m.€¦ · Journal of Applied Business and Economics Editors Dr. Adam Davidson Dr. William Johnson...

Journal of Applied Business and Economics

North American Business Press

Atlanta - Seattle – South Florida - Toronto

Journal of Applied Business and Economics

Editors Dr. Adam Davidson Dr. William Johnson

Editor-In-Chief

Dr. David Smith

NABP EDITORIAL ADVISORY BOARD

Dr. Andy Bertsch - MINOT STATE UNIVERSITY Dr. Jacob Bikker - UTRECHT UNIVERSITY, NETHERLANDS Dr. Bill Bommer - CALIFORNINA STATE UNIVERSITY, FRESNO Dr. Michael Bond - UNIVERSITY OF ARIZONA Dr. Charles Butler - COLORADO STATE UNIVERSITY Dr. Jon Carrick - STETSON UNIVERSITY Dr. Mondher Cherif - REIMS, FRANCE Dr. Daniel Condon - DOMINICAN UNIVERSITY, CHICAGO Dr. Bahram Dadgostar - LAKEHEAD UNIVERSITY, CANADA Dr. Deborah Erdos-Knapp - KENT STATE UNIVERSITY Dr. Bruce Forster - UNIVERSITY OF NEBRASKA, KEARNEY Dr. Nancy Furlow - MARYMOUNT UNIVERSITY Dr. Mark Gershon - TEMPLE UNIVERSITY Dr. Ph ilippe Gregoire - UNIVERSITY OF LAVAL, CANADA Dr. Donald Grunewald - IONA COLLEGE Dr. Samanthala Hettihewa - UNIVERSITY OF BALLARAT, AUSTRALIA Dr. Russell Kashian - UNIVERSITY OF WISCONSIN, WHITEWATER Dr. Jeffrey Kennedy - PALM BEACH ATLANTIC UNIVERSITY Dr. Jerry Knutson - AG EDWARDS Dr. Dean Koutramanis - UNIVERSITY OF TAMPA Dr. Malek Lashgari - UNIVERSITY OF HARTFORD Dr. Priscilla Liang - CALIFORNIA STATE UNIVERSITY, CHANNEL ISLANDS Dr. Tony Matias - MATIAS AND ASSOCIATES Dr. Patti Meglich - UNIVERSITY OF NEBRASKA, OMAHA Dr. Robert Metts - UNIVERSITY OF NEVADA, RENO Dr. Adil Mouhammed - UNIVERSITY OF ILLINOIS, SPRINGFIELD Dr. Roy Pearson - COLLEGE OF W ILLIAM AND MARY Dr. Sergiy Rakhmayil - RYERSON UNIVERSITY, CANADA Dr. Robert Scherer - CLEVELAND STATE UNIVERSITY Dr. Ira Sohn - MONTCLAIR STATE UNIVERSITY Dr. Reginal Sheppard - UNIVERSITY OF NEW BRUNSWICK, CANADA Dr. Carlos Spaht - LOUISIANNA STATE UNIVERSITY, SHREVEPORT Dr. Walter Amedzro ST-Hilaire - HEC, MONTREAL, CANADA Dr. Ken Thorpe - EMORY UNIVERSITY Dr. Robert Tian - MEDIALLE COLLEGE Dr. Calin Valsan - BISHOP'S UNIVERSITY, CANADA Dr. Anne Walsh - LA SALLE UNIVERSITY Dr. Thomas Verney - SHIPPENSBURG STATE UNIVERSITY Dr. Christopher Wright - UNIVERSITY OF ADELAIDE, AUSTRALIA

Volume 13(3) ISSN 1499-691X Authors have granted copyright consent to allow that copies of their article may be made for personal or internal use. This does not extend to other kinds of copying, such as copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. Any consent for republication, other than noted, must be granted through the publisher:

North American Business Press, Inc. Atlanta - Seattle – South Florida - Toronto ©Journal of Applied Business and Economics 2012 For submission, subscription or copyright information, contact the editor at: [email protected] Subscription Price: US$ 360/yr Our journals are indexed by UMI-Proquest-ABI Inform, EBSCOHost, GoogleScholar, and listed with Cabell's Directory of Periodicals, Ulrich's Listing of Periodicals, Bowkers Publishing Resources, the Library of Congress, the National Library of Canada. Our journals have been used to support the Academically Qualified (AQ) faculty classification by all recognized business school accrediting bodies.

This Issue

Canadian Interest Rates and Yield Spread Dynamics ..................................................................... 11 Ilona Shiller We apply two commonly used cointegration techniques to study the relation between corporate yields and government yields and derive implications for the relation between yield spreads and government yields. Due to the stationary nature of the yield spread data, based on results of the conventional unit root tests, we cannot use cointegration theory to test this directly. The results of the unrestricted impulse response analysis provide evidence, which contradicts the results of cointegration analysis applied to corporate and government yields. Our expectation of a positive long-run relation between yield spreads and government yields is only slightly realized for AA yield spreads. The effect of a shock to the 10-year government yield appears to have a consistently negative impact on A and BBB yield spreads, both over the short-run and the long-run. The negative yield spread - government rate relation is induced due to the over-representation of callable bonds in the sample of bond indices. Moreover, yield spreads appear to exhibit characteristics similar to long-memory processes, for which the order of integration lies between zero and one. The hypothesis of fractional integration has to be tested using a completely different set of statistical tools and is not examined in this paper. Quality Award and Market Performance: An Empirical Investigation about Chinese Stock Market...................................................................................... 25 Xiangzhi Bu, Jinmei Tang, Robert Guang Tian Based on the empirical data from the Chinese stock market and by using an event study method, this paper investigates the relation between the quality award and the market performance of the publicly listed firms that have won quality awards from 2001 to 2009 in China. Our findings show that in the short-term the winners would get significantly accumulated abnormal returns, which differed because of the companies’ size, risk in investment and the prestige of the awards. In the short-run, firms with larger sizes, higher debt ratios, and the China Quality Award (CQA) winners can get accumulated abnormal high returns. Prior leakage of the awards information announcement has played a certain role in the process of abnormal high returns. The Impact of the “BRIC Thesis” and the Rise of Emerging Economies on Global Competitive Advantage: Will There Be a Shift from West to East? ................................................ 36 Richard T. Mpoyi The paper examines the thesis that by mid-21st century BRIC economies of Brazil, Russia, India and China (the “East”) would be wealthier than today’s seven largest developed economies of the G7 (the “West”). After analyzing the thesis, the study proposes the following. First, economic power is likely to shift from West to East because the combined GDP of the BRICs would be larger than that of the G7. Second, competitive advantage is less likely to shift from West to East, as after reaching G7’s income levels, BRIC economies would simply at best become as competitive as G7 economies.

The Effect of Product Demand Decline on Investments in Innovations: Evidence from the U.S. Defense Industry........................................................................................ 48 Donald W. Gribbin, Hong Qian, Ke Zhong The end of the Cold War led to a substantial decline in defense product demand. This study investigates the effects of product demand decline on defense firms’ investments in R&D for innovations. Our evidence indicates significant lower levels of R&D intensity for the low demand period (1993 to 1998) than for the high demand period (1984 to 1989). We also find significant declines in the defense firms' return on assets over the period, which is mainly attributable to a significant decrease in the firms' efficiency of using assets to produce sales. The defense firms, despite decline in defense product sales, generally maintained their total sales by partially shifting their capacity to commercial markets, which might be at the sacrifice of profitability, operating efficiency, and R&D investments for innovations. The Symmetry of Demand and Supply Shocks in Monetary Unions ............................................... 63 Maru Etta-Nkwelle, Carlton Augustine, Youngho Lee This study has been motivated by the numerous proposals for greater monetary and economic integration in Africa. We investigate the correlation of shocks between the exiting members of the West African Economic and Monetary Union and potential entrants in the region. In Southern Africa, we examine the pair wise correlations using the hub and spoke framework with South Africa as the hub. We observe large demand shock asymmetry amongst the countries in the west than in the south, suggesting more economic homogeneity amongst the members of the South African Development Community. However, there seem to be more supply shock symmetry amongst the countries in West Africa than their counterparts in the south. Determinants of Steady State Income: Is a Linear Specification Too Simple? ................................ 73 Kathleen E. Odell This paper examines whether the Solow model of economic growth and steady state income is supported by available cross-section data. The paper argues that while income per worker is correlated positively with investment in capital (both physical and human) and negatively with population growth, a simple log-linear empirical model derived from the Cobb-Douglas production function oversimplifies these relationships. A nonlinear specification is presented as an alternative. Estimation of the nonlinear specification shows that for the period 1960-2000, the strongest marginal effects of education occur in countries where education levels are already relatively high. The Influence of Legitimacy and Multi-level Environments: A Case Study of Taiwanese Subsidiary’s Entry Mode Choice in Mainland China ...................................................................... 91 Shaodong Hu, Hongxin Yao, Zongling Xu This research studies Taiwanese subsidiaries’ entry-mode choice from the angle of institutional theory. The findings show that Taiwanese subsidiaries in Mainland China imitate each other to gain legitimacy when choosing an entry mode at different levels of institutional environments. Taiwanese subsidiaries prefer to imitate the entry mode of their own group, rather than institutional environments at other levels, and prefer to imitate the entry mode of local industry institutional environment, rather than national industry institutional environment and region institutional environment, which indicates that Taiwanese subsidiaries tend to seek legitimacy in a narrow institutional environment among a multi-level institutional environments in Mainland China.

Great Recession of 2008-2009: Causes and Consequences............................................................. 107 Alexander Katkov Last four years of the turbulent economic performance raised more questions than provided answers about causes of the Great Recession of 2008-2009. The real GDP has declined about 3.7% and the full recovery has been achieved only recently. Most economists are blaming the real estate market collapse and the followed it financial crisis as main causes of the Recession. But the real macroeconomic cause was the change of the national macroeconomic policy from the Demand support to Supply support strategy in the middle of 1980s. This article is discussing the cause of the Recession and its consequences. The Effect of Fringe Benefits on the Paid Overtime Hours in Japan ............................................. 123 Hui-Yu Chiang In this paper, we adapt Bell and Hart’s model (1999) to examine paid overtime work by using Japanese Survey of Company Fringe Benefits data which includes information on the employer’s provision of fringe benefits as well as paid overtime hours for individuals. By including a crucial labor demand variable- the quasi-fixed cost which is omitted from Bell and Hart’s (1999) report, the present study provides more complete documentation of the structure of labor costs with paid overtime work.

GUIDELINES FOR SUBMISSION

Journal of Applied Business and Economics (JABE)

Domain Statement

The Journal of Applied Business and Economics is dedicated to the advancement and dissemination of business and economic knowledge by publishing, through a blind, refereed process, ongoing results of research in accordance with international scientific or scholarly standards. Articles are written by business leaders, policy analysts and active researchers for an audience of specialists, practitioners and students. Articles of regional interest are welcome, especially those dealing with lessons that may be applied in other regions around the world. This would include, but not limited to areas of marketing, management, finance, accounting, management information systems, human resource management, organizational theory and behavior, operations management, economics and econometrics, or any of these disciplines in an international context.

Focus of the articles should be on applications and implications of business, management and economics. Theoretical articles are welcome as long as their focus is in keeping with JABE’s applied nature. Objectives Generate an exchange of ideas between scholars, practitioners and industry specialists Enhance the development of the Business and Economic disciplines Acknowledge and disseminate achievement in regional business and economic development thinking Provide an additional outlet for scholars and experts to contribute their ongoing work in the

area of applied cross- functional business and economic topics. Submission Format

Articles should be submitted following the American Psychological Association format. Articles should not be more than 30 double-spaced, typed pages in length including all figures, graphs, references, and appendices. Submit two hard copies of manuscript along with a disk typed in MS-Word.

Make main sections and subsections easily identifiable by inserting appropriate headings and sub-headings. Type all first- level headings flush with the left margin, bold and capitalized. Second- level headings are also typed flush with the left margin but should only be bold. Third-level headings, if any, should also be flush with the left margin and italicized.

Include a title page with manuscript which includes the full names, affiliations, address, phone, fax, and e-mail addresses of all authors and identifies one person as the Primary Contact. Put the submission date on the bottom of the title page. On a separate sheet, include the title and an abstract of 200 words or less. Do not include authors’ names on this sheet. A final page, “About the authors,” should include a brief biographical sketch of 100 words or less on each author. Include current place of employment and degrees held.

References must be written in APA style. It is the responsibility of the author(s) to ensure that the paper is thoroughly and accurately reviewed for spelling, grammar and referencing. Review Procedure

Authors will receive an acknowledgement by e-mail including a reference number shortly after receipt of the manuscript. All manuscripts within the general domain of the journal will be sent for at least two reviews, using a double blind format, from members of our Editorial Board or their designated reviewers. In the majority of cases, authors will be notified within 60 days of the result of the review. If reviewers recommend changes, authors will receive a copy of the reviews and a timetable for submitting revisions. Papers and disks will not be returned to authors. Accepted Manuscripts

When a manuscript is accepted for publication, author(s) must provide format-ready copy of the manuscripts including all graphs, charts, and tables. Specific formatting instructions will be provided to accepted authors along with copyright information. Each author will receive two copies of the issue in which his or her article is published without charge. All articles printed by JABE are copyrighted by the Journal. Permission requests for reprints should be addressed to the Editor. Questions and submissions should be addressed to:

North American Business Press 301 Clematis Street, #3000

West Palm Beach, FL USA 33401 [email protected]

866-624-2458

Canadian Interest Rates and Yield Spread Dynamics

Ilona Shiller University of New Brunswick

We apply two commonly used cointegration techniques to study the relation between corporate yields and government yields and derive implications for the relation between yield spreads and government yields. Due to the stationary nature of the yield spread data, based on results of the conventional unit root tests, we cannot use cointegration theory to test this directly. The results of the unrestricted impulse response analysis provide evidence, which contradicts the results of cointegration analysis applied to corporate and government yields. Our expectation of a positive long-run relation between yield spreads and government yields is only slightly realized for AA yield spreads. The effect of a shock to the 10-year government yield appears to have a consistently negative impact on A and BBB yield spreads, both over the short-run and the long-run. The negative yield spread - government rate relation is induced due to the over-representation of callable bonds in the sample of bond indices. Moreover, yield spreads appear to exhibit characteristics similar to long-memory processes, for which the order of integration lies between zero and one. The hypothesis of fractional integration has to be tested using a completely different set of statistical tools and is not examined in this paper. INTRODUCTION

Most models predict that the rates on corporate bonds are positively related to the corresponding rates on government bonds (Merton, 1974; Longstaff and Schwartz, 1995). However, the precise relation depends on the methodology used in deriving the model. There has been little empirical research done in this area. Extant papers focus on studying how interest rates set by the government impact corporate yield spreads.

There is a disagreement in the area of fixed-income securities with respect to the nature of the yield spread -- government rate relation. This is the focus of the current study, which is closely related to that of Duffee (1998), Jacoby, Liao, and Batten (2009), and Morris, Neal, and Rolph (2000, hereafter referred to as MNR), who investigated yield spread dynamics by estimating the effect of government rates on yield spreads not only over the long-run, but also over the short-run. The approaches of the above papers to modelling yield spread dynamics and their results differ.

Most researchers in this area apply standard regression analysis on changes over time in yield spreads as a function of changes in government rates (Duffee, 1998 and Jacoby et al, 2009). However, some (see for example, MNR) argue that this approach lacks empirical power since they find evidence that the time series of yields on corporate and government bonds are nonstationary. This may invalidate the results of the standard regression analysis. To avoid problems associated with nonstationarity, Duffee (1998) and Jacoby et al (2009) apply statistical analysis on changes in bonds' interest rates rather than their levels. However, MNR argue that by using changes in rates one may lose valuable information regarding the

Journal of Applied Business and Economics vol. 13(3) 2012 11

long-term relationship between yield spreads and government yields, which is reflected in the levels of the two variables. To overcome this problem one can apply cointegration analysis to test whether the variables tend to move together over the long run.

In this paper we use cointegration methodology to estimate the relation between corporate and government yields and we also derive implications for the relation between yield spreads and government yields.

We utilize Canadian corporate bond data to estimate the relationship between corporate yield spreads and government yields. Canadian bond data has several important advantages over U.S. bond data (Jacoby et al, 2009). Unlike U.S. data, Canadian bond data allows accounting for factors that may bias the estimated relationship. These factors include callability provisions commonly attached to corporate bonds, coupon and tax differential between government and corporate bonds. Thus, an analysis of the relationship between yield spreads and government bond yields using Canadian bond data should provide clean and robust results.

Longstaff and Schwartz (1995, hereafter referred to as LS) find that yield spreads are negatively related to interest rates. They account for this result by presuming that the correlation between the value of the firm's assets and the risk-free rate is negative. Their regression analysis yields a negative yield spread - treasury rate relation, decreasing in magnitude as credit rating of the bond issue increases.

Jacoby et al (2009) issues two warnings regarding to LS regression analysis. The first warning relates to the presence of callable bonds in the LS sample that might have influenced their results. He claims that by overlooking the issue of callability one may mistakenly conclude that the negative yield spread - government rate relation stems solely from the inherent default risk. The second warning concerns the LS regression model applied to relative yield spreads, a model that was shown to produce spurious results.

Duffee (1998) finds that the relation between both callable and noncallable bond yield spreads and treasury yields is negative. However, this negative relation is much stronger for callable bond issues. The relation is also found to be more negative for high-priced callable bonds than for low-priced callable bonds. This is because the call option for high-priced bonds is deeper in the money. Duffee explains the observed weak yield spread - treasury yield relation for noncallable bonds by the coupon level effect. Everything else being equal, lower coupon rates of treasury bonds as compared with corporate bonds reflect the higher duration of treasury bonds. This implies that treasury bonds will be more sensitive to changes in treasury yields and leads, he explains, to the observed negative yield spread - treasury yield relation for non-callable bonds.

Duffee (1998) also finds that the relation between the yield spreads and the slope of the treasury term structure is negative. This relation turns out to be weaker for short- and medium-maturity bonds. He then estimates the persistence of changes in the yield spreads as a result of changes in treasury yields by running unrestricted vector autoregression models (VAR) for the yield spreads, treasury yields, and the slope of the term structure. Analyzing the impulse response functions, Duffee finds that impulse error bands are quite large to produce reliable inferences. The general pattern of impulse responses indicates that it takes a long time for yield spreads and bond prices to adjust to new information.

MNR (2000) use cointegration approach to model the relation between corporate and treasury yields. They use monthly averages of daily yields for 10-year constant maturity treasury bonds and Moody's Aaa and Baa seasoned bond indices obtained from the Board of Governors of the Federal Reserve System. The sample size of each data series sums to a total of 456 observations. They show that corporate rates are cointegrated with treasury rates. Theorectically, this result suggests that the dynamics of this relationship is time-varying - the relation between corporate and treasury rates is positive in the long-run and negative in the short-run. Intuitively, this pattern implies the same time-dependent relation for the relation between yield spreads and treasury yields. To confirm this, they compute the separate impulse response functions for corporate and treasury yields as a result of a shock in the treasury yield. Then, they find the implied change in yield spreads by taking the difference between the two functions. Although innovative, this approach lacks theoretical and econometric underpinning. This is because their data for yield spreads appears to be stationary based on conventional unit root tests and, thus, unrestricted VAR impulse response function analysis is called for.

12 Journal of Applied Business and Economics vol. 13(3) 2012

Their estimated long-run positive relation between corporate yields/yield spreads and treasury rates is consistent with models of Duan et al (1995), Lesseig and Stock (1998), and Bernanke and Gertler (1989). For example, the model of Bernanke and Gertler (1989) predicts that higher treasury rates will increase the agency problems for borrowers, widening the difference between internal and external financing costs and the firm's yield spread.

The paper is organized as follows. The second section applies both univariate and multivariate cointegration procedures to Canadian corporate and government yields. In this section we also establish the implied relation between yield spreads and government rates based on the dynamics of the corporate yield - government yield relation. Summary and conclusions are offered in the third section. DATA, SAMPLE PROPERTIES, AND RESULTS

The sample of corporate and treasury yields is based on the month-end, yield-to-maturity indices from the Scotia Capital Markets (SCM) taken from the CANSIM database. The sample covers the December 1985 to April 2002 period. SCM provides long-term corporate indices for four different investment-grade ratings: AAA, AA, A, and BBB. Since the data for the AAA investment-grade rating is discontinued in March 1993, we exclude the AAA rating from the analysis for reason of limited sample size. SCM's long-term corporate bond indices include bonds with maturity greater or equal to 10 years. The yield spreads data is constructed as the difference in the index yield and 10-year constant maturity Government of Canada index yield. Thus, average portfolio maturity of 10 years is assumed for long-term corporate bond index. The slope of the term structure variable is proxied with the difference between the yields of the 30-year and 10-year constant maturity Government of Canada indices.

Jacoby et al (2009) argues that the SCM data are suitable to study the yield spread - government yield relation. He argues that these indices provide the flexibility to control for callability risks, coupon level effects, and effects arising from taxation.

Table 1 reports the summary statistics for corporate yields, government rates, constructed series of yield spreads and yield spread changes during the sample horizon. In general, the lower the credit rating of an index, the wider is the spread due to higher probability of default. Over the estimated period, AA yields averaged 9.54%, A yields averaged 9.78%, and BBB yields averaged 10.61%. Also, AA yields have the highest volatility among all ratings: the standard deviation of AA yields is 2.42% as compared with the standard deviation of 2.16% of the BBB yields. The AA yield spreads averaged 0.97% over the sample period, A yield spreads averaged 1.22%, and BBB yield spreads averaged 2.05%.

Table 1 also reports the results of the Jarque-Bera test of normality. Recall, if the p value of the chi-square statistic is sufficiently low, one can reject the hypothesis that the residuals are normally distributed. Looking at table 1, it is seen that all series, including corporate yields, government yields, yield spreads and the slope of the term structure, are not normally distributed. Almost all of the estimates of kurtosis statistics are less than three and this implies that probability density functions are platykurtic (thick-tailed). The estimates of kurtosis for the mean changes in corporate yields, government yields, the slope of the term structure, and BBB yield spreads show a value greater than three that implies leptokurtic (a slim or long-tailed) probability function. The normal distribution is, however, characterised by zero skewness and kurtosis of three. Thus, normal distribution is symmetric and mesokurtic. Turning to skewness measures, Table 1 shows that all corporate yields and yield spreads in levels have positive skewness. And positive skewness implies that the left tail of the distribution have more probability than a normal distribution.

We examine the quantile-quantile (QQ)-plots for corporate yields and yield spreads. These figures (available upon request) plot the quantiles of the chosen series against the normal distribution. If the two distributions are the same, the QQ-plot should lie on a 450 line. The pattern of deviation from linearity provides an indication of the nature of the mismatch; the two figures clearly indicate that yield spreads have distributions closer to the normal than corporate yields. Since all estimated plots curve downward at the left, and upward at the right, it is an indication that respective distributions are platykurtic and have a thicker tail than the normal distribution. Comparing yield spread distributions with corporate yield


distributions, it can be seen that the plots are slightly convex, indicating that distributions of yield spreads are slightly negatively skewed compared with the distributions of corporate yields.

TABLE 1 SUMMARY STATISTICS FOR CORPORATE YIELDS AND YIELD SPREADS OF SCM

LONG-TERM CORPORATE BOND INDICES

This table reports the summary statistics for corporate yields, government yields, the slope of the term structure, and yield spreads. Summary statistics are reported for levels and changes in the above variables. We report the mean, the standard deviation, the skewness, the kurtosis, and the Jarque-Bera test of normality with its p-value. Level_TS is the level of the term structure represented by the 10-year Government of Canada constant maturity index. Slope_TS is the slope of the term structure represented by the difference between the yields in the 30-year and 10-year Government of Canada constant maturity indices.

The statistics are based on a monthly SCM corporate bond indices datafrom 1985:12 to 2002:4.

