would take into account the interactions between thecross-sectional data, its usage resulted to loss of unchanging explanatory space variables and the loss of asignificant amount of degrees of freedom. (Dougherty, 2006) The formula based from Torres-Reyna (2009) forthe space-varying fixed effects model is stated below: ... Where y^sub it^ = dependent variable where i =company and t = year β^sub 0^ =intercept of the model x^sub k,it^ = independent variable β^sub k^ = coefficientof the independent variable Y^sub n^ =coefficient for the binary regressors (companies) E^sub n^ = dummyvariable representing the companies u^sub it^ = error term Time- Varying Fixed Effects Model - Test for Effectsof Time-Specific Attributes (LSDV2) The second type of fixed effects panel data regression is the time-varyingfixed effects model, which also called the first differences regression model or LSDV2. LSDV 2 is similar toLSDV 1, except that it enables the intercept to vary over time. However, the problem in the first fixed model,which is the loss of a significant amount of degrees of freedom, is still present in this model (Dougherty, 2006).According to Torres-Reyna (2006), the general formula of time-varying fixed effects model is as follows: ...Where y^sub it^ = dependent variable where i = company and t = year β^sub 0^ = intercept of the model x^subit^t = independent variable β^sub k^ = coefficient of the independent variable a^sub n^ = coefficient for thebinary time regressors T^sub m^ = dummy variable representing the time period u^sub it^ = error term Space-and Time- Varying Fixed Effects Model - Test for Effects of Both Company and Time Specific Attributes(LSDV3) The last type of fixed effects panel data regression is the space- and time-varying fixed effects modelor LSDV3. This model is a combination of the two previously discussed fixed effect models. LSDV 3 allows theintercept to vary across each individual company and over time. However, consistent with the two previousLSDV models, LSDV3 also suffers from the loss of a significant amount of degrees of freedom (Dougherty,2006). Looking at the work of Torres-Reyna (2006), the general formula of the space- and timevarying fixedeffects model is: ... Where y^sub it^ = dependent variable where i = company and t = year β^sub 0^ = interceptof the model x^sub k,it^ = independent variable β^sub k^ = coefficient of the independent variable Y^sub n^ =coefficient for the binary regressors (companies) E^sub n^ = dummy variable representing the companiesδ^sub N^ = coefficient for the binary time regressors T^sub n^ = dummy variable representing the time periodu^sub it^ = error term Test Parm for Determining the Most Appropriate Fixed Effects Model After creating thethree fixed effects models, the command testparm in Stata was used to determine which among the threemodels would best predict the dependent variables. After entering the command, the model with the highest F-statistics or the lowest p-value was chosen to represent fixed-effects model. Wald's Test for Comparing NaïveModel and Fixed Effects Model After determining the best type of Fixed Effects Panel Data Regression model,

08 March 2013 Page 22 of 145 ProQuest

the naive panel data regression model and the time-varying fixed effects panel data regression was comparedusing the Wald's Test. The F statistics and the critical F-value, which are the two elements that would becompared in Wald's test, is manually computed using results from both the OLS and the chosen FEM modelwith the use of the equation provided in the following page. ... Where, RSS^sub R^ = Pooled RegressionModel/Ordinary Least Squares RSS^sub UR^ = Fixed Effects Model M = # of parameters D^sub f^ = # ofobservations - # of parameters (degree of freedom lost) The degrees of freedom is computed as follows: ...Where m = number of time variables or years G = number of space variables or companies k = total number ofvariables including constant and dependent variables After computing the Fstat, the Fcrit is then determined bylooking at the critical F values with m variables, df degrees of freedom and 0.05 level of significance. Finally, thegroup compared the Fstat and the Fcrit. If the F statistics is greater than the critical f-value, then FEM ispresumed to be the better model. In contrast, if the F statistics is less than the critical f-value, then the nullhypothesis that OLS is the better model is accepted. Random Effects Panel Data Regression Model Unlike thefixed effects model, random effects model assumes that the entity's error term is not correlated with theindependent variables, which would allow time-invariant variables to be included as explanatory variables(Torres-Reyna, 2006). The random effects model treats the variables previously unobserved in the fixed effectsmodel as being randomly drawn and that these unobserved variables are independent from the variablesincluded in the fixed effects model (Dougherty, 2006). According to Torres-Reyna (2006), the random effectmodel general equation is: ... Where y^sub it^ = dependent variable where i = company and t = year β^sub k^ =coefficient of the independent variable a = unknown intercept for each company u^sub it^ = between entity errorE^sub it^ = within entity error Breusch and Pagan Lagrangian Multiplier Test for Random Effects - ComparisonBetween Naïve Model and Random Effects Model The Breusch and Pagan lagrangian multiplier test determineswhether there is significant evidence against the null hypothesis that the variances of groups in the one-wayrandom group effect model or the naive model are zero. Thus, if the variances of the groups in the naive modelare not zero, the random effects model will be better than the one-way random group effect model. The Statacommand xttestO allowed the researchers to determine if the p-value of the Breusch and Pagan multiplier testwould be lower than 0.05. If the p-value is lower than 0.05, then the random effects would be the better model. Ifnot, the naive model would be the more appropriate model. EMPIRICAL RESULTS AND ANALYSIS Thisportion of the study answers the research problem and objectives of the paper. This chapter presents thesummary statistics of the data gathered, the result of the regression analyses made using the IFRS disclosureindex as the dependent variable and various profitability measures of the firm as the independent variable. Thepredicted variables are return on equity and return on total assets while the predictor variable is merger andacquisition. The mean value of return on equity is 93.41, which indicates that the companies, on average, havenet income that is larger than their common stockholders' equity. On the other hand, the return on total assetssignified an average of -59.02. The negative sign is the result of companies that had obtained a net loss. Lastly,the exogenous variable, merger and acquisition, provided an average of 0.30. This means that there are moreobservations that did not undergo merger or acquisition, than there are samples that had engaged. Thecorrelation table revealed the relation of the different variables with one another. The association of return onassets to return on equity is the only relation that indicated a moderate correlation. It actually signified amoderately negative relation, as indicated by having a correlation coefficient that is above 0.20 but below 0.80.The relation of merger and acquisition to either of the dependent variables, on the other hand, connoted a lowcorrelation. Low negative relations with regards to return on equity while a low positive relation to return onassets ratio. Panel analysis was conducted in order to determine which among the three models: ordinary leastsquares, fixed effects model and random effects model was appropriate for the study. The data was first runusing the OLS model and the appropriate sum of the squares of the residuals was noted. Afterwards, the datawas run under each of the three variations of the fixed effects model, namely LSDV 1, LSDV 2 and LSDV 3. Themodel with the highest f-value or lowest p-value among the three LSDV models was used to represent the FEM.

08 March 2013 Page 23 of 145 ProQuest

The test of overall significance of the dummies revealed that LSDV 2 has the highest f-statistics and the lowestpvalue. As a consequence, LSDV2 was the model used to represent FEM. In order to compare and determinewhich between OLS and FEM is the more appropriate model for the data, Wald test was conducted. Theresulting Wald' s statistics revealed a value of 0.9872 while the resulting f-statistics showed 2.3797. It is evidentfrom the results that the Wald statistics is lower than the critical f-value. As a result, the null hypothesis that OLSis the better model is accepted. Hence, OLS is considered the better model, as compared to FEM. Since theresult of the Wald' s test indicate that OLS is the better model, Breusch and Pagan Lagrangian multiplier testwas subsequently conducted to examine whether OLS or REM is the better model. The result, as shown in theabove figure, indicated a p-value of approximately 0.86, which is significant at a equals 0.05. This implies thefailure to reject the null hypothesis that OLS is the better model. Hence, OLS is the resulting model used frompanel analysis, which indicate that this is the model that would be used for the regression analysis. Merger andacquisition displayed a significant negative relation with the return on equity, as evidenced by the significant p-value. The t-statistic of -2.04 connotes a p-value of 0.04, which is significant at a equals 0.05. Consequently, thenull hypothesis that the independent variable has no relation with the dependent variable is rejected. Thecoefficient of merger and acquisition implies that, on average, having a merger or acquisition would decreasethe return on equity by - 153.23. Hence, there is a significant negative relation between merger and acquisitionand the return on equity. The significant negative relation of return on equity to merger or acquisition isconsistent with the study of Kemal (201 1) and Yen and Andre (2010), as well as the apriori expectation, butcontrary to the findings of Wong et. al (2009) and Mantravadi and Reddy (2008). The expectation beforehand isthat having a merger or acquisition would result to the decline in the profitability of the company because of thedisruption on the company's operation, as well as possible desynergistic effects. This expectation is taken fromthe study of Kemal (2011), Yen and Andre (2010) and Williams (2010). Hence, the finding of this variableactually reinforced the results of many previous studies, which promulgates that most mergers and acquisitionsresult to the decline in the profitability of the company. The significant negative relation of having a merger oracquisition to return on equity implies that most mergers and acquisitions do harm to the financial well-being ofthe companies, rather than good. Panel analysis was conducted in order to determine which among the threemodels: ordinary least squares, fixed effects model and random effects model was appropriate for the study.The data was first run using the OLS model and the appropriate sum of the squares of the residuals was noted.Afterwards, the data was run under each of the three variations of the fixed effects model, namely LSDV 1,LSDV 2 and LSDV 3. The model with the highest f-value or lowest p-value among the three LSDV models wasused to represent the FEM. The test of overall significance of the dummies revealed that LSDV 2 has thehighest f-statistics and the lowest pvalue. As a consequence, LSDV2 was the model used to represent FEM. Inorder to compare and determine which between OLS and FEM is the more appropriate model for the data, Waldtest was conducted. The resulting Wald' s statistics revealed a value of 0.5522 while the resulting f-statisticsshowed 2.3797. It is evident from the results that the Wald statistics is lower than the critical f-value. As a result,the null hypothesis that OLS is the better model is accepted. Hence, OLS is considered the better model, ascompared to FEM. Since the result of the Wald' s test indicate that OLS is the better model, Breusch and PaganLagrangian multiplier test was subsequently conducted to examine whether OLS or REM is the better model.The result, as shown in the above figure, indicated a p-value of approximately 0.81, which is significant at aequals 0.05. This implies the failure to reject the null hypothesis that OLS is the better model. Hence, OLS is theresulting model used from panel analysis, which indicate that this is the model that would be used for theregression analysis. Unlike in the previous variable, merger and acquisition provide an insignificant relation tothe return on total assets, as evidenced by the insignificant p-value. The t-statistic of 0.36 translates to a p-valueof approximately 0.72, which is insignificant at a equals 0.05. This suggests that the true coefficient of mergerand acquisition with regards to return on total assets is 0. Hence, merger and acquisition has no relation withthe return on total assets ratio. As opposed to previous expectation, merger and acquisition provided an

08 March 2013 Page 24 of 145 ProQuest

insignificant relation with the return on assets ratio. This finding is consistent with the result of the study of Hu(2009) regarding the relationship of merger and acquisition to the profitability of the company over a 1 yearperiod. It is, however, inconsistent with the results of Altunbas and Ibanes (2004) and Williams (2010). Altunbasand Ibanes (2004) obtained a positive relation between merger and acquisition and return on capital whileWilliams (2010) found a negative relation between merger and acquisition and the profitability of a company.The anticipated relation is that having a merger or acquisition would result to the decline in the profitability of thecompany because of the many researches (Kemal, 2011; Yen &Andre, 2010 and Williams, 2010) suggesting anegative relation between those variables. As a result, the finding of this variable provides empirical evidencethat having a merger and acquisition does not affect the return on assets ratio of companies in the Philippines.CONCLUSIONS The major types of business combination as follows: mergers, consolidations and stockacquisitions. Merger is the combination of two or more entities by purchase acquisition whereby the identity ofone of the entities remain while the others are being dissolved. The reasons behind the merger transactions arebasically gaining market share, competitive advantage, increasing revenues and risk and productdiversifications. With the global financial crises, it is noticeable that mergers and acquisitions have considerablyincreased. Corporations employed such combination not only for the sake of competitiveness but to maintain afirm foothold in the industry as well. This has lead to the significant transformation in the business landscape.Merger and acquisition displayed a significant negative relation with the return on equity, as evidenced by thesignificant p-value. The t-statistic of -2.04 connotes a p-value of 0.04, which is significant at a equals 0.05.Consequently, the null hypothesis that the independent variable has no relation with the dependent variable isrejected. The coefficient of merger and acquisition implies that, on average, having a merger or acquisitionwould decrease the return on equity by - 153.23. Hence, there is a significant negative relation between mergerand acquisition and the return on equity. Unlike in the previous variable, merger and acquisition provide aninsignificant relation to the return on total assets, as evidenced by the insignificant p-value. The t-statistic of 0.36translates to a p-value of approximately 0.72, which is insignificant at a equals 0.05. This suggests that the truecoefficient of merger and acquisition with regards to return on total assets is 0. Hence, merger and acquisitionhas no relation with the return on total assets ratio. The significant negative relation of having a merger oracquisition to return on equity implies that most mergers and acquisitions do harm to the financial well-being ofthe companies, rather than good. Furthermore, merger and acquisition provide an insignificant relation to thereturn on total assets, as evidenced by the insignificant p-value. As a result, the finding of this variable providesempirical evidence that having a merger and acquisition does not affect the return on assets ratio of companiesin the Philippines. References REFERENCES Altunbas, Y. &Ibanez, D. (2004). Mergers and Acquisitions andBank Performance in Europe. The Role of Strategic Similarities. Retrieved September 14, 2011from:http://www.ecb.int/pub/pdf/scpwps/ecbwp398.pdf Dougherty, C. (2007). Introduction to Econometrics. NewYork: Oxford University Press. Hennart (2001). Chapter 5: Theories of the Multinational Enterprise. OxfordHandbook of International Business,127-149. Retrieved from Business Source Complete. Holmstram, B. (2001).Corporate Governance and Merger Activity in the U.S.: Making Sense of the 1980s and 1990s. Journal ofEconomic Perspectives, 15. Retrieved September 14, 2011 from:http://www.hss.caltech.edu/~camerer/BEMEcl46/ Winter07/kaplanholmstromjpe0 1 .pdf Hu, Y. (2009). TheLong-Term Performance of Acquiring Firms: Evidence from Chinese Stock Markets. Retrieved September 14,2011 from: http://arno.unimaas.nl/show.cgi?fid=16865 Jones (1999). Bounded Rationality. Annual ReviewPolitical Science, 1999 (2), 297-321. Retrieved September 7, 2011 from:http://www.princeton.edu/~smeunier/JonesBoundedl.pdf Kemal, M. (2011). Post-Merger Profitability: A Case ofRoyal Bank of Scotland (RBS). International Journal of Business and Social Science, 2. Retrieved September14, 2011 from: http://www.ijbssnet.com/journals/Vol._2_No._5_fSpecial_Issue_-_March_2011J/20.pdf Klein, P.(2006). Does Transaction Cost Economics Need Opportunism? Retrieved September 7, 2011 from:http://organizationsandmarkets.com/2006/10/06/does-transaction-cost-economics-need-opportunism/ Lafosse,

08 March 2013 Page 25 of 145 ProQuest

E. (1999). Consequences of the Choices on the Accounting Methods and the Payment Methods for Acquirersand Targets. A comparative Study between NYSE/AMEX. 4nd NASDAQ for the Nineties. Retrieved September14, 2011 from: http://www.collectionscanada.gc.ca/obj/s4/f2/dskl/tape3/PQDD_0017/MQ47807.pdf Markusen, J.(2009). Internalization Theory. The Princeton Encyclopedia of the World Economy ,2653-659. Retrieved fromProQuest: ABMNFORM Global (XML Gateway). Mkkonen, I. Real Property Transaction Costs from Slovenia.Retrieved September 7, 2011 from: http://www.theslovenian.com/articles/mikkonen.htm Mantravadi, P. &Reddy,A. (2008). Post-Merger Performance of Acquiring Firms from Different Industries in India. InternationalResearch Journal of Finance and Economics. Retrieved September 14, 2011 from: http: //www. euroj ournals .com/irj fe_22_ 1 5 . pdf Ollinger, M., Nguyen, S., Blaynay, D., Chambers, B. &Nelson, K. (2006). Food IndustryMergers and Acquisitions Lead to Higher Labor Productivity. Retrieved September 14, 2011 from:http://www.ers.usda.gov/Publications/ERR27/ Torres-Reya, O. (2006). Panel Data Analysis Fixed and RandomEffects (using StatalO.x). Retrieved July 25, 2011, from Princeton University Web Site:http://dss.princeton.edu/training/Panell01.pdf Singh, H. &Zollo, M. (1999). Post-acquisition Strategies,Integration Capability, and the Economic Performance of Corporate Acquisitions. Retrieved September 14,2011 from: http://openlibrary.org/books/OL18319482M/Postacquisition_strategies_integration_capability_and_the_economic_performance_of_corporate_acquisitions Williams, J. (2010). A Case of a Merger and Acquisition Mega Blind Spot. RetrievedSeptember 14, 2011 from: http://www.alliedacademies.org/public/conferences/Presentations/AMS-Williams.pptWilliamson (1981). The Economics of Organization: The Transaction Cost Approach. The American Journal ofSociology, 87(3). Retrieved from Jstor. What is Transaction Cost Theory? Retrieved September 7, 201 1 from:http://www.businessmate .org/Article.php?ArtikelId=182 Williams, J. A Case Of a Merger and Acquisition MegaBlind Spot. Allied Academies International Conference. Academy of Marketing Studies. Proceedings. Sun-Times News Group. 2010. Retrieved September 13, 2011 at: http://www.highbeam.com/doc/lP3 Wong, ?.,Cheung, K. &Mun, T. (2009). The Effects of Merger and Acquisition Announcements on the Security Prices ofBidding Firms and Target Firms in Asia. International Journal of Economics and Finance, 2. RetrievedSeptember 14, 2011 from: http://www.ccsenet.org/journal/index.php/ijef/article/view/ 3409/3091 Yen, T. &Andre,P. (2010). The Effects of Ownership Structure on Operating Performance of Acquiring Firms in EmergingMarkets. Journal of Business and Policy Research, 5. Retrieved September 14, 2011 from:http://www.essec.edu/faculty /showDeclFileRes.do?declId=9749&key=Publication-Content Yurtoglo, B. TheEffects of Merger. Retrieved September 13, 2011 from: homepage.univie.ac.at/besim.yurtoglu/Mergers_Effects.ppt AuthorAffiliation Rodiel C Ferrer, De La Salle University Subject: Studies; Regression analysis; Acquisitions&mergers; Profitability; Return on equity; Return on assets Location: Philippines Publication title: Academy of Accounting and Financial Studies Journal Volume: 16 Issue: 3 Pages: 31-55 Number of pages: 25 Publication year: 2012 Publication date: 2012 Year: 2012

08 March 2013 Page 26 of 145 ProQuest

Publisher: The DreamCatchers Group, LLC Place of publication: Cullowhee Country of publication: United States Journal subject: Business And Economics--Accounting ISSN: 10963685 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Diagrams;Equations;Tables;References ProQuest document ID: 1036697076 Document URL: http://search.proquest.com/docview/1036697076?accountid=86413 Copyright: Copyright The DreamCatchers Group, LLC 2012 Last updated: 2012-09-14 Database: Accounting&Tax

_______________________________________________________________ Document 3 of 10 Does reduced trade tax revenue affect government spending patterns? Author: Moore, Michael O; Zanardi, Maurizio Publication info: International Tax and Public Finance 18. 5 (Oct 2011): 555-579.ProQuest document link Abstract: Many skeptics of trade liberalization in the developing world argue that lowering trade taxes can causesignificant fiscal pressures in countries particularly reliant on these taxes and result in a reallocation ofresources away from important development goals. This paper evaluates whether there is evidence that centralgovernments systematically change the composition of spending priorities in the wake of lowered trade taxrevenues as a share of total government revenues. We find no systematic evidence for this concern in a sampleof 51 developing countries for the 1991 through 2005 period.[PUBLICATION ABSTRACT] Full Text: Int Tax Public Finance (2011) 18:555579 DOI 10.1007/s10797-011-9191-y Published online: 28 July2011 Springer Science+Business Media, LLC 2011 Abstract Many skeptics of trade liberalization in thedeveloping world argue that lowering trade taxes can cause signicant scal pressures in countries particularlyreliant on these taxes and result in a reallocation of resources away from important development goals. Thispaper evaluates whether there is evidence that central governments systematically change the composition ofspending priorities in the wake of lowered trade tax revenues as a share of total government revenues. We ndno systematic evidence for this concern in a sample of 51 developing countries for the 1991 through 2005period. Keywords Government expenditure Tariff revenue Trade liberalization JEL Classication H7 F13Opponents and proponents of trade liberalization long have argued about the economic impact of open tradeincluding the impact on job losses in import-competing sectors, wage pressures, productivity, and growth. Whileskepticism remains, there nonetheless has been a remarkable degree of liberalization over the last threedecades, especially in countries that traditionally have had high trade barriers. Whatever ones view of theeconomic globalizations impact, a very real issue for policymakers in many developing countries is the impact

08 March 2013 Page 27 of 145 ProQuest

that trade liberalization may have on the public nance position of a government. In particular, many poorercountries have long relied on trade taxes as a major source of revenue for central M.O. Moore Institute forInternational Economic Policy and Department of Economics, George Washington University, Washington, DC,USA M. Zanardi ( ) ECARES, Universit Libre de Bruxelles, Avenue F.D. Roosevelt 50, CP 114, 1050 Brussels,Belgium e-mail: mzanardi@ulb.ac.be Does reduced trade tax revenue affect government spending patterns?Michael O. Moore Maurizio Zanardi 556 M.O. Moore, M. Zanardi government spending. If trade liberalizationreduces tariff and export tax revenues, then governments must either nd other sources of revenue (e.g.,increasing sales or income taxes or faster economic growth) or readjust public expenditure patterns. The lattercould be a reduction across the board with little change in the shares across categories or a substitution awayfrom one group of sectors to others. There is particular concern in some quarters that reduced trade taxrevenues might result in lower public investment in critical areas (e.g., education, health, and infrastructure)thereby arguably undercutting longer term growth and generating domestic social problems. Some observersalso argue that vulnerable groups in societies with less political power might see their share of resources drop.In addition, increased exposure to international competition could also mean that adjustment costs rise asimport-competing rms reduce output and lay off workers. These disruptions could increase the demands on asocial safety net just as government nancial resources may be falling. Critics also have pointed to this problemas a loss of economic sovereignty and have used this as a reason to push back against calls for more tradeliberalization. In short, skeptics argue that increased global economic integration reduces the ability ofgovernments to choose domestic policies they believe are in their national interests. The possibility of scalpressures arising out of trade liberalization is a frequently voiced concern of governments resisting reductions intrade barriers, especially in the developing world. A typical version of this was given by the Saint Kitts and Nevisrepresentative at the Cancun WTO ministerial conference in 2003: Another matter of great concern to smalleconomies and the Caribbean is the threat of tariff reductions and the effect on government revenue. Importtaxes account for as much as 50% of government revenue in some Caribbean countries. It is therefore criticalthat high tariff revenue dependent countries be exempted from further tariff reduction to provide them with theexibility to take the decisions consistent with their development needs. (Saint Kitts and Nevis 2003) A Jamaican2004 Trade Policy Review made a similar argument: The benets of trade liberalization can be outweighed insituations where governments are dependent on these duties as a key component of government revenue. Adrastic reduction of government revenue from customs-related duties and charges could have a negative effecton the scal balance. (Jamaica 2004) We will examine one aspect of this issue in this paper using data from1991 to 2005 for 51 developing countries. In particular, we will explore whether there is any evidence that tradetax revenues as a share of total revenues can explain central government spending patterns. We do notexamine whether trade liberalization is a good or bad policy or whether increased globalization in generalresults in lower spending levels in particular categories. Nor do we examine whether trade liberalization hasresulted in a reduction in overall government revenues. All of these important issues have been addressed inthe previous literature, some of which is discussed below in more detail. Instead, we take trade tax revenue as ashare of total revenue as a given Does reduced trade tax revenue affect government spending patterns? 557and examine whether governments have changed the pattern of spending shares in a systematic way giventhose revenue levels. In other words, we are looking to see if there is a substitution effect away from certainsectors of government spending and toward other sectors as the share of resources from trade taxes changes.The primary source of data for this study is the International Monetary Funds Government Financial Statistics(GFS), which is a compilation of the public nance position of dozens of countries across the globe for the scalyear of each country, obtained through detailed questionnaires sent to each government. The databaseincludes information on the sources and uses of revenue, in principle for national and subnational levels ofgovernment. However, we will only examine spending patterns for central government activities, which is thelevel of government that typically collects trade tax revenue. This is both a practical decision (since data at the

08 March 2013 Page 28 of 145 ProQuest

subnational level are often missing) and substantive (since we are interested in particular in how centralgovernment revenue pressures from trade tax changes affects spending at the national level). While the datahave been collected for some decades, changes in procedures mean that recent data are only comparable from1991 forward.1 Table 1 provides a context for the issue we are addressing. We show a subset of developingcountries experience with trade taxes (i.e., tariffs plus export taxes) as a share of total central governmentrevenues. The table includes countries from Latin America (Argentina, Chile, and Mexico), Asia (India,Indonesia, South Korea, and Thailand), and two island nations (Mauritius and Seychelles). These countriesrepresent a wide variety of experiences, including signicant as well as minimal trade liberalization, steadyeconomic growth and economic crisis, small and large economies, and geographic variation. We see that thisset of countries began the period with an important reliance on trade taxes, with an average of 19.7% of totalrevenues in 1991 with a high of 46% for the Seychelles. There is also a marked reduction in this reliance acrossthese countries over time. For example, the share fell around 65% in South Korea (1991 to 2005), and theSeychelles (1991 to 2002) and approximately 50% for Argentina, Mexico and India (1991 to 2005). Thesecountries, therefore, have all faced important reductions in trade taxes as a share of overall governmentrevenue, which raises the possibility of a realignment of expenditure priorities. Although we do not focus ondeveloped countries in this paper, it is instructive to compare these gures with the experience of industrializedcountries. For example, revenue for trade taxes in the US fell from2.7% to 2% from 1990 to 2003 and from 4%to 1.7% in Canada. Thus, although there are nontrivial percentage changes in the reliance on trade taxes, thelow levels in developed countries throughout the period suggest that government expenditures would not likelybe signicantly affected by changes in trade tax revenues, at least in recent years.2 Table 1 also includes thesum of social protection, health, and education spending for the countries as a share of total central governmentspending. We include these 1For more details about the GFS, please see the Government Financial StatisticsManual available at http://www.imf.org/external/pubs/ft/gfs/manual/index.htm Web End=http://www.imf.org/external/pubs/ft/gfs/manual/index.htm Accessed April 1, 2011.2Similar patterns can bediscerned in Australia (from 4.5% in 1990 to 2.8 in 1998) and in Switzerland(3.0% in 1991 and 2.0 in 2005).Relevant statistics for EU member states are unavailable since customs duties are collected at the EU levelrather than at the national level. 558 M.O. Moore, M. ZanardiTable1Percentageoftradetaxrevenueandpercentageofsocialprotection,health,andeducationspendingArgentinaChileMexicoIndiaIndonesiaSouthKoreaThailandSeychellesMauritius(a)(b)(a)(b)(a)(b)(a)(b)(a)(b)(a)(b)(a)(b)(a)(b)(a)(b) 1991856105784126451392719314635463819928571059847244512727173346344038 199385910597522245136321832403241411994762960654224618631173342334142 19955629614472444187301633433034441996762962447253319630153338293343 1997860862444224316632123241283045199875986345221441443292943343046 1999657765451214314432927243026452000557667450194434113722472746 20014555671643144321039193925452002467154312434103216312146 20031543134282044 20041543114252046 20051453272049Source:GovernmentFinancialStatisticsNotes:Tariffrevenueasashareoftotalcentralgovernmentrevenuesdisplayedincolumn(a).Sumofthethreeexpenditurecategoriesasashareoftotalcentralgovernment expendituresdisplayedincolumn(b) Does reduced trade taxrevenue affect government spending patterns? 559 categories because they are likely candidates for thoseparts of government spending that critics of trade liberalization fear would be squeezed as a result of a drop intrade tax revenues, i.e., governments might substitute away from these sectors since they support morevulnerable and politically less inuential constituents. These also are sectors where one might expect higherspending shares if governments were supplementing efforts to deal with the adjustment costs associated with astronger presence by international competitors subsequent to trade liberalization.3 The patterns are much lessconsistent than in the rst columns, with some countries experiencing a small decrease from 1991 to the last

08 March 2013 Page 29 of 145 ProQuest

year for which data are available (Argentina), while others experience an increase in this broad measure ofsocial spending (e.g., Chile, Mexico, and Mauritius). At the very least, these data suggest that there may not bea simple relationship between a reduced share of trade tax revenues and a broad measure of social spendingas some have feared. We will see below that more detailed descriptive statistics as well as formal econometricresults for 51 developing countries also do not support the argument that central government spending patternsgenerally can be explained by variations in shares of trade tax revenues except in isolated spending categoriesfor specic country groups. These results suggest that developing country governments have not changed thecomposition of how they allocate resources as a consequence of lower import and export tax revenues. Thisindicates that the governments have not reduced the share of social spending, even if the trade liberalizationmight lower the overall level of scal resources. In other words, at least one of the worries often expressed abouttrade liberalization, i.e., that the vulnerable will bear a disproportionate amount of the burden from scalpressures, nds little support in this study. However, another way to interpret this result is that thesegovernments may not have been able to increase the share of spending in these categories, presuming thatneeds for social spending had risen. The key to distinguishing between these two interpretations is what thegovernments optimal composition of spending might be subsequent to a trade liberalization reform. In order tobe able to disentangle these effects empirically, a structural model of government spending would be required,which is outside the scope of this paper but that would be an interesting direction for future research. This paperis organized as follows. After a brief literature review in Sect. 1, we turn to a more detailed discussion ofrelevant descriptive statistics for our sample of developing nations in Sect. 2. Section 3 includes a discussion ofthe econometric methodology and results while robustness checks are reported in Sect. 4. We offer someconcluding remarks in Sect. 5. 1 Relevant literature This work analyzes one component of the broad question ofhow globalization affects domestic economic policies. This vast literature, which includes the impact of 3Analternative expectation would arise out of an application of the Stolper-Samuelson theorem. Since developingcountries are likely to be unskilled labor abundant, trade liberalization would result in higher real wages to thoseworkers, perhaps obviating the need for some social spending. Researchers have found, however, that manydeveloping countries have not experienced a reduction in the wage premium for higher skilled workers. See, forexample, Attanasio et al. (2004) and Cragg and Epelbaum (1996). 560 M.O. Moore, M. Zanardi internationaleconomic integration on monetary policy, tax policy, and environmental policy, just to name a few, is far too vastto summarize here. Our work should be seen in the context of the potential constraints on scal policy whentrade is liberalized. This constraint is composed of a series of possible cause and effect relationships. The rst(Link 1) is that globalization, including reductions in trade tax rates (through international negotiations such asGATT rounds, unilateral liberalization, or those conducted as part of an IMF program), might reduce revenues tothe central government. The central government subsequently might be forced to raise domestic revenues inother ways to make up for the shortfall (Link 2a) or reduce and adjust spending plans (Link 2b).4 The linkbetween trade tax rate reduction and overall tax revenue is a complicated one. First, the tax rate on trade couldbe so high as to trigger a Laffer-curve type effect so that reductions of tariff rates might actually increase tradetax revenues. Second, it is possible that trade liberalization could spur economic growth that could result inhigher overall revenues. Third, governments might readjust the tax burden onto domestic sources so that thescal position might not deteriorate or even improve. Although the issue is crucial for developing countries andcomplex because of the various channels, there is surprisingly little research on the specic link between tradeliberalization and tax revenues. Khattry and Rao (2002) examine one version of Link 1 by evaluating therelationship between a broad measure of openness to the international economy and tax revenues in a panel of80 countries over the period 1970 through 1998. They nd that increased openness (dened here as the ratio oftrade taxes to the volume of total trade) is related both to reduced trade tax revenues and falling overall taxrevenues as shares of GDP. They argue that the reduced state capacity associated with lowered trade taxesand the subsequent reduction in general tax revenues have undercut the ability of developing countries to use

08 March 2013 Page 30 of 145 ProQuest

scal policies to deal with important development problems. However, they do not explore the effects of thedropping revenues on particular spending categories. Baunsgaard and Keen (2010) focus on the relationshipbetween declining trade revenue and increased revenues from other sources but focus specically on whethercountries are able to replace falling trade tax revenues with revenue from other sources (i.e., they examine Link2a). Their study utilizes data on over one hundred countries for 1975 to 2006.5 The authors nd evidence thathigh- and middle-income countries generally have been able to nd ways to make up for any loss of trade taxrevenues through other sources. Developing countries, on the other hand, have had 4Another related researchquestion concerns the relationship between trade openness and governments size. Rodrik (1998) argues thatthe higher income risks associated with an open economy can lead to higher levels of government spending tomitigate those risks. However, globalization may also lead to the shrinking of governments size to ensure highercompetitiveness in the international market (i.e., efciency hypothesis). We do not address this issue since wetake governments size as given and investigate possible changes it its composition. See Schulze and Ursprung(1999) for a survey of the broad issues surrounding international economic integration and the functions ofnational governments.5They augment GFS data with other information provided to the IMF as part of its generalmonitoring programs under Article IV of the IMFs charter. Does reduced trade tax revenue affect governmentspending patterns? 561 much more difculty doing so. This problem is particularly acute for low-income countrieswhere they estimate that governments were less able to replace lost trade tax funds through other revenue.6These results suggest that concerns about scal pressures arising out of trade liberalization episodes reectrealities, at least for poorer countries. This in turn may create problems for poor countries efforts usinggovernment spending to achieve development goals or to compensate losers from trade liberalization. However,the authors do not examine whether such a change in the composition of public spending is evident in the data.Dreher et al. (2008) focus on a combination of Link 1 and Link 2b. They examine how disaggregated categoriesof government spending have been affected by increased globalization measured in a variety of ways.7 They ndvery little evidence that government spending has been constrained by increased globalization. In particular,they nd that four different measures of globalization have very little explanatory power in predicting the spendingstructure in a broad sample of 60 countries from 19712001 as well as a more detailed (in terms of expenditurecategories) but smaller (i.e., only ten countries) OCED sample from 1990 through 2001. These results suggestthat international economic integration may have relatively little impact on how governments spend their funds,i.e., independence of scal action generally remains robust. We contribute to the literature by focusing specicallyon Link 2b, i.e., we focus exclusively on how the share of trade tax revenues in total revenue affects thecomposition of central government spending across nine separate functional categories. Baunsgaard and Keen(2010) have provided evidence that there may be a scal shock, especially in the poorest countries, when tradetaxes are lowered. We are interested, however, in whether this scal shock constrains how central governmentsallocate resources across alternative uses. For example, do governments substitute away from social spendingas the budget constraint becomes tighter? Evidence of such a substitution might suggest a further deteriorationin the governments pursuit of development goals. The absence of such evidence would suggest a continuedability of governments to allocate resources (albeit potentially smaller) subsequent to trade liberalization.Alternatively, do governments shift resources toward social spending in order to reinforce a safety net to dealwith the consequences of reallocation of resources subsequent to trade liberalization in what Gemmell et al.(2008) call the compensation effect? These questions mean that our work is similar to Dreher et al. (2008)although we look directly at trade taxes and not broadly dened globalization measures. Note also that we do notspecically account for whether globalization and trade liberalization per se are the causes of any falling trade taxrevenue. Moreover, we are able to focus on 51 developing countries, which we argue are more at risk whenfaced with trade-related scal shocks, over a longer sample period. Gemmell et al. (2008), in a closely 6Emranand Stiglitz (2005) examine the issue of replacing trade taxes with domestic income taxes. They argue that thelarge informal sectors in many low income countries means that increased VAT taxes on the formal sector could

08 March 2013 Page 31 of 145 ProQuest

actually reduce national welfare relative to retaining trade taxes.7The globalization measures include: (1)exports plus imports as a share of GDP; (2) inward and outward FDI share of GDP; (3) restrictions on capitalaccounts; and (4) the Swiss Economic Institutes index of globalization. 562 M.O. Moore, M. Zanardi relatedpaper, focus on the impact of globalization (measured by the stock of inward FDI as well as trade as a share ofGDP) on government expenditure patterns. They nd evidence that increased FDI shifts government spendingtoward social spending. Their paper also differs substantially from ours in that it focuses solely on OECDcountries. 2 Overview of trade revenue and spending patterns The basis of this study is information on thedisaggregated level of central government spending as reported in the IMFs GFS database. The categories forthe functional classications of central government spending used as the dependent variables in the analysisinclude the following nine sectors: social protection (e.g., unemployment, old age pensions, disability), health,education, economic affairs (e.g., economic and labor affairs, infrastructure, communications), general publicservices (e.g., executive and legislative branch costs, foreign aid), defense, public order (e.g., police, re,prisons, and law courts), housing and community amenities (e.g., housing development, water supply), andrecreation.8 The data used in the analysis are developed using a cash-accrual basis, i.e. ows reect the year inwhich a change in the economic value occurs.9 Table 2 displays basic information from our data set about tradetax revenues (as a share of total central government revenue) and spending patterns for a subset of spendingcategories (as a share of central government expenditures). These tables make clear that there have beensubstantial differences across countries.10 For example, African countries were almost twice as reliant on tradetax revenue as the entire group of countries in the sample. This African reliance does fall over time though thelevel remains high relative to other developing countries in the dataset. Latin American and Asian countriesreliance on trade taxes also fell over time. Table 2 also includes information based on countries income level, asdened by the World Bank. We see a clear increase in the reliance on trade taxes as countries average incomelevel decreases. All income groups experienced a reduction in the percentage of revenues derived from tradetaxes. However, this drop was greatest for lower-middle income countries and the smallest for low-incomecountries. This also reects the pattern that the poorest countries in the world have liberalized the least. 8Inprinciple, the GFS data also report environmental spending. However, we do not include it in our analysisbecause of missing data in many countries.9One downside to this data is that there is spotty availability ofinformation for many of the countries covered by the GFS in principle. Consequently, it is important to note thatthe results of this study do not necessarily apply to all developing countries but only to those actually used in theeconometric study. 10Note that these data are for all developing countries reporting information in the GFS andare not identical to the data used in the formal econometric work. (See Appendix for the list of countriesincluded in the descriptive statistics and in the econometric analysis.) The broad patterns are quite similar in thetwo samples. Detailed statistics for the observations included in the regressions are reported with the estimationresults. Does reduced trade tax revenue affect government spending patterns? 563 Table 2 Shares of trade taxand selected expenditure categories Trade tax share 19912005 19911997 19982005 All countries 11.7 14.5 8.7Latin America 10.3 11.5 8.5 Africa 22.4 24.4 16.5 Asia 12.1 12.8 11.5 Low income 16.8 19.1 14.0 Lower middleincome 11.8 15.0 8.1 Upper middle income 7.8 9.5 6.1 Social protection 19912005 19911997 19982005 Allcountries 20.4 18.1 22.9 Latin America 28.4 29.2 27.2 Africa 10.7 7.8 19.3 Asia 4.4 3.6 5.1 Low income 2.7 2.82.6 Lower middle income 23.3 18.7 28.4 Upper middle income 27.1 29.1 25.4 Education 19912005 1991199719982005 All countries 11.6 12.2 10.9 Latin America 14.7 13.6 16.4 Africa 15.3 14.6 17.4 Asia 12.2 13.5 10.9Low income 10.9 11.0 10.7 Lower middle income 12.0 13.1 10.5 Upper middle income 11.5 11.3 11.7 Health19912005 19911997 19982005 All countries 7.4 7.1 7.7 Latin America 9.3 9.3 9.4 Africa 5.6 5.1 6.8 Asia 4.1 4.23.9 Low income 4.7 4.2 4.9 Lower middle income 8.2 7.9 8.3 Upper middle income 8.1 7.7 8.5 Economic affairs19912005 19911997 19982005 All countries 16.9 19.0 14.5 Latin America 13.1 14.3 11.2 Africa 17.1 18.0 14.2Asia 25.6 28.9 22.5 Low income 24.5 27.0 21.2 Lower middle income 16.3 18.4 13.9 Upper middle income 12.513.2 11.9 Source: Government Financial Statistics Notes: Trade tax revenue as a share of total central

08 March 2013 Page 32 of 145 ProQuest

government revenues. Spending categories as share of total central government expenditure. See Appendix forcountry group membership 564 M.O. Moore, M. Zanardi These reductions in trade tax revenues clearly mightcause some changes in spending patterns especially if, as Baunsgaard and Keen (2010) have argued, the dropin trade tax revenue have not always been matched by increases from other sources for low-income countries.We focus our discussion here on four categories of expenditures about which globalizations skeptics might beparticularly concerned: social protection, education, health, and economic affairs. We see in the lower fourremaining panels of Table 2 that there are marked differences across country groups in central government scalexpenditures. For example, Latin America has by far the highest percentage of spending in social protectionand health of the three continents displayed for the period as a whole and also for the two subsamples. Aloneamong these three, Latin America is the only one where social spendings share of total expenditures fellbetween the two periods (but only by 2%). Average spending on economic affairs, a very broad category thatincludes items such as labor, agriculture, mining, transport, and communications fell by 3 percentage pointsacross the two time periods while education spending rose by2.8 percentage points in our sample. In short,Latin American countries reliance on trade tax revenues fell substantially over the years of the sample and thereis some evidence of reallocation of spending during this same period, though there are few indications of broadexpenditure retrenchment for these sectors as a whole. In fact, the total share of these four sectors remainedbroadly unchanged at 65% of central government spending. African countries in the sample allocated a muchlarger share of their spending to education, social protection, and health in the latter period while economicaffairs spending in Africa fell during this same timeframe. On the surface, these results are suggestive thatAfrica, with less trade liberalization and less scal pressure from trade tax revenue decreases, may have beenable to devote more resources to scal categories supportive of human development. Our sample of Asiancountries shows a different pattern, with lower trade tax revenues as in Latin America but markedly differentspending patterns. We see that social protection spending shares increased, while education, health, andeconomic affairs spending shares decreased. The panels of Table 2 show that the reliance on trade taxrevenues varies substantially across country income categories. However, there has been a broad reduction intrade tax share for all income categories over time. On the other hand, there has been very little change in theshare of the four spending categories as a whole, with the total more or less around 60% for all three incomegroups. These aggregate numbers do not reect some important changes in individual categories. For example,lower-middle income countries experienced an increase in social spendings share while education spendingfell. Upper-middle income countries spent more on health in the 19982005 period compared to 19911997. Insum, we see clear evidence that these groupings of countries in our sample have experienced a decrease,sometimes sharp, in the share of total revenue that central governments obtain from trade taxes. These guresmirror those displayed for individual countries in Table 1. However, we see less specic evidence that thereDoes reduced trade tax revenue affect government spending patterns? 565 has been a systematic reduction insensitive spending sectors that might come under pressure with a nancial squeeze, at least for these broadcategories of countries. There is also little evidence that trade liberalization episodes have resulted in a greatershare of spending to deal with dislocation from increased international competition. For example, LatinAmerican governments, typically implementing signicant trade liberalization in this period, do not seem to bemoving the share of government spending toward categories that would provide a strengthened safety net. Wewill now turn to a more detailed econometric study of individual countries while controlling for other factors thatcan help explain government spending patterns. 3 Econometric methodology and results We do not develop aformal theoretical model for the determination of government spending but instead posit a reduced formrelationship. However, we make the reasonable assumption that government spending shares in a particularcategory are not independent of the previous years share. Thus, the basic econometric approach is verystraightforward. The dependent variable is country is government spending share (out of total central spending)in each expenditure category j (out of the nine categories for which data is available) in year t, which we assume

08 March 2013 Page 33 of 145 ProQuest

can be expressed as a linear function of relevant covariates: Gi,j,t = Gi,j,t1 + Xi,t1 + i + t + i,j,t (1) where Gi,j,t1 isthe lagged value of the dependent variable, Xt1 is a matrix of (one-year lagged) explanatory variables withassociated slope coefcients and , respectively. The disturbance term is given by i,j,t , while i and t are coun tryand time xed effects (reecting unobservables), respectively. Because of the lagged dependent variable, thecoefcients on the explanatory variables represent the short-run adjustment of government spending while /(1 ) isthe long run effect.11 The inclusion of the lagged dependent variable together with xed effects lead to apossible bias, especially because the time dimension of our panel is not particularly long. To deal with this bias,we will use as our basic specication the bias-correction 11There may be concerns that the dependent variablesare not stationary. In statistical tests not reported here, we performed a battery of tests to explore thispossibility. In particular, we followed Choi (2001) and used the inverse chi-squared, inverse normal, and inverselogit tests for two to seven lags (chosen according to the rules suggested by Schwert 1989) for the dependentvariable. We also followed Levin et al. (2002) and conducted the tests with and without demeaned series. Theresults of these tests provide little formal evidence that the dependent variables are non-stationary. 566 M.O.Moore, M. Zanardi estimator proposed by Bruno (2005) for unbalanced panels.12 We also present the systemGMM estimator suggested by Arellano and Bover (1995) and Blundell and Bond (1998) as an alternative modelfor our baseline dataset for the entire sample of countries.13 In the robustness section, we will also considerspecications without the countries xed effects. Our principal focus is on the impact of trade taxes revenue shareon these government expenditure categories. We turn once again to the GFS for these data. Table 3 containsdetailed variable denitions and sources while summary statistics for the observations included in theregressions are reported with the estimation results. In particular, trade tax revenue is the sum of revenue fromimport tariffs, export taxes, prots of exports, or import monopolies, etc., divided by total revenues of the centralgovernment. In the econometric analysis, a positive coefcient on trade tax revenue would indicate that a drop intrade tax revenue share would result in a diminished share of that particular sector in overall central governmentspending. We will also control for a number of other factors. In the basic specication, we include, as notedabove, the lagged dependent variable for each spending category, which measures the persistence ofspending. In addition, we include GDP per capita (calculated in natural logs for constant 2005 US $) to controlfor the possibility that the level of economic development could affect spending in the various categories. Wefollow Dreher et al. (2008) and include total government expenditures as a share of GDP (expenditure share) toaccount for the possibility that the level of total government spending in an economy might inuence the share ofspending. Two more regressors control for macroeconomic conditions: annual real growth in GDP (growth) andination (ination), the latter of which equals the annual change in the GDP deator for time t(t) divided by (1 + t1),which will help adjust for countries experiencing high levels of ination. A countrys age population distributionalso might inuence spending in many of these categories. Consequently, we include the age dependency ratio,which is the ratio of the population younger than 15 and older than 65 years to the population between 15 and64 (i.e., the working population). Finally, the basic specication also includes country and year xed effects toaccount for unobserved variables. Table 4 includes the basic set of regressions for all countries in the data setas well as descriptive statistics for the data used in the regressions. All estimated equations include country andyear xed effects.14 The rst four columns of this and the other regression tables include the results for thesensitive sectors discussed above: social protection, health, education, and economic affairs. The coefcient ontrade tax revenues is not signicantly different from zero for any of these four sectors so that there is no evidencein the base specication 12The standard errors are found by bootstrapping. We report results with 500repetitions. Results are identical with 200 repetitions.13The use of the system (instead of difference) GMMestimator is justied because of the likely persistence of the expenditure shares.14See Appendix for a list of thecountries included in the sample. Does reduced trade tax revenue affect government spending patterns? 567DependentVariablesa IMF Government Financial StatisticsSocialprotectionSocialprotectionexpenditureIMFGovernmentFinancialStatistics

08 March 2013 Page 34 of 145 ProQuest

HealthHealthexpenditureIMFGovernmentFinancialStatisticsEducationEducationexpenditureIMFGovernmentFinancialStatisticsEconomicaffairsEconomicaffairsexpenditureIMFGovernmentFinancialStatisticsPublicservicesGeneralpublicservicesexpenditureIMFGovernmentFinancialStatisticsDefenseDefenseexpenditureIMFGovernmentFinancialStatisticsPublicorderPublicorderandsafetyexpenditureIMFGovernmentFinancialStatisticsHousingHousingandcommunityamenitiesexpenditureIMFGovernmentFinancialStatisticsRecreationRecreation,culture,andreligionexpenditureIMFGovernmentFinancialStatistics ExplanatoryVariablesTradetaxrevenueTradetaxrevenue(%oftotalcentralgovernmentrevenue)IMFGovernmentFinancialStatisticsTotalexpenditures(%GDP)TotalcentralgovernmentexpendituresasashareofGDPWorldDevelopmentIndicatorsGDPpercapitaLogofGDPpercapita,PPP(constant2005international$)WorldDevelopmentIndicatorsInation/(1+)whereistheannualchangeinGDPdeatorWorldDevelopmentIndicatorsGrowthAnnualgrowthrateofGDPWorldDevelopmentIndicatorsAgedependencyratio(populationyoungerthan15andolderthan65)/populationbetween15and64WorldDevelopmentIndicatorsTotalgovernmentrevenue(%GDP)TotalcentralgovernmentrevenueasashareofGDPWorldDevelopmentIndicatorsAID(%GNI)Aid(%ofGNI)WorldDevelopmentIndicatorsIBRDandIDAcredit(%GDP)IBRDloansandIDAcredits(%GDP)WorldDevelopmentIndicatorsPolityIndexrangingfrom10(autocracy)to+10(democracy)MontyandJaggers(2002) Inter-stateconictsDummyvariableequalto1forinter-stateconicts(ofhostilitylevel>2)CorrelatesofWarproject Intra-stateconictsNumberofconcurrentlyactiveintra-stateconictsUCDP/PRIOArmedConictdataset a All dependentvariables refer to central governments expenditures and are calculated as percentage shares of total centralgovernment expenditures Table3Datadescription DescriptionSource 568 M.O. Moore, M. ZanardiLaggeddependentvariable0.704***0.679***0.753***0.758***0.867***0.665***0.645***0.842***0.764***(0.051)(0.061)(0.049)(0.053)(0.055)(0.053)(0.059)(0.054)(0.055)Tradetaxrevenue0.0320.0480.0320.0350.0570.0490.0100.0110.017(0.10/0.09)(0.051)(0.033)(0.029)(0.066)(0.092)(0.040)(0.022)(0.030)(0.012)Totalexpenditures(%GDP)0.0740.0060.0220.1310.0060.0090.0050.0030.008(0.27/0.10)(0.064)(0.041)(0.036)(0.082)(0.112)(0.050)(0.027)(0.036)(0.014)GDPpercapita0.051*0.0110.0070.0260.102**0.069***0.0150.0060.002(8.61/0.95)(0.028)(0.018)(0.016)(0.036)(0.049)(0.022)(0.012)(0.016)(0.006)Ination0.0200.0010.0040.0230.0350.029**0.0100.0050.001(0.15/0.20)(0.018)(0.012)(0.011)(0.024)(0.032)(0.014)(0.008)(0.011)(0.004)Growth0.0280.0040.0370.0330.1370.083**0.0160.0310.008(0.04/0.05)(0.048)(0.030)(0.027)(0.061)(0.084)(0.037)(0.020)(0.027)(0.011)Agedependencyratio0.100*0.073*0.088**0.0440.1200.0250.0220.0030.008(0.59/0.15)(0.059)(0.040)(0.036)(0.077)(0.107)(0.047)(0.026)(0.035)(0.013)Yeareffects(Prob.>F)0.020.000.030.060.000.230.020.680.93 Observations418418418418418418418418418Notes:Resultsofbias-correctedleast-squaredummyvariableestimations(seeBruno2005).Bootstrappedstandarderrorsinparenthesiscalculatedwith500repetitions;***,**,and*denotesignicanceat1,5,and10%,respectively.Prob.>Freportstheprobabilityofthejointtestthattheyearxedeffectsarezero;allvariablesarelaggedbyoneyear.51countriesincluded;countrieswithfewerthan3observationsexcluded Table4AllcountriesSocialHealthEducationEconomicPublicDefensePublicorderHousingRecreation protectionaffairsservicesExplanatoryvariable

08 March 2013 Page 35 of 145 ProQuest

(mean/standarddeviation)(0.22/0.16)(0.08/0.05)(0.12/0.59)(0.17/0.10)(0.24/0.13)(0.08/0.07)(0.05/0.03)(0.03/0.04)(0.02/0.02) Does reduced trade tax revenue affect government spending patterns? 569 that scal pressuresfrom trade liberalization have caused a retrenchment in spending in these categories nor do governments moveresources into those sectors.15,16 The other results in Table 4 generally correspond to our expectation that thelagged value of the dependent variable is an important predictor of later spending in that category. We nd thatGDP per capita and age dependency ratio are the regressors most frequently statistically signicant. Forexample, an increase in those dependent on the working population will increase spending on health andeducation while causing a reduction on expenditures for social protection. The coefcient on GDP per capita isstatistically signicant (at least at a 10% level) and negatively related to expenditure on social protection anddefense and positively related to public services. The inclusion of the lagged dependent variable with countriesxed effect in a panel with a short time dimension may lead to a bias and the correction proposed by Bruno(2005) that we employ is designed to deal with this problem. As an alternative we have reestimated thespecications in Table 4 employing the system GMM estimator proposed by Arellano and Bover (1995) andBlundell and Bond (1998). However, the use of the system GMM estimator poses problems of its own,especially because it is designed for samples with a large number of individuals (i.e., countries in ourapplication). For this reason, we will only employ it on the full sample of 51 countries since the subsamples (byincome and geographical levels) that we will consider below are based on a limited number of countries. Theresults using system GMM are reported in Table 5. Also in this case, the trade tax revenue does not helpexplain government spending shares. Only the coefcients on education and public services are signicantlydifferent from zero and in both cases positive. Thus, we do not nd consistent nor systematic patterns for tradetax revenues explaining expenditure shares across econometric specications. Table 5 also suggests verymodest effects for the variables with signicant coefcients. The short-run effect of 1 percent reduction in tradetaxes on education is 0.07 while the long run effect is 0.11. As for the performance of the system GMMestimator, the tests for rst- and second-order autocorrelation do not reject the specication. We follow thesuggestion by Roodman (2009) for a parsimonious choice of instruments: we use only two lags and alsocollapse the instruments.17 We report the results based on only two lags because including more than two lagswould lead to signicant values of the Sargan tests that would reject the validity of the instruments for severalspending categories. With two lags, the Sargan test of overidentifying restrictions rejects the null hypoth- 15Thepoint estimates also suggest that the short run effects are quite small at the margin irrespective of the level ofstatistical signicance. For example, a 1% drop in tariff revenue reduces central government spending on healthonly by 0.0485%. The long-term effects (i.e. /(1 )) is a 0.150% drop in health expenditures for every 1% drop intrade tax revenue.16It is possible that the effects of reduced trade revenue might take more than one year towork their way into expenditure patterns. In results not reported here but available on request, we included 2-year lags for trade tax shares, which leads to a reduction in the number of observations. Once again, there is nosystematic relationship between spending categories and falling trade tax revenues.17When collapsing theinstruments, the columns of the instrumenting matrix are added in a horizontal way. 570 M.O. Moore, M.Zanardi Laggeddependentvariable0.930***0.595***0.591***1.146***0.906***0.697***0.826***0.763***0.909***(0.170)(0.123)(0.140)(0.122)(0.087)(0.222)(0.285)(0.187)(0.073)Tradetaxrevenue0.0070.0230.065**0.0020.062***0.0170.0090.0330.004(0.10/0.09)(0.054)(0.021)(0.031)(0.031)(0.019)(0.038)(0.014)(0.047)(0.005)Totalexpenditures(%GDP)0.0330.066**0.0210.0430.0310.0290.0070.0070.002(0.27/0.10)(0.044)(0.030)(0.032)(0.033)(0.027)(0.035)(0.017)(0.014)(0.003)GDPpercapita0.0030.006**0.011**0.0030.0040.0100.0010.0030.001***(8.61/0.95)(0.007)(0.003)(0.005)(0.003)(0.005)(0.008)(0.003)(0.002)(0.000)Ination0.0300.0030.034**0.0130.039**0.0060.0060.0020.001(0.15/0.20)(0.033)(0.009)(0.013)(0.015)(0.015)(0.014)(0.008)(0.007)(0.002)