No. obs. Mean St.Dev. Skewness Kurtosis Jarque-Bera Prob.AA 239 9.5400 2.4300 0.3300 2.7000 5.14 0.0770A 239 9.7800 2.3900 0.3500 2.8500 5.15 0.0760

BBB 239 10.6100 2.1600 0.1100 2.8900 0.62 0.7331Level_TS 239 8.5600 2.3900 0.3700 2.6600 6.67 0.0356Slope_TS 239 0.3000 0.2700 -0.2400 2.6700 3.36 0.1863

ΔAA 238 -0.0500 0.3500 -0.3700 4.2800 21.78 0.0000ΔA 238 -0.0500 0.3400 -0.5200 4.4700 31.88 0.0000

ΔBBB 238 -0.0400 0.4100 0.2900 6.3800 116.61 0.0000ΔLevelTS 238 -0.0400 0.3700 -0.0600 4.3400 17.86 0.0001ΔSlopeTS 238 0.0010 0.1100 -0.2100 3.3000 2.69 0.2600AA_CS 239 0.9800 0.2800 0.5300 2.9800 11.00 0.0100A_CS 239 1.2200 0.3300 0.3600 2.4800 7.80 0.0200

BBB_CS 239 2.0500 0.8300 1.3800 5.2900 127.92 0.0000ΔAACS 238 -0.0010 0.1800 0.2300 4.7700 33.09 0.0000ΔACS 238 -0.0020 0.1800 0.0800 4.4100 20.03 0.0001

ΔBBBCS 238 0.0030 0.3100 1.7300 20.0700 3007.30 0.0000

We now summarize the dynamics of the corporate yield - government yield relation and derives some

implications with respect to the yield spread - government yield relation. The estimation and inference will be drawn from unit root testing, cointegration, and error-correction estimation. We look at both univariate and multivariate testing procedures and compare the performance of univariate residual-based tests of Engle-Granger to that of Johansen maximum likelihood system-based tests. The existence of a bivariate cointegrating vector between corporate yields and government yields is analyzed. This section provides evidence that corporate yields and the level of the term structure are non-stationary and thus can indeed be on the same wavelength. Yield spreads appear to be stationary, meaning that standard cointegration theory framework cannot be applied to test the yield spread - government yield relation directly.

Impulse response functions to the shocks in risk-free rate can provide a useful and valid way to analyze the short- and long-run dynamics of these relations. A standard proposition in the related literature is that real disturbances (shocks in government rates) may cause permanent (long-run) and/or temporary short-run deviations of actual yield spreads from equilibrium spreads. The data show that


impulse response analysis of the corporate yield - government yield relation has to be estimated by restricted VAR analysis. However, due to the small sample size of the estimated sample, as well as the drawbacks of cointegration testing procedures themselves, impulse response functions in this context were not attempted in this study. Moreover, the yield spread data appear to be stationary and unrestricted VAR analysis is called for to find out the response of yield spreads to a one standard deviation innovation in the 10-year government rate.

The primary objective of the following discussion is to apply cointegration testing procedures to corporate and government yields of the SCM sample and to make intuitive inferences about the impact of government yields on corporate yield spreads. Corporate Yield - Government Yield Relation

We find autocorrelations for corporate and government yields, and the slope of the term structure. For all of the above-mentioned variables, autocorrelation coefficients are quite large up to a lag 6, meaning that the series under consideration exhibit some persistence.

Augmented Dickey-Fuller and Phillips-Perron tests that have unit root as the null hypothesis investigate the presence of a random walk component in each series. To save space, we do not report the unit root test results but they are available upon request. The null hypothesis of a unit root was accepted for all series using both the augmented Dickey-Fuller tests and the Phillips-Perron tests. For the yield spreads, the results of the unit root tests are different. The Dickey-Fuller test results show that yield spreads contain unit root. But using the Phillips-Perron test, we are able to reject the null hypothesis of a unit root at all lags for yield spreads. Applying the Kwiatkowski et al (1992) test, we are also unable to reject the hypothesis of level stationarity and trend stationarity for AA and A yield spread data series at the 5% level and for BBB yield spread data at the 1% level. This confirms the preliminary evidence that yield spreads are stationary. The monthly changes in corporate yields, the level, and the slope uniformly exhibit stationarity. This means that corporate yields, the level, and the slope of the term structure are integrated of order 1, whereas the yield spread data appears to be stationary.

Table 2 presents the results of Engle-Granger equilibrium regressions for three corporate yields: AA, A, BBB. Table 3 reports the results of unit root testing procedures applied to residuals formed from the equilibrium regressions. Since the estimated residuals from the equilibrium regressions have a zero mean and do not have a time trend, the tests include only lagged residuals and their differenced lagged values as regressors. An iterative procedure is followed to determine the appropriate lag length. 12 lagged difference terms are added to the regression equation. Then the lag length is determined by minimizing the Schwartz criterion for augmented Dickey-Fuller tests and by selecting the bandwidth parameter ℓ for the kernel-based estimators of Johansen system-based tests. The bandwidth selection is based on the Newey-West procedure. Comparing the estimated τ-values with calculated critical values for the null of no cointegration from J.G. MacKinnon (1993), it can be seen that in absolute terms the estimated τ-values exceed the critical values at least at the 5% level of significance. Thus, there is sufficient evidence that corporate and government yields are cointegrated based on the Engle-Granger cointegration methodology. The Engle-Granger approach indicates that the relation between corporate and government yields is time-varying: positive in the long-run and negative in the short-run.


TABLE 2 ENGLE-GRANGER EQUILIBRIUM REGRESSIONS FOR CORPORATE AND

GOVERNMENT YIELDS

This table reports the results of the first step of the Engle-Granger cointegration procedure. We report the results of a simple OLS regression analysis for corporate and government yields.

AA A BBBVariable Coefficientt-statistic Coefficien t-statistic Coefficien t-statisticConstant 0.8858 13.1068 1.2781 16.2038 3.3254 18.4097Level_TS 1.0107 132.917 0.993 111.8946 0.8512 41.8811R² 0.9868 0.9814 0.881DW 0.4246 0.3017 0.1614

TABLE 3 ENGLE-GRANGER TESTS FOR COINTEGRATION BETWEEN CORPORATE AND

GOVERNMENT YIELDS

The residuals from each equilibrium regression were checked for unit roots. The unit root tests in a cointegration context are computationally easy to do. Since the estimated residuals from the equilibrium regressions have a zero mean and do not have a time trend, the tests included only lagged residuals and their differenced lagged values as regressors. An iterative procedure was followed to determine the appropriate lag length. 12 lagged difference terms were added to a regression equation. Then the lag length was determined by min imizing the Schwartz criterion for the augmented Dickey-Fuller tests and by selecting the bandwidth parameter ℓ for the kernel-based estimators of Ω₀. The bandwidth selection was based on the Newey-West procedure. **, *, and *** means that the variable under consideration is stationary at the 10% level, 5%, and 1% level respectively.

Indices AA A BBBDF PP DF PP DF PP

Long-term -5.3437** -4.9375** -4.5492** -4.1361** -3.2589* -3.1259***

Table 4 presents the results of Johansen maximum likelihood approach to test for the existence of a

bivariate cointegrating vector between corporate and government yields. These results are based on the appropriate lag length determined by minimizing the Schwartz Criterion. We find the evidence of 2 cointegrating vectors for AA rates, and no cointegrating vectors for A and BBB rates at the 5% level using this approach. At the 1% significance level, however, we find no cointegrating vectors for all three rates. Consider, for example, the results for the A-rated index. For this index, the first eigenvalue statistic is not significant at the 5% level. Therefore, we are unable to reject the hypothesis of zero cointegrating vectors in favour of the alternative that there exists one cointegrating vector. The second maximum eigenvalue statistic is significant and supports the existence of two cointegrating vectors. However, since the existence of one cointegrating relation is rejected, the existence of two cointegrating relations can be eliminated. The results for other series are very similar.


TABLE 4 MULTIVARIATE COINTEGRATION RESULTS FOR CORPORATE AND

GOVERNMENT YIELDS

This table presents the results of the Johansen multivariate test. The statistics are based on a monthly long-term SCM corporate bond indices data from 1985:12 to 2002:4. Level_TS is the level of the term structure represented by the 10-year Government of Canada constant maturity index. Slope_TS is the slope of the term structure represented by the difference between the yields in the 30-year and 10-year Government of Canada constant maturity indices.

Cointeg.Group Lags Maximal Eigenvalue Statistic Trace StatisticEigenvalues Statistic 5% critical Statistic 5% critical

AA, Level 1.0258 0.0712 17.5035* 14.07 26.6033* 15.41λ 0.0377 9.0999* 3.76 9.09986* 3.76A, Level 0.0572 13.9492 14.07 22.8490* 15.41λ 0.9368 0.0369 8.8998* 3.76 8.8998* 3.76BBB, Level 0.0416 10.0679 14.07 17.1443* 15.41λ 0.4554 0.0294 7.0764* 3.76 7.0764* 3.76

If the two variables are cointegrated, then they are on the same wavelength, or the trends in corporate

and government yields cancel out. This would have a twofold meaning. First, if we plot the cointegrating relations between corporate yields and the 10-year Government of Canada yield, it should not on average significantly deviate away from a zero line. Examining cointegrating relations between corporate yields and the Government of Canada yield (see Figure 1), it can be seen that all of these lines are close on average to the zero line. Thus, it can be concluded that our corporate yields and government rates are cointegrated, a result that implies that an Engle-Granger test procedure has more power than the Johansen test procedure. This is in line with theoretical research in this area. Second, although the Johansen test shows no cointegration for corporate and government yields, the estimates of the long-run relation between corporate and governmet rates based on Engle-Granger and Johansen test procedures are approximately equal in magnitude. For instance, the cointegrating vector for AA rate using the Engle-Granger methodology is (1, -1.0107) vs (1, -1.0258) using the Johansen approach.

FIGURE 1 COINTEGRATING RELATIONS BETWEEN CORPORATE AND GOVERNMENT YIELDS

These graphs sketch the stationary linear combinations of corporate and government yields or, in other words, cointegrating equations. Cointegrating relations sketched below represent the long-run equilibrium relat ionship among the variables, which should not significantly deviate from a zero line. The data covers the period from December 1985 to April 2002.

-.8

-.4

.0

.4

.8

86 88 90 92 94 96 98 00

Cointegrating relation between A YS and GR

-.8

-.6

-.4

-.2

.0

.2

.4

.6

.8

86 88 90 92 94 96 98 00

Cointegrating relation between AA YS and GR

-2

-1

0

1

2

3

86 88 90 92 94 96 98 00

Cointegrating relation between BBB YS and GR


The results of the more powerful Engle-Granger approach indicate that corporate yields and government yields are cointegrated. Since a 1% increase in the government yield generates an increase in corporate yields, it can be claimed that as interest rates rise, yield spreads will eventually widen as well.

The results of the Johansen maximum likelihood approach are puzzling, taking into consideration that this test is biased towards the acceptance of the alternative hypothesis of cointegration. But we argue that the Engle-Granger approach is better suited for the purpose of testing the corporate yield - government yield relation in our sample. This is because our corporate and government yield series are not normally distributed and only one lag was included in the estimation. The Johansen tests are inferior to the Engle-Granger tests when the data deviate from the assumption of normally, identically and independently distributed disturbances and when the lag length/sample size is small.

An alternative explanation for the non-existence of a bivariate cointegrating vector between corporate and government yields based on the Johansen test results and for the weak existence of this vector based on the Engle-Granger test results is as follows. Plotting the autocorrelation function for the error-correction term of the equilibrium regression for the AA rate in the Figure 2 below, it can be observed that although autocorrelation coefficients appear to be stationary, they exhibit some long-memory attributes. Autocorelation coefficients do not exhibit the same type of persistence as they do for the original series.

FIGURE 2 CORRELOGRAM FOR THE AA YIELD ERROR-CORRECTION TERM

This figure plots the 200 sample autocorrelation coefficients for the estimated error-correction term of the equilibrium regression involving the AA rate. The data covers the period from December 1985 to April 2002.

Thus, the error-correction term decays more slowly to zero as a result of a shock than the usual

exponential decay of the autocorrelation function for the covariance stationary and invertible autoregressive moving average process. This means that the effect of a shock is long lasting and that the error-correction term might be following a fractionally integrated process of the form: ,)()1( tt

d ueL =−− µ where ut is integrated of order 0.

Several researchers have looked at the power of cointegration tests in the presence of fractionally integrated series and concluded that the Johansen test has very little power against fractional alternatives (see, for example, Gonzalo (1998); Cheung and Lai (1993)). Thus, we argue that the Engle-Granger procedure is more robust to the misspecification of the long-memory components of variables entering the model as well as to the fractionally integrated deviations from the long-run equilibrium.

-0.4-0.2

00.20.40.60.8

1

1 19 37 55 73 91 109

127

145

163

181

199

Lag

Aut

ocor

rela

tion


Based on the results of the Engle-Granger tests, we estimate the corporate yield - government yield error-correction models. We use the residuals obtained from the equilibrium regressions as an instrument for the error-correction terms, as proposed by Engle and Granger. Other than the error-correction term, obtained from the equilibrium regression, the error-correction model constitutes VAR in first differences and will be applied in the VAR framework. Thus, OLS is an efficient estimation strategy since each equation contains the same set of regressors. Moreover, since all terms in the error-correction model are stationary, the test statistics used in traditional VAR analysis can also be used here. The lag length for our error-correction models was gauged by minimizing the Schwartz criterion.

Table 5 reports the results of the error-correction models. The point estimates of the error-correction terms in the model are all insignificantly different from zero. This might indicate that movements in corporate yields are independent of movements in government yields. Moreover, the point estimate for the AA rate error-correction term is positive, casting doubt on the existence of cointegration between AA corporate yield-government yield relation altogether.

TABLE 5

ESTIMATES OF THE ERROR-CORRECTION MODEL FOR CORPORATE AND GOVERNMENT YIELDS

This table presents the results of the error-correction bivariate tests, which are based on the Engle-Granger univariate method. The following erro r-correct ion regression was estimated.

Dependent variableAA A BBB

Variable Coefficient t-statistic Coefficient t-statistic Coefficient t-statisticConstant -0.0371 (-1.6535) -0.0367 (-1.6547) -0.0364 (-1.3640)

ΔAA_t-1 0.1212 -0.8626ΔA_t-1 0.1652 -1.2024

ΔBBB_t-1 -0.025 (-0.2755)ΔLevelTS_t-1 0.0432 -0.3308 0.0012 -0.0091 0.1086 -1.0968

ECterm 0.0223 -0.2642 -0.0232 (-0.3303) -0.0626 (-1.7278)R² 0.0275 0.0282 0.021

Impulse response function analysis is not attempted here due to the expected biased nature of

cointegration relationship that arises when the variables entering the system are not normally distributed and the sample size is small. The small power of cointegration tests against the fractional alternatives leads us to believe that the cumulative impulse response corresponding to a shock in the infinite past will be zero for the processes integrated of order less than one.

Phyllips (1998) shows that the estimated impulse responses in a cointegrated VAR model when based on the reduced rank regression are consistent if the cointegrating rank is consistently estimated. Since we feel that the cointegrating rank is underestimated, the impulse response function analysis is not appropriate in this setting. Moreover, Kilian (1998) stresses that the small sample distribution of impulse responses can be significantly biased and skewed. He shows that when this is true, it can make traditional confidence intervals extremely inaccurate.


Yield Spread - Government Yield Relation Since yield spreads are integrated of order zero, it is inappropriate to estimate the yield spread -

government yield relation directly by applying standard cointegration theory. Considering the distributional patterns of yield spreads, it can be seen from the Jarque-Bera normality

test results and QQ-plots that yield spreads deviate from the normal distribution. Although this deviation is small (the skewness and kurtosis statistics are close to zero and three respectively), the kurtosis statistics are all less than three meaning that probability distribution functions are fat-tailed. Pedrosa and Roll (1986) document similar results for yield spreads and their changes. Heavier tails compared to the standard normal density mean that there is a higher probability of extreme observations. Given the later finding and knowing that critical values of the unit root tests and cointegration tests are biased in the presence of non-normal, non-independent, and non-identical distribution of disturbances, we cannot use the standard cointegration theory to test the yield spread - government yield relation.

The direct analysis of the yield spread - government yield relation is more reasonable from a theoretical perspective. Researchers are interested in how the yield spread responds to the shifts in the government yield. Some researchers apply regression analysis to yield spread changes instead of levels (see for example, Duffee, 1998 and Jacoby et al, 2009). However, the theory discusses the relation between the levels of these variables and not their changes. Therefore, it is important to apply unit root testing procedures to the levels of yield spreads and government yields. If the variables are non-stationary then the researcher should not difference the non-stationary series to apply conventional regression analysis, since important and valuable information can be lost.

Figures 3 and 4 demonstrate that the relation between changes in spreads and changes in the government yield is different from the relation between the levels of these variables. The relation between changes of yield spreads and government yields seems to be negative for AA, and A indices and there seems to be no relation between these variables for the BBB index. The relation between the AA yield spread in level form and the government yield in the level form appears to be positive, whereas that for the A and BBB indices appears to be negative.

FIGURE 3

CHANGES IN YIELD SPREADS VERSUS CHANGES IN GOVERNMENT RATES (LONG-TERM INDICES)

These figures plot changes in yield spreads as a function of changes in government yields. The data covers the period from December 1985 to April 2002.

-.6

-.4

-.2

.0

.2

.4

.6

-1.2 -0.8 -0.4 0.0 0.4 0.8 1.2

Change in Government Rate (%)

Cha

nge

in A

A S

prea

d (%

)

D(AA_CS) vs D(T10)

-.6

-.4

-.2

.0

.2

.4

.6

-1.2 -0.8 -0.4 0.0 0.4 0.8 1.2

Change in governmrnt rate (%)

Cha

nge

in A

Spr

ead

(%)

D(A_CS) vs D(T10)

-2

-1

0

1

2

3

-1.2 -0.8 -0.4 0.0 0.4 0.8 1.2

Change in Government Rate (%)

Cha

nge i

n B

BB

Spr

ead

(%)

D(BBB_CS) vs D(T10)


FIGURE 4 YIELD SPREADS VERSUS GOVERNMENT YIELDS (LONG-TERM INDICES)

These figures plot yield spreads as a function of government yields. The data covers the period from December 1985 to April 2002.

Since the unit root results suggest that yield spreads are stationary, whereas the level of the term

structure is non-stationary, cointegration theory cannot be used to understand the nature of the relationship between yield spreads and government yields. However, unrestricted VAR analysis can and should be applied to estimate impulse responses of yield spreads to a one standard deviation innovation in a 10-year government rate. The ordering of the variables is: 10-year government yield, the appropriate yield spread. Figure 5 plots impulse responses of yield spreads as a result of a one standard deviation innovation in a government rate. First, the standard error bands are wider for BBB-rated bonds than those for A-, and AA-rated bonds. This means that reliable inferences can be made, especially at shorter horizons. The relation between the A, BBB yield spreads and the government rate seems to be consistently negative at long horizons, slowly pulling towards zero as time progresses. Consistent with the AA corporate yield - government yield estimated relation, the AA yield spread falls initially by 0.06 standard deviations, but ultimately increases by 0.0004 standard deviations by month 16 as a result of a shock in the 10-year government yield. Consistent with the findings of Duffee (1998), the impulse response results show that the responses of yield spreads to shocks in government yields take a long time to dissipate.

The results of the unrestricted impulse response analysis are puzzling. The long-run effect of a shock to the 10-year government yield appears to have a positive long-run impact on the AA yield spreads but negative long-run impact on the A and BBB yield spreads. An explanation can be provided by Jacoby et al (2009), who claims that most Canadian corporate bonds issued starting from the year of 1987 carry the "doomsday" call provision. This call provision provides means to control for the callability bias. Specifically, he asserts that starting from the year 1995, vast majority of bonds in the SCM long-term indices category carry a doomsday call provision instead of a standard call provision. Recalling that it will typically be economically suboptimal for firms to call BBB-rated bonds carrying a doomsday call provision, he argues that the BBB index starting from 1995 can be assumed to be economically non-callable. However, since the sample period from 1985 to 1995 is dominated by bond issues carrying a standard call provision, the callability bias is applicable to our estimation period. Duffee (1998) argues that the presence of callable bonds in the sample makes the relation between yield spreads and government yields negative. This implies that the cumulative impact of callability in our sample is larger for the A and BBB yield spreads rather than for the AA yield spreads.

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

4 5 6 7 8 9 10 11 12

Government Rate (%)

AA

Spr

ead

(%)

AA_CS vs T10

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

4 5 6 7 8 9 10 11 12

Government Rate (%)

A S

prea

d (%

)

A_CS vs T10

0

1

2

3

4

5

6

4 5 6 7 8 9 10 11 12

Government Rate (%)

BB

B S

prea

d (%

)

BBB_CS vs T10


FIGURE 5 IMPULSE RESPONSES OF YIELD SPREADS TO A SHOCK IN THE 10-YEAR

GOVERNMENT YIELD (LONG-TERM INDICES)

Each graph represents the impulse response of long-term AA, A and BBB yield spreads to the shock in the 10-year constant maturity Government of Canada yield implied by a vector autoregression with two lags of 10-year constant maturity Government of Canada yields, and the given yield spread, in that order. The standard error bounds on the impulse responses are also displayed. The data covers the period from December 1985 to April 2002.

Since it is known that unit root tests have very small power to detect fractional integration root parameters, we have hypothesized that yield spread data is fractionally cointegrated. Empirically, if the root differs from one in either direction, then a different set of statistical tools is called for. This drawback of the unit root tests is also shared by the cointegration testing procedures, implying that conventional tests are unable to capture the existence of possible fractional cointegration between yield spreads and government yields. Next, we examine the autocorrelation functions of yield spreads by plotting the autocorrelation coefficients against 200 lag parameters for the AA yield spread data in Figure 6. The

FIGURE 6 CORRELOGRAM FOR THE AA YIELD SPREAD

The figure plots the first 200 sample autocorrelation coefficients for the AA yield spread. The monthly long-term SCM yield spread data covers the period from December 1985 to April 2002.

-.10

-.08

-.06

-.04

-.02

.00

.02

.04

5 10 15 20 25 30 35 40

Response of AA_CS to CholeskyOne S.D. T10 Innovation

-.12

-.10

-.08

-.06

-.04

-.02

.00

.02

.04

25 50 75 100

Response of A_CS to CholeskyOne S.D. T10 Innovation

-.20

-.15

-.10

-.05

.00

.05

.10

25 50 75 100

Response of BBB_CS to CholeskyOne S.D. T10 Innovation

-0.4-0.2

00.20.40.60.8

1

1 19 37 55 73 91 109

127

145

163

181

199

Lag

Aut

ocor

relat

ion o

f AA


following pattern emerges: the autocorrelation coefficient function appears to be generated by a stationary process, but this process seems to exhibit long memory characteristics. Thus, yield spreads appear to: (i) possess non-normal distribution with fat tails; (ii) have autocorrelations that decay to zero very slowly; and (iii) have cycles, which are not periodic. This might indicate that yield spreads have a fractional order of integration. CONCLUSION

This paper applies two commonly used cointegration techniques to study the relation between corporate yields and government yields. This sign and significance of this relation has important implications for the relation between yield spreads and government yields.

The hypothesis of non-stationarity for corporate and government yields is not rejected. However, first differences of these time series are uniformly stationary. This has an important implication for academic circles, namely that one should be extremely precautious to use corporate yields in the level form in statistical applications. One can only do that in the case if corporate yields and the other series under consideration are cointegrated.

We find preliminary evidence of the existence of cointegrating relation between corporate and government yields based on the Engle-Granger method. The Engle-Granger residual-based tests provide evidence that corporate yields and the 10-year Government of Canada yield are cointegrated, implying that there is a long-run equilibrium relationship between these variables. Unfortunately, error-correction estimates are all insignificantly different from zero. This might indicate that movements in corporate yields are independent of movements in government yields. Moreover, the point estimate for the AA rate error-correction term is positive, casting doubt on the existence of cointegration between AA corporate yield and the 10-year government yield altogether.

Consistent with the findings of Duffee (1998), our results also show that the responses of yield spreads as a result of shocks in the 10-year government rate take a long time to dissipate. This can be partly explained by the possible long memory characteristics exhibited by the yield spread series, for which the order of integration lies between 0 and 1. Estimates of the exact order of integration of yield spreads should be of interest to policy makers for at least two reasons. First, this will enable them to determine whether shocks to yield spreads are short-lived, long-lived or infinitely lived. Second, if the order of integration is less than one, then we can suspect that the cointegrating relation involving yield spreads may not be precisely of zero order, with the consequence that adjustments to re-establish the long-run equilibrium state may follow long-memory processes as it appears to be. In case of yield spreads, significant shocks to interest rate expectations, such as those resulting from changes in policy regimes, may indeed take a long time to dissipate. This implies that it will take a long time for yield spreads to revert to their respective means. Thus, forecast accuracy might be improved only within the framework of longer-term forecasts. Even though estimates of the exact order of integration are extremely important, they have to be calculated using a completely different set of statistical tools and are not looked at in this paper. REFERENCES Bernanke, B.& Gertler, M. (1989). Agency Costs, Net Worth and Business Fluctuations. American Economic Review, 79, 14-31. Cheung, Y., & Lai, K. S. (1993). Finite Sample Size of Johansen's Likelihood Ratio Test for Cointegration. Oxford Bulletin of Economics and Statistics, 55, 313-328. Dickey, D. A. & Fuller, W. A. (1979). Distribution of the Estimators for Autoregressive Time Series with a Unit Root. Journal of American Statistical Association, 74, 427-431.