08 March 2013 Page 36 of 145 ProQuest

Growth0.0140.0100.0000.0310.0160.0360.0010.0150.003(0.04/0.05)(0.034)(0.019)(0.035)(0.044)(0.059)(0.027)(0.024)(0.015)(0.006)Agedependencyratio0.0240.0090.054*0.0090.0080.0020.0010.0110.001(0.59/0.15)(0.053)(0.015)(0.029)(0.022)(0.021)(0.038)(0.012)(0.014)(0.001)Yeareffects(Prob.>F)0.060.000.440.010.010.020.600.010.00 Instruments232323232323232323 P-valueAR(1)rstdiff.0.000.000.000.000.000.210.020.020.03 P-valueAR(2)rstdiff.0.110.070.160.130.470.440.260.810.85 P-valueSargantestofoverid.0.010.450.940.170.900.230.010.390.12 P-valueHansentestofoverid.0.180.670.970.500.950.780.130.750.77Observations418418418418418418418418418 Notes:System-GMMregressionresultsusing2lagsforinstrumentingthelaggeddependentvariableandcollapsingtheinstruments.Standarderrorsinparenthesis;***,**,and*denotesignicanceat1,5,and10%,respectively.Yearxedeffectsincludedineveryregression;Prob.>Freportstheprobabilityofthejointtestthatthexedeffectsarezero;allvariablesarelaggedbyoneyear.51countriesincluded;countrieswithfewerthan3observationsexcluded Table5Allcountries(system-GMMregressions)SocialHealthEducationEconomicPublicDefensePublicorderHousingRecreation protectionaffairsservicesExplanatoryvariable(mean/standarddeviation)(0.22/0.16)(0.08/0.05)(0.12/0.59)(0.17/0.10)(0.24/0.13)(0.08/0.07)(0.05/0.03)(0.03/0.04)(0.02/0.02) Does reduced trade tax revenue affect government spending patterns? 571 esis of joint validity ofthe instruments for only one expenditure category (health) while the Hansen test never rejects theoveridentifying restrictions.18 Tables 6, 7, and 8 include the results when the full sample is split into low-, lowermiddle-, and upper-middle-income countries, respectively. The results are broadly similar to those of Tables 4and 5. There are isolated instances where the coefcient on trade tax share for a particular sensitive spendingcategory share is signicantly different from zero for specic country income groups (e.g., health for lower-middleincome countries at a 10% level, and housing for upper-middle income countries). Most importantly, we see noconsistent evidence across income groups for spending categories. We focus now on the results for low-incomecountries in Table 6, which may be under particular scal distress given that the evidence reported byBaunsgaard and Keen (2010) and Khattry and Rao (2002) that low-income countries in particular have haddifculties in nding revenue substitutes for falling trade tax revenues. The results of Table 6 suggest, however,that the trade tax revenue share generally does not explain the spending share in the expenditure categories(with recreation spending as the sole exception) for low-income countries. This does not mean that falling tradetax revenues might not result in reduced spending in any of these particular categories. However, on averagethese governments seem to share the burden equally across sectors. In results not reported here but availableupon request, we also run the basic spec-ication separately for Latin America, Asia, and Africa. We nd noinstance where the level of trade tax revenue as a share of total revenue has statistically signicant explanatorypower for the various spending categories. However, each of these sub-samples has a small number ofcountries so that the results should be treated with caution. 4 Robustness checks We have undertaken twotypes of robustness checks. On a methodological front, we want to mitigate the possibility of a bias due to theinclusion of the lagged dependent variable with countries xed effect in a panel with a short time dimension.Since the use of a system GMM estimator (e.g., Table 5) is not without caveats, we have reestimated thespecication in Table 4 without countries xed effect. In another set of robustness checks, we have added variousregressors to the basic specication. Neither set of sensitivity test alters the basic conclusion reached in theprevious section that tariff revenue shares generally do not affect central government expenditure shares acrossa variety of categories. The results are omitted here but are available upon request. The exclusion of thecountries xed effects does not lead to any major qualitative change for the effect of tariff revenue on expenditure

08 March 2013 Page 37 of 145 ProQuest

shares, which remains insignificant with the exception for education (at 5% level) and public services (at 10%18The former is robust to a large number of instruments but not to heteroskedasticity while it is the opposite forthe latter. 572 M.O. Moore, M. ZanardiLaggeddependentvariable0.900***0.557***0.905***0.916***0.836***0.480***0.764***0.683***0.599***(0.154)(0.156)(0.149)(0.149)(0.155)(0.157)(0.150)(0.153)(0.143)Tradetaxrevenue0.0200.0060.0680.2260.1340.1660.0310.0270.051*(0.13/0.10)(0.092)(0.054)(0.084)(0.261)(0.388)(0.213)(0.057)(0.130)(0.030)Totalexpenditures(%GDP)0.0840.0950.0400.3180.0920.1050.0060.0310.037(0.22/0.11)(0.118)(0.070)(0.113)(0.309)(0.484)(0.291)(0.075)(0.171)(0.042)GDPpercapita0.0090.0070.0290.0280.1480.239**0.0070.0720.001(7.26/0.82)(0.053)(0.030)(0.049)(0.137)(0.218)(0.121)(0.033)(0.076)(0.017)Ination0.0230.0180.0290.0280.1800.0700.0360.0290.016(0.13/0.20)(0.048)(0.029)(0.046)(0.127)(0.192)(0.117)(0.030)(0.070)(0.017)Growth0.0780.0080.1190.1040.695*0.3590.0110.0320.044(0.04/0.05)(0.104)(0.060)(0.099)(0.275)(0.419)(0.251)(0.065)(0.155)(0.036)Agedependencyratio0.0070.0520.1240.0270.0100.0060.0300.1430.051(0.75/0.14)(0.144)(0.083)(0.131)(0.384)(0.608)(0.348)(0.092)(0.216)(0.055)Yeareffects(Prob.>F)0.650.000.360.530.950.740.540.390.45 Observations797979797979797979Notes:Resultsofbias-correctedleast-squaredummyvariableestimations(seeBruno2005).Bootstrappedstandarderrorsinparenthesiscalculatedwith500repetitions;***,**,and*denotesignicanceat1,5,and10%,respectively.Prob.>Freportstheprobabilityofthejointtestthattheyearxedeffectsarezero;allvariablesarelaggedbyoneyear.10countriesincluded;countrieswithfewerthan3observationsexcluded Table6Low-incomecountriesSocialHealthEducationEconomicPublicDefensePublicorderHousingRecreation protectionaffairsservicesExplanatoryvariable(mean/standarddeviation)(0.03/0.03)(0.05/0.03)(0.11/0.06)(0.23/0.13)(0.36/0.18)(0.13/0.10)(0.03/0.3)(0.05/0.05)(0.01/0.02) Does reduced trade tax revenue affect government spending patterns? 573Laggeddependentvariable0.782***0.712***0.764***0.756***0.869***0.768***0.400***0.964***0.604***(0.073)(0.081)(0.068)(0.083)(0.074)(0.062)(0.086)(0.060)(0.083)Tradetaxrevenue0.0600.094*0.0600.0030.1630.0170.0070.0150.006(0.11/0.10)(0.089)(0.056)(0.041)(0.095)(0.115)(0.045)(0.032)(0.025)(0.010)Totalexpenditures(%GDP)0.0000.0240.0270.0470.0030.0340.0020.0130.012(0.26/0.09)(0.115)(0.073)(0.055)(0.123)(0.151)(0.059)(0.044)(0.032)(0.013)GDPpercapita0.0780.0000.0020.0310.0870.0080.0260.0200.001(8.65/0.59)(0.058)(0.037)(0.028)(0.062)(0.074)(0.029)(0.022)(0.016)(0.007)Ination0.0250.0030.0060.0220.0410.028*0.0160.0030.000(0.14/0.19)(0.030)(0.019)(0.014)(0.031)(0.039)(0.015)(0.011)(0.008)(0.003)Growth0.0010.0270.0720.1200.0480.0200.070**0.0280.005(0.04/0.05)(0.096)(0.062)(0.046)(0.100)(0.125)(0.049)(0.035)(0.028)(0.011)Agedependencyratio0.0590.0720.093*0.0340.1110.0320.0080.0070.003(0.59/0.14)(0.099)(0.063)(0.053)(0.103)(0.129)(0.047)(0.034)(0.026)(0.011)Yeareffects(Prob.>F)0.000.000.190.380.000.430.000.380.12 Observations214214214214214214214214214Notes:Resultsofbias-correctedleast-squaredummyvariableestimations(seeBruno2005).Bootstrappedstandarderrorsinparenthesiscalculatedwith500repetitions;

08 March 2013 Page 38 of 145 ProQuest

***,**,and*denotesignicanceat1,5,and10%,respectively.Prob.>Freportstheprobabilityofthejointtestthattheyearxedeffectsarezero;allvariablesarelaggedbyoneyear.26countriesincluded;countrieswithfewerthan3observationsexcluded Table7Lower-middleincomecountries SocialHealthEducationEconomicPublicDefensePublicorderHousingRecreationprotectionaffairsservices Explanatoryvariable(mean/standarddeviation)(0.25/0.13)(0.08/0.05)(0.12/0.06)(0.16/0.09)(0.21/0.10)(0.07/0.06)(0.06/0.02)(0.03/0.03)(0.01/0.01) 574 M.O. Moore, M. ZanardiLaggeddependentvariable0.586***0.594***0.633***0.686***0.720***0.661***0.924***0.767***0.712***(0.094)(0.107)(0.102)(0.117)(0.110)(0.102)(0.090)(0.119)(0.108)Tradetaxrevenue0.0120.0200.0180.2170.2360.0080.0450.143**0.003(0.05/0.04)(0.120)(0.086)(0.074)(0.147)(0.214)(0.059)(0.048)(0.070)(0.045)Totalexpenditures(%GDP)0.1660.0150.0020.282*0.1540.0690.0150.0160.025(0.30/0.10)(0.117)(0.085)(0.075)(0.149)(0.217)(0.058)(0.047)(0.073)(0.042)GDPpercapita0.0530.0180.0020.0280.0450.0220.0090.0350.016(9.40/0.50)(0.050)(0.036)(0.031)(0.064)(0.090)(0.025)(0.020)(0.031)(0.018)Ination0.0030.0160.0190.0280.0280.0050.0060.0210.000(0.17/0.23)(0.034)(0.024)(0.021)(0.042)(0.061)(0.017)(0.014)(0.021)(0.012)Growth0.0560.0160.0290.0200.0790.0320.0220.0380.013(0.04/0.06)(0.079)(0.056)(0.048)(0.097)(0.138)(0.038)(0.032)(0.046)(0.028)Agedependencyratio0.2050.0500.0640.0650.0500.0170.0260.1540.040(0.50/0.08)(0.261)(0.192)(0.170)(0.346)(0.500)(0.129)(0.106)(0.170)(0.096)Yeareffects(Prob.>F)0.130.570.730.170.510.400.940.440.97 Observations125125125125125125125125125Notes:Resultsofbias-correctedleast-squaredummyvariableestimations(seeBruno2005).Bootstrappedstandarderrorsinparenthesiscalculatedwith500repetitions;***,**,and*denotesignicanceat1,5,and10%,respectively.Prob.>Freportstheprobabilityofthejointtestthattheyearxedeffectsarezero;allvariablesarelaggedbyoneyear.15countriesincluded;countrieswithfewerthan3observationsexcluded Table8Upper-middleincomecountries SocialHealthEducationEconomicPublicDefensePublicorderHousingRecreationprotectionaffairsservices Explanatoryvariable(mean/standarddeviation)(0.29/0.15)(0.09/0.05)(0.12/0.06)(0.13/0.06)(0.23/0.11)(0.05/0.04)(0.06/0.04)(0.02/0.03)(0.02/0.02) Does reduced trade tax revenue affect government spending patterns? 575 level). However, theestimates of the coefcients on the lagged dependent variables are much higher because of their positivecorrelation with the error. As a matter of fact, the true parameter should lie in the range between theseestimates (i.e., upper bound) and those found when not including countries xed effects (not reported here tosave on space). The estimates in Table 4 do fall within these intervals. Similar results hold for all but onecoefcient for the lagged dependent variable when using the GMM system estimator. The one exception is forEconomic Affairs in Table 5, which is larger than one for the lagged dependent variable, and above the upperbound, which suggests possible problems with this specication. To investigate whether the results in Table 4are robust to inclusion of additional regressors, we considered the role of total government revenue, foreignnancial aid, democratic institutions, and conicts. In the rst case, we added total government revenue (as a shareof GDP) since only better-funded governments may engage in spending in nonnecessities (e.g., education,especially for poor countries). As for the nancial controls, we added foreign aid and IBRD and IDA credits, allexpressed as a share of GDP. These additional nancial resources might affect the patterns of governmentspending by relaxing a government budget constraint but perhaps with conditionality that might affect spendingpatterns. We controlled for the level of democracy by employing the Polity index, which ranks countries on a

08 March 2013 Page 39 of 145 ProQuest

scale ranging from 10 to +10 scale, where 10 represents the highest score for a democratic institutions. Itsinclusion reects the possibility that a nation with more responsive government to its citizens needs might have adifferent expenditure share pattern than less democratic nations.19 Finally, we considered whether a country isinvolved in inter or intrastate conicts. All things equal, a government facing such conicts might alter its spendingpatterns to deal with the associated military and security exigencies. The inclusion of each of these regressorsreduces the number of observations (e.g., including the aid variables reduces the sample from 418 to 325observations) so that we do not include them all together in the specications.20 The results show that thecoefcient for total revenue is rarely signicant but, more importantly, its inclusion does not affect our qualitativeresults in any systematic way. The foreign nancial aid variables generally provide little explanatory power(though the coefcient for IBRD/IDA is positive and signicant for defense). Only the coefcient for tariff revenue fordefense is now signicantly different from zero at 1%. More democratic countries do spend more for publicservices but less on education, economic affairs, and housing but there is no change in the role of trade taxes.As expected, intrastate conicts lead to a larger share of government expenditures for defense purposes to thedetriment of spending on economic affairs and housing.21 With respect to the trade revenue shares, they arepositive and signi- 19A governments ideological disposition towards the public spendings appropriate role in theeconomy could also play an important role in expenditure decision. Unfortunately, we are not aware of agenerally acceptable and appropriate metric to measure ideological positions.20See Table 3 for description andsources of all extra regressors. 21Interstate conicts surprisingly do not provide signicant explanatory power fordefense spending. 576 M.O. Moore, M. Zanardi cant (at a 5% level) for education and negative and signicant (ata 10% level) for defense.22 5 Conclusions Many developing country governments traditionally have reliedextensively on import and export taxes as means to nance government operations. These governments face thepossibility that trade liberalization will result in lower overall revenues, especially if domestic capacity of raisingfunds through other means cannot be found. This in turn can lead to a reduction of spending in areas that thesegovernments and societies might nd important. These concerns are not merely academic. Trade negotiators,such as the ones quoted in the Introduction from Jamaica and St. Kitts and Nevis, have explicitly referred to thespending constraints arising out of trade liberalization as a justication for resisting calls for reduced barriers.Assessing these concerns is therefore important in understanding both the effects of trade liberalization moregenerally but also whether poorer countries might have legitimate fears about the scal pressures they might beunder if they reduce this traditional means of nancing central government spending. Researchers (e.g.,Baunsgaard and Keen 2010; Khattry and Rao 2002) have found evidence for one important part of this fear. Inparticular, their work suggests that developing countries, especially those with the lowest average income, mayhave a particularly hard time in nding alternative sources for falling trade tax revenues. Gemmell et al. (2008) ndthat increased inward foreign direct investment, another measure of greater exposure to the global economy,increases the share of social spending among high-income OECD countries. In this paper, we take this line ofreasoning one step forward and ask how the share of government funds derived from import and export taxesaffects the spending shares of central government funding. If lower trade tax revenue shares shift resourcesaway from, for example, spending on health and education, then trade liberalization unmatched by increases inother funds could be even more problematic. If instead, shares of government spending categories areunaffected by lower trade tax revenues, then the problem is less acute. In a sense, we start with the premisethat there might be an income effect from lower overall revenues associated with trade liberalization but we askwhether there is a substitution effect that moves spending away from sensitive areas as a consequence of thescal pressures. We nd little evidence indeed that such a substitution effect is taking place, at least in the sampleof countries for which data is available from the IMFs Government Financial Statistics database. Simpledescriptive statistics show that there has been a broad reduction in the reliance of trade tax revenues across thedeveloping world since 1991 through 2005. But there is little systematic evidence that this decrease has causeda shift in the composition of spending across central government spending 22It is important to notice that the

08 March 2013 Page 40 of 145 ProQuest

differences in the signicance of the trade revenue shares are due to the sample selection due to the availabilityof the extra regressors and not to the role of these extra regressors. Does reduced trade tax revenue affectgovernment spending patterns? 577 categories, especially at the expense of items such as health, education,and social protection. This is true both for subsamples of countries based on average income as well as acrossLatin America, Asia, and Africa. The results of formal econometric analysis are consistent with these generalobservations. We use a dynamic panel approach with a bias correction method proposed by Bruno (2005) andsystem-GMM techniques to analyze nine spending categories, using trade revenue shares along with othercovariates likely to be helpful in explaining expenditure patterns. We nd very little evidence that these traderevenue shares affect spending categories as a percentage of the whole. This is true whether we consider all ofthe developing countries simultaneously or whether we divide the sample into average income categories orregional groupings. There is, however, some scattered evidence for such a spending share shift as a result oftrade tax reductions for particular specications but these results do not carry across econometric methodologiesor subsamples. The evidence presented here, in short, suggests that developing countries in the past have notsystematically altered the pattern of spending as a consequence of falling revenues from import and exporttaxes. This of course does not undercut the fears of government ofcials, consistent with ndings of researchers,that poorer developing countries might face a reduction in overall scal resources if trade liberalization occurs.But our results do suggest that governments in this sample of countries have not moved resources out of onesector and into others as a consequence of these pressures. Of course, it is possible that governments mighthave preferred to increase the share of spending in these categories to deal with the social dislocation resultingfrom resource reallocation associated with trade liberalization. If this is true, then our results suggest that thesegovernments have not been able to pursue a different composition of spending when faced with the scalpressures and reduced trade tax revenues. Only further research can determine which interpretation is correctsince it depends on determining what the optimal composition of spending might be for these governments.Acknowledgements The authors would like to thank Ron Davies, participants at the ETSG 2010 conference andtwo anonymous referees for comments on a previous version of the paper. All remaining errors are our own.Maurizio Zanardi gratefully acknowledges nancial support from the PEGGED Collaborative Project funded bythe European Commissions 7th Framework Programme for Research (Grant Agreement No. 217559). Lowincome Lower middle income Upper middle income Africa Africa Africa Burundia Mauritiusa Dem. Rep. ofCongoa Moroccoa Egypta Tunisiaa LesothoMadagascar Appendix: List of countries 578 M.O. Moore, M.Zanardi Continued Low income Lower middle income Upper middle income Asia Asia Asia BangladeshMalaysiaa South Koreaa Bhutana ThailandaIndiaaIndonesiaaNepalaPakistana Latin America Latin AmericaLatin America Argentinaa Brazila Boliviaa Mexicoa Chilea Uruguaya Costa Ricaa Venezuelaa DominicanRepublicaEl SalvadorNicaraguaaPanamaa Others Others Others Maldivesa Albania Bahraina MyanmaraAzerbaijana Belarusa Bulgariaa EstoniaaCameroon HungaryaCroatiaa LatviaaCzech RepublicaLithuaniaaGeorgiaa MacaoaIrana Maltaa Kazakhstana Netherlands Antilles Lebanon Russian FederationMoldovaa Seychelles Mongoliaa SloveniaaPolanda Trinidad and Tobagoa RomaniaaSlovakRepublicaSyriaaTajikistanTurkeyUkraineaVanuatuZimbabwe aIndicates that the country is included in theeconometric analysis Does reduced trade tax revenue affect government spending patterns? 579 ReferencesAttanasio, O., Goldberg, P., &Pavcnik, N. (2004). Trade reforms and wage inequality in Colombia. Journal ofDevelopment Economics, 74(2), 331366. Arellano, M., &Bover, O. (1995). Another look at the instrumentalvariable estimation of error-components models. Journal of Econometrics, 68(1), 2951. Baunsgaard, T., &Keen,M. (2010). Tax revenue and (or?) trade liberalization. Journal of Public Economics, 94, 563577. Blundell, R.,&Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models.Journal ofEconometrics, 87(1), 115143. Bruno, G. S. F. (2005). Estimation and inference in dynamic unbalanced paneldata models with a small number of individuals. Stata Journal, 5(4), 473500. Choi, I. (2001). Unit root tests forpanel data. Journal of International Money and Finance, 20(2), 292372. Correlates of War project. Available at

08 March 2013 Page 41 of 145 ProQuest

http://www.correlatesofwar.org/ Web End =http://www.correlatesofwar.org/ . Cragg, M., &Epelbaum, M. (1996).Why has wage dispersion grown in Mexico? Is it the incidence of reforms or the growing demand for skills?Journal of Development Economics, 51(1), 99116. Dreher, A., Sturm, J.-E., &Ursprung, H. (2008). The impactof globalization on the composition of government expenditures: evidence from panel data. Public Choice,134(34), 263292. Emran, S., &Stiglitz, J. (2005). On selective indirect tax reform in developing countries.Journal of PublicEconomics, 89(4), 599623. Gemmell, N., Kneller, R., &Sanz, I. (2008). Foreign investment,international trade and the size and structure of public expenditures. European Journal of Political Economy,24(1), 15171. Jamaica (2004). Jamaica trade policy review (WT/TPR/G/139). Khattry, B., &Rao, M. (2002).Fiscal faux pas? An analysis of the revenue implications of trade liberalization. World Development, 30(8),14311444. Levin, A., Lin, C.-F., &Chu, C.-S. (2002). Unit root tests in panel data: asymptotic and nite-sampleproperties. Journal of Econometrics, 108(1), 124. Monty, G. M., &Jaggers, K. (2002). Polity IV data set.[Computer le; version p4v2004] College Park, MD: Center for International Development and ConictManagement, University of Maryland. Rodrik, D. (1998). Why do more open economies have biggergovernments. Journal of Political Economy, 106(5), 9971032. Roodman, D. (2009). A note on the theme of toomany instruments. Oxford Bulletin of Economics andStatistics, 71(1), 135158. Schulze, G., &Ursprung, H.(1999). Globalisation of the economy and the nation state. World Economy,22(3), 295352. Saint, K., &Nevis(2003). Statement by Mr Horatio Versailles of Permanent Secretary, Ministry of International Trade andCARICOM Affairs (WT/MIN(03)/ST/123). Schwert, W. G. (1989). Tests for unit roots: a Monte Carloinvestigation. Journal of Business &EconomicStatistics, 7(2), 147159. Uppsala Conict Data Program(UCDP/PRIO). Available at http://www.prio.no/CSCW/Datasets/Armed-Conflict/UCDP-PRIO/ Web End=http://www.prio.no/CSCW/Datasets/ Armed-Conict/UCDP-PRIO/. Subject: Government spending; Tariffs; Revenue; Trade liberalization; Taxes; Analysis; Studies Classification: 1120: Economic policy&planning, 1300: International trade&foreign investment, 4200: Taxation,2600: Management science/operations research, 9130: Experimental/theoretical Publication title: International Tax and Public Finance Volume: 18 Issue: 5 Pages: 555-579 Publication year: 2011 Publication date: Oct 2011 Year: 2011 Publisher: Springer Science&Business Media Place of publication: Boston Country of publication: Netherlands Journal subject: Business And Economics--Public Finance, Taxation ISSN: 09275940 Source type: Scholarly Journals Language of publication: English Document type: Feature

08 March 2013 Page 42 of 145 ProQuest

DOI: http://dx.doi.org/10.1007/s10797-011-9191-y ProQuest document ID: 888716133 Document URL: http://search.proquest.com/docview/888716133?accountid=86413 Copyright: Springer Science+Business Media, LLC 2011 Last updated: 2012-01-13 Database: Accounting&Tax

_______________________________________________________________ Document 4 of 10 THE DECLINING USE OF INTERNAL SERVICE FUNDS: HOW LOCAL GOVERNMENTS ARECHANGING THE ALLOCATION OF INDIRECT COSTS Author: Modlin, Steve Publication info: Journal of Public Budgeting, Accounting & Financial Management 23. 2 (Summer 2011): 151-165.ProQuest document link Abstract: Studies involving the allocation of funds among local governments usually are broad in nature with focibased on a variety of factors ranging from service demand to performance outcomes. The conundrum ofindirect costs allocation associated with service demand continues to confront local governments. The internalservice fund (ISF) has been the primary device used in this endeavor, but over the past two decades, itsutilization has decreased. County finance officers in the southeastern United States were surveyed to determinewhy the ISF is not as prevalently used as in previous years and what has happened to indirect costs as a resultof these changes. Findings suggest many reasons for ISF's usage decline including limited usefulness andreallocation of indirect costs to departments. In addition, county governments with a cost allocation plan andlarger budget sizes continue to use the ISF as an accounting device. [PUBLICATION ABSTRACT] Full Text: Headnote ABSTRACT. Studies involving the allocation of funds among local governments usually arebroad in nature with foci based on a variety of factors ranging from service demand to performance outcomes.The conundrum of indirect costs allocation associated with service demand continues to confront localgovernments. The internal service fund (ISF) has been the primary device used in this endeavor, but over thepast two decades, its utilization has decreased. County finance officers in the southeastern United States weresurveyed to determine why the ISF is not as prevalently used as in previous years and what has happened toindirect costs as a result of these changes. Findings suggest many reasons for ISF's usage decline includinglimited usefulness and reallocation of indirect costs to departments. In addition, county governments with a costallocation plan and larger budget sizes continue to use the ISF as an accounting device. INTRODUCTION Thecontinual struggle to provide goods and services with minimum revenue increases has created tremendouscapacity issues for local governments. To account for many of the indirect costs which go into providingservices, many local governments employ internal service funds (ISFs). These proprietary funds allow for oneunit of government to charge for services requested by another unit. Local governments have much discretionconcerning the use and application of internal service funds creating a litany of cost allocations and applicationsamong units. This article examines the use of ISFs among county governments in North Carolina, SouthCarolina, and Tennessee. The declining use of ISFs suggests that county governments are trying to distributeindirect costs in a more discretionary manner. The emergence of service-providing functional departments, lackof resources necessary for implementation, and alternative accounting standards have all contributed to less

08 March 2013 Page 43 of 145 ProQuest

use of the ISF. DEFINING INTERNAL SERVICE FUNDS Internal Service Funds are one of two forms ofproprietary funds utilized by local governments. Internal service funds engage in business-type activities thatare usually not rendered for public use (Holder, 2004). However, there are cases in which the public canpurchase particular services from ISF providing departments directly. Traditionally, ISFs provide services toother departments within the government organization on a cost reimbursement basis (Granof &Wardlow,2003). This means that the department providing the services charges the receiving department. Often, thesedepartments correspond to related organizational units, such as data processing, vehicle repair centers,janitorial functions, and capital asset leasing. However, they can be established to account for activities in whichthere is no comparative organizational unit such as self-insurance (Granof &Wardlow, 2003). ISFs are designedto provide efficiency in the acquisition, distribution, service provision, and the accounting of goods and services.Ives, Razek, and Hosch (2004) suggest that the main reasons for establishing ISFs are (1) to reduce the costsof obtaining goods and services; and, (2) to improve the distribution of goods and services within thegovernmental unit. Often, the general fund is billed for the service incurred by the ISF, which creates revenuefor the ISF. A separate fund is usually used for each unit identified as an object of expenditure since theaccumulation of service provision costs must be associated with revenue earned from these same provisions.ISFs usually engage in the same form of accounting practices as enterprise funds. The full accrual basis ofaccounting is traditionally used for identifying revenues and expenses that occur between the ISF and theagency receiving services. ISF costs are full costs suggesting the billing rate reflects all operating costs,including depreciation and interests if applicable, and other indirect costs. This creates the opportunity for thegovernment unit to monitor the full cost of providing a good or service with little anticipation for change (Granof&Wardlow, 2003). Financial statements for ISFs are (1) the statement of net assets; (2) statement of revenues,expenses, and change in net assets; and (3) the statement of cash flows. UTILIZATION AND INTERNALSERVICE FUND DEVELOPMENT Proper implementation of the internal service fund provides manyadvantages for financial officers. First, it provides information regarding common specific services required bydepartments. From health insurance to motor pools and even to electricity usage, officials can determine theprecise usage of assets via departmental service provisions, and in some cases depreciate some of the fixedassets on a regular basis. Second, the use of the internal service fund allows for service cost comparisons(Gianakis, 1995). In most cases, the ISF is comprehensive in nature with full cost application to the servicesprovided. If these costs appear to be creating excessive expenditures, governments can turn to Request forProposals (RFP) to examine the provision cost as compared to private sector vendors. For state governments,Thurmaier (1990) suggests that many of these costs moved out of the more traditional "overhead cost" areasand into Object Classification Schemes that can be adjusted according to budget or financial needs. Third, ISFsare also very useful in determining the overhead costs to grants (GFOA, 1988). With the broad structure ofmany grant programs, especially those associated with community development, variable costs increases forimplementation are captured through the ISF, therefore acting as a catalyst for managers to sustain the need forincreased revenues through grant programs. Hendrick (1998) found that some departments can expectequivalent appropriations annually despite the number of grants received. Choosing to implement a costallocation plan with the utilization of the internal service fund has interdepartmental limitations. First, mostdepartments prefer to be self-sustaining thus recalcitrant to accept many provisional services from otherdepartments. Many agencies and departments relish the opportunity to provide services "in-house" withadditional responsibilities and revenues to match. Second, units which have maintained a support role fortraditional services can experience capacity issues when particular services are required in an expeditiousmanner (Davis, 1991). Services stemming from natural disasters and large-scale emergencies requireexpeditious allocations that local government support departments cannot provide. Third, the full allocationcosts surrounding a service can be difficult to determine. User charges are relatively easily accounted for incustomer-agencies, but not very transparent upon examination through the ISF department (Gianakis, 1995).

08 March 2013 Page 44 of 145 ProQuest

Service recipients have many incentives to include all costs, both direct and variable upon estimatingcomprehensive expenditures for future service provision. The ISF department, which usually is the onlydepartment or agency providing particular services, may not fully document all related costs. In many of thesecases, managers and decision makers are forced to examine outside contractors for services using equivalentbenchmarks for both parties to ensure the most efficient provisions will be available (Ammons, 2002).Government officials, especially those in local government settings, are continually exhausting methods toimprove financial practices due to criticisms of continual expense increases (Rubin, 1993). As a result of thesepractices, the more traditional ISF fund departments are facing challenges from both in-house departments thatare interested in obtaining technology to provide the service and privatesector counterparts. FINDINGSRELATED TO ISF IMPLEMENTATION Most of the research surrounding the use of ISFs by local governmentsis rather limited. The studies have mainly included case studies demonstrating anomalies within ISF uses. Themost comprehensive study of ISF use by a plethora of local governments was conducted by Coe and O'Sullivan(1993), who examined all U.S. cities with a population of more than 500,000 and 25 percent of cities between25,000-499,999. Professionally managed governments (council-manager form) were found to be more likely toimplement the ISF compared to governments with the elected chief executive. This was expected since council-manager governments have been found to outperform other forms of local governments on many levels offinancial and accounting practices along with providing higher transparency among financial devices (Giroux&McLelland, 2003; Ingram &DeJong, 1987). The findings of Coe and O'Sullivan also suggested that 72 percentof cities did implement an ISF and further attempted to ascertain overhead costs of providing services, oftenthrough some cost accounting mechanism. The primary use of the ISF of responding cities was through the useof motor pools. Costs associated with the operation of a motor fleet are extremely diverse: depreciation, fuelcosts, insurance, maintenance, and optimal resale value. Finance officers also responded that insurance wasthe other common use of the ISF. Health insurance is becoming one of the more scrutinized areas concerningexpense reduction; however, this is expected to be an ongoing activity supported by ISFs. Cities that did notutilize ISFs cited reasons ranging from inadequate information to negative costbenefit ratios. Some of therespondents even suggested the ISF was not needed. Much diversity exists with the use of ISFs amongmunicipalities, but questions remain as to how cost allocations have changed among local governments. Forinstance, there is no research concerning the use of cost allocation methods among county governments. Asadministrative arms of the state with increasing responsibilities surrounding service delivery, these contributionsare vital. Also, there is expected to be even more of a disparity among different forms of government thanindicated by the Coe and O'Sullivan findings. Professionally administered county governments are expected tobe more financially diverse than other forms. From a population standpoint, rural government has been definedin a number of ways by researchers. Much research focuses on governments with at least 100,000 citizens in alocality, but this does not include many rural areas, especially in the case of county governments in which thecounty is the predominant level of government responsible for services. Finally, there is no researchsurrounding on why the internal service fund has devolved. Questions remain as to where these costs are nowallocated and if departments have increased in size and manner as a result of in-house provision of services.METHODOLOGY To examine the changes and uses of the ISF among county governments, a survey was sentto all North Carolina county (finance officers), South Carolina county (auditors/finance officers), and Tennessee(trustees). All of the local governments in North Carolina and a very high proportion of South Carolina countygovernments are professionally administered with a county manager/administrator while Tennessee hadmultiple forms of local county governments with the majority having a chief executive. Approximately 241surveys were sent with a response rate of 40% with the vast majority coming from the professionallyadministered governments. Of the 97 surveys returned, 61% of North Carolina county governments wererepresented in the sample as well as 35% of South Carolina county governments. Only 20 surveys werereturned from Tennessee county governments. The response rate was expected to be slightly higher since

08 March 2013 Page 45 of 145 ProQuest

public officials were the source of information, but the segregation of financial responsibilities among countyofficials created convolution. For instance, in North Carolina, the finance officer is responsible for virtually allaspects of local government finance from budgeting, financial management, and investment practices. Incounties with the elected trustee, the responsibility of daily budget transactions including accompanyingaccounting practices are the responsibility of the finance officer with cash management left to the trustee. Inthese instances, requests were made to have the responsible party complete the survey. Other information wasobtained through external sources. Information consisting of county budget size was obtained through stateentities such as the North Carolina Department of State Treasurer (2008), South Carolina State Budget andControl Board (2007), and the Tennessee Comptroller of the Treasury (2008). County forms of governmentwere obtained through the National Association of Counties. FINDINGS Initial findings did support a substantialdecline in the use of the internal service fund as compared to the Coe and O'Sullivan study. Only 26% of thesampled county governments responded that they employed internal service funds as a method of determiningcosts of an activity. Of the counties which responded that they did use the internal service fund, only 23% usedit for just one area while the remainder of counties utilized it more frequently. The three most common areas forISF use were health insurance and workers compensation followed by the county garage or motor pool. Forhealth insurance, all county employees usually contribute to the fund in some manner although use of theinsurance by employees is not nearly as equitable. Workers compensation usually just covers employees thatwork in more hazardous departments such as buildings and grounds and even county garages. The countygarage covers a multitude of indirect costs as mentioned earlier, but a key feature of the county garage is that itprovides department heads with a mechanism to enhance departmental budgets through automobileacquisitions. Service repairmen in garages can make recommendations for the replacement of automobilesbased on criteria such as high mileage, poor performing automobiles, or even an automobile that has beenpoorly maintained due to continual neglect by department personnel. The installation of an ISF is usually analternative used by individuals that are aware of alternative accounting mechanisms used to detect indirectcosts. For that reason, there were two general expectations. First, larger county governments were expected toutilize the ISF more than do smaller county governments. The complexity surrounding the variety of methodsused to sustain services by larger county governments was expected to incur higher indirect costs and thereforean increased use of ISFs. Second, professionally administered governments, those with thecommissionmanager or council-manager forms of county government were expected to utilize the ISF morethan the other forms. Of the county governments, the professionally administered governments used the ISFoverwhelmingly when compared to other government types. Table !supports both hypotheses. ISFs arepredominantly used to determine the indirect costs of a particular service. However, when counties which usedthe ISF were asked if they depreciated assets, only seven responded. Depreciation took place only within twoareas: county garages and information systems. For county garages that do not depreciate, vehicles are rotatedbased on different criteria such as an approximate age of the vehicle or other equipment and if the serviceprovider has determined replacement to be more beneficial. Information systems replacement is usually doneon an in-house basis as needed. In an effort to determine why so many county governments are not using theISF, respondents were asked to submit reasons why the ISF was not utilized. Most of the respondents (27%)maintained that they were unable to implement ISFs due to limited resources (Table 2). Nearly 25% claimed thecosts associated with ISF implementation exceeded the benefits while another 24% claimed a vast array ofreasons including lack of necessity, year-end accounting of indirect costs, and GASB 34 substitution. Inaddition, another 20% were unfamiliar with the ISF. Findings such as this generally lead to the generalpresumption of non-clarity in local government financial auditing (Wallace, 1981). The other portion of theresearch question focused on identifying key areas that have supplanted the use of ISFs. There were only 12responses to the question with the majority indicating that many of these costs are attributed to a specificdepartment which has evolved from just simply functioning as a support area. Other answers suggested that the

08 March 2013 Page 46 of 145 ProQuest

activity covered by the fund had been contracted out, such as the case with health insurance, or the countyused general fund monies to cover the costs. More than 67% of the respondents have been in their currentposition for more than five years thus providing the conclusion that these costs have been allocated in thismanner for an extended period of time. The lack of use of the ISF provided the supposition that localgovernments are actively using cost allocation plans to detect indirect costs encountered during the fiscal year.Surprisingly enough, nearly half of the respondents (47%) stated that they did not use a cost allocation plan todetermine indirect costs. The majority of county governments that utilized cost allocation plans (36%) chargedother administrative departments on a regular basis. One respondent did account for indirect costs during thecourse of the fiscal year, but did not change the general ledger to reflect those costs while another respondentused a cost allocation plan only for federal and state grants. To determine what factors contribute to the use ofISFs, Table 3 presents a model with construction of the ISF, a dummy variable in this case, as the dependentvariable. With the exception of state use of the ISF, which were dummy variables representing each state in thestudy, the other explanatory variables were coded on a 1-5 scale. Cost allocation plan responses were allocatedbased on responses ranging from using the ISF to account for indirect costs to using miscellaneous methods toaccount for indirect costs. Finance officer experience, finance officer education level, and county budget sizewere all coded relative to their responses as well as available information. There were also five different formsof county government: council/commission-manager, council, county executive, charter, and consolidated.While there were just minimal responses from charter and consolidated county governments, these anomaliesprovided an additional facet to the study. Table 3 presents a logit model of likely ISF use by local governmentsbased on five different independent variables. In the model, the use of a cost allocation plan is significant andpositive indicating that local governments that do not use a cost allocation plan are less likely to employ the useof the ISF compared to those that use a cost allocation plan. Also significant were budget size of the localgovernment and local government type. Local governments with larger budgets were 61% more likely to use theISF compared to smaller local governments. Council-manager and commissionmanager governments that wereprofessionally administered utilized the ISF much more frequently compared to council, charter, andconsolidated governments. This finding corresponds with state use of the ISF. All 100 North Carolina countiesas well as 41 of 46 South Carolina counties have professionally administered governments. Only 5 respondentsoutside of North Carolina stated they used the ISF. Education and experience were expected to be significantfactors in determining the use of ISFs among county governments, but the null hypothesis was supported inboth cases. Those with higher education levels and experience levels used the ISF less frequently.DISCUSSION The research has illustrated a vast array of practices that county governments engage in whenencountering indirect costs. The declining use of the internal service fund has been the result of many factorsincluding more progressive accounting measures, departmental growth, and even managers covering costs inthe most expethent manner possible. However, there definitely appears to be no decisive way to handle indirectcosts - although in many cases, they may be increasing among local governments. Caution has to be exercisedwhen trying to generalize these findings across other local governments. First, a high percentage of theresponses came from professionally administered governments under the council-manager form. There is notreally sufficient information to determine the indirect costs processes of other forms of county governments.Second, much information which may be needed to determine indirect costs was not accessible such ascomprehensive annual reports, journal and/or ledger entries, or even specific third party contracts. Third, therewere believed to be more respondents that did not know the definition of an internal service fund than what wasindicated suggesting that there are many finance officers conducting daily financial transactions without fullknowledge of accounting alternatives. There are also many implications associated with the findings. First, withthe absence of indirect cost allocation and the internal service fund, the environment for continued departmentalbudget increases is realized. The ability for managers and elected officials alike to question departmentalfunding is severely limited. Second, elected officials actually see themselves very involved in the budget

08 March 2013 Page 47 of 145 ProQuest

process in general, although the amount of time they actually spend on intimate accounting details is verylimited (Modlin, 2008). These findings support those conclusions; elected officials will now be even moredisadvantaged when attempting to rationalize departmental expenses. Finally, the findings demonstrate that it isvery important to prepare students with the best training possible prior to public sector financial endeavors.Many texts do not cover many of the accounting facets necessary to develop strong cost finding skills (Finkler,2005; Lee, Johnson &Joyce, 2004; Mikesell, 2007; Smith &Lynch, 2004). CONCLUSION This study hasattempted to answer two basic questions. First, to determine whether or not local governments, especiallycounty governments are using the ISF less now as compared to previous years; and second, if there is lessusage, to determine where these costs are allocated. There is definitely less usage of the ISF compared to whatwas found in previous studies. Only 27% of respondents were using the ISF as a method of allocating costs. Inaddition to this finding, many finance officers responded there was no need to use the ISF because costs weredistributed by other means through interdepartmental mechanisms or just accounted for at the end of the fiscalyear. The use of the cost allocation plan and the size of the county budget were significant factors indetermining use of the ISF. Indirect costing is an important facet within the daily financial activities of localgovernments. For many reasons, the practice has become quite limited with few participants. Localgovernments should attempt to utilize as many approaches as necessary in order to increase financialtransparency. The ISF can still be a very useful tool and a tremendous accounting asset for finance officers aswell as a method to determine areas that have problems associated with cost effectiveness. References REFERENCES Ammons, D. (2002). Tools for Decision Making: A Practical Guide for Local Government.Washington, DC: Congressional Quarterly Press. Coe, C., &O'Sullivan E. (1993). "Accounting for Hidden Costs:A National Study of Internal Service Funds and Other Indirect Costing Methods in Municipal Governments."Public Administration Review, 53 (2): 59-64. Davis, T. (1991). "Internal Service Organizations: Strategies forIncreasing their Effectiveness and Controlling their Costs." Organizational Dynamics, 20 (2): 5-22. Finkler, S.(2005). Financial Management for Public, Health, and Not-for-Profit Organizations. (2nd ed.). Upper SaddleRiver, NJ: Pearson Prentice-Hall: 443-444. Gianakis, G. (1995). "Productivity Issues in Internal Service FundAgencies." Public Productivity &Management Review, 18 (4): 349-363. Giroux, G., &McLelland, A. (2003)."Governance Structures and Accounting at large Municipalities." Journal of Accounting and Public Policy, 22(3): 203-230. Government Finance Officers Association (GFOA). (1988). Government Accounting, Auditing, andFinancial Reporting. Chicago, IL: Author. Granof, M., &Wardlow, P. (2003). Core Concepts of Government andNot-for-Profit Accounting. Hoboken, NJ: John Wiley &Sons: 218-224. Hendrick, R. (1998). "The Impact ofFederal Grants and Other Funds on General Fund Expenditure Decisions: A Detailed Analysis of One City."Journal of Public Administration Research and Theory, 8 (3): 353-389. Holder, W. (2004). "FinancialAccounting, Reporting, and Auditing." In J.Richard Aronson and Eli Scwartz (Eds.), Management Policies inLocal Government Finance (5th ed.) (pp. 207-222). Washington, DC: International City/County ManagementAssociation. Ingram, R., &DeJong, W. (1987). "The Effect on Local Government Disclosure Practices." Journalof Accounting and Public Policy, 6 (4): 245-269. Ives, M., Razek, J., &Hosch, G. (2004). Introduction toGovernment and Not-for-Profit Accounting. (5th ed.). Upper Saddle River, NJ: Pearson Prentice-Hall. Lee Jr.,R., Johnson, R., &Joyce, P. (2004). Public Budgeting Systems. (7th ed.). Boston, MA: Jones &Bartlett.Mikesell, J. (2007). Fiscal Administration. (7th ed.). Belmont, CA: Thomson Wadsworth. Modlin, S. (2008)."Defining Involvement of County Commissioners in the Budget Formulation Process." Politics &Policy, 36 (6):56-72. North Carolina Department of State Treasurer (2008). County and Municipal Information. Raleigh, NC:Author. [Online]. Available at http://www.nctreasurer.com/lgc/units/unitlistjs.htm. Rubin, I. (1993). "Who InventedBudgeting in the United States?" Public Administration Review, 53 (5): 438-444. Smith, R. &Lynch, T. (2004).Public Budgeting in America (5th ed.). Upper Saddle River, NJ: Pearson Prentice-Hall. South Carolina StateBudget &Control Board (2007). Office of Research &Statistics. Local Government Finance Report. Columbia,SC: Author. [Online]. Available at http://www.ors.state.sc.us/economics/ economics.asp. Tennessee Comptroller

08 March 2013 Page 48 of 145 ProQuest

of the Treasury (2008). Division of County Audit. Annual Financial Reports. Nashville, TN: Author. [Online].Available at http://www.tn.gov/comptroller/ca/. Thurmaier, K. (1990). "Information Resources ManagementExpenditures in the States: A Case for Integrating Accounting and Budgeting Systems." Public Budgeting&Finance, 10 (4): 28-46. Wallace, W. (1981). "Internal Control Reporting Practices in the Municipal Sector." TheAccounting Review, 56 (3): 667-688. AuthorAffiliation Steve Modlin* AuthorAffiliation * Steve Modlin, Ph.D., /san Assistant Professor, Department of Political Science, East Carolina University. His research interests includelocal government budgeting and finance. Subject: Studies; Local government; Overhead costs; Cost allocation; Counties Location: United States--US Classification: 9190: United States, 9130: Experiment/theoretical treatment, 9550: Public sector Publication title: Journal of Public Budgeting, Accounting&Financial Management Volume: 23 Issue: 2 Pages: 151-165 Number of pages: 15 Publication year: 2011 Publication date: Summer 2011 Year: 2011 Publisher: PrAcademics Press, Florida Atlantic University Place of publication: Boca Raton Country of publication: United States Journal subject: Business And Economics--Banking And Finance ISSN: 10963367 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Tables;References ProQuest document ID: 877017307 Document URL: http://search.proquest.com/docview/877017307?accountid=86413 Copyright: Copyright PrAcademics Press, Florida Atlantic University Summer 2011 Last updated: 2011-07-23 Database: Accounting&Tax

_______________________________________________________________ Document 5 of 10 International differences in IFRS policy choice: a research note

08 March 2013 Page 49 of 145 ProQuest

Author: Kvaal, Erlend; Nobes, Christopher Publication info: Accounting and Business Research 40. 2 (2010): 173.ProQuest document link Abstract: Building on literature that suggests motives and opportunities for national versions of IFRS practice,we examine whether there are systematic differences in IFRS accounting policies between countries. Usinginformation from the annual reports of companies in the blue chip indices of the largest five stock markets thatuse IFRS, we reject a null hypothesis that IFRS practice is the same across countries. For 16 accounting policyissues, we find instead significant evidence that pre-IFRS national practice continues where this is allowedwithin IFRS. By this, we document the existence of national patterns of accounting within IFRS. We also pointout some policy implications that arise from our findings. [PUBLICATION ABSTRACT] Full Text: Headnote Abstract - Building on literature that suggests motives and opportunities for nationalversions of IFRS practice, we examine whether there are systematic differences in IFRS accounting policiesbetween countries. Using information from the annual reports of companies in the blue chip indices of thelargest five stock markets that use IFRS, we reject a null hypothesis that IFRS practice is the same acrosscountries. For 16 accounting policy issues, we find instead significant evidence that pre-IFRS national practicecontinues where this is allowed within IFRS. By this, we document the existence of national patterns ofaccounting within IFRS. We also point out some policy implications that arise from our findings. Keywords:international standards; international differences; policy choice 1. Introduction It has been suggested that thereare motives and opportunities for the survival of international differences under International Financial ReportingStandards (IFRS) (Ball, 2006; Nobes, 2006; Zeff, 2007). This paper seeks to answer two questions relating tothis. First, are there systematic differences between countries with respect to the accounting policies thatcompanies use within IFRS, so that one can identify national IFRS patterns? Second, if there are, can weexplain how policies were chosen on transition to IFRS? We investigate these questions using the 2005- 06IFRS annual reports of companies based in five countries: Australia, France, Germany, Spain and the UK. In allthese countries, IFRS is compulsory,1 at least for the consolidated statements of listed companies. Strictlyspeaking, it is EU-endorsed IFRS2 that is compulsory for the EU companies, and IFRS-based Australianstandards that are compulsory in Australia. This point presents one of the drivers of different practices. Otheropportunities for variety arise from options clearly available within IFRS, and we concentrate on these. Giventhe motives and opportunities for national versions of IFRS, we expect to find such differences in practice. Thispaper contributes to the literature in a number of ways. First, we document formally that there are differentnational versions of IFRS practice. Related to this, we show that companies not only have an opportunity topursue pre-IFRS practices originating in their national GAAP,3 but also extensively use this opportunity. Thesefindings are important for several reasons. For financial statement users, they imply that full internationalcomparability has not yet arrived. Therefore, it has been suggested, investors might be misled by an apparentuniformity (Ball, 2006: 15). As long as accounting standards contain options and require use of judgment, somevariation in accounting practice is inevitable. However, the existence of systematic differences in practicerelated to national borderlines is clearly in conflict with the objective of international harmonisation and maymislead financial statement users who do not pay attention to them. Some differences within IFRS practice areobservable and can be adjusted for by alert analysts (e.g. the location of dividends in a cash flow statement);other differences are easily observable but cannot be adjusted for without a large degree of estimation (e.g. theeffects of the inventory flow method on profit, or the absence of a gross profit figure in a by-nature incomestatement); yet others are not observable (e.g. the application of criteria for making impairments or forcapitalising development costs). Some users of financial statements might be misled by even the first type ofdifferences, but many might be misled by the third type. The second and third types create difficulties forinternational comparative analysis. There are also policy implications. First, the IASB aims not just to issue

08 March 2013 Page 50 of 145 ProQuest

standards but to facilitate comparable information (IASCF, 2005). This paper illustrates topics on which morework would be needed to achieve this objective. Second, the Securities and Exchange Commission (SEC) ismonitoring the use of IFRS for foreign registrants on US exchanges (SEC, 2008, II, D). Part of this considerationincludes an assessment of IFRS practice from 2005, but we show that there are several national versions ofIFRS practice. The paper proceeds as follows: Section 2 summarises relevant literature; Section 3 draws on thisto state our main hypothesis and outline our research design; Section 4 explains our selection of countries,companies and accounting topics for this; Sections 5 and 6 present detailed hypotheses and results; andSection 7 draws interpretations and conclusions. 2. Literature One strand of literature that is relevant to whatfollows is research on the motives and opportunities for international differences in accounting before theadoption of IFRS. This is examined in many papers and textbooks. The objective of our paper is not to try toexplain pre-IFRS accounting differences. We ask, instead, whether there is evidence that country-specificvariables affect choices within IFRS. To our knowledge there is no scientific literature that addresses this issue.Nobes (2006) summarises the literature on the reasons for pre-IFRS accounting differences, asking whetherthese reasons might continue to operate in the context of transition to IFRS. A large number of factors has beenproposed as pre-IFRS influences. The most proximate to accounting itself are legal systems, taxation systemsand financing systems. These could still be relevant to IFRS practice. As examples of the three influences inturn: monitoring and enforcement of IFRS still depends on national regulatory institutions; tax motivations canstill affect practice in unconsolidated statements, and some of this might flow through to consolidatedstatements; and companies in equity-finance countries might be the more interested in voluntary disclosures.The national literature on IFRS is also likely to perpetuate national practices (e.g. PwC (2005) on formats).4 Ball(2006: 15) suggests about IFRS that: 'The fundamental reason for being sceptical about uniformity ofimplementation in practice is that the incentives of preparers (managers) and enforcers (auditors, courts,regulators, boards, block shareholders, politicians, analysts, rating agencies, the press) remain primarily local.'Commentators sometimes even argue in favour of attempting to preserve a national flavour of IFRS (Küting,2007: 2557). The international differences in accounting policies that we study mostly result from companies'policy choices, and research on this subject is potentially relevant to our work. Much of it is directed at revealingthe incentives and motivations of such choices, e.g. in the context of earnings management (for comprehensiveliterature reviews, see Healy and Wahlen, 1999; Fields et al., 2001). These perspectives are not so importantfor the policy choices studied in this paper, because only a few of our issues (pension accounting, fair valueoption) affect the inter-period allocation of net income. Closer to our study is the research that explores thecauses and effects of companies' adopting high-quality GAAP. It is often argued that companies accept thecosts of such adoption in order to reduce their cost of capital (Leuz and Verrecchia, 2000; Ashbaugh andPincus, 2001; Cuijpers and Buijink, 2005). Although there is ample evidence that voluntary adoption of IFRShas enhanced accounting quality (Barth et al., 2006; Gassen and Sellhorn, 2006), the benefits of mandatoryadoption are more doubtful (Daske et al., 2007; Christensen et al., 2008). The importance of an adequateinstitutional framework for reporting incentives has also been emphasised (Ball, 2001; Ball et al., 2000;Bushman and Piotroski, 2006). Although this paper does not address the extent to which companies reap therewards of IFRS reporting, the tendency to preserve national practice that we document may be one of thephenomena that limit the benefits of common reporting requirements. There is some professional literature onIFRS practices from 2005 onwards. KPMG and von Keitz (2006) focus on 199 IFRS reports of the largestcompanies of ten countries (seven of them in the EU), using year-ends of 2005 or before. The use of thoseyear-ends excludes the first implementation by many UK companies,5 and also means that countries such asAustralia6 were excluded. The KPMG study reports on the choice of options, in some cases including abreakdown by country. However, that study is not designed to produce a formal comparison of practicesbetween countries. ICAEW (2007) reports on a survey of 200 listed companies of all sizes across 24 EUcountries for 2005-06. A similar report for 2006-2007 has also been published (European Commission, 2008). In

08 March 2013 Page 51 of 145 ProQuest

general, the data in these reports on the choice of options are aggregated rather than shown by country,although there are some exceptions to that. There is also some literature that records pre- IFRS nationalpractices (rather than investigating motives for international differences in them) and the differences betweennational GAAP and IFRS. To explore this, we have consulted national laws and standards, and analyses ofthem, such as TRANSACC (2001). We have also looked at surveys of practice, such as FEE (1991).Differences between national rules and IFRS were analysed by Nobes (2001), whose data form the basis of astudy of factors influencing the scale of these differences by Ding et al. (2007). 3. General hypothesis andresearch design The differences in IFRS practice that we study relate to policy choices. We base ourhypotheses on the literature (of the previous section) that suggests that companies tend to continue with theirprevious national practices where this is possible under IFRS. However, we note that there are fourdistinguishable reasons for this. First, as explained above, the underlying causes of previous differencesbetween national accounting practices (such as enforcement systems) may still have scope to affect IFRSpractice. Essentially, many drivers of policy choice remain national. Second, and relatedly, IFRS consolidatedstatements are drawn up from unconsolidated statements. So, for example, the practices required or chosen inthe unconsolidated financial statements of a German parent or a German subsidiary under German law mightflow through to the consolidated IFRS financial statements where the practices are permitted under IFRS. Athird reason is that directors of a group might try to maintain consistent accounting policies over time, despitethe transition to IFRS, so as to create as much continuity as possible for the users of the financial statements.Fourth, and relatedly, the directors might wish to minimise the number of changes to their accounting systems,thereby reducing the company's costs of transition to IFRS, by retaining pre-IFRS practices where possible. Apotential explanation for a particular predominant pre-IFRS policy in a country might be the importance ofcertain sectors in that country. For example, perhaps a particular sector mostly uses first in, first out (FIFO)inventory valuation whichever country it is in, and this sector is especially strong in one country, making FIFOmore than averagely common there. Our prediction is that FIFO would continue to be common in that countryunder IFRS. However, this might mean that the option was not being chosen in a way that reduced internationalcomparability among similar companies. Nevertheless, as will be shown, many of the international differencesare so strong that sectoral imbalances cannot explain them. For example, no German company in our sampleuses only FIFO in its IFRS statements whereas half of the UK companies do. In order to discover whetherinternationally different versions of IFRS practice exist, we selected large companies from five major stockmarkets and examined their IFRS policies for 16 issues. We propose the following null hypothesis: IFRSpractice is the same across all countries. We test the null hypothesis against an alternative of non-homogeneityby chi-square tests for each topic. We further test the validity of the null hypothesis against a number ofalternative hypotheses that predict national practice relating to each issue. The predictions implied in thealternative hypotheses are based on our presumption that companies, in the absence of strong incentives to dootherwise, will pursue a policy previously adopted if it is still allowed. We by no means exclude acompanyspecific motivation for any choices previously made under the national GAAP (see Section 2), but ourfocus is only on the company's behaviour on transition to IFRS. 4. Selection of countries, companies and policyissues Nobes (2006) suggests a series of hypotheses about international differences under IFRS, mainlyexpressed by using Germany and the UK as exemplars of previously different accounting 'systems'. We studycompanies from these two countries, but add Australia, France and Spain. The rationale for this list of five isthat, of the countries where IFRS are compulsory for listed companies, they had the five largest stock markets.7Australia is different from the other four countries in not being a member of the EU. We do not expect that, byitself, to cause differences in IFRS practice.8 However, one particular feature of Australian IFRS is relevanthere: for two of the 16 accounting issues that we study, there was no option in Australian IFRS in 2005-2006.So, Australian policies on these issues in that period were not choices. Nevertheless, the requirements inAustralian IFRS continued previous national requirements, so this is consistent with our general hypothesis that

08 March 2013 Page 52 of 145 ProQuest

IFRS practice will preserve national practice. Further, the IFRS options were re-inserted in Australia for 2007-2008 reports onwards, so we investigate whether Australian companies continue with the 'Australian' policies onthese two issues even when they are not required to. From each of the five countries, we select the largestlisted companies by examining the members of the 'blue chip' indices, respectively the ASX 50, CAC 40, DAX30, IBEX 35 and FTSE 100. To some extent, the different number of companies in the indices adjusts fordifferences in the size of stock markets. We exclude foreign9 companies and those that do not use IFRS. Theonly country for which the last point was a significant issue was Germany where seven of the DAX 30 used USGAAP. After these exclusions, we have a sample of 232 IFRS reports. This is a much larger set of companiesfor our five countries than used by KPMG and von Keitz (2006) or by ICAEW (2007). Also, our sample is acomplete set of domestic IFRS reporters in the indices, whereas the samples in the professional studies arelikely to suffer from some selection bias, as already noted. Table 1 shows the sectoral distribution of the samplecompanies, analysed by country. The selection of large companies is justified for both conceptual and practicalreasons. Large companies are probably more attentive than smaller companies to the requirements andexpectations of the global investor community (e.g. Chaplinsky and Ramchand, 2000;Wu and Kwok, 2002).Therefore, international notions about 'best practice' under IFRS will spread more rapidly among the largecompanies. For that reason, whenever we observe national differences in practice among the largestcompanies, we expect that similar differences exist among smaller firms, which are less likely to feelinternational influences. For the topics discussed in this paper, we can make inferences from the samples oflarge companies to the whole IFRSreporting population that we could not make as easily the other way round.As noted earlier, sectoral issues affect some accounting policies. For example, in the EU there are threedifferent versions10 of the Fourth Directive (for banks, insurance companies and others) which contain pre-IFRS requirements on many presentation and policy issues. For such (and other) reasons, many empiricalstudies exclude banks and other financial institutions. As a result, these companies are under-researched. Weinclude them. However, for several of the policy issues that we study, it is obviously appropriate to treat thebanks or financial institutions separately. We examine the annual reports for 2005-2006, that is those relating toaccounting years starting in 2005. Many of these years begin on 1 January, but some UK companies havechosen other dates (especially 1 April), and many Australian companies use 1 July. The 2005-2006 reportswere the first for which IFRS was compulsory11 in our five countries, and they were also the last full setavailable12 when we collected our data. All the companies were subject to the same requirements, as therewere no changes to IFRS in this period. Nobes (2006) identifies eight types of opportunity for internationalvariations in IFRS practice: different versions of IFRS; different translations of IFRS; gaps in IFRS; overtoptions; covert options; measurement estimations; transitional issues; and imperfect enforcement. For severalof these, detailed lists are provided: e.g. 18 overt options, 21 covert options. From these lists, we identified allthe issues13 for which data are observable in published annual reports. The resulting 16 issues of accountingpolicy are shown as Table 2. Nine of these relate to presentation and seven to measurement. Of thepresentation issues, some are cosmetic (such as issues 4, 5 and 6), whereas others (such as issues 2, 3, 8 and9) directly affect the content of key items within the income and cash flow statements. For our purpose, it isimportant to collect all the available information on international differences in IFRS policies. This is because onsome other major issues, e.g. the criteria for assessing impairments (Ball, 2006: 17), it is not possible to detectand measure differences. The more that we can demonstrate systematic international differences for issues thatcan be observed (however important or otherwise), the more we can be confident that there will be differencesfor important issues that cannot be observed. Our policy issues are not, of course, a random selection. They aredeliberately chosen as those for which IFRS offers a choice and for which the chosen policy is observable. Weare not claiming that the adoption of IFRS has led to no standardisation of practice. We are investigatingwhether there remain substantial systematic international differences in practice even under IFRS. The datarelating to the 16 accounting policy issues of Table 2 are not available on any database and were hand-picked

08 March 2013 Page 53 of 145 ProQuest

from the annual reports14 for the 232 companies in our sample. For many of the issues, a full set of data wasobtained. For a few issues no data were available for some companies,15 because the issue did not apply orbecause of poor disclosure. 5. Hypotheses As explained in Section 3 we have a general null hypothesis ofsimilar IFRS practice across countries that we analyse by a chi-square test. In addition, we make pair-wisecomparisons between countries on all of the 16 issues covered by our study. The hypotheses underlying thesecomparisons are explained below. Our expectation is that pre-IFRS national practices will continue. We brieflyreview these practices and then set out our predictions for the 16 issues of Table 2. In nearly all of ourhypotheses below, the pre-IFRS practices that we refer to result from national requirements. We assume thatpractices conformed with requirements (especially for these easily visible practices of listed companies, whichwere all audited by Big Four audit firms). In three cases (issue 13 for Spain, and issues 4 and 14 for the UK), werefer to predominant pre-IFRS practice. Strictly speaking, we should refer, company by company, to the actualpre-IFRS practices. So, in Section 6, we do ask whether particular companies continued with their pre-IFRSpolicies, but that detailed approach is not necessary for the general prediction of the IFRS practices ofcompanies. 1. Presentation of income statements (non-financials). The Spanish law of 1989 sets out a by-nature format for the income statement. By contrast, the pre-IFRS rules in all the other countries allowedbynature or by-function.We therefore predict for IFRS practice that: H1: Spanish companies are more inclinedthan others to use the by-nature format of the income statement. This is an important issue for analysis becauseit is not possible for users to obtain the same information16 from the two different formats. 2. Operating profitshown (non-financials). Pre- IFRS national regulations on formats differ on whether a sub-total for 'operatingprofit' should be shown. There is such a line in the French plan comptable général (section I.III.III) and in theSpanish plan derived from it. Similarly, the formats found in the German Handelsgesetzbuch (HGB } 275) andthe UK Companies Act17 show operating items (specifically thus labelled) separately from others, althoughwithout specifically showing the subtotal. By contrast, there were no such headings or subtotals in Australia inASRB 1018 (para. 4.1). We therefore predict: H2: Australian companies are less likely than other companies toshow a line for operating profit. 3. Treatment of equity-accounted profits (nonfinancials). IAS 1 has few formatrequirements for the income statement. However, its non-mandatory implementation guidance shows equity-accounted profit after finance costs and therefore outside of operating profit. The same applies to the Frenchplan comptable général (Appendix to Chapter IV) and the related Spanish requirements. By contrast, theGerman HGB (} 275) and the UK standard (FRS 9) show such profits after operating but before finance items.There has been no clear tradition in Australia. Many Australian companies do not show an 'operating' heading(see 2 above). Pre-IFRS guidance from AASB 1018 (para. 4.1) showed equity-accounted profits after financecosts, as do IFRS illustrations18 from Australian audit firms. Given that the French and Spanish nationalrequirements show equity-accounted profits lower down the income statement than in the other countries whichhave a concept of 'operating', and given that only in those countries are the requirements mandatory (for non-IFRS reporting), we predict: H3: French and Spanish companies are more inclined than others to show equity-accounted profits after finance items. 4. Presentation of balance sheets (non-financials). The pre-IFRSrequirement in Australia was in the Corporations Law (and AASB 1034) which specified a format that showed'net assets' but no total of credit balances such as total shareholders' funds and liabilities. The same applied inFormat 1 of the UK Companies Act 1985 (CA 1985), which also showed 'net current assets' and did not showtotal assets. This was the predominant format used in practice (Gordon and Gray, 1994: 76; and our own surveyof pre-IFRS policies of our companies, discussed later). These can be called 'financial position' formats,although the UK's was a purer form. By contrast, the accounting plans of France and Spain showed a two-sidedT-account format, and the German Handelsgesetzbuch (HGB } 266) had a vertical version of this. In all threecontinental cases, there is no heading for 'net assets' but there is a heading for the total of the credit balances.We therefore predict: H4: Australian and UK companies are more inclined than others to use a version of afinancial position format. 5. Liquidity order (non-financials). The pre-IFRS regulations for balance sheets

08 March 2013 Page 54 of 145 ProQuest

(referred to above) show items in order of decreasing liquidity in Australia but (except for banks) increasingliquidity in the other four countries. Therefore, we predict: H5: Australian companies are more inclined thanothers to present liquidity-decreasing balance sheets. 6. Statement of changes in equity. Only in the UK did pre-IFRS rules (FRS 3) require a performance statement in addition to the income statement. This UK statementwas the model for IAS 1's statement of recognised income and expense (SORIE) - equivalent to the 'othercomprehensive income' of a later version of IAS 1- as opposed to the alternative statement of changes in equityof IAS 1 (para. 97).19 So, we predict: H6: UK companies are more inclined than others to present a SORIE. 7.Method of calculation of operating cash flow. Pre-IFRS rules on cash flow statements were lacking in detail inGermany and were noticeably different from IAS 7 in France, Spain and the UK. However, only in Australia wasthe direct method required (AASB 1026) and this found its way into the Australian version20 of IAS 7 that was inforce in 2005-2006. However, IAS 7's choice of the indirect method was re-inserted into the version in force in2007-2008, so we used data for that period for the Australian companies (typically periods ending on 30 June2008). Given that the direct method is more onerous for preparers, we predict a continued21 avoidance of itelsewhere: H7: Australian companies are more inclined than others to use the direct method to calculateoperating cash flows. 8/9. Presentation of dividends received and interest paid in cash flow statements (non-financials). IAS 7 (para. 33) suggests that dividends received might be either operating or investing flows andthat interest paid might be either operating or financing, except that financial companies 'usually' treat them bothas operating. In the UK, FRS 1 (para. 14) requires both dividends received22 and interest paid to be shown as'returns on investments and servicing of finance'. In Australia, AASB 1026 (para. 7.1) required both cash flowsto be shown as operating. The French requirement23 for consolidated statements is also to show bothdividends received and interest paid as operating. There is no requirement for a cash flow statement in Spain;rather the law24 requires a statement of sources and applications of funds. In Germany, cash flow statementsare required for listed companies (from 1999 onwards) but the pre- IFRS rules lack detail. There is therefore noclear national practice for Germany or Spain, but we can predict: H8: UK companies are less likely thanAustralian or French ones to show dividends received as operating. H9: UK companies are less likely thanAustralian or French ones to show interest paid as operating. 10. Use of fair value to measure property, plantand equipment (PPE). Pre-IFRS requirements in France, Germany and Spain were to base measurement onhistorical cost except for occasional revaluations in France and Spain according to government regulations(TRANSACC, 2001: 1162, 2263). Only in Australia (AASB 1041) and the UK (FRS 15) was revaluation freelyallowed. We predict: H10: Australian and UK companies are more inclined than others to measure PPE at fairvalue. 11. Use of fair value to measure investment property. As for other PPE (above), pre-IFRS requirementsin France, Germany and Spain were generally to measure investment property on a cost basis. However, as forother PPE, there was an option to use fair value in Australia. By contrast, continuous valuation25 is requiredunder UK GAAP by SSAP 19. We therefore predict: H11: The tendency to measure investment property at fairvalue will be found in decreasing order in the UK, Australia and continental Europe. 12. Designation of financialassets to fair value (non-financials). Pre-IFRS requirements concerning the measurement of financial assets bynonfinancial companies differed by country. German law required measurement at cost or lower for all assets(HGB } 253). French and Spanish accounting laws were less resolutely opposed to measurement above cost,26so we use Germany in the hypothesis below. UK law allowed various versions of market value (CA 1985, Sch.4, para. 31). UK standards and Australian law and standards had no requirements in this area. We thereforepredict for nonfinancial companies: H12: Australian and UK companies are more inclined than Germancompanies to designate financial assets to fair value. Financial institutions had different laws (for example,different Directives; see Section 4 and Hypothesis 5) allowing marking to market. We do not test hypotheses forfinancial institutions because of the small number of such companies in our sample of continental countries. 13.Interest capitalisation. The pre-IFRS requirement in Australia was to capitalise interest (AASB 1036). In Spain,the ICAC Resolution of 30 July 1991 deals with the issue in some detail, and it was almost universal pre-IFRS

08 March 2013 Page 55 of 145 ProQuest

practice of our companies. 27 By contrast, in the other three countries, capitalisation of interest was allowed28but was not covered in detail in the regulations and was less common.29 We predict: H13: Australian andSpanish companies are more inclined than others to capitalise borrowing costs on construction. 14. Inventoryflow assumptions (non-financials). Excluding consideration of last in, first out (LIFO) (which is not allowed in IAS2), the UK and Germany stand out as having predominant flow assumptions in pre-IFRS national practice. Inthe UK, FIFO was the normal practice (FEE, 1991: 164; as also confirmed in our own survey of pre- IFRSpolicies). In Germany, weighted average was generally required by tax law (TRANSACC, 2001: 1293). In Spain,although there was no legal favouring of weighted average, there was also evidence of a clear pre-IFRSpreference for it (FEE, 1991: 167; Gonzalo and Gallizo, 1992: 114). In the other two countries, no predominantbasis was clear. We therefore predict: H14A: German companies are more inclined than others (exceptSpanish) to use weighted average only. H14B: UK companies are more inclined than others to use FIFO only.15. Actuarial gains and losses. Most German DAX companies were already using IFRS before 2005 when anextra option was added to IAS 19 (para. 93A) to allow actuarial gains and losses to be taken in full to theSORIE. Therefore, they were already using the corridor approach (IAS 19.92/93). By contrast, the pre-IFRSrequirement in the UK (under FRS 17) was the same as the SORIE option. Neither of these options wasavailable in the laws of the other three countries. So, we predict: H15A: German companies are more inclinedthan others to use the corridor approach. H15B: UK companies are more inclined than others to use the SORIEapproach. 16. Proportional consolidation. Pre-IFRS rules in Australia (AASB 1006) and the UK (FRS 9) did notallow proportional consolidation of interests in joint venture entities. By contrast, pre-IFRS French regulationsrequired proportional consolidation (Loi sur les Sociétés Commerciales, Art. 357-3). In Spain, the method wasrequired30 in some industries and common in others (Gonzalo and Gallizo, 1992: 168). In Germany,proportional consolidation was allowed and used by some groups (TRANSACC, 2001: 1389). However, it wasnot typical practice, as it had been banned in Germany until 1987. We, therefore, predict: H16: The tendency touse proportional consolidation is found in the following countries in decreasing order: France, Spain, Germany,UK and Australia. As in policy issue 7 above, there is a complication with the data for Australia. In the Australian2005-2006 version of IFRS (i.e. AASB 131 in this case), the proportional option in IAS 31.30 was deleted. Sowe cannot measure policy choice for 2005-2006. However, the option was re-inserted for 2007-2008, so we usedata for that period. 6. Results 6.1. Tests of hypotheses Table 3 shows the results of testing the abovehypotheses. First, it summarises the data collected for the five countries relating to all the issues of Table 2. Foreach country and issue, the table shows the policies used, as percentages of the companies for which thepolicy was observable (see the 'N'). In most cases, the data can be reduced to the percentages using one policyout of two available in IFRS, although in a few cases (e.g. issue 3) we record the scores for three possibilities.As explained earlier, we conduct two sorts of statistical tests on these data. The chi-square test measures theoverall independence between policy choice and country for each of the 16 issues. The null hypothesis ofsimilar practice is rejected at the 1% level for 14 issues and at the 5% level for two of them. Table 3 also showsthe results of the binomial tests. The testing of issues that have two choices is carried out with conventionalmethods of approximations to the normal distribution. In practice we do the same tests for the issues that havemore than two choices, by formulating the hypothesis with respect to one specific alternative. It follows from theidea underlying the alternative hypotheses that, for testing purposes, the sample of companies from eachcountry should be treated as separate populations. When, for example, it is claimed that Spanish companiesare more inclined than others to present an income statement by nature, the related testing consists of pair-wisecomparisons between the scores of the Spanish sample and the scores of each of the other samples, i.e. a totalof four tests. If we had been certain that the null hypothesis were true for all companies except Spanish ones,we might, of course, have pooled the scores of the latter four for the purpose of the testing. However, whetherthe statistical distributions are identical or not is precisely the question we seek to answer, and the consequenceis that all samples are treated separately. As we have designed the statistical analysis, each of the four tests

08 March 2013 Page 56 of 145 ProQuest

should result in a non-rejection of the null hypothesis if it is true. One problem that we encounter by this pair-wise approach is that some of the samples compared are under the threshold recommended for approximationsto the normal distribution (typically 25, see for example, Bhattacharyya and Johnson, 1977: 295). In this studyeach single test is not essential for the conclusion, so we report all results, being aware that some of them maybe based on insufficient sample sizes. Thus, hypothesis H1, which proposes that Spanish companies have agreater tendency than others to use an income statement by nature, was tested pair-wise for each of the otherfour countries. In all four cases, the null hypothesis that Spanish choices are the same as others is rejected atthe 1% level. We ran 82 binomial tests. Of these, 62 led to the rejection of the null hypotheses at the 1% level,and a further 12 tests did so at the 5% level. The remaining eight tests did not enable rejection of the nullhypotheses but in six of these cases the data were consistent with our alternative hypotheses. In sum, there is alarge amount of highly significant evidence that policy choice under IFRS varies internationally and is notrandom. Furthermore, we have shown that the national profile of IFRS practice is explained by national pre-IFRS requirements (or predominant practice). 6.2. Comparisons with pre-IFRS practices The above hypotheseslargely concern the continuation of policies previously required by national GAAP. The only cases where werelied on predominant national choices for our hypotheses related to Spain (issue 13) and the UK (issues 4 and14). A more precise hypothesis is that a particular company continued with its particular pre-IFRS policy choice.To test this, we looked at all the 2004-05 (pre-IFRS) reports of our Spanish and UK companies for these issues.For issue 13 (capitalisation of interest), 94% of Spanish companies31 maintained their pre-IFRS practice ofcapitalisation. For issue 14 (FIFO, weighted average or a mixture), all 6932 UK non-financial companies madeexactly the same policy choice under IFRS as they had done pre-IFRS. For issue 4 (balance sheet format), 88%of the UK companies maintained their policies.33 On the assumption (defended earlier) that, pre- IFRS, ourcompanies complied with national requirements, Table 4 shows the policy switches under IFRS. As can beseen, there are few such switches (e.g. less than 3% of policies were switched by Spanish companies). If weadd in the other policy issues for which there was no national requirement, by studying the pre-IFRS practicesof the particular companies, we find similar results.34 7. Conclusions The central objective35 of the IASB is tofoster the provision of comparable financial information for participants in the world's capital markets. This wouldbe achieved if similar transactions were measured and presented similarly throughout the markets, i.e. uniformpractice. The existence of systematic differences in accounting policies due to non-economic characteristics -such as country of incorporation - is clearly contrary to this ambition. This paper highlights 16 accounting issuesfor which the literature identifies international differences in pre-IFRS reporting, and for which variation withinIFRS is allowed and is observable if it occurs. An examination of the policies used by all the domestic IFRSreporters in the stock indices of five major capital markets for the first year of compulsory IFRS adoption allowsus confidently to conclude that IFRS practice is subject to systematic differences across countries. Thecontinuation of national traditions seems to explain variations in IFRS policies between countries. However, thatis merely a proxy for our more precise hypothesis that a particular company continued with its pre-IFRS policieson transition to IFRS. For each non-financial company of the five countries, we compare its practice under IFRSwith pre-IFRS requirements and find that there were few deviations from those earlier requirements. When weextend this to look at policy switches even where there was no pre-IFRS requirement, we again confirm thepreservation of previous practices. Our research shows that systematic differences exist both in trivial matters(such as the liquidity order of the balance sheet) and in more complex matters (such as the composition of cashflows from operations or the treatment of actuarial losses). Whereas the former are hardly any obstacle tocomparability, the latter most likely are. Some of our policy issues are not as important as others but theybolster the evidence for the existence of national versions of IFRS practice. This allows strong inferences to bemade about variation in practices that cannot easily be measured, e.g. the tendency to make impairments. Webelieve that our results are extensible in various ways. First, we examined very large listed companies. Theseare probably the least likely to evince national practices.We expect that test results would be at least as strong

08 March 2013 Page 57 of 145 ProQuest

for other companies, but this can be examined. Second, our choice of blue chip companies limited the size ofthe sample, especially for Germany. We do not expect that a larger sample would change our results, except tomake them even stronger, but it would enable an extended analysis of whether a company's sector affects itspolicy choices. This would add to our findings of some differences between financial and non-financialcompanies. Third, we examined 16 areas of policy choice. There are many others which are less observable.For example, there are several covert options and estimations in the issue of impairment, such as whether torecognise impairments, how to measure cash flows, and what discount rate to use. National traditions (and,specifically, previous practices of companies) are likely to continue in some of these areas. It is not clear how toexamine these covert options, but other researchers might try. Fourth, we examined five countries, but wouldexpect national versions of IFRS to be observable in smaller IFRS-using capital markets. Fifth, weconcentrated36 on one year's worth of annual reports; mostly37 the first year of IFRS adoption. Later researchcould address whether companies gradually exploit options more fully, at least up to reports published in 2010when some38 IFRS options will no longer be available. To the extent that we look at data after 2005-2006 (for2007-2008 for two Australian issues), we find national practice continuing. The five points above are limitationsof our research. A more general limitation is that, although we can largely explain why particular companiesadopted particular IFRS policies in the context of the transition to IFRS, more work is needed to go deeper intowhy, over a long period, particular policies have been preferred in particular countries or by particularcompanies. For some issues (e.g. the German preference against measurement above cost), the literature isextensive. For others, 'accidents' are the apparent cause (such as the German use of the 'corridor' pensionsmethod because German companies adopted IFRS before other countries did, when the 'SORIE' method wasnot available). For yet other issues (e.g. why the Spanish prefer to capitalise interest), there is no convincingtheory. It might be necessary to theorise about each policy choice, one by one, such that no overall theoreticalmodel would be explanatory. We remind readers of other caveats. First, for most of our hypotheses, we assumethat, pre-IFRS, national requirements were followed. We believe that this is highly likely to be a correctassumption for our large listed companies for our easily observable policy choices. However, there might besome exceptions. Second, for two of the policy issues for Australia, companies had no choice in 2005-2006.Although this does not alter our findings about the continuation of pre-IFRS practices, these are different casesfrom all the other countries and all the other issues. Nevertheless, by substituting data for 2007-2008 (whenthere was a choice), we can rectify that. This time, we need to assume that a company would have chosen in2005-2006, what it chose in 2007-2008. We know of no reason to doubt that. In addition to the possibleextensions to our research resulting from the limitations mentioned above, another opportunity is to examinewhether a two-class model (Anglo-Saxon versus Continental European) that was discussed39 before theadoption of IFRS exists under IFRS. That is, for example, do Australian and UK companies tend to chooseIFRS options in the same way (and perhaps in conformity with US GAAP), at least compared to continentalEuropean companies? A further possibility is to ask why certain companies deviate from pre-IFRS practices andnational profiles, and why this is more common for UK companies than for Spanish ones (see Table 4). Anotherissue to be investigated is whether market participants are able to see through the different policy choices. If itturns out that they can, then the differences are less important for users and regulators. However, investigationsin this area are complex, and previous studies (e.g. on LIFO adoption) are both numerous and inconclusive.There are policy implications from our findings. We believe that options in accounting standards are justified tothe extent that they enable companies with different economic characteristics to produce a fair presentation oftheir activities. The systematic differences in practice between countries, that we document in this paper, are anunwanted corollary. In our view the disadvantages of systematic differences outweigh the advantages of havingoptions, so we encourage the IASB to continue its efforts to remove options. Second, analysts of financialstatements should be alert to the continuing international differences within ostensibly 'international' standards.Analysts might benefit from knowing the national profiles when trying to construct comparable figures. For

08 March 2013 Page 58 of 145 ProQuest

example, it might be helpful to know that French IFRS companies tend to proportionally consolidate the cash ofjoint ventures, to show interest paid as an operating cash flow and to charge actuarial losses as expense,whereas UK IFRS companies tend not to do those things. It might also be helpful to know that other, potentiallymore important, country-related differences (e.g. on impairment) exist beneath the surface of IFRS practice.Third, this variation in IFRS practice is likely to be of interest to regulators, especially to the SEC as it monitorsIFRS practice of foreign registrants and perhaps, in future, of US companies. The SEC's acceptance of IFRShas been made on the assumption40 of further progress in removing options. Sidebar Table 2 IFRS policychoices 1 (a) income statement by function (b) by nature (c) neither 2 (a) inclusion of a line for EBIT oroperating profit (b) no such line 3 (a) equity accounting results included in 'operating' (b) immediately after (c)after finance 4 (a) balance sheet shows assets = credits (b) showing net assets 5 (a) liquidity decreasing inbalance sheet (cash at top) (b) liquidity increasing 6 (a) Statement of Changes in Equity, including dividendsand share issues (b) SORIE, not including them 7 (a) direct operating cash flows (b) indirect 8 (a) dividendsreceived shown as operating cash flow (b) as investing 9 (a) interest paid shown as operating cash flow (b) asfinancing 10 (a) only cost for PPE (b) some fair value 11 (a) investment property at cost (b) at fair value 12 (a)some designation of financial assets at fair value (b) none 13 (a) capitalisation of interest on construction (b)expensing 14 (a) FIFO for inventory cost (b) weighted average 15 (a) actuarial gains and losses to SORIE (b) toincome in full (c) corridor 16 (a) proportional consolidation of some joint ventures (b) only equity methodFootnote 1 In some countries, e.g. Germany, certain companies were allowed to wait until 2007. However, nocompanies that took advantage of this have been included in our study. 2 The main difference between IFRSand EU-endorsed IFRS is greater permission to use hedge accounting in the latter. There are also lags inendorsement. However, none of these differences affects our study. 3We use this acronym to mean 'generallyaccepted accounting practices', i.e. those practices that result from national requirements or from predominantchoices. 4 This publication shows a financial position form of balance sheet (like Format 1 in the UK CompaniesAct) as an example of IFRS practice. 5 Many UK companies do not have 31 December year-ends, so their firstIFRS reports related to years ending in 2006. 6 Australian usage of IFRS began, for most companies, on 1 July2005. 7 For example, see data from the World Federation of Exchanges, as at June 2005. 8 As explainedearlier, the difference between EU-IFRS and IFRS on the subject of IAS 39 is not relevant in our study. 9 Wedefine 'foreign' as meaning not legally registered in the country. That is, for example, we exclude from theFrench sample Belgian-registered companies that prepare IFRS statements in the context of Belgian law.Weexclude Rio Tinto from the Australian sample because it is also in the FTSE 100 and prepares IFRS statementsin the context of UK law. We also exclude AXA Asia Pacific from the Australian sample because it is asubsidiary of AXA (France). 10 The Directives for banks (1986) and insurance companies (1991) are derivedfrom the fourth company law Directive 'on the annual accounts of certain types of companies' of 1978. 11 Of ourfive countries, only Germany contained companies voluntarily using IFRS immediately prior to 2005. A majorityof our sample of German companies used IFRS before 2005. However, we do not anticipate that this wouldaffect policy choices except where new options were introduced in 2005 or shortly before. In our list of policyareas, the only new option was to take actuarial gains and losses to the SORIE, introduced in 2005. It istherefore possible that this recent option was more likely to be ignored by German companies. However, this isstill a country-specific factor. 12 For example, reports for years ending in November 2007 were not availableuntil well into 2008.We collected data during the second half of 2007. 13 The 16 issues are all 'overt options'.Weexcluded six issues from Table 1 of Nobes (2006) because they related to unconsolidated statements (theoptions in IASs 27, 28 and 31 concerning investor statements) or to rare issues on which little or no data wasavailable (commodity broker traders (IAS 2), government grants (IAS 20) and revaluation of intangibles (IAS38)). Similarly, it is not possible to gauge the use of 'covert options' by using published annual reports. However,a few of the overt options in Nobes (2006) cover several issues (e.g. the treatments of interest and dividends incash flow statements). So we have separated them. 14We used the English language reports in all cases, but

08 March 2013 Page 59 of 145 ProQuest

we do not expect that this would affect our data. 15 See the 'N' numbers in Table 3. 16 For example, gross profitcannot be calculated from the bynature format. 17 Schedule 4 to CA 1985, now replaced by 'CompanyRegulations' in Statutory Instruments. 18 For example, KPMG's Reporting Under Australian AccountingStandards, Example Public Company Limited (for 2007), p. 21; and PwC's Value AIFRS Holdings, p. 63. 19Werefer to the version of IAS 1 in force in 2005 and 2006. 20 That ruling in 2005-2006. 21 For example, all our UKcompanies used the indirect method under UK GAAP in 2004-2005; see discussion later. 22 Except thatdividends received from associates and joint ventures are shown separately, also outside of operating. 23Second Methodology, } 426. 24 Law 19/1989. 25 SSAP 19 (para. 11) requires measurement at 'open marketvalue' which is similar to fair value. 26 For example, revaluations of various assets were required for listedcompanies in France in 1978 and in Spain in 1996. 27We surveyed the 2004 annual reports for the companiesthat specified their practice in 2005. Of the 17 companies, 16 capitalised interest, and one company capitalisedsome interest. 28 For example, by AktG } 255(3) in Germany, or CA 1985, Sch. 4, para. 26(3)(b) in the UK. 29For example, our survey of UK reports of 2004-2005 showed that 35% of companies disclosed a policy ofcapitalisation. Only 13% disclosed a policy of non-capitalisation, but our expectation is that this would havebeen the policy of the non-disclosers. 30 At least, information on a proportional basis had to be included in thebalance sheet for joint ventures in the construction industry (TRANSACC, 2001: 2314). 31 Seventeencompanies disclosed a policy for both years. 32 There were 72 non-financial companies in our UK data, butthree of them did not publish UK GAAP reports for 2004-2005. 33 Of the 69 companies, seven changed fromshowing net assets to not doing so, and one changed the other way. 34 For example, for the UK, we add sixmore issues, and find 18.8% switches for all 15 issues.We omit issue 12 because there was no pre-IFRSrequirement or practice on designation. 35 IASB's Preface to International Financial Reporting Standards, para.6. 36 Although, we examined 2008 reports for two Australian issues. 37 Except for the German companies. 38Of our 16 issues, the presentation of the income statement, the capitalisation of interest, and perhaps thetreatment of joint ventures will be affected, but not compulsorily until 2009 yearends or later. 39 See, forexample, the disagreement between d'Arcy (2001) and Nobes (2004). 40 SEC, 2008, II, B. References References Ashbaugh, H. and Pincus, M. (2001). 'Domestic accounting standards, international accountingstandards, and the predictability of earnings'. Journal of Accounting Research, 39(3): 417-434. Ball, R. (2001).'Infrastructure requirements for an economically efficient system of public financial reporting and disclosure'.Brookings-Wharton Papers on Financial Services: 127-169. Ball, R. (2006). 'International Financial ReportingStandards (IFRS): pros and cons for investors'. Accounting and Business Research, International AccountingForum: 5- 27. Ball, R., Kothari, S. and Robin, A. (2000). 'The effect of international institutional factors onproperties of accounting earnings'. Journal of Accounting and Economics, (29): 1-51. Barth, M., Landsman, W.,Lang, M. and Williams, C. (2006). 'Accounting quality: international accounting standards and US GAAP',SSRN. Bhattacharyya, G. and Johnson, R. (1977). Statistical Concepts and Methods. Hoboken, New Jersey,NJ: John Wiley &Sons. Bushman, R. and Piotroski, J. (2006). 'Financial reporting incentives for conservativeaccounting: the influence of legal and political institutions'. Journal of Accounting and Economics, 42: 107-148.Chaplinsky, S. and Ramchand, L. (2000). 'The impact of global equity offerings'. Journal of Finance, 55(6):2767- 2789. Christensen, H., Lee, E. and Walker, M. (2008). 'Incentives or standards: what determinesaccounting quality changes around IFRS adoption?', SSRN. Coopers &Lybrand (1997). AccountingComparisons: UK, Belgium, Italy and Spain. London: Accountancy Books. Cuijpers, R. and Buijink, W. (2005).'Voluntary adoption of non-local GAAP in the European Union: a study of determinants and consequences'.European Accounting Review, 14(3): 487-524. Daske, H., Hail, L., Leuz, C. and Verdi, R. (2007). 'MandatoryIFRS reporting around the world: early evidence on the economic consequences', SSRN. D'Arcy, A. (2001).'Accounting classification and the international harmonisation debate - an empirical investigation'. Accounting,Organizations and Society, 26(4): 327-349. Ding, Y., Hope, O-K., Jeanjean, T. and Stolowy, H. (2007).'Differences between domestic accounting standards and IAS: measurement, determinants and implications'.

08 March 2013 Page 60 of 145 ProQuest

Journal of Accounting and Public Policy, 26(1): 1-38. European Commission (2008). Evaluation of theApplication of IFRS in the 2006 Financial Statements of EU Companies, at http://ec.europa.eu/internal_market/accounting/docs/studies/2009-report_en.pdf. FEE (1991). European Survey of Published Accounts 1991.London: Routledge. Fields, T., Lys, T. and Vincent, L. (2001). 'Empirical research on accounting choice'. Journalof Accounting and Economics, 31: 255-307. Gassen, J. and Sellhorn, T. (2006). 'Applying IFRS in Germany -determinants and consequences'. Betriebswirtschaftliche Forschung und Praxis, 58(4). Gonzalo, J.A. andGallizo, J.L. (1992). European Financial Reporting: Spain. London: Routledge. Gordon, P.D. and Gray, S.J.(1994). European Financial Reporting: United Kingdom. London: Routledge. Healy, P. and Wahlen, J. (1999). 'Areview of the earnings management literature and its implications for standard setting'. Accounting Horizons,13(4): 365-383. IASCF (2005). Constitution, London: International Accounting Standards CommitteeFoundation, para. 2 (a). ICAEW (2007). EU Implementation of IFRS and the Fair Value Directive. London:Institute of Chartered Accountants in England and Wales. KPMG and von Keitz, I. (2006). The Application ofIFRS: Choices in Practice. London: KPMG. Küting, K. (2007). 'Unterschiedliche Erfolgs- und Gewinngrössen inder internationalen Rechnungslegung: was sollen diese Kennzahlen aussagen?'. Der Betrieb, 47, 23November: 2549-2557. Leuz, C. and Verrecchia, R. (2000). 'The economic consequences of increaseddisclosure'. Journal of Accounting Research, 38 (Supplement): 91-124. Nobes, C.W. (2001). GAAP 2001: ASurvey of National Accounting Rules Benchmarked against International Accounting Standards. London: ArthurAndersen and other firms. Nobes, C.W. (2004). 'On accounting classification and the internationalharmonisation debate'. Accounting, Organizations and Society, 29(2): 189-200. Nobes, C.W. (2006). 'Thesurvival of international differences under IFRS: towards a research agenda'. Accounting and BusinessResearch, 36(3): 233-245. PwC (2005). IFRS/UK Illustrative Financial Statements for 2005. London:PricewaterhouseCoopers. SEC (2008). Acceptance from Foreign Private Issuers of Financial StatementsPrepared in Accordance with International Financial Reporting Standards without Reconciliation to U.S. GAAP,RIN 3235-AJ90. Washington, DC: Securities and Exchange Commission. TRANSACC (2001). TransnationalAccounting, edited by D. Ordelheide and KPMG, New York, NY: Palgrave. Wu, C. and Kwok, C. (2002). 'Whydo US firms choose global equity offerings?'. Financial Management, Summer: 47-65. Zeff, S. (2007). 'Someobstacles to global financial reporting comparability and convergence at a high level of quality'. BritishAccounting Review, 39(4): 290-302. AuthorAffiliation Erlend Kvaal and Christopher Nobes* *Erlend Kvaal is atthe Norwegian School of Management, BI and Christopher Nobes is at Royal Holloway, University of London.The authors are grateful for comments on previous drafts from Steinar Kvifte (Ernst &Young), John ChristianLangli (BI), R.H. Parker (University of Exeter) and Christian Stadler (Royal Holloway), from participants inworkshops at the UK's Financial Reporting Council, BI and the 2008 European Financial Reporting Conferencein Lund, and from two reviewers and the editor of this journal. Christopher Nobes acknowledges researchsupport from ACCA. Correspondence should be addressed to: Prof. C.W. Nobes, School of Management,Royal Holloway, Egham TW20 0EX, UK. E-mail: chris.nobes@rhul.ac.uk. This research note was accepted forpublication in August 2009. Subject: Studies; International Financial Reporting Standards; Accounting policies; International standards Classification: 4120: Accounting policies&procedures, 9180: International, 9130: Experiment/theoreticaltreatment Publication title: Accounting and Business Research Volume: 40 Issue: 2 First page: 173

08 March 2013 Page 61 of 145 ProQuest

Number of pages: 15 Publication year: 2010 Publication date: 2010 Year: 2010 Publisher: Taylor&Francis Ltd. Place of publication: Abingdon Country of publication: United Kingdom Journal subject: Business And Economics--Accounting ISSN: 00014788 CODEN: ACBRB5 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Tables;References ProQuest document ID: 501760689 Document URL: http://search.proquest.com/docview/501760689?accountid=86413 Copyright: Copyright Wolters Kluwer (UK) Ltd. 2010 Last updated: 2011-10-04 Database: Accounting&Tax

_______________________________________________________________ Document 6 of 10 CITIZEN PARTICIPATION IN THE BUDGET PROCESS: THE EFFECT OF CITY MANAGERS Author: Zhang, Yahong; Yang, Kaifeng Publication info: Journal of Public Budgeting, Accounting & Financial Management 21. 2 (Summer 2009): 289-317.ProQuest document link Abstract: Much of the literature on citizen participation in the budget process links the council-manager form ofgovernment with higher levels of citizen participation, assuming the council-manager form representsprofessional administration. This is contradictory to the reality that different forms of government have"borrowed" features from each other and many now have mixed forms of government (i.e., adapted). Theliterature also contains ambiguities about city managers' role in participatory budgeting. We review the literatureand identify three competing theories about the role of professional managers in the budget process. Wedirectly examine the effect of city managers in terms of their professional dimensions, institutional environment,and individual willingness to represent citizens. Using survey data from Florida, we demonstrate that managers'professionalism, perceived political environment, and attitude toward citizen input are important factorsexplaining local governments' adoption of participatory budgeting. [PUBLICATION ABSTRACT]

08 March 2013 Page 62 of 145 ProQuest

Full Text: Headnote ABSTRACT. Much of the literature on citizen participation in the budget process links thecouncil-manager form of government with higher levels of citizen participation, assuming the council-managerform represents professional administration. This is contradictory to the reality that different forms ofgovernment have "borrowed" features from each other and many now have mixed forms of government (i.e.,adapted). The literature also contains ambiguities about city managers' role in participatory budgeting. Wereview the literature and identify three competing theories about the role of professional managers in the budgetprocess. We directly examine the effect of city managers in terms of their professional dimensions, institutionalenvironment, and individual willingness to represent citizens. Using survey data from Florida, we demonstratethat managers' professionalism, perceived political environment, and attitude toward citizen input are importantfactors explaining local governments' adoption of participatory budgeting. INTRODUCTION Participatorybudgeting is a process of democratic policy-making in which the government invites citizen inputs during thebudget process and allow their influence in budget allocations. Participatory budgeting has drawn significantattention from public administration practitioners and scholars in recent years. According to the WorldwatchInstitute (2007), about 1200 municipalities around the world had adopted participatory budgets by 2007. Basedon a survey of U.S. cities with populations greater than 50,000 in late 1999 and early 2000, Wang (2001) foundthat 46.2% of the respondents reported that their cities involved citizens or citizen activists in the budgetingfunction. In a more recent 2004 survey, Yang and Callahan (2005) found the adoption rate increased to 66% forcounties/municipalities with populations from 25,000 to 49,999 and from 250,000 to 499,999. Ebdon andFranklin (2006) studied factors that affect the adoption of citizen budgets and proposed an impact model ofcitizen participation in budgeting (see also Ebdon, 2000; Franklin &Ebdon, 2005). However, evidence isinconclusive as to why some local governments include citizen participation in the budget process while othersdo not. Ebdon and Franklin (2006, p.445) acknowledge "we have relatively little generalizable empiricalknowledge about the use of participation in budgeting". We argue that the question of what factors drive a localgovernment to invite citizen participation deserves more attention from public administration scholars. Even forsome topics that Ebdon and Franklin (2006) consider "what we already know" rather than knowledge gaps, theevidence is not definite and more systematic examinations are necessary. For example, Ebdon and Franklinassert that the council-manager form of government is more likely to solicit citizen input, but the statement,largely based on results from interviews (Ebdon, 2002) or case studies (Franklin &Ebdon, 2005), is difficult togeneralize. Furthermore, the assertion is not consistent with the empirical results from Wang's 2001 study thatfinds no effect of the council-manager form. We submit that the controversy of the impact of the form ofgovernment reveals the necessity to investigate in-depth mechanisms and motivations within the localgovernment context, especially the role of city managers in the decision process. As an initial step, this articleaims to clarify the linkages between city managers and citizen participation in budgeting, examining whetherand how city managers' professional characteristics and attitudes affect local governments' adoption of citizenparticipation in the budget process. After the literature review, the second section of this article outlines themethodology, followed by the findings and discussions. The final section draws the conclusion. THEAMBIGUITY IN THE LITERATURE In a much-needed article, Ebdon and Franklin (2006) develop an impressivetypology of elements and variables that are important in describing and explaining citizen participation in thebudget process in terms of its adoption, process design, mechanisms, goals, and outcomes. They suggest thatthe form of government makes a difference in participatory budgeting adoption in the way that "the council-manager form of government appears to be more likely to solicit input" (p. 439). To substantiate this conclusion,Ebdon and Franklin (see also Franklin &Ebdon, 2005) refer to several prior studies. One is the book by Kweitand Kweit (1981), which observes that "with the presence of a full-time professional administrator, [it] is morelikely to seek citizen input than other forms of government" (Ebdon &Franklin, 2006, p. 169). Another is anarticle by Ebdon (2002), who finds council-manager cities are more likely to use formal budget-participationmethods. They also make reference to Nalbandian (1991; 1999), who finds that city managers respond to a

08 March 2013 Page 63 of 145 ProQuest

variety of community values and have increasingly treated community participation as an importantadministrative value and task. However, those studies do not provide evidence that is strong enough to supportthe argument. Kweit and Kweit's (1981) observation of the role of the form of government is based on casestudies. Ebdon's (2002) results are based on interviews with budget directors in only 28 Midwestern cities.Nalbandian (1999) does conclude that "community building has become part of the city managementprofessional's role and responsibility" and that "managers are increasingly expected to facilitate participation" (p.187). However, Nalbandian also states "there is less adherence to the council-manager plan as the 'one bestform' of government" (p. 187). He points out that adaptations to both council-manager and strong mayor formsof government have moderated the distinctions between the two and questions whether the remainingdifferences do have an impact. In the literature of citizen participation in general (as opposed to the literature inthe budget process in particular), the form of government is treated as an important predictor (e.g., Cole, 1974;Greenstone &Peterson, 1971; Streib, 1992; Wang, 2001). However, empirical research has left ambiguitiesabout the impact of government form on citizen participation. Greenstone and Peterson (1971), for instance,conclude that council-manager cities have more citizen participation than strong-mayor cities becauseinformation necessary to mobilize and empower citizens is often withheld in the latter (also see Streib, 1992). Incontrast, Wang (2001) does not find evidence of such a correlation, regardless of the dimensions ofparticipation. Cole (1974) even observes that council-manager governments have less, rather than more,participation. Yang and Callahan (2005) reveal that the impact of the council-manager form depends on thedimensions of citizen involvement: council-manager governments are more likely to adopt involvementmechanisms such as public hearings, community meetings, and citizen surveys; but they do not differ fromother governments with regard to citizen involvement in strategic decision making, management, and servicedelivery. There are many potential solutions, theoretical and methodological, to deal with the ambiguities in theliterature, but one alternative is to question whether the form of government is a strong predictor of governmentbehaviors, especially when different forms of government have "borrowed" features from each other.Frederickson, Johnson, and Wood (2004) find that in the past twenty years mayorcouncil cities have rapidlyincreased the use and the powers of chief administrative officers (CAO), powers similar to those of the citymanager in the council-manager plan. They call this type of cities "adapted" ones or mixed forms ofgovernment. As they correctly observe, a majority of the cities are "adapted" and have a professional managerposition. Thus, the claim that the councilmanager form is better fitting for citizen participation does not squarewith the reality of "adapted cities." In addition, the underlying assumption for using the form of government as apredictor is that city managers, who are centralized professional executives in charge of local governments'daily operation, have a distinct role (Lubell, Feiock, &Ramirez, 2005). It is somewhat perplexing why theliterature has largely used the dichotomous variable of the form of city government rather than including directvariables about city managers. It would seem natural to directly assess the mechanisms and motivations withinthe position of city manager, which may shape the adoption of citizen participation. Unlike most of the previousstudies, we are particularly interested in the role of city managers in involving citizens in the budget process. Weare not alone in paying attention to city managers' characteristics and perceptions. For example, Yang andCallahan (2007) perceive that chief administrative officers' attitude toward citizen involvement is perhaps themost important explanatory factor in accounting for the level of citizen involvement in local governments. Wang(2001) observes that managers' and employees' willingness to be accountable is positively associated with theadoption of citizen participation. Marlowe and Portillo (2006) assume that city managers are important for citizenparticipation in local governments because if they "do not view participation as adding value to budget decisionprocesses, they may discount it or even discourage it" (p. 180). Marlowe and Portillo (2006) focus on an after-adoption issue: how city managers perceive citizen inputs once the inputs are produced. None of the previousstudies, however, include variables such as city managers' professional education, professional networking,professional experience, and institutional authority. This article attempts to fill that gap. THE ROLE OF CITY

08 March 2013 Page 64 of 145 ProQuest

MANAGERS Competing Perspectives There are very different perspectives about how city managers'characteristics might impact citizen involvement. Some scholars hold a "positive" perspective and believe thatcity managers are likely to encourage citizen participation. One reason is that city managers tend to be"modernizers" or public entrepreneurs who seek to experiment with scientific management tools (Berman&West, 1995; Feiock, 2003; Poister &Streib, 1989). Citizen participation in budgeting could be viewed as amanagement innovation. Another reason is that community building and participation have become aprofessional norm for management professionals in local government. Therefore, appointed managers mayemphasize citizenship values over technocratic values (Nalbandian, 1991; 1999). We can label this firstperspective as the "citizen leadership" model. Another perspective is "negative" in that it is concerned with thetension between professional administration and citizen involvement (DeSario &Langton, 1984; Fischer, 2000;Kweit &Kweit, 1981; Simonsen &Robbins, 2000). For instance, Fischer (2000) indicates that "the tensionbetween professional expertise and democratic governance is an important political dimension of our time" (p.ix). As public problems become highly sophisticated in modern society, policy processes are increasinglydominated by professional experts. Such technocratic dominance, however, is likely to hamper citizenparticipation because administrative decision-making based on expertise and professionalism may leave littleroom for participatory processes. We can call this perspective the "technocratic expert" model. From thisperspective, one might argue that since budgeting is a central and complex management function (O'Tool&Marshall, 1988; Simonsen &Robbins, 2000), professional administrators may fear that citizen involvementreduces administrative efficiency, and, as a result, they may discourage citizen involvement in budgeting (Bland&Rubin, 1997). The technocratic model echoes the writings on bureaucratic personality and bureaucraticexperience. In Hummel's (1994) description, bureaucracies are in a "cold" environment in which employees aresupposed to have no personal feelings, emotions, or judgments and treat various clients as cases without anydistinction. Following Hummel (1994), Alkadry (2003) contends that professional administrators becomeindifferent to citizen needs because of their bureaucratic personality. That is, their responsiveness to citizens isconstrained by their inability to take action or their unwillingness to take action given that they are constantlywatched by their supervisors and governed by strict rules and job descriptions. Alkadry (2003) and Hummel(1994) aim to build a general theory and treat all bureaucratic administrators as the same regardless of thelevels of government. We can call their theory the "bureaucratic indifference" model. According to this model,city managers' personality and behaviors are shaped by their professional experience in a way that theirtendency toward citizen participation in the budget process is constrained by their inability and theirunwillingness to involve citizens. Yang and Callahan (2007) try to integrate the citizen leadership model and thetechnocratic expert model in examining the factors driving citizen participation in local governments. In contrastto the bureaucratic indifference model, they suggest that chief administrative officers may internalize theprofessional values promoting community building and civic engagement as Nalbandian (1991; 1999) observes,and in turn, proactively seek citizen input. However, Yang and Callahan (2007) acknowledge the technocraticexpert model may also play a role, indicating that the citizen leadership model may explain better whether thereare citizen participation activities while the technocratic expert model may explain better whether citizen inputwill actually make a difference in decision outcomes: It is likely that professional managers treat involvementmechanisms as professional management tools and use them to obtain customer feedback and improve servicequalityAfter the mechanisms are put into place, however, whether and how citizen input is used in strategicdecisions depends on the political and institutional dynamics of a particular community. In particular,professional managers may fear that citizen involvement in strategic decisions will reduce their authority andpower (Yang &Callahan, 2007, p. 259). Hypotheses The models of "citizen leadership," "technocratic expert,"and "bureaucratic indifference" provide different theoretical perspectives to think about how professionaladministration affects city managers' behavior in regard to involving citizens in the budget process. Consideringthe three competing perspectives, we are interested in empirically testing three questions: 1. As city managers