Duan, J., Moreau, A. & Sealey. C. (1995). Deposit Insurance and Bank Interest Rate Risk: Pricing and Regulatory Implications. Journal of Banking and Finance, 19, 1091-1108. Duffee, G. R. (1998). The Relation between Treasury Yields and Corporate Bond Yield Spreads: an Empirical Analysis. Journal of Finance, 53, 2225-2241. Engle, R. F. & Granger, W. J. (1987). Cointegration and Error Correction: Representation, Estimation, and Testing. Econometrica, 55, 251-276. Ericsson, N. R. & MacKinnon, J. G. (1999). Distributions of Error-Correction Tests for Cointegration, Board of Governors of the Federal Reserve System, International Finance Discussion Papers, No. 655. Gonzalo, J. & Lee, T. (1998). Pitfalls in Testing for Long-Run Relationships. Journal of Econometrics, 86, 129-154. Jacoby, G., Liao, C. & Batten, J. (2009). Testing for the Elasticity of Corporate Yield Spreads. Journal of Financial and Quantitative Analysis, 44, (3), 641-656. Johansen, S. (1991). Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models. Econometrica, 59, (6), 1551-1580. Johansen, S. & Juselius, K. (1990). Maximum Likelihood Estimation and Inference on Cointegration- with Applications to the Demand for Money. Oxford Bulletin of Economics and Statistics, 52, (2), 169-210. Kilian, L. (1998). Small-sample Confidence Intervals for Impulse-Response Functions. The Review of Economics and Statistics, 80, 218-30. Kwiatkowski, D., Phyllips, P. C. B., Schmidt, P. & Shin, Y. (1992). Testing the Null Hypothesis of Stationarity against the Alternative of a Unit Root. Journal of Econometrics, 54, 159-178. Lessig, V. & Stock, D. (1998). The Effect of Interest Rates on the Value of Corporate Assets and the Risk Premia of Corporate Debt. Journal of Portfolio Management, 17, 52-64. Longstaff, F., & Schwartz, E. (1995). A Simple Approach to Valuing Risky Fixed and Floating Rate Debt. Journal of Finance, 50, 789-819. MacKinnon, J.G. (1993). Critical Values for Cointegration Tests, Cointegrated Time Series, 267-276. Merton, R. C. (1974). On the Pricing of Corporate Debt: the Risk Structure of Interest Rates. Journal of Finance, 29, 449-470. Morris, C., Neal, R. & Rolph, D. (2000). Interest Rates and Yield Spreads Dynamics, Discussion Paper. Pedrosa, M. & Roll, R. (1998). Systematic Risk in Corporate Bond Yield Spreads. Journal of Fixed Income, 12, 7-26. Phyllips, P.C.B. (1998). Impulse Response and Forecast Error Variance Asymptotics in Nonstationary VARs. Journal of Econometrics, 83, 21-56.


Quality Award and Market Performance: An Empirical Investigation about Chinese Stock Market

Xiangzhi Bu

Shantou University

Jinmei Tang Shantou University

Robert Guang Tian Shantou University

Based on the empirical data from the Chinese stock market and by using an event study method, this paper investigates the relation between the quality award and the market performance of the publicly listed firms that have won quality awards from 2001 to 2009 in China. Our findings show that in the short-term the winners would get significantly accumulated abnormal returns, which differed because of the companies’ size, risk in investment and the prestige of the awards. In the short-run, firms with larger sizes, higher debt ratios, and the China Quality Award (CQA) winners can get accumulated abnormal high returns. Prior leakage of the awards information announcement has played a certain role in the process of abnormal high returns. INTRODUCTION

In an increasingly competitive global economy, it is believed that quality is one of the important

components for most of Chinese companies to survive. So, in their attempts to remain in a competitive position, companies implement with quality management programs one after another, such as ISO certification, total quality management (TQM), Six Sigma, Quality Award etc. As is well known, ISO 9000 is a set of international standards that establish procedures and requirements for the management of quality, and TQM is a set of principles and tools, encouraging continuous improvement and prevention of defects. In business terms, Six Sigma(means 3.4 defects per million opportunities, DPMO), is defined as a business improvement strategy used to improve business profitability, to decrease wastes, to reduce costs of poor quality, and to improve the effectiveness and efficiency of all operations to meet or even exceed customers’ needs and expectations (Antony and Banuelas, 2001). Quality Award is an excellent performance criterion. One important objective of the quality reward is to recognize firms that have done an outstanding job in implementing effective quality improvement programs.

There are many quality awards all over the world, among them, Malcom Baldrige National Quality Award (MBNQA), Deming Prize and European Quality Award are the most famous three. Like MBNQA, China Association for Quality (CAQ) determine China Quality Award (CQA) using the “outstanding performance evaluation criteria” (GB/T19580-2004), a 1000-point scoring system, to evaluate the


candidate leadership, resource management, process management, information and operational results. In 2004, China Association for Quality (CAQ) launched another quality award to the advanced firms for implementing “Performance Excellent Model” (PEM). Currently, CQA and PEM award are the most authoritative quality awards in China granted to organizations that have gained obvious quality, economic and social benefits. For a company to win such an award, it must carry out a strict and effective quality program. Winning a quality reward by a firm provides evidence that it has implemented an effective quality program, such as TQM (Hendricks and Singhal, 1996).

There is a large body of literatures about the influence of implementing effective TQM on firms’ financial performance. Saraph and Benson（1989）are among the first to investigate the effect of TQM on company’s performance, they concluded that there was a significant positive correlation between TQM and business performance. Landmark studies by Hendricks and Singhal (1997, 2001) show that effectively implement TQM can improve companies’ financial performance. Motwani and Mahmoud (1994) concluded that parts of the elements of TQM have a remarkable impact on enterprise operation performance. Terziovki and Samson (1998) also state that TQM has a remarkable impact on enterprise operational performance.

By contrast, less work has been done on estimating the impact of quality improvement program on the market value of the firm, which is a widely accepted measure of market performance. Docking (1999) studies the relationship between ISO9000 certification and stock market abnormal returns by the sample of American listed companies, and they find that firms can get significantly abnormal returns in stock market on the day of passing ISO certification, and smaller companies can get more abnormal returns than bigger ones. Taking 187 Spanish listed companies which have passed ISO9000 certification from 1993 to 1999 as samples, Nicolau (2002) finds that on the announcement day alone, the firms can get 62.9 percent of significantly abnormal returns due to the decrease of market information asymmetry by ISO9000 certification.

Hendriks and Singhal (1996) as well as Adams et al. (1999) are in the same line. Using a sample of 91 announcements of winning quality awards, Hendricks and Singhal (1996) report that the mean abnormal return on the announcement day is a statistically significant 0.64%. Adams et al. (1999) find marginally positive stock price responses on the announcement day for the 20 publicly traded firms that won the Baldrige Award from 1988 to 1997, but they didn’t find evidence of positive abnormal returns for the period from 1992 to 1997.

We find there is no empirical research on the relationship between China Quality Award(CQA) and market value of firms, except BU Xiang-zhi (2007). By extending the samples and inspecting the mean and accumulated abnormal returns of winners, this paper will investigate two issues related to the market value of the firms which won CQA and advanced firms for implementing PEM. First, it investigates whether the firm has positive abnormal returns on the day of the CQA announcement. Second, it examines the factors that affect abnormal returns of the firm after winning the CQA.

In the following sections, we will describe the hypotheses and issues examined. Then we will discuss the process of sample selection and research methodology. Followed by a discussion of the empirical results, and finally summarize the paper. THEROY DEVELOPMENT AND HYPOTHESIS Quality Award and Abnormal Stock Returns

The existing theories and practices demonstrate that the implementation of an effective quality improvement program can significantly improve an enterprise's operational and financial performance. Winning a quality award, which is regarded as a symbol of undergoing a strict and effective quality improvement program, indicates the winners will have better operational and financial performance in the future. Therefore, winning a quality award conveys the good information for the investors that the firm has implemented an effective quality program. Analogue to the commonly TQM literature, improving quality is likely to increase the expected net cash flow and the market value of the firm. As such, the market value and stock price will increase.


However, it is anticipated that the extent of stock market reaction to an announcement of a quality award depends on how well the information contained in winning a quality reward. While winning a quality award is evidence to the investors that the firm has implemented an effective quality improvement program, this information may not be a surprise to the stock market because previous actions of the firms could have indicated to the stock market that the firm is trying to implement a quality improvement program. Those actions include the quality propaganda weeks, leader's speech, media reports about corporate quality and management, etc., If information pre-revealing does exist, the effect of awards to stock price will be underestimated by the short-term performance of stock market (Hendricks and Singhal, 1996). Anyway, winning a quality award has proved the expectations to the past performance of the firm, and it will get a better guarantee for its future operation. Therefore, we expect a positive average abnormal change in the stock prices of a sample of firms that wining quality awards. We give our first hypothesis as follows:

H1: winning a quality award leads to a positive abnormal stock return on the day of the award announcement.

Influence Factors on Abnormal Returns

Firm size is one of the main factors that affect abnormal returns. However, there is no common belief between firm size and abnormal returns. There exist two contradictive theories about it, namely, the cost hypothesis and information hypothesis.

Information hypothesis holds the belief that large firms are more likely to be tracked closely by the public media and financial analysts, and, as such, information about a large firm's quality improvement effort would be more publicly available, and the marketplace may be informed ahead of information about whether quality improvement programs in large firms are effective or not. Thus, the announcement of winning quality awards of large firms may cause smaller reaction in the stock market. Hendricks and Singhal (1996) find that the information of winning a quality award of small firms can produce more active responses than large in American stock market. Siu Y Chan (2001) draws the same conclusion by studying the reaction of Hong Kong stock market to ISO certification. Their research conclusions support the above information hypothesis.

According to cost hypothesis, a firm must pay a fix cost, such as assessment and training cost, for passing a quality certification and/or winning a quality reward. However, for large sized firms the unit revenue will be greater than that for small sized ones. Therefore, in the stock market, the larger firms will have much higher abnormal returns.

However, based on the study of Taiwan’s stock market, Jo-hui Chen (2001) finds that ISO certification of large and small firms can both bring positive abnormal returns, but that is not the same case for medium sized firms, the ISO certification of medium sized ones cause little market reaction. Whatever cost hypothesis and information hypothesis, the firm size can affect the abnormal returns. Accordingly, we give our second hypothesis:

H2A: The size of the quality award winners will have an impact on the magnitude of the mean abnormal return generated.

Apart from size of the firm, we believe that debt ratio is also one of the important factors influencing

abnormal returns. There are two reasons to support our belief. First, according to the capital structure tradeoff theory (also known as MM theory), company’s debt can bring tax shield returns, however market debt cost will increase as debt ratio increasing. When debt ratio reaches some extent, the probability of bankruptcy will increase and the risk of investment will increase too. For this reason, one company's best capital structure is a tradeoff between the tax shield returns of debt with the risk of bankruptcy. Secondly, most of previous studies have proven that effectively implementing quality management program can improve corporate performance, so the bankruptcy risk of a quality award winner is relatively lower. Reasonable debt ratio of winners not only brings tax shield returns, but also increases investors’


expectation on the company’s operation. Based on this, we give the third hypothesis:

H2B: Quality award winners with greater financial leverage can get greater stock abnormal returns.

Although CQA and PEM are the most authoritative quality awards in China, the review process and

the criterion of CQA is much stricter, and the CQA is more famous than PEM in China. We believe that the award prestige has impact on abnormal returns, which brings our fourth hypothesis:

H2C: The CQA winners can get more abnormal returns than the PEM winners.

DATA SOURCES, SAMPLE SELECTION AND RESEARVH DESIGN Data Sources and Sample Selection

The award winners refer to the ones which win the CQA award (including consolation and nomination award) and honorary title of advanced firms of implementing “performance excellent model” (PEM). The samples all come from the announcement in the net of China Association for Quality (CAQ), and award winners listed in the journal of China Quality. In order to keep the data correctness and effectiveness, we selected the samples as follows: (1) Firms have already been listed for more than half a year before winning award. (2) There is no other important announcement in the stock market in the research period. (3) If a firm gets more than one award, we will take the day of winning the first award as the event day. We finally obtain 79 samples from 2001 to 2009, 38 of which are CQA winners (including winners of both consolation and nomination award), 41 of which are winners of PEM. The samples distribution is showed in figure 1. While due to the financial data is not completion, we only get 61 samples from 2001 to 2007 in the long run performance study.

TABLE 1 THE INDUSTRY DISTRIBUTION OF THE SAMPLES

Manufacturing Wholesale

and retail Real

Estate Construction Industry

Transportation and warehouse

Extractive industry

Information Technology

Services Total

CQA 31 1 1 1 1 2 1 38 CEP 27 2 1 6 5 41

FIGURE 1 THE YEARLY DISTRIBUTION OF THE SAMPLES

910

2 2 21

4 4 4

0 0 0 0

12

10 10

8

1

0

2

4

6

8

10

12

14

2001 2002 2003 2004 2005 2006 2007 2008 2009

年份

数量 CQA

CEP

Year

samples


Research Methodology This paper applies the event study methodology to study the market performance of the firms. Event

study methodology is a technique to isolate the component of price change due to firm-specific events by adjusting them for other factors. The component attributed to firm-specific events is typically referred to as the "abnormal" return. The basic idea is to test for the statistical significance of the average abnormal returns on an event date for a sample of firms experiencing the same type of firm-specific event. The average abnormal return is interpreted as capturing the valuation impact of that event (Barber and Lyon, 1997). In short term, we calculate abnormal returns using the market adjusted returns model.

For each observation, calendar time is translated to event time using the following conventions: The quality award announcement event day is denoted day 0 in event time. The next calendar day on which trading took place is denoted day +1 in event time, and the trading day preceding the announcement event date is day -1, and so on. In this paper, the research event window is from the 15th trading day preceding the announcement day to the 15th trading day after that, a total of 31 trading days. In addition, we calculate respectively cumulative abnormal returns of winners for 5 event window periods, which refer to CAR(-10,-1), CAR(-5,-1), CAR(1,5), CAR(1,10), CAR( (-15,15). We tend to make a relatively comprehensive investigation of stock price changes of winners during their announcement periods in a longer time (ZHU, 2006) and to check for any leakage or tardiness of information.

The Market Adjusted Returns Model:

mtitit RRAR −= where itAR is the abnormal return of stock i at day t, itR is the return of stock i at day t, mtR is the

market return on day t. The mean abnormal return, tAAR ,on day t is then computed as:

1

1 N

t itt

AAR ARN =

= ∑

where N is the number of stocks with return information. Thus we can compute the cumulative abnormal return in 1 , ][ Tt as followed:

1

,

T

i T itt t

CAR AR=

=∑

Therefore, the mean cumulative abnormal return is computed as:

1

,1

1 T N

i T itt tt

ACAR ARN = =

= ∑ ∑

EMPIRICAL RESULTS AND ANALYSIS The Overall Performance

Table 2 and Figure 2 gives the empirical results and the trend of mean abnormal return in event period [-5,5] and cumulative abnormal return(CAR) in each window period. The empirical results show that we can get positive cumulative abnormal returns(CAR) of 5.77% in the 31 trading days which begin from the 15th trading day before winning award to the 15th trading day after that and the result is significant at level 1%. We get 1.02% and 1.47% of the statistically significant and positive abnormal returns in the fifth and second trading day respectively preceding winning award, and 2.58% and 1.96% of CAR in the window period (-10,-1) and (-5,-1) respectively.


FIGURE 2 TREND OF MEAN ABNORMAL & CUMULATIVE ABNORMAL RETURN

Table 2 shows that on announcement day the abnormal return is negative and statistically significant

(-0.76%), while the abnormal return preceding announcement day is positive and significant (1.47%), and that in the 2nd day after announcement is also positive and significant (1.15%). At the same time, the cumulative abnormal returns (CAR) in window period (1,10) and (1,5) are -0.63% and 1.76%, both lower than the pre-announcement CAR, which indicate that there is prior leakage of award announcement information, and the stock market is over-reacting to this internal information.

TABLE 2 THE MEAN ABNORMAL RETURN IN EVENT PERIOD [-5,5] AND CAR IN EACH

WINDOW PERIOD (%)

Abnormal Return

Time Period

-5 -4 -3 -2 -1 0 1 2 3 4 5

AAR 1.020 -0.689 -0.054 0.200 1.474 -0.762 0.470 1.152 -0.419 -0.20 0.776

T-value 3.78*** -2.12** -0.21 0.81 4.52*** -2.71*** 1.65 3.60*** -1.11 -0.50 2.45**

P-value 0.000 0.037 0.835 0.422 0.000 0.008 0.103 0.001 0.270 0.616 0.016

Cumulative Abnormal Return(CAR)

Time Period (-10,-1) (-5,-1) (1,5) (1,10) (-15,15)

ACAR 2.583 1.955 1.758 -0.629 5.770 T-value 2.10** 2.80*** 1.96* -0.44 2.88*** P-value 0.039 0.006 0.053 0.663 0.005

① ***,**and* indicate paired t-test significance level at 0.01,0.05 and 0.1 respectively. Influence Factors on Abnormal Returns

Table 3 gives the mean abnormal returns and cumulative abnormal returns for the sample broken down into awards to "large" and "small" firms. A firm was classified as a small firm if it had total assets less than the median total assets of firms in our sample. The results show that large firms get greater CAR, which have obtain 8.76% of CAR in window period ( -15,15), however, the CAR of small firms is not significant in all of the 5 event window periods, which indicates that the value contribution of awards to

-0.04

-0.02

0

0.02

0.04

0.06

0.08

-15 -12 -9 -6 -3 0 3 6 9 12

Period

AAR,ACAR

AAR

ACAR


large firms is greater than to small ones in the window periods. In addition, table 3 shows that stock market has greater over-reaction to the pre-announcement information of small firms than that of large ones, the abnormal return of small firms is -0.96%, its absolute value is larger than that of large firms(-0.56%), which suggests the stock market has drastic over-reaction to small firms during award announcement period. Apparently, large firms can get more abnormal returns in certification announcement period.

TABLE 3 TEST RESULTS OF AR IN GROUP (%)

Window period

-5 -4 -3 -2 -1 0 1 2 3 4 5

Small firmsa（N=40） AAR 1.110 -1.011 -0.142 -0.050 0.970 -0.960 0.148 1.202 0.081 -0.543 0.158 T-value 3.60*** -2.07** -0.43 -0.18 2.17** -2.21** 0.34 2.44** 0.16 -1.00 0.37 P-value .001 0.046 0.667 0.856 0.037 0.033 0.737 0.019 0.874 0.323 0.712

Large firmsb（N=39） AAR 0.935 -0.359 0.036 0.459 1.991 -0.560 0.800 1.070 -0.933 0.147 1.408 T-value 2.07** -0.84 0.09 1.10 4.26*** -1.56 2.23 2.68** -1.67 0.25 3.12** P-value 0.045 0.405 0.930 0.276 0. .000 0.126 0.032 0.011 0.103 0.807 0.003

Low debt ratio firmsc（N=40） AAR 0.597 -0.980 0.026 0.058 1.002 -0.795 0.640 0.821 -0.298 -0.114 0.814 T-value 1.67 -2.00* 0.09 0.20 2.40** -2.06** 1.66 1.71* -0.60 -0.25 2.11** P-value 0.102 0.052 0.933 0.843 0.021 0.046 0.105 0.096 0.550 0.804 0.042

High debt ratio firmsd（N=39） AAR 1.46 -0.39 -0.13 0.35 1.96 -0.73 0.30 1.46 -0.55 -0.29 0.74 T-value 3.65*** -0.91 -0.32 0.86 3.94*** -1.76* 0.70 3.59*** -0.94 -0.44 1.45 P-value 0.001 0.366 0.749 0.398 0.000 0.087 0.489 0.001 0.354 0.666 0.156

CQA（N=38） AAR 0.645 -0.915 0.633 0.508 1.293 0.021 0.542 1.115 -0.70 0.595 1.595 T-value 1.765 -2.31** 1.94* 1.48 2.79*** 0.06 1.70* 2.11** -1.35 1.24 3.55*** P-value 0.086 0.027 0.060 0.147 0.008 0.954 0.098 0.042 0.186 0.222 0.001

CEP（N=41） AAR 1.375 -0.480 -0.691 -0.088 1.643 -1.488 0.403 1.154 -0.157 -0.941 0.016 T-value 3.50*** -0.94 -1.84* -024 3.55*** -3.82*** 0.87 3.16*** -0.29 -1.52 0.04 P-value 0.001 0.353 0.073 0.814 0.001 0.000 0.392 0.003 0.776 0.135 0.969

② Firm size measured by the total asset prior one year winning the award. Debt ratio measured by the asset-debt ratio prior one year winning the award

a Large firm: total assets more than the median total assets of firms in our sample b Small firm: total assets less than or equal to the median total assets of firms in our sample c High debt ratio firms: total debt ratio more than the median total debt ratio of firms in our sample d High debt ratio firms: total debt ratio less than or equal to the median total debt ratio of firms in our sample。

As to debt ratio, we categorize a firm to be a firm with high debt ratio if it had debt ratio that is

greater than the median debt ratios of firms in our sample. Table 4 gives the result that firms with high debt ratio can get 7.78% of abnormal returns in window period (-15,15), and firms with high debt ratio can get 1.96% of abnormal returns on the day prior the award announcement day while low debt ratio firms can get only 1.0% of that. Thus, the third hypothesis is tested, the debt ratio has impact on abnormal return on the event of quality award announcement.


TABLE 4 TEST RESULTS OF CAR IN GROUP (%)

Window period (-10,-1) (-5,-1) (1,5) (1,10) (-15,15)

Small firms（N=40） ACAR 1.462 0.876 1.046 -1.779 2.855 T-value 0.77 0.93 0.83 -0.87 0.95 P-value 0.449 0.360 0.415 0.391 0.348

Large firms（N=39） ACAR 3.733 3.062 2.489 0.551 8.760 T-value 2.42** 3.03*** 1.96* 0.27 3.37*** P-value 0.021 0.004 0.057 0.787 0.002

Low debt ratio firms（N=40） ACAR 0815 0704 1.864 0.487 3.804 T-value 0.42 0.83 1.59 .026 1.49 P-value 0.676 0.414 0.120 0.800 0.146

High debt ratio firms（N=39） ACAR 4.397 3.238 1.650 -1.650 7.788 T-value 2.96*** 2.99*** 1.20 -0.82 2.51** P-value 0.003 0.007 0.416 0.416 0.035

CAQ（N=38） ACAR 2.089 2.163 3.144 2.028 6.300 T-value 0.98 2.38** 2.21** 0.88 2.62** P-value 0.336 0.023 0.034 0.386 0.013

PEM（N=41） ACAR 3.041 1.763 0.475 -3.091 5.280 T-value 2.30** 1.66 0.44 -1.82** 1.68 P-value 0.027 0.104 0.666 0.076 0.105

The test results also show that both CQA and CEP winners can get positive and significant abnormal

return on the prior day before winning award. The abnormal return of PEM winners is 1.64%, greater than PEM winners (1.29%), but the abnormal return of PEM winners are significant negative on award announcement day(-1.49%), which indicates that the stock market is over-reacting to PEM information. The reason for this phenomenon could be that there is prior leakage of information of the awards in the review process, for instance, the selection criteria of quality award mentions that quality winners must have won an award of PEM or excellent customer satisfaction in the prior 3 years before winning award. However, as is shown by the CAR test results, CQA winners can get more abnormal returns, which get significant 6.3% of cumulative abnormal returns(CAR) in window period (-15,15). This suggests that the award prestige does have impact on stock market. Long Term Performance

Some actions of a firm before winning award may have indicated to the stock market that the firm is successful in management, in such case the effect of winning a quality award on stock short-term performance will be underestimated if we just take stock market short-term reaction into account. So we have also tested the long term market performance of the award winners using the buy-and-hold abnormal return (BHAR) method proposed by Barber and Lyons (1997). However, we do not find significant results of long-term abnormal return. Hendriks and Singhal (2001) indicate that a firm is beginning to get abnormal returns in the third year after award. While due to the sample limitation, we just analyze the abnormal returns of the prior year before winning award and two years after that, a total of 3 years. Although we have not got significant test results, this finding tells us that the firms undergoing an effective quality program should be patient, for it will take a long time for the quality program to have a significant impact.


FIGURE 3 THE TREND OF MEAN AND MEDIAN LONG-TERM RETURN

TABLE 5 THE TEST RESULT OF LONG-TERM STOCK PRICE (%)

Period window （-12，-1）（-12，0）（-1，0）（0，24）（-12，24） BHAR（mean） 0.128 0.092 -2.120 0.371 0.304 T-value 0.905 0.695 -0.838 1.587 1.204 P-value 0.369 0.489 0.405 0.118 0.233 BHAR（median） 0.064 0.047 -0.001 0.104 0.248 T-value 0.449 0.343 -0.801 1.143 0.224 P-value 0.655 0.733 0.426 0.258 0.824

CONCLUSIONS

Based on the sample of listed firms from Shanghai and Shenzhen stock markets, we test the effect of

winning a quality award on the abnormal return of listed firms in stock market by studying the change of market return of these firms in the period of award announcement; we can draw the conclusions below:

(1) Quality awards can bring abnormal returns to winners. The cumulative abnormal returns (CAR) is significant and positive 5.77% in the event window period [-15,15]. However, the average abnormal return is negative on the day of award announcement. So we believe that there exists prior leakage of award information, and stock market is over-reacting to the information leakage.

(2) Different with Hendricks and Singhal (1996), we find that the larger firms will get more abnormal returns, and the firms with high debt ratio will get more CAR in the event window [-15,15].

(3) Similar with Hendricks and Singhal (1996), the award prestige have impact on abnormal returns in stock market and the firms winning a higher prestige award (such as CQA) can get more abnormal returns.

For investors, this paper indicates that holding stocks of award winners for short term is profitable, especially firms with large-scale investment and high debt ratio can get more abnormal returns in the short run.

The limitation of this paper is that some of the sample firms winning the CAQ or PEM are branch offices or subsidiaries of listed firms instead of the entire listed firms, which may cause the analysis deviate from reality. In addition, as mentioned above, we just study the award data of the prior year before winning award and two years after that from 2001 to 2007, so if keep a certain amount of samples in the research of long-term stock price, we may get better analysis result if we take a longer time span.

-0.2

-0.1

0

0.1

0.2

0.3

0.4

-12 -9 -6 -3 0 3 6 9 12 15 18 21 24

Period

BHAR Mean

Median


ACKNOWLEDGEMENT The authors would like to thank the referees for their helpful suggestions. This research is supported

in part by National Natural Science Foundation of China under grant NSFC 70701022，71002070；Humanities and Social Sciences Research Project by Ministry of Education(MOE) under grant 11YJC630006, 09YJC630244; Philosophy and Social Science the 12th Five-Year Plan Project of Guangdong Province GD11YGL01; Natural Science Foundation of Guangdong Province 9151503101000015; Academic Innovation Team Construction foundation of Shantou University ITC10004; Liberal Arts Research Foundation of Shantou University SR10003. REFERENCES Adams G, McQueen G and K. Seawright (1999). Revisiting the stock price impact of quality awards, The International Journal of Management Science, 27,595-604. Antony, J. and R.Banuelas (2001). A strategy for survival, Manufacturing Engineer, 80,119-121. Barber, B. M., J. D. Lyon. (1997). Detecting Long-Run Abnormal Stock Returns: The Empirical Power and Specification of Test-Statistics. [J].Journal of Financial Economics. 43, 341-372. BU Xiang-zhi, CHEN Rong-qiu, et al. (2007) .An Empirical Study of China Quality Award on Firm’s Market Value Based on the dada from Chinese Stock Market ,IEEE EXplore, 4230– 4232. Chan , Siu Y.(2001). The Usefulness of International Quality Standards ( ISO) 9000 Certification , BRC Working Papers ,Hong Kong Baptist University , Series No.WP200103. Chen , Jo – Hui. (2001). ISO Certification and Abnormal Return of Stock Price —The Study of the Taiwan Stock Market [J]Review of Pacific Basin Financial Markets and Policies ,2(2) ,109-126. Danny Samson, Mile Terziovski.(1999).The link between total quality management practice and organizational performance [J].International Journal of Quality & Reliability Management, 13(3), 226-237. Docking , Diane Scott and Dowen Richard .(1999). Market Interpretation of ISO9000 Registration [J].The Journal of Financial Research, 22 (2), 147 - 160. Eurico J.F, Amit S., Dale V. (2007). Long-run performance following quality management certification [J]. Review of quantitative finance and accounting, 27, 93~109. George S. Easton., Sherry L. Jarrell (1998). The effects of total quality management on corporate performance: an empirical Investigation [J]. Journal of Business, 71, 253-307. Hendricks Kevin B., Singhal Vinod R.(1996). Quality Awards and he Market Value of the Firm: An empirical investigation [J]. Management Science, 42, 415–436. Hendricks Kevin B., Singhal Vinod R.(1997). Does implementing an effective TQM program actually improve operating performance [J]. Management Science, 9(43), 1258-1274. Hendricks Kevin B., Singhal Vinod R. (2001a). Firm characteristics, total quality management ,and financial performance[J]. Journal of Operations Management, 19, 269-285.


Hendricks Kevin B., Singhal Vinod R.(2001b). The long-Run stock price performance of firm with effective TQM programs [J] Management Science, 47, 359-368. M otwani J.G., Mahmoud E., Rice G. (1994). Quality practice of Indian organizations: An empirical analysis[J].International Journal of Quality and Reliability management , 11, 38-52. Nicolau Juan Luis, Ricardo Sellers. (2002). The stock market’s reaction to quality certification: empirical evidence from Spain [J]. European Journal of Operational Research, 142, 632–641. Saraph J.V., Benson P.G. et al.(1989). An instrument for measuring the critical factors of quality management [J]. Decision Sciences, 20, 810-829. ZHU Tao. (2006). The short-term and long-term stock price performance of merger and acquisition of listed companies [J] Modern Economic Science. l3(28), 31-36(In Chinese).


The Impact of the “BRIC Thesis” and the Rise of Emerging Economies on Global Competitive Advantage:

Will There Be a Shift from West to East?

Richard T. Mpoyi Middle Tennessee State University

The paper examines the thesis that by mid-21st century BRIC economies of Brazil, Russia, India and China (the “East”) would be wealthier than today’s seven largest developed economies of the G7 (the “West”). After analyzing the thesis, the study proposes the following. First, economic power is likely to shift from West to East because the combined GDP of the BRICs would be larger than that of the G7. Second, competitive advantage is less likely to shift from West to East, as after reaching G7’s income levels, BRIC economies would simply at best become as competitive as G7 economies. INTRODUCTION

Over the last ten years, developed countries have grown very little. Comparatively however, several developing economies have expanded at unprecedented rates. As developing countries’ share of the world economic output has been rapidly rising, economists have started to predict that in the next few decades, global competitive advantage will likely shift from West to East. In the context of this prediction, West refers to the United States, Canada, most of Europe, and developed countries of Asia (primarily Japan). On the other hand, East involves a few developing countries of Asia, Latin America and Eastern Europe. The shift in global competitive advantage is a proposition advanced by organizations such as the Brookings Institution (Kharas, 2010; Lieberthal, 2010), Morgan Stanley (Morgan Stanley Capital International, 2011), the Financial Times (Pilling, November 22, 2010), and Investopedia (2011). Of a particular significance is the “BRIC thesis” formulated by Goldman Sachs’ economists (O’Neill, 2001, July 2011; O’Neill & Stupnytska, 2009; Wilson & Purushothaman, 2003; Wilson, Kelston, & Ahmed, 2010). BRIC is an acronym for four largest developing economies of Brazil, Russia, India, and China. The thesis advances that by 2032, the combined GDP of BRIC economies would be as large as that of G7. G7 are seven biggest developed economies (the United States, Japan, Germany, France, the United Kingdom, Italy and Canada). The thesis also suggests that by 2050 BRIC countries would be wealthier than most of current developed countries. The purpose of this research is to study the impact of the growing economic power of developing countries, especially BRIC nations, on the ability of developed countries and their companies to sustain their current competitive edge. In particular, the paper will examine the proposition that competitive advantage is in the process of shifting from West to East.


GLOBAL ECONOMIC ENVIRONMENT

World economies can be divided into two broad categories, developed countries and developing countries (World Bank, 2011). Developed countries are high income economies with a per capita income of US$12,275 or more, and developing countries are low to middle income economies with a per capita income of US$12,274 or less. Increasingly, institutions such as the International Monetary Fund (IMF) refer to developed and developing countries as advanced and emerging economies. For practical purposes, this paper will use developed and developing countries. Later the term emerging economies (as opposed to advanced economies) will be introduced and briefly discussed.

TABLE 1 GDP FOR MAJOR DEVELOPED AND DEVELOPING COUNTRIES

Countries 1980 1990 2000 2010 US$ (Billions) % US$

(Billions) % US$ (Billions) % US$

(Billions) %

Developed: United States Japan Germany France Others

8,163 2,788 1,071 826 691 2,787

76.0 26.0 10.0 7.7 6.5 25.8

17,669 5,801 3,058 1,547 1,249 6,014

80.0 26.2 13.8 7.0 5.6 27.4

25,694 9,951 4,667 1,906 1,333 7,837

80.0 30.9 14.5 5.9 4.2 24.5

41,531 14,658 5,459 3,316 2,583 15,515

66.0 23.3 8.7 5.2 4.1 24.7

Developing: China Brazil India Russia Others

2,544 202 163 182 N/A 1,997

24.0 1.9 1.5 1.7 N/A 18.9

4,511 390 508 326 N/A 3,287

20.0 1.8 2.3 1.5 N/A 14.4

6,533 1,198 642 480 260 3,953

20.0 3.7 2.0 1.4 0.8 12.1

21,378 5,878 2,090 1,538 1,465 10,407

34.0 9.3 3.3 2.4 2.3 16.7

Total World 10,707 100.0 22,180 100.0 32,227 100.0 62,909 100.0 Source: Data retrieved from the International Monetary Fund online database (April 2011)

In 2000, most economic activities originated in developed countries (Table 1). The GDP of developed countries represented 80% of global output. With close to one-third (30.9%) of world GDP, the U.S. was by far the largest economy in the world. The other developed countries that accounted for a significant share of world economy in 2000 were Japan (14.5%), Germany (5.9%), and France (4.2%).

Not surprisingly, companies from these countries also dominated the competitive arena in 2000. Based on Global Fortune 500 (see Table 2), 95.4% (477 companies) of the largest and arguably most competitive companies were from developed countries. The US was home to more than one-third (37%) of those companies. Japan, France and Germany followed with respectively 20.8%, 7.4%, and 6.8%.

As companies grew through diversification strategies (either industry and/or international diversification), competition intensified in most industries (Chandler, 1990). Firms that had achieved powerful advantages in their industry used their organizational capabilities to compete in other industries (inter-industry competition). In an effort to sustain their existing competitive strength in the long term, companies from developed countries also expanded their activities in various regions around the world (international competition). Regional diversification helped major Western companies to take advantage of national differences in cost and quality of factors of production (e.g. labor, energy, land, and capital). For Hill (2009), massive investments in global markets, especially in developing countries of Asia and Latin America, contributed to lowering overall cost structure and/or improving the quality or functionality of product offerings.


TABLE 2 GLOBAL FORTUNE 500 FOR MAJOR COUNTRIES

Countries 2000 2005 2010 # firms % # firms % # firms %

Developed: United States Japan France Germany Others

477 185 104 37 34 117

95.4 37.0 20.8 7.4 6.8 23.4

455 170 70 38 35 142

91.0 34.0 14.0 7.6 7.0 28.4

425 136 71 39 37 142

85.0 27.2 14.2 7.8 7.4 28.4

Developing: Brazil Russia India China Others

23 3 2 1 12 5

4.6 0.6 0.4 0.2 2.4 1.0

45 4 5 6 20 10

9.0 0.8 1.0 1.2 4.0 2.0

75 7 6 8 46 8

15.0 1.4 1.2 1.6 9.2 1.6

Total World 500 100.0 500 100.0 500 100.0 Source: Data retrieved from CNNMoney (2011)

TABLE 3 LOCATION OF GLOBAL PRODUCTION FOR LEADING AUTOMAKERS

Companies 2000 2009 Location Units (000) % Location Units (000) %

Toyota

Japan US/Canada European Union Indonesia/Thai Others Total

4,151 1,103 178 169 354 5,955

70 19 3 2 6 100

Japan US/Canada Indonesia/Thai China Others Total

3,543 1,190 721 601 1179 7,234

49 17 10 8 16 100

General Motors

US/Canada European Union Brazil/Mexico Australia Others Total

5,186 1,955 778 133 81 8,133

64 24 9 2 1 100

China US/Canada European Union Brazil/Mexico Others Total

1,769 1,540 1,138 950 1062 6,459

27 24 18 15 16 100

Volkswagen

European Union Brazil/Mexico China Others Total

3,769 935 316 87 5,107

74 18 6 2 100

European Union China Brazil/Mexico Others Total

3,612 1,244 1,100 111 6,067

60 20 18 2 100

Ford

US/Canada European Union Brazil/Mexico China Others Total

4,430 2,249 385 27 232 7,323

60 31 5 0 4 100

US/Canada European Union Brazil/Mexico China Others Total

1,629 1,660 579 446 371 4,685

35 35 12 10 8 100

Source: Data retrieved from the OICA online database (2011).


Foreign direct investments allowed global companies such as Coca Cola, McDonald’s, General Motors, Sony, Siemens, and many more, to become well recognized names throughout the world. The automobile industry is an example of how global companies have been redirecting their investments away from developed countries toward developing nations.

In 2000, leading automakers were performing well over half of their production activities in their home countries. Toyota, General Motors and Ford made respectively 70%, 64% and 60% of their vehicles in their domestic markets (Table 3). Volkswagen (VW) produced 74% of its automobiles in the European Union. Also, most of their remaining operations were located in other developed countries. By 2009 however, domestic production for Toyota, General Motors and Ford dropped below 50%. Although the percentage of VW’s production in the European Union was still above 50%, it declined from 74% in 2000 down to 60% in 2009.

As developing countries attracted more foreign direct investments, major automobile companies gradually shifted an increasing share of their production to these nations. In 2009 for example, General Motors had moved a sizable portion of its activities out of North America to China. Of the 8.1 million vehicles that the company produced in 2000, 5.2 million (64%) were made in North America, and only 30,000 (nearly 0%) in China. But in 2009, China had become the single most important location for General Motors’ activities. In that year, General Motors made more automobiles in China (1.8 million or 27% of its production) than in the United States and Canada combined (1.5 million). The situation was not much different for the other leading vehicle manufacturers. RISING GLOBAL COMPETITIVE ADVANTAGE OF BRIC COUNTRIES

The shift in foreign direct investments toward developing countries occurred, not only in the automobile industry, but also in many other sectors. Although developed countries were still the main destination for foreign direct investments, developing nations have been attracting a growing portion of these investments (Hill, 2009). Western companies were increasingly locating their operations in developing countries of Asia (particularly China and India), Latin America (for example Brazil), and to some extent, Eastern Europe (e.g. Russia).

Though necessary for Western companies to sustain their global competitive advantage in the long term, geographic diversification of investments also had undesirable consequences. It led to a transfer of managerial and technological know-how to these countries. To effectively operate in foreign countries, multinational firms needed to hire locals who were, over time, able to master technical and organizational skills. Subsequently, some developing economies saw an increase in the pool of highly qualified workforce that could contribute to the start of local businesses. At the same time, the superiority of the West came to be perceived as a benchmark and therefore a target to emulate. As democratic systems and free market mechanisms were key characteristics of most developed countries, several developing countries carried out political and/or economic reforms, in the hope to facilitate the emergence of their economies and companies. Massive investments in developing countries, coupled with both the transfer of managerial and technological capabilities, and the transformation of political and economic systems, laid the ground for the rise of new economic powers, primarily the BRICs.

During the first decade of the twenty-first century, the size of BRICs as well as that of some other developing economies grew so rapidly that the global economic and competitive environment began to drastically change. In 2010, the United States was still the largest economy in the world. However, one by one, the other developed countries were surpassed by China, and the remaining BRICs (Brazil, Russia and India) were quickly closing the gap. As can be seen in Table 1, in only ten years, the share of global GDP from developing economies has gone from 20% in 2000 to 34% in 2010. A closer look at China can better highlight the emergence of BRICs and other developing countries. In 1990, China, then a minor player on the global economic scene, accounted for a mere 1.8% of world output. Its share increased to 3.7% in 2000 and to an astonishing 9.3% in 2010, making China the second largest economy in the world, ahead of economic giants such as France, Germany, and Japan.


The economic emergence of developing countries can further be noticed in the standing of these countries’ firms among the largest companies in the world. As discussed earlier, companies from developed countries accounted for 95.4% (477) of the largest 500 companies in Global Fortune 500 in 2000 (see Table 2). In 2010 however, their share went down to 85%, a 10% decline in just ten years. During the same period, the percentage of companies from developing countries more than tripled. From only 4.6% (23 companies) in 2000, their share increased to 15% (75 firms) in 2010. It is worth noting that BRICs were home to most companies from developing countries. The number of companies from BRICs on Global Fortune 500 was 67 (13.4%) in 2010, up from 18 companies (3.6%) in 2000. Over two third (46 firms) of BRIC companies were from China.

To reflect the new and significantly changed global marketplace, the terminology used in the classification of world economies has been evolving. Rapidly rising developing countries (with BRICs in the lead) came to be known as “emerging economies”. Currently, the International Monetary Fund (2011) groups world economies into two categories, advanced economies (developed and high income countries), and emerging and developing economies (low to middle income countries). Recently, Goldman Sachs Asset Management has proposed the term “growth markets” (O’Neill, April 2011). According to O’Neill, to describe the countries that are driving most of the positive momentum behind the world economy as emerging economies is no longer appropriate. Growth markets would be countries outside the developed world with a growth rate well above the world average. These economies include BRICs, as well as such developing nations as Mexico, Indonesia and Turkey.

The emergence of BRICs and other developing countries raises questions about how the global marketplace would look like in the next several decades. The first decade of the twenty-first century however offers some early indications. During the last ten years (2000-2010), BRICs have made their mark on the global economy by contributing to over a third of world GDP growth (Wilson, Kelston & Ahmed, 2010). Perhaps the sectors that best illustrate the rapid growth of developing nations are energy consumption and automobile production. According to a team of economists at BP (BP Statistical Review of World Energy, 2011), global energy consumption rebounded strongly and grew by an average of 5.6% in 2010, the highest increase in percentage since 1973. While energy demand from advanced economies grew by 3.5%, it increased by 7.5% in developing regions, especially in BRICs. With an expansion of 11.2% in 2010, China has become the globe’s largest energy consumer, a position the United States held from the early 1900s until 2009 (Barr, 2011; Swartz & Oster, 2010). In 2010, the Chinese share of global energy consumption reached 20.3%, against 19% for the United States. Looking ahead, the International Energy Agency (2011) estimated that developing economies would account for 93% of projected increase in global energy demand, as the economic activity of these nations will grow at much faster rates.

Besides the energy sector, the global automobile production also exemplifies how developing economies are gaining economic ground. Table 3 above showed that automakers had been shifting their production facilities from developed economies to developing countries of Asia and Latin America.

As a consequence of the new trend in foreign direct investments, developing nations have gradually emerged as major motor vehicle manufacturing nations. In 2000, only one BRIC country, China, and another developing economy, Mexico, were among the top ten manufacturers of automobiles (Table 4). Ten years later, in 2010, three of four BRICs (Brazil, India, and China), and Mexico, were on the list of the top ten. Remarkably, China moved from being number 8 in 2000 (when it accounted for 3.5% of world production) to becoming by far the biggest automobile producing country in the world. In 2010, close to one quarter (23.5%) of all motor vehicles were made in China. Japan, the second largest producing country, manufactured a little over half of China’s level. Clearly developed economies have been losing their dominance in the automobile industry.


TABLE 4 TOP TEN AUTOMOBILE PRODUCING COUNTRIES

2000 2010 Countries Units (000) % Countries Units (000) % 1. United States 2. Japan 3. Germany 4. France 5. South Korea 6. Spain 7. Canada 8. China 9. Mexico 10. UK Others

12,800 10,141 5,527 3,348 3,115 3,033 2,962 2,069 1,936 1,814 11,629

21.9 17.4 9.5 5.7 5.3 5.2 5.1 3.5 3.3 3.1 20.0

1. China 2. Japan 3. United States 4. Germany 5. South Korea 6. Brazil 7. India 8. Spain 9. Mexico 10. France Others

18,265 9,626 7,761 5,906 4,272 3,648 3,537 2,388 2,345 2,228 17,882

23.5 12.4 10.0 7.6 5.5 4.7 4.5 3.1 3.0 2.9 23.0

Total 58,374 100.0 Total 77,858 100.0 Source: Data retrieved from the OICA online database (2011)

The production of automobiles was among the factors contributing to increased energy consumption in developing countries. Both automobile production and high energy consumption were signs of a rapidly growing demand for goods and services from the middle class in developing economies. The emergence of a sizable middle class in developing countries was a significant development, because the middle class has been seen as having attributes that result in increased consumption (e.g. Murphy, Schleifer, & Vishny, 1989; Schor, 1999). Specifically, in addition to a constant and upscaling of lifestyle norms, the middle class is characterized by the pervasiveness of conspicuous and status goods, and by its willingness to pay a little extra for quality. In Murphy et al.’s view, these characteristics constitute a force that drives economic growth since they feed investment in innovation, production and marketing. The fact that the United States had the largest middle class throughout the twentieth century led Kharas (2010) to suggest that the global economy relied on US consumption for its growth. Recently, the middle class in BRICs and a few other developing nations has been rising quickly (Kharas, 2010; Wilson, Kelston & Ahmed, 2010). There were for example 150 million middle class consumers in China in 2010, and by 2021, this number is predicted to top 670 million (Kharas, 2010). Kharas also estimated that by 2015 the size of middle class consumers in Asia will probably equal that of North America and Europe combined.

The changes in the energy and the automobile sectors, along with the growing middle class in developing nations, may be indicative of broader economic shifts that the global competitive environment is likely to go through in the future. Recently, phrases such as ‘the BRIC decade’ (Wilson, Kelston, & Ahmed, May 2010) or ‘a cross-over from West to East’ (Kharas, 2010) have been used by economists to highlight the changes that are increasingly taking place in the global economic environment. The Wall Street Journal (WSJ) for example suggested that China could surpass the United States as the main driver of the global economy in 2012 (WSJ, November 10, 2010). O’Neill (2011) predicted that by 2020, BRICs would be responsible for close to 50% of the increase in global GDP. In a 2003 report, economists from Goldman Sachs advanced the proposition that by 2050, BRIC economies would be wealthier than most of today’s developed countries. They called this proposition the “BRIC thesis” (Wilson & Purushothaman, 2003). DOES THE BRIC THESIS MAKE SENSE?

It is still possible that brain drain and entrepreneurship could help to revitalize businesses in developed economies. Also, one may conceive of systemic problems that may slow down BRICs’ rapid growth. Assuming that BRICs will progress the way they do now, in the long term, the standing of current


developed countries is likely to be threatened by two factors. The first factor is the size and quality of developing countries’ population. Based on estimates from the International Monetary Fund (IMF, 2011), there were 6.8 billion people in the world in 2009. About 80% of them were from developing countries, with nearly 42% living in BRICs. China, the most populous country, had more than 1.3 billion people. The other BRICs, India, Brazil and Russia had respectively 1.2 billion, 193 million, and 140 million. In addition to population size, developing nations have improved the quality of their education system, as illustrated by the results of a 2009 exam given by PISA (Program for International Student Assessment). The exam was taken by some 470,000 students in 65 countries. Scores from the PISA test showed not only that the students from a number of developing nations performed very well, but also that educational attainments in countries such as China were higher than those of wealthy nations like the United States (Armario, 2010). Because of their large population, and the growing number of highly skilled workers, BRIC countries may forge ahead as far as the size of their GDP.

The second factor is economic convergence. According to the convergence theory, in any long period, the gap in per capita income levels across countries tends to close. The basis for economic convergence is the catch-up hypothesis, which asserts that being backward carries a potential for rapid advance (Abramovitz, 1986). Rapid advance occurs because high income countries’ superiority in technology provides poor countries with a target to emulate, thus an opportunity for rapid growth. The larger the technological gap between high and low income countries, the stronger the low income countries’ potential for high growth in productivity.

TABLE 5 PER CAPITA INCOME FOR MAJOR COUNTRIES (IN US$)

Countries 1980 1990 2000 2010 Developed economies: United States Japan Germany France

12,249 9,172 10,759 12,865

23,198 24,774 19,610 22,017

35,252 36,800 23,220 22,574

47,284 42,820 40,631 41,019

Developing economies: China Brazil India Russia

205 1,372 263 N/A

341 3,464 378 N/A

946 3,751 460 1,775

4,382 10,816 1,265 10,437

Source: Data retrieved from the International Monetary Fund online database (April 2011). Empirical evidence has supported the occurrence of economic convergence in the past (Ball, Hallahan, & Nehring, 2004). For example, studies have found that per capita income levels among today’s developed countries converged over the last century, particularly after World War II (e.g. Abramovitz, 1986; Baumol, 1986; Nelson, 1991). Because income differences between developed and developing nations are significant (see Table 5), economic convergence should take place.

Since institutional reforms fuel growth (Mpoyi, 2008; Yedder, 2005), and therefore facilitate the process of catching up, developing countries such as BRICs have restructured their economic and educational systems to make their environments more favorable for business. For example, planned and controlled economy was replaced by free market in countries such as China and Russia. By promoting entrepreneurship while also attracting foreign investments, the reforms opened up opportunities for rapid growth and modernization in developing nations. Indeed, as Table 6 shows, BRIC economies achieved GDP growth rates significantly higher than those of developed economies. The outcome of such high growth rates has been a rapid increase in BRICs’ per capita incomes. From 2000 to 2010, the per capita


income of all BRIC economies more than doubled. China’s per capita income went from $946 in 2000 to $4,382 in 2010, more than fourfold increase (Table 5).