08 March 2013 Page 65 of 145 ProQuest

become more professional, are cities less likely to open the budget process to citizen involvement? (thetechnocratic expert model) 2. As city managers are more constrained by their inability to take action within thegovernment structure, are cities less likely to open the budget process to citizen involvement? (the bureaucraticindifference model) 3. As city managers become more willing to listen to citizens, are cities more likely to openthe budget process to citizen involvement? (the citizen leadership model) City managers' professionalism canbe indicated by their professional education, participation in professional associations, and professionalexperience. Professional education is an important component of professionalism because it is supposed toenrich students with professional skills and professional ethics (DiMaggio &Powell, 1989). Evidence shows thatthe values and preferences of city managers are shaped by the modern norms of professional associations andpublic administration schools, which may stress ideas differently from traditional perceptions (e.g., Lubell,Feiock, &Ramirez, 2005; Nalbandian, 1999). Considering the fact that masters of public administration (MPA)programs' faculty members generally support democratic administration and a larger role for citizenparticipation, we are inclined to propose that city managers' professional education (MPA degree) wouldincrease their willingness to promote citizen participation in the budget process. H1: City managers' professionaleducation is positively associated with the cities' adoption of citizen involvement in the budget process.Professional affiliation or networking has been found to shape physicians' value and ideology about theirprofessional service (e.g., Del Gaudio, Stein, Ansley, &Carpenter, 1975; Roback, Purdon, Ochoa, &Bloch,1993). While city managers and physicians are in distinctly different professions, it is reasonable to assume thatnetworking with other professionals in the same profession can shape participants' cognition, attitudes, andbehaviors regardless of the specific profession (DiMaggio &Powell, 1989). Considering the fact that ICMA andits branches have been advocating for community building and citizen participation for years, we propose thatcity managers' professional networking would increase city managers' willingness to promote citizenparticipation in the budget process. H2: City manager's professional networking is positively associated with thecity's adoption of citizen involvement in the budget process. Experience is an important indicator ofprofessionalism because it translates into immediate outcomes of work-related knowledge, skills, attitudes, andemotions (Alkadry, 2003; Tesluk &Jacobs, 1998). In political science, the length of time that state legislatorsserve in the legislatures is treated as a major aspect of their professionalism and is assumed to impactlegislative behaviors (Fiorina, 1994; 1999). In other fields, on-the-job experience has been widely used topredict individuals' behavior in the work place (e.g., Quinones, Ford, &Teachout, 1995; Ruth-Sahd &Hendy,2005; Tuggle &Sneed, 1998). According to the bureaucratic indifference model and the technocratic expertmodel, city managers' professional experience may be negatively associated with their inclination to adoptparticipatory budgeting. Admittedly, as the citizen leadership model suggests, appointed managers may start tohave more experience of participatory governance and management because community building andparticipation has increasingly become a professional norm for management professionals in local government.Nevertheless, traditional governance styles persist in many jurisdictions, and few studies have paid attention tohow city managers' professional experience influences citizen involvement. We thus tentatively follow thetechnocratic expert model and bureaucratic indifference model, and hypothesize: H3: City manager'sprofessional experience is negatively associated with the city's adoption of citizen involvement in the budgetprocess. City managers' inability or ability is affected by their institutional authority and political environment.Institutional authority or autonomy refers to the legitimate power delegation that a professional exercises withinthe organization. One of the main claims made in the medical literature is that physicians have a strongprofessional identity and value their professional autonomy. Thus physicians will try to preserve their autonomyand professional control and oppose any reform that threatens their autonomy (Gross, Tabenkin, &Brammli-Greenberg, 2007), a dynamic that is consistent with the technocratic expert and the bureaucratic indifferencemodels. Based on this dynamic, there would be a negative relationship between citizen participation and citymanagers' institutional authority. Again, realizing that such a statement conflicts with the modern perception of

08 March 2013 Page 66 of 145 ProQuest

city managers' leadership role in community development, we cautiously hypothesize: H4: City manager'sinstitutional authority is positively associated with the city's adoption of citizen involvement in the budgetprocess. City managers' institutional authority is delegated by elected officials and their ability in the position isgreatly influenced by the local political environment-whether the local politics is healthy and easy to deal with,and whether there is political stability. With unhealthy and unstable politics, city managers are likely to bedistracted from substantive managerial issues, become risk-averse, avoid entrepreneurial actions, and notexercise their autonomy (O'Toole &Meier, 2003; Thompson, 1967; Yang &Pandey, 2008). As Yang (2008)shows, public managers are less likely to involve stakeholders, including citizens, in measuring publicorganizational performance if public managers work in a hostile political environment. Similarly, we submit that ahealthy and stable political environment should positively impact city managers' inclination to involve citizens inthe budget process. H5: Healthy and easy politics is positively associated with the city's adoption of citizeninvolvement in the budget process. H6: Political stability is positively associated with the city's adoption ofcitizen involvement in the budget process. As mentioned above, evidence shows that city managers' personalattitude toward citizen involvement is an important predictor of the adoption of citizen participation mechanisms(Yang &Callahan, 2007). Since a positive attitude toward citizen involvement reflects strong felt-accountabilitytoward citizens, this attitude-adoption link is consistent with Wang's (2001) finding that there is a positiveassociation between city managers' willingness to be accountable and the adoption of citizen participation. Wethus hypothesize: H7: City managers' willingness to represent the community they serve is positively associatedwith the city's adoption of citizen involvement in the budget process. We include two groups of control variables.The first group is about city government. Following Ebdon (2000), we consider cities' representative structureand government size, despite the fact that Ebdon (2000) does not find strong evidence to support the effect ofrepresentative structures. We also consider whether the mayor is fulltime because a full-time position can betaken as an indicator of professionalism; full-time mayors have more policy knowledge and skills than part-timemayors. This logic has been used, for example, in measuring the professionalism of state legislators (Fiorina,1994). In addition, we control for racial representation on city council as a diverse council is more likely toinvolve different groups of citizens to participate. Ebdon (2000) uses racial diversity of city population as apredictor but does not find empirical support. We contend that racial representation on the council is a betterpredictor than population diversity because it is council members who possess legislative power and determinecity policies. The second group of control variables consists of cities' demographic factors: population,population growth, poverty, and education. While Ebdon and Franklin (2006) assert that larger population sizeslead to more support for participatory budgeting, the relationship between population and participation isdebatable in the literature, with some scholars contending that larger communities are associated with greaterparticipation while others argue for the contrary (Kelleher &Lowery, 2004; Yang &Callahan, 2005). The influenceof population growth has rarely been considered in previous studies on citizen participation. A rapid growth ofpopulation may lead to a resources shortage in a community. In the meantime, growing communities are morelikely to confront newly-emerged interests and conflicts; they thus have to pay more attention to parochialdevelopment issues than do other communities, which further constrains governments' ability to involve citizensin the budget process. We also control for socio-economic status of cities, as more wealthy and educatedcitizens tend to demand more participation opportunities (Weber, 2000), while low-income and less educatedresidents usually have less desire to participate because of their work and family priorities (Irvin &Stansbury,2004). METHOD Sample The primary data source is from a mail survey that was designed for a broaderproject. Two different surveys with overlapping questions were sent to Florida mayors and city managersrespectively from October 2006 through February 2007. The mayor survey focused on questions of institutions,budget process, and mayors' individual-level information. The city manager survey included questions aboutmanagers' interactions with the council, economic development policies, and managers' individual-levelinformation. Florida is an excellent study site because of the great political, economic, and demographic

08 March 2013 Page 67 of 145 ProQuest

variation among its cities. Among Florida's 404 cities, 276 (68 percent) have a city manager, city administrator,or chief administrative officer position so the questionnaires were sent to the mayors and the managers of those276 cities. Responses were received from 203 (74%) of the city managers and 200 (73%) of the mayors. .Thenumber of cities with both a mayor and manager who responded is 151 (55%). Due to some missing data, ourworking sample for this study consists of 141 cities (51%). The sample is relatively representative. For example,among 276 targeted cities, 36% are small cities with a population less than 6,000, 39% are medium cities with apopulation between 6,000 and 30,000, and 25% are large cities with a population of more than 30,000. In ourworking sample, the three types of cities make up 28%, 42%, and 30%, respectively. The sample containsslightly more medium and large cities but fewer small cities. This is probably because small cities have fewerstaff members to assist their mayors and city managers, who would then be less likely to respond. Weconducted t tests to compare the sample cities with the targeted cities on variables such as population size,median income, education, and ethnicity; no statistically significant difference was found. MeasurementDependent Variables The level of citizen participation in the budget process was measured by respondents'evaluation of two statements in the mayor survey: (1) the council considers formal recommendations on theproposed budget from citizen groups or committees; and (2) the council coordinates with local media to highlightthe community input process. Both statements had a 7-point scale (1=never, 7=always). Adapted from the 1996ICMA survey on roles and relationships of local government officials (see Ebdon, 2000), the two statementscaptured the levels of formal citizen participation and less formal citizen participation in budgeting at the stage ofbudget consideration (as opposed to budget preparation). Our survey did not include the other three items thatappeared in the ICMA survey because in our pre-test, the responses to them were highly skewed with verysmall data variation. For example, "making the proposed budget document or summary available to the publicprior to adoption" was almost universally used by local governments (35 out of 37 or 95%). We first treatresponses to the two questions as two separate dependent variables, "consideration of formalrecommendations" and "coordination with media for input"; then we add up the values of these two questionsand create a third dependent variable, the "general involvement." The first two dependent variables aremeasured at the ordinal level and their distributions are presented in Table 1. The third dependent variable hasquasi-continuous values ranging from 2 to 14, with the mean at 9.4 and standard deviation at 3.3. IndependentVariables Regarding the professional factors, professional education was measured by a dichotomous item: acity manager holding the MPA degree was coded 1, and 0 means no MPA degree. Among the respondents, 52(37%) have the MPA degree and 89 (63%) do not. Professional networking was measured by the level ofrespondents' involvement in the activities of ICMA and Florida City and County Management Association(FCCMA). We asked how often they attended meetings and other activities organized by ICMA and FCCMArespectively, using a 4-point scale (1=never, 4=very often). We then constructed an index by averaging thevalues of the two items (M=2.2; SD=0.9). Professional experience was measured by the number of years amanager has served as a management professional in local governments. The score for this variable rangedfrom 1 to 48 years (M=20; SD=10). As to city managers' ability, we used the number of department heads that acity manager can directly appoint or remove to measure managers' institutional authority. Among the 141 cities,the value of this variable ranges from 0 to 7 (M=4.9; SD=1.7). Healthy and easy political environment ismeasured by the statement in the manager survey: "The local politics is easy to deal with," with 1 being"strongly disagree" and 7 being "strongly agree" (M=3.8, SD=1.6). Political stability was measured by mayors'response to the statement "The council sets long-term goals and provides direction for the city manager," with 1representing "never" and 7 representing "always" (M=5.9; SD=1.5). City managers' attitudes toward citizenparticipation was measured by city managers' evaluation of the statement "I gather community input and usethat information to determine community needs during budget preparation," with a 7-point scale (1=never,7=always). The responses range from 1 to 7 (M=4.4; SD=1.7). Control Variables As an indicator ofrepresentative structure of city government, elected mayor is a dichotomous variable: 0 being "directly elected

08 March 2013 Page 68 of 145 ProQuest

by citizens" while 1 being "selected from the council". In the sample, 31% of the cities directly elect the mayor,while 69% select the mayor among council members. Similarly, full-time mayor is also a dichotomous variablewith 1 being "full-time" and 0 being "part-time". 95% of the cities have a part-time mayor while only 5% have afulltime mayor. At-large election of council is measured by the percentage of at-large election seats. Racialrepresentation on city council is measured by two terms: percentage of Black members and the percentage ofHispanic members on the council. Wang (2001) uses the number of full-time government employees, notadjusted by population, as the measure of government size. Since government size varies along with citypopulation, the number of government employees per se may not accurately capture governments' resourcesand capacities. The simple use of the number of government employees may also cause a collinearity problemwith city population in the model. We assume a curvilinear relationship between the number of governmentemployees and the population size, so we adjusted government size by the natural logarithmic term ofpopulation and called it relative government size. Population size was measured by the city's population in2006. Again population size may have a curvilinear relationship with the dependent variables, so we used itsnatural logarithmic term in the model. Population growth was calculated in terms of the growth rate from 1999 to2006. The socio-economic status was measured by two items: the percentage of households below the povertyline and the percentage of citizens holding high school diplomas and above. The percentage of householdsbelow the poverty line may also be seen as an indicator of economic stress faced by city government. Thedemographic data were collected from city-data.com and reflected the 2000 census results. Test ProcedureSince the first two dependent variables are ordinal, we used ordered logistic MLE models to estimate the extentto which a city government decides to consider formal recommendations from citizen groups or committees inthe budget process (Model 1) and to coordinate with local media to highlight the community input processrespectively (Model 2).1 With the "general involvement," which has quasi-continuous values due to theaggregation, the third model uses OLS regression. Prior to running the models, a correlation matrix wasproduced (not presented in this article) and no multicollinearity threat seemed to exist. According to Hamilton(2006), a better assessment of multicollinearity is to look at the variance inflation factor (VIF). The highest VIFvalue in our models was 4.6 (for the poverty variable) and most of the other values were below 2.0, suggestingthat multicollinearity is not an issue in this study. FINDINGS AND DISCUSSIONS Table 2 presents the resultswith the coefficients and standard errors of the three models, while Table 3 reports the odds ratios of the twoordered logistic models. The results show that the two MLE models are statistically significant at the .001 levelbased on the likelihood ratio chi-square;2 and they have Count R2 40% and 34% respectively.3 The OLS modelis significant at the .001 level regarding the F value as well; its adjusted R2 for the OLS model is 22%. Theseresults suggest that the three models generally fit the data well and have acceptable explanation power. Ourdiscussion will focus first on Model 1 and then compare Model 1 with the two other models. All explanatoryvariables in Model 1 are significant at the .05 level and have the relationships with the dependent variable ashypothesized. First, professional education, or holding the MPA degree, is found to have a positive associationwith seeking formal recommendation from citizen groups in the budget process, with an odds ratio of 2.4 Theresult is likely to suggest that city managers with MPA experience are more equipped with knowledge and skillsnecessary for citizen participation, and thus more confident to handle citizen involvement in the budget process.It is also likely that the socialization process in MPA programs may shape managers' cognition, attitudes, anddecision making. Given that public administration scholars have been advocating for more citizen participationand democratic citizenship, observing a positive relationship between the MPA degree and greater motivation toinvolve citizens is not surprising. Professional networking is significantly and positively associated with formalcitizen involvement in the budget process with an odds ratio of 1.5.5 Professional networking is part ofmanagerial networking from which county/city managers can get access to training, best practices,opportunities, and socialization. Meier and O'Toole (2003) find that managerial networking leads to betterorganizational performance. Our study further confirms the role of ICMA and its local branches in promoting

08 March 2013 Page 69 of 145 ProQuest

modern professional values such as citizen participation, transparency, and democratic governance. The resultsuggests that professional networking does not reinforce the sense that budget decisions are territories only forprofessionals; rather, it helps managers appreciate the positive role of participatory budgeting. While publicadministration associations such as ICMA were perceived as merely voluntary with little or no influence over theentrance, promotion, training standards, and ethical performance of individual city managers (Stillman, 1977),their positive effect on participatory budgeting seems to be substantive at present. Professional experience isnegatively associated with cities' use of formal citizen recommendations in the budget process. The result canbe explained by the bureaucratic indifference model which treats bureaucratic experience as a barrier for citizenparticipation because it produces a bureaucratic personality that is antithetical to individual responsiveness,participation, and flexibility (Hummel, 1994). This explanation is not necessarily contradictory with the citizenleadership model which observes that recent city managers may embrace community building and citizenengagement as a new professional norm (Nalbandian, 1991; 1999). It is reasonable to expect a high correlationbetween managers' age and their professional experience, so the impact of professional experience may reflectthe influence of age. In other words, this finding may imply that younger managers are more likely to supportcitizen involvement than do senior managers because younger managers are more likely to be shaped by thenew professional norm of citizen engagement. Nevertheless, while the data of managers' age are not availablefor this article, we expect future study to further explore the relationship between citizen involvement andmanagers' age. Managerial authority is significantly and negatively associated with the adoption of citizenparticipation in budgeting (odds ratio of 0.8), as predicted in the bureaucratic indifference and the technocraticexpert models. When a city manager has greater institutional power, s/he is more likely to rely on formaladministrative channels for decision making and less likely to open the administrative process to citizeninvolvement. A manager with great managerial authority is more likely to rely on professional knowledge,subordinate support, and hierarchical control. In contrast, a front-line professional administrator, given his/herconstant interaction with citizens/clients, may be more likely to appreciate local knowledge and embrace citizenparticipation. This is particularly true for administrative functions that are complex and highly sophisticated suchas budgeting. However, considering the measure of this variable-the number of department heads that the citymanager can appoint and remove-the result may alternatively suggest that when the city manager has a greaterspan of control, s/he may have less motivation or time to involve citizens in the budget process. Easy and stablelocal politics encourages managers to consider formal citizen participation in government budgeting, as bothvariables (easy politics and stable politics) are significantly and positively associated with the dependentvariable (odds ratio of 1.3 and 1.4 respectively). In such an environment, citizen involvement is less likely to becontroversial because citizens' preferences and expressions are likely to be stable, so city managers may feelfewer risks in involving citizens. In addition, managers facing easy and stable politics can concentrate onmanagerial and policy issues without fears about job security and dirty politics, so they are more likely to beentrepreneurial and to be responsive to citizens. This finding is in line with Yang's (2008) observation that asupportive political environment enables organizations to involve their stakeholders in performancemeasurement. While the public management literature has emphasized the importance of a stable environmentfor organizational performance (O'Toole &Meier, 2003; Thompson, 1967; Yang &Pandey, 2008), our resulthighlights the positive effect of such an environment in fostering open government, transparency, andparticipation. As hypothesized, managers' attitude toward citizen input is found to be significantly and positivelyassociated with local governments' adoption of formal citizen recommendations (odds ratio of 1.2). Thisconfirms Yang and Callahan's (2007) observation that managerial attitude is a significant predictor of citizeninvolvement efforts by local governments (see also Yang, 2006). More generally, the finding is consistent withthe recent public management and bureaucratic politics literature, which demonstrates that bureaucratic valuesare far more important than external political factors in explaining bureaucratic decisions, outputs, andoutcomes (Meier &O'Toole, 2006). Public managers do not just passively respond to external political

08 March 2013 Page 70 of 145 ProQuest

pressures; rather, they make judgments about what is best for the community and strive to solve communityproblems. Regarding the control variables, elected mayor is not statistically significant in the model, but full-timemayor is significantly and positively associated with the dependent variable. This finding is consistent with theresult regarding the significant and negative influence of managers' institutional authority because the full-timemayoral position means that greater power is possessed by the political leader and less authority is delegatedto the manager. Taken together, the results indicate that a more politically representative government is morelikely to adopt formal citizen participation in the budget process, while a more managerially oriented governmentis less likely to do so. The results are in line with Yang and Callahan's (2007) conclusion that strong electedofficials may be strong advocates for more citizen participation in the administrative process and decisionmaking. Relative government size is not statistically significant in the model. While Wang (2001) finds that largergovernment sizes are associated with higher levels of participation, our result suggests that, after adjustment forpopulation size, larger government sizes are not associated with governments' adoption of formal citizenrecommendations. Larger relative sizes may mean more resources and stronger capacity, which support morecitizen participation, but they may also mean more red tape and stronger hierarchical control, which preventcitizen participation. For council selection method, Black council member percentage, and Hispanic councilmember percentage, none is found to be statistically significant in the model. The result regarding councilselection method is consistent with Ebdon's (2000) finding. Although Ebdon hypothesizes that cities with morerepresentative structures- elected mayor and council members elected by district rather than atlarge- may bemore inclined to involve citizens in the budget process, she finds no evidence supporting the claim. The lack ofeffect from minority council representation suggests that ethnicity may not be a good predictor of councilmembers' attitude toward formal citizen participation. Certainly, this warrants more systematic examination byfuture studies. Regarding the influence of demographic factors, population size and population growth arestatistically significant in Model 1 at the .1 level. Formal citizen recommendations are more likely to beconsidered in larger cities than in smaller cities, a result that is consistent with Ebdon and Franklin's (2006)contention. It also echoes Yang and Callahan's (2007) conclusion that large population categories areassociated with higher levels of involvement mechanisms. Two theories can be offered to account for thisrelationship: a larger population means more conflicts that lead to greater demands for participation, or it meansstronger capacity and more resources supporting participation. The significant and negative impact ofpopulation growth may suggest that the governments' abilities and resources are greatly constrained by therapid growth of the communities, which leaves little room for them to adopt community-wide citizen participationin the budget process. In Model 2, the dependent variable, the extent to which city governments coordinate withlocal media to highlight the community input process, is a less formal and less substantive approach of citizeninvolvement than the dependent variable in Model 1. It is not surprising that the results from these two modelsare not identical. In Model 2, only 3 out of the 7 explanatory variables-managers' institutional authority, stablepolitics, and managers' willingness-are statistically significant at the .1 level, while the other 4 variables do notmake a difference in shaping cities' inclination to highlight the citizen input process through local media. Inparticular, none of the three professional factors is statistically significant in the model. It might suggest thatprofessional managers are more likely to adopt a formal approach of citizen involvement rather than informalones. From the perspective of the citizen leadership model, professional education and networking enrichmanagers with professional norms supporting citizen engagement. Internalized norms and values may leadmanagers to support substantial involvement efforts such as considering formal citizen recommendations, buttheir effect on less substantive involvement activities may be tenuous. Highlighting citizen participationprocesses through local media may be conducted for procedural and public-relations purposes. Such lesssubstantive efforts may not be in the decision domain of senior managers, which explains why professionalexperience is not significant in the model. In Model 2, elected mayor rather than full-time mayor is significantlyand positively associated with citizen participation that is less formal and less substantive. This suggests that

08 March 2013 Page 71 of 145 ProQuest

elected mayors, regardless of the time they spent on the job, are likely to appeal to constituencies by promotingcitizen participation and transparency through local media. Population size is significant and positive probablybecause large population sizes indicate stronger capacity and more resources that are necessary to launch apublic relations campaign. It is also likely that when population size is large, it is necessary to use the localmedia to make sure the information about the participation process can reach most residents. In Model 3, thedependent variable is created by adding the dependent variables in Models 1 and 2. The aggregation issomewhat arbitrary since the two original variables were quite distinctive, but not exhaustive-that is, there areother activities or dimensions of participatory budgeting. Hence, the inclusion of Model 3 is only for comparisonpurposes. The results show several variables remain statistically significant, such as stable politics, managers'attitude, and population size. This makes sense because perceived political environment directly affectsmanagers' calculations on stakeholder involvement (Yang, 2008), and managerial attitude toward citizenparticipation is probably the most important predictor of citizen involvement decisions (Yang, 2006; Yang&Callahan, 2007). External political environment and internal bureaucratic values are two of the most significantfactors in explaining bureaucratic decisions (Meier &O'Toole, 2006). CONCLUSION While a common belief inthe participatory budgeting literature is that council-manager governments are more supportive of citizenparticipation in the budget process than other forms of government, we argue that this belief does not squarewith the reality that many cities have adopted mixed forms of government. In an era when adaptive or mixedforms of government are increasingly popular (Frederickson, Johnson, &Wood, 2004), the form of governmentalone cannot capture the institutional complexities of local governments, and it is more appropriate to directlymeasure the dimensions of a city manager's professional status and background in order to better understandhow city managers relate to citizen participation in the budget process. In the meantime, the current literature isambiguous on the relationship between professional administration and citizen participation as it providescontradictory arguments. This article discusses the competing theoretical perspectives and examines how theadoption of citizen participation in the budget process is associated with city managers' professional factors,institutional environment, and willingness to represent citizens. The results suggest that this is a usefulapproach to study the adoption of participatory budgeting. Specifically, the results strongly support Nalbandian's(1991; 1999) observation that community building and citizen engagement have become professional norms forlocal government managers, as professional education and networking are positively associated with theconsideration of formal citizen recommendations in the budget process. Even when the negative effect ofprofessional experience seems to support the bureaucratic indifference and technocratic expert models, it maywell just reflect the fact that new professional norms have emerged only since the mid-1990s and it is youngermanagers who are more likely to be deeply influenced. Furthermore, our results are consistent with the publicmanagement literature in finding that external political environment and managerial attitude toward citizenparticipation are important factors in accounting for local governments' decisions in the area of citizeninvolvement. The study has limitations. The sample is from Florida, so caution must be taken in generalizing theresults beyond the sample. Moreover, the results of this study depend on the way variables are measured. Thedependent variables measure the extent to which local governments use participation mechanisms inbudgeting, but does not assess whether citizen input really makes a difference in budgetary outcomes. Whetherlocal governments allow citizen participation in the budget process and whether such participation really makesa difference in budgetary outcomes are different questions, for which the effects of professional administrationmay vary. Future studies may choose different samples and different ways of measurement to verify our results.In addition, the actual level of citizen participation is determined by both the extent to which governmentsprovide involvement opportunities and the extent to which citizens are willing and competent to participate.Studies occasionally used the demand-side (citizen-side) mechanisms to develop hypotheses and interpretresults, but the focus of this article is on the supply- (government) side. Future inquiries may pay more attentionto citizen-side factors such as the accessibility of issues critically important to citizens in the budgetary process.

08 March 2013 Page 72 of 145 ProQuest

Despite the limitations, our results are informative considering the purpose of this exploratory study, which is toshow that different dimensions of professional administration may have different effects on citizen participation.Footnote NOTES 1. We also operated ordered probit MLE models. The results are very similar to those in theordered logistic MLE models that we present in this article. 2. Likelihood ratio chi-square evaluates the nullhypothesis that all coefficients in the model, except the constant, equal zero. 3. Count R2 measures thedifference between the predicted and actual outcomes on the dependent variable. 4. It indicates that in the citieswith managers holding MPA degrees, the odds ratio of being in a higher level to a lower level of citizeninvolvement (in a 1-7 scale) is 100 percent higher than the odds ratio in the cities where the managers do nothold MPA degrees, controlling other factors constant. 5. The odds ratio of being in a higher level to a lower levelof citizen involvement (in a 1-7 scale) will increase 50 percent as the index of professional networking increasesone unit, holding other factors constant. References REFERENCES Alkadry, M. G. (2003). "DeliberativeDiscourse Between Citizens And Administrators: If Citizens Talk, Will Administrators Listen?" Administration&Society, 35 (2): 184-209. Berman, E., &West, J. (1995). "Municipal Commitment to Total Quality Management:A Survey of Recent Progress." Public Administration Review, 55 (1): 57-66. Bland, R. L., &Rubin, I. (1997).Budgeting: A Guide for Local Government. Washington, DC: ICMA. Cole, R. (1974). Citizen Participation andthe Urban Process. Lexington, MA: Lexington Books. Del Gaudio, A. C., Stein, L. S., Ansley, M. Y., &Carpenter,P. J. (1975). "Community Mental Health Ideology as a Function of Professional Affiliation and Social ClassBackground." Journal of Community Psychology, 3 (4): 341-345. DiMaggio, P., &owell, W. (1989). The NewInstitutions in Organizational Analysis. Chicago, IL: University of Chicago Press. DeSario, J., &Langton, S.(1984). "Citizen Participation and Technocracy." Review of Policy Research, 3 (2): 223-233. Ebdon, C. (2002)."Beyond the Public Hearing: Citizen Participation in the Local Government Budget Process." Journal of PublicBudgeting, Accounting and Financial Management, 14 (2): 273- 94. Ebdon, C. (2000). "The Relationshipbetween Citizen Involvement in the Budget Process and City Structure and Culture." Public Productivity andManagement Review, 23 (3): 383-393. Ebdon, C., &Franklin, A. (2006). "Citizen Participation in BudgetingTheory. Public Administration Review, 66 (3): 437-47. Feiock, R. (2003). "Politics, Governance, and theComplexity of Local Land Use Regulation." Urban Studies, 41 (2): 363-77. Fiorina, M. P. (1999). "FurtherEvidence of the Partisan Consequences of Legislative Professionalism." American Journal of Political Science,43 (3): 974-977. Fiorina, M. P. (1994). "Divided Government in the American States: A Byproduct of LegislativeProfessionalism?" The American Political Science Review, 88 (2): 304-316. Fischer, F. (2000). Citizens,Experts, and the Environment. Durkham, NC: Duke University Press. Franklin, A., &Ebdon, C. (2005). "Are WeAll Touching the Same Camel? Exploring a Model of Participation in Budgeting." American Review of PublicAdministration, 35 (2): 168-85. Frederickson, G., Johnson, G. A., &Wood, C. (2004). The Adapted City. Armonk,NY: M.E. Sharpe. Greenstone, D. &Peterson, P. (1971). "Reformers, Machines, and the War on Poverty." In B.T. Downs (Ed.), Cities and Suburbs: Selected Readings in Local Politics and Public Policy (pp. 377- 99).Belmont, CA: Wadsworth. Gross, R., Tabenkin, H., &Brammli-Greenberg, S. (2007). "Factors Affecting PrimaryCare Physicians' Perceptions of Health System Reform in Israel: Professional Autonomy versus OrganizationalAffiliation." Social Science &Medicine, 64 (7): 1450-1462. Hamilton, L. C. (2006). Statistics with STATA.Thomson Press. Hummel, R. (1994). The Bureaucratic Experience: A Critique of Life in the ModernOrganization (4th ed.). New York: St. Martin's. Irvin, R. A., &Stansbury, J. (2004). "Citizen Participation inDecision Making: Is It Worth the Effort?" Public Administration Review, 64 (1): 55-65. Kelleher, C., &Lowery, D.(2004). "Political Participation and Metropolitan Institutional Context." Urban Affairs Review, 39 (6): 720-57.Kweit, M. G., &Kweit, R. W. (1981). Implementing Citizen Participation in a Bureaucratic State. Pittsburgh, PA:University of Pittsburgh Press. Lubell, M., Feiock, R., &Ramirez, E. (2005). "Political Institutions andConservation by Local Governments." Urban Affairs Review, 40 (6): 706-29. Marlowe, J., &Portillo, S. (2006)."Citizen Engagement in Local Budgeting: Does Diversity Pay Dividends?" Public Performance &ManagementReview, 30 (2): 179-202. Maser, S. M. (1998). "Constitutions as Relational Contracts: Explaining Procedural

08 March 2013 Page 73 of 145 ProQuest

Safeguards in Municipal Charters." Journal of Public Administration Research and Theory, 8: 527-64. Meier, K.J., &O'Toole, Jr., L. J. (2006). "Political Control versus Bureaucratic Values: Reframing the Debate." PublicAdministration Review, 66 (2): 177-192. Meier, K. J., &O'Toole, Jr., L. J. (2003). "Public Management andEducational Performance: The Impact of Managerial Networking." Public Administration Review, 63: 675-685.Nalbandian, J. (1999). "Facilitating Community, Enabling Democracy: New Roles for Local GovernmentManagers." Public Administration Review, 59 (3): 187-197. Nalbandian, J. (1991). Professionalism in LocalGovernment: Transformations in the Roles, Responsibilities, and Values of City Managers. San Francisco, CA:Jossey-Bass. O'Toole, D. E., &Marshall, J. (1988). "Citizen Participation through Budgeting." The Bureaucrat,17 (2): 51-55. O'Toole, Jr. L. J., &Meier, K. J. (2003). "Plus ça Change: Public Management, Personnel Stability,and Organizational Performance." Journal of Public Administration Research and Theory, 13 (1): 43-64 Poister,T., &Streib, G. (1989). "Management Tools in Municipal Government: Trends over the Past Decade." PublicAdministration Review, 52: 240-48. Quinones, M., Ford, K., &Teachout, M. (1995). "The Relationship betweenWork Experience and Job Performance: A Conceptual and Meta-Analytical Review." Personnel Psychology, 48:887- 905. Roback, H. B, Purdon, S. E., Ochoa, E. &Bloch, F. S. (1993). "Effects of Professional Affiliation onGroup Therapists' Confidentiality Attitudes and Behaviors." The Bulletin of the American Academy of Psychiatryand the Law, 21 (2): 147-153. Ruth-Sahd, L., &Hendy, H. (2005). "Predictors of Novice Nurses' Use of Intuitionto Guide Patient Care Decisions." Journal of Nurse Education, 44 (10): 450-458. Simonsen, W., &Robbins, M.D. 2000. Citizen Participation in Resource Allocation. Westview Press. Stillman, R. J., 1977. The city manager:professional helping hand, or political hired hand? Public Administration Review, 37: 659-670. Streib, G. (1992)."Professional Skill and Support for Democratic Principles." Administration &Society, 24 (1): 22-40. Tesluk, P. E.,&Jacobs, R. R. (1998). "Toward an Integrated Model of Work Experience." Personnel Psychology, 51 (2): 321-55. Thompson, J. D. (1967). Organizations in Action: Social Science Bases of Administrative Theory. New York:McGraw-Hill. Tuggle, C. A., &Sneed, D. (1998). "Faculty in Professional Programs: The Mix of Experience andDegrees." Journalism and Mass Communication Educator, 53 (1):14-22. Wang, X. (2001). "Assessing PublicParticipation in U.S. Cities." Public Performance &Management Review, 24 (4): 322-36. Weber, E. P. (2000). "ANew Vanguard for the Environment: Grass- Roots Ecosystem Management as a New EnvironmentalMovement." Society and Resources, 13 (3): 237-59. Worldwatch Institute (2007). State of the World 2007: OurUrban Future. New York: W. W. Norton. Yang, K. (2008). "Examining Perceived Honest Performance ReportingBy Public Organizations: Bureaucratic Politics And Organizational Practice." Journal of Public AdministrationResearch and Theory, 19 (1): 81-105. Yang, K. (2006). "Trust and Citizen Involvement Decisions: Trust inCitizens, Trust in Institutions, and Propensity to Trust?" Administration &Society, 38 (5): 573-595. Yang, K.,&Callahan, K. (2007). "Citizen Involvement Efforts and Bureaucratic Responsiveness: Participatory Values,Stakeholder Pressures, and Administrative Practicality." Public Administration Review, 67 (2): 249-64. Yang, K.,&Callahan, K. (2005). "Assessing Citizen Involvement Efforts by Local Governments." Public Performance andManagement Review, 29 (2): 191-216. Yang, K., &Pandey, S. (2008). "How Do Perceived Political Environmentand Administrative Reform Affect Employee Commitment." Journal of Public Administration Research andTheory, 19 (1): 335-60. AuthorAffiliation Yahong Zhang and Kaifeng Yang* * Yahong Zhang, Ph.D., is anAssistant Professor, School of Public Affairs and Administration, Rutgers University-Newark. Her researchinterests include local government, public policy, and gender issues. Kaifeng Yang, Ph.D., is an AssociateProfessor, Askew School of Public Administration and Policy, Florida State University. His research interestsinclude citizen participation, public management, and organizational theory. Subject: Studies; City managers; Budgeting; Professional responsibilities; Citizen participation Location: United States--US Classification: 9190: United States, 9130: Experiment/theoretical treatment, 9550: Public sector

08 March 2013 Page 74 of 145 ProQuest

Publication title: Journal of Public Budgeting, Accounting&Financial Management Volume: 21 Issue: 2 Pages: 289-317 Number of pages: 29 Publication year: 2009 Publication date: Summer 2009 Year: 2009 Publisher: PrAcademics Press, Florida Atlantic University Place of publication: Boca Raton Country of publication: United States Journal subject: Business And Economics--Banking And Finance ISSN: 10963367 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Tables;References ProQuest document ID: 500961686 Document URL: http://search.proquest.com/docview/500961686?accountid=86413 Copyright: Copyright PrAcademics Press, Florida Atlantic University Summer 2009 Last updated: 2011-06-23 Database: Accounting&Tax

_______________________________________________________________ Document 7 of 10 Auditor Disaffiliation Program in China and Auditor Independence Author: Gul, Ferdinand A; Sami, Heibatollah; Zhou, Haiyan Publication info: Auditing 28. 1 (May 2009): 29-51.ProQuest document link Abstract: This study examines whether the disaffiliation program introduced by the Chinese governmentimproved auditor independence and whether auditor quality affects this relationship. Auditor independence ismeasured in terms of the likelihood of receiving a qualified report and the level of earnings management(measured by noncore operating income). The results show that the likelihood of receiving qualified auditopinions for listed companies significantly increased, and noncore operating earnings significantly decreased,after auditors were disaffiliated. However, companies audited by auditors without any affiliation also showed anincrease in the likelihood of receiving qualified opinions and a decrease in noncore operating earnings, possiblybecause of the increased surveillance by the regulatory bodies that accompanied the act of disaffiliation. The

08 March 2013 Page 75 of 145 ProQuest

results also show that the association between the disaffiliation program and the likelihood of receiving qualifiedaudit opinions is stronger for small auditors than for large auditors, possibly because of the initial lower auditquality of small auditors. Auditor size, however, did not significantly affect the association between thedisaffiliation program and noncore operating earnings. [PUBLICATION ABSTRACT] Full Text: Headnote SUMMARY: This study examines whether the disaffiliation program introduced by theChinese government improved auditor independence and whether auditor quality affects this relationship.Auditor independence is measured in terms of the likelihood of receiving a qualified report and the level ofearnings management (measured by noncore operating income). The results show that the likelihood ofreceiving qualified audit opinions for listed companies significantly increased, and noncore operating earningssignificantly decreased, after auditors were disaffiliated. However, companies audited by auditors without anyaffiliation also showed an increase in the likelihood of receiving qualified opinions and a decrease in noncoreoperating earnings, possibly because of the increased surveillance by the regulatory bodies that accompaniedthe act of disaffiliation. The results also show that the association between the disaffiliation program and thelikelihood of receiving qualified audit opinions is stronger for small auditors than for large auditors, possiblybecause of the initial lower audit quality of small auditors. Auditor size, however, did not significantly affect theassociation between the disaffiliation program and noncore operating earnings. Keywords: auditorindependence; audit opinion; audit quality; earnings management; noncore operating income; emerging market.Data Availability: Data used in this study are available from public sources. INTRODUCTION Prior to theprivatization efforts of the Chinese government, most of the accounting and auditing firms in China were insome way affiliated with government agencies, government-sponsored bodies, or universities and researchinstitutions. As part of the privatization policy, the Chinese government undertook various policy changes aimedat improving auditor independence. One of the changes was the act of auditor disaffiliation coupled with relatedregulations during the period 1997-1998 (the disaffiliation program).1 The program was prompted, in part, byunprecedented corporate scandals such as the Yuanye scandal (1991), the Great Wall Fund Raising scandal(1992), the Wangfujing scandal (1993), the Jianfeng scandal (1993), and the Nantong Machine Tool event(1995), all of which were audited by affiliated auditors.2 In these cases, managers issued fraudulent financialstatements or falsified capital investment accounts to mislead investors, while affiliated CPA firms either failedto detect or detected but did not report such managerial misconduct. Even more worrisome was the fact thatsome CPAs aided in and abetted the falsification of the accounts (Xiao et al. 2000). These ethical and qualityproblems involving auditors directly created a credibility crisis for the Chinese accounting profession. Accordingto Liu and Zhang (1996), a survey conducted in 1995 showed that more than 60 percent of investors had noconfidence in the audited accounting information. To protect professionals from this credibility crisis, theChinese Institute of Certified Public Accountants (CICPA) and the Ministry of Finance (MOF) launched a seriesof programs, such as nationwide investigation of CPA firms, establishment of the Division of PracticeSupervision in CICPA, as well as the implementation of the Chinese Independent Auditing Standards (CIAS)(Xiao et al. 2000). The CICPA and the MOF also realized that the most important step in the reform of theaccounting profession was to improve auditor independence in fact and in appearance. For this purpose, theCICPA and the MOF required that all CPA firms sever their links with their sponsoring bodies, such asgovernment agencies, government-owned institutes or universities. This act of disaffiliation involved about 110nationwide CPA firms. As part of this disaffiliation program, the CICPA and the MOF also establishedregulations and guidelines designed to improve auditor independence. The objective of this study is toinvestigate whether the act of disaffiliation and/or the disaffiliation program improved auditor independence. Weuse the likelihood that companies would receive qualified audit opinions after the disaffiliation program and thelevel of noncore operating income (our proxy for earnings management) as measures of auditor independence.We also investigate whether auditor quality moderates the relationship between the disaffiliation program andauditor independence. In an earlier study, Yang et al. (2001) reported an increase in the number and

08 March 2013 Page 76 of 145 ProQuest

percentage of qualified audit opinions subsequent to the disaffiliation program in China. They discussedpossible explanations such as the impact of Big 5 (now Big 4) accounting firms, the implementation of newauditing standards, auditor experience and competence, increases in the number of listed companies andbusiness activities, and the auditor disaffiliation program. Yang et al. (2001) concluded that such an increase inqualified audit reports resulted from the disaffiliation program. Unfortunately, their study left several empiricalquestions unanswered. First, did the changes in the firm-specific factors and auditor-specific factors affect auditopinions? That is, did the act of disaffiliation affect audit opinions after controlling for the effects of firm-specificand auditor-specific factors such as the financial condition of the firms and the size of the auditor?3 Second, didthe act of disaffiliation deter or prevent earnings management behavior of managers in publicly heldcompanies? Third, was it the act of disaffiliation alone or me act of disaffiliation combined with other, morestringent regulations and sanctions (the disaffiliation program), which were simultaneously or subsequentlyenacted, that led to improved auditor independence? Finally, did auditor quality play a role in the effectivenessof the disaffiliation program? By attempting to answer these questions, this paper provides a morecomprehensive analysis and in so doing provides a better understanding of the impact of this disaffiliationprogram on auditor independence. The results of our study suggest me following. First, we find that thelikelihood of receiving qualified audit opinions by public companies significantly increased after the disaffiliationprogram. This is true not only for companies previously audited by affiliated auditors, but also for companiespreviously audited by nonaffiliated auditors. We do not find a significant difference between these two groups ofauditors. Second, we find that the increase in the likelihood of issuing qualified audit opinions is greater for smallauditors compared with large auditors. Third, we find that in the post-disaffiliation period, companies have lowernoncore operating income than in the pre-disaffiliation period. As in the earlier case, there is no difference innoncore operating income between companies previously audited by affiliated auditors compared withcompanies previously audited by nonaffiliated auditors. Finally, unlike the case of qualified reports, we find noevidence of the moderating role of auditor size on the earnings management behavior of companies. Overall,these results indicate that it was the disaffiliation program but not the act of disaffiliation by itself that improvedauditor independence, and that auditor quality moderates the relationship between the disaffiliation program andme likelihood of receiving a qualified audit report. The study makes several contributions. First, it providesevidence on whether the Chinese government's disaffiliation program was successful in improving auditorindependence. This information should be of interest to Chinese policy makers. Second, the evidence in thispaper should be useful to Chinese domestic investors since it is asserted that in emerging markets like theChinese markets, accounting information is considered noisy (Fox 1998; Rask et al. 1998) and auditors lackindependence and experience (Xiang 1998; Graham 1996). The evidence that the disaffiliation programimproved auditor independence is likely to provide investors with more confidence in audited financialstatements, ceteris paribus. Third, our findings regarding auditor independence may be of interest tointernational accounting firms and other institutions interested in mis emerging market, especially with Chinarecently joining the World Trade Organization (WTO). Last but not least, our paper adds to the evolvingliterature on audit quality in China (Xiang 1998; Graham 1996; Yang et al. 2001) by showing mat thedisaffiliation program improved auditor independence. The paper proceeds as follows. In the next section, wepresent the institutional background and develop the hypotheses. The third section contains the researchdesign. The data and empirical analyses are provided in the fourth section, and the fifth section summarizes ourfindings and discusses the limitations. INSTITUTIONAL BACKGROUND AND HYPOTHESES History ofAffiliation of CPAs with Government Agencies Accounting and auditing practices in China have experienceddramatic changes over the past two decades. Both the Shanghai Stock Exchange and the Shenzhen StockExchange have developed rapidly, with listed companies increasing from 14 in 1991 to 1,059 in 2000 andannual trading volume increasing from 300 million shares in 1991 to 455.6 billion shares in 2000. This periodalso witnessed important changes in auditing practice, including the transformation of auditors from government

08 March 2013 Page 77 of 145 ProQuest

agencies to regulated independent providers of assurance on accounting information (Xiang 1998). During theperiod 1949-1980, no independent auditing practice existed in the former planned economy, because allenterprises were owned and managed by the state and accounting practice simply focused on reportingcompliance with state economic plans, using a set of accounts specially structured on the basis of sources anduses of funds (Graham 1996). Auditing practice resumed in the 1980s as a result of privatization of state-ownedenterprises and the separation of government from enterprises (Xiang 1998). However, the major providers ofauditing services were state-owned auditing firms or auditing bureaus (Xiao et al. 2000). Before the 1990s, theaccounting firms only provided independent auditing service to a handful of companies with foreign investment.In 1995, auditing and accounting firms were brought under the auspices of the CICPA, and Certified PublicAccounting (CPA) emerged as the only certificate program available to independent accountants.4 Thesponsoring government agencies or state-owned institutions continued to own the CPA firm, manage itspersonnel, and supervise or even directly develop its operating policy (for a detailed discussion, see Yang et al.2001). Effects of Affiliation with Government Agencies The affiliation of the audit firms with the state was themajor impediment to auditor independence (DeFond et al. 2000; Yang et al. 2001). Government agencies coulduse their administrative powers and/or ownership to assign their CPA firms to audit those listed companiesunder their control. According to DeFond et al. (2000), government-affiliated CPA firms audited about 70percent of listed companies. The government affiliation also protected CPA firms from litigation risk (Yang et al.2001; DeFond et al. 2000) because the ultimate liability was borne by the sponsoring agencies. The continuingexpansion of stock markets, the introduction of foreign capital, and the privatization of state-owned enterprisesled to increasing demand for independent auditing services. In particular, there were concerns that earningsmanagement was widespread. The major reasons for earnings management included the need for listedcompanies to meet the target profitability level for an IPO or to avoid being delisted (Jian and Wong 2004) andthe fact that managers' compensation was often based on their firms' financial performance, which is usuallybased on accounting numbers (Yang et al. 2001). In an environment where both auditors and clients hadaffiliation with government agencies, auditors were often subject to substantial pressure from governmentagencies or managers to comply with their requests. Institutional Factors Associated with Disaffiliation Theperiod prior to the disaffiliation program was characterized by a series of financial scandals and reforms. One ofthe earliest attempts to regulate accounting firms was the CPA Act of 1994, which made the CPAs liable toclients and other parties who may suffer a loss due to audit failures. Such liability was further emphasized bythe Chinese High Court in Document No. 56, issued in 1996 (GuI et al. 2003). During this period, investors alsodemonstrated that they were not hesitant to protect their interests through litigation against managers andauditors. According to the CPA Newsletter (1998) quoted by Yang et al. (2001), starting from the first case in1996, litigation against auditors increased to about 400 cases in 1998. Further, according to the same source,the cases of warnings, fines, withdrawal of certificate, and other sanctions increased dramatically, from about100 in 1994 to about 500 in 1997. Finally, in anticipation of China's entering the World Trade Organization(WTO),5 the Chinese government decided to reorganize the CPA firms. The reorganization centered onrequiring Chinese CPA firms to separate themselves from their sponsoring bodies and to reorganize themselvesas partnerships or limited liability companies similar to their international counterparts. In addition to the aboveinstitutional and regulatory changes related to the disaffiliation program, there were other changes in theinstitutional and regulatory environment after 1997 that might have impacted auditor independence including thefrequency of qualified audit opinions. For example, to provide managers more latitude to make judgments, theaccounting system was revised in 1998 to eliminate some of the differences between the InternationalAccounting Standards (IAS) and the Chinese accounting standards (Zheng and Stone 2005). This newaccounting system, which was more complex, might have caused disagreement between auditors andmanagers of listed firms, leading to an increase in the frequency of qualified audit opinions. Further, in 1999, thereturn-on-equity (ROE) threshold was changed from an initial 10 percent for each of three consecutive years to

08 March 2013 Page 78 of 145 ProQuest

6 percent with an average of 10 percent over three consecutive years. This might account for lower earningsmanagement in 1999-2000 compared with 1996-1998. We address these possible explanations in the sectiontitled "Alternative Explanations and Sensitivity Tests." To the extent that we have not been able to do this, weacknowledge the limitation that changes in audit qualification rates and earnings management could be theoutcome of both the disaffiliation program and associated improvements in enforcement and other governanceimprovements during our test period. Effects of Disaffiliation Program As a result of disaffiliation, all firms wererequired to take the form of either a limited liability company or a partnership (GuI et al. 2003; Yang et al. 2001).The CPA firms no longer had any financial affiliations, such as an investment, profit, or payroll relationship, withany government agencies. More importantly, CPA firms had to face the risk of litigation as a result of becomingindependent legal entities subsequent to the disaffiliation program. Whether this act of disaffiliation and thedisaffiliation program has achieved its claimed effect on auditor independence is the empirical issue this studyattempts to resolve. Hypotheses Development We expect the act of auditor disaffiliation together with the seriesof regulations during that period to be associated with more auditor independence. Independent auditors areless likely to yield to client pressures to agree to issue clean audit opinions when qualified opinions areappropriate, or to agree to accounting numbers that are inappropriate. Assuming that the disaffiliation programimproves auditor independence, we expect that the relative frequency of qualified audit opinions will increase,while client earnings management will decrease, subsequent to the implementation of the disaffiliation program.Qualified Audit Opinions Prior studies have examined the effects of policy changes in accounting and auditing inChina. DeFond et al. (2000), for example, investigate the impact of new auditing standards on audit opinions,and their impact on the market concentration. Comparison of a sample of the pre-standard period 1993-1994with the post-standard period 1995-1996 shows a higher frequency of qualified audit opinions in the post-standard period. They also find a decline in the IPO audit market share for large auditors that have greaterpropensity to issue qualified opinions. Thus, they conclude that the higher frequency of qualified audit opinions,reflecting the improved auditor independence, results in a decreasing IPO market share for large auditors inChina.6 In this study, we extend their first research question to examine whether the disaffiliation programimproved auditor independence and whether auditor quality moderates this relationship. Yang et al. (2001)investigate the number and percentage of qualified audit opinions subsequent to the disaffiliation program andobserve an increase in the number of qualified audit opinions for fiscal years 1997 and 1998, concluding thatthe disaffiliation program improved auditor independence. Whether the increase in the number of qualifiedopinions is due to the disaffiliation program or the deterioration in clients' financial status or performance is notclear. Also, because large auditors tend to act more independently and hence moderate the impact of thedisaffiliation program on auditor independence, it is not clear whether the increase in qualified opinions resultedfrom clients' switching auditors from small auditors to large auditors. In addition, the listed companies audited bygovernment-affiliated CPA firms (the group directly affected by the disaffiliation program) were not directlyexamined. Thus, it is not clear whether the increase in qualified opinions comes from affiliated and/ornonaffiliated auditors.7 The gaps and unanswered questions in prior research provide a motivation for a morecomprehensive analysis of the effects of the disaffiliation program on auditor independence. From theperspective of policy makers, auditor affiliation may impair auditor independence, and auditor disaffiliation -coupled with or without regulation - may improve auditor independence. However, from the perspective ofimplementation, some possible competing effects of the disaffiliation program might exist. On one hand, it maybe more difficult for the sponsoring agencies to influence the affiliated auditors after disaffiliation, and thus,auditors may become more independent. On the other hand, after disaffiliation, because the CPA firm is ownedby the CPAs but not government agencies, auditors will have much stronger incentives to compete for clients bycompromising independence. Further, the sponsoring agencies and auditors may maintain close relationshipsto pursue their own interests. These competing effects may make the disaffiliation program ineffective inimproving auditor independence, or it may reduce auditor independence.8 Since there are two possible

08 March 2013 Page 79 of 145 ProQuest

opposing expectations on the effect of the disaffiliation program, we opt to test the following hypothesis in a nullform: H1a (in null form): The likelihood of receiving qualified opinions by public companies does not changesubsequent to the disaffiliation program. Role of Auditor Size Similarly, we also provide the followingnondirectional hypomesis to explore the moderating role of auditor size on the linkage between the frequency ofqualified opinions and the disaffiliation program: H1b (in null form): The magnitude of change in the likelihood ofreceiving qualified opinions is me same for companies audited by large versus small auditors. EarningsManagement Our second hypothesis investigates whether the improved auditor independence subsequent tothe disaffiliation program is associated with higher quality of earnings (lower earnings management). Priorresearch has offered evidence on the inverse relationship between earnings management and auditorindependence (see, for example, Davis et al. 2002). After the disaffiliation, auditors could independentlydetermine their own business strategies, design operating policies, and perform their audit services. At thesame time, they had to face business risks, in particular the litigation risk when violating professional standardsand practice rules (Yang et al. 2001). Since auditors face less pressure from government agencies afterseparating from their sponsoring bodies but more litigation risk due to enhanced professional regulations (seeearlier sections), they are expected to be more independent and, therefore, more likely to constrain aggressiveearnings management. On the other hand, as discussed earlier, if the disaffiliation program results in lessauditor independence, they are less likely to constrain aggressive earnings management. Based on thesediscussions, we test the following null hypothesis: H2a (in null form): Earnings management behavior ofmanagers of public companies does not change subsequent to the disaffiliation program. Role of Auditor SizeDeFond et al. (2000) found that large auditors have a higher propensity to issue qualified opinions in theChinese emerging market. However, it is not clear whether large auditors in China are also more likely to deterearnings management behavior of their clients. As argued in DeFond et al. (2000), larger auditors tend to actmore independently because (1) they suffer large losses in a litigation crisis and a credibility crisis when meyare in violation of auditing regulations (De Angelo 1981), and (2) mey tend to invest more resources, such asprofessional education, to increase audit quality (Dopuch and Simunic 1980). Therefore, since large auditorsare already inherently more independent (DeAngelo 1981; Becker et al. 1998), they are likely to be less affectedby the disaffiliation program. We set up the following null hypothesis to test this expectation: H2b (in null form):The magnitude of change in earnings management is the same for companies audited by large versus smallauditors. RESEARCH DESIGN Qualified Audit Opinions In our univariate analysis of the first group ofhypotheses, we compare the relative frequency of types of audit opinions of all companies as well as affiliatedand nonaffiliated auditors in the pre- and post-disaffiliation periods. In addition, we use auditor size as ameasure of audit quality and compare the relative frequency of types of audit opinions for large auditors andsmall auditors in each period. Following DeFond et al. (2000) and Gulf et al. (2003), we define large auditors asthe first ten auditors with the largest market shares based on the total assets of clients (see the Appendix for thelist of top-ten auditors). Since auditor size is related to auditors' independence (DeFond et al. 2000), controllingfor auditor size helps to isolate the effect of auditor size from that of the disaffiliation program on auditorindependence.

08 March 2013 Page 80 of 145 ProQuest

If the act of the disaffiliation by itself improves auditor independence, then the coefficients of PERIOD, DISHIS,and their interaction would be positive and significant. Further, if the coefficient of PERIOD is positive andsignificant but the coefficients of DISHIS and its interaction with PERIOD are insignificant, this would imply thatthe disaffiliation program (the disaffiliation action coupled with the related regulations) improved auditorindependence. TOPAUD is expected to have a positive and significant coefficient, while the interactioncoefficient for TOPAUD*DISHIS is expected to be negative and significant. Earnings Management While priorstudies have used discretionary accruals to proxy for earnings management (Dechow et al. 1995; Guay et al.1996), in this study, following Chen and Yuan (2004), we use noncore operating earnings (net income minuscore operating income) as a proxy for earnings management, for the following reasons (see also Chen andYuan 2004; Jian and Wong 2004). First, the tradition of rule-based accounting in China permits only limitedopportunities for accruals. Second, the Chinese Securities Regulatory Commission (CSRC) has set a clearbenchmark of return on equity for stock issues and trading. For instance, the publicly listed companies wererequired to have a minimum of 10 percent ROE for three consecutive years prior to any public offering before1998 (CSRC Notice No. 17, 1996), which was later modified to a minimum of 10 percent average ROE and aminimum of 6 percent ROE in each year (CSRC Notice No. 12, 1999). In addition, when a listed company incurslosses for two consecutive years, its stock shares will receive a prefix of "ST" (special treatment) as a warningfrom the stock exchanges; if it sustains losses for three consecutive years, its shares will be temporarily delistedwith a prefix of "PT" (particular transfer) and allowed to trade only on Fridays, with many other transferlimitations (Article 157, Company Law). Under these circumstances, listed companies in China have a tendencyto manipulate earnings through nonrecurring items. Regulators also regard nonrecurring items as potentialearnings management items. Since 1999, the CSRC has been adopting a different method that excludes

08 March 2013 Page 81 of 145 ProQuest

infrequent items - such as gains and losses from the investment and sales of fixed assets - from the calculationof ROE. However, management may still use noncore operating income to manage earnings for other reasons,such as keeping stockholders happy, generating positive impact on stock prices, and managementcompensation based on net income or stock prices. We deflate noncore operating income by year's end totalequity to calculate a noncore return on equity (NROE).

If the coefficients of PERIOD, DISHIS, and their interaction are negative and significant, this would imply thatthe act of disaffiliation improved auditor independence. However, if the coefficient of PERIOD is negative andsignificant but the coefficients of DISHIS and its interaction with PERIOD are insignificant, this would imply thatthe disaffiliation program reduced earnings management. TOPAUD is expected to have a negative andsignificant coefficient as well as a positive and significant interaction with PERIOD. EMPIRICAL RESULTSSample Selection The original sample consists of all companies listed on the Shanghai and Shenzhenexchanges with sufficient data for the period 1995-2000. The periods of 1995-1996 and 1998-2000 representthe pre-disaffiliation and post-di saffi liation periods, respectively. The year 1997 is excluded because auditorshad two years (1997 and 1998) to comply with the disaffiliation rule, making 1997 a transition year. To eliminatethe effect of new auditing standards, we exclude observations of years 1993 and 1994 from the pre-disaffiliationperiod. The financial data are collected from the Taiwan Economic Journal (TEJ) database. As shown in PanelA of Table 1, the sample selection starts with 6,019 firm-year observations (1,248 firms) that comprise thepopulation of Chinese listed companies over the period 1995-2000. Data on auditor name and opinion type arecollected from annual reports of these companies. We delete 1,660 firm-year observations that are prior tofirms' listing dates and 13 observations for which audit opinion or auditor name is not available.10 We furtherexclude 711 firm-year observations in the transition year of 1997, yielding 3,635 firm-year observations (1,074firms) for our analysis on audit opinions. Panel B shows the distribution of sample firms over each year of thesample period. To test for the earnings management hypothesis, we require that firms must have operatingincome data. This yields 2,765 firm-year observations (838 firms) for our analysis of the noncore operatingincome. Again, Panel B shows the distribution of sample firms over each year of the sample period. Since theliterature documents that the IPO firms tend to manage their earnings upward to meet the criteria for IPO on theexchanges (Aharony et al. 2000), the inclusion of IPO firms usted in the post-adoption period might make itdifficult to detect the effect of the disaffiliation program on earnings management. Hence, we conduct asensitivity test using only the non-IPO firms. Qualified Opinions Table 2 presents univariate tests of our firsthypothesis by comparing the frequency of qualified opinions in the pre- and post-disaffiliation periods. Panel Ashows that an average of 10.50 percent of firms received qualified opinions prior to the disaffiliation programand 17.58 percent afterward. The t-test indicates that the difference in the frequency of qualified opinions issignificant (p <0.01). Panel B shows that the average relative frequency of qualified opinions increased from08 March 2013 Page 82 of 145 ProQuest

3.52 percent to 14.05 percent for nonaffiliated auditors and from 11.96 percent to 18.11 percent for affiliatedauditors. The t-tests indicate mat the frequency of qualified opinions for both groups significantly increased. Thecombined results of Panels A and B suggest that me improvement in auditor independence may be attributed tothe disaffiliation program (the act of disaffiliation coupled wim related regulations) rather than the act ofdisaffiliation by itself. Panel C shows mat the average relative frequency of qualified opinions increased from15.53 percent to 17.72 percent for top-ten auditors and from 7.45 percent to 17.53 percent for non-top-tenauditors. All the changes are significant (p <0.01). The Chi-square test indicates mat the non-top-ten auditorsexperienced significantly larger changes than the top-ten auditors (Chi-square = 30.62, p <0.001). Thus, theresults of the univariate tests suggest a rejection of the null H1a and H1b in favor of the alternative hypotheses.

Table 3 reports descriptive statistics on client firm characteristics relevant to qualified opinions for the pooledsample and for each period. Similar to the findings of DeFond et al. (2000), top-ten auditors have smallermarket shares subsequent to the disaffiliation program. Client firms tend to have a longer listed history; be lesslikely to have overseas investors (less likely to issue B- or ?-shares); and have higher market value, lower returnon equity ratio, lower inventory (all significant at ? <0.01) and lower accounts receivable (significant at p <0.10).

08 March 2013 Page 83 of 145 ProQuest

08 March 2013 Page 84 of 145 ProQuest

Table 4 presents the Spearman correlation coefficients for major variables in the multivariate analysis. Thelower triangular matrix presents the Spearman correlation coefficients for the variables in the audit opinionmodel - Model (1), while the upper triangular matrix presents those for the variables in the earningsmanagement model - Model (2). No correlation coefficient exceeds 0.30, except those between current ratio(CURRAT) and debt to equity ratio (DETRAT) and between accounts receivable (REQ and inventory (INV).Since these are only control variables, we do not expect this correlation to affect the results of our test variables.However, as a robustness check, we drop current ratio and/or accounts receivable and find the results to bequalitatively similar to those reported. Table 5 reports the logistic regression results of Model (1). The resultsshow that the coefficient on the PERIOD dummy is significantly positive (p <0.01). However, neither thecoefficient of DISHIS nor its interaction with PERIOD is significant. The significant results on other controlvariables are mostly consistent with the literature (Dopuch et al. 1987; DeFond et al. 2000). Based on theseresults, we reject the null HIa in favor of the alternative that the likelihood of receiving qualified audit opinions forpublic companies increased subsequent to the disaffiliation program. That is, the disaffiliation program, ratherthan the act of disaffiliation itself, resulted in an increase in firms' likelihood of receiving qualified opinions fromtheir auditors, indicating an increase in auditor independence. The coefficient on the variable for auditor size(TOPAUD) is positively significant, and that of its interaction with the PERIOD is significantly negative. Theseresults reject the null HIb and indicate that the magnitude of increase in the likelihood of receiving qualifiedopinions, subsequent to the disaffiliation program, is lower for firms audited by large audit firms. In a similar test,when only companies with affiliated auditors rather than a pooled sample are examined, we find similar results.Combined with the decreasing market shares by top auditors, the negative coefficient for TOPAUD interactionwith PERIOD suggests that the disaffiliation program largely impacted smaller auditors by increasing theirmarket share and encouraging them to issue qualified opinions to their clients when such opinions areappropriate. Earnings Management Table 6 presents the univariate tests on the difference in noncore operatingincome between different periods and between companies audited by different types of auditors. It shows thatthe noncore operating income is significantly lower after the disaffiliation period for all companies, thosecompanies audited by affiliated auditors, and those audited by nontop-ten auditors. These results indicate thatthe act of disaffiliation by itself is associated with the decline in earnings management via noncore operatingincome. However, univariate test only suggests a significant difference for the effect of the act of disaffiliation byitself. It should be noted that this result is without controlling for the effect of other firm characteristics, throughmultivariate analyses, that might affect noncore operating earnings. Table 7 reports the multivariate regressionresults. It shows that the coefficient on PERIOD is negative and significant, but neither the coefficient of DISHISnor that of its interaction with PERIOD is significant. These results support a rejection of null H2a in favor of thealternative that there is less earnings management subsequent to the disaffiliation program, leading to theconclusion that the disaffiliation program, rather than the act of disaffiliation itself, resulted in a decrease inmanagers' earnings management.

08 March 2013 Page 85 of 145 ProQuest

08 March 2013 Page 86 of 145 ProQuest

08 March 2013 Page 87 of 145 ProQuest

The coefficients of variables TOPAUD and the interaction between TOPAUD and PERIOD are not significant,thus suggesting that auditor quality did not affect the relationship. Among the control variables, the coefficientsof AGE and FOREIGN are significant with expected signs, while the coefficient of ROE is significant with anunexpected sign. Although the coefficients of other control variables (except LNSIZE and CURRAT) have theexpected signs, these coefficients are not significant.

Alternative Explanations and Sensitivity Tests Changes in the Accounting System Although the evidence fromour study supports the contention that the disaffiliation program rather than the act of disaffiliation itselfimproves auditor independence, an alternative explanation for the results could be the change in the institutional

08 March 2013 Page 88 of 145 ProQuest

environment. In 1998, the accounting system of stock companies was revised for the first time to eliminate thedifference between Chinese accounting and IAS and to provide managers more opportunities to makejudgments (Zheng and Stone 2005). These revisions included: (1) more rights for managers to decide thepercentage for doubtful accounts, depreciation rate, and amortization period; (2) new regulations on debtrestructuring; and (3) lowering the requirement of equity method on long-term investment and consolidation.Since the new accounting system was more complex and would require more professional estimation andjudgment, disagreements between auditors and listed firms were more likely to occur, thereby increasing thenumber of qualified opinions. However, no audit report issued in 1998 contained qualifications related to the"new accounting system," which suggests that auditors treated the 1998 system as a continuous form of the1992 system. In addition, as a sensitivity test, we investigated whether any audit opinion was associated withthe items under revision. Our results are not affected after deleting the cases of disagreement between auditorsand managers due to the use of new accounting methods under the revised accounting system.11 Changes inthe Profitability Requirements As already mentioned, the public offering requirement of ROE was changed inour sample period. Before 1998, ROE was required to be at least 10 percent for each of three consecutiveyears. Starting from 1999, ROE was required to be at least 6 percent for each of three consecutive years, withthe three-year average no less than 10 percent. Thus, the profitability requirement was lower in 1999-2000 thanin 1996-1998, which might explain the reduction in earnings management of newly listed firms in 1999-2000.Moreover, since the CSRC made more efforts to stamp out earnings management in the process of approv- ingthe rights issues of listed firms after 1998 (Chen and Yuan 2004), instances of earnings management by newlylisted firms would be expected to decrease. To test these possibilities, we conducted a sensitivity test bycomparing earnings management data (NROE) for observations in 1995-19% with those in 1998 and foundsimilar results as those reported. In addition, we deleted all BPO firms in our sample, and the results arequalitatively the same as reported. These results provide us with some indications that the changes in earningsmanagement are not introduced by the changes in the CSRC benchmark on rights issue and the relevantenforcement employed after 1998.

08 March 2013 Page 89 of 145 ProQuest

Changes in the Delisting Regulation There was a change in the regulation on delisting around 1998. Forexample, from 1998, the Shanghai and Shenzhen Stock Exchanges passed the regulation on ST (specialtreatment) for the loss firms. The regulation on ST passed in 1998 could work against our finding any results,because after this regulation the firms are expected to engage in more earnings management to preventdelisting. Nevertheless, to detect such an effect, we conducted a sensitivity test by deleting all ST firms, and theresults were qualitatively the same as those reported. Different Earnings Management Models Our earningsmanagement model is a modified version of the Chen and Yuan (2004) model. We did not include the samevariables, primarily to avoid the multicollearity problem. For instance, when industry-adjusted nonoperatingincome (nonoperating income industry median) or managed operating income (10 percent-industry-adjustednonoperating income) is added to our model, we could not get any results, since the model is not fully ranked,i.e., some variables are the combinations of other variables. However, we did test these variables as alternativemeasures of earnings management and found similar results as those reported. In addition, we included auditopinions and lagged noncore operating income in the earnings management models as a sensitivity test andfound no significant difference in the results compared with those reported. We also used discretionary accrualsbased on the modified Jones model as a measure of earnings management. We find that with a paired sampledesign, there is a marginal decrease in discretionary accruals in univariate test. In multivariate test, we alsofound a similar decrease in the group of companies audited by smaller auditors but not in the other group. Thus,using discretionary accruals as a measure of earnings management, we provide some evidence (albeit weak) ofdecline in earnings management in the post-disaffiliation period. Endogeneity Given that auditor choice andearnings management could be affected by the same group of firm characteristics' variables, such as firm size,ownership structure, financial performance, and the complexity of operations, we conducted a two-stage least-squares regression analysis as a sensitivity test. The first model is our earnings management model - Model(2), and the second model is auditor choice model, including firm age, prior audit opinion, foreign ownership,firm size, return on equity, debt-to-equity ratio, current ratio, receivables, and inventory (deflated by total assets)as explanatory variables. The results for the earnings management model remain qualitatively the same asthose reported. Government' versus Non-Government-Owned Clients To compare the group directly affected bythe disaffiliation program with others not affected (less affected), we investigated whether non-government-owned clients with affiliated auditore are less affected than government-owned clients with affiliated auditors.12We found that both groups were affected by the disaffiliation program, with a significant coefficient for thePERIOD dummy. No significant difference is found between the two groups, as the coefficients for the dummyfor government-owned clients and its interaction term with PERIOD are insignificant. We further examined theeffect of auditor size in each group of clients separately. We found that in both groups, the likelihood of smallauditors issuing modified audit opinions increased after the disaffiliation program much more than largeauditors. This indicates that the disaffiliation program (the act of disaffiliation coupled with relevant regulations),rather than the act of disaffiliation alone, resulted in increased auditor independence. Other Tests Our definitionof top-ten auditor is based on the market share in each year, thus the list of top-ten auditors varies over time. Analternative measure is to use a consistent list based on either the market share before the disaffiliation program(e.g., Year 1995) or the list of top-ten auditors after the disaffiliation program. Similar results are found for thesealternative measures. The prior-period qualifications could be a good predictor of current-period qualificationswhen it can a take longer period of time to resolve a qualified opinion, such as a going concern opinion. Toaddress this issue, we included a dummy for prior audit opinions (1 for qualified opinions and 0 otherwise) andfound our results to be qualitatively the same. Finally, since firms in the utilities and finance industries maybehave differently, we dropped these firms from our analysis and found that the results are qualitatively similarto those reported. CONCLUSIONS AND LIMITATIONS This paper investigates the effect of implementing thedisaffiliation program in China on auditor opinions and earnings management. We hypothesize that the

08 March 2013 Page 90 of 145 ProQuest

disaffiliation program increases auditor independence, which leads to the issuance of more qualified auditopinions and reduction of earnings management. Our results in general support our expectations. We find thatthe likelihood of receiving qualified audit opinions significantly increased after auditors were separated from theirsponsoring bodies. Companies audited by auditors without any affiliation also experienced significant increasesin the likelihood of receiving qualified opinions. These results indicate that the disaffiliation program, rather thanthe act of disaffiliation itself, increased auditor independence. The results further indicate that the magnitude ofthe effect of the disaffiliation program is more pronounced for smaller auditors compared with larger auditors. Inaddition, we find that noncore operating income decreased subsequent to the disaffiliation program. Neither thedisaffiliation history of auditors (whether an auditor was disaffiliated before the disaffiliation program) nor auditorsize significantly affects the results, again indicating that the disaffiliation program, rather than the act ofdisaffiliation alone, resulted in increased auditor independence, which in turn helped reduce earningsmanagement. This study suffers from the usual limitations of this type of empirical investigation in a rapidlychanging economy like China. For example, it is not possible to completely rule out the effects of other possibleinstitutional factors that have evolved in the Chinese corporate landscape and that are unknown to us. We alsorecognize the difficulties of completely disentangling the effects of the disaffiliation program from those of otherregulation reforms, constituting a limitation on the overall scope of our study. Despite these limitations, thisstudy improves our understanding of the effects of various policy initiatives undertaken by the Chinesegovernment to improve auditor independence. Footnote 1 Throughout this paper, we use the terra "disaffiliationprogram" to refer to the act of disaffiliating auditors from their sponsoring bodies as well as related and morestringent regulation that was simultaneously or subsequently enacted to assure that the program achieved itsobjective of improving auditor independence. 2 For details, see Xiao et al. (2000), Lin and Su (1998). 3 Size isbased on the DeAngelo (1981) quasi-rents argument that firms with a large client base have more to lose interms of quasi rents in the event of an audit failure, and, merefore, are more likely to provide higher qualityaudits. Footnote 4 Hereafter, the term "CPA firms" refers to both accounting and auditing firms. Footnote 5 Withthe opening of the Chinese accounting and auditing markets to international accounting finns, Chinese CPAfirms will have to compete with international CPA firms as financially and operationally independent entities(Xiao et al. 2000). Footnote 6 However, DeFond et al. (2000) do not consider me emerging market features inexplaining the increase in qualified opinion and decrease in IPO market share of the top ten auditors. The lossof IPO market share for the top ten auditors might be due to the emergence of new CPA firms. The new CPAfirms typically have difficulties in competing with the existing CPA firms, and they access the auditing market byproviding service to the IPO firms. In addition, clients in the IFO market tend to have their accounting reportsaudited by their original auditors when diese auditors receive authorization to audit listed companies. 7 Forinstance, DeFond et al. (2000) reported that of 63 CPA firms auditing listed Chinese companies in 1995, 56CPA firms were affiliated with government agencies (including six firms affiliated with governmentsponsoringcolleges/institutions), and only seven CPA firms were joint ventures. In our study, we include companies auditedby either nonaffiliated or government-affiliated auditors. 8 We thank one of the anonymous reviewers forsuggesting the possible "opposing" effects of the disaffiliation program. Footnote 9 In addition to shares issuedto domestic investors (called A-shares), companies were allowed to issue shares to foreign investors (called B-shares) through the two national exchanges or through other exchanges (such as H-shares, traded on the HongKong Stock Exchange). Both A- and B- (or H-) shares convey equal rigiats to the same company, although theyan different in terms of ownership. However, A-share investors receive accounting information prepared underChinese (MAP and mostly audited by local CPA firms, while B-share (or H-share) investors receive accountinginformation prepared under lAS (or Hong Kong GAAP) and mostly audited by international accounting firms. Inthis study, we focus on the impact of auditing regulations on audit quality and earnings quality. Therefore, weonly examine the A-shares where there are direct impacts. Footnote 10 TEJ database includes financialinformation on public companies prior to their initial public offerings, which was collected from their first annual

08 March 2013 Page 91 of 145 ProQuest

reports. Footnote 11 We find only vely few cases in which auditors reported amortization issues in theexplanatosy paragraph of the audit report. Footnote 12 One may expect that nonlocal clients of affiliatedauditors will be less affected by the disaffiliation program than their local clients (those in the same region as theauditor). Because of the limitation on the availability of data on locations of auditor offices and companies'headquarters, we did not examine such a possibility in our paper. We consider this issue to be a limitation of ourstudy and would like to leave it for future research. References REFERENCES Aharony, J., J. Lee, and T. J.Wong. 2000. Financial packaging of IPO firms in China. Journal of Accounting Research 38 (1): 103-126.Becker, C, M. DeFond, J. Jiambalvo, and K. R. Subramanyam. 1998. The effect of audit quality on earningsmanagement. Contemporary Accounting Research 15 (1): 1-24. Chen, C. W. K., and H. Q. Yuan. 2004.Earnings management and capital resource allocation: Evidence from China's accounting-based regulation ofrights issue. The Accounting Review 79: 645-665. Davis, L. R., B. Soo, and G. Trompeter. 2002. Auditor tenure,auditor independence and earnings management. Working paper, Boston College and Michigan TechUniversity. DeAngelo, L. 1981. Auditor size and audit quality. Journal of Accounting and Economics 3 (3):183199. Dechow, P. M., R. G. Sloan, and A. P. Sweeney. 1995. Detecting earnings management. TheAccounting Review 70 (2): 193-226. DeFond, M. L., T. J. Wong, and S. Li. 2000. The impact of improved auditorindependence on audit market concentration in China. Journal of Accounting and Economics 28: 269-305.Dopuch, N., and D. Simunic. 1980. The nature of competition in the auditing profession: A descriptive andnormative view. The Illinois Auditing Symposium. _____ , R. Holthausen, and R. Leftwich. 1987. Predictingaudit qualifications with financial and market variables. The Accounting Review 62: 431-454. Fox, J. 1998. Thegreat emerging market rip-off. Fortune 137 (May 11): 98-110. Graham, L. 1996. Setting a research agenda forauditing issues in the People's Republic of China. The International Journal of Accounting 31 (1): 19-37. Guay,W. R., S. P. Kothari, and R. L. Watts. 19%. A market-based evaluation of discretionary accrual models. Journalof Accounting Research 34 (Supplement): 83-105. Gul, F. A., S. Y. Sun, and J. S. Tsui. 2003. Audit quality,earnings, and the Shanghai Stock Market reaction. Journal of Accounting, Auditing &Finance 18 (3): 41 1-427.Jian, M., and T. J. Wong. 2004. Earnings management and tunneling through related party transactions:Evidence from Chinese corporate groups. Working paper, Nanyang Technological University and Hong KongUniversity of Science and Technology. Lin, Z. Y., and Y. R. Su. 1998. Accounting disclosure: Motivation,development, problems and suggestions. In The Finance and Accounting Issues in Stock Markets (in Chinese),edited by J. L. Zhang and S. H. Li, 41-80. Publishing House of Shanghai University of Finance and Economics.Liu, K. C, and W. G. Zhang. 19%. Contemporary Accounting Issues in China: An Analytical Approach. NewYork, NY: Prentice Hall. Rask, R., D. Chu, and T. Gottschang. 1998. Institutional change in transitionaleconomics: The case of accounting in China. Comparative Economic Studies 40 (4): 76-100. Xiang, B. 1998.Institutional factors influencing China's accounting reforms and standards. Accounting Horizons 12(2): 105-119.Xiao, J. Z., Y. Zhang, and Z. Xie. 2000. The making of independent auditing standards in China. AccountingHorizons 14 (1): 69-89. Yang, L., Q. Tang, A. Fulgore, and Y. Jiang. 2001. Auditor-government associations andauditor independence in China. The British Accounting Review 33 (2): 175-189. Zheng, L., and M. Stone. 2005.The information content of earnings and cash flows under new accounting system. Working paper, NorthwestIllinois University. AuthorAffiliation Ferdinand A. Gul is a Professor at The Hong Kong Polytechnic Universityand is the Maybank Chair at the University of Malaya, Heibatollah Sami is a Professor at Lehigh University, andHaiyan Zhou is an Assistant Professor at The University of Texas-Pan American. We thank Dan Simunic (theeditor) and two anonymous reviewers for helpful suggestions. Also, we thank Niranjan Chipalkatti (discussant ofthe paper during the 2005 Annual Meeting of the AAA) and workshop participants at the City University of HongKong for their comments and suggestions. An earlier version of this paper was presented at the 2005 AnnualMeeting of the American Accounting Association and the 2005 Annual Conference of the British AccountingAssociation. Editor's note: Accepted by Dan Simunic. Submitted: September 2005 Accepted: June 2008Published Online: May 2009

08 March 2013 Page 92 of 145 ProQuest

Subject: Accountant independence; Auditors opinions; Quality; Earnings management; Correlation analysis;Affiliates; Studies Location: China Classification: 9179: Asia&the Pacific, 9130: Experiment/theoretical treatment, 8305: Professional services notelsewhere classified Publication title: Auditing Volume: 28 Issue: 1 Pages: 29-51 Number of pages: 23 Publication year: 2009 Publication date: May 2009 Year: 2009 Publisher: American Accounting Association

08 March 2013 Page 93 of 145 ProQuest

Place of publication: Sarasota Country of publication: United States Journal subject: Business And Economics--Accounting ISSN: 02780380 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Equations;Tables;References ProQuest document ID: 216738337 Document URL: http://search.proquest.com/docview/216738337?accountid=86413 Copyright: Copyright American Accounting Association May 2009 Last updated: 2012-07-12 Database: Accounting&Tax

_______________________________________________________________ Document 8 of 10 Corporate governance, firm characteristics and risk management committee formation in Australiancompanies Author: Subramaniam, Nava; McManus, Lisa; Zhang, Jiani Publication info: Managerial Auditing Journal 24. 4 (2009): 316-339.ProQuest document link Abstract: The purpose of this paper is to examine how a risk management committee (RMC), as a newlyevolving sub-committee of the board of directors, functions as a key governance support mechanism in theoversight an organisation's risk management strategies, policies and processes. However, empirical evidenceon the factors associated with the existence and the type of RMCs remains scant. Using an agency theoryperspective, this study investigates the association between board factors such as proportion of non-executivedirectors, Chief Executive Officer duality, and board size; as well as, other firm-related factors (e.g. auditor type,industry, leverage, and complexity), and the existence of a RMC, and the type of RMC (namely, a separateRMC versus one that is combined with the audit committee). Data was collected from the annual reports of thetop 300 Australian Stock Exchange (ASX)-listed companies. The results, based on logistic regression analyses,indicate that RMCs tend to exist in companies with an independent board chairman and larger boards. Further,the results also indicate that in comparison to companies with a combined RMC and audit committee, those witha separate RMC are more likely to have larger boards, higher financial reporting risk and lower organisationalcomplexity. The paper presents data limited to top 200 top ASX-listed companies, thus restrictinggeneralisability of the results. The findings of this study provide additional information on the use and design ofRMCs in a voluntary setting. Full Text: Introduction In the aftermath of the recent corporate collapses, numerous governance initiatives havebeen proposed for improving corporate governance with significant emphasis placed on the role of riskmanagement. An effective risk management system is seen to help the organisation achieve its businessobjectives, enhance its financial reporting as well as safeguard its reputation. In Australia, the Corporate

08 March 2013 Page 94 of 145 ProQuest

Governance Council of the Australian Stock Exchange (ASX) has set guidelines for risk management withinAustralian public listed organisations and the board of directors are seen to hold the primary responsibility overthe establishment and implementation of a proper risk management system. In 2007, the ASX released itsrevised Corporate Governance guidelines with specific amendments broadening its recommendations on riskmanagement and its disclosure. For example, the best practice guideline 7.1 requires not only the board (or anappropriate board committee) to establish policies on risk oversight and management, but also to disclose asummary of the policies. Further, best practice guideline 7.2 specifically advises that, it is the board'sresponsibility to ensure management designs and implements the firm's risk management and internal controlsystem in order to manage the firm's material business risks ([5] ASX, 2007, p. 33). In addition, the board isrequired to disclose whether it has received assurance from the Chief Executive Officer (CEO) and the ChiefFinancial Officer that the financial statements as reported by the company are founded on a sound system ofrisk management and internal compliance and control. It is also specifically noted that "a board committee is anefficient mechanism for focusing the company on appropriate risk oversight, risk management and internalcontrol" ([5] ASX, 2007, p. 33), and that an appropriate board committee may be the audit committee, the riskmanagement committee (RMC) or other relevant committee, although ultimate responsibility for riskmanagement would still rest with the full board. Organisations however tend to differ in their approaches, thestructures and processes adopted towards managing risks. While traditionally significant attention has fallen onthe audit committee for achieving proper risk management ([27] Harrison, 1987; [31] Korosec and Horvat,2005), in more recent times there has been significant growth in RMCs which are specialised risk-focused boardcommittees. A RMC is defined as a sub-committee of the board of directors that provides enterprise riskmanagement education at board level, establishes buy-in at board level for risk appetite and risk strategy,develops "ownership" of risk management oversight by the board, and reviews risk reports of the enterprise([32] KPMG, 2001). Such a committee is potentially a critical resource for boards in meeting their riskmanagement responsibilities. Yet, empirical evidence on the formation and nature of RMCs remain scant andlimited. Objectives of the study The objectives of the present study are twofold: To examine the factorsassociated with the establishment of a RMC in Australian companies. More specifically, the study examineswhether selected board factors and firm characteristics are significantly associated with the existence of a RMCas a sub-committee of the board. Namely, the study examines whether the existence of a RMC is associatedwith board factors, such as the proportion of non-executive directors, CEO duality, and board size, as well as,related firm characteristics, such as the type of auditor, organisational complexity and risk variables includingfinancial reporting risk and leverage. The objective of the study is to investigate the association between theprior mentioned board and related firm characteristics, and the type of RMC; namely, a separate and distinctRMC versus a RMC combined with the audit committee. Motivation for the study A key motivation for this studyis that while the establishment of RMCs have significantly grown in recent years, presently there is littleempirical evidence on factors associated with the existence of RMCs. For example, a recent survey of 80directors and senior executives from the top 200 ASX/NZSX companies, various government and privateorganisations by [33] KPMG (2005) revealed that over half (54 per cent) the respondent organisations hadestablished a RMC. Of these, 70 per cent were integrated with the board audit committee. However, thefindings of the study remain largely descriptive; with little information on factors determining an organisation'sdecision to set-up a RMC and to disclose its existence in the annual report. Such information is important giventhat having a well-designed board structure including the existence of appropriate sub-committees haveimplications for organisational accountability and performance ([42] Roberts et al. , 2005; [43] Ruigrok et al. ,2006; [27] Harrison, 1987). Another key motivation for this study is that there is increasing concern over boardstructures, particularly the establishment board sub-committees and the inter-relationships among suchcommittees. For example, a number of prior studies have focused on the formation of audit committees ([16]Chau and Leung, 2006; [39] Piot, 2004; [14] Carson, 2002; [2] Adams, 1997; [10] Bradbury, 1990), nomination

08 March 2013 Page 95 of 145 ProQuest

committees ([43] Ruigrok et al. , 2006; [14] Carson, 2002), and remuneration committees ([14] Carson, 2002).The results in general suggest that the formation of such sub-committees is systematically associated withselected organisational and corporate governance factors such as board composition and leadership,ownership and organisational size. For example, [16] Chau and Leung (2006) based on data from 397 publiclytraded firms in Hong Kong found a positive association between the proportion of independent non-executivedirectors on the corporate board and audit committee existence. [14] Carson (2002) found that remunerationcommittees were associated with Big Six auditors, intercorporate relationships and institutional investment.However, no study to date has undertaken a systematic analysis of the association between the existence of aRMC and factors such as board and other company characteristics. The third motivation for this study is thatRMCs are found to be generally integrated or combined with audit committees ([33] KPMG, 2005). However,the expanding roles and responsibilities of audit committees raise various criticisms and doubts as to their abilityto function effectively ([4] Alles et al. , 2005; [27] Harrison, 1987). Given that boards have commonly relegatedboth the financial reporting and risk management oversight responsibilities to audit committees, it can be arguedthat increasing workload pressure on such committees would raise the potential for inefficiencies. For example,[4] Alles et al. (2005, p. 22) contend that audit committee members "however well qualified, often have full-time,high-level responsibilities elsewhere which inhibit their desire and ability to get more involved with the firm".Furthermore, since risk management oversight generally requires significant understanding of evolvingorganisation wide structures and processes and the related risks, it is arguable that a separate RMC is likely tobe more efficient than one that is combined with an audit committee ([61] Collier, 1993; [43] Ruigrok et al. ,2006; [53] Turpin and DeZoort, 1998). Yet, there is little empirical evidence on both corporate governance andfirm-related factors associated with an organisation's decision to establish a separate RMC as opposed to acombined RMC committee. No doubt, such an understanding of the determinants of RMCs is important as theASX requires listed companies to disclose corporate governance practices in their annual reports in a clear andtransparent manner. The remainder of this paper is organised as follows. In the next section, a brief overview isprovided on risk management in business organisations, followed by a literature review of board committeesand the theoretical explanations for their formation. This is followed by hypotheses development, a delineationof the research method and the results of the data analysis. The final section of the paper covers discussion ofthe results and conclusions of the study. Background - risk management Business risks are defined by theInstitute of Internal Auditors Research Foundation (IIARF) as "threats to achieving the entity's objectives" ([52]IIARF, 2003). They are uncertainties about events and/or their outcomes that could have a material effect onthe goals of the organisation ([45] Selim and McNamee, 1999). The management of risks is an integral part ofgood business practice. It has been carried out on an ongoing and informal basis by many organisations.Traditionally, risk management has developed as a professional and technical discipline in a number of keyareas, namely finance, health and safety, clinical and environmental areas. However, organisations areincreasingly facing a variety of risks including financial, operational, reputation, regulatory and information risk([11] Burlando, 1990; [32] KPMG, 2001). Information about an organisation's risk is not only important tomanagement and shareholders, but also to suppliers, creditors, employees and other stakeholders. Theinformation is useful to management and shareholders as it indicates the stability of the organisation'sprocesses and expected results. Further, such information is also useful for creditors for assessing a company'sability to settle its financial liabilities, for suppliers in relation to their decisions about future credit terms, and foremployees assessing their future prospects in the organisation ([31] Korosec and Horvat, 2005). Therefore,proper risk management support structures are likely to help in managing business risks more effectively and indisclosing the risk management outcomes to the organisation's stakeholders. Board committees Committees ofthe board of directors exist to assist the board perform its role more effectively. In particular, with the expandingrole of the board, such committees are increasing viewed to be essential rather than preferable. According to[27] Harrison (1987), there are two types of board committees. One type of committee undertakes a more

08 March 2013 Page 96 of 145 ProQuest

strategic role in terms of advising management and the board on major business decisions, e.g. a strategicplanning committee. The other type of committee relates to the monitoring or oversight function of the boardsuch as audit, remuneration, and nomination committees. These committees are seen to specifically enhancethe accountability of the board as they provide independent oversight of various board activities ([27] Harrison,1987). The [12] Cadbury Committee (1992) has strongly advocated the appointment of oversight committees bythe board, noting that the effectiveness of a board is buttressed by such structures and procedures. In morerecent years, RMCs have gained popularity as an important oversight board committee ([24] Fields and Keys,2003). Commonly, the broad areas of responsibilities of a RMC include: - determining the organisational riskmanagement strategies; - evaluating the organisational risk management operations; - assessing theorganisational financial reporting; and - ensuring the organisation is in compliance with the laws and regulations([19] COSO, 2004; [44] Sallivan, 2001; [46] Soltani, 2005). The committee members are expected to discusswith senior management the state of the organisation's risk management, review the adequacy andmanagement of the risk procedures, and report to the board on its findings. For example, the annual report ofAdelaide Bank Limited states that its RMC "shall review and approve the risk strategy of the company, establishand maintain policies which reflect the risk strategy and monitor the management of credit risk, liquidity risk,market risk and operational risk" ([3] ADB, 2005). Principle 7.2 of the revised Corporate Governance guidelinesas released by the ASX ([5] ASX, 2007) likewise notes that a board committee specifically focusing on riskmatters (such as a RMC) can be an effective mechanism in supporting the full board meet its responsibilities ofrisk oversight, risk and internal control management. A RMC entailing members who are specialists in riskmanagement would be better able to support corporate governance through undertaking an in-depth anddetailed analysis and review of risks and internal controls. Further, specialist boards such as a RMC will be ableto devote more time and effort towards integrating the various risks organisation-wide and evaluating the relatedcontrols as a whole. As such, the role of RMCs in supporting corporate governance is potentially a critical one.Theoretical perspectives on board committee formation A literature review of research on the formation andstructure of governance mechanisms such as board committees reveals agency theory as the dominantparadigm used by prior studies. However, it is increasingly argued that there is a need for a more multi-theoreticapproach towards understanding board composition, roles, and their outcomes ([58] Daily et al. , 2003; [43]Ruigrok et al. , 2006). As such, the development of the research hypotheses pertinent to this study is largelyguided by three key theoretical perspectives, namely: agency theory; corporate legitimacy; and signallingtheory[1] . Agency theory Agency theory provides a rich theoretical premise for understanding organisationalprocesses and designs from a principal-agent perspective ([49] Subramaniam, 2006). An agency relationshipmay be defined as a contract under which one party (the principal) engages another party (the agent) to performsome service on their behalf. The agent is generally assumed to act based on his/her self-interest ([29] Jensenand Meckling, 1976; [34] Lambert, 2001), and the principal has two major avenues for mitigating such costs:monitoring the agent's behaviour by adopting auditing and other governance mechanisms that aligns theagent's interest with that of the principal's; and/or providing attractive employment incentives to the agent andsetting up reward structures that encourage the agent to act in the principal's best interests. The commoninternal monitoring device is the board of directors and the external monitoring mechanism is the externalauditors ([1] Adams, 1994; [6] Baiman, 1990; [29] Jensen and Meckling, 1976; [34] Lambert, 2001; [49]Subramaniam, 2006). The use of agency theory has been predominant in prior studies on board committeessuch as the audit, nomination and remuneration committees ([43] Ruigrok et al. , 2006; [57] Benz and Frey,2007). In general, monitoring board committees are seen to provide better quality monitoring, leading to loweropportunistic behaviour by managers. Such board committees are thus predicted to exist in situations whereagency costs are high, e.g. high leverage and greater firm complexity and size. Furthermore, agency theorysuggests that board characteristics such as its independence and the existence of an independent chairmanare potential factors affecting board committee structures ([16] Chau and Leung, 2006; [14] Carson, 2002; [10]

08 March 2013 Page 97 of 145 ProQuest

Bradbury, 1990). However, agency theory tends to predominantly focus on the motives of human behaviour,particularly from self-interest and ignore other reasons that may guide organisational decisions. For instance,organisational decisions may also be undertaken to conform to institutional norms or to meet selectedstakeholder pressures, thus enhancing organisational legitimacy. Corporate legitimacy Legitimation has beendefined as "the process whereby an organisation justifies to a peer or super-ordinate system its right to exist,that is, to continue to import, transform, and export energy, material, or information" ([36] Maurer, 1971, p. 361).Legitimacy theory is another common perspective that has been adopted to understand organisational formsand structures based on the assumption that a corporation has to maintain its legitimacy for its survival ([59]Meyer and Rowan, 1977). In recent years, there has been increasing focus on the structure and strategiesadopted by a board in meeting various stakeholder needs, and the adoption of monitoring sub-committees maybe viewed as one such strategy for maintaining corporate legitimacy. A monitoring board committee such as aRMC is likely to enhance corporate accountability by providing a mechanism for independent oversight ofcorporate activities, thus promoting corporate legitimacy ([27] Harrison, 1987). Furthermore, with increasingscrutiny from regulatory agencies and industry watchdogs, the use of more visible forms of legitimisation suchas a board committee has become more attractive and prevalent. Key stakeholders such as the externalauditors may also play an important role in encouraging the adoption of such governance mechanisms.Signalling theory Signalling theory is widely used to address problems of information asymmetry in the market([37] Morris, 1987; [15] Certo, 2003). When applied to organisational disclosure practices, signalling theoryproposes that it would be generally beneficial for organisations to disclose good or improved corporategovernance initiatives and practices so as to create a favourable image in the market. For example, currentlythere is no mandatory regulatory requirement for companies to establish RMCs. However, according tosignalling theory, a firm may set-up a RMC so as to flag its commitment to good corporate governance. In turn,such a disclosure is expected to minimise any potential for investors' devaluation of the company (oralternatively, to maximise the potential for firm value enhancement. In particular, according to signalling theory,firms with high complexity or in highly dynamic or uncertain industries are more likely to employ such strategiesso as to flag their commitment to good governance. In the following sections, we provide an overview of thenature of RMC adoption, followed by the development of several hypotheses based on selected board and firm-related characteristics. Hypotheses development Nature of RMC adoption In this study, the conceptualisation ofthe existence and nature of RMC is classified under three headings: Nil or non-existent - where a company hasnot set up nor reported the existence of a RMC. A combined committee - where the annual report discloses theexistence of a board committee under the heading of audit and RMC. A separate RMC - where the annualreport discloses the existence a distinct board committee that specifically oversees risk management and istitled as a "RMC". It is argued that both the actual and perceived quality of internal monitoring with respect torisk management is likely to be higher when a RMC exists compared to a situation when there is no RMC.Likewise, it is expected that the actual and perceived quality of internal monitoring in relation to riskmanagement will be higher when a separate and distinct RMC exists compared to when a combined committeeis present. This rationalisation is based on two key reasons: Risk management is a complex process ofidentifying, managing, monitoring and minimising business risks. Thus, having a RMC will enable a board ofdirectors to more effectively deal with assessing the various threats and opportunities faced by an entity. Thus,the actual internal monitoring of risk management will be of a higher standard when a RMC exists. Furthermore,it can be argued that having a separate and distinct RMC will allow committee members to fully focus on thevarious risk processes and reports, and as such provide better quality internal monitoring than when having acombined committee. For instance, a combined RMC and audit committee would not only have to oversee therisk management function but would need to be actively involved with the financial reporting and related auditoversight function as well ([4] Alles et al. , 2005). As such, time constraints and fatigue are more likely to occurin combined committees, which consequently may inhibit the committee members' desire and ability to

08 March 2013 Page 98 of 145 ProQuest

undertake a more rigorous review of the various reports and processes. The establishment and disclosure of aRMC may flag the board's commitment to high quality corporate governance. In other words, a board committeemay exist merely as a signal to the outside world. As argued by [27] Harrison (1987, p. 113), given that it is verydifficult to observe what work these committees actually do, "there is the possibility that monitoring committeeswill be established to create a favourable appearance". Furthermore, the disclosure of a separate RMC willmore strongly reflect and flag the presence of better quality internal risk monitoring mechanism than a combinedcommittee. As such, the perceived quality of risk management monitoring would be highest for firms with aseparate RMC and lowest when there is no RMC. Board characteristics Proportion of non-executive directors The board of directors is an important mechanism for monitoring management behaviour, resulting in bettercorporate accountability and disclosure. The proportion of non-executive directors on the board is seen as a keyindicator of the independence of the board from management. [38] Pincus et al. (1989, p. 246) argue that thepresence of non-executive directors on the board "should increase the quality of monitoring because they arenot affiliated with the company as officers or employees, and thus are independent representatives of theshareholders' interests". Furthermore, it is argued that non-executive directors tend to be more concerned abouttheir reputation, and as such will more actively question and seek higher quality governance than executivedirectors. Prior research has also shown that firms with a higher proportion of non-executive directors havebetter governance in terms of having fewer fraud allegations ([54] Uzun et al. , 2004); lower earningsmanagement ([60] Klein, 2002) and quality and extent of financial disclosure ([17] Chen and Jaggi, 2000).Based on the above discussion, we argue that the board with a large proportion of non-executive directors islikely to more actively enquire about risks, and view that the establishment of a RMC to be an important sourceof support to help them meet their risk management oversight responsibilities than one with a small proportionof non-executive directors. Furthermore, boards with larger non-executive members are also predicted to prefera separate RMC over a combined committee as the former will have greater focus and capacity to more fullyreview the organisation's risk management policies and procedures. Thus, the first set of hypotheses is asfollows: H1(a). The existence of a RMC is significantly and positively associated with the percentage of non-executive directors on the board. H1(b). The existence of a separate RMC is significantly and positivelyassociated with the percentage of non-executive directors on the board. Independent chairman CEO dualityrefers to the same person being the CEO and the chairman of the board. A firm with an independent chairman(i.e. when the CEO and the board chairman are different) is generally viewed to have lower agency costs ([54]Uzun et al. , 2004; [42] Roberts et al. , 2005). For instance, an independent chairman is seen to provide betterboard monitoring by undertaking an independent check on the CEO. Furthermore, an independent chairmancan be seen to have strong motivations to maintain his/her own reputation ([29] Jensen and Meckling, 1976). Assuch, an independent chairman is more likely to seek high quality monitoring, where possible so that there isless chance of organisational failure ([35] Matolcsy et al. , 2004). Empirical evidence likewise suggests that anindependent chairman is associated with audit committee diligence and lower earnings management ([24]Fields and Keys, 2003; [41] Raghunandan et al. , 2001). However, [16] Chau and Leung (2006) found anegative correlation between audit committee existence and the presence of an independent chairman. In thisstudy, we predict that the independent chairman is more likely to promote the establishment of a RMC as itwould enable better monitoring of the risks of the organisation. Furthermore, since a separate RMC is likely tobe specifically focused on risk management matters when compared to a combines RMC, it is expected that aseparate RMC would be significantly and positively associated with an independent chairman. Thus, second setof hypotheses for the study are as follows: H2(a). The existence of a RMC is significantly and positivelyassociated with the use of an independent chairman on the board. H2(b). The existence of a separate RMC issignificantly and positively associated with the use of an independent chairman on the board. Board size Theexistence of a RMC may also be associated with the size of the board. Prior research suggests a positiveassociation between the number of directors and the existence of an audit committee ([10] Bradbury, 1990; [39]

08 March 2013 Page 99 of 145 ProQuest

Piot, 2004). It can be argued that a larger board is likely to entail more resources for the board to allocate. Forexample, the larger the number of members on the board, the greater the opportunity to find directors with thenecessary skills to coordinate and be involved in a sub-committee devoted to risk management. Subsequently,it becomes easier to establish a separate RMC as well. With the greater levels of resources offered by largerboards, there would be less pressure to establish a combined risk management and audit committees. Asdiscussed above, the following hypotheses are thus proposed: H3(a). The existence of a RMC is significantlyand positively associated with board size. H3(b). The existence of a separate RMC is significantly and positivelyassociated with board size. Other predictors Auditor reputation Auditors are a key external monitoringmechanism of an organisation, and in recent years have come to pay significant attention to risk management.Audit firms are generally able to influence their client's internal control systems by making post-auditrecommendations on improving the design of such systems. Big Four audit firms have been found to encouragehigher quality internal monitoring mechanisms among their clients than non-Big Four firms ([18] Cohen et al. ,2004)[2] . This push is seen to be motivated by the need to maintain audit quality and to protect their brandname. For example, prior research has found a positive link between the Big Four audit firms and higher qualityfinancial reporting ([18] Cohen et al. , 2004). It is thus expected that there will be greater pressure amongorganisations with a Big Four firm organisations to establish a RMC than in firms with a non-Big Four. Having anRMC is likely to be viewed as an additional support when an audit firm is in the process of assessing theinternal risk monitoring systems, which in turn is likely to minimise the loss of reputation through audit failure.Likewise, it can be argued that rather than a combined committee, a separate RMC would be preferred by theBig Four firms it is more likely to enhance the quality of risk monitoring and assessment. Thus, the fourthhypotheses-set is as follows: H4(a). The existence of a RMC is significantly and positively associated with theuse of a big four external auditor. H4(b). The existence of a separate RMC is significantly and positivelyassociated with the use of a big four external auditor. Industry type The type of industry may affect the level ofbusiness and related risks of a company ([7] Beasley et al. , 1999). [55] Wallace and Kreutzfeldt (1991) identifythat financial companies are more likely to have an internal control function as they are more risky than otherindustries. Both [26] Goodwin and Kent (2006) and [13] Carcello et al. (2005) identify a positive relationshipbetween the existence an internal audit function and finance companies. Both studies argued some industriesare considered more risky than other industries as they face substantial regulatory and related requirement. Inparticular, financial industry is highly regulated and has compliance risks that exceed those in many otherindustries. In this study, it is predicted that financial companies are more likely to establish RMCs, and also thatsuch RMCs will be separate committees rather than combined with their audit committees. This is because aseparate RMC is expected to provide a higher level of oversight over financial reporting than a combinedcommittee. Thus, the following set of hypotheses is proposed: H5(a). The existence of a RMC is significantlyand positively associated with the organisation in a financial industry. H5(b). The existence of a separate RMCis significantly and positively associated with the organisation in a financial industry. Organisational complexity In general, organisational complexity can be seen to increase with the number of business segments ([13]Carcello et al. , 2005). Organisations with a large number of business segments usually have more productionlines, departments or marketing strategies. As a result, greater complexity increases risks at different levelsincluding operational and technological risks, leading to a greater demand for monitoring such risks. It is thusexpected that an organisation is more likely to have a RMC in order to dilute the risks presented byorganisational complexity. Furthermore, a separate RMC will be preferred over a combined committee inorganisations with greater complexity as separate committees facilitate better quality oversight of risks, both intime and effort of the members. Therefore, based on the above discussion the following hypotheses aresuggested: H6(a). The existence of a RMC is significantly and positively associated with larger number ofbusiness segments. H6(b). The existence of a separate RMC is significantly and positively associated withlarger number of business segments. Financial reporting risk Companies with large proportion of assets that are

08 March 2013 Page 100 of 145 ProQuest

in the form of accounts receivable and inventory are seen to entail higher financial reporting risks due to thehigh levels of uncertainty in the accounting data ([31] Korosec and Horvat, 2005). For instance, the larger theproportion of accounts receivable, the higher the risk of bad and doubtful debts being not properly recognised.Likewise, the valuation of inventory obsolescence is higher in larger inventory balances and thus there is higherfinancial reporting risks. The establishment of a RMC, and particularly a separate RMC, will facilitate betteroversight of such risks. These arguments lead to the following set of hypotheses: H7(a). The existence of RMCis significantly and positively associated with the portion of accounts receivable and inventory. H7(b). Theexistence of a separate RMC is significantly and positively associated with the portion of accounts receivableand inventory. Leverage Companies that have a large proportion of long-term liabilities entail greater financialrisk ([26] Goodwin and Kent, 2006). Highly leveraged firms are more likely to have debt covenants and highergoing concern risks. Lenders are more likely to demand better internal controls and related monitoringmechanisms. Consequently, it can be argued that there will be a greater demand for such companies to have aRMC to oversight such risks. Further, a separate RMC may function more effectively in risk oversight. Thus, thefollowing hypotheses are proposed: H8(a). The existence of a RMC is significantly and positively associatedwith the portion of long-term debts. H8(b). The existence of a separate RMC is significantly and positivelyassociated with the portion of long-term debts. Research method The data was collected from the annualreports of companies listed in the top 300 companies in the S&P all ordinaries index on the ASX for the 2005financial year. The sample of companies comprised a number of industries such as materials, transportation,food and beverage, financial and other service industries. The annual reports were sourced from eachcompany's official web site. Sample From the top 300 publicly listed companies, all trusts, funds and groups andforeign companies, ASX delisted companies, and companies with missing information were eliminated. This lefta final sample of 200 companies. The details of the elimination process are outlined in Table I [Figure omitted.See Article Image.]. The industry breakdown of the final sample of 200 companies is presented in Table II[Figure omitted. See Article Image.]. A wide cross-section of industries is represented in the sample. Dataanalysis Data analysis was conducted using logistic regression analysis. Logistic regression is suited to modelswhere the dependent variable is dichotomous. In the present study, the two dependent variables, RMCexistence and separate and distinct risk management committee (SRMC) existence, are dichotomous. Twomodels were tested regressing nine independent variables against each dependant variable. The followinglogistic regression equation was tested: Equation 1 [Figure omitted. See Article Image.] The two dependentvariables were measured by: RMC existence: a dichotomous variable where 1 = the existence of a RMC (eithera separate committee or a committee combined with the audit committee) and 0 = no RMC. SRMC existence: adichotomous variable where 1 = the existence of a separate RMC and 0 = no separate RMC (i.e. a combinedRMC with the audit committee). The nine independent variables were measured by: INDEPCHAIR: adichotomous variable coded as 1 = company has an independent chairman and 0 = no independent chairman.NONEXECDIR: the percentage of non-executive directors on the board calculated by the number of non-executive directors divided by the total number of directors on the board. BOARDSIZE: the total number ofdirectors on the board. BIGFOUR: a dichotomous variable where 1 = external auditor was a member of the "BigFour" accounting firms and 0 = otherwise. TYPE: a dichotomous variable coded as 1 = company is a financialcompany and 0 = company is a non-financial company. This measure follows the approach adopted by [26]Goodwin and Kent (2006) and [13] Carcello et al. (2005). In the present study, companies in both the bankingand diversified financial industries are considered financial industry firms. BUSSEGMENT: the number ofbusiness units in a company. REC&INV/ASSET: the proportion of assets that are in the form of accountsreceivable and inventory. The variable was calculated as the sum of the accounts receivable and inventorybalances divided by total assets. DEBT/ASSET: the proportion of total long-term debt to total assets. Thevariable was calculated as the total long-term liabilities divided by total assets. SIZE: the total assets of thecompany. Size was included as a control variable in the study. Since agency costs are expected to be higher in

08 March 2013 Page 101 of 145 ProQuest

larger organisations, it is suggested that increased agency costs is likely to lead to greater monitoring and thusthe need for risk management ([13] Carcello et al. , 2005; [26] Goodwin and Kent, 2006). Therefore, companysize is included as a control variable in this study. Results Descriptive statistics Of the sample of 200companies, 44 per cent have a RMC (n =88) and 56 per cent do not have a RMC (n =112). Of the 88companies that have a RMC, 25 per cent have a separate RMC (n =22) and 75 per cent have a combined RMC(n =66). For the 22 firms with a separate RMC: - Average number of members is 4.6, with three being thesmallest, ten being the largest. - Average number of meetings, 5.8 meetings a year, range being 1-13. - Onaverage there were 0.8 independent directors, with ten firms comprising fully independent members. - Overlapbetween AC and RMC was minimal, however, in most cases at least one member from the AC was present onthe RMC. - There were also executive directors on the RMC in 50 per cent of the firms. The charter of theseparate RMC generally appear to cover a whole of organisation approach to risk management. Threeexamples of a RMC charter as reported in the annual report are as follows: The risk committee is responsiblefor ensuring that risks, and also opportunities, are identified on a timely basis and that the group's objectivesand activities are aligned with the risks and opportunities identified by the board. Areas of risk which areconsidered by the RMC include: - safety; - the environment; - the community in which the company operates;and - minimization of business risk. The responsibilities of the Risk Committee include: - reviewing the group'srisk profile within the context of the risk - return profile determined by the board; - implementing and reviewingrisk management and internal compliance and control systems throughout the group; - reviewing the adequacyand effectiveness of the group's compliance management framework; - reviewing the balance sheet riskmanagement framework and strategies; - overseeing the group's credit policies; - assessing operational risks; -reviewing business risk management; - reviewing country lines of credit; and - reviewing the liquidity policies ofthe group. The RMC is responsible for the review of risk in all aspects of the business. It is responsible foroverseeing, monitoring and reviewing the group's risk management principles and policies, strategies,processes and controls including credit, market, balance sheet, operational risk and compliance. It may approvecredit transactions and other matters beyond the approval discretion of executive management. The descriptivestatistics of the independent variables are presented in Panel A, Table III [Figure omitted. See Article Image.].The average percentage of non-executive directors on the board is 75.4 per cent with a minimum and maximumpercentage of 33.3 and 100 per cent, respectively. The minimum number of directors on the board is 3 and themaximum board size is 13. Among the 200 sample companies, the number of business segment ranges from 1to 9. The average proportion of assets that are in the form of accounts receivable and inventory and long-termdebt are 0.372 and 0.186, respectively. The average total assets figure for the sample is $8.026 billion, with aminimum and maximum asset value of $0.016 and $302.327 billion, respectively. Fifteen per cent of the sampleare financial companies ( n =30) and 85 per cent are from other industries (n =170). Seventy-five per cent of thecompanies have an independent chairman (n =150) and 89 per cent of the companies are audited by one of theBig Four audit firms (n =178). The descriptive statistics of the independent variables are presented in Panel B ofTable III [Figure omitted. See Article Image.]. Table IV [Figure omitted. See Article Image.] provides theSpearman's ρ correlations between the two independent variables and the dependent variables. The variablesare not too highly correlated suggesting that multicollinearity is not a problem. Logistic regression The Hosmer-Lemeshow goodness-of-fit statistic was interpreted to determine whether the logistic regression modelsreasonably approximate the behaviour of the data ([20] Cox, 1970; [28] Hosmer and Lemeshow, 1989; [30]Kleinbaum et al. , 1982). If the statistic is >0.05, the null hypothesis of no difference between the observed andmodel predicted values is accepted, signifying that the model's estimates fit the data at an acceptable level.Panel A of Table V [Figure omitted. See Article Image.] provides the results of the RMC existence logisticregression model. The Hosmer-Lemeshow goodness-of-fit statistic is non-significant (χ2 =9.393, df = 8, p >0.05)suggesting that the overall model fit is acceptable. In addition, the model yielded a Nagelkerke R2 of 13.3 percent indicating that the independent variables make a contribution to the variance in the existence of a RMC.