TABLE 6 GDP GROWTH RATES FOR MAJOR COUNTRIES

Countries 1981-1985 1986-1990 1991-1995 1996-2000 2001-2005 2006-2010 Developed: United States Japan Germany France

3.29 4.28 1.18 1.59

3.25 5.01 3.45 3.27

2.52 1.41 2.15 1.16

4.30 0.97 2.01 2.82

2.40 1.31 0.56 1.63

0.96 0.18 1.18 0.75

Developing: China Brazil India Russia

10.78 1.20 5.23 N/A

7.92 2.09 5.95 N/A

12.28 3.10 5.00 N/A

8.62 2.02 6.18 1.77

9.76 2.80 6.51 6.13

11.20 4.41 8.57 3.61

Source: Data retrieved from the International Monetary Fund online database (April 2011).

Implied in the convergence theory is a progressive slowdown in backward economies’ growth rates as the gap between their incomes and those of advanced economies closes. So as they catch up, BRIC economies should anticipate slowly declining growth rates. Once convergence is completed, BRIC nations will become high income countries and will therefore be part of advanced economies. At that point, probably a few decades from now, the global economic environment would be a different marketplace that can be described as follows. First, with the addition of new developed countries (e.g. BRICs), the pool of advanced economies will be much larger. Second, the combined GDP of the East will be bigger than that of the West because of the population factor (i.e. shift from West to East as far as GDP). Third, national competitive advantage across the larger pool of advanced will rather be similar to the one that exists among current developed countries (i.e. no shift from West to East when it comes to competitive advantage). In the end, companies’ ability to achieve global competitive advantage will be primarily dependent on organizational capabilities, regardless of firms’ home countries. DISCUSSION

There is little doubt that given their population and increasingly qualified workforce, BRIC countries’ GDP will be larger than that of developed Nations. Therefore, economic power will shift from West to East. Also, the East will probably close the competitiveness gap with the West. However, for several reasons, it is difficult to see how competitive advantage will shift from West to East (i.e. East becoming more competitive than West). First, the history of past economic convergence provides evidence that countries that were successful in catching up did not gain a significant competitive edge over previously advanced economies. For instance, after their income levels converged toward those of the United States, Western Europe and later Japan did not become more competitive than the United States. Consistent with the convergence theory, the East should expect a gradual slowdown in growth rate. In fact, a decline in growth rates may have already started in the largest BRIC economy. The Wall Street Journal (Back, October 16, 2011) and the Associated Press (McDonald, October 18, 2011) have reported that China’s economic growth eased to 9.1% in the third quarter of 2011, slowing from 9.5% in the second quarter, 9.7% in the first quarter, and 10.3% in 2010. A 9.1% growth rate is still robust, and the slight slowdown may be the result of troubles that China’s main trading partners are going through, the European Union debt crisis, and high US unemployment. Nonetheless, BRICs should start to see a slight but gradual moderation in their economic growth rates.


There is a second reason why competitive advantage is less likely to shift from West to East. China for example will need to address the consequences of an aging population (the result of a one-child policy). In the long term, this might cause a shortage of qualified workforce. Third, China would have to face global responsibilities that come with economic power. Because of its wealth, China will be expected to play a much greater role in addressing global concerns. It is however unclear how China would respond to growing demand to spend more resources in order to deal with problems such as solving military conflicts, intervening in humanitarian crises, or bailing out economies in difficulties. All these issues may ultimately limit the ability of BRIC economies (China in particular) to take over global competitive advantage in the future. While BRIC nations would be confronting these multiple issues, Ferguson (2011) argues that the West would have the opportunity to reform itself, upgrade its civilization, and in his own words ‘reboot the software before it is too late’.

The fourth and much more important reason is related to structural rigidities. It is well known that Japan went through a lost decade in the 1990s, considerably limiting its economic expansion. The lost decade was caused in part by societal orientations such as reliance on an inter-organizational cooperative form called Keiretsu which involves numerous inefficiencies. Likewise, BRIC nations are and/or will be facing systemic issues, such as those that the largest BRIC economy, China, will confront in years to come. China has been quickly catching on, and also catching up, essentially because of its ability to emulate the West. Specifically, China successfully imitated five of six social developments that historian Ferguson (2011) identified as institutions that allowed the West to become the preeminent political and economic force in the modern world. The five social developments that China has mastered include competition, science, medicine, consumption and work ethic. However, as Ferguson pointed out, China has failed to effectively integrate into its political and legal system the rule of law that is the basis for the sixth social development, private property rights. For him, the reason for that failure is because private property rights are the outcome of a democratic system of representative government. Ferguson’s argument is in line with an assumption that is at the core of the convergence theory. For rapid growth to occur, the convergence theory assumes that a backward country needs to have enlarged social capabilities. Social capabilities are tenacious societal characteristics such as education system, and political and economic institutions (Abramovitz, 1986; Nelson, 1991; Lusigi, Piesse, & Thirtle, 1998). Social capabilities are considered to be enlarged (or advanced) when a nation has an effective education infrastructure, a stable political environment, and a market economic system. As is the case with other BRIC nations, China has for the most part achieved an effective education system and successfully adopted free market system. Despite its economic success, China has been holding on to its communist regime. The China Daily, a Chinese newspaper, has referred to China’s political system as social democracy, and to the interesting combination of social democracy and free market as socialist market economy (O’Neill, July 2011). Ferguson’s argument that economic success should be supported by a democratic political system implies that without true democratic reforms, China would someday reach a point where its totalitarian system led by the communist party would clash with its growing wealth. For O’Neill (July 2011), the fact that the clash hasn’t happened yet is the greatest global contradiction in a long time. If and when economic success and communist system collide, the extraordinary economic expansion of China will slow down or even come to a halt. Convinced that a clash between economic success and totalitarianism is certain, Ferguson does not believe that the future belongs to China.

For these reasons, the East will at best reach the competitiveness level of the West, but it will be less likely to take over global competitive advantage. Still, as discussed earlier, BRIC economies have the potential to grow faster in the few decades to come. The rise of high growth markets of BRICs may have a few competitive implications. The anticipated momentous shift of economic strength (in terms of GDP) from West to East led Newsweek magazine to suggest that Africa will have an opportunity to become the new Asia (Guo, March 2010). About two decades ago, entrepreneurship powered by the influx of returning skilled workers, the frenetic urbanization, a big push in services and infrastructure were among the key factors that transformed the economies of China and India. These factors are now driving a rapid emergence of middle class in Africa’s most robust economies (Majahan, 2009). A growing middle class in Africa will increase consumption, a necessary condition to attract outside talent and capital (Guo,


March 2010). Furthermore, as incomes increase in the East (e.g. BRICs), multinationals will redirect some of their foreign direct investments towards Africa as it is still abundant in low labor costs. To become the new Asia however, Africa would need to undertake profound institutional transformations in its political, legal and economic systems (Mpoyi, Festervand, & Sokoya, 2006). CONCLUSION

This paper was an attempt to examine the consequences of the unprecedented rise of developing economies, particularly BRIC nations, on the future of the global competitive landscape. Unless unexpected events drastically slow down the rapidly expanding GDP of BRICs, the income levels of developing economies of Brazil, Russia, India and China are predicted to converge toward those of developed economies of the West. Given their disproportionally large population, it appears likely that the combined GDP of BRICs will be bigger than that of Western countries. So the thesis that in the next few decades BRIC nations will be wealthier than today’s developed countries has merit.

Although possible, the suggestion that the East will gain competitive edge over the West is questionable. It is hard to make the case that the West will lose global competitive advantage in favor of the East. This did not happen in the past. After World War II, the United States remained the only economic power. Then income levels of Western Europe and Japan gradually converged toward those of the United States. Once they converged, Western Europe and Japan did not become more competitive than the United States. So, in itself, economic convergence does not lead to a shift in competitive advantage. Also, as it catches up, the East will encounter structural difficulties such as inadequate social capabilities. These difficulties have the potential to slow the rapid growth of the East in the long-term. REFERENCES Abramovitz, M. (1986). Catching up, Forging Ahead, and Falling Behind. The Journal of Economic History, 46: 385-406. Armario, C. (December 07, 2010). Wake-Up Call: U.S. Students Trail Global Leaders. Associated Press, http://abclocal.go.com/wpvi/story?section=news/national_world&id=7829170 [accessed December 07, 2010]. Back, A. (October 16, 2011). China’s Economic Growth Slows. The Wall Street Journal, http://online.wsj.com/article/SB [accessed October 18, 2011]. Ball, V.E., C. Hallahan, & R. Nehring (2004). Convergence of Productivity: An Analysis of the Catch-Up Hypothesis within a Panel of States. American Journal of Agricultural Economics, 86 (5): 1315-1321. Barr, R. (June 08, 2011). Developing Countries Lead Surge in Energy Demand. Associated Press, http//:www.rdmag.com/News/FeedsAP/2011/06/energy-developing-countries [accessed August 23, 2011]. Baumol, W. (1986). Productivity Growth, Convergence, and Welfare: What the Long-Term Data Show. American Economic Review, 76 (5), 1072-1085. BP Statistical Review of World Energy (2011). Energy in 2010: A Strong Rebound. Statistical Review. http://bp.com/statisticalreview [accessed July 09, 2011]. Chandler, A.D. (1990). Scale and Scope. Cambridge, MA: The Belknap Press.


CNNMoney (2011). Global Fortune 500. http://money.cnn.com/magazine/fortune/global500/ [accessed July 04, 2011]. Ferguson, N. (2011). Civilization: The West and the Rest. New York, NY: Penguin Group Inc. Guo, J. (February 19, 2010). How Africa is Becoming the New Asia. Newsweek . http://www.newsweek.com/id/233501/output/print [accessed February 25, 2011]. Hill, C. (2009). International Business. New York, NY: McGraw-Hill Irwin. International Energy Agency (2011). World Energy Outlook 2010. http://www.iea.org [accessed June 28, 2011]. International Monetary Fund (April 2011). World Economic Outlook. http://www.imf.org/weoforum [accessed June 07, 2011]. Investopedia (2011). What does Brazil, Russia, India and China Mean? ValueClick, http://investopedia.com/corp/about.asp [accessed July 06, 2011] Kharas, H. (2010). The New Global Middle Class: A Cross-over from West to East. In C. Li (Ed.), China’s Emerging Middle Class: Beyond Economic Transformation. (Chapter 2). Washington, DC: Brookings Institution Press. Lieberthal, K. (2010). Is China Catching Up with the US? Ethos, 8: 12-16. Lusigi, A., J. Piesse, and C. Thirtle (1998). Convergence of Per Capita Income and Agricultural Productivity in Africa. Journal of International Development, 10 (1): 105-115. Majahan, V. (2009). Africa Rising. Saddle River, NJ: Wharton School Publishing. McDonald, J. (October 18, 2011). China Economic Growth Slows. The Associated Press, http://articles.boston.com/2011-10-18/business/30293196_1_sheng-laiyun-national-statistics-bureau-global-growth [accessed October 18, 2011]. Morgan Stanley Capital International (2011). SCI Emerging Markets Indices. Morgan Stanley, http://www.msci.com/products/ [accessed July 06, 2011]. Mpoyi, R.T. (2011). Advanced Economies’ Competitive Advantage Under Threat: Are Emerging Economies Catching Up or Forging Ahead? Competition Forum, 9 (1): 49-57. Mpoyi, R.T. (2008). Impact of National Institutions on Companies’ Global Competitive Advantage: Dominance of Companies from Countries with Advanced Social Capabilities. Journal of Business Administration Online, 7 (2). http://jbao.atu.edu/ [accessed March 19, 2009]. Mpoyi, R.T., T.A. Festervand, & S.K. Sokoya (2006). Creating a Global Competitive Advantage for Sub-Saharan African Companies. Journal of African Business, 17 (1): 123-142. Murphy, K., A. Schleifer, & R. Vishny (1989). Income Distribution, Market Size and Industrialization. Quarterly Journal of Economics, 104 (3): 537-564.


Nelson, R.R. (1991). Diffusion of Development: Post-World War II Convergence among Advanced Industrial Nations. The American Economic Review, 81: 271-275. OICA (French acronym for International Organization of Motor Vehicle Manufacturers) (2011). World Motor Vehicle Production. http://oica.net/category/production-statistics/ [accessed July 06, 2011]. O’Neill, J. (2001). Building Better Global Economic BRICs. Goldman Sachs Global Economics, Paper No 66, November 30, 2001. O’Neill, J. (April 2011). Introducing “Growth Markets”. Goldman Sachs Outlook , http://www2.goldmansachs.com/our-thinking/global-economic-outlook/ [accessed October 18, 2011]. O’Neill, J. (July 2011). China, China, China. Viewpoints, July 03, 2011. O’Neill, J., & A. Stupnytska (2009). The Long-Term Outlook for the BRICs and N-11 Post Crisis. Goldman Sachs Global Economics, Paper No 192, December 4, 2009. Pilling, D. (November 22, 2010). Asia: Poised for a Shift. Financial Times, http://ft.com/cms/s/682dccb6-f66e- [accessed December 07, 2010]. Schor, J. (1999). The New Politics of Consumption: Why Americans Want so Much More Than they Need. Boston Review, Summer: 1-8. Swartz, S., & S. Oster (July 18, 2010). China Tops the U.S. in Energy Use. The Wall Street Journal, http://online.wsj.com/article/ [accessed July 02, 2011]. The Wall Street Journal (November 10, 2010). China Could Surpass the U.S. in 2012. http://wsj.com/economics/2010/11/10 [accessed December 07, 2010]. The World Bank Group (July 01, 2011). Changes in Country Classifications. http://data.worldbank.org/about/country-classifications [accessed July 03, 2011]. Yedder, O.B. (2005). Reforms Fuel Growth Boom. African Business, 308: 18-19. Wilson, D., A. Kelston, & S. Ahmed (2010). Is This the ‘BRICs Decade’? BRICs Monthly, 10 (3), May 20, 2010. Wilson, D., & Purushathaman, R. (2003). Dreaming with BRICs: The Path to 2050. Goldman Sachs Global Economics, Paper No 99, October 1st, 2003.


The Effect of Product Demand Decline on Investments in Innovations: Evidence from the U.S. Defense Industry

Donald W. Gribbin

Western Michigan University

Hong Qian Oakland University

Ke Zhong

Central Washington University

The end of the Cold War led to a substantial decline in defense product demand. This study investigates the effects of product demand decline on defense firms’ investments in R&D for innovations. Our evidence indicates significant lower levels of R&D intensity for the low demand period (1993 to 1998) than for the high demand period (1984 to 1989). We also find significant declines in the defense firms' return on assets over the period, which is mainly attributable to a significant decrease in the firms' efficiency of using assets to produce sales. The defense firms, despite decline in defense product sales, generally maintained their total sales by partially shifting their capacity to commercial markets, which might be at the sacrifice of profitability, operating efficiency, and R&D investments for innovations. INTRODUCTION

In the organizational decline literature, researchers debate on whether the decline in product demand will inhibit or stimulate the innovations of an organization (e.g., the review by Mone, McKinley, and Barker, 1998). One stream of studies contends that organizational decline including product demand decline would inhibit innovations. Resource scarcity due to product demand decline restricts information processing of an organization and imposes an urgency to conserve resources, which leads to organizational rigidity and undermines its capacity to innovate (Staw, Sandelands, and Dutton, 1981; Cameron, Whetten, and Kim, 1987; D’Aunno and Sutton, 1992; Ocasio, 1995; Barker and Mone, 1998). In contrast, the other stream of studies suggests that product demand decline could serve as a stimulus for innovations. They contend that poorer performance caused by product demand decline pressures an organization to be more risk-seeking in its investment decisions, and motivates it to search for innovative solutions for improving its performance (Miles and Cameron, 1982; Cameron, 1983; Lant, Milliken, and Batra, 1992; Haveman, 1993; Hundley, Jacobson, and Park, 1996; Wiseman and Bromiley, 1996).

The end of the Cold War in 1989 led to a substantial decline in US defense procurement spending. US Defense procurement from the defense industry had declined from more than 120 billion for 1984 to around 45 billion for 1998, both in constant 1999 US dollars (the Department of Defense, Green Book 1999). The defense industry during 1980s-1990s provides a natural context to investigate the effects of


product demand decline. Thus, the purpose of this study is to empirically examine the effects of declining product demand during 1980s-1990s on R&D investments of US defense-contracting firms. The evidence provided by this study helps to test which of the competing views could better explain the US defense industry’s response to the declining demand subsequent to the end of the Cold War.

The declining demand for defense products substantially deteriorated the operating environment of US defense firms (Lundquist, 1992). There have been extensive discussions on defense firms’ strategy alternatives for rapidly declining demand in defense markets (Lundquist, 1992; Minnich, 1993; Dial and Murphy, 1995; Gholz and Sapolsky, 2000). The strategies adopted by major defense contractors include “acquisitions to achieve critical mass; diversification into nondefense areas, or converting defense operations to commercial products and services; globalization, i.e., finding international markets for defense operations; downsizing and consolidation; and exit” (Dial and Murphy, 1995, page 293). However, there is little evidence on the defense industry’s strategic responses to the decline in defense products demand, and the effect on their investments in innovations. Our study intends to close this gap in the literature.

We believe the paper has importance to R&D researchers, as it finds that a substantial decline in product demand could significantly undermine firms’ motivation and capacity for R&D investments for innovations. Keeping a R&D lead in developing advanced defense products is a critical strategy for the U.S. national security (Rogerson, 1989). This study could provide evidence helpful to assess the trends in defense R&D investments for innovations. Many technological innovations initially developed for military purposes have been later adapted for commercial purposes. Software and the internet are two classical examples (Campbell-Kelly, 2003). These technological innovations developed by the defense industry have since become critical to competitive advantages of the US economy. Thus, the effect of product demand decline on R&D investments for innovations of defense contractors could have implications beyond the area of national security. With the Great Recession since 2008 and subsequent deleveraging of U. S. customers, our study could also be relevant to the current issues in general with R&D investments in the U.S.

The organization of the remaining of this study is as follows. Section 2 develops the major hypotheses. Section 3 describes the sample firms and data characteristics. Section 4 presents the empirical evidence. Section 5 concludes with a summary and implications of the findings.

HYPOTHESES

One body of organizational decline research suggests that product demand declines would inhibit

innovations of an organization. Staw, Sandelands, and Dutton (1981) develop a theoretical model on the effect of external “threat” on organizational behavior. They define threat as “an adverse condition in the environmental, such as resources scarcity, competition, or reduction in the size of the market” (page 515). Their model, called by them “threat-rigidity effects”, suggests that threat could affect organizational behavior in three ways. First, it restricts an organization’s capacity for processing information, and results in a lower number of innovative alternatives to be considered. Second, threat increases an organization’s concern with improving control and coordination of organizational activities, which generally leads to centralization of authority and more formalized procedures. Third, resource scarcity due to the environmental adversity imposes an urgency to conserve resources through cost cutting and results in a “dominance of efficiency concerns”. All these lead to more rigidity in organizational behavior, which consequently inhibits organizational innovations. The theoretical model of “threat-rigidity effects” have been supported by empirical evidence from studies such as Sutton and D’Aunno (1989), D’Aunno and Sutton (1992), Ocasio (1995), and Barker and Mone (1998). Based on this stream of research, we expect that defense product demand decline subsequent to the end of the Cold War would decrease defense contractors’ innovations. We call this “inhibiting view”.

R&D intensity is commonly used as a proxy for R&D investments for innovations in prior research. Thus, the general method we use to test for a change in R&D investments involves a comparison of defense firms’ R&D intensity for the high demand (1984-1989) and the low demand (1993-1998) period.


A decrease in R&D intensity, vis-à-vis nondefense firms in the same industry, reflects the decreased investment in R&D by defense firms. We provide the following hypothesis:

H1a: Industry-adjusted R&D intensity for defense firms during the low-demand period (1993-1998) was lower than that during the high-demand period (1984-1989).

In contrast with the inhibiting view, some studies on organizational decline suggest that external

threat could serve as a stimulator for organizational innovations. This stream of research, based on organizational learning or prospect theories, contends that organizational decline such as lower financial performance, makes organizations to be more risk-seeking for changes or adaptations (Singh, 1986; Bromiley, 1991; Haveman, 1993; Wiseman and Bromiley, 1996), and consequently stimulates managers to search out innovative solutions to problems of their organizations (Miles and Cameron, 1982; Lant, Milliken, and Batra, 1992). The organizational learning theory suggests that organizations are more tended to change and adapt “when their performance is below aspiration level, or perceived as failure” (Lant and Mezias, 1992). Firms are more likely to incur performance below aspiration when facing a substantial decline in product demand. According to the prospect theory, decision makers tend to be more risk- averse “in choices involving sure gains” whereas to be more “risk seeking in choices involving sure losses” (Kahneman and Tversky, 1979). Following this line of research, the severe decline in defense product demand could motivate defense firms to engage in riskier R&D projects as a way to gain competitive advantage in the more challenging defense or commercial product markets. Both the organizational learning and the prospect theories suggests that defense firm would be more motivated to engage in innovation activities for surviving the intensified competition resulted from the decline in defense product demand. Based on this stimulating view, we provide the following hypothesis:

H1b: Industry-adjusted R&D intensity for defense firms during the low-demand period (1993-1998) was higher than that during the high-demand period (1984-1989).

SAMPLE AND DATA

Following McGowan and Vendrzyk (2002), our sample period is from 1984 to 1989 for the period of

high demand for defense product, and from 1993 to1998 for the period of low demand for defense product. We consider 1990 to 1992 as the transition period from high to low demand since most defense contracts last for more than one year, and many defense contractors might begin to have substantial declines in defense sales a few years later than the declines in the awarded volume of contracts (Lundquist, 1992). The low demand sample period ends with 1998 because most firms stopped reporting segment sales to the government subsequent to 1998 due to the change in disclosure requirements on segment reporting by the SEC in January1999 (SEC Final Rule 33-7620).

Our initial sample is 56 firms (or their parent companies) included in the annual DOD report, 100 Companies Receiving the Largest Dollar Volume of Prime Contract Awards of 1989, with nonzero sales to the U. S. government included in Compustat segment files of 1989.1 Please refer to the Appendix for the list of the sample firms. The majority of firms with defense sales also made sales to commercial markets. We examine R&D investments at the firm level, instead of segment level, because the data for R&D expenses are available only at the firm level for most sample firms. Another reason is that operations of government and commercial business are difficult to be clearly separated due to the existence of “externalities” (Bohi 1973) and “cost shifting” (Rogerson 1992; Lichtenberg 1992). We deal with this issue in our regression analysis by including defense dependence, the government sales as a percentage of total sales, as a control variable.

We report sample breakdowns by industry in Table 1, Panel A. Most of the sample defense firms (86% of the 56 firms) concentrate in manufacturing industries with four-digit SIC codes from 2000 to 3999. We report sample breakdowns by year in Table 1, Panel B. Facing with declining product demand,


some defense firms could be delisted due to mergers & acquisitions or even bankruptcies. Thus, the number of the sample firms decreased from 56 in 1989 to 42 in 1998.

TABLE 1 PANEL A: INDUSTRY BREAKDOWN FOR SAMPLE FIRMS OF 1989

First two digits of Industry SIC codes Number Percentage Building construction contractors 15 1 1.8% Chemicals & allied products 28 3 5.4% Primary metals 33 1 1.8% Fabricated metal products 34 2 3.6% Industrial and commercial machinery and computer equipment 35 7 12.5% Electronic and other electrical equipment 36 5 8.9% Transportation equipment 37 23 41.1% Measuring, analyzing and controlling instruments 38 7 12.5% Communication services 48 1 1.8% Wholesale durable goods 50 1 1.8% Business services 73 2 3.6% Services-engineering, accounting, research, and management 87 3 5.4% Total 56 100%

PANEL B: YEAR BREAKDOWN FOR SAMPLE FIRMS

High-Demand Period Low-Demand Period Year 1984 1985 1986 1987 1988 1989 1993 1994 1995 1996 1997 1998 Total

Firms 52 54 54 55 55 56 52 50 48 46 44 42 608

The substantial decline in demand for defense products posted a challenging operating environment for US defense industry. As Lundquist (1992) pointed out, “The cuts will be deeper and longer than any in our history, deeper and longer than anyone in Washington or industry wants to admit. The cuts will force industry to retrench because they will reduce revenue by a greater margin than defense contractors can make up through globalization, diversification, or commercialization.” To have an overview of the effects of declining product demand subsequent to 1989 on operating environment of defense industry, our empirical examination covers the defense firms’ profitability, capital intensity, financial flexibility, and operating efficiency variables, in addition to R&D investment variables. Firms in different industries exhibit different operating and financial characteristics. Thus, we examine these variables both without adjustment and being adjusted by their respective industry medians. For each of those variables, the industry-adjusted measure for each sample year is the difference between a firm and the median of Compustat firms that have the same first two digits of SIC codes with the firm. The definitions for the financial, efficiency and innovations variables used in our empirical examination are included in Table 2.