08 March 2013 Page 102 of 145 ProQuest

According to the Wald statistic, an independent chairman and a larger board size significantly predict theexistence of a RMC (z =2.938, p <0.05; z =2.706, p =0.050, respectively). Therefore, H2(a) and H3(a) aresupported. In addition, the existence of a RMC is marginally associated with company size (z =1.620, p =0.10).However, the remainder of the set of (a) hypotheses are not supported. The results of the SRMC existencelogistic regression model are also shown in Panel B of Table V [Figure omitted. See Article Image.]. The overallmodel fits the data (χ2 =4.034, df = 8, p =0.854). The Nagelkerke R2 is 38.6 per cent which indicates that theindependent variables in the model make a significant contribution to the existence of a separate RMC. Theexistence of a separate RMC is significantly associated with size of the board (z =7.756, p <0.050) andmoderately associated with an independent chairman (z =1.919, p <0.10), the business segment (z =3.018, p<0.10) company risk factors (z =2.388, p <0.10) and company size (z =2.242, p <0.10), showing support forH2(b) , H3(b) and H7(b) . However, contrary to the predicted sign, the coefficient for the number of businesssegments signifies a negative association with the existence of a separate RMC. Discussion of results The twomain objectives of the study are to: identify factors associated with the establishment RMCs; and those relatedwith the set-up of a separate RMC as opposed to one combined with the audit committee. Guided by an agencytheory perspective, the study namely investigated the role played by agency-cost related factors such as boardindependence, CEO duality and board size, as well as company-related factors such as type of industry,organisational complexity and leverage on the establishment and nature of RMCs. The results, based on logisticregression analyses, indicate that several board factors and organisational characteristics are associated withthe existence of a RMC and the type of RMC. Of the first three sets of hypotheses of this study, two of the setswere supported (i.e. H2 and H3 ). The results reveal a significant and positive association between two types ofboard characteristics (an independent chairman on the board and board size), and the existence of a RMC.Both board characteristics were also found to be positively associated with the establishment of a SRMC. Thesupport of H2(a) and H2(b) suggest that an independent chairman is likely to reduce agency costs byestablishing control mechanisms, such as a RMC. The results also indicate that having an independentchairman promotes not only having a RMC but also a separate and distinct RMC. H3(a) and H3(b) are bothsupported suggesting larger boards are able to offer the resources needed to maintain and operate boardcommittees. This finding is in line with the [56] ASX's (2003) Principles of Good Corporate Governance and theBest Practice Recommendations , where larger boards are advised to set-up support mechanisms such as anRMC. The results also indicate that organisations would be willing to not only set up RMCs, but also separateRMCs when they have sufficient board resources. Further, this result is consistent with [39] Piot's (2004) studywhere a study of 285 listed companies in France revealed the presence of an audit committee to be positivelycorrelated with board size. It is likely that having a large sized board provides greater resources through higherprobability of members with requisite risk expertise being on the board so as to form sub-committees. The resultalso somewhat supports the notion based on signalling theory that when agency costs (e.g. board size) is high,separate RMCs are established and disclosed so as to flag to the stakeholders the intentions of the board toinstall high quality monitory mechanisms. However, H1(a) and H1(b) are not supported, thus indicating thatthere is no significant association between the existence of a RMC and the proportion of non-executivedirectors. This finding is also similar to [39] Piot's (2004) study where there was no significant found betweenthe proportion of non-executive directors and the existence of an audit committee. A possible explanation forthis finding may relate to the level of non-executive directors membership and the concept of boardindependence, in that although a board may have a high level of non-executive directors, it may not directlyrelate to the board being independent. In this study, non-executive directors are identified as being not involvedin day-to-day decisions, but their independence is not fully clear in terms of their financial link with theorganisation. It is possible some directors may be termed non-executives, but may not meet the criteria offinancial independent, e.g. a non-executive director may still hold more than 5 per cent of the voting shares ofthe company[3] . The results of the present study provide very little support for the effect of the type of external

08 March 2013 Page 103 of 145 ProQuest

auditor on RMC establishment and its nature. H4(a) and H4(b) are not supported, indicating that the existenceof a RMC is not significantly correlated with the type of external auditors (Big Four firms versus non-Big Fourfirms). One explanation for this finding is that only a small number of firms have a non-Big Four auditor, i.e. 22(11 per cent) of the cases, and of these many are medium tiered and thus may have similar preferencestowards RMCs. The results for the next four sets of hypotheses (H5 , H6 , H7 and H8 ) which proposesignificant relationships between the four company characteristics: industry type; organisational complexity;financial reporting risk and leverage; and the establishment and type of RMC appear mixed. More specifically,none of the four company characteristics are found to have a significant association with the existence of anRMC, thus H5(a) , H6(a) , H7(a) and H8(a) are not supported. On the other hand, in relation to the type of RMC,two of the four hypotheses are supported. First, there is marginal support for H7(b) (p <0.10), indicating anassociation between the existence of a separate RMC and financial reporting risks. It appears that theprobability of the establishment of a separate RMC increases with the proportion of accounts receivable andinventory to total assets. The findings thus support the argument that as the risk of financial reporting increases,organisations may view greater value in establishing board committees such as a RMC that is solely dedicatedto risk oversight. Second, H6(b) is supported but the direction of the relationship is not as proposed. Thefindings suggest that as complexity of an organisation based on the number of business segments increases,the probability of a firm establishing a combined committee increases, rather than a separate committee. Apossible reason for this finding is that a combined committee may offer certain advantages over a separateRMC. In particular, with increasing complexity arising from coordinating a larger number of business segments,the involvement of audit committees jointly in overseeing risk management may be more advantageous. Furtherstudy in this regard however is needed to better understand the specific advantages and disadvantages ofseparate and combined RMCs in such situations. The size of the organisation is also found to be significantlyassociated with the existence of RMCs, but not with the nature of the RMC. The results indicate that theprobability of an organisation establishing a RMC increases with its size but size has no significant impact onwhether a separate or combined committee is set-up. The findings thus support prior studies that have arguedlarger companies are expected to have higher agency costs, and thus will be pressured to adopt better internalrisk monitoring ([13] Carcello et al. , 2005; [26] Goodwin and Kent, 2006). Interestingly, industry type in terms offinance versus non-finance firms was not found to be a significant predictor of the existence of RMCs nor thetype of RMC. A possible reason for this result is that with the escalating levels of complexity and volatility inbusiness environments and increasing expectations for higher quality governance mechanisms withinorganisations, both non-finance and finance firms may be equally pressured to set up RMCs. For instance, thepush for enterprise-wide risk management systems such as that advocated by [19] COSO (2004) have gainedmuch popularity as a means to align risks from an organisation-wide perspective, internal controls and the goalsof the organisation. Conclusions, limitations of the study and suggestions for future research With the increasinggrowth in RMC establishment and the disclosure of such board committees in the annual reports of companies,the need to better understand the factors that motivate their existence is clearly important. The present studyprovides some illumination towards the factors that affect RMC establishment and structure, thus contributing tothe literature in several ways. First, the findings of this study provide systematic empirical evidence on RMCs inAustralian companies where the adoption of RMCs is still voluntary. The study clearly identifies that boardleadership and board size are key elements in the establishment of RMCs. Secondly, the present findingsclearly highlight the significant roles played by board leadership and board size, as well as financial reportingrisks and organisational complexity in the decision to set up and disclose separate RMCs. As such, the findingsof this study should be of interest to institutional policymakers as the quality of risk management supportmechanisms are likely to be related to other governance factors such as board leadership and size. There arehowever several limitations of the study. First, the sample of the study is only 200, and only a small number ofthe companies (22 out of 200 companies) were found to have a separate RMC. Furthermore, the study involved

08 March 2013 Page 104 of 145 ProQuest

only the top 300 Australian firms listed on ASX. There are presently around 2,000 public listed companies inAustralia. Thus, the generalisability of the results is limited. Future studies may extend their investigation to alarger sample. The second limitation of the study is that the data were collected from companies' annualreports. Thus, only firms that have reported having a RMC in the annual reports have been included in thestudy. It is possible that companies may use other structures for managing their risks and not call themselves aRMC. Unfortunately, there is limited information about companies' risk management structures in annualreports. Therefore, future studies may review the existence of alternate structures, and other research methodssuch as a questionnaire survey or interviews of board representatives may help to better understand the use ofsuch alternate risk management structures in an organisation. The third limitation of the study is that thevariables such as business segment and proportion of non-executive directors in regression models may not begood proxies for board independence level and company complexity factors that are measured in the study,respectively. For example, as [26] Goodwin and Kent (2006, p. 96) argued in their study that "the number ofbusiness segments may not reflect the true complexity of the firm, while the proportion of non-executivedirectors on the board may not be a sound measure of independence". The fourth limitation is the assumptionthat a combined RMC is less efficient than a separate committee. Future research needs more empiricalevidence to substantiate this assumption. As an added suggestion for future studies, alternative researchmethods such as interviews may help to further explain the reasons why companies choose to use RMCs. In-depth interviews may also help better assess the motivation for establishing RMC in an organisation. Inparticular, signalling theory suggests that organisations may undertake such committees to merely flagmanagement's commitment to corporate governance. Finally, exploration of the complex interactions betweenthe various governance mechanisms of RMCs, audit committees, external audit and internal audit is alsoneeded. The role of RMC in risk management function is relatively unexplored and thus presents a fruitfulavenue for future research. In conclusion, this study indicates board resources and leadership are key factors inthe set-up and disclosure of RMCs. Given the growing demands on risk management, further studies on suchgovernance support mechanisms are vital. No doubt, there is mounting pressure on boards of directors toensure that effective corporate governance in an increasingly dynamic and challenging business environment.Footnote 1. Another explanation for the establishment of a board committee is that it may provide liabilityadvantages to some of the directors ([27] Harrison, 1987). For example, the American Bar Association hasemphasised the liability advantages of using monitoring subcommittees whereby the liabilities of directors whoare not members of a given committee can be construed to be more limited in comparison to those on thecommittee. 2. Big Four audit firms in this study are KPMG, PWC, Ernst and Young, and Deloitte and Touche. 3. According to [56] ASX (2003) listing rules, an independent director may be construed as one who:* holds lessthan 5 per cent of the voting shares of the company;* has not within the last three years been employed in anexecutive capacity by the company; and* is free from any interest and any business or other relationship whichcould, or could reasonably be perceived to, materially interfere with the director's ability to act in the bestinterests of the company. References 1. Adams, M.B. (1994), "Agency theory and the internal audit", Managerial Auditing Journal, Vol. 9No. 8, pp. 8-12. 2. Adams, M. (1997), "Determinants of audit committee formation in the life insurance industry: New Zealandevidence", Journal of Business Research, Vol. 38, pp. 123-9. 3. Adelaide Bank Limited (2005), Adelaide Bank Limited Annual Report 2005. 4. Alles, M.G., Datar, S.M. and Friedland, J.H. (2005), "Governance-linked D&O coverage: leveraging the auditcommittee to manage governance risk", International Journal of Disclose and Governance, Vol. 2 No. 2, pp.114-29. 5. Australian Stock Exchange (2007), Corporate Governance Principles and Recommendations, 2nd ed., ASX

08 March 2013 Page 105 of 145 ProQuest

Corporate Governance Council, Sydney. 6. Baiman, S. (1990), "Agency research in managerial accounting: a second look", Accounting, Organizationsand Society, Vol. 15 No. 4, pp. 341-71. 7. Beasley, M.S., Carcello, J.V. and Hermanson, D.R. (1999), Fraudulent Financial Reporting: 1987-1997, AnAnalysis of US Public Companies, COSO, New York, NY. 10. Bradbury, M.E. (1990), "The incentives for voluntary audit committee formation", Journal of Accounting andPublic Policy, Vol. 9 No. 1, p. 19. 11. Burlando, A.J. (1990), "The 1990s: the decade of risk management", Risk Management, Vol. 37 No. 3, p.50. 12. Cadbury Committee (1992), Report of the Committee on the Financial Aspects of Corporate Governance,Gee &Co., London. 13. Carcello, J.V., Hermanson, D.R. and Raghunandan, K. (2005), "Factors associated with US publiccompanies' investment in internal auditing", Accounting Horizons, Vol. 19 No. 2, pp. 69-84. 14. Carson, E. (2002), "Factors associated with the development of board sub-committees", CorporateGovernance: An International Review, Vol. 10 No. 1, pp. 4-18. 15. Certo, S.T. (2003), "Influencing initial public offering investors with prestige: signaling with board structures",Academy of Management Review, Vol. 29 No. 3, pp. 432-46. 16. Chau, G. and Leung, P. (2006), "The impact of board composition and family ownership on audit committeeformation: evidence from Hong Kong", Journal of International Accounting, Auditing and Taxation, Vol. 15, p. 1. 17. Chen, J.P.C. and Jaggi, B. (2000), "Association between independent non-executive directors, familycontrol and financial disclosures in Hong Kong", Journal of Accounting and Public Policy, Vol. 19 Nos 4-5, pp.285-310. 18. Cohen, J., Krishnamoorthy, G. and Wright, A. (2004), "The corporate governance mosaic and financialreporting quality", Journal of Accounting Literature, Vol. 23, pp. 87-152. 19. (The) Committee of Sponsoring Organisations of the Treadway Commission (2004), Enterprise RiskManagement-Integrated Framework, The Committee of Sponsoring Organisations of the TreadwayCommission, New York, NY. 20. Cox, D.R. (1970), Analysis of Binary Data, Methuen, London. 24. Fields, M.A. and Keys, P.Y. (2003), "The emergence of corporate governance from Wall St. to Main St.:Outside Directors", Board Diversity, Earnings Management, and Managerial Incentives to Bear Risk. TheFinancial Review, Vol. 38, pp. 1-24. 26. Goodwin, J. and Kent, P. (2006), "The use of internal audit by Australian companies", Managerial AuditingJournal, Vol. 21 No. 1, pp. 81-101. 27. Harrison, J.R. (1987), "The strategic use of corporate board committees", California Management Review,Vol. 30 No. 1, p. 109. 28. Hosmer, D.W. and Lemeshow, S. (1989), Applied Logistic Regression, Wiley, New York, NY. 29. Jensen, M.C. and Meckling, W.H. (1976), "Theory of the firm: managerial behavior, agency costs andownership structure", Journal of Financial Economics, Vol. 3 No. 4, pp. 305-60. 30. Kleinbaum, D.G., Kupper, L.L. and Morgenstern, H. (1982), Epidemiologic Research: Principles andQuantitative Methods, Van Nostrand Reinhold, New York, NY. 31. Korosec, B. and Horvat, R. (2005), "Risk reporting in corporate annual reports", Economic and BusinessReview for Central and South-Eastern Europe, Vol. 7 No. 3, p. 217. 32. KPMG (2001), Enterprise Risk Management: An Emerging Model for Building Shareholder Value, KPMG. 33. KPMG (2005), Strategic Risk Management Survey: A Survey of Contemporary Strategic Risk ManagementPractices in Australia and New Zealand, KPMG. 34. Lambert, R.A. (2001), "Contracting theory and accounting", Journal of Accounting Economics, Vol. 32, pp.

08 March 2013 Page 106 of 145 ProQuest

3-87. 35. Matolcsy, Z., Stocks, D. and Wright, A. (2004), "Do independent directors add value?", AustralianAccounting Review, Vol. 14 No. 1, p. 33. 36. Maurer, J.G. (1971), Readings in Organization Theory: Open-System Approaches, Random House, NewYork, NY. 37. Morris, R.D. (1987), "Signalling, agency theory and accounting policy choice", Accounting and BusinessResearch, Vol. 18 No. 69, pp. 47-56. 38. Pincus, K., Rusbarsky, M. and Wong, J. (1989), "Voluntary formation of corporate audit committee amongNAS-DAQ firms", Journal of Accounting and Public Policy, Vol. 8 No. 4, p. 239. 39. Piot, C. (2004), "The existence and independence of audit committee in France", Accounting and BusinessResearch, Vol. 34 No. 3, p. 223. 41. Raghunandan, K., Read, W.J. and Rama, D.V. (2001), "Audit committee composition 'gray directors', andinteraction with internal auditing", Accounting Horizons, Vol. 15 No. 2, pp. 105-18. 42. Roberts, J., McNulty, T. and Stiles, P. (2005), "Beyond agency conceptions of the work of the non-executivedirectors: creating accountability in the boardroom", British Journal of Management, Vol. 16, p. 5. 43. Ruigrok, W., Peck, S., Tacheva, S., Greve, P. and Hu, Y. (2006), "The determinants and effects of boardnomination committees", Journal of Management Governance, Vol. 10, pp. 119-48. 44. Sallivan, L. (2001), "Enterprise wide: building a risk management program from the ground up", RiskManagement, Vol. 48 No. 12, p. 24. 45. Selim, G. and McNamee, D. (1999), "Risk management and internal auditing: what are the essentialbuilding blocks for a successful paradigm change?", Internal Journal of Audinting, Vol. 3, p. 147. 46. Soltani, B. (2005), Factors Affecting Corporate Governance and Audit Committees in Selected Countries(France, Germany, the Netherlands, the United Kingdom, and the United States), The Institute of InternalAuditors Research Foundation, Altamonte Springs, FL, p. 192. 49. Subramaniam, N. (2006), "Agency theory and accounting research: an overview of some conceptual andempirical issues", in Hoque, Z. (Ed.), Methodological Issues in Accounting Research: Theories and Methods,Spiramus, London, pp. 55-82. 52. (The) Institute of Internal Auditors Research Foundation (2003), Research Opportunities in Internal Auditing,The Institute of Internal Auditors Research Foundation, Altamonte Springs, FL. 53. Turpin, R.A. and DeZoort, F.T. (1998), "Characteristics of firms that include an audit committee report intheir annual report", International Journal of Auditing, Vol. 2, pp. 35-48. 54. Uzun, H., Szewczky, S.H. and Varma, R. (2004), "Board composition and corporate Fraud", FinancialAnalysts Journal, Vol. 60, pp. 33-43. 55. Wallace, W. and Kreutzfeldt, R. (1991), "Distinctive characteristics of entities with and internal auditdepartment and the association of the quality of such departments with errors", Contemporary AccountingResearch, Vol. 7 No. 2, pp. 485-512. 56. Australian Stock Exchange (2003), Principles of Good Corporate Governance and Best PracticeRecommendations, ASX Corporate Governance Council, Sydney. 57. Benz, M. and Frey, B.S. (2007), "Corporate governance: what can we learn from public governance?",Academy of Management Review, Vol. 32 No. 1, pp. 92-104. 58. Daily, C., Dalton, D. and Rajagopalan, N. (2003), "Governance through ownership: centuries of practice,decades of research", Academy of Management Journal, Vol. 46, pp. 151-9. 59. Meyer, J.W. and Rowan, B. (1977), "Institutionalized organizations: formal structure as myth andceremony", American Journal of Sociology, Vol. 83, pp. 340-63. 60. Klein, A. (2002), "Audit committee, board of director characteristics, and earnings management", Journal ofAccounting and Economics, Vol. 33 No. 3, pp. 375-400.

08 March 2013 Page 107 of 145 ProQuest

61. Collier, P.A. (1993), "Audit committees in major UK companies", Managerial Auditing Journal, Vol. 8 No. 3,pp. 25-30. Further Reading 1. Berenson, M.L., Levine, D.M. and Krehbiel, T.C. (2006), Basic Business Statistics: Conceptsand Applications, 10th ed., Pearson Education, Upper Saddle River, NJ. 2. Bies, S.S. (2004), "Current issues in corporate governance", Vital Speeches of the Day, Vol. 70 No. 14, p.424. 3. Crawford, M. and Stein, W. (2004), "Risk management in UK local authorities: the effectiveness of currentguidance and practice", The International Journal of Public Sector Management, Vol. 17 Nos 6/7, p. 498. 4. DeZoort, F.T., Hermanson, D.R., Archambeault, D.S. and Reed, S.A. (2002), "Audit committee effectiveness:a synthesis of the empirical audit committee literature", Journal of Accounting Literature, Vol. 21, pp. 38-75. 5. Duggan, O. (2006), "Enterprise risk management the challenge for the public sector", Accountancy Ireland,Vol. 38 No. 4, p. 25. 6. Francis, G. (2004), Introduction to SPSS for Windows: Versions 12.0 and 11.0 - With Notes for Studentware,Pearson Education, Frenchs Forest. 7. PricewaterhouseCooper (2004), Strengthening the Business: Optimizing the Benefits of Enterprise-wide RiskManagement for Audit Committee, PricewaterhouseCooper. 8. Spira, L.F. and Bender, R. (2004), "Compare and contrast: perspectives on board committees", paperpresented at the 6th International Conference on Corporate Governance and Direction, The HenleyManagement College, October. 9. Spira, L.F. and Page, M. (2003), "Risk management: the reinvention of internal control and the changing roleof internal audit", Accounting, Auditing and Accountability Journal, Vol. 16 No. 4, pp. 640-61. 10. Tabachnick, B.G. and Fidell, L.S. (2001), Using Multivariate Statistics, Allyn and Bacon, Boston, MA. 11. The Institute of Chartered Accountants in England &Wales (1999), Internal Control: Guidance for Directorson the Combined Code, The Institute of Chartered Accountants in England &Wales, London. Appendix Corresponding author Nava Subramaniam can be contacted at: nava.subramaniam@deakin.edu.auAuthorAffiliation Nava Subramaniam, Department of Accounting, Economics and Finance, Deakin University,Burwood, Australia Lisa McManus, Department of Accounting, Finance and Economics, Griffith BusinessSchool, Griffith University, Gold Coast, Australia Jiani Zhang, Griffith University, Gold Coast, AustraliaIllustration Equation 1 Table I: Sample selection process Table II: Sample companies industry segment TableIII: Descriptive statistics Table IV: Correlation matrices Table V: Logistic regression analysis of RMC existence (n =200) and SRMC existence (n =88) Subject: Studies; Corporate governance; Risk management; Executive committees; Organization development Location: Australia Classification: 3300: Risk management, 2110: Boards of directors, 9130: Experimental/theoretical, 9179:Asia&the Pacific Publication title: Managerial Auditing Journal Volume: 24 Issue: 4 Pages: 316-339 Publication year: 2009 Publication date: 2009 Year: 2009

08 March 2013 Page 108 of 145 ProQuest

Publisher: Emerald Group Publishing, Limited Place of publication: Bradford Country of publication: United Kingdom Journal subject: Business And Economics--Management ISSN: 02686902 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Tables;Equations;References DOI: http://dx.doi.org/10.1108/02686900910948170 ProQuest document ID: 274547062 Document URL: http://search.proquest.com/docview/274547062?accountid=86413 Copyright: Copyright Emerald Group Publishing Limited 2009 Last updated: 2010-06-14 Database: Accounting&Tax

_______________________________________________________________ Document 9 of 10 A Methodology for Evaluating the Cost-Effectiveness of Alternative Management Tools in Public-Sector Institutions: An Application to Public Education Author: Mensah, Yaw M; Schoderbek, Michael P; Werner, Robert H Publication info: Journal of Management Accounting Research 21 (2009): 203-239.ProQuest document link Abstract: The shift toward performance budgeting and outcome measures for public-sector institutions in recentdecades has created a need to formally link inputs consumed and outcomes achieved. Given the inherentproblems of cost accounting systems in public-sector institutions, we propose a statistical approach to identifythe most cost-effective management tools that also recognize the endogeneity between costs and outcomes.The model developed allows for the examination of possible trade-offs that can be exercised by public-sectorinstitutions facing multiple stakeholders with conflicting objectives. Using public schools in New Jersey and a setof variables identified from the education economics literature, we estimate cost and outcome functions todemonstrate empirically the choices made by school district superintendents that trade off the interests ofvarious stakeholders, while seeking to meet the core objectives of the institutions. Our empirical results provideinsight on the variables controllable by the superintendents that appear to be used inefficiently, or are subject toinstitutional constraints that limit the flexibility in input choice assumed by the proposed method. From amanagement accounting standpoint, the identification of such variables narrows the areas to be focused on inthe search for improvements in performance. [PUBLICATION ABSTRACT] Full Text: Headnote ABSTRACT: The shift toward performance budgeting and outcome measures for public-sector institutions in recent decades has created a need to formally link inputs consumed and outcomesachieved. Given the inherent problems of cost accounting systems in public-sector institutions, we propose a

08 March 2013 Page 109 of 145 ProQuest

statistical approach to identify the most cost-effective management tools that also recognize the endogeneitybetween costs and outcomes. The model developed allows for the examination of possible trade-offs that canbe exercised by public-sector institutions facing multiple stakeholders with conflicting objectives. Using publicschools in New Jersey and a set of variables identified from the education economics literature, we estimatecost and outcome functions to demonstrate empirically the choices made by school district superintendents thattrade off the interests of various stakeholders, while seeking to meet the core objectives of the institutions. Ourempirical results provide insight on the variables controllable by the superintendents that appear to be usedinefficiently, or are subject to institutional constraints that limit the flexibility in input choice assumed by theproposed method. From a management accounting standpoint, the identification of such variables narrows theareas to be focused on in the search for improvements in performance. (ProQuest: ... denotes formula omitted.)INTRODUCTION Over the past two decades, considerable attention has been focused on performancemeasurement and reporting for public-sector institutions. Much of this effort has been directed at moving frominput-based control budgets to output- and outcomebased performance budgets. At the federal governmentlevel, the Government Performance and Results Act of 1993 initiated the most recent and comprehensiveattempt at reforming the financial planning, management, and control system (Kravchuk and Schack 1996;McNab and Melese 2003). Similar contemporaneous initiatives have been undertaken at the state and localgovernment levels in states such as Texas, Oregon, Minnesota, Virginia, and Florida (Broom 1995).1 Thesedevelopments parallel the adoption by the Governmental Accounting Standards Board (GASB) of ConceptsStatement No. 2, Service Efforts and Accomplishments Reporting in 1994 (GASB 1994), and subsequent GASBStandards No. 33 (GASB 1998) and No. 34 (GASB 1999). Yet, despite the comparative advantages ofaccounting researchers in knowledge of the measurement process (Kinney 2001), accounting researchcontributions to the public debate on how to increase the cost-effectiveness of public-sector institutions havebeen sparse, given the importance of that sector in national and state economies. The move from input-orientedcontrol budget systems to outcome-oriented performance budgets is motivated by several considerations, themost important of which are making public-sector institutions more efficient, more effective, and holding publicmanagers accountable for policy outcomes. Control budgets can then be loosened so that public managers canmore effectively employ their managerial discretion in the pursuit of desired outcomes (Gianakis 2002). Forpublic managers to be more effective in the discharge of their duties, a systematic method of linking resourceinputs to final outcomes on which management performance can be evaluated is needed. Although activity-based costing (ABC) has been touted in the literature as offering a means to link inputs used to the outputs ofpublic-sector institutions (Mosso 1999; Brown et al. 1999; Mullins and Zorn 1999), the relationship of their inputsto eventual outcomes as perceived by the public cannot be defined in an a priori manner (McNab and Melese2003). Establishing the ex post relationship between inputs and measured outcomes via statistical means,however, is feasible. Thus, assuming the temporal stability of the relationship between inputs and the finaloutcomes, public managers can develop a knowledge base of managerial tools that are cost-effective inachieving the desired outcomes. At the same time, certain aspects of public-sector institutions are pertinent inunderstanding the constraints in which they operate, and consequently, the choices they can make among thefeasible set of inputs and outputs. Hansmann (1996) has presented a theoretical framework for understandingorganizational structures as one where the evolved structure minimizes the combined costs of ownership and ofcontracting. Hansmann (1996, 228- 230) argues that non-profit institutions arise where the primary beneficiariesof the services provided are frequently in a poor position to determine, without significant effort or cost, thequality and/or quantity of services provided. As a result, assigning ownership of the institution to any other groupof stakeholders would create severe agency problems (i.e., incentive and opportunity for the owners to exploitthe beneficiaries). In addition, the cost to the beneficiaries to exercise effective control would be quite largerelative to the value of their transactions with the institution. Thus, creating a non-profit institution allows themanagers to operate the institution as fiduciary trustees for the intended beneficiaries. However, because there

08 March 2013 Page 110 of 145 ProQuest

are multiple stakeholders, the managers must also be sensitive to the concerns of these other parties. Thus, thechoices they can make in terms of the selection of inputs and managerial tools reflect multiple influences. Thepurpose of this paper is to present empirical evidence of the trade-offs available to public-sector managersfacing multiple objectives and conflicting demands. Because of the difficulty of measuring performance, public-sector institutions do not have high levels of bonus-based compensation. However, as noted by Rose-Ackerman (1996), the values, motives, and incentives of managers of non-profit institutions may be lessmercenary than those of for-profit organizations. Thus, even though incentive contracting can align the interestsof the stakeholders and the public-sector managers, the incentive contracts must include non-pecuniary as wellas pecuniary returns (Baber et al. 2002). By identifying some performance measures that can be chosen bypublic-sector managers under specific circumstances, and further differentiating between those contributing tocost efficiency and outcome effectiveness while simultaneously controlling for endogeneity among them, thisstudy contributes to an understanding of how the effectiveness of control tools available to public-sectormanagers can be measured in multiple-objective situations. The method we develop in this paper is applied to asample of 521 New Jersey school districts. Two models are estimated simultaneously: a cost function (with per-pupil expenditures as the dependent variable) and an outcome function (with student achievement scores asthe dependent variable). The two dependent variables are treated as jointly endogenous. Explanatory variablesinclude those controllable by the school district superintendent (e.g., student-faculty ratio) and those that areuncontrollable (e.g., percent students receiving federal meal aid). The focus of analysis is on the controllableexplanatory variables included in both the cost and outcome equations. By examining the statistical significanceand signs (positive or negative) of the estimated coefficients, this paper develops a measure of relative cost-effectiveness of the controllable variables. Consider the case where a particular variable is significant andpositive in the outcome equation, but lacks significance in the cost equation. The positive coefficient suggeststhat a marginal increase in its use is associated with a marginal increase in test outcomes, but has no marginal(off-setting) effect on total costs. On the other hand, variables that are significant with positive signs in both thecost equation and the outcome equation indicate a trade-off where increased levels of that variable will increaseboth test scores and total costs. To evaluate these types of trade-offs, we compute a trade-off ratio from theestimated coefficients as a guide to school district administrators. The paper is organized as follows. The nextsection presents the theoretical background and our methodology to evaluate the cost-effectiveness of thechoices made by publicsector managers. The following section is a review of the educational economicsliterature on the determinants of school spending and achievement scores from which our choice of variablesare based. Next, we provide the institutional setting of New Jersey's public school system and standardizedtests. The following section includes specification of our cost and outcome functions. Afterwards, we present ourempirical results, including an analysis of the relative cost-effectiveness of alternative management tools andcontribution of our proposed methodology. A sensitivity analysis of our main results is provided in the nextsection, and the conclusions and recommendations follow. THEORETICAL DEVELOPMENT AND PROPOSEDMETHODOLOGY Theoretical Development Theories of the economic foundation for non-profit institutions reston the difficulty of measuring organizational performance in the settings in which they exist. For example,Hansmann (1996) argues that ''public good'' institutions arise in contexts where multidimensional criteria areneeded to evaluate performance because of multiple stakeholders. As an example of such ''public good''institutions, consider the multiple stakeholders in the public education system: (1) the external stakeholders likethe state legislature, state and local government officials, and taxpayers, and (2) the internal or beneficiarystakeholders like parents, school district administrators, and teachers. The state legislature has an interest inachieving high reported levels of educational outcomes subject to financing constraints. State and localgovernment administrators are similarly concerned with outcomes, but the budgetary constraints are even morepressing since they must deal with local taxpayers and the specific financing of schools through such fundingmechanisms as local property taxes. Taxpayers (whether through income taxes, sales taxes, or local property

08 March 2013 Page 111 of 145 ProQuest

taxes) are very much focused on the tax burden imposed by the school system and thus on the efficiency withwhich the public schools are run. In contrast to these external stakeholders, internal stakeholders such asparents are more concerned about the quality of the educational outcomes, and much less focused on thebudgetary aspects. Teachers, on the other hand, are expected to be equally concerned about their personalremuneration, the quality of students, the instructional resources provided, and the ultimate educationaloutcomes as measured by objective measures such as statewide test scores. School district administratorstherefore form the apex of these convergent forces where a balance must be maintained between the externaland internal stakeholders. It is worthwhile to note that the very key role played by district superintendents inmoving urban school districts forward in terms of measurable achievements required under the federal No ChildLeft Behind Act of 2001 (U.S. Department of Education 2002) (under the auspices of Harvard University's PublicEducation Leadership Project) has been observed by Childress et al. (2006). Three other pertinent aspects ofthe non-profit (including the public sector) environment are: (1) the lack of well-defined production functions thattie inputs with outcomes; (2) the frequent inability to measure outcomes quantitatively or over a short-termhorizon; and (3) the tendency of some stakeholders to focus on the more easily measured quantitativeoutcomes with more limited attention to outcome quality.2 Our objective in this paper is to contribute to the firstarea by presenting a methodology that ties inputs with outcomes (or reasonable proxies for difficult-to-measureoutcomes.) Our objective is thus similar to Dopuch and Gupta (1997), except that we seek to exploit the specificpeculiarity of publicsector and non-profit institutions-the lack of a well-defined production or associated costfunction. Thus, in contrast to the single equation approach adopted by Dopuch and Gupta (1997), we use a two-equation approach. In addition, instead of their stochastic frontier estimation approach, we examine the use ofboth seemingly unrelated regressions and simultaneous equations approaches. Our approach is thus focusedon identifying the average behavior of public-sector administrators engaged in actions designed to satisfymultiple constituencies, rather than an approach that assumes that such administrators are necessarilyconcerned with cost-minimization or output-maximization.3 Proposed Methodology For public-sector institutions,a key resource constraint that managers face is the level of funds available for expenditure purposes. Given thatthe funds available are typically appropriated by legislatures and the services provided are frequently either freeor at less than full cost to the users, the effective management of expenditures is of paramount importance. Atthe same time, the funding agency or legislature, as well as taxpayers, are well-attuned to perceptions of thequality of services delivered by the institution. These countervailing forces suggest that managers of public-sector institutions must necessarily be concerned with their expenditure levels and the outcomes they deliver totheir stakeholders. Managers of public-sector institutions face a multi-dimensional objective function, but asnoted above, the two key variables are the cost of providing the services and the outcomes they are able toachieve. We propose a methodology that ties together a possible set of management control tools to assist inmanagement's decision that also deals with the potential endogeneity inherent in this setting. Consider a simpletwo-equation system given by: Y = a + (a X + .... + a X +a Y) + (= Z + .... = Z ), (1) 1 0 1 1 n n n+1 2 1 1 m mand: Y = c + (b W + .... + b W +b Y) + (+ Z + .... + Z ), (2) 2 0 1 1 s s s+1 1 1 1 m m where: Y1 = cost; Y2 =outcome; X1 ... Xn = exogenous variables that are theorized to drive cost levels; W1 ... Ws = exogenousvariables that are theorized to determine outcome levels; and Z1 ... Zm = management decision variables thatmay affect both costs and outcome levels. In Equations (1) and (2), the variables denoted X and W could be thesame variables, but they differ from the Z variables in that they are presumed to be truly exogenous in that theyare not subject to management or administrative control. So X and W could be input prices (where the institutionis a price-taker for that input), socio-economic variables that drive either costs or demand for the public servicesprovided, or other exogenous variables. As long as the set of X variables are not the same as the W variables,the simultaneity implied by the two system of equations can be resolved since the variables unique in eitherequation would serve as the instruments for the dependent variable in that equation. The two-system ofequations can be estimated using ordinary least-squares (OLS) on the reduced form equations, jointly using

08 March 2013 Page 112 of 145 ProQuest

seemingly unrelated regressions (SUR), or simultaneously using either two-stage least-squares (2SLS) orthree-stage least-squares (3SLS). The endogeneity presumption embodied in the system of equations can beevaluated in a three-step process: (1) a test of the fit of the instruments to deal with the weak instrumentsproblem (Staiger and Stock 1997; Stock et al. 2002); (2) a test of over-identified restrictions if either of the twoequations is overidentified (Bound et al. 1995; Hahn and Hausman 2003); and (3) a Hausman (1978) test of thenull hypothesis that no endogeneity problem exists (1978). The three-step process described above is astandard econometric technique for estimating simultaneous equations. The innovation in this paper is inrelating the coefficients in the cost equation to the coefficients in the outcome equation in a systematic mannerto derive the cost-effectiveness of the various controllable variables. In the standard neoclassical productionsetting, the normal approach is to estimate either a cost function (where cost is the dependent variable) or aproduction function (where output is the dependent variable). Estimating both functions together in a system ofequations is never done in practice to our knowledge because, in such setting, the cost function is the dual ofthe production function, so no new information can be gleaned from such a system of equations in which boththe cost and production functions are estimated together (see, for example, Shephard 1970; Varian 1992, 83;Mundlak 1996). What is unique about the public-sector setting is that the lack of a well-defined productiontechnology defies the closed production system implied in neoclassical production settings where such a uniquetransformation function between the cost and production functions exist.4 Thus, although cost and productionfunctions have been estimated previously in the educational economics literature to describe the public schooleducation system, this paper is the first to estimate both jointly and to explore the resulting coefficients toprovide insight into the cost-effectiveness of the controllable variables. Assuming the econometric problems inestimating Equations (1) and (2) are resolved, the relationship between =I and +> provides the primary focus foranalysis. Note that Equation (1) is a cost function, so a positive sign for =I implies that a factor is cost-increasing. At the same time, Equation (2) is an outcome function, so that a positive sign for +> indicates thatvariable is effective in increasing outcome levels. From a management standpoint, the most interesting insightsare gained by comparing the statistical significance of variables across the cost and outcome equations.Variables that are statistically significant with positive signs in both the cost and test outcome equations signifythat the tools, though effective, are also costly. Thus, a trade-off is involved between cost and outcomes. Tohelp in identifying such variables, the following labels will be used in the empirical section of the paper: BC =variable that is beneficial to costs, but has no effect on outcomes; BO = variable that is beneficial to outcomes,but has no effect on costs; DB = variable that is doubly beneficial (i.e., beneficial effects on both costs andoutcomes); DN = variable that is doubly negative (i.e., negative effect on both costs and outcomes); CI =variable that has an increasing effect on costs, but no effect on outcomes; NO = variable that has a negativeeffect on outcomes, but no effect on costs; and TO = variable with a trade-off effect (i.e., either a beneficial[negative] effect on costs offset by a negative effect on outcomes, or vice versa). In general, variables labeledDB are the most effective, followed by BC and BO variables. TO variables are useful policy tools, but the costsand effects must be weighed carefully. Finally, DN variables in particular (but also CI and NO variables) are tobe avoided if possible, or minimized in their use. For variables given a rating of TO, the ratio ==+i=/==>= (i.e.,the absolute values of the outcome equation coefficient to the cost equation coefficient) is derived and labeledan index of relative cost-effectiveness. This index is given a sign (=) that corresponds to the sign that appearedin the coefficients for the cost and outcome equations (since TO variables have the same sign in bothequations). Higher values for the index with positive signs signify that higher outcome is achieved for eachdollar of spending (with due allowance made for the measurement units of the variable under consideration). Incontrast, indices with negative signs signify that outcomes are sacrificed for each cost dollar injected. That is,when the index value has a negative sign, the variable in question lowers both costs and outcomes. Thus, thehigher the absolute value of the cost relative to the outcome, the more cost-effective that variable is (in thesense of a trade-off). The reverse is true for indices with positive signs. The models presented in Equations (1)

08 March 2013 Page 113 of 145 ProQuest

and (2) also permit the decisions made by public-sector managers to be evaluated in an ex post setting.Specifically, assuming the incentive contracts under which they are operating take into account the differing(and potentially conflicting) objectives of the multiple stakeholders, the model proposed permits the resultingpattern of choices to be analyzed. Such an analysis would permit the costliness of the constraints faced bypublic-sector managers to be better understood. For example, the consistent appearance of CI (cost increasing)and NO (no benefit to outcomes) controllable management tools would imply the existence of institutionalfactors that limit the managers' ability to influence the usage of those tools. Such constraints may exist iflegislatures, donors, or the public, in well-meaning but misguided efforts, impose constraints to the ability of themanagers to change the mix of inputs.5 Research Questions As an empirical demonstration of the proposedmethod, we apply our methodology to the public school system in the State of New Jersey. Public schoolsdominate private schools in relative numbers of students, and education represents a very important public-sector institution. Continuing concerns exist on the cost-effectiveness of public education, with frequent calls forreform of the education system. Thus, the following research questions are investigated using our methodology:(1) What factors are posited to influence the levels of school expenditures and test score outcomes in the publicschools districts? Which of these factors can be deemed controllable by the school district superintendent? (2)Which of these factors are common and which are unique in their influence on the two dependent variables(per-pupil expenditures and student achievement scores)? (3) Given the signs and magnitudes of thecoefficients on the controllable variables in the cost and test outcome equations, which factors are cost-effective, costineffective, or involve a trade-off between costs and benefits. To investigate these questions, thenext section reviews the public education literature to identify the Xi, Wi, and Zi variables specified in ourmodels. Subsequently, the institutional setting of the New Jersey public schools are described, and Equations(1) and (2) are operationalized. PRIOR LITERATURE ON PERFORMANCE MEASUREMENT IN PUBLICSCHOOLS Since the Coleman report (Coleman et al. 1966),6 the literature on performance measurement inpublic schools has largely focused on the effect of school inputs on student achievement scores. This research,centered on the ''Does Money Matter'' debate, is relevant for the selection of variables for our outcome function.However, it should be noted that variables posited to influence test scores (e.g., the teacher-student ratio), moreoften than not, affect school expenditures as well. Thus, most of the variables discussed are relevant for ourcost function as well as our test outcome function. It is this endogeneity between school inputs and testoutcomes that motivates our attempt to develop a methodology to analyze the beneficial trade-offs between thetwo. The predicted influence on costs and outcomes of the variables identified below is discussed later in modelspecification. Hanushek (1986) surveyed research from 38 different articles that utilized an educationalproduction function approach to explain standardized test scores. A set of core explanatory variables wereincluded in these production functions, including the teacher-pupil ratio, teacher education, teacher experience,teacher salary, and expenditures per pupil (Hanushek 1986, 1161). Although Hanushek (1986, 1162) ultimatelyconcluded: ''There appears to be no strong or systematic relationship between school expenditures and studentperformance,''7 combinations of these variables have been widely used in the production functions of nearly allstudies on the school expenditures-test score debate. Thus, these core explanatory variables, with theexception of teacher education,8 are used in our models as controllable factors to explain achievement scores.In addition to these teacher-related factors, more recent studies have also included administrative-relatedfactors in their production functions. Brewer (1996), who uses number of school district personnel by functionalarea to explain test scores, includes separate regressors for district administrators and building administrators.Jaggia and Kelly-Hawke (1999) include per-pupil administration expenditures and total per-pupil expendituresas determinants of test scores, whereas Dee's (2005) model includes instructional expenditures per pupil andnon-instructional expenditures per pupil as explanatory variables. To capture resources devoted toadministration, we include the median administrative salaries and the median years of administrator'sexperience. These two variables, both controllable by the school district, parallel the construction of our two

08 March 2013 Page 114 of 145 ProQuest

teacher-related variables: median faculty salaries and median years of faculty experience. Additionally, toexamine the effect of resources deployed on instruction vis-a`-vis administration, we include the relative costshares for instruction (instructional expenditures/ total expenditures) and administration (administrativeexpenditures/ total expenditures) in our models. It should be noted that, by using relative input cost shares, wemitigate potential multicollinearity problems from deflating, for example, both instructional expenditures andadministrative expenditures by number of students.9 We include two additional controllable factors in our model,the number of students per computer and the student attendance rate. The number of students per computerproxies for the degree to which modern instructional technologies have been introduced into the classroom.Elliott (1998) used teacher survey data to construct a variable for computer usage. Student attendance rate,while not used as a regressor in the studies we surveyed, is identified as an indicator of efficiency by the GASB(1989) in its research report, Service Efforts and Accomplishments: Elementary and Secondary Education.Myriad variables have been used in the literature to control for the effect of socioeconomic factors onachievement scores. Sander (1993) includes family income and three indicator variables (Black, Hispanic, andPoor) in his model to explain scores on the ACT college entrance exams. In their model to explainMassachusetts test scores, Jaggia and Kelly-Hawke (1999) include four variables to capture students' familybackground: percent rental units within the community, percent single mothers, community crime rate, andpercent professionals and managers living in the community. Dee (2005), who explains highschool graduationrates, uses five variables to control for socio-economic priors: percent children at risk, percent children whospeak English ''not well'' or ''not at all,'' percent householders with high-school degrees, percent householderswith some college, and median income in households with children. Elliott (1998), who uses hierarchical linearmodeling to explain math and science achievement scores, includes indicator variables for gender, minoritystatus, and urban school; and continuous variables for socio-economic status (constructed from parent'seducation, occupation, and income), percent students receiving free lunch, percent students in specialeducation programs, and percent students with limited English proficiency. It should be noted that many of thevariables used above capture similar dimensions of socio-economic background. The studies surveyed andvariables utilized in these studies are not intended to be all-inclusive. The selection of socio-economic factorsused in empirical studies on the school input-output relation is often driven by data availability, and this currentstudy is no exception. The variables we used to capture socio-economic factors are discussed in the section inmodel specification. In contrast to the number of studies that examine the effect of school inputs on studentoutcomes, relatively few papers have modeled educational cost functions. Dopuch and Gupta (1997; hereafterD&G), used stochastic frontier estimation (SFE) to examine the efficiency in providing education for 446Missouri school districts. D&G (1997) model total school expenditures as a function of number of students,standardized test scores, percentage of aid dependent or orphan students, percentage of minority students,percentage of students continuing on to higher education, and average income level. OLS regression estimationfound all independent variables to be significant in the hypothesized direction in explaining school expenditures.Like D&G (1997), we specify a cost function to investigate the determinants of perpupil spending. But ourmethodology also employs a production (outcome) function to explain test scores, like the earlier studiessynthesized by Hanushek (1986). We then use the estimated coefficients from these models to analyze themost cost-effective management tools by school district superintendents in providing education in the presenceof multiple stakeholders. INSTITUTIONAL SETTING New Jersey School Districts In New Jersey, the schoolsystem is highly decentralized with multiple levels of controls. At the top of the hierarchy, there is a county-widesuperintendent who oversees all the schools in the county (of which there are 21 in the state). Within the countyare the local school boards and the school district superintendents. Each locality or municipality has its ownschool districts, although some regional school districts (particularly at the high-school level) also exist. Thefocus of interest in this study is the local school district superintendents. The state has five distinct types ofschool districts: elementary school districts (Kindergarten to Grade 6 = ELEM), middle school districts

08 March 2013 Page 115 of 145 ProQuest

(Kindergarten to Grade 8 = MIDD), K-12 school districts (Kindergarten to Grade 12 = K12), high-school districts(Grades 7 to 12 and Grades 9 to 12 = HIGH), and county vocational high-school districts (Grades 8 to 12). Ofthe total 572 school districts in the state, there are 66 ELEM, 223 MIDD, 215 K12, 47 HIGH, and 21 vocationalschool districts. Because the vocational schools are quite different from the other four districts in terms of theiremphasis on occupational as opposed to academic achievement, this study excludes the vocational schooldistricts. The school district superintendent who oversees the districts operations is hired by the Board OfEducation on a multi-year contract, typically three to five years. The superintendent hires additionaladministrative personnel at the district-level, such as Business Administrators, Directors of Curriculum, Directorsof Human Resources, etc., depending on the type of district (e.g., K-12, K-8) and size of the district. Schoolprincipals are also hired on multi-year contracts by the district superintendent, with consultation by the Board ofEducation, and are granted tenure after three years of service. Teachers are hired on renewable annualcontracts by the principals, with consultation by the District Superintendent, and also earn tenure after threeyears. Teachers in New Jersey are members of the American Federation of Teachers (AFT), and operate underunion rules. Thus, seniority is a large component of teachers' salary increases, although principals and schooldistrict superintendents may also reward teachers for superior performance in the classroom. In controllinginstructional costs, district superintendents have a wide array of tools at their disposal, notwithstanding teachertenure and seniority pay scales. According to the New Jersey Department of Education (2004; hereafterNJDOE), approximately one-forth of New Jersey's full-time teachers have zero to three years of teachingexperience and are thus not tenured, nor are part-time teachers that comprise 2.2 percent of the teachingworkforce. There is also wide flexibility in the use of teacher aids (non-tenure-track positions), which numberone-fourth the number of full-time classroom teachers (NJDOE 2004), and educational support servicepersonnel, that include teachers with special skills for which state certification is required (handicap learningspecialists). In controlling costs Superintendents can shift teaching personnel to various assignments,depending on their educational needs and cost objectives. As an example of a cost-reduction mechanism,enrichment programs or early intervention programs can be curtailed or eliminated, and the teaching staff fromthese programs can be re-assigned to full classroom duties to replace retired teachers. Thus, in this studyinstructional costs as well as administrative costs are treated as controllable by the school districts. DistrictFunding Mechanisms Public schools in New Jersey are funded almost exclusively through property taxes leviedat the school district level. The exception to this rule is the school districts designated as Abbott districts. Abbottschool districts have been identified by the New Jersey State Supreme Court (or through subsequent legislativeprocesses) as ''poorer urban districts'' with both ''poverty and educational inadequacy'' so substantial that''poorer disadvantaged students'' cannot compete with ''relatively advantaged students'' (Librera 2003). Forthese 30 Abbott districts (all K-12), the state provides ''Abbott parity aid'' that is calculated to provide them withthe same per-pupil operating budget as would be found in New Jersey's wealthiest school districts. In addition,under the New Jersey State Supreme Court Abbott decisions, school districts that show financial inability but donot meet the Abbott criterion are eligible for Discretionary Education Opportunity Aid.10 StandardizedAchievement Tests and Performance Measurement Public school students are required to sit for three sets ofstandardized tests given in Grade 4, Grade 7, and Grade 11 (shifted to Grade 12 in 2002). Since our interest ison capturing a single measure of outcomes across all four types of school districts, we adopt an outcomemeasure based on Callan and Santerre (1990) and D&G (1997). School district output is defined as the productof quantity of output (total student enrollments = ENROLL) and quality of output (measured by test scoreoutcomes = CTEST).11 The composite test score (CTEST) for the jth school district is constructed as follows:... (3) where: Sij = the number of students in the jth district taking TESTi; TESTi = the standardized state-administered test given in Grades 4, 7, and 11-12; TESTiMAX = the maximum score for TESTi, attained by anydistrict for that year; and Kc = 1 for ELEM and 9-12 HIGH school districts, 2 for MIDD and 8-12 HIGH schooldistricts; and 3 for K-12 school districts. As calculated above, CTEST is the weighted average percentage score