TABLE 2 VARIABLE DEFINITIONS

Variable Definition____________________________________ Profitability Variables: NI/AT: return on assets, measured as net income divided by total assets at the end of fiscal

year. NI/SALE: return on sale, measured as net income divided by net sales. Financial Flexibility Variables: DT/AT: debt leverage, measured as total liability divided by total assets. CHE/AT: financial resources, measured as cash and short-term investments divided by total

assets. Capital Intensity Variables: PPENT/AT: capital intensity, measured as net book value of property, plant, and

equipment divided by total assets at the end of fiscal year. CAPX/SALE: capital intensity, measured as capital expenditures (CAPX) divided by net sales. Operating Efficiency Variables: SALE/AT: efficiency of assets to produce sales, measured as net sales divided by total

assets at the end of fiscal year. SALE/PPENT: efficiency of plant assets to produce sales, measured as net sales divided by net

property, plant & equipment assets at the end of fiscal year. SALE/EMP: efficiency of employees to produce sales, measured as net sales divided by

number of employees at the end of fiscal year. NI/EMP: efficiency of employees to produce profits, measured as net income divided by

number of employees at the end of fiscal year. Innovations Investments Variables: RD/AT: research and development intensity, measured as research and development

expenses scaled by total assets. RD/SALE: research and development intensity, measured as research and development

expenses scaled by net sales. RD/S&GA: research and development intensity, measured as research and development

expenses scaled by selling and general administrative expenses. Other Variables: GovSal: sales made by a firm to the domestic government SALE: control variable for the size effect, measured by net sales in millions of 1998

constant dollar . Post: a dummy variable. It equals 1 if a firm year is during 1993 to 1998 (i.e.,

post the defense product demand decline), and equals 0 if a firm year is during 1984 to 1989 (i.e., pre the demand decline).

Control: a dummy variable, which equals 1 if a firm is one of the non-defense firms in the control group; otherwise, if a firm is one of the sample defense firms, it equals 0.

Defense: the degree to which a defense firm depends on government contracts for its Dependence: sales, measured as sales to domestic government divided by net sales RESEARCH METHOD AND EMPIRICAL RESULTS Evidence for Defense Firms' Financial and Operational Characteristics

To provide evidence on changes in the operating environment of defense industry, we first compared profitability, total sale, defense sales, and employment of defense firms for high-demand period (1984-1989) and for low-demand period (1993 to 1998). We use non-parametric tests, median tests, for the


comparisons instead of mean tests because the financial and operational variables usually are not normally distributed. The results are reported in Table 3, Panel A. We found that median defense firms earned a significantly lower return on assets for both unadjusted (χ2=5.24, p<0.05) and industry-adjusted returns (χ2=20.74, p<0.01) during the low-demand period, but had no significant change in return on sales over the period. For median defense firms, the defense sales for the low-demand period (1993-1998) are less in amounts (in 1998 constant dollars, χ2=30.58, p<0.01) and consist of a lower percentage of their total sales (χ2=16.53, p<0.01) than for the high-demand period (1984 to 1989). Although the median amount of total sales (in 1998 constant dollars) declined from $4,936 million to $4,352 million, the change is not statistically significant for the median tests (χ2=1.30, p>0.10). The results suggest that the defense firms partially compensate their loss of sales from defense products by making more sales to their commercial markets or foreign markets. The median number of employees for the sample defense firms declined significantly from 37,966 to 28,500 (χ2=2.94, p<0.10). Additionally, the substantial decline in defense product demand seemed to have significantly undermined defense firms' financial flexibility. The median defense firm exhibited significantly higher debt leverage, measured as total liabilities divided by total assets, for both unadjusted (χ2=5.20, p<0.05) and industry-adjusted measures (χ2=4.49, p<0.05). As for financial resources measured by cash and short-term investments scaled by total assets, the industry-adjusted measure exhibited a significant decrease in median (χ2=5.19, p<0.05) although no significant change in unadjusted measure. Our findings indicate that the product demand declines after 1989 had a substantial negative effect on the sample defense firms’ profitability and operations.

With the decline in products demand, one strategy predicted by theories that could be adopted by defense firms is to substantially lower its investment level, called “milking or harvesting the investment” (Perry, 1986). We examined the defense firms’ capital intensity using two measures, net property, plant, and equipment divided by total assets (PPENT/AT) and capital expenditures divided by net sales (CAPX/SALE). The results are also reported in Table 3, Panel A. The median defense firm experienced a significant decline in both the unadjusted and the industry-adjusted capital intensity measures. The results indicate that consistent with the theories, the defense firms indeed cut capital investments as a response to the declining defense product demand.

Prior studies suggest that firm managers have strong incentives to maintain the size of their firms, and could delay downsizing and restructuring to the declined product demands at the cost of operating efficiency (Dial and Murphy, 1995; Sanders, 2001). To investigate the defense firms’ operating efficiency, we compared medians of total asset sale efficiency (net sale/total assets), plant asset sale efficiency (net sales/net property, plant, & equipment), employee sale efficiency (net sales /number of employees), and employee profit efficiency (income before extraordinary items /number of employees) for the high- and low-demand period. The results are reported in Table 3, Panel B. The median defense firm experienced a significant decline in efficiency of assets to produce sales for both the unadjusted (χ2=15.24, p<0.01) and the industry-adjusted measure (χ2=13.99, p<0.01) from the high- to low-demand period. According to the DuPont Model, a firm's return on assets can be represented by its return on sale multiplied by its efficiency of using assets to produce sales (NI/AT =NI/Sale × Sale/AT). Thus, the change of NI/AT can result from changes in NI/Sale and/or changes in Sale/AT. Our evidence suggests that the significant decline in defense firms' median return on assets, as reported in Table 3, Panel A, can be mainly attributable to deterioration in their efficiency of using assets to produce sales since there is no significant change in defense firms' median return on sales. For a firm's efficiency of using plant assets to produce sales, we found no significant change in the median. For efficiency of employee to produce sales, the median defense firms improved from $139.37 thousand to $166.43 thousand for unadjusted measure (χ2=32.72, p<0.01) over the period. For industry-adjusted efficiency of employee to produce sales, however, the defense firms experienced no significant change in the median (χ2=0.81, n. s.). The results are similar for efficiency of employee to produce income. The results indicate that although defense firms improved their employee efficiency to produce sales and income over the period, the improvement was not better than what were achieved by their peer firms operating in the respective industries. Our evidence suggests that the defense firms in general cut their capital investments and workforce to adapt to the substantially lower demand for their defense products. But some of them might have not shrunk


TABLE 3 RESULTS FOR MEDIAN TESTS

Panel A:

Variables High-demand

(median) Low-demand

(median) Median Tests (χ2)

Profitability Return on Sale (NI/SALE) n=326 n=282

Unadjusted 0.0417 0.0433 0.24 Industry-Adjusted 0.0127 0.0109 0.60

Return on Assets (NI/AT) n=326 n=282

Unadjusted 0.0578 0.0477 5.24** Industry-Adjusted 0.0229 0.0057 20.74***

Total Asset, Sale, and Government Sale n=326 n=282 Total Assets (AT) b 4185.1 3916.9 0.03 Total Sale (SALE)b 4936.2 4351.9 1.30 Government Sale (GovSal)b 1416.4 681.01 30.58*** Government Sale/Total sale (GovSal/SALE) 0.3740 0.2090 16.53*** Employment n=322 n=281 Number of employees (EMP) 37966 28500 2.94* Financial flexibility Debt leverage (DT/AT) n=326 N=280

Unadjusted 0.1780 0.2076 5.20** Industry-Adjusted -0.0388 -0.0095 4.49**

Financial resources (CHE/AT) n=326 n=282

Unadjusted 0.0492 0.0474 0.11 Industry-Adjusted -0.0200 -0.0322 5.19**

Capital Intensity PPENT/AT n=326 n=282

Unadjusted 0.2873 0.2425 9.55*** Industry-Adjusted 0.0591 0.0413 5.19**

CAPX/SALE n=322 n=278

Unadjusted 0.0488 0.0390 18.12*** Industry-Adjusted 0.0073 0.0004 8.96***


Panel B:

Variable High-demand

(median) Low-demand

(median) Median Tests (χ2)

Operating Efficiency Asset Efficiency (SALE/AT) n=326 n=282

Unadjusted 1.3407 1.1847 15.24*** Industry-Adjusted 0.0709 -0.0739 13.99***

Plant Asset Efficiency (PPENT/AT) n=326 n=282

Unadjusted 4.4246 4.6394 0.66 Industry-Adjusted -0.5136 -0.6939 1.30

Employee Sale Efficiency (SALE/EMP)a n=322 n=281

Unadjusted 139.37 166.43 32.72*** Industry-Adjusted 9.2938 12.398 0.81

Employee Profit Efficiency (NI/EMP)a n=322 n=281

Unadjusted 5.6051 6.7573 6.41** Industry-Adjusted 1.7122 1.6583 0.01

R&D Intensity R&D/AT n=288 n=255

Unadjusted 0.0421 0.0301 9.60*** Industry-Adjusted 0.0051 -0.0038 10.21***

R&D/SALE n=288 n=255

Unadjusted 0.0323 0.0254 5.81** Industry-Adjusted 0.0067 -0.0036 11.26***

R&D/S&GA n=275 n=235

Unadjusted 0.2040 0.1829 4.93** Industry-Adjusted 0.0582 0.0389 4.17**

*, **, *** represent significance level 10%, 5%, and 1%, respectively. a. Employee sale and profit efficiency variables are in thousands of 1998 constant dollar. b. Amounts of total sale and government sale are in millions of 1998 constant dollar.

sufficiently. Given the evidence of no significant change in median net sales and total assets from the high demand to the low demand period in spite of a significant decline in defense sales (refer to Table 3, Panel A), the results suggest that many manufacturing defense firms responded to defense product demand declines by partially shifting or diversifying their resources to commercial business. Being successful with the commercial markets requires defense firms to utilize new skills such as marketing and sales. And relative to commercial firms, defense firms tend to have high cost structures and low operating efficiencies due to longtime serving the government. Therefore, defense firms, in the process of


commercialization or diversification, could be forced to enter unfamiliar territories, and suffer lower efficiency and profitability (Lundquist, 1992; Anand and Singh, 1997). Evidence for Testing the Hypotheses Results from Median Tests

Following prior studies (e.g., Hitt et al. 1996; Lev and Sougiannis 1996), we use R&D intensity as a proxy for R&D investments for innovations, which is generally measured as reported R&D expenses divided by net sales of a firm.2 Large fluctuations of a firm’s net sales could result in substantial variations in the measure of R&D divided by total sale, which does not necessarily represent significant changes in R&D activities. To mitigate potential bias from this, we also examined two additional variables for a firm’s R&D intensity: reported R&D expenses divided by total assets and R&D expenses divided by selling and administrative expenses. A firm’s total assets are generally more stable than its net sales. Selling and administrative expenses of a firm are usually more stable than its net sales due to the short-term “stickiness” of these expenses. The results of median tests are reported in Table 3, Panel B. From the high-demand to the low-demand period, the median of R&D over net sales (R&D/SALE) significantly decreased from 0.0323 to 0.0254 for the unadjusted measure ((χ2=5.81, p<0.05), and from 0.0067 to -0.0036 for the industry-adjusted measure ((χ2=11.26, p<0.01). The median of R&D expenses scaled by total assets (R&D/AT) also declined over the period, and the decline was significant for both the unadjusted (χ2=9.60, p<0.01) and the industry-adjusted measure ((χ2=10.21, p<0.01). The median of R&D scaled by selling and administrative expenses also declined over the period. The decline is significant for both the unadjusted (χ2=4.93, p<0.05) and the industry-adjusted measure (χ2=4.17, p<0.05). In summary, the above results indicate a significant decline in R&D intensity for the defense firms from the high demand (1984-1989) to the low demand period (1993-1998). These results are consistent with our H1a, "the inhibiting view" that the decline in defense product demand over the period in general undermined the defense firms’ ability to innovate through investments in research and development. Results from Multivariate Regression Analyses

The preliminary evidence from the median tests suggests that decline in defense product demand significantly undermined defense firms’ R&D investments. In the following section, we employ multivariate regression analysis to examine the effects of defense product demand decline on the defense firms. In the regression analysis, we included a control group of Compustat firms that had the same first two digits of SIC industry codes with our sample defense firms but reported no sales to the government in 1989. The control group is used to control for the change in R&D intensity due to confounding factors common to the respective industries in which the sample defense firms operate. The following regression model is employed:

Dependent Variable = β0 + β1 Post + β2 Post*Control + β3 Control + β4 Defense Dependence β5 Industry + β6 Sale + ε (1) where ε is a random error term. Dependent variables are the three variables for R&D intensity (R&D/AT, R&D/Sale and

R&D/S&GA). Post is a dummy variable equal to 1 if the observation is in the low-demand period (1993 to 1998); it equals 0 for the high demand period (1984 to 1989). Control is a dummy variable which equals 0 if the observation is one of our sample defense firms; otherwise, it equals 1. R&D intensity for a firm's defense business could be systematically different from that for its commercial business (Lichtenberg, 1987). Thus, Defense dependence, which is measured as a firm’s sales to the government scaled by its total net sale, is included to control the degree to which a firm depends on defense business for its total sales. Industry variable, measured based on a firm's first two digits of SIC code, is used to control the effect of industry characteristics on R&D intensity. Sale variable (in constant 1998 dollars) is included to control size effects. The definitions of the variables are included in Table 2. The coefficient of


the Post variable indicates the difference in R&D intensity from the high demand (1984-1989) to the low demand period (1993-1998) for the sample defense firms. The coefficient of the Control variable indicates the difference between the sample defense firms and non-defense firms in the control group for the high demand period. The coefficient of the interaction term Post×Control indicates the difference between the sample defense firms and non-defense control firms in the change of R&D intensity from high demand to low demand.

Our sample defense firms tend to be relatively large in sizes. The minimum values of total assets and net sales are 97 million and 260 million, respectively (both in 1998 constant dollars). The non-defense firms with the same first two digits of SIC codes in the control group include many firms of small sizes. To improve the comparability of the sample defense firms and the control group, we exclude the observations with total assets and net sales smaller than 97 million and 260 million (in 1998 constant dollars), respectively. Different from median test, regression analysis is subject to the influence of extreme values. Thus, we trimmed the largest 3% of the three R&D intensity variables to address the undue effect of extreme values in the dependent variables.

TABLE 4 REGRESSION ESTIMATIONS RESULTS FOR TESTING THE HYPOTHESES a

R&D Intensity

Dependent Variables R&D/AT R&D/SALE R&D/S&GA

Independent Variables Coeff. Coeff. Coefficients (t-stat.) (t-stat.) (t-stat.) Constant 0.0465*** 0.0392*** 0.2048*** (12.33) (10.49) (16.92) Post -0.0119*** -0.0073** -0.0197* (-3.56) (-2.18) (-1.78) Post × Control 0.0173*** 0.0135*** 0.0419*** (4.90) (3.84) (3.60) Control -0.0133*** -0.0090*** -0.0422*** (-3.94) (-2.69) (-3.89) Defense Dependence -0.0116** -0.0228*** 0.0522*** (-2.12) (-4.21) (2.85) SALEb 0.0002*** 0.0002*** 0.0019*** (3.07) (4.86) (10.99) Industry Yes Yes Yes Adjusted R2 0.010 0.017 0.049 F-Statistics 9.54*** 15.70*** 42.04*** N 5060 5060 4768 *, **, *** represent significance level 10%, 5%, and 1%, respectively. a. We exclude the observations in the largest 3% of the respectively R&D intensity variables to address the potential undue influence of ext reme values. b. in billions of 1998 constant dollars


Table 4 reports the results for the regression tests. For the R&D intensity variables (R&D/AT, R&D/SALE and R&D/S&GA), the coefficients for the Post variable are significantly negative (t= –3.56, –2.18, and –1.78; significant at 1%, 5%, and 10% level, respectively). The results indicate that the sample defense firms experienced a significant decline in R&D intensity from the high demand (1984-1989) to the low demand period (1993 to 1998). The coefficients for the Control variable are significantly negative for all the three dependent variables at 1% level. The results suggest on average the sample defense firms have higher R&D intensity than the non-defense firms. The coefficients for Post×Control are significantly positive (t=4.90, 3.84, and 3.60, respectively; all significant at 1% level), which indicates that change in R&D intensity is significantly more negative for sample defense firms than that for the non-defense firms in the control group. Further examination (not reported in the table) indicates that for the non-defense firms in the control group, the coefficients of the Post variable (coeff.=0.0054 for R&D/AT, 0.0062 for R&D/SALE, and 0.0222 for R&D/XSGA) are significantly positive (t=4.84, 5.60, 6.10, respectively; all significant at 1% level). The results suggest that the non-defense firms in the control group actually experienced significant increases in R&D intensity over the period of 1984 to 1998, which is consistent with the general trend of increasing knowledge-orientation for U.S. economy. In summary, the evidence above suggests that declined demand in defense products significantly undermined defense firms’ investments in R&D for innovations, which is consistent with our H1a, the “inhibiting view” and inconsistent with the competing "stimulating view". CONCLUSIONS AND IMPLICATIONS

This study investigated the effects of product demand declines subsequent to the end of the Cold War

on defense firms’ profitability, financial flexibility, capital intensity, operating efficiency, and R&D intensity. We found that from the high demand (1984-1989) to the low demand period (1993 to 1998), defense firms in general experienced a substantial decline in profitability as measured by return on assets, which could be attributable to their lower efficiency of using assets to produce sales for the low demand period. The challenging environment faced by the defense firms also put some stress on their financial flexibility. As a response to the declined defense product demand, the defense firms downsized their capital investments and workforce. However, we also found that defense firms, although they experienced significant declines in sales to defense markets, generally managed to maintain their size in terms of total assets and net sales by diversifying into commercial markets or shifting their sources to their existing commercial business. Consistent with Dial and Murphy (1995), our evidence suggests that a significant portion of defense firms might not have downsized and restructured sufficiently, which might be at the sacrifice of profitability and operating efficiency.

We also found that defense firms in general experienced a significant decline in R&D intensity from the high demand (1984-1989) to the low demand period (1993-1998). This trend is in contrast with their respective industry non-defense peers which experienced a significant increase in R&D intensity over the period as the U.S. economy has become increasingly intellectual oriented. Our results suggest that the declines in defense product demand significantly undermined the defense firms’ motivation and capacity to investing in R&D for technological innovations. The inhibiting view in the organization decline literature could better explain defense firms’ responses to product demand decline than the stimulating view. As pointed out by Harbison, Moorman, Jones, and Kim (2000), the substantial lowered demand for defense products lead to a risky consequence for defense industry --- “the industry is eating its ‘seed corn’ in terms of reinvesting in innovations”. Given the high importance of R&D innovations to the national security of the U.S. (Rogerson 1989), this might have produced undesired effects.

Defense industry is highly regulated with only one major buyer which is also the regulator: the Department of Defense. The contingency framework developed by Mone, McKinley, and Barker (1998) suggests that the effect of product demand decline on innovations depends on the institutions in which a company operates. Firms in regulated industries could face legal and political constraints that limit managers’ capacity to seek innovative solutions. Regulations could also motivate firms in the regulated industries, especially defense industry, to compete in alternative ways such as active lobbying. This


potentially undermines defense firms’ incentive to cope with the declining product demand through innovations and improving operating efficiency. ENDNOTES

1. We treat all the sales to the government as defense sales since defense sales account for the majority portion of all the sales to the U.S. government for the firms in the list of index (Lichtenberg, 1992; McGowan and Vendrzyk, 2002).

2. The reported R&D expenses do not include R&D expenditures funded by a firm’s customers including the Department of Defense.

REFERENCES Anand, J., and Singh, H. (1997), “Asset Redeployment, Acquisitions and Corporate Strategy in Declining Industries”, Strategic Management Journal Vol.18 (special issue), pp.99-118. Barker, V. L., III, and Mone, M. A. (1998), “The Mechanistic Structure Shift and Strategic Reorientation at Declining Firms Attempting Turnarounds”, Human Relations Vol. 51 (10), pp. 1227-1258. Bohi, D. R. (1973), “Profit Performance in the Defense Industry”, Journal of Political Economy Vol. 81 (3), pp.721-728. Bromiley, P. 1991. Testing a causal model of corporate risk taking and performance. Academy of Management Jour-nal, 34: 37-59. Cameron, K. S. (1983), “Strategic Responses to Conditions of Decline: Higher Education and the Private Sector”, Journal of Higher Education Vol. 54, pp.359-380. Cameron, K. S., Whetten, D. A., and Kim, M. U. (1987), “Organizational Dysfunctions of Decline”, Academy of Management Journal, Vol. 30, pp.126-138. Campbell-Kelly, Martin (2003), “From Airline Reservations to Sonic the Hedgehog: A History of the Software Industry”, The MIT Press, Cambridge, Massachusetts. Department of Defense, National Defense Budget Estimates for FY 1999 (Green Book). D'Aunno, T. and Sutton, R. I. (1992), “The Responses of Drug Abuse Treatment Organizations to Financial Adversity: A Partial Test of the Threat-rigidity Thesis”, Journal of Management, Vol. 18, pp. 117-131. Dial, J. and Murphy, K.J. (1995), “Incentives, Downsizing, and Value Creation at General Dynamics”, Journal of Financial Economics Vol. 37 (3), pp. 261–314. Gholz, E. and Sapolsky, H. M. (2000), “Restructuring the US Defense Industry”, International Security Vol. 24 (3), pp. 5-51. Harbison, J.R., Moorman, T.S., Jones, M.W., and Kim, J. (2000), “U.S. Defense Industry Under Siege-An Agenda for Change”, Booz, Allen & Hamilton Publishing Inc. Haveman, H. A. (1993), “Organizational Size and Change: Diversification in the Savings and Loan Industry after Deregulation”, Administrative Science Quarterly vol.38, pp. 20-50.


Hitt, M. A., Hoskisson, R. E., Johnson, R. A., and Moesel, D. D. (1996), “The Market for Corporate Control and Firm Innovations”, Academy of Management Journal Vol. 39 (5), pp.1084-1119. Hundley, G., Jacobson, C. K., and Park, S. H. (1996), “Effects of Profitability and Liquidity on R&D Intensity: Japanese and U.S. Companies Compared”, Academy of Management Journal Vol. 39, pp. 1659-1674. Karpoff, J. M., Lee, D. S., and Vendrzyk, V. P. (1999), “Defense Procurement Fraud, Penalties, and Contractor Influence”, The Journal of Political Economy Vol.107 (4), pp.809-842. Lant, T. K., & Mezias, S. J. (1992), “An Organizational Learning Model of Convergence and Reorientation”. Organization Science Vol.3, pp. 47-71. Lant, T. K., Milliken, F. J., and Batra, B. (1992), “The Role of Managerial Learning and Interpretation in Strategic Persistence and Reorientation: An Empirical Exploration”. Strategic Management Journal Vol.13, pp.585-608. Lev, B. and Sougiannis, T. (1996), “The capitalization, amortization, and value-relevance of R&D”, Journal of Accounting & Economics Vol. 21 (1), pp.107-138. Lichtenberg, F. R. (1987), "The effect of government funding on private industrial research and development: a re-assessment", The Journal of Industrial Economics Vol. 36, pp.97-104. Lichtenberg, F. R. (1992), “A Perspective on Accounting for Defense Contracts”, The Accounting Review Vol. 67 (4), pp.741-752. Lundquist, J. T. (1992), “Shrink Fast and Smart in the Defense Industry”, Harvard Business Review, November-December, pp.74-85. McGowan, A. S. and Vendrzyk, V. P. (2002), “The Relation between Cost Shifting and Segment Profitability in the Defense-Contracting Industry”, The Accounting Review Vol. 77 (4), pp.949-969. Miles, R. H., and Cameron, K. S. (1982), “Coffin Nails and Corporate Strategies”, Englewood Cliffs, NJ: Prentice-Hall. Minnich, Richard T. (1993), “U.S. defense industry in transition: Can the leopard change its spots?”, Business Forum, Winter/Spring. Mone, M., McKinley, W., and Barker, V. (1998), “Organizational Decline and Innovations: A Contingency Framework”, Academy of Management Review Vol. 23, pp.115-132. Ocasio, W. (1995), “The Enactment of Economic Adversity: A Reconciliation of Theories of Failure-induced Change and Threat-rigidity”, In L. L. Cummings & B. M. Staw (Eds.), Research in organizational behavior Vol. 17, pp. 287-331. Greenwich, CT: JAI Press. Perry, L. T. (1986), “Least-Cost Alternatives To Layoffs In Declining Industries”, Organizational Dynamics Vol. 14 (4), pp.48-61. Rogerson, W. P. (1989), “Profit Regulation of Defense Contractors and Prizes for Innovations”, Journal of Political Economy Vol. 97 (6), pp.1284-1305.


Sanders, W. G. (2001), “Behavioral Responses of CEOs to Stock Ownership and Stock Option Pay”, The Academy of Management Journal Vol. 44: pp. 477-492. Singh, J. V. (1986), “Performance, slack, and risk taking in organizational decision making”, The Academy of Management Journal, Vol. 29, pp. 562-585. Staw, B. M., Sandelands, L. E., and Dutton, J. E. (1981), “Threat-rigidity Effects in Organizational Behavior: A Multilevel Analysis”, Administrative Science Quarterly, Vol. 26, pp.501-524. Kahneman, D., and Tversky, A. (1979), “Prospect Theory: An Analysis of Decision under Risk”, Econometrica, Vol. 47 (2), pp. 263-292. Wiseman, R. M., and Bromiley, P. 1996. Toward a model of risk in declining organizations: An empirical examination of risk, performance and decline. Organization Science, 7: 524-543. APPENDIX

SAMPLE DEFENSE FIRMS OF 1989 Company Name Primary SIC Code

1 ALLEGHENY TECHNOLOGIES INC 3724 2 ALLIANT TECHSYSTEMS INC 3480 3 ARVIN INDUSTRIES INC 3714 4 AVONDALE INDUSTRIES INC 3730 5 BOEING CO 3721 6 CBS CORP 3585 7 CERIDIAN CORP 3571 8 COMPUTER SCIENCES CORP 7373 9 CONTEL CORP 4813

10 CORDANT TECHNOLOGIES INC 3760 11 CRAY RESEARCH 3571 12 DIGITAL EQUIPMENT 3570 13 DIRECTV GROUP INC 3812 14 DYNCORP INC 8744 15 E-SYSTEMS INC 3812 16 EATON CORP 3714 17 FMC CORP 2800 18 GENCORP INC 3760 19 GENERAL DYNAMICS CORP 3721 20 GENERAL ELECTRIC CO 3600 21 GRUMMAN CORP 3721 22 HARRIS CORP 3663 23 HARSCO CORP 3440 24 HENLEY GROUP INC/DEL 3821


25 HERCULES INC 2821 26 HONEYWELL INC 3822 27 HONEYWELL INTERNATIONAL INC 3724 28 INTL BUSINESS MACHINES CORP 3570 29 KAMAN CORP -CL A 5080 30 LITTON INDUSTRIES INC 3812 31 LOCKHEED MARTIN CORP 3760 32 LOGICON INC 7371 33 LORAL CORP 3812 34 LTV AEROSPACE & DEFENSE CO 3728 35 LTV CORP 3312 36 MARTIN MARIETTA CORP 3760 37 MCDONNELL DOUGLAS CORP 3721 38 MORRISON KNUDSEN CORP OLD 1540 39 MOTOROLA INC 3663 40 NORTH AMERICAN PHILIPS CORP 3640 41 NORTHROP GRUMMAN CORP 3721 42 OLIN CORP 2800 43 OSHKOSH TRUCK CORP 3711 44 PERKINELMER INC 8711 45 RAYTHEON CO 3812 46 ROCKWELL AUTOMATION 3760 47 SCIENCE APPLCTNS INTL 8700 48 SEQUA CORP -CL A 3724 49 SUNDSTRAND CORP 3728 50 TENNECO AUTOMOTIVE INC 3523 51 TEXAS INSTRUMENTS INC 3674 52 TEXTRON INC 3720 53 TRACOR INC 3728 54 TRW INC 3760 55 UNISYS CORP 3570 56 UNITED TECHNOLOGIES CORP 3724


The Symmetry of Demand and Supply Shocks in Monetary Unions

Maru Etta-Nkwelle Howard University

Carlton Augustine

The American University

Youngho Lee Howard University

This study has been motivated by the numerous proposals for greater monetary and economic integration in Africa. We investigate the correlation of shocks between the exiting members of the West African Economic and Monetary Union and potential entrants in the region. In Southern Africa, we examine the pair wise correlations using the hub and spoke framework with South Africa as the hub. We observe large demand shock asymmetry amongst the countries in the west than in the south, suggesting more economic homogeneity amongst the members of the South African Development Community. However, there seem to be more supply shock symmetry amongst the countries in West Africa than their counterparts in the south. INTRODUCTION

Recently, several initiatives towards greater financial and monetary integration have been proposed on the African continent. For example, the South African Development Community (SADC) plans to establish a common currency by 2018. In the East of Sub-Saharan Africa (SSA), the East African Community (EAC) is considering a monetary union, while in the west, the West African Economic and Monetary Union (WAEMU) plans to expand its membership to include all the members of the West African Monetary Zone (WAMZ). On a larger scale, the African Union has proposed to implement a single monetary zone (and possible single currency) for all of SSA and other African countries by 2028.

A central theme in most of the empirical literature on financial integration is whether the cost of integration outweighs the benefits. The origins of this debate can be traced back to Mundell’s (1961) seminal paper on optimum currency areas. Mundell writes that the benefits of financial integration are realized more amongst countries with similar terms of trade shocks than amongst those with asymmetric shocks. The presence of asymmetric shocks, he argues, makes financial integration harder and more costly to manage because the respective countries may need different monetary policies to respond effectively to their individual shocks. In addition, Houssa and Leuven, (2004) argue that in the SSA, one of the concerns with the proposed expansion of the WAEMU to include WAMZ countries hinges on the fact that the high inflationary countries (WAMZ) may lose competitiveness in a monetary union with the lower inflationary countries (WAEMU). This implies that a similarity in terms of trade shocks is a


necessary pre-condition for monetary integration because it prevents the need for individualized adjustment tools among the countries and allows for the implementation of a common monetary policy.

Apart from the lack of symmetry in terms of trade shocks, others have examined in general the welfare benefits and costs of belonging to a monetary union. One such study by Fieldings and Shields (2001) on the CFA zone suggests that the welfare benefits of low inflation, lower exchange rate variability, lower transaction costs and greater macroeconomic integration enjoyed by existing monetary unions certainly surpasses the most recognized cost which is the loss of exchange rate flexibility as an adjustment tool. Other proponents of monetary unions have argued that the loss of exchange rate flexibility should not impede adjustment to macroeconomic shocks, since there exist sufficient mitigation instruments to achieve the same objective (Devarajan and De Melo, 1987). There is also ample evidence suggesting that some of the welfare losses experienced by the exiting monetary unions in SSA stem from the fact that monetary unions are formed in the region without regard to the economic profiles of the respective countries. Fielding and Shields (2001) argue that the nations make commitments to join single currency and monetary unions with countries with whom they have different economic characteristics. Similarly, Etta-Nkwelle and Jeong (2009) add that unlike the European Monetary Union, where members were required to satisfy some pre-requisite conditions prior to admission, in Africa formation was exogenous and without any convergence criteria. As a result, these countries frequently subject themselves to regional macroeconomic policy directives which may not necessarily be optimal to their specific economy. In an effort to address some of these concerns, the WAEMU, WAMZ and SADC established, in 1994, 2001 and 2002 respectively, a set of convergence guidelines to reduce the divergence in macroeconomic characteristics between the members. The guidelines emphasized both monetary and fiscal convergence amongst the members, but, as reported by Ghosh, Gulde and Wolf (2008), progress has been slow and inconsistent, and the WAMZ has pushed its targeted date for meeting their criteria three times to 2015.

Furthermore, over the last decade, as integration in Europe has progressed, several authors have examined the macroeconomic characteristics and symmetry of shocks amongst African nations to determine the feasibility of monetary integration in the region (Bayoumi and Ostry, 1998; Guillaume and Stasavage, 2000; Fielding and Shields, 2003; Houssa and Leuven, 2004; and Houssa, 2008). The study by Bayoumi and Ostry (1998) on Sub-Saharan African countries using an AR (2) model of output disturbances found no correlations amongst the members. In contrast, the studies by Fielding and Shields (2003) and Houssa and Leuven (2004) on the CFA zone members found negative output shock correlations amongst the members. The authors conclude that the asymmetry of supply shocks amongst the members in the region reflects the lack of uniformity of macroeconomic policies and the diversity in commodity specialization which exist in the region. The conclusion drawn by these studies is that the cost of monetary unions amongst the members in the region may be high, implying that these countries are poor candidates for monetary union at this time (Bayoumi and Eichengreen, 1994).

Although a body of evidence exists indicating that the cost of monetary union in Africa may outweigh the benefits, Guillaume and Stasavage (2000) suggest that monetary integration amongst the nations in SSA is a necessary mechanism through which the commitment to price stability and financial discipline can be achieved. Therefore, in this paper, we are mindful of the commitments that these nations have made toward macroeconomic convergence, and we seek to investigate if these efforts have reduced the degree of asymmetric shocks between the memberships. In other words, this paper updates existing evidence on the feasibility of monetary integration in SSA by examining the correlation of demand and supply shocks between members of existing monetary unions and non-monetary union countries in Sub-Saharan Africa. Specifically, in the south of Africa, we expand the Common Monetary Area (CMA) and analyze the formation of a monetary union for the 14 members of the SADC. Likewise, in the west, the proposed expansion of the WAEMU to include five (WAMZ) accession members of ECOWAS is also analyzed. We do this to assess the degree of symmetry of shocks and to determine if the proposed expansion of the existing monetary unions to include other SSA countries and possible formation of a single currency zone for the region is feasible.


Like previous studies we employ a structural vector auto-regression (SVAR) model of output growth and inflation to recover supply and demand shocks. Unlike previous studies we use the hub and spoke framework in analyzing the SADC with South Africa (the largest economy in the SADC) as the hub on which the correlation of the shocks of the other countries are measured. We feel that this approach is justified by the virtue of South Africa’s economic strength. But in the west of Africa, we compared the correlations of demand and supply shocks of the proposed accession countries (WAMZ) to that of the existing WAEMU nations. We see this update as necessary because of the dynamic nature of macro-economic performance on the continent. Therefore, the results of this study may provide additional evidence on the readiness of the SSA countries to form a monetary union and or whether there is a need for mandatory fulfillment of pre-conditions before accession into the proposed monetary zone.

The rest of the paper is organized as follows: Section 2 presents the analytical framework; Section 3 discusses the data; Section 4 presents the results; section 5 concludes with a summary of the main findings. ANALYTICAL FRAMEWORK

Several of the existing studies that have examined the correlation of demand and supply shocks amongst groups of countries as a precondition for monetary integration have employed the structural vector autoregression procedure developed by Blanchard and Quah’s (1989). This method uses a two variable VAR model of prices and output to identify demand and supply shocks with the restriction that only supply shocks have a permanent effect on output since demand shocks are transitory. Albeit, the fact that this procedure has been widely used in literature, it has its critics, among them: Bayoumi and Eichengreen, (1992); Kawai and Okumura, (1996); Faust and Leeper, (1997); Cooley and Dwyer, (1998); Demertzis, Hallett and Rummel (2000); and Gottschalk, (2001) and Houssa (2008) to name a few. Demertzis, Hallett and Rummel (2000) are concerned that this model does not necessarily identify purely stochastic shocks because estimated demand shocks tend to include the effect of macroeconomic policies, but estimated supply shocks are less likely to include the impact of the implemented policies ( Zhang, McAleer and Sata, 2004). Houssa (2008) on the other hand is concerned about the large number of restrictions needed to recover the shocks. He adds that the number of parameters estimated with VAR models have a tendency to grow with the square of the number of variables, which may lower the degrees of freedom of the estimation.

Despite its limitations, the structural vector autoregression procedure is still widely used in the literature with slight modifications and/or extensions. In this paper, we follow Fidrmuc and Korhonen’s (2003) version of the Blanchard and Quah (1989) procedure which uses an infinite framework to decompose output into its temporary and permanent components. The infinite framework overcomes the lack of uniqueness that is often used to criticize the univariate procedure of other research (such as that by Beveridge and Nelson, 1991). Like in Bayoumi and Eichengreen (1994a), we make the following assumptions: there is another variable that is affected by the same set of shocks as income; demand shocks increase the price level, while supply shocks decrease it, and output and prices are stationary. Therefore, the two variables VAR with output and inflation can be specified as an infinite moving average representation of demand and supply shocks:

0 0 1 0 2 0 30

.... it t t t t i t

iY A A A A L Aξ ξ ξ ξ ξ

∞

− − −

=

= + + + + =∑ (1.1)

In this model, Yt is a vector of differences of logs of output and prices, ξ is a vector of demand and supply shock, Ai are the 2x2 matrices which transfer the effect of disturbances to the variables and Li is the lag operator. As mentioned above, it is also assumed that the shocks are uncorrelated, their variance is unity and demand shocks have no long run impact on output. A finite VAR specification of model (1.1) is estimated and used to identify demand and supply disturbances. Since the Yt is stationary, the VAR specification can be inverted to obtain the moving average specification, where Ω is the vector of residuals from the two estimated equations:


(1.2)

According to Fidrmuc and Korhonen (2003), the variance-covariance matrix of the residuals is VAR (Ω) = α. From the two equations above (1.2), the relationship between the estimated residuals (Ω) and the original shocks (ξ): Ωt = A0ξ is derived. The matrices Bi are known from estimation but that of A0 need to be known in order to calculate the underlying supply and demand shocks. Knowing that Ai = Bi A0 and

that = helps with the identification of A0. To recover the four elements of Ao, we need

four restrictions. Two of the restrictions are normalizations defining the variance of the shocks ξdt. and ξst. The third restriction is the assumption that demand and supply shocks are orthogonal, which, with our notation, means that A0A0

* = α. The fourth restriction, as mentioned above, is that demand and supply shocks equal zero. These restrictions identify the elements of A0 and allow for the extraction of demand and supply shocks from the residuals of the estimated VAR. DATA

The sample used in this study consists of twenty-one Sub-Saharan African countries. Annual observations of real GDP and the consumer price index (CPI) for the period 1980 to 2007 were extracted from the World Bank’s World Development Indicator (2009), International Monetary Fund’s Regional Economic Outlook: Sub-Saharan Africa (2009), and the International Monetary Fund’s International Financial Statistics (2008). The annual change in the CPI and the annual change in the real GDP have been used as proxies for inflation and growth rate respectively. All data have been converted to natural logarithms. EMPIRICAL RESULTS

The feasibility of a monetary union is largely determined by the symmetry of shocks among the potential members. As a measure of symmetry, we use the correlation of demand and supply shocks. In particular, for the WAEMU, our focus is on the correlation of shocks between potential entrants (WAMZ) and existing members of the monetary union. For the SADC, we examine the pair wise correlation coefficients using the hub and spoke framework and measure the correlation of members, vis a vis South Africa. The results of the demand shock analysis of the WAEMU (Table 1) shows that the only significantly positive correlation is between Guinea Bissau and Gambia. These two coastal countries share some homogeneity in terms of export commodities, implying similarity in terms of trade shocks. That is, the countries are both exporters of agricultural products such as cashews and peanuts and have a relatively large fishing industry. In fact, in 2007, shelled cashew was the main export crop from both countries. Most of the other correlations are negative and insignificant. In a few cases, such as Benin and Ghana; and Guinea Bissau and Nigeria the correlations are negative and significant. This negative correlation has been found by other studies (Houssa, 2008). The authors suggest that this negative correlation implies a loss of competitiveness for the WAMZ if they join a monetary union with the low inflation countries in the WAEMU. For the countries that belong to the SADC, the only positive and significant demand correlations are between DR of Congo and Botswana, and Malawi and Botswana (Table 2). The existence of demand shock symmetry between DR of Congo and Botswana is not surprising, as both countries have diamond and other precious metals as their primary export commodity and thus may have similar responses to diamond price volatility. On the other hand, the symmetry of demand shocks between Malawi and Botswana is interesting given that the former exports mostly agricultural products such as tobacco, sugar and tea. Mauritius and Malawi, and Swaziland and DR of Congo, do show a significant but negative correlation implying diverse economic characteristics. Interestingly, we observe that more countries in the south exhibit positive but insignificant correlation

1 1 2 2 3 30

.... ti

t t t t t ii

Y L∞

Ω Ω − Ω − Ω − Ω

=

= +Β +Β +Β + = Β∑

0i

iA

∞

=∑ 0

0i

iA

∞

=

Β∑


TABLE 1 DEMAND SHOCK CORRELATION OF WAEMU AND WAMZ MEMBERS

Gambia Ghana Liberia Nigeria Sierra Leone Benin -0.03 -0.29*** -0.11 0.17 0.04 (-0.27) (-2.46) (-0.63) (1.03) (0.21) Burkina Faso 0.07 0.17 0.09 0.37 -0.15 (0.44) (0.10) (0.53) (0.22) (-0.87) Cote d’Ivoire -0.16 -0.17 0.00 -0.01 -0.10 (-0.97) (-0.99) (0.00) (-0.04) (-0.60) Guinea Bissau 0.31* -0.19 -0.16 -0.41*** -0.10 (1.98) (-1.15) (-0.93) (-2.61) (-0.59) Mali -0.20 0.23 0.06 0.26 0.00 (-1.26) (1.37) (0.34) (1.59) (0.00) Niger -0.05 -0.18 -0.08 -0.16 -0.26 (-0.35) (-1.06) (0.49) (-0.95) (-1.59) Togo -0.24 -0.18 0.12 -0.19 0.16 (-1.44) (-1.06) (0.73) (-1.09) (0.97) Senegal -0.10 -0.04 0.00 0.14 -0.15 (-0.58) (-0.22) (0.00) (0.83) (-0.87) __________________________________________________________________________________ Notes: The values of the "t" statistics are in parentheses. (*, **, *** denotes significance at the 10 percent level, 5 percent level and at the 1 percent level respectively).

TABLE 2

SADC: CORRELATION OF DEMAND SHOCKS

Correlation t-Statistics Probability Botswana South Africa 0.164246 0.985073 0.3313 DR Congo DR Congo

South Africa Botswana

-0.103324

0.286523

-0.614564

1.769274*

0.5428

0.0856 Malawi Malawi

South Africa Botswana

0.238744

0.292304

1.454488

1.808270*

0.1547

0.0792 Mauritius Mauritius Mauritius Mauritius

South Africa Botswana DR Congo Malawi

0.020845 0.041181 0.171018 -0.371911

0.123345 0.243838 1.026884 -2.370282

0.9025 0.8088 0.3115 0.0234

Swaziland Swaziland Swaziland Swaziland

South Africa Botswana DR Congo Mauritius

0.174335 -0.155652 -0.260836 0.006142

1.047421 -0.932211 -1.598457 0.036335

0.3021 0.3576 0.1189 0.9712

Zambia Zambia Zambia Zambia Zambia Zambia

South Africa Botswana DR Congo Malawi Mauritius Swaziland

0.011937 0.033199 0.086212 0.022169 -0.202858 -0.041750

0.070625 0.196517 0.511944 0.131185 -1.225608 -0.247211

0.9441 0.8453 0.6119 0.8964 0.2285 0.8062

The only positive and significant correlations are Malawi and Botswana and DR Congo and Botswana


coefficients while in the west the coefficients are mostly negative and insignificant. Although demand shocks can be influenced by monetary and fiscal policy, we interpret the demand shock results on (Table 2) as suggesting that the SADC countries are relatively more economically homogenous and thus better suited to begin talks of monetary integration than their counterparts in the west. In an effort to support our interpretation, we examined some macroeconomic indicators in the respective regions and find little support for our suggestion. We find more divergence from the regional average (in terms of GDP growth, terms of trade shocks, inflation, current account balance and external debt) among the SADC countries than amongst the WAEMU and WAMZ nations. That is, we observe more uniformity in the economic characteristics of the WAEMU and WAMZ countries than amongst the countries in the SADC. The uniformity of the latter countries can be attributed to the fact that these countries have belonged to a customs union for many decades and thus have made some efforts towards the harmonization of trade policies albeit with little success.

On the supply side, we find more supply shock symmetry (twenty six positive relationships of which 57 percent are significant) amongst the WAEMU and WAMZ members. The results (Table 3) suggest positive and significant supply shock correlations exist for the following countries: Benin and Ghana; Benin and Nigeria; Burkina Faso and Liberia; Cote d’Ivoire and Gambia; Mali and Liberia; Niger and Ghana; Senegal and Gambia; and Togo and Sierra Leone. Benin and Gambia and Guinea Bissau and Liberia are the pairs of countries with negative and significant supply shocks.

TABLE 3 SUPPLY SHOCK CORRELATION OF WAEMU AND WAMZ MEMBERS

Gambia Ghana Liberia Nigeria Sierra Leone Benin -0.44*** 0.31* 0.18 0.28 -0.11 (-2.82) (1.93) (1.10) (1.64) (-0.66) Burkina Faso -0.03 -0.08 0.26 0.24 -0.00 (-0.18) (-0.47) (1.61) (1.43) (-0.04) Cote d’Ivoire 0.30* 0.03 -0.00 0.04 -0.10 (1.91) (0.20) (-0.02) (0.25) (-0.57) Guinea Bissau 0.14 0.13 -0.28*** -0.03 0.19 (0.84) (0.74) (-2.40) (-0.20) (1.14) Mali 0.24 -0.20 0.28* 0.08 0.02 (1.46) (-1.19) (1.76) (0.47) (0.15) Niger 0.02 0.32* 0.09 0.09 0.05 (0.11) (1.97) (0.53) (0.51) (0.27) Togo -0.18 0.23 -0.06 0.18 0.50*** (-1.1) (1.34) (-0.38) (1.06) (3.33) Senegal 0.37*** 0.03 0.05 -0.19 -0.01 (2.33) (0.19) (0.28) (1.11) (-0.04) ____________________________________________________________________________________ Notes: The values of the "t" statistics are in parentheses. (*, **, *** denotes significance at the 10 percent level, 5 percent level and at the 1 percent level respectively)

These results are contrary to the asymmetric supply shock observed earlier by Houssa (2008) between WAEMU and WAMZ members. Our results suggest that based on recent data the WAMZ has made some progress in terms of macro-economic policy harmonization with at least one WAEMU country and thus a monetary union may now be feasible. For the SADC (Table 4) we observe a significant positive pair wise correlation among: Zambia and South Africa; Mauritius and Zambia; DR Congo and South Africa and Madagascar and Mauritius. In addition, there is a significant negative correlation between the supply shocks of Botswana and Mauritius.


TABLE 4 SADC: CORRELATION OF SUPPLY SHOCKS

Correlation t-Statistics Probability

Swaziland South Africa 0.151324 0.905675 0.3713 Zambia Zambia

South Africa Swaziland

0.349018* -0.161059

2.203374 -0.965443

0.0342 0.3409

Mauritius Mauritius Mauritius

South Africa Swaziland Zambia

0.251851 -0.221924 0.440883*

1.539595 -1.346497 2.905971

0.1327 0.1868 0.0063

Malawi Malawi Malawi Malawi

South Africa Swaziland Zambia Mauritius

0.031565 -0.011576 0.112287 0.180027

0.186832 -0.068490 0.668525 1.082744

0.8529 0.9458 0.5082 0.2863

Madagascar Madagascar Madagascar Madagascar Madagascar

South Africa Swaziland Zambia Mauritius Malawi

0.069329 0.163600 0.065922 0.328714* 0.247338

0.411147 0.981090 0.390853 2.059123 1.510195

0.6835 0.3333 0.6983 0.0470 0.1400

DR Congo DR Congo DR Congo DR Congo DR Congo DR Congo

South Africa Swaziland Zambia Mauritius Malawi Madagascar

0.332751 0.073743 -0.209900 0.127809 -0.254902 0.047168

2.087541 0.437459 -1.270080 0.762380 -1.559535 0.279361

0.0442 0.6645 0.2124 0.4509 0.1279 0.7816

Botswana Botswana Botswana Botswana Botswana Botswana Botswana

South Africa Swaziland Zambia Mauritius Malawi Madagascar DR Congo

-0.137319 0.022484 -0.025130 -0.315859 0.113105 -0.112139 0.026130

-0.820158 0.133054 -0.148717 -1.969471 0.673461 -0.667636 0.154641

0.4177 0.8949 0.8826 0.0569 0.5051 0.5087 0.8780

*Positive and significant correlations are: Zambia and South Africa; Mauritius and Zambia; and Madagascar and Mauritius and DR Congo and South Africa.

Another aspect of symmetry relates to the size and the speed of adjustment to the shocks (Table 5).