08 March 2013 Page 116 of 145 ProQuest

(relative to the maximum score attained in that year), where the weights are the total number of students whotook that test in that district. Scaling by the maximum score attained in that year was necessary because themaximum scores differ across grade levels, and lack of scaling would have biased the composite score in favorof districts that did comparatively better on the tests with higher maximum scores. The use of standardized testscores as the sole measure of school district performance has been criticized as inadequate since publicschools pursue multiple objectives (Hanushek 1979, 1986). Alternative measures employed in the literature arethe dropout rate, highschool graduation rates, the college placement rate, and post-graduation earnings.However, measures based on eventual outcomes are applicable only to the K-12 and HIGH districts, and notapplicable to this study since we pooled data from all four types of school districts. Moreover, the state of NewJersey itself uses the results of the standardized tests as a summary measure of the public educationalprocess, and the public accountability reports issued for the schools emphasize these results almostexclusively. Thus, as a basis for contracting, this composite test score measure is quite appropriate.EDUCATIONAL COST AND OUTCOME FUNCTIONS Specification of the Cost Function To address theresearch questions above, we operationalize the model in Equation (1) by identifying the factors that arecontrollable and uncontrollable by the school district superintendents. The seven uncontrollable factors in ourcost function include some of the main variables found to affect costs or outcomes in prior studies. Theuncontrollable factors are: (1) weighted district factor grouping index (WDFG); (2) geographic cost index(GEOCEI); (3) an indicator variable for Abbott districts (ABBOT); (4) the proportion of students in the schooldistrict not receiving federal meal aid (HIGHINC); (5) school size (measured as the natural log of the studentpopulation = ENROLL); (6) the number of students enrolled in special education programs (SPED); and (7) thenumber of students in limited English proficiency programs (LEP). The first two variables are from year 2000Census data and are constant for all three years. The New Jersey Department of Education used factoranalysis to construct a District Factor Group index that reflects socio-economic factors linked in the educationalliterature to test score performance.12 Based on the factor scores, the state classified the school districts intoone of eight ordinal rankings, with one denoting the most disadvantaged, and eight the most privileged. We usethe same ordinal rankings for the variable WDFG.13 The second variable, GEOCEI, captures geographicdifferences in the cost of living across the state that impact on the cost of providing education. Both WDFG andGEOCEI are expected to be positively related to costs in our cost equation. Because the Abbott mandatecreated by the New Jersey Supreme Court may create a lesser incentive for cost efficiency, the coefficient onthe dummy variable, ABBOT, is expected to be positive. The variable HIGHINC is computed as the complementof the percentage of the student body from poor families who qualify for federal food aid.14 High values (cappedat 100 percent) reflect a concentration of higher-income families in the district. In general, such school districtswould be expected to spend higher amounts per pupil because they can raise more property taxes to financethe school district.15 However, because the state of New Jersey attempts to equalize spending by directingmore funds to needy districts, the observed spending per pupil may be more reflective of cost efficiency. Totalstudent enrollments (ENROLL) are expected to have the most significant effect on total expenditures. As shownlater, when the cost function is defined in terms of spending per pupil, ENROLL is expected to have a negativesign, reflecting the existence of economies of scale. In addition, to allow for possible diseconomies of scale, thesquare of the natural log of ENROLL is included in the model. This squared term is expected to have a positivesign if diseconomies of scale exist (D&G 1997; Greene 1980). Students enrolled in special education programsor in language proficiency programs can substantially increase the cost of educating students, so positive signsare predicted on the variables SPED and LEP. The controllable factors (i.e., controllable by the districtsuperintendent) included in the model are more often employed in production functions to explain test scores(see ''Prior Literature on Performance Measurement in Public School'') but are also posited in this study toinfluence costs. These controllable factors (grouped into student-related, faculty-related, and administrativefactors) are summarized below: (1) Student-Related Factors: These factors consist of the student-faculty ratio

08 March 2013 Page 117 of 145 ProQuest

(STUFAC), the number of students per computer (STUCOMP), and the student attendance rate (ATTD). (2)Faculty-Related Factors: These factors consist of the median level of faculty salaries (FACSAL), the ratio ofinstructional expenses to total expenditures (CSINS), and the median years of faculty experience (FACEXPR).(3) Administrative-Related Factors: These factors consist of the median level of administrative salaries(ADMSAL), the ratio of administrative costs to total operating expenditures (CSADM), and the median level ofadministrators experience in years (ADMEXPR). STUFAC reflects the average number of students per teacherin the school district. Since higher ratios indicate larger class sizes, the coefficient of STUFAC is expected to benegative in the cost function. Higher ratios of STUCOMP reflect lower degrees of penetration of moderninstructional technology. We expect the coefficient for STUCOMP to be negative in the cost equation, reflectingthe expectation that the introduction of modern instructional technology is costly. The coefficient on ATTD(student class-attendance rate) in the cost equation is expected to be negative, because as school district'sfixed costs are spread over higher volumes, costs per pupil should decrease. FACSAL is predicted to bepositive in the cost function because higher median salaries for the teachers should translate into higher overallper-pupil spending. No prediction is made on the sign of the cost share for instructional spending (CSINS),because it is difficult to determine ex ante whether greater resources spent on instruction (relative to the otherfunctional cost areas) will lead to higher or lower overall spending.16 The median years of teaching experience(FACEXP) is expected to have a positive coefficient in the cost equation, reflecting the expectation that moreseasoned teachers will earn higher pay than teachers with less experience. For administrative-related factors,ADMSAL is expected to exert an upward pressure on spending, leading to an expected positive coefficient.There are two possible ways to view the cost share for administrative spending (CSADM). One view is that themore highly paid and hence more competent the administrators are, the greater the degree to which costefficiency can be achieved, leading to a negative coefficient in the cost equation. The alternative view is thatCSADM should have a positive sign, reflecting the frequently held view that the ratio of administrative costs tototal spending is a measure of inefficiency in not-for-profit settings. The expected sign of the final administration-related variable, ADMEXPR, is also indeterminate in the cost function, because there is no strong a priorireason to expect that ADMEXPR will necessarily lead to lower overall costs. In addition to these uncontrollableand controllable factors, the quality outcome measure CTEST may be considered as an explanatory (potentialjointly endogenous) variable in the cost equation. Communities not satisfied with the level of studentachievement in their public schools are likely to be motivated to adopt measures (including increased spendingvia pressure on the state government and increased local taxes) to raise test scores. Thus, a positive coefficientis predicted on CTEST. Based on the foregoing, and using D&G (1997) as a guide for the functional form, thecomplete cost equation can be presented as: TCOST = ELEM1MDD2HIGH 3CTETS 4ENROLL5(ENROLLlnENROLL) 6 j 0 WDFG 7GEOCEI 8SPED 9LEP 10HIGHINC 11ABBOT 12STUFAC=1STUCOMP=2ATTD=3FACSAL=4CSINS=5FACEXPR=6 ADMSAL=7CSADM=8ADMEXPR=9e (4)After dividing through both sides of Equation (4) by ENROLL, and taking the natural log of both sides, theaverage cost function estimated can be written as: ln(EXPPP) = + ELEM + MIDD + HIGH + ln(CTEST) 0 1 2 3 4+ (1 >)ln(ENROLL) + (SQ ENROLL) + ln(WDFG) 5 6 7 + ln(GEOCEI) + ln(SPED) + ln(LEP) 8 9 10 +ln(HIGHINC) + ABBOT + = ln(STUFAC) 11 12 1 + = ln(STUCOMP) + = ln(ATTD) + = ln(FACSAL) 2 3 4 + =ln(CSINS) + = ln(FACEXPR) + = ln(ADMSAL) 5 6 7 + = ln(CSADM) + = ln(ADMEXPR) + e (5) 8 9 where:EXPPP = total expenditures per pupil; and SQ ENROLL = [In (ENROLL)].2 Note that the coefficient for ENROLL(1 >5) would be negative if there are economies of scale in the initial ranges of school district size. A significantpositive value for SQ ENROLL is expected if diseconomies of scale exist within the range.17 Specification of theOutcome Function As noted earlier, the output of public schools is a product of quantity of students andoutcome quality. In the operationalization of our outcome function in Equation (2), we include the potential jointlyendogenous variable EXPPP (expenditures per pupil) as a determinant of test scores. Given the previous(though inconsistent) findings in the literature (Hanushek 1986; Hedges et al. 1994), we expect EXPPP to have

08 March 2013 Page 118 of 145 ProQuest

a positive coefficient. Estimating a quality outcome function that includes EXPPP as an explanatory variableintroduces mutual simultaneity into the relationship. The selection of explanatory variables must consider thesimultaneous equation identification issue, and the need for suitable instrumental variables. Six of thesocioeconomic variables included in the cost equation are also included in the outcome equation: SPED, LEP,HIGHINC, ABBOT, WDFG, and GEOCEI. SPED is expected to have a negative coefficient in the outcomefunction because it is a handicap in the learning process. The expected sign of LEP is uncertain. Althoughlimited English proficiency may be an initial handicap, effective teaching can overcome this initial disadvantage.HIGHINC is expected to have a positive coefficient because students from wealthier districts have consistentlybeen shown to perform better on standardized tests. The coefficient on ABBOT is expected to be negativebecause the criterion for inclusion in that category is a priori poor performance. Since higher ordinal rankings(i.e., 1-8) of WDFG capture more privileged students, WDFG is expected to be positively associated with testscore performance. Finally, GEOCEI has an expected positive sign because it may proxy for family incomedifferences in the different geographic areas of the state. To aid in identification of the model, an additionaluncontrollable socio-economic explanatory variable, student mobility, was added to the test outcome function.Student mobility (STMOB) is known to have a strong negative influence on test score outcomes (Hanushek etal. 2004). The set of controllable management control tools in the outcome equation is the same as was used inthe cost equation. Although the evidence on the pupil-teacher ratio is mixed (Hanushek 1986), we expect thecoefficient on STUFAC to be negative according to conventional wisdom that smaller classes will enhancelearning. The variable STUCOMP is also expected to be negative under the assumption that greater penetrationof computers in the classroom enhances learning. The attendance rate (ATTD) is predicted to be positivelyrelated to test scores. The three faculty-related variables, FACSAL, CSINS, and FACEXP, all have predictedpositive signs in the test outcome equation under the expectation that higher-paid teachers, more resourcesdevoted to teaching (relative to other functional areas), and more experienced teachers will all result in higherstudent achievement. For similar reasons, the coefficients on the three administrative-related factors, ADMSAL,CSADM, and ADMEXPR are also predicted to be positive. The full test (quality) outcome function estimated canbe written as: ln(CTEST) = b + b ELEM + b MIDD + b HIGH + b ln(EXPPP) 0 1 2 3 4 + b ln(WDFG) + bln(GEOCEI) + b (lnSTMOB) 5 6 7 + b ln(SPED) + b ln(LEP) + b ln(HIGHINC) 8 9 10 + b ABBOT + +ln(STUFAC) + + ln(STUCOMP) 11 1 2 + + ln(ATTD) + + ln(FACSAL) + + ln(CSINS) 3 4 5 + + ln(FACEXPR) + +ln(ADMSAL) 6 7 + + ln(CSADM) + + ln(ADMEXPR) + e (6) When estimated together, the instrumental variablesfor EXPPP in Equation (5) are ENROLL and SQ ENROLL, so these two variables do not appear in Equation (6).There is evidence in the literature to suggest that small schools are more cost-effective than larger schools(Toch 2003), but our preliminary regressions (consistent with results reported by other studies such asChildress et al. 2006) did not find such a relation. Conversely, the instrumental variable for CTEST is STMOB,which appears in Equation (6) but not in Equation (5).18 There is no reason to believe that student mobilitywould have cost implications, other than student orientation costs, since no additional services need to beoffered to students transferring into a school district or transferring out. Given the apparent endogeneity ofEXPPP and CTEST, we estimated several variations of Equations (5) and (6): as single equations individuallyusing ordinary least-squares (OLS), jointly as a system of equations under seemingly unrelated regressions(SUR), and lastly as a system of simultaneous equations under three-stage least-squares (3SLS). We thenperformed the test of the fit of the instrumental variables, a statistical test for overrestriction of the instrumentalvariables, and the Hausman (1978) test for endogeneity.19 We also tested for heteroscedasticity andmulticollinearity. RESULTS Summary Statistics Summary statistics on key variables used in the study arepresented in Table 1 for year 2000. Because the profile of the school districts by type is similar in the other 2years, they have been omitted for the sake of brevity. Table 1 presents summary statistics on the distributions ofper-pupil expenditures, total enrollments, and median faculty salaries, by type of school district. The followingmay be noted from these results. First, per-pupil expenditures vary systematically by type of school district. The

08 March 2013 Page 119 of 145 ProQuest

lowest mean per-pupil expenditures occur in elementary school districts ($7,798), and increase progressivelythrough middle school districts ($7,982), Kindergarten- Grade 12 districts ($8,512), and high-school districts($9,918). Mean school district enrollments tend to be lowest for the ELEM and MIDD school districts, largest forthe K-12 districts, with the HIGH school districts in between. Finally, median faculty salaries are highest forHIGH, followed by K-12, and lowest for the MIDD and ELEM school districts. Table 2 presents the summarystatistics of the variables used in this study pooled across school district types by year. Total studentenrollments varied from 83 to 41,378 students in 2000, with similar ranges observed in 2001 and 2002. Theproportion of special education needs students (SPED) ranged from 5 percent to a maximum of 28 percent in2000, with the mean at 13 percent. By 2002, the maximum SPED had increased to 35 percent, although themean was unchanged. Similarly, the proportion of limited English proficiency students (LEP) ranged from 0 to27 percent in 2000 with the mean at 2 percent. The mean was virtually unchanged in 2001 and 2002, althoughthe maximum increased to 39 percent in 2002. Among the controllable management tools, the student-facultyratio averaged 12.69 in 2000, 12.48 in 2001, and 12.28 in 2002, a steady decline that is also noticeable withstudents-computers ratio (6.35, 5.89, and 4.98 in the three years, respectively). However, the attendance rateheld steady at an average of 95 percent in all the three years. The cost shares for instruction and administration(CSINS and CSADM, respectively) also remained constant, with averages of 61 percent and 13 percentrespectively for all the three years. The level of faculty experience (FACEXPR) experienced a decline from amean of 14.67 years in 2000, to 13.97 years in 2001, and finally to 13.17 years. This slight decline is alsonoticeable for administrative experience (ADMEXPR), which averaged 23.91, 23.86, and 23.53 for the threeyears, respectively, although the medians were the same. These results may reflect the result of personnelattrition as well as intra-state and inter-state mobility. As noted previously, the ability of school districtadministrators to initiate changes in, for example, the average teacher experience may be limited by teacher-union contracts, tenure rules, and so forth. Nevertheless, flexibility exists for district administrators to enforceinstitute changes by hiring only inexperienced teachers, shifting resources toward instructional costs or otherfunctional areas, and enacting policies to increase student attendance, for example. The effectiveness of suchpolicies relative to the cost of enforcement is one focus of this study. Results from Estimating the Cost FunctionThe results of estimating the cost function specified in Equation (5) using the SUR approach are presented inTable 3. Because the results obtained using OLS, 2SLS, and 3SLS are similar to the SUR results, we omit themfor the sake of brevity.20 For ease of referencing, the natural log transformation of the variables will be ignoredwhen referring to these variables in the regression. Thus, ln(EXPPP), for example, will be referred to simply asEXPPP. Panels A, B, and C of Table 3 present the results for years 2000, 2001, and 2002, respectively. Thereported significance levels are one-tailed for variables where a specific sign was expected, and two-tailedwhere there was no a priori expectation. The results in all three panels show that, with the exception of thelimited English proficiency ratio (LEP), all the uncontrollable variables are significant and consistent with a prioriexpectations, where such expectations existed. Since the results are consistent across all three years, we willuse the year 2000 results to illustrate the main findings in this section. The negative coefficient for ENROLL(student enrollments) is consistent with the expectation of increasing economies of scale (i.e., 5 = 0.628 since 1>5 = >0.372 in 2000). At the same time, the positive coefficient for SQ ENROLL (although with a smallercoefficient of 0.02) reflects diseconomies of scale for the largest school districts. The negative coefficient forHIGHINC (complement of percent students receiving meal aid) is consistent with the interpretation that schooldistricts in high-income brackets tend to be more cost-efficient than districts in poor income areas when otherfactors are controlled for. Of the nine management control variables, all but three are significant across all threeyears. The three exceptions are the attendance rate (ATTD), faculty experience (FACEXPR), and administratorexperience (ADMEXPR). Of the six controllable variables with significant signs, STUFAC (average class size)and STUCOMP (number students per computer) are negative as expected. Similarly, average faculty salary(FACSAL) and average administrator's salary (ADMSAL) are both positively signed as expected. Of the

08 March 2013 Page 120 of 145 ProQuest

remaining two variables whose sign could not be established on an a priori basis, both the cost shares forinstruction and administration (CSINS and CSADM, respectively) have negative coefficients. This implies thatcost efficiency is associated with higher spending on both instruction and administration at the expense of othercosts (such as operating and maintenance, and student support costs). The intriguing question is whetherhigher levels of these two inputs are related to achievement (quality) outcomes. This issue is addressed in thenext section. The final column of Table 3 presents the variance inflation factors (VIFs) obtained in the costequation regression for 2000. The VIFs gauge the degree of multicollinearity in the regression estimates, andare defined as the inverse of the multiple correlation coefficients from the regression of all the independentvariables on all the other independent variables (Belsley et al. 1980). The VIFs for years 2001 and 2002 aresimilar and have been omitted. Results indicate that except for the high degree of correlation between ENROLLand SQ ENROLL, none of the other variance inflation factors are a cause for concern. In order to investigate thepossible simultaneous equation bias issue, Equation (5) was estimated jointly with Equation (6) using the 3SLSapproach. The results using 3SLS (not reported) are substantially identical to those presented in Table 3 basedon SUR. The main difference is the coefficient on weighted-average test scores (CTEST). This variable wassignificant in the SUR model in Table 3 (p 0.001), but is not significant under 3SLS. Specification tests todetermine if simultaneous equation bias existed were performed on the 3SLS results. The test for the fit of theinstrumental variables is based on partial F-statistics, which measures the incremental contribution of theinstruments in the firststage regressions. The partial F-statistics are 37.27, 27.54, and 19.45 for years 2000,2001, and 2002, respectively. Thus, for all three years, the instrument used for CTEST (namely STMOB-studentmobility) was found to be strong. Since Equation (5) was exactly identified, there was no need to perform thetest for over-identified restrictions. Proceeding to the Hausman (1978) test shows that the null hypothesis of noendogeneity cannot be rejected in any of the three years, with F ratios of 0.52, 0.06, and 0.00, respectively.These results indicate that the SUR results are less biased than the 3SLS results for Equation (5). Since theSUR results in Table 3 shows CTEST to be statistically significant while the 3SLS results did not, the resultssupport the conclusion that test scores have an independent effect on spending levels. Results from Estimatingthe Test Outcome Function Results of estimating the outcome function in Equation (6) under SUR arepresented in Table 4, with years 2000, 2001, and 2002 in Panels A, B, and C, respectively. All three panels ofTable 4 indicate that three of the uncontrollable variables (WDFG, STMOB, and HIGHINC) are significant in allyears. WDFG (weighted district factor group index) and HIGHINC have the expected positive sign, whileSTMOB is negative (as expected). Of the remaining four uncontrollable variables, the dummy variable forAbbott districts (ABBOT), the geographical cost of education index (GEOCEI), and the ratio of special educationstudents to the student population (SPED) are all negatively signed and significant in at least one of the threeyears. LEP is positively signed but significant in only year 2000. Of the nine management control variables,three are consistently significant across all three years: ATTD and ADMSAL (with expected positivecoefficients), and ADMEXPR (with an unexpected negative coefficient). Of the remaining six controllablevariables, STUFAC, STUCOMP, and CSINS are significant in 2 of the three years. In contrast to expectations,STUFAC has a significant positive coefficient in 2001 and 2002, suggesting that students perform better inlarger classes. FACSAL, CSADM, and FACEXPR have statistically insignificant coefficients in all three years.Finally, the jointly endogenous variable, expenditures per-pupil (EXPPP), is strongly significant in all threeyears. The final column of Table 4 presents the variance inflation factors (based on the 2000 data) to permit thedegree of multicollinearity to be assessed. Since none of the observed values exceed six (the thresholdprovided by Belsley et al. (1980)), and the results for 2001 and 2002 are similar to those reported here,multicollinearity is not a problem in these regressions. Like the cost function, tests for specification of theoutcome function were conducted by estimating Equation (6) simultaneously with Equation (5) using 3SLS. Thepartial F-statistics for evaluating the incremental contribution of the instrumental variables are 65.21, 80.31, and55.49 for 2000, 2001, and 2002, respectively. The instruments have high explanatory power, alleviating any

08 March 2013 Page 121 of 145 ProQuest

concern about potentially weak instruments. In addition, the χ2 values from the over-identifying restrictions testwere 0.21, 0.05, and 0.36 for the three years respectively, none of which are significant at reasonableprobability levels. Thus, the instruments can be regarded as exogenous. Finally, the F-ratios from the Hausman(1978) test were 0.03, 0.31, and 0.010, respectively, all of which are statistically insignificant. The nullhypothesis of no endogeneity cannot be rejected. In conclusion, the SUR results are less biased than the 3SLSresults, consistent with the earlier findings. For this reason, presentation of the 3SLS results are omitted fromthis paper.21 Summary and Synthesis of Results Before proceeding to an analysis of the results, it is importantto note that the leastsquares regressions used here, unlike the stochastic (and/or deterministic) frontierestimation techniques used by D&G (1997) and Mensah and Li (1993), among others, are not based onextreme values. That is, the least-squares method used in this paper estimates the average effects of unitchanges in the independent variables on the dependent variable within a relevant range. This choice isdeliberate since we seek to provide insight into the trade-off decisions made by the average schoolsuperintendent who is facing the multiple constituencies and objectives. We do not make the assumption thatthe average school superintendent is a cost-minimizer or outcome-maximizer. To synthesize the resultsprovided earlier, the relative cost-effectiveness indicators (coefficients) for the controllable variables estimatedunder SUR are summarized in Table 5 for each of the three years. For the cost function results, positivecoefficients indicate costincreasing factors, while negative coefficients are cost-beneficial. In contrast, for thetest outcome function, positive coefficients are outcome beneficial, while negative signs have the oppositeeffect. The results in Table 5 show that, of the various factors that are controllable by school administrators, fourdeserve special favorable attention. First, CSINS is doubly beneficial (DB = outcome-increasing and cost-decreasing at the same time) in 2000 and 2001, and rates a BC (beneficial cost effect) in 2002. This is closelyfollowed by ATTD, which has a BO (beneficial effect on outcomes with no offsetting cost-increasing effect) in allthree years. Somewhat surprisingly, STUFAC earns a BC (cost-decreasing effect with no offsetting negativeeffect on outcomes) rating in 2000, and a DB rating in 2001 and 2002. However, we should note that the stateof New Jersey has a mandatory maximum student/ faculty ratio for the different types of school districts andgrades. Subject to this state mandate, it appears that, for the data and the years covered, school districts whosestudent/ faculty ratios were above the state-wide means (but which met the state-wide maximum limits) actuallydid better from both a cost-efficiency and test outcome standpoint than schools with lower student/ facultyratios.22 This finding suggests that school districts significantly below the state-mandated maximum student/faculty ratios did not benefit much from those decisions. The fourth variable of note is CSADM, which earned aBC (cost-reducing effect with no negative effect on outcomes) rating in all three years. Once again, it isimportant to note that the New Jersey Department of Education monitors the administrative cost ratios and hasguidelines regarding them. Granted this context, the results suggest that schools whose administrative costratios were above the state-wide means did, in fact, achieve better control over costs without paying a penalty interms of lower test score performance. A closer examination of this finding and its implications clearly meritattention in future research.23 At the opposite end of the spectrum, two variables stand out as highly negative intheir impact on either costs or outcomes. ADMEXPR has a uniformly NO interpretation (negative outcome effectwith no offsetting effect on costs) in all years. This result suggests that, at the margin, an increase in ADMEXPRis associated with a decrease in test scores, but simultaneously has a positive effect on costs. Thus, oneinference from these results is that schools with ADMEXPR above the state-wide mean may benefit from theflexibility to hire administrators with less experience who command lower salaries, with all other factors heldconstant. The retirement of older, more experienced administrators, if they are replaced by less experienced butlower-paid administrators, may actually be beneficial to such school districts. The other noticeable negativefactor is FACSAL, which is rated CI (cost increasing with no offsetting benefit to outcomes) in all three years.These results suggest that school districts with faculty salaries above the state-wide means do not obtain anyextra payoff from the higher salaries. One intuitive interpretation for this finding is that collective bargaining

08 March 2013 Page 122 of 145 ProQuest

agreements with the teachers' union may limit the ability of administrators to control teacher salaries. Finally,two of the nine management tools have ratings of TO (trade-off of costs against outcomes) in the first 2 years,with a change to either BC or BO in 2002. The first variable, STUCOMP, is the number of students percomputer. STUCOMP has a negative TO rating in 2000 and 2001, implying that greater utilization of computersin the classroom increases total costs, but also increases test scores. The magnitude of the trade-off forSTUCOMP decreased from 0.286 in 2000, to 0.208 in 2001, but in 2002 STUCOMP had a rating of BC(beneficial to costs with no offsetting reduction of outcomes). As noted earlier, the student/computer ratioexhibited a steady decrease over the sample period, reflecting a trend toward the introduction of morecomputers in the classroom. The change in the rating for STUCOMP may be related to this decrease, butadditional years of data and/or further investigation would be required to establish any causal relation. The othervariable with a TO rating is ADMSAL. The positive TO ratings in 2000 and 2001 indicate that the increasedeffect on outcomes due to greater administrative salaries is offset by its cost-increasing effect. The TO rating forADMSAL increases from 0.102 in 2000 to 0.216 in 2001, and is finally BO (beneficial on outcomes only) in 2002under the SUR results. This was accomplished through a steady increase in the outcome coefficient, and asimilarly steady decrease in the cost coefficient. However, since one cannot rule out the possibility that thetrends in the ratios over time are merely statistical noise, no further interpretation can be offered withoutadditional years of data. Contribution of Joint Estimation Methodology As noted in the introduction, the aim ofthis paper is to propose the joint estimation of cost and outcome functions to enable management accountantsto understand the coping mechanisms adopted by non-profit managers who face multiple constituents andmultiple objectives, some of which may be in conflict. The results show that, for controllable variables denotedBC, BO, or DB, decision unit managers appear to be utilizing low levels of beneficial resources, within the rangeof our data set. In still other cases, some managers utilize high levels of resources identified as NO or CI that onaverage provide no further benefit within the range. Clearly, one possible explanation for such behavior mightbe if the decision unit managers face institutional constraints that limit their freedom of action in important ways.The proposed joint estimation methodology thus allows specific resources or control tools whose levels appearto be under or over utilized, within the relevant range of our data set, to be identified. Further investigation bythe management accountants would permit the exact nature of the institutional constraints (and the associatedopportunity costs) to be identified and possibly dealt with by the decision unit manager. Since the primarycontribution of the paper is intended to be methodological, and the school district example was only used forillustrative purposes, we provide below only a brief overview of how the method can be beneficial in this context.The literature on public school education is vast and increasing in scope, particularly following nationaleducation initiatives such as the No Child Left Behind Act of 2001 (U.S. Department of Education 2002).However, what is most striking about this vast literature is the consistency with which contradictory results havebeen found. In a thorough review of the education literature, Hanushek (1986) observed the following: (1) Out of112 studies of outcome functions surveyed, 23 studies found the Teacher/ Pupil ratio to be a statisticallysignificant variable, while 89 found it to be insignificant. Of the 23 that reported statistical significance, ninereported positive coefficients, while 14 reported negative coefficients. (We found STUFAC to be positive andsignificant in our outcome function in 2001 and 2002, but not significant in 2000).24 (2) In the 109 studies whereFACEXPR was examined in the outcome function, 40 studies found it to be statistically significant, while 69found it to be insignificant. Of these 40 studies, 33 reported a positive coefficient, while 7 reported a negativecoefficient. (We found FACEXPR to be statistically insignificant across all years in the outcome function). (3) In60 studies where FACSAL was included as an independent variables in the outcome function, 10 reportedstatistically significant coefficients (nine positive signs, and one negative sign) while 50 reported no statisticallysignificant coefficients. (We found the variable FACSAL had no effect on test scores in all three years). (4) In 65studies that included per-pupil expenditures in the outcome function, 16 studies found it to be a significantexplanatory variable. Of these 16, 13 studies had positive coefficients and three had negative coefficients. (We

08 March 2013 Page 123 of 145 ProQuest

found EXPPP to be significant and positive across all three years in our outcome function under SURestimation). These studies summarized by Hanushek (1986) span a number of decades, and reflect institutionalfactors specific to different states and even evolving national policies. As a result, the findings of educationalstudies may not be generalizable beyond the time periods and samples from which the data are obtained.25Additionally, the studies may use different measures for explanatory variables and employ different statisticalmodels. Educational outcomes are also often based on different achievement tests. These differencesundermine any possibility that this study (or any other, for that matter) can yield authoritative findings that canresolve the overwhelming conflicts reported in the literature. CHECKS FOR ROBUSTNESS OF RESULTS Inthis section, we perform checks to determine the robustness of our results. First, given the correlation betweensome of the independent variables, we examine how the omission of any of the variables affects theexplanatory power of the remaining variables in the cost and outcome equations. Second, we evaluate morecarefully the frequently held view that devoting more resources to instruction is the best way to make publiceducation more cost-effective. Finally, we test the models for the underlying assumptions of linearity (in logterms) and additivity of the variables. All robustness tests are conducted by pooling the data cross-sectionallyover years 2000-2002. Incremental Explanatory Power The coefficients on all the variables are essentiallymarginal effects obtained with all the other variables held constant at their mean values (Johnston 1972, 132-135). To aid in understanding the relative importance of the individual variables, we estimated the SURregressions with one variable omitted at a time. The R2s without the omitted variable and the incrementalchange in the OLS and system-weighted R2s are disclosed in Table 6. The results in Table 6 are sorted by themagnitude of the overall effect on the SUR system-weighted R2. But for analytical purposes, the incrementaleffect of the omission of each variable on the individual OLS R2s are more relevant. The results for the costequation show that the controllable variables with the greatest incremental contribution in explaining totalexpenditures are the student-faculty ratio (7.5 percent), mean faculty salaries (4.6 percent), and total cost shareof instruction (3.2 percent). None of the other controllable variables exceeds 1.0 percent. Of the noncontrollablevariables, the economies of scale effect (ENROLL &SQ ENROLL), had the most impact on the observedexplanatory power of the model (7.4 percent). Additional insights are provided in Table 6 by examining the mostimportant individual factors that contribute to explaining test score performance. The most important variables inthe outcome equation are the district factor groupings (WDFG, 0.033), level of district income (0.017), studentmobility (0.017), and the existence of special education programs (0.017). None of the controllable variableshave an incremental explanatory contribution to the model that exceeds 1.0 percent. This finding is consistentwith the literature summarized by Hanushek (1986), which shows that once socio-economic factors arecontrolled for, school inputs such as the student-faculty ratio, teacher salaries, and per-pupil expendituresexplain a relatively small portion of student achievement scores. Linearity The results presented thus far do notallow for possible non-linearity or curvatures in the model. However, it can be argued that for many of thesevariables, it is likely that increases in the quantities may push the relationship identified beyond levels that arebeneficial. One good example is the cost share assigned to instruction. Although our results, along with the priorliterature, accords with the view that as much resources as possible should be devoted to instruction, it ispossible to envision situations where too many resources are devoted to instruction, and insufficient resourcesare made available to other functional areas. To examine this issue in more detail, we divided the pooledsample into three equal groups as follows: (1) CSINS 1 denotes those observations with instructional cost shareranging from the minimum of 47.04 percent to 59.2 percent; (2) CSINS 2 denotes observations with CSINSranging from 59.2 percent to 62.2 percent; and (3) CSINS 3 denotes those observations with instructional costshares ranging from 62.2 percent to the maximum of 73.2 percent. Equations (5) and (6) were re-estimated withthe three partitioned variables for CSINS replacing the original single variable. The results of these regressionsare reported in Table 7. For the sake of brevity, only the regression statistics for the three partitioned variablesfor CSINS are reported in Table 7 (along with the original variable in the pooled sample). As the results in Table

08 March 2013 Page 124 of 145 ProQuest

7 show, the results for CSINS 1 and CSINS 2 are consistent with the prior findings of a DB (double-benefit)variable. That is, for the first two groups, increased spending on instruction is associated with both lower overallspending and improved test scores. However, for the third partition (where the instructional cost share is above62.2 percent), the situation reverses, with higher spending on instruction associated with both higher costs andlower test scores. This unexpected finding suggests that rule-ofthumb prescriptions that call for increaseddiversion of resources to hire more teachers, without due attention to their proper deployment, the developmentof appropriate curriculum, etc., can have adverse consequences. This kinked cost-behavior, whereby thebenefits of additional spending on instruction are lost, may explain why there are conflicting findings reported inthe educational literature on the importance of direct spending on instruction (Hanushek 1986), and deservesfurther study. Additivity Effects No attempt was made in our main results to determine whether our findings varycrosssectionally between high- and low-income school districts. In this section, we divide the sample into twosub-samples: (1) low-income districts (districts where the variable HIGHINC is below the median of 0.92) and(2) high-income districts (districts where HIGHINC is above the median). Equations (5) and (6) were re-estimated for these two subsamples. Only the results for the outcome equation are reported in Table 8 (the costfunction results are omitted), but the last two columns apply our methodology to show how the two sub-samplesdiffer with respect to the cost-effectiveness of the management tools. We focus the discussion on the last twocolumns of Table 8, where the cost-effectiveness labels on the independent variables differ between high-income and low-income districts. There are five variables with inconsistent cost-effectiveness ratings. Of thesefive, two are uncontrollable factors (LEP and HIGHINC). LEP has a trade-off rating (TO) for low-income districts,but a benefits only (BO) effect for high-income districts. This relation is reversed for HIGHINC. Greater incomeswithin the low-income district sub-sample are associated with higher test scores but not lower costs. But in thehigh-income district sub-sample, a trade-off is observed with respect to HIGHINC (higher test scores but alsogreater costs). Among the controllable factors, three variables have differing signs on costeffectiveness(STUCOMP, FACEXPR, and CSADM). Increasing the proportion of students to computers in the low-incomesub-sample reduces costs with no effect on test score outcomes (BC). In contrast, in the high-income sub-sample, STUCOMP has a trade-off effect (TO), with lower test scores offset by lower spending. The level offaculty experience has a decreasing effect on student achievement in the low-income sub-sample with nooffsetting cost reduction (NO), but no effect within the high-income sub-sample (NS). Our finding for the low-income sub-sample contradicts the results of some prior studies that found a positive or no relation betweenteacher experience and achievement scores (see for example, Elliott 1998; Hanushek 1986).26 This findingmay in part be explained by Hanushek (1986, 1162), who posits that student achievement affects teacherselection. In this case, it may be that experienced teachers with better teaching skills select schools with higher-achieving students residing in high-income districts. Lastly, greater costs shares devoted to administration has abeneficial effect on total costs with no offsetting in reduction in test scores (BC) in the low-income sub-sample,but a double benefit (both cost-reducing and test performance-increasing) in the high-income districts. Thisresult is contrary to the results of Dee (2005), who finds a negative relation between non-instructional per-pupilexpenditures and high-school graduation rates, and Brewer (1996), who finds no significant relation betweennumber of school district administrators and achievement scores. Our finding on how changing cost shares fordifferent functions can have such marked effects on costs and outcomes represents an avenue of futureresearch. We performed additional robustness tests to check for possible interaction effects. For example, itcould be argued that teacher experience may be beneficial to test score outcomes in the poorer school districts(e.g., the Abbott districts), even if not significant overall. Likewise, lower student-faculty ratios could be morebeneficial to test scores in districts marked by high student mobility. Neither of these two possible interactioneffects were significant when estimated. We also surveyed the educational economics literature for examples ofinteraction terms estimated by other researchers, but failed to find evidence of interactions among the variablesused in these studies examined. Given the large number of possible interaction terms that could be conceived

08 March 2013 Page 125 of 145 ProQuest

of, without a theory to guide why particular interaction terms should be expected, further tests for potentialinteractions among variables did not seem practical. CONCLUSIONS In conclusion, the technique presented inthis paper for identifying cost-effective management tools has the potential to yield valuable insight into thetrade-offs made by the management of public-sector institutions with their multiple constituencies andobjectives. The methodology proposed does not require any skills beyond some statistical knowledge, and theability to identify and quantify some of the key management tools that may be applicable to a given context. Ourempirical illustration of the methodology using public school data has demonstrated that the technique isfeasible in an actual decision-making setting. By applying this methodology in a public education setting, schoolofficials can identify factors controllable by school district superintendents to institute reforms to improve theiroperations and achieve greater cost efficiency and effectiveness. Among the limitations of this study, perhapsthe most important is that the empirical illustration draws on the public school data for only one state. As notedearlier, the results shown here may not be generalizable to other states due to institutional constraints andtaxpayer preferences. Another potential limitation is that the illustration uses only one type of public-sectorinstitution, public schools. Clearly, the application of the methodology to public school data in other states and toother types of public-sector institutions would help to establish the wider applicability of the suggestedtechnique. Finally, the method proposed in the paper assumes linear additivity of the independent variables. Tothe extent that there are significant interaction terms, this assumption may be invalid, limiting the generalapplicability of the proposed method. Sidebar We gratefully acknowledge Yut'se Thomas and Allen DuPree ofthe New Jersey Department of Education for providing us with data used in this study. This study has benefitedfrom funding by the Center for Governmental Accounting Education and Research (CGAER) of RutgersUniversity. This paper was presented at the Management Accounting Section of the American AccountingAssociation's 2007 Annual Meeting in Chicago. We thank the discussant (James Hesford) and the otherconference participants for their helpful comments and suggestions. Finally, we are grateful to Joan Luft (editor)and the anonymous reviewer of this journal whose very insightful comments have contributed greatly in therevisions of this paper. Footnote 1 For a review of performance-based budgeting at the state level, seeWilloughby and Melkers (1998, 2000). 2 As examples of this third aspect, the output of hospitals is oftenmeasured in terms of the number of surgical procedures performed, number of patients treated, total patientdays, and number of outpatient visits. However, their final outcome measures focus on mortality rates, patientrecovery rates, and longevity of patients after treatment. Similarly, for correctional institutions, output measurestend to focus on the total number of inmates and days of incarceration, while final outcome measures focus onthe effectiveness of incarceration as captured by the rate of recidivism (Mensah and Li 1993). 3 Specifically, weestimate a cost function using the inputs as traditionally defined, and also estimate an outcome function usingmost of the same explanatory variables as in the cost function. Thus, from the estimated coefficients in the costand outcome equations, we are able to derive directly comparable measures of the marginal cost and marginalbenefits of the input factors. 4 In this paper we define a ''closed production system'' as one in which all inputsand outputs are known and measurable and included in the equation estimated. For example, in the standardneoclassical system, the typical inputs of labor, materials, and capital are known and measurable, just as theoutputs of the system are. This results in a closed system. The inclusion of price information allows theestimation of a cost function that is the dual of the production function. In the public-sector setting, theproduction system is an open system. Beyond the labor, supplies, and capital inputs, the environmental, socio-economic, and political settings can greatly influence the outcomes, which they themselves are not easilyquantified. 5 See, for example, Mensah and Li (1993) on the effects of constraints such as line-item budgetingon allocative efficiency. 6 The Coleman report, based on a database of 3,000 schools and over a half millionstudents, concluded that cross-sectional differences in family background and student characteristics were theprimary determinants of student achievement, not school inputs (e.g., per-pupil expenditures). 7 Hedges et al.(1994; hereafter HL&G), who criticize Hanushek's (1986, 1989) ''vote-counting'' methodology, reanalyzed

08 March 2013 Page 126 of 145 ProQuest

Hanushek's data using a meta-analysis approach. The results of HL&G's (1994) combined significance testsfind a systematic positive relation between school resource inputs and student achievement tests. 8 Of the 106educational production functions summarized by Hanushek (1986, 1161), teacher education was positivelyrelated to test scores in only 6 percent of the models, and negatively related in 5 percent of the models. 9Jaggia and Kelly-Hawke (1999, 193) acknowledge that mutilcollinearity may exist among their school inputvariables since per-pupil expenditures are mostly a function of the teacher-pupil ratio and per-pupiladministrative expenditures. 10 A legislative history of the Abbott vs. Burke court cases and the criteriaemployed by the New Jersey Commissioner of Education in the designation of Abbott Districts can be found athttp://www.nj.gov/ njded / abbotts / regs / criteria.htm. 11 The Grade 11 exam was replaced in 2002 by adifferent exam given in Grade 12. For all three tests, the basis skills evaluated during the period 2000 to 2002are language and mathematics. 12 The factors used in the NJDOE factor analysis are (1) the number of singlemothers in the school district; (2) the average income level in the district; and (3) the concentration ofdisadvantaged minorities in the school district. 13 The exact variables included in the factor analysis programand the factor loadings are considered proprietary by the state, so we could not obtain more detailedinformation. 14 The complement of LOWINC (percentage of students whose families qualify for federal food aid)is used in order to avoid dealing with negative values in the subsequent log transformation of the variable. 15 Awell-known economic theory offered by Tiebout (1956) posits that spending on government services will varywidely as a function of differences in wealth. Thus, wealthier areas can spend more from local resources onschools, and although court mandates and public policy may seek to equalize such spending across districts,differences may still survive. 16 The five major functional cost classifications adopted by the NJDOE areInstructional, Administration, Student Support, Operations and Maintenance, and Extra-Curricular Activities, andare derived from Financial Accounting for Local and State School Systems (U.S. Department of Education,Office of Educational Research and Improvement 1990). 17 See D&G (1997) for the optimal size formula. 18Note there are two instrumental variables for EXPPP, so Equation (6) is over-identified. There is oneinstrumental variable for CTEST so Equation (5) is exactly identified. 19 Following Larcker and Rusticus (2005)and Hahn and Hausman (2002, 2003), we applied these three tests in succession. First, we checked for theweak instruments problem identified by Hahn and Hausman (2003), followed by the test for over-identifyingrestrictions for Equation (6). Finally, where the test for over-identifying restrictions indicated that the instrumentswere valid, we conducted a standard Hausman (1978) test to determine if a simultaneity problem existed. 20Seemingly unrelated regression (SUR) is a full information statistical approach that exploits the likelycrossequation correlation of the error terms. For well-specified models, SUR is more efficient than OLS (Greene1990). Its simultaneous equation counterpart is three-stage least-squares (3SLS), which applies the SURapproach to the 2SLS estimates. We applied all four possible statistical methods, and have chosen to focus onthe SUR results because the coefficient estimates are more efficient and less bias. 21 In contrast to the strongresults observed for EXPPP in Table 4, the coefficient on EXPPP from the 3SLS regression is not significant.However, the statistical insignificance of the Hausman test for endogeneity suggests that these results can berejected in favor of the SUR results in Table 4. 22 The case for STUFAC is more complicated than in the resultsreported here because the data is pooled across the four different types of school districts. In a more detailedanalysis based on disaggregated data, we found STUFAC in the K12 and ELEM school systems to be a TOwith a negatively signed relative cost-effectiveness measure. However, the trade-off ratios (=+i= / ==i=) werequite small, indicating that the dampening effect of higher student / faculty ratios on outcomes is modestcompared to their cost savings effect. (That is, they contribute more to cost savings than they induce lower test-score performance.) In the HIGH and MIDD school districts, STUFAC earned a straight BC (cost-savings withno off-setting negative effect on outcomes). This suggests that relatively high levels of STUFAC may actually bemore cost-effective than excessively low ratios. 23 It should be noted that, in addition to institutional constraintsplaced on the cost share for administration by the state of New Jersey, taxpayer preferences may influence

08 March 2013 Page 127 of 145 ProQuest

other outcomes not captured by our model. For example, the desire to produce high-quality athletic teams mayeffectively limit the ability of school district administrators to reallocate funds to administration and instructionand away from extra-curricular activities. 24 The variable used in our models, STUFAC, is the inverse of theTeacher / Pupil ratio. Thus, our finding of a positive coefficient on STUFAC in years 2001 and 2002 in ouroutcome function is consistent with the 14 studies summarized by Hanushek (1986) that found a negativecoefficient on the Teacher / Pupil ratio. 25 Our assertion that the results of educational studies may not bereplicable outside of the sample periods and locations is supported by the fact that when we attempted toreplicate D&G (1997) using data from the state of New Jersey (instead of the state of Missouri data), thecoefficients we derived (as well as the level of cost inefficiency) were very different from the ones they found.26 It should be noted that the prior works cited do not partition their school samples on measures of medianfamily income, making comparisons difficult with regards to teacher experience. References REFERENCESBaber, W. R., P. L. Daniel, and A. A. Roberts. 2002. Compensation to managers of charitable organizations: Anempirical study of the role of accounting performance measures of program activities. The Accounting Review77 (3): 679-693. Belsley, D., E. Kuh, and R. Welsch. 1980. Regression Diagnostics: Identifying Influential Dataand Sources of Collinearity. New York, NY: John Wiley &Sons. Bound, J., D. Jaeger, and R. Baker. 1995.Problems with instrumental variables estimation when correlation between the instruments and the endogenousexplanatory variables is weak. Journal of the American Statistical Association 90 (430): 443-450. Brewer, D.1996. Does more school district administration lower educational productivity? Some evidence on the''administrative blob'' in New York public schools. Economics of Education Review 15 (2): 111-124. Broom, C.A. 1995. Performance-based government models: Building a track record. Public Budgeting &Finance 15(Winter / 4): 3-17. Brown, R. E., M. J. Myring, and C. G. Gard. 1999. Activity-based costing in government:Possibilities and pitfalls. Public Budgeting &Finance 19 (Summer / 2): 3-21. Callan, S. J., and R. E. Santerre.1990. The production characteristics of local public education: A multiple product and input analysis. SouthernEconomic Journal 57 (2): 468-480. Childress, S., R. Elmore, and A. Grossman. 2006. How to manage urbanschool districts. Harvard Business Review (November): 55-68. Coleman, J., E. Campbell, C. Hobson, J.McPartland, A. Mood, F. Weinfeld, and R. York. 1966. Equality of Educational Opportunity. Washington, D.C.:Government Printing Office. Dee, T. 2005. Expense preference and student achievement in school districts.Eastern Economic Journal 31 (1): 23-44. Dopuch, N., and M. Gupta. 1997. Estimation of benchmarkperformance standards: An application to public school performance. Journal of Accounting and Economics 23(2): 141-161. Elliott, M. 1998. School finance and opportunities to learn: Does money well spent enhancestudents' achievement? Sociology of Education 71 (3): 223-245. Gianakis, G. A. 2002. The promise of publicsector performance measurement: Anodyne or placebo? Public Administration Quarterly 26 (Spring): 35-64.Governmental Accounting Standards Board (GASB). 1989. Research Report, Service Efforts andAccomplishments Reporting: Elementary and Secondary Education. Norwalk, CT: GASB. _____. 1994. ServiceEfforts and Accomplishments Reporting (April). Concepts Statement No. 2. Norwalk, CT: GASB. _____. 1998.Accounting and Financial Reporting for Nonexchange Transactions (December). Statement No. 33. Norwalk,CT: GASB. _____. 1999. Basic Financial Statements-and Management's Discussion and Analysis-for State andLocal Governments (June). Statement No. 34. Norwalk, CT: GASB. Greene, W. 1980. Maximum likelihoodestimation of econometric frontier production functions. Journal of Econometrics 13 (1): 27-56. _____. 1990. Agamma-distributed stochastic frontier model. Journal of Econometrics 46 (1-2): 141-163. Hahn, J., and J.Hausman. 2002. A new specification test for the validity of instrumental variables. Econometrica 70 (1)(January): 163-189. _____, and _____. 2003. Weak instruments: Diagnosis and cures in empiricaleconometrics. The American Economic Review 93 (2): 118-125. Hansmann, H. 1996. The Ownership ofEnterprise. Cambridge, MA: Belknap Press. Hanushek, E. 1979. Conceptual and empirical issues in theestimation of educational production functions. The Journal of Human Resources 14 (3): 351-388. _____. 1986.The economics of schooling: Production and efficiency in public schools. Journal of Economic Literature 24 (3):

08 March 2013 Page 128 of 145 ProQuest

1141-1177. _____. 1989. The impact of differential expenditures on school performance. EducationalResearcher 18 (4): 45-62. _____, J. F. Kain, and S. G. Rivkin. 2004. Disruption and Tiebout improvement: Thecosts and benefits of switching schools. Journal of Public Economics 88 (9-10): 1721-1746. Hausman, J. 1978.Specification tests in econometrics. Econometrica 46 (6): 1251-1271. Hedges, L., R. Laine, and R. Greenwald.1994. Does money matter? A meta-analysis of studies of the effects of differential school inputs on studentoutcomes. Educational Researcher 23 (3): 5-14. Jaggia, S., and A. Kelly-Hawke. 1999. An analysis of thefactors that influence student performance: A fresh approach to an old debate. Contemporary Economic Policy17 (2): 189-198. Johnston, J. 1972. Econometric Methods. 2nd edition. New York, NY: McGraw-Hill. Kinney, W.2001. Accounting scholarship: What is uniquely ours? The Accounting Review 76 (2): 275-284. Kravchuk, R. S.,and R. W. Schack. 1996. Designing effective performance: Measurement systems under the GovernmentPerformance and Results Act of 1993. Public Administration Review 56 (4): 348-358. Larcker, D. F., and T. O.Rusticus. 2005. On the use of instrumental variables in accounting research. SSRN Working paper. Availableat: http: / / ssrn.com/abstract=694824. Librera, W. 2003. Designation of Abbott districts: Criteria and process.New Jersey Department of Education (April 11). Available at: http: / /www.nj.gov/njded/ abbotts / regs /criteria.htm. McNab, R. M., and F. Melese. 2003. Implementing the GPRA: Examining the prospects forperformance budgeting in the federal government. Public Budgeting &Finance (Summer / 2): 73-95. Mensah, Y.M., and S.-H. Li. 1993. Measuring productive efficiency in a not-for-profit setting: An extension. The AccountingReview 68 (1): 66-92. Mosso, D. 1999. Accounting for the business of government-New goals, old myths. PublicBudgeting &Finance 19 (Winter / 4): 65-74. Mullins, D. R., and C. K. Zorn. 1999. Is activity-based costing up tothe challenge when it comes to privatization of local government services? Public Budgeting &Finance 19(Summer / 2): 37- 58. Mundlak, Y. 1996. Production function estimation: Reviving the primal. Econometrica 64(March): 431-438. New Jersey Department of Education (NJDOE). 2004. Vital Education Statistics 2004-2005.Available at: http://www.state.nj.us/njded/ data / vitaled/0405/ . Rose-Ackerman, S. 1996. Altruism, nonprofits,and economic theory. Journal of Economic Literature 34 (June): 701-728. Sander, W. 1993. Expenditures andstudent achievement in Illinois: New evidence. Journal of Public Economics 52 (3): 403-416. Shephard, R.1970. Cost and Production Functions. Princeton, NJ: Princeton University Press. Staiger, D., and J. Stock.1997. Instrumental variable regression with weak instruments. Econometrica 65 (3): 557-586. Stock, J., J.Wright, and M. Yogo. 2002. A survey of weak instruments and weak identification in generalized method ofmoments. Journal of Business &Economic Statistics 20 (4): 518-529. Tiebout, C. M. 1956. A pure theory of localexpenditures. The Journal of Political Economy 64 (5): 416-424. Toch, T. 2003. High Schools on a HumanScale: How Small Schools Can Transform American Education. Boston, MA: Beacon Press. U.S. Department ofEducation, Office of Educational Research and Improvement. 1990. Financial Accounting for Local and StateSchool Systems. Washington, D.C.: National Center for Education Statistics. U.S. Department of Education.2002. No Child Left Behind Act of 2001. Public Law 107-110. Available at: http://www.ed.gov/ policy/ elsec/leg/esea02/index.html. Varian, H. 1992. Microeconomic Analysis. 3rd edition. New York, NY: Norton.Willoughby, K. G. and J. E. Melkers. 1998. The state of the states: Performance-based budgeting in 47 out of50. Public Administration Review 58 (Spring/ 1): 66-73. _____, and _____. 2000. Implementing PBB: Conflictingviews of success. Public Budgeting &Finance 20 (Spring/ 1): 105-120. AuthorAffiliation Yaw M. MensahMichael P. Schoderbek Robert H. Werner Rutgers, The State University of New Jersey Subject: Studies; Cost accounting; Accounting systems; Public schools; Federal budget; Performanceevaluation; Models Location: United States--US, New Jersey Classification: 9130: Experiment/theoretical treatment, 4120: Accounting policies&procedures, 5240:Software&systems, 9550: Public sector, 8306: Schools and educational services, 9190: United States

08 March 2013 Page 129 of 145 ProQuest

Publication title: Journal of Management Accounting Research Volume: 21 Pages: 203-239 Number of pages: 37 Publication year: 2009 Publication date: 2009 Year: 2009 Publisher: American Accounting Association Place of publication: Sarasota Country of publication: United States Journal subject: Business And Economics--Accounting ISSN: 10492127 Source type: Scholarly Journals Language of publication: English Document type: Feature Document feature: Equations;Tables;References ProQuest document ID: 210179154 Document URL: http://search.proquest.com/docview/210179154?accountid=86413 Copyright: Copyright American Accounting Association 2009 Last updated: 2012-07-05 Database: Accounting&Tax

_______________________________________________________________ Document 10 of 10 The provision of non-audit services and earnings conservatism: Do New Zealand auditorscompromise their independence? Author: Zhang, Beilei; Emanuel, David Publication info: Accounting Research Journal 21. 2 (2008): 195-221.ProQuest document link Abstract: Purpose - This study seeks to examine the relation between non-audit services provided by incumbentauditors and an important aspect of earnings quality - conservatism, using New Zealand data. Conservatism isdefined as the adoption of accounting policies that accelerate expenses towards the current period and/or deferrevenues to later periods. If the provision of non-audit services undermines auditor independence andencourages the condoning of clients' aggressive accounting practices, this will be revealed by a reduction inaccounting conservatism. Design/methodology/approach - The method aims to determine whether Basu'sconservatism is affected by the level and proportion of non-audit services, where the dataset is partitioned intomultiple levels of these services. Findings - Using 528 firm-year observations, evidence is presented for theexistence of earnings conservatism in New Zealand. However, no evidence of a negative association between

08 March 2013 Page 130 of 145 ProQuest

non-audit services and earnings conservatism is found. Research limitations/implications - Although widelyaccepted in the conservatism literature, the Basu model still suffers from some drawbacks. Issues ofendogeneity may also contribute to the insignificant results. Practical implications - The results are consistentwith factors such as reputational penalties and litigation risk constraining auditor behaviour. Auditors in NewZealand may therefore still maintain their independence "in fact", irrespective of the level of non-audit fees(NAFs) purchased by clients. Originality/value - This paper contains the first tests of conservatism using NewZealand data. It adds to the body of knowledge about the relationship, or the lack thereof, between NAF anddesirable attributes of accounting. Full Text: 1 Introduction The recent collapse of several major corporations in the USA and elsewhere has led tothe allegation that auditors of these companies failed to detect and/or persuade their clients to recogniseeconomic losses in a timely fashion ([52] Wall Street Journal , 2002). In a number of cases, the auditor providedsubstantial consulting services to the troubled client. Given this, critics have been concerned that the financialdependence and responsibility that accompany the provision of non-audit services[1] may reduce the ability ofauditors to perform objective monitoring. As auditor independence is crucial in maintaining the credibility offinancial reporting, recent research has examined the relationship between non-audit services and clients'earnings quality. However, the evidence to date remains inconclusive. While, [28] Frankel et al. (2002) and [21]Dee et al. (2001) suggest that the provision of non-audit services leads to more aggressive accountingpractices, other studies ([17] Chung and Kallapur, 2003; [47] Reynolds et al. , 2004; [22] DeFond et al. , 2002)contradict their findings. There are several motivations for this particular study. The first is that issuesassociated with the role of non-audit fees (NAFs) are only going to be resolved by the weight of evidence, ratherthan by one study. The second is that there is, as far as we are aware, only one prior study looking at this issueusing New Zealand data ([13] Cahan et al. , 2008). The third is that New Zealand data on audit and NAFs havebeen available since the mid-1990s and therefore there is a panel of data available for analysis. The last year ofdata is 2001, the year before Andersen was indicted over the Enron affair. The fourth is that New Zealand'ssmall size can be an advantage if the argument is advanced that this smallness and proximity could lead tocloser interaction between auditors and clients. For example, there is only one professional accounting body,and there are only two major cities which are the sites for most audits. Finally, many prior studies focus onattributes of accounting that are suggestive of earnings management, but the methods used to identify earningsmanagement have relatively low-explanatory power. Rather than using earnings management proxies, thisstudy examines whether the provision of non-audit services by incumbent auditors is associated with areduction in one aspect of earnings quality - conservatism. Conservatism is the application of a higher degree ofverification for favourable information, via the adoption of accounting policies that accelerate expenses to thecurrent period, and/or defer revenues to later periods ([7] Basu, 1997). Conservatism fills an ex ante efficientrole in contracting between the firm and its stakeholders. [7] Basu (1997) and [54], [55] Watts (2003a, b) arguethat conservatism assists in monitoring important contracts involving the entity and its stakeholders (for exampledebt contracts, and contracts associated with remuneration), and is a central feature of corporate governance([5] Ball et al. , 2000). In the context of this paper, conservatism is a desirable attribute of accountinginformation. Financial statements are the result of both management's representations and the auditor'sassurance to outsiders about the validity of those representations ([39] Krishnan, 2005). If non-audit servicesincrease auditors' economic dependence on clients and this impairs independence, auditors will be less willingto challenge accounting that might be "aggressive". If so, then this will be observable through a reduction inconservatism. Therefore, it is hypothesised that clients purchasing higher levels of non-audit services will tendto report earnings of lower quality, which means lower levels of conservatism. The empirical tests used toconduct this study require measurement of the economic bond that non-audit services create, as well asmeasures that capture variation in earnings conservatism. In regard to the former, we use two proxies that havebeen commonly used in the literature. The first is the ratio of NAFs to total fees (non-audit fee ratio - NAFR),

08 March 2013 Page 131 of 145 ProQuest

which captures the relative value of the audit versus non-audit services provided by the auditor. Although thismeasure is of interest to investors and regulators, it may not reflect the economic importance of the client to theauditor ([28] Frankel et al. , 2002; [4] Ashbaugh et al. , 2003). A second measure, the dollar value of NAFs, isalso employed. In regard to measures that capture variation in earnings conservatism, the [7] Basu (1997)model is adopted. This will provide some evidence on the existence of earnings conservatism in New Zealand.Earnings are regressed on contemporaneous stock returns. That is returns equals economic income, and thisreflects good (bad) news for positive (negative) economic income. To test the main hypothesis, in a mannersimilar to [48] Ruddock et al. (2006), a modified [7] Basu (1997) model is then introduced to capture theincremental effect of high NAFR (or NAF) on earnings conservatism. Using this model as a base, later modelsinclude three control variables, Big N auditors, client size, and leverage, to account for possible confoundingeffects on conservatism. Based on 528 firm-year observations from 1995 to 2001, this study provides evidencefor the existence of earnings conservatism in New Zealand. This is consistent with the findings of [5], [6] Ball etal. (2000, 2003), which suggest that earnings in common law countries display significant conservatism. As faras we are aware this is the first evidence of [7] Basu (1997) conservatism in New Zealand. The insignificance ofkey coefficients in the models when NAFR and NAFs are added does not support the idea that NAFs affect thedegree of accounting conservatism. Additional tests provide qualitatively similar results to the main analyses.Overall, higher levels of non-audit services are not found to be associated with reduced earnings conservatism.One plausible explanation for this finding is that auditors in New Zealand have market-based incentives tomaintain their independence. Concerns about reputation and litigation may serve to discipline auditors,preventing them from compromising their independence to protect NAFs. Auditors providing high levels of non-audit services may therefore still maintain their independence "in fact", requiring the same degree ofconservatism from all clients regardless of the level of non-audit services purchased. Of course, weacknowledge that a study that fails to reject the null hypothesis is open to a number of alternativeinterpretations. The remainder of the paper is organised as follows. Section 2 briefly reviews prior literature inthe areas of non-audit services, auditor independence and earnings conservatism, while Section 3 develops thestudy's main hypothesis. Section 4 describes the sample selection and Section 5 provides descriptive statisticsfor the data. Section 6 discusses the research design. Section 7 summarises the results of the main analyses,while sensitivity analyses are presented in Section 8. Section 9 concludes. 2 Literature review 2.1 Auditorindependence Auditors aid investor-management contracting by attesting to the reliability of management'sfinancial reporting. An audit is an efficient monitoring device that increases a firm's value by mitigating agencyproblems that arise from these contracts ([36] Jensen and Meckling, 1976). However, this monitoring will onlybe valuable if the auditor is independent ([57] Watts and Zimmerman, 1986). According to [56], [57] Watts andZimmerman (1983, 1986), auditor independence is "the probability that the auditor will report a discoveredbreach in the financial reports". An independent auditor will therefore keep an objective perspective and be ableto withstand clients' pressure to accept substandard reporting. As this is the foundation for trust in theattestation function, auditor independence is often considered to be the cornerstone of the auditing profession.However, recent accounting scandals such as at Enron, WorldCom, Adelphia, HIH, and Global Crossing haveput the issue of auditor independence in the newspaper headlines. This has been particularly so as the firm ofArthur Andersen was auditor of a disproportionately large number of failed firms, and it has been noted thatArthur Andersen earned very large amounts from NAFs from at least some of these firms. Since capital marketsneed to make investment decisions with the assumption of auditor independence, the question that arises fromthese scandals is whether the auditor is truly independent. In response to negative publicity surrounding majoraudit failure, the Securities and Exchange Commission - [50] SEC (2000) reinforced the importance ofindependence by defining it as the absence of interests that may cause bias in appearance[2] as well as infact[3] . This was followed by the Sarbanes-Oxley Act of 2002, which banned a wide range of services thathistorically auditors may have provided. The [42] New Zealand Institute of Chartered Accountants (NZICA)

08 March 2013 Page 132 of 145 ProQuest

(2005) offers a similar definition of auditor independence - independence in "mind" and in "appearance". Notonly must auditors be independent in practice, but they should also be seen to be independent[4] . Accountingscandals in the USA had an effect on several New Zealand listed companies and their accounting serviceproviders. In 2002, Telecom New Zealand Limited announced the separation of its auditing and consulting work.PricewaterhouseCoopers continued to consult on tax and financial services, but Telecom's auditing contractwas given to KPMG. As [25] Edlin (2002) commented, "the move has been made in response to confidence-shaking concerns raised by an epidemic of accounting scandals in the United States and other countries". Thiswas also the case at The Warehouse Group Limited, where "global pressure" led Ernst and Young to give up itsexternal auditing contract with the group. The New Zealand Shareholders' Association had claimed that auditorindependence was at risk because external auditors were permitted to perform internal audit work ([16]Chapple, 2003). Further, some New Zealand audits are now directly affected by the provisions of the Sarbanes-Oxley Act. Others are affected through the development of new independence policies within the globalaccounting firms. As well, in the light of these events, some New Zealand companies have accepted theprovisions of Sarbanes-Oxley as reflecting best practice. Both the "mind" and "appearance" dimensions ofindependence can significantly affect the ability of an audit to mitigate agency costs. Users may place little valuein the audit opinion if they perceive low independence, even if independence has been maintained "in fact" ([14]Chan et al. , 2002). While independence in appearance is an important issue in its own right, most research hasaddressed the determinants of independence impairment "in fact". Several researchers have examined thepotential link between NAFs and auditor independence, but the empirical evidence is mixed. Some researchershave examined whether the provision of independent audit services creates an economic bond that jeopardizesauditor independence. Others have taken this further by investigating the relationship between non-auditservices and clients' earnings management using different proxies. Each of these avenues of research is brieflydiscussed below. 2.1.1 Non-audit services and auditor independence Both the agency and the behaviouralliterature suggest that, intentionally or unconsciously, auditors are more likely to accede to a client's wishes,including pressure to consent to earnings management, when the provision of non-audit services is substantial([28] Frankel et al. , 2002). The original paper in this area is [51] Simunic (1984). He finds that an increase in theprovision of non-audit services is associated with a significant increase in audit fees. He argues that the jointprovision of audit and non-audit services creates "knowledge spillovers[5] ", which may strengthen an auditor'seconomic bond with the client. In response to the view that the information gained from non-audit servicescomplements the audit function and hence improves financial reporting quality, Simunic claims that thisinformation advantage would not enhance the quality of financial reporting if the economic bonding is significantenough to compromise the auditor's independence. He argues that this will be especially the case when theperformance of audit and non-audit services is "tied" together and when the specialised resources from the non-audit services cannot be transferred to an equally profitable alternative use. Regulators and standard-settersoffer similar arguments. In particular, the SEC is concerned about two effects that non-audit services may havein threatening auditor independence and financial reporting quality. It is argued that the provision of non-auditservices may make auditors more financially dependent on their clients, and hence increase the likelihood theauditor will not draw attention to earnings management, fearing the loss of lucrative fees. The Public OversightBoard - [45] POB (2000, p. 121) of the American Institute of Certified Public Accountants addressed this issue inits Report and Recommendations from the Panel on Audit Effectiveness - "prospective revenues from theprovision of non-audit services, extending into the future, create precisely the kind of financial stake thatproduces a conflict of interest capable of impairing independence". Second, difficulties may arise if the auditorperforms two roles that have the potential to be in conflict. It is argued that consulting services can put auditorsin managerial roles, potentially threatening their ability to monitor the client objectively. While the agencyliterature views auditor bias as intentional, the behavioural literature suggests that auditors unconsciously biastheir judgements. [10] Beeler and Hunton (2001) investigate the effect of contingent economic rents on

08 March 2013 Page 133 of 145 ProQuest

professional judgement and impairment of auditor independence. Auditors appear motivated to develop andmaintain ongoing relationships with clients in order to realize future economic rents, yet still believe that theirprofessional judgement is objective and unbiased. Beeler and Hunton conclude that independence impairmentis typically the result of subconscious biases that distort audit evidence and its interpretation, rather thandeliberate opportunistic behaviour. A competing viewpoint suggests that concerns about reputation and thethreat of lawsuits provide auditors with incentives to be independent. Therefore, the provision of non-auditservices will not compromise auditor independence; instead it may strengthen audit quality and reduce clients'earnings management. [3] Arruñada (1999) claims that the provision of non-audit services increases auditors'investment in reputational capital. Using arguments similar to those of [18] DeAngelo (1981), he shows thatconcern about reputation constrains the behaviour of audit firms. This is because the gains from agreeing to anyone client's demands are outweighed by the loss of reputational capital that would be imposed by other clientswho value a reputation for independence. Similarly, [24] Dopuch et al. (2003) demonstrate that this increase inreputational capital will increase auditor independence and audit quality, provided that the probability ofmisstatement is low. Further, [43] Palmrose (1988) finds that "Big 5" auditors reduce litigation exposure byincreasing their independence. This finding is consistent with [27] Francis et al. (1999) and [9] Becker et al.(1998), who show that "Big 5" auditors appear to constrain clients' accounting discretion. 2.1.2 Auditorindependence, NAF and earnings management The above competing arguments derive from a cost-benefittrade-off - whether the economic rents generated by retaining clients dominate the expected costs of sacrificingauditor independence[6] . But it is not possible to quantify directly the benefits or costs. Therefore, it is ultimatelyan empirical question whether auditors compromise their independence in order to retain non-audit serviceclients, and to what extent the credibility of financial reports is eroded ([22] DeFond et al. , 2002). Over the lastfour or five years an extremely large number of papers have examined this issue. In this summary we brieflymention some of them. [28] Frankel et al. 's (2002, henceforth FJN) analysis shows that the extent of NAF isassociated with several proxies for clients' earnings management. In particular, they find that firms purchasinghigher levels of non-audit services are more likely to report small earnings surprises and larger discretionaryaccruals and are better at meeting analysts' earnings forecasts. Because these clients appear to manageearnings to a greater extent, FJN conclude that the provision of non-audit services adversely impacts auditorindependence. However, subsequent studies have challenged their conclusions. The criticisms include thefailure of FJN to include a firm performance variable in the discretionary accruals model ([4] Ashbaugh et al. ,2003), failure to account for the endogeneity of firm's reporting and their decisions to purchase services fromauditors ([1] Antle et al. , 2006), and failure to control for industry effects ([17] Chung and Kallapur, 2003). [47]Reynolds et al. (2004) conclude that FJN's results are primarily due to small to medium firms, firms having initialpublic offerings, and firms in the e-commerce, biomedical, telecommunications and pharmaceuticals industries.[40] Larcker and Richardson (2004) attribute FJN's results to a small subset of their data. The drawbacks ofearnings management proxies are well documented in the accounting choice literature. A popular proxy forearnings management is the level of discretionary accruals, since it represents managers' discretionaryaccounting policy choices, which in turn could be motivated by self-interest. However, abnormal accruals are anoisy proxy for earnings management ([34] Healy and Wahlen, 1999; [20] Dechow et al. , 1995). Studies thatexamine firms just beating (or meeting) earnings benchmarks also face significant problems. For example, firmsmay meet benchmarks through means other than earnings management, such as improvements in operations([19] Dechow et al. , 2003). Non-earnings management proxies also have limitations. Using auditors' propensityto issue going concern opinions and restatements restricts samples to distressed or restating firms, potentiallyreducing external validity. 2.2 Earnings conservatism Conservatism has influenced accounting practice andtheory for more than 500 years ([7] Basu, 1997). Recent research documents not only the existence ofconservatism, but also its rise in the last four decades ([54] Watts, 2003a; [32] Givoly and Hayn, 2000).Accounting conservatism has been traditionally expressed by the rule "anticipate no profits but anticipate all

08 March 2013 Page 134 of 145 ProQuest

losses" ([11] Bliss, 1924). [7] Basu (1997) interprets this as representing "the accountant's tendency to require ahigher degree of verification to recognize good news as gains than to recognize bad news as losses."Differential verification produces earnings that reflect bad news in a more timely fashion than good news. Itfurther implies that the effect of bad news on the earnings time-series will tend to be transitory, whereas positiveearnings changes will tend to persist. The Framework for the Preparation and Presentation of FinancialStatements Section 37 ([35] International Accounting Standards Board, 2005) provides an official definition ofthe concept - conservatism (prudence) requires preparers to exercise a degree of caution when makingestimations of uncertain transactions and events, "to ensure that assets or revenues are not overstated andliabilities or expenses are not understated." Explanations for the existence of conservatism point to the benefitsit provides users of financial reports. As a desirable attribute of accounting earnings, conservatism plays an exante efficient role in contracting between the firm and its stakeholders ([7] Basu, 1997; [54], [55] Watts, 2003a,b). They have limited horizons and limited liability, and as a consequence asymmetric payoffs. Conservatismproduces accounting numbers that can be used in contracts among various parties to alleviate the conflictscreated by these asymmetries ([54] Watts, 2003a). Shareholder litigation is another explanation forconservatism. Overstatements of earnings or assets are more likely to generate litigation costs for the firm thanunderstatements. Owing to this asymmetry, managers and auditors have incentives to adopt conservativeaccounting so as to reduce the firm's expected litigation costs. This litigation explanation suggests that non-contracting parties in society also value conservatism's constraint on managers' dysfunctional behaviour ([54]Watts, 2003a). And taxation and regulation are two additional explanations for conservatism ([49] Shackelfordand Shevlin, 2001; [54] Watts, 2003a). The importance of accounting conservatism calls for empirical researchon the topic. In re-examining the conservatism principle, [7] Basu (1997) develops a reverse regression modelthat estimates earnings timeliness towards positive and negative stock returns. Annual earnings are regressedon current stock returns, where positive (negative) returns proxy for good (bad) news[7] . His results areconsistent with the existence of conservatism in US financial reporting. Basu's model has been widely used byresearchers in the area. As commented by [5] Ball et al. (2000), Basu's reverse regression captures earnings'timeliness in reflecting publicly available news, and hence has implications for earnings quality and reportingtransparency. In their study, they predict that Basu measures of conservatism vary across common lawcountries (the USA, the UK, Australia and Canada) and code-law countries (France, Germany and Japan)according to their different institutional factors[8] . Consistent with this prediction, [5] Ball et al. (2000) find thataccounting income in common law countries is significantly more conservative than in code law countries. 3Hypothesis development Critics of the profession have alleged that the high margin of NAFs increases theauditor's economic dependence on the client firm, which in turn increases the auditor's incentives to agree toclients' wishes and condone aggressive accounting. Where there is an increase in the probability of an auditornot truthfully reporting the results of the audit investigation, there is a threat to independence ([51] Simunic,1984). Therefore, the provision of non-audit services may jeopardize auditors' independence, which could inturn reduce audit quality and ultimately result in reported earnings of lower quality. This study examines theimpact of non-audit services on a specific attribute of audited financial statements - earnings conservatism. It issimilar in motivation to [48] Ruddock et al. (2006), although it uses different proxies for economic bond andeconomic income, and is applied to a different data set[9] . To the extent that non-audit services impair auditorindependence, auditors are likely to be less willing to challenge clients' aggressive accounting. This will berevealed by a reduction in accounting conservatism, as measured. Therefore, clients purchasing higher levels ofthese services are expected to report earnings of lower quality, and in particular earnings of lower conservatism.This reasoning leads to the main hypothesis: H1. The higher the level of non-audit services provided by theincumbent auditor, the less conservative the client's accounting practices will be. 4 Sample selection and data The basic data set is provided by the Department of Accounting and Finance at The University of Auckland. Theinitial sample consists of 814 firm-year observations from 1995 to 2001, representing 176 New Zealand

08 March 2013 Page 135 of 145 ProQuest

companies. For each firm-year, the data file provides information on the firm's auditors, including fees paid foraudit and non-audit services, as well as key measures of financial performance and position. In order tocalculate the market return of each firm, annual adjusted share prices[10] are obtained from DataStream. Firm-years with insufficient data to estimate the empirical models are eliminated. To remove the effect of outliers,observations are winsorized to the 1 and 99 per cent levels. Financial institutions and regulated firms are alsoexcluded, as these firms may have different financial reporting incentives ([12] Burgstahler and Dichev, 1997).The final sample consists of 528 observations, representing 118 New Zealand listed companies. 5 Researchdesign 5.1 Economic dependence This study uses two measures to proxy for the economic bond between aclient and its auditor. The first is the ratio of NAF to total fees paid by the client (NAFR), as used by FJN andseveral other researchers. This measure captures the relative value of the audit versus non-audit servicesprovided by the audit firm to the client, which may affect investors' and regulators' perceptions of independence.To the extent that non-audit services have a higher profit margin, it is alleged that auditors are more concernedwith avoiding the loss of audit clients that generate relatively large amounts of NAF, and use audits as aplatform from which they sell non-audit services[11] . This specification is, however, subject to a number ofcriticisms. Specifically, this ratio does not capture the financial importance of the client to the audit firm ([28]Frankel et al. , 2002; [4] Ashbaugh et al. , 2003). Although the Enron-Arthur Andersen scandal is a frequentlycited example of significant client-auditor economic bonding, as pointed out by [38] Kinney and Libby (2002)and others, an analysis of Enron's fee disclosures (audit fees of $US 25 million and NAF of $US 27 million)would result in a NAFR of only 0.52, which is close to the median (0.51) reported by FJN in Table II [Figureomitted. See Article Image.]. Therefore, the second measure for economic dependence is the dollar amount ofNAF paid by each client to its auditor[12] . This measure is specified to incorporate the payment size, and thusattempts to capture the clients' financial importance to the audit firm. 5.2 Measuring conservatism 5.2.1 Model1: the Basu model, applied to pooled data and sub-samples Following [7] Basu (1997), the first test utilizes a"reverse" regression of annual earnings on contemporaneous stock returns, which is stated as follows: Equation1 [Figure omitted. See Article Image.] where, NIit - net income after tax for firm i in fiscal year t , deflated by themarket capitalisation at the beginning of the fiscal year; RETit - annual stock return on firm i for fiscal year t ,that is (Pt -Pt -1 )/Pt -1 , where prices have already been adjusted for dividends; DRETit - 1 if RETit ≤0, and 0otherwise. Model 1 is first estimated for the pooled sample, and then separately for firm-years with high NAFRs(NAFR values in the top third[13] of the pooled sample) and low NAFRs (NAFR values in the bottom third of thepooled sample). Using the [7] Basu (1997) model, both the pooled data, and sub-samples, can provide insightinto the existence of earnings conservatism in New Zealand. Positive (negative) annual stock returns are usedas a proxy for good (bad) news. Assuming that news is uncorrelated through time, the intercept α0 captures thepostponed recognition of news from previous periods. Conservative accounting implies that α0 is positive, asmost unrealised losses are captured immediately. Little bad news is postponed to future periods, while therecognition of gains is delayed. ß0 measures the responsiveness (sensitivity) of earnings to current good news.The interactive slope coefficient (ß1 ), which is of interest here, represents the differential sensitivity of earningsin recognising bad news over good news. Conservatism means that earnings are more concurrently sensitive inincorporating economic losses relative to economic gains, thus implying that ß1 is positive. 5.2.2 Model 2:modified Basu model including the incremental effect of high NAFRs on earnings conservatism, applied to sub-samples To capture the incremental effect of high NAFRs on earnings conservatism, additional intercept andslope coefficients are included. A dummy variable (DNAFR) is introduced, denoting the level of firm-years'NAFRs in the pooled sample distribution. Based on the former classification, firm-years with a NAFR in the topthird of the pooled sample have a value of one for DNAFR, while firm-years with a NAFR in the bottom thirdtake a value of 0. Firm-years in the middle third are excluded from the analysis. The following modifiedregression model is estimated: Equation 2 [Figure omitted. See Article Image.] where, DNAFR it - 1 for firm-years with NAFRit in the top third of the pooled sample, and 0 for firm-years with NAFRit in the bottom third.

08 March 2013 Page 136 of 145 ProQuest

The interactive coefficients ß3 and ß2 , which are the focus here, measure the incremental responsiveness ofearnings to bad news and good news when NAFR is high. If conservatism is reduced when the ratio of NAF tototal fees is high, ß3 is expected to be significantly negative. Also, if the responsiveness of earnings to goodnews is increased when NAFR is high, consistent with reduced asymmetrical timeliness, ß2 is expected to besignificantly positive. 5.2.3 Model 3: modified Basu model including the incremental effect of high NAFRs onearnings conservatism and control variables, applied to sub-samples Prior literature identifies several factorsthat may influence a firm's earnings conservatism. To account for possible confounding effects on earningsconservatism, three control variables are introduced. First, [8] Basu et al. (2001) document that Big 8 clientsreport earnings that are more conservative than non-Big 8 clients, consistent with Big 8 auditors being moreconservative in order to protect their brand name reputations. Likewise, [9] Becker et al. (1998) find that Big 6auditors constrain accruals-based earnings management more than non-Big 6 auditors. Therefore, Big N (nowBig 4[14] ) auditors are expected to enhance earnings conservatism by constraining aggressive accruals andpersuading clients to report economic losses in a timely fashion. A dummy variable (BIGN) is included to controlfor the impact that Big N auditors may have on the conservatism of reported earnings. BIGN takes the value of1 if the firm's auditor is one of the Big N , and 0 otherwise. Second, the size hypothesis predicts that larger firmswill tend to make income-decreasing accounting choices, so as to reduce their political visibility ([53] Watts andZimmerman, 1978)[15] . A control variable for client size, proxied by the natural logarithm of firms' total assets(in $'000), is employed to take account of this effect. Third, firms facing severe debtholder-shareholder conflictsover dividend policy are inclined to adopt more conservative accounting ([2] Ahmed et al. , 2002). Following [2]Ahmed et al. (2002), leverage, measured as the ratio of long-term debt to total equity, is used as a proxy for theseverity of debtholder-shareholder conflicts. Firms with higher leverage are expected to have greater earningsconservatism. Including the above three control variables, the regression model is now estimated as follows:Equation 3 [Figure omitted. See Article Image.] where, BIGNit - 1 if the firm i 's auditor is a Big N audit firm inyear t ; and 0 otherwise; SIZEit - natural logarithm of firm i 's total assets in year t ; LEVit - ratio of firm i 's long-term debt to its total equity in year t . All control variables are interacted with the test variables in Model 2. Thisnot only accounts for the effects those control variables may have on the intercept, but also allows the slopes tovary. Consistent with prior literature that larger firms and higher leveraged firms are predicted to have moreconservative earnings, the interactive coefficients on these three control variables ( γ3 , δ3 , λ3 , respectively)are expected to be positive. 5.2.4 Models 4 and 5: replication of Models 2 and 3 using a different proxy foreconomic bonding (NAF), applied to sub-samples As previously mentioned, although NAFR is of interest toinvestors and regulators, it may not capture the importance of clients to the audit firm. Therefore, anothermeasure of economic dependence (NAF) is employed, and the above regressions (Models 2 and 3) arereplicated using this new measure. Firm-years are re-ranked according to the magnitude of NAF, and a dummyvariable (DNAF) takes a value of 1 if a firm-year's NAF is in the top third of the pooled sample, and 0 if it is inthe bottom third. Firm-years in the middle third are excluded from the analysis. With the variables defined asbefore, Model 1 is re-estimated using this new classification of sub-groups, and Models 2 and 3 are re-specifiedas follows: Equation 4 [Figure omitted. See Article Image.] Equation 5 [Figure omitted. See Article Image.]where, DNAF it - 1 for firm-years with NAFit in the top third of the pooled sample, and 0 for firm-years with NAFitin the bottom third. 6 Descriptive statistics Descriptive statistics for the pooled sample are reported in Table I[Figure omitted. See Article Image.]. The median of the dependent variable, net income deflated by thebeginning-of-the-year market capitalization (NI), is 0.063. The first quartile (0.014) of deflated net income showsthat the large majority of the firm-years report a surplus. However, some firm-years experienced severe losses,which reduces the mean to a value of 0.016. The median annual stock return is 0, telling us that half of all firm-years encounter economic losses. Table I [Figure omitted. See Article Image.] also shows considerablevariation in the consumption of non-audit services. Eighteen percent of all firm-years have zero NAF, indicatingthat the auditor provided no additional services[16] , while the maximum fee is $5 million. On average, firm-year

08 March 2013 Page 137 of 145 ProQuest

NAF are $197,860, while the median fee is only $35,393. Some firm-years have substantial fee levels, whichpushes the mean well above the third quartile ($135,000). This variation makes the NAF proxy for economicdependence particularly interesting, as the sub-sample design will capture the significant differential betweenlow and high NAF levels. NAFR is fairly evenly distributed, with a mean of 0.331 and median of 0.310. Both themean and median ratios are lower than those reported in FJN (0.331 vs 0.49 and 0.310 vs 0.51, respectively),suggesting that the provision of non-audit services is less prevalent in New Zealand than in the USA. Summarystatistics for the three control variables are also presented in Table I [Figure omitted. See Article Image.]. Themean for BIGN is 0.887, showing that Big N auditors dominate New Zealand's listed firm market. The mean forclient size (SIZE) and leverage (LEV) are 11.713 and 0.477, respectively. Correlation coefficients are presentedin Table II [Figure omitted. See Article Image.], along with a χ2 statistic for the two binary variables, DNAFR (orDNAF) and BIGN. Panel A contains the results for the sub-samples when using NAFR as a proxy for economicdependence, while Panel B reports the correlations for the sub-samples when using the alternative NAF proxy.In Panel A, the point biserial correlation coefficient shows that the dummy variable for high and low NAFRs(DNAFR) is positively and significantly correlated with RET and SIZE. SIZE is positively correlated with BIGNand LEV, as expected. A similar pattern of correlations is observed in Panel B. Despite many independentvariables being significantly correlated, only the association between DNAF and SIZE is large in economicterms (point biserial correlation of 0.60). Large firms use the consulting services that the auditor has availablemore than small firms do, in part as non-BIGN auditors are more constrained in the range of services that theycan provide. In this setting, multicollinearity is unlikely to be a problem in the regression models. 7 Results Table III [Figure omitted. See Article Image.] reports the regression results for Model 1 (for both pooled and subsamples). Table IV [Figure omitted. See Article Image.] reports the results for Models 2 and 3 using NAFR toproxy for the economic bond between an auditor and its client, while Table V [Figure omitted. See ArticleImage.] presents the results for Models 4 and 5 using the alternative NAF proxy. Recall that Model 1 is a simple"reverse" regression of earnings on contemporaneous stock returns, with a dummy variable that equals one fornegative returns. Model 2 (4) includes an additional dummy variable DNAFR (DNAF) that accounts for thoseinstances where NAFR (NAF) is relatively high. Using Model 2 (4) as a base, Model 3 (5) includes three furthervariables to control for possible confounding effects on earnings conservatism. In Table III [Figure omitted. SeeArticle Image.], regression results for Model 1 are presented initially for the pooled sample, and then separatelyfor firm-years with high and low values of NAFR (NAF). Evidence from the full sample of firm-years is consistentwith earnings conservatism. The intercept ( α0 ), and the incremental coefficient on negative stock returns (ß1 )are both positive and highly significant. The significantly positive intercept (α0 ) indicates that the recognition ofunrealized gains is postponed to future periods, and the significantly positive interactive slope coefficient (ß1 )suggests a differential treatment of economic losses. The magnitude of ß1 is far greater than ß0 (0.444 vs0.093), showing that earnings is about eight times (7.7=[0.444+0.093]/0.093) as sensitive to negative returns asit is to positive returns. Overall, these results suggest that, in New Zealand, earnings are more responsive tocontemporaneous bad news. This is consistent with international studies which find that earnings in commonlaw countries display significant conservatism ([5], [6] Ball et al. , 2000, 2003). When Model 1 is estimatedseparately for firm-years with high and low NAFR (NAF), conservatism is evident in both sub-samples. Underboth economic bond proxies, the coefficient ß1 is positive and highly significant, indicating that earnings is moretimely in recognising economic losses irrespective of the level of NAF. Panel A of Tables IV and V [Figureomitted. See Article Image.] summarises the regression results for Models 2 and 4. As discussed earlier, if firmsconsuming higher values of NAFR (NAF) report less conservative earnings, ß3 is expected to be significantlynegative and ß2 is expected to be significantly positive[17] . The signs of the interactive slope coefficients ß3and ß2 are mixed, and none of the estimates is significantly different from zero. ß1 continues to be positive andsignificant, indicating an asymmetrically higher responsiveness to bad news. Panel B of Tables IV and V [Figureomitted. See Article Image.] presents the regression results for Models 3 and 5. The inclusion of control

08 March 2013 Page 138 of 145 ProQuest

variables increases the explanatory power of the models, with adjusted R2 rising above 20 per cent. In bothcases ß2 and ß3 are statistically insignificantly different from 0. The interactive coefficient for leverage (λ3 ) issignificantly positive, which is consistent with the prediction that highly leveraged firms may adopt moreconservative accounting to reduce shareholder-debtholder conflicts. Overall, the insignificant results for ß3 andß2 in Models 2 to 5 fail to support the main hypothesis. No evidence is found for an incremental reduction inconservatism where NAFR or NAF is relatively high. This finding is consistent with other recent research, suchas [17] Chung and Kallapur (2003), [47] Reynolds et al. (2004), and [48] Ruddock et al. (2006) which fail todocument a negative association between NAF and earnings quality. 8 Sensitivity analyses The primary testsare based on firm-years that include no purchase of non-audit services from the incumbent auditor. With zeronon-audit fee observations comprising 18 per cent of the pooled sample, most "low NAFR (NAF)" sub-samplefirm-years are more accurately described as "zero NAFR (NAF)". To examine whether this specification drivesthe insignificant results of the main tests, additional analyses are conducted. Regressions of Models 2 to 5 arere-run after the elimination of zero non-audit fee observations. The results for the NAFR and NAF sub-samplesare presented in Tables VI and VII [Figure omitted. See Article Image.], respectively. The coefficients of interest,ß3 and ß2 , remain insignificant except for ß2 in Panel B of Table VI [Figure omitted. See Article Image.], butthat estimate is negative. Interestingly, ß3 in Table VII [Figure omitted. See Article Image.] Panel B is evensignificantly positive, which suggests that auditors constrain clients' aggressive accounting practice when theprovision of non-audit services is high. ß1 is still positive but is significantly different from zero in only two of thefour regressions shown in these tables. To further check the robustness of the primary results, the secondsensitivity analyses are conducted by changing the classification of high/low NAFR/NAF. In the primary tests,firm-years in the top (bottom) third of NAFR/NAF distribution are compared to see if the high level of provision ofnon-audit services leads to clients' less conservative accounting practices. This comparison now becomesstricter by replacing the classification of high and low NAFR/NAF based on the upper and bottom quartile andquintile of the NAFR/NAF distribution. These involve some extreme comparisons and the effect of the level ofthe provision of non-audit services have on the clients' accounting practices might be observed. For example,firm-years in the lower quintile are nearly "zero NAFR/NAF" observations. Models 2 to 5 are re-run based on thenew classification of high/low NAFR/NAF and results are reported in Table VIII [Figure omitted. See ArticleImage.]. Nevertheless, ß3 and ß2 are still insignificant, except in one case, where ß2 is significantly positiveusing the Model 2 regression based on the observations in the upper/lower quintile of the NAFR/NAFdistribution. Overall, the results of these additional analyses are qualitatively similar to the main test results,suggesting that the primary analyses are not sensitive to the inclusion of zero NAF observations[18] . Finally,we also identified all audit qualifications in the dataset, as it could be argued that large NAF revenues mightinduce a lower level of audit qualification than might otherwise be the case. There are only three qualifications inthe "high" NAF group and seven in the low-NAF group, so we are dealing with extremely small numbers.Further, the high-NAF group of companies is significantly more profitable and has significantly higher returns onaverage than the low-NAF group. Therefore, we cannot attribute the lower level of qualifications in the high-NAFgroup to the NAF. 9 Conclusion This study examines the relation between non-audit services provided byincumbent auditors and one important aspect of earnings quality - conservatism. Conservatism is defined as theasymmetric recognition of economic losses. It plays an important role in the efficient monitoring of contracts,and helps mitigate agency costs in organisations. The provision of non-audit services by incumbent auditors hasbeen subject to much recent criticism. In particular, the economic bond created by these services is alleged tocompromise auditor independence, and hence lower the quality of clients' reported earnings. It is hypothesizedthat a negative relation exists between the provision of non-audit services and clients' earnings conservatism.Two measures derived from the prior literature are employed to proxy for the economic bond between auditorsand their clients. The NAFR relates to the concerns of investors and regulators, while the total of NAF capturesthe client's economic importance to the auditor. Using the [7] Basu (1997) model, evidence is presented for the

08 March 2013 Page 139 of 145 ProQuest

existence of earnings conservatism in New Zealand. This is consistent with international studies which suggestthat earnings in common law countries show significant conservatism ([5], [6] Ball et al. , 2000, 2003). However,a modified [7] Basu (1997) model that includes the incremental effect of high NAFR (NAF) on earningsconservatism fails to support the main hypothesis. The further inclusion of three control variables, Big Nauditors, client size and leverage, also leads to insignificant results. The sensitivity analyses, which eliminatezero NAF observations and change the classification of high/low NAFR/NAF provide results that arequalitatively similar to the main analyses. Overall, this study provides no evidence that higher levels of non-auditservices are associated with reduced earnings conservatism. One plausible explanation is that auditors in NewZealand have market-based incentives to maintain their independence. Concerns about reputation and litigationmay dominate the economic dependence created by higher NAF. Auditors of clients purchasing high levels ofnon-audit services may therefore still maintain their independence "in fact", requiring the same degree ofconservatism from these clients as for those purchasing low levels of non-audit services. Although widelyaccepted in the conservatism literature, the [7] Basu (1997) model still suffers from some drawbacks. Theefficient market hypothesis may not always hold, and clear interpretations of the model are only possible ifreturns cause earnings and not the reverse ([23] Dietrich et al. , 2002). Finally, issues of endogeneity may alsocontribute to the insignificant results. Accounting choices, including the level of conservatism exercised, may beassociated with the same factors that determine a firm's demand for non-audit services. The problem here isthat it is not clear what those factors might be. Hence, this limitation is inherent in all prior research examiningthe relation between non-audit services and firms' accounting choices. It should also be noted that the inclusionof control variables in Models 3 and 5 may mitigate this concern[19] . Nevertheless, there is scope for futureresearch to develop more sophisticated methods[20] to address the issue of endogeneity. Footnote 1. Non-auditservices refer to fee based services performed by the auditing firm which are not related to the auditengagement. They include tax services, management consulting, systems planning, systems design andreview, personnel services, risk management, actuarial services, and acquisition services. ([14] Chan et al. ,2002). 2. The [50] SEC (2000, Section I) will not consider an auditor to be independent with respect to a particularclient "if a reasonable investor, with knowledge of all relevant facts and circumstances, would conclude that theauditor is not capable of exercising objective and impartial judgement". 3. An auditor who is independent in fact has the ability to make independent audit decisions, even if there is aperceived lack of independence. 4. New Zealand Financial Reporting Standards (FRS-9) clause 6.13(e) also mandates disclosure of audit feesand NAF. 5. Knowledge spillovers refer to information generated while performing non-audit services that can produceeconomic rents by reducing auditing costs. 6. As discussed above, the expected costs of sacrificed independence include the loss of reputation andlitigation costs associated with audit failure. 7. This model relies on the "efficient market hypothesis". Stock prices are assumed to efficiently reflect value-relevant information as it becomes public. When this condition holds, stock prices will reflect informationreceived from sources other than current earnings, and may lead accounting earnings. 8. Other studies using the [7] Basu (1997) model to examine cross-country variation in conservatism include[44] Pope and Walker (1999), [31] Giner and Rees (2001), [29] GarciaLara and Mora (2004) and [30]GarciaLara et al. (2005). 9. In the review of [48] Ruddock et al. (2006) by [26] Francis (2006), three major comments are advanced.Francis is critical of Ruddock et al. 's extrapolation of their Australian results to the US audit market. The secondis that he objects to Ruddock et al. 's use of "unexpected" NAF. The third is that Francis is a "skeptic" when itcomes to the [7] Basu (1997) earnings conservatism story. Only the third of those comments applies to this

08 March 2013 Page 140 of 145 ProQuest

paper. 10. Prices are adjusted for subsequent capital actions, including dividend payouts. Share prices at the end offiscal year 1994 are also gathered to calculate annual stock returns for fiscal year 1995. 11. As claimed by a former SEC chairman, "the auditing function is simply being used as a springboard to morelucrative consulting services" ([41] Levitt, 2000). 12. Several researchers use this proxy, including [1] Antle et al. (2006), [22] DeFond et al. (2002), [37] Kinneyet al. (2004) and [47] Reynolds et al. (2004). An alternative approach would be to examine the client'simportance within the auditor's portfolio of clients, by comparing the NAF paid by the client with the audit firm'stotal revenue in that year. However, data on the total fee income of audit firms is not available; hence, thismeaningful measure of individual client significance cannot be calculated. 13. This follows [48] Ruddock et al. (2006). 14. Given the time period covered in this study, top tier auditors are hereafter referred to as the "Big N". N startsas 6 and ends at 5. 15. See, [58] Zmijewski and Hagerman (1981) and [46] Reynolds and Francis (2001) for evidence supportingthis position. 16. This issue will be addressed in the sensitivity analyses. 17. These regressions were replicated using total fees in place of NAF. While NAF are the focus of ourattention, the reason for this replication was to determine the similarity of the results between total and NAF.The (unreported) results also have ß2 and ß3 estimates that are insignificantly different from 0. The regressionswere also replicated using a ratio of NAF to the total of all fees for that auditor in each year with similar(insignificant) results. 18. We also examined all auditor changes. There are 44 such changes in the (814 firm year) database, and 25are retained in the smaller (528 observations) dataset that is used for this research. Only 20 make it throughinto the high- and low-NAF analysis, with ten in each group. Of the ten high-NAF cases, eight show increases inNAFs in the year of the switch. Nine of these ten changes are from one Big 5 to another Big 5 auditor. In allcases changes in both audit and non-audit fees are small. 19. For example, larger firms may be both more likely to report conservative earnings and purchase more non-audit services. The inclusion of client size as an exogenous variable in Models 3 and 5 controls for thispossibility. 20. For example, [1] Antle et al. (2006) uses a simultaneous equations method to overcome endogeneityproblems. References 1. Antle, R., Gordon, E.A., Narayanamoorthy, G. and Zhou, L. (2006), "The joint determination ofaudit fees, non-audit fees and abnormal accruals", Review of Quantitative Finance and Accounting, Vol. 27, pp.235-66. 2. Ahmed, A., Billings, B., Morton, R. and Stanford-Harris, M. (2002), "The role of accounting conservatism inmitigating bondholder-shareholder conflicts over dividend policy in reducing debt costs", The AccountingReview, Vol. 77, pp. 867-90. 3. Arruñada, B. (1999), "The provision of non-audit services by auditors: let the market evolve and decide",International Review of Law and Economics, Vol. 19, pp. 513-31. 4. Ashbaugh, H., LaFond, R. and Mayhew, B. (2003), "Do non-audit services compromise auditorindependence?", The Accounting Review, Vol. 78, pp. 611-39. 5. Ball, R., Kothari, S. and Robin, A. (2000), "The effect of international institutional factors on properties ofaccounting earnings", Journal of Accounting and Economics, Vol. 29, pp. 1-52. 6. Ball, R., Robin, A. and Wu, J. (2003), "Incentives versus standards: properties of accounting income in fourEast Asian countries", Journal of Accounting and Economics, Vol. 36, pp. 235-70. 7. Basu, S. (1997), "The conservatism principle and the asymmetric timeliness of earnings", Journal of

08 March 2013 Page 141 of 145 ProQuest

Accounting and Economics, Vol. 24, pp. 3-37. 8. Basu, S., Hwang, L.S. and Jan, C.L. (2001), "Differences in conservatism between big eight and non-Bigeight auditors", working paper, City University New York, NY and California State University, Hayward, CA. 9. Becker, C., DeFond, M., Jiambalvo, J. and Subramanyam, K. (1998), "The effect of audit quality on earningsmanagement", Contemporary Accounting Research, Vol. 15, pp. 1-24. 10. Beeler, J.D. and Hunton, J.E. (2001), "Contingent economic rents: precursors to predecisional distortion ofclient information", working paper, University of South Carolina, Columbia, SC. 11. Bliss, J. (1924), Management through Accounts, The Ronald Press Co., New York, NY. 12. Burgstahler, D. and Dichev, I. (1997), "Earnings management to avoid earnings decreases and losses",Journal of Accounting and Economics, Vol. 24, pp. 99-126. 13. Cahan, S.C., Emanuel, D.M., Hay, D. and Wong, N. (2008), "Non-audit fees, long-term auditor-clientrelationships, and earnings management", Accounting and Finance, Vol. 48, pp. 181-207. 14. Chan, K., Wong, J. and Wong, N. (2002), "Non audit services and auditor independence", CharteredAccountants Journal of New Zealand, Vol. 81 No. 5, pp. 27-32. 16. Chapple, I. (2003), "Warehouse, Sky City challenged on audits", New Zealand Herald, December 2. 17. Chung, H. and Kallapur, S. (2003), "Client importance, nonaudit services, and abnormal accruals", TheAccounting Review, Vol. 78, pp. 931-55. 18. DeAngelo, L. (1981), "Auditor size and audit quality", Journal of Accounting and Economics, Vol. 3, pp. 183-99. 19. Dechow, P., Richardson, S. and Tuna, I. (2003), "Why are earnings kinky? An examination of the earningsmanagement explanation", Review of Accounting Studies, Vol. 8, pp. 355-84. 20. Dechow, P., Sloan, R. and Sweeney, A. (1995), "Detecting earnings management", The Accounting Review,Vol. 70, pp. 193-225. 21. Dee, C., Lulseged, A. and Nowlin, T. (2001), "Earnings quality and auditor independence: an examinationusing non-audit fee data", working paper, Florida State University, Tallahassee, FL. 22. DeFond, M., Raghunandan, K. and Subramanyam, K. (2002), "Do non-audit service fees impair auditorindependence? Evidence from going concern audit opinions", Journal of Accounting Research, Vol. 40, pp.1247-74. 23. Dietrich, J.R., Muller, K.A. and Riedl, E.J. (2002), "Using stock returns to determine bad versus good newsto examine the conservatism of accounting earnings", working paper, Ohio State University, Columbus, OH. 24. Dopuch, N., King, R. and Schwartz, R. (2003), "Independence in appearance and in fact: in experimentalinvestigation", Contemporary Accounting Research, Vol. 20, pp. 79-114. 25. Edlin, B. (2002), "Like Caesar's wife, Telecom knows virtue lies in the eyes of the beholder", TheIndependent, July 31. 26. Francis, J. (2006), "Are auditors compromised by nonaudit services? Assessing the evidence",Contemporary Accounting Research, Vol. 23, pp. 747-60. 27. Francis, J., Maydew, E. and Sparks, H. (1999), "The role of Big 6 auditors in the credible reporting ofaccruals", Auditing: A Journal of Practice &Theory, Vol. 18, pp. 17-34. 28. Frankel, R., Johnson, M. and Nelson, K. (2002), "The relation between auditors' fees for nonaudit servicesand earnings management", The Accounting Review, Vol. 77, pp. 71-105. 29. GarciaLara, J. and Mora, A. (2004), "Balance sheet versus earnings conservatism in Europe", TheEuropean Accounting Review, Vol. 13, pp. 261-92. 30. GarciaLara, J., Osma, G. and Mora, A. (2005), "The effect of earnings management on the asymmetrictimeliness of earnings", Journal of Business Finance &Accounting, Vol. 32, pp. 691-726. 31. Giner, B. and Rees, W.P. (2001), "On the asymmetric recognition of good and bad news in France,Germany and the United Kingdom", Journal of Business Finance &Accounting, Vol. 28, pp. 1285-331.

08 March 2013 Page 142 of 145 ProQuest

32. Givoly, D. and Hayn, C. (2000), "The changing time series properties of earnings, cash flows and accruals:has financial reporting become more conservative?", Journal of Accounting and Economics, Vol. 29, pp. 287-320. 34. Healy, P. and Wahlen, J. (1999), "A review of the earnings management literature and its implications forstandard setting", Accounting Horizons, Vol. 13, pp. 365-83. 35. International Accounting Standards Board (2005), The Framework for the Preparation and Presentation ofFinancial Statements, Section 37, International Accounting Standards Board, London. 36. Jensen, M. and Meckling, W. (1976), "Theory of the firm: managerial behavior, agency costs and ownershipstructure", Journal of Financial Economics, Vol. 3, pp. 305-60. 37. Kinney, W., Palmrose, Z. and Scholz, S. (2004), "Auditor independence, non-audit services, andrestatements: was the US government right?", Journal of Accounting Research, Vol. 42, pp. 561-88. 38. Kinney, W. and Libby, R. (2002), "Discussion of the relation between auditors' fees for nonaudit servicesand earnings management", The Accounting Review, Vol. 77, pp. 107-14 (supplement). 39. Krishnan, G.V. (2005), "The association between Big 6 auditor industry expertise and the asymmetrictimeliness of earnings", Journal of Accounting, Auditing &Finance, Vol. 20, pp. 209-28. 40. Larcker, D. and Richardson, S. (2004), "Fees paid to audit firms, accrual choices, and corporategovernance", Journal of Accounting Research, Vol. 42, pp. 625-58. 41. Levitt, A. (2000), "Renewing the covenant with investors", speech at The New York University Center forLaw and Business, New York, NY, May 10. 42. New Zealand Institute of Chartered Accountants (NZICA) (2005), New Zealand Financial ReportingStandards (FRS) No. 9 Information to be Disclosed in Financial Statements, Clause 6.13(e), New ZealandInstitute of Chartered Accountants - NZICA, Wellington. 43. Palmrose, Z. (1988), "An analysis of auditor litigation and audit service quality", The Accounting Review,Vol. 63, pp. 55-73. 44. Pope, P. and Walker, M. (1999), "International differences in the timeliness, conservatism, and classificationof earnings", Journal of Accounting Research, Vol. 37, pp. 53-87. 45. POB (2000), Panel on Audit Effectiveness: Report and Recommendations, American Institute of CertifiedPublic Accountants, Public Oversight Board - POB, Stamford, CT. 46. Reynolds, K. and Francis, J. (2001), "Does size matter? The influence of large clients on office-level auditorreporting decisions", Journal of Accounting and Economics, Vol. 30, pp. 375-400. 47. Reynolds, J., Deis, D. and Francis, J. (2004), "Professional service fees and auditor objectivity", Auditing: AJournal of Practice &Theory, Vol. 23, pp. 29-52. 48. Ruddock, C., Taylor, S. and Taylor, S. (2006), "Non audit services and earnings conservatism: is auditorindependence impaired?", Contemporary Accounting Research, Vol. 23, pp. 701-46. 49. Shackelford, D. and Shevlin, T. (2001), "Empirical tax research in accounting", Journal of Accounting andEconomics, Vol. 31, pp. 321-87. 50. SEC (2000), Final Rule: Revision of the Commission's Auditor Independence Requirements, Securities andExchange Commission - SEC, Washington, DC. 51. Simunic, D. (1984), "Auditing, consulting, and auditor independence", Journal of Accounting Research, Vol.22, pp. 679-702. 52. Wall Street Journal (2002), "Accounting faces calls for change", Wall Street Journal, February 6. 53. Watts, R.L. and Zimmerman, J.L. (1978), "Towards a positive theory of the determination of accountingstandards", The Accounting Review, Vol. 53, pp. 112-34. 54. Watts, R.L. (2003a), "Conservatism in accounting Part I: explanations and implications", AccountingHorizons, Vol. 17, pp. 207-21. 55. Watts, R.L. (2003b), "Conservatism in accounting Part II: evidence and research opportunities", Accounting

08 March 2013 Page 143 of 145 ProQuest

Horizons, Vol. 17, pp. 287-301. 56. Watts, R.L. and Zimmerman, J.L. (1983), "Agency problems, auditing and the theory of the firm: someevidence", Journal of Law &Economics, Vol. 26, pp. 613-34. 57. Watts, R.L. and Zimmerman, J.L. (1986), Positive Accounting Theory, Prentice-Hall, Englewood Cliffs, NJ. 58. Zmijewski, M. and Hagerman, R. (1981), "An income strategy approach to the positive theory of accountingstandard setting or choice", Journal of Accounting and Economics, Vol. 3, pp. 129-49. Further Reading 1. Chaney, P. and Philipich, K. (2002), "Shredded reputation: the cost of audit failure", Journalof Accounting Research, Vol. 40, pp. 1221-45. 2. Healy, P. and Palepu, K. (2003), "The fall of Enron", Journal of Economic Perspectives, Vol. 17, pp. 3-26. Appendix Corresponding author David Emanuel can be contacted at: d.emanuel@auckland.ac.nzAuthorAffiliation Beilei Zhang, BDO Spicers, Auckland, New Zealand David Emanuel, Department of Accountingand Finance, The University of Auckland Business School, Auckland, New Zealand Illustration Equation 1Equation 2 Equation 3 Equation 4 Equation 5 Table I: Descriptive statistics for pooled sample Table II: Pearson(Spearman) correlation coefficients above (below) the diagonala point biserial correlation coefficient where onevariable is binary χ2 statistic where both variables are binary Table III: Results from pooled cross-sectionalregressions of deflated earnings on contemporaneous stock returns Table IV: Results from pooled cross-sectional regressions of deflated earnings on contemporaneous stock returns (NAFR sub-samples) Table V:Results from pooled cross-sectional regressions of deflated earnings on contemporaneous stock returns (NAFsub-samples) Table VI: Results of sensitivity analyses (excluding zero NAF observations) (NAFR sub-samples)Table VII: Results of sensitivity analyses (excluding zero non-audit fee observations) (NAF sub-samples) TableVIII: Results of sensitivity analysis (changing the classification of high/low NAFR/NAF) Subject: Studies; Regression analysis; Nonaudit services; Auditors; Accountant independence; Earningsmanagement; Accounting procedures Location: New Zealand Publication title: Accounting Research Journal Volume: 21 Issue: 2 Pages: 195-221 Publication year: 2008 Publication date: 2008 Year: 2008 Publisher: Emerald Group Publishing, Limited Place of publication: Bingley Country of publication: United Kingdom Journal subject: Business And Economics--Accounting, Business And Economics--Banking And Finance,Business And Economics--Investments ISSN: 10309616 Source type: Scholarly Journals Language of publication: English Document type: Feature

08 March 2013 Page 144 of 145 ProQuest

Document feature: Equations;Tables;References DOI: http://dx.doi.org/10.1108/10309610810905953 ProQuest document ID: 214448303 Document URL: http://search.proquest.com/docview/214448303?accountid=86413 Copyright: Copyright Emerald Group Publishing Limited 2008 Last updated: 2011-07-23 Database: Accounting&Tax

_______________________________________________________________ Contact ProQuest Copyright 2012 ProQuest LLC. All rights reserved. - Terms and Conditions

08 March 2013 Page 145 of 145 ProQuest