The impulse response functions are generated from the structural VARs and these facilitated the recovery of the supply and demand shocks, under the assumption that demand disturbances have no long-run impact on output, while the supply disturbances have permanent effects on output. This assumption provides the basis for measuring the sizes of the disturbances. Following Bayoumi and Eichengreen (1994), we use the long-run output effect of supply shock, and also measure the demand shock as the accumulated long run effect of demand shocks on the price level. Meanwhile for both shocks, the speed of adjustment is measured as the cumulative response after 2 years as a share of the long run effect. We find that the average size of demand shock to the SADC is almost twice (.26) that of the WAEMU and WAMZ combined (.15), while the supply shocks to both regional groups is relatively low and uniform (.05 and .05 respectively). Similarly, an examination of the individual nations reveals that the demand shocks to the southern nations are higher than that to the western nations. In the south, we find shocks of .56 (Zambia) and 1.03 (DR of Congo), while in the west, the highest shock is .48 (Guinea Bissau). The high shock to DR of Congo is not surprising given the internal conflicts which plagued the nation in the


1980s and 1990s. Interestingly, although the SADC are more prone to demand shocks than their counterparts in the west, they also seem to respond to these shocks relatively faster. In contrast, the speed of adjustment to supply shocks of the nations in the west is relatively faster than that of the southern nations. Albeit, slight difference, we conclude that both regions seem to adjust fairly quickly to both demand and supply shocks.

TABLE 5 THE SIZE AND SPEED OF ADJUSTMENT OF SADC, WAEMU AND WAMZ MEMBERS

Demand Shocks Supply Shocks Size Speed of Adjustment Size Speed of Adjustment SADC 0.26 0.78 0.05 0.78 Botswana 0.02 0.83 0.09 0.51 DR Congo 1.03 0.86 0.14 0.50 Madagascar 0.10 0.85 0.04 0.99 Malawi 0.18 0.84 0.04 0.98 Mauritius 0.07 0.92 0.03 0.90 South Africa 0.07 0.67 0.04 0.79 Swaziland 0.04 0.93 0.005 0.91 Zambia 0.56 0.36 0.05 0.69 WAEMU 0.15 0.80 0.05 0.88 AND WAMZ Benin 0.08 0.75 0.03 0.99 Burkina Faso 0.08 0.83 0.02 0.94 Cote d’Ivoire 0.11 0.85 0.08 0.56 Gambia 0.11 0.83 0.03 0.92 Ghana 0.16 0.96 0.06 0.92 Guinea Bissau 0.48 0.48 0.08 0.99 Mali 0.09 0.92 0.05 0.98 Nigeria 0.12 0.99 0.05 0.99 Senegal 0.02 0.95 0.03 0.95 Sierra Leone 0.41 0.44 0.13 0.54 Togo 0.01 0.84 0.05 0.99 ____________________________________________________________________________________ CONCLUSION

Although our results suggest that demand and supply shocks in Sub-Saharan Africa are still largely asymmetric, some symmetry is beginning to emerge especially amongst the countries in the west. We find that large demand shock asymmetry still exists amongst the WAEMU and WAMZ countries. Only Guinea Bissau and Gambia are positively correlated. However, in the south, although the SADC countries also have low demand shock symmetry, more countries seem to exhibit a positive (but insignificant) correlation coefficient, suggesting the movement towards homogeneity. Positive and significant demand shock symmetry exists between DR of Congo and Botswana and Malawi and Botswana. Therefore, in terms of demand shock, this study suggests that the southern African countries are better ready to form a monetary union than their counterparts in the west.

Interestingly, on the supply side, we observe more supply shock symmetry amongst the WAEMU and WAMZ countries than their counterparts in the south (SADC). This finding is interesting and different


from previous studies such as that by Fielding and Shield (2001) which found more demand shock correlations among the countries in the west. This observed progress towards more symmetric supply shocks amongst the countries in the west is not surprising given that the WAEMU and WAMZ institutional arrangements have existed longer than that of the south and efforts have been made over the last decade towards macroeconomic convergence. What is surprising is that progress has been slow in all the groupings towards macroeconomic convergence as the region continues to deal with low interregional trade, civil unrests, famine and differences in external shocks. REFERENCES Beveridge, S. & Nelson, C. R. (1981). A New Approach to Decomposition of Economic Time Series into Permanent and Transitory Components with Particular Attention to Measurement of the Business Cycle. Journal of Monetary Economics, 7, 151-174. Baxter, M, & Kouparitsas, M. A (2005). Determinants of Business Cycle Co-movement: a Robust Analysis. Journal of Monetary Economics, 52, (1), 113-157. Bayoumi, T. & Eichengreen, B. (1993). Adjustment and Growth in the European Monetary Union, Cambridge University Press, Cambridge, 193-229. Bayoumi, T. & Eichengreen, B. (1994a). Macroeconomic Adjustment under Bretton Woods and the Post-Bretton Woods: An Impulse Response Analysis. The Economic Journal 104, 813-827. Bayoumi, T. & Eichengreen, B. (1994b). One Money or Many? Analyzing the Prospects for Monetary Unification in Various Parts of the World. Princeton Studies in International Finance, 76. Bayoumi, T. & Ostry, J. (1997). Macroeconomic Shocks and Trade Flows within Sub-Saharan Africa: Implications for Optimum Currency Arrangements. Journal of African Economies 6(3), 412-444. Blanchard, O. & Quah, D. (1989). The Dynamic Effects of Aggregate Demand and Aggregate Supply Disturbances. American Economic Review 79, 655-673. Cooley, T. & Dwyer, M. D. (1998). Business Cycle Analysis without Much Theory: A Look at Structural VARs. Journal of Econometrics 83, 57-88. Darvas, Z. Rose, R. & Szapary, G. (2005). Fiscal Divergence and Business Cycle Synchronization: Irresponsibility is Idiosyncratic. NBER International Seminar on Macroeconomics 2005. Demertzis, M. Hallett, A. H. & Rummel, O. (2000). Is the European Union a National Currency Area, or Is It Held Together by Policy Makers? Weltwirtschaftliches Archiv, 136 (4), 657-679. Devarajan, S. & De Melo, J. (1987b). Adjustment with a Fixed Exchange Rate: Cameroon, Cote d’Ivoire and Senegal. The World Bank Economic Review, 1 (3), 447-487. Etta-Nkwelle, M. & Jeong, G. (2009). Post devaluation real exchange rates and the pact of convergence in the West African Economic and Monetary Union. Washington Business Research Journal, 1 (1), December 2009, 61-71. Faia, E. (2007). Financial Differences and Business Cycle Co-Movements in a Currency Area. Journal of Money, Credit and Banking, 39(1), 151-185.


Faust, J. and Leeper, E.M. (1997). When do long-run identifying assumptions give reliable results? Journal of Business and Economic Statistics, 15, 345-353. Fidrmuc, J. & Korhonen, I. (2003). Similarity of Supply and Demand Shocks between the Euro Area and the CEECs. Economic Systems, 27, 313-334. Fieldings, D., Lee, K. & Shields, K. (2001). Modeling macroeconomic shocks in the CFA franc zone. Journal of Development Economics, 66, 199-223. Fieldings, D., Lee, K. & Shields, K. (2003). The Characteristics of macroeconomic shocks in the CFA franc zone. Journal of African Economies. 13, 488-517. Fielding, D. & Shields, K. (2005). The Impact of Monetary Union on Macroeconomic Integration: Evidence from West Africa. Economica 72, 682-704. Frankel, J. & Rose, R. (1998). The endogeneity of the optimum currency area criterion. Economic Journal, 108, 1009–25. Ghosh, A., Guide, A. & Wolf, H. (2008). Monetary Union in Central and Western Africa. Journal of Financial Transformation, CAPCO Institute, 20. Guillaume, D. M., & Stasavage, D. (2000). Improving Policy Credibility: Is There a Case for African Monetary Unions? World Development 28, 8, 1391-1407. Houssa, R. (2008). Monetary union in West Africa and Asymmetric shocks: A Dynamic Structural Factor Model Approach. Journal of Development Economics, 85 (1-2), 319-347. Kawai, M. & Okumura, T. (1996). Financial and Capital Markets in Asia, Tokyo: Nihon Keizai Shinbunsha, 217-237. Mundell, R. (1961). Theory of optimum currency areas. American Economic Review, 51, 509-17. Zhang, Z., Sato, K. & McAleer, M. (2002). Asian Monetary Integration: A Structural VAR Approach. Mathematics and Computers in Simulation (forthcoming). Zhang, Z., Sato, K. & McAleer, M. (2004). Is a Monetary Union Feasible for East Asia? Applied Economics, 36, 1031-1043.


Determinants of Steady State Income: Is a Linear Specification Too Simple?

Kathleen E. Odell

Dominican University

This paper examines whether the Solow model of economic growth and steady state income is supported by available cross-section data. The paper argues that while income per worker is correlated positively with investment in capital (both physical and human) and negatively with population growth, a simple log-linear empirical model derived from the Cobb-Douglas production function oversimplifies these relationships. A nonlinear specification is presented as an alternative. Estimation of the nonlinear specification shows that for the period 1960-2000, the strongest marginal effects of education occur in countries where education levels are already relatively high. INTRODUCTION

The economic growth literature contains numerous studies offering empirical support for the neoclassical growth model, but estimation in these studies is almost always conducted using a linear model and ordinary least squares (OLS) regression. However, the results of OLS estimation provide average coefficients, and these averages may or may not accurately represent the relationships in the data. There are two key implications of the assumption of linear coefficients. First, this assumption implies that all countries included in the cross-section follow identical linear growth processes. Second, the assumption implies that the coefficients on all explanatory variables are constant across all levels of those variables. Where income and growth are concerned, these are strong and testable implications. This paper argues that the simple log-linear model of steady state income is too simple. In particular, there is evidence of increasing returns to education in the cross-section data.

Empirical support for the neo-classical Solow (1956) model of economic growth was presented by Mankiw, et. al. (1992) (henceforth MRW). MRW used a log-linear estimation equation derived from the Cobb-Douglas production function. Their finding was that nearly 60% ( 2 0.59R = ) of the cross-country variation in income per worker could be explained by a “Textbook Solow Model” containing only the savings rate (as a proxy for investment in physical capital) and the population growth rate. Further, MRW developed an “Augmented Solow Model” which included both physical and human capital accumulation, and this augmented model explained nearly 80% ( 2 0.78R = ) of the variation in income per worker. Bernanke and Gurkaynak (2001) and Karras (2008) have determined that the Textbook Solow Model performs as well or better when additional data, available since the publication of the MRW paper, is included. Bernanke and Gürkaynak also re-estimate the Augmented Solow Model with updated data and find results similar to MRW.

Given the theoretical consistency, high explanatory power, and robustness of the log-linear model (estimated with OLS), it is tempting to conclude that this specification is an accurate description of the data. However, the importance of the linearity assumption is increasingly under scrutiny. As a result,


there now exists a substantial body of empirical growth research which supports the argument that a standard linear model (estimated with OLS) is inconsistent with the historical cross-sectional growth data. Both theoretical (Azariadis and Drazen (1990), Galor and Weil (2000)) and empirical (Durlauf and Johnson (1995), Hansen (2000), Liu and Stengos (1999)) work has suggested that a linear model is inconsistent with reasonable explanations of the economic growth process. However, few non-linear studies have been conducted using updated versions of the cross-sectional data. Further, there is little consensus in the literature about the most effective method for testing for nonlinearity.

This study demonstrates that for the period 1960-2000, the well-documented support for the Textbook Solow Model of growth is robust to the relaxation of the linearity assumption. However, the Augmented Solow Model, which adds the effect of human as well as physical capital to the model, is better described by a model that is nonlinear in the coefficient of human capital. This paper outlines a simple testing methodology for nonlinearity in the coefficients. The Generalized Additive Model (GAM) is a straightforward approach to performing a specification check on the linear specification. This methodology can be easily extended to test the specification of equations describing convergence in income; however, in this paper the testing is limited to a specification of the effect of savings, population growth, and education on steady state levels of income.

The analysis begins with re-estimation of MRW’s basic empirical equations using OLS. OLS estimation is used to establish a baseline for comparison and to test whether updated data confirms findings from the original MRW study. Initial nonlinear testing is conducted using GAM together with the original MRW data on 98 countries for the period 1960-1985. Using the MRW data, both the population growth variable and the human capital accumulation variable are significantly nonlinear. This finding also appears using an updated version of the data for the 1960-1985 period.

When a longer period (1960-2000) is estimated using both OLS and GAM, the OLS results are qualitatively similar to the original MRW results. However, GAM specification testing suggests that the human capital accumulation variable is significantly nonlinear for this period, indicating that the linear specification over-simplifies the relationships between income and education. There is a notable disparity between the results for the 1960-2000 period and the results for the 1960-1985 period. For the 1960-1985 period, the growth rate of the working age population has a nonlinear effect on steady state income in both the Textbook and the Augmented Solow Models. This result is robust to the data set used for estimation. For the period 1960-2000, this nonlinear effect of population growth disappears. This result is surprising but supports Temple (2000), who observes that “we must wait until…the year 2005 to see whether empirical models originally estimated using data from 1960-1985 perform well over the 1985-2005 period” (p. 202). The results in this paper suggest major differences in the models describing these two periods.

The strongest finding of the paper is that the effect of human capital on income is nonlinear, increasing in the level of human capital. This result appears across different sets of countries, the two different time periods under study (1960-1985 and 1960-2000), and different measurement strategies for human capital.

As a test of the performance of the 1960-1985 models, these models are applied to the 1960-2000 data. With both linear and nonlinear models for the 1960-1985 period in hand, it is possible to compare the performance of these two models when applied to the extended updated data. The nonlinear model of steady state income estimated with GAM forecasts better out of sample than the linear model estimated with OLS. GAM reduces the sum of squared forecasting error by about 10% relative to OLS for both the Textbook and the Augmented Solow Models.

The rest of the paper is organized as follows. Section 2 outlines the empirical methodology, including the estimating equations for steady state income levels. Section 3 discusses data definitions and sources. The results of the OLS and GAM testing are presented in Section 4. The results of the out-of-sample forecasting are presented in Section 5. Section 6 offers conclusions and directions for further research.


EMPIRICAL METHODOLOGY MRW study the effects of the growth determinants on standard of living. The basic OLS estimation in

this paper follows the approach of MRW. For the Textbook Solow Model, assume a Cobb-Douglas production function, so that production at time t is given by the Equation 1.

EQUATION 1

COBB DOUGLAS PRODUCTION FUNCTION, TEXTBOOK MODEL

1( ) ( ) ( ( ) ( )) 0 1 Y t K t A t L tα α α−= < <

The notation is as follows: Y is output, K is physical capital, L is labor, and A is the level of technology. L and A are assumed to grow exogenously at rates n and g (respectively). In addition, capital depreciates at a constant rate δ. Assuming that a constant fraction of output, s, is saved, MRW show that log steady state income per capita will be given by the equation in Figure 2.

EQUATION 2

LOG STEADY STATE INCOME PER CAPITA, TEXTBOOK MODEL

( )ln ln( ) ln( )( ) 1 1

Y t a s n gL t

α α δα α

= + − + + − −

Equation 3, the empirical equation used to estimate the model, follows from Equation 2.

EQUATION 3 EMPIRICAL ESTIMATING EQUATION, TEXTBOOK MODEL

0 1 2( )ln ln( ) ln( )( ) i i i

i

Y t s n gL t

γ γ γ δ ε

= + + + + +

Here, is is the savings rate for country i averaged over the time period under study, in is the growth

rate of the working age population for country i averaged over the time period under study, g+δ is assumed to be a constant 0.05, and ε is an error term assumed to be independent of s and n. The expected results are 1 0γ > and 2 0γ < .

Similarly, MRW derive an empirical estimating equation for the Augmented Solow Model which splits capital into its physical and human components. The augmented model begins with the Cobb-Douglas production function, as outlined in Equation 4.

EQUATION 4

COBB DOUGLAS PRODUCTION FUNCTION, AUGMENTED MODEL

1( ) ( ) ( ) ( ( ) ( ))Y t K t H t A t L tα β α β− −=

Here, H(t) is the stock of human capital at time t and all other variables are defined as above. From the production function an equation similar to Equation 2 above can be derived, where sk is the fraction of income invested in physical capital and sh is the fraction of income invested in human capital.


EQUATION 5 LOG STEADY STATE INCOME PER CAPITA, AUGMENTED MODEL

( )ln ln (0) ln( )( ) (1 )

ln( ) ln( )(1 ) (1 )k h

Y t A gt n gL t

s s

α β δα βα βα β α β

+= + − + + − −

+ +− − − −

Equation 5 is the basis for the empirical specification of the Augmented Solow Model, which is given

by Equation 6. EQUATION 6

EMPIRICAL ESTIMATING EQUATION, AUGMENTED MODEL

0 1 2 3( )ln ln( ) ln( ) ln( )( ) i ik i h i

i

Y t s n g sL t

γ γ γ δ γ ε

= + + + + + +

Here,

ihs is the average rate of human capital accumulation (or level of human capital) for country i. The expected results are 1 0γ > , 2 0γ < , and 3 0γ > . If the level of human capital, rather than the accumulation rate of human capital, is used the theoretical predictions of the magnitude of the coefficients on all variables in the model differ, while the predicted signs remain the same. If the level of human capital is used, Equation 5 changes slightly. See MRW (p. 417-418) for a full discussion.

In order to test whether the linear specification is appropriate, a nonlinear model is specified and GAM is used for estimation. Originally proposed by Hastie and Tibshirani (1990), a generalized form of GAM is given in Equation 7, where ( , )j jf x m is a polynomial smoothing function degree m fit to the jx

series. If 1m = then ( ,1)j j jf x x≡ , but if 1m > then the x variable is smoothed. Once the ( , )j jf x m

vectors for 1,j k= are calculated, OLS is used to estimate , 0,j j kγ = . The GAM algorithm also allows for significance tests of the estimated coefficients using asymptotic standard errors corrected for degrees of freedom. See Hastie and Tibshirani (1990, p. 127) for a full discussion.

EQUATION 7

THE GENERALIZED ADDITIVE MODEL

1 2 0 1 1 1 2 2 2( ) ( , , , ) ( , ) ( , ) , , ( , )k k k kE y g x x x f x m f x m f x m eγ γ γ γ= = + + + + + A key feature of the GAM algorithm is that the gain from removing the linearity assumption of one

variable, j, in an otherwise linear model can be tested. This is accomplished by calculating 'e e when 1jm > and comparing this to 'e e when 1m = . A statistical test is used to determine whether the

increase in 'e e when 1m = , holding the functional form of all other variables fixed, is statistically significant.

Stokes (2008) suggests that GAM should be used as a routine diagnostic procedure. Stokes further suggests that it is theoretically sound to use the GAM coefficients to estimate out of sample. A critical question is whether a nonlinear model estimated with GAM provides better out of sample forecasts than a linear model estimated with OLS. If the relationship between the variables is truly nonlinear, and if GAM has accurately estimated the functional form of the nonlinear relationship, then the model estimated with


GAM should provide better out of sample predictions. On the other hand, if GAM has over-fit the data, the out of sample predictions from the nonlinear model will be poor.

The nonlinear estimating equations are variants of Equations 3 and 6, where the right-hand side variables are smoothed as described above. The nonlinear equation for the Textbook Solow Model is given by Equation 8, where fi is a smoothing function on the ith variable. Similarly, the nonlinear equation for the Augmented Solow Model is given by Equation 9. The basic GAM results reported in subsequent sections are insensitive to the choice of the degree of smoothing, so degree three has been used throughout.

EQUATION 8

GAM ESTIMATING EQUATION, TEXTBOOK MODEL

0 1 1 2 2( )ln (ln( ), ) (ln( ), )( ) i i i

i

Y t f s m f n g mL t

γ γ γ δ ε

= + + + + +

EQUATION 9

GAM ESTIMATING EQUATION, AUGMENTED MODEL

0 1 1 2 2 3 3( )ln (ln( ), ) (ln( ), ) (ln( ), )( ) i ik i h i

i

Y t f s m f n g m f s mL t

γ γ γ δ γ ε

= + + + + + +

The following approach is used to generate the out-of-sample forecasts. For forecasting using the

linear specification, Equations 3 and 6 are estimated with OLS using data for the period 1960 to 1985. The coefficients are then saved and applied to the 1960-2000 data, as in Equation 10, where _ˆOLS forecasty is a vector of forecasted log incomes for the 1960-2000 period, 1960 2000X −

is the matrix of explanatory variables for the 1960-2000 period, and

1960 1985

OLS

−Γ is the vector of coefficients from the linear model of the

1960-1985 period.

EQUATION 10 OLS OUT OF SAMPLE FORECASTING

1960 1985_ 1960 2000ˆ OLSOLS forecasty

−−= Χ Γ

Similarly, nonlinear forecasts are obtained by first generating coefficients and smoothing functions by estimating (with GAM) Equations 8 and 9 with the 1960-1985 data. Then, using updated data, the estimated smoothing functions are applied to the explanatory variables, and the smoothed vectors are multiplied by the 1960-1985 coefficients, as in Equation 11, where 1960 2000[ ]s −Χ is the matrix of smoothed explanatory variables obtained by applying the 1960-1985 smoothing functions to the 1960-2000 data, and 1960 1985

GAM−Γ is the vector of coefficients from the nonlinear model.

EQUATION 11

GAM OUT OF SAMPLE FORECASTING

1960 1985_ 1960 2000ˆ [ ] GAMGAM forecasty s

−−= Χ Γ


DATA For comparison of the results over different samples and time periods, five distinct cross-section data

sets are defined and named Data Set A through Data Set E. These data sets are described in detail below and summarized in Table 1.

MRW use a cross-section of 98 non-oil producing countries for the period 1960-1985. The MRW data are from an early version of the Penn World Table (PWT) constructed by Summers and Heston (1988). MRW measure n as the average rate of growth of the working age (15-64) population, s as the average share of real investment in real GDP, and Y/L as real GDP in 1985 divided by the working age population in that year. MRW measure human capital accumulation using a proxy that measures the approximate percentage of the working age population that is in secondary school. Data Set A is the MRW data as published in the 1992 paper.

In order to update the analysis to include years after 1985, a data set for the period 1960-2003 is constructed primarily from the PWT, Version 6.2, documented in Summers, et. al. (2006). PWT 6.2 is the source of data on real GDP per capita, investment share of GDP, and population size. Data on the percentage of the total population that is of working age was obtained from the World Bank World Development Indicators (2008). Data Set B contains 97 countries for which all data series are available for the 1960-2003 period.

Two measures of human capital accumulation augment the PWT 6.2 data. First, making an assumption that the approximate rate of human capital accumulation has not changed significantly since 1985, the MRW human capital accumulation measure is transplanted into the updated data set. This augmented data set, Data Set C, contains 83 countries which are contained in both the MRW data and the updated PWT 6.2 data. For comparability, Data Set D is a subset of the original MRW data (Data Set A), but contains only the 83 countries which are also contained in the PWT 6.2 data. The second measure of human capital uses data on the average years of education for the population over 25, averaged over the study period, from Barro and Lee (2001). This second augmented data set, Data Set E, contains 75 countries; these are countries contained in the 97 country PWT 6.2 data set for which the Barro-Lee data is also available.

Two measures of human capital accumulation are used because measures of this variable are thought to be flawed. Although the Barro-Lee data is acknowledged to be an improvement upon MRW’s rough measure, it still is imperfect. Measuring human capital in two ways provides a robustness check on both the linear and the nonlinear results for this variable. It is important to note, however, that the MRW measurement corresponds to the rate of accumulation of human capital, while the Barro-Lee data corresponds to the level of human capital. While this does not change the basic expected results of the empirical equations, it does change the expected magnitude of the coefficients. See Mankiw, et. al. (1992, p. 418) for a discussion. The data sets used for estimation are briefly described in Table 1. RESULTS OF ESTIMATING THE LINEAR AND NONLINEAR MODELS

To establish a baseline of comparability between the data sets under study, estimation of the linear

specification of both the Textbook and the Augmented Solow Models for the period 1960-1985 is replicated. Table 2 gives the results of the OLS estimation of Equation 3, the empirical specification of the Textbook Solow Model, using various data sets. Columns 1 and 2 give the original MRW results alongside the results when Data Set D, which contains the 83 countries that overlap the PWT 6.2 data, is used. Reducing the sample size from 98 to 83 countries does not make a substantive difference in the OLS results, as the coefficients are all statistically equal. The explanatory power of the model improves slightly ( 2R = 0.59 vs. 0.64), but this is the only notable change.

Columns 3 and 4 of Table 2 report the OLS results of Equation 3 using data from the updated version (6.2) of the Penn World Table for the period 1960-1985. A comparison of Columns 1 and 3 shows that


TABLE 1 DESCRIPTION OF DATA SETS A THROUGH E

Data Set Description N

A Full sample from Mankiw, Romer, and Weil (1992);

Period 1960-1985

98

B Full sample from Penn World Table, v. 6.2; Period 1960-2003

97

C Reduced sample from Penn World Table, v. 6.2, containing countries from Data Sets A and B that overlap. Augmented with the variable “SCHOOL” from Mankiw, Romer, and Weil. Period 1960-2003

83

D Reduced sample from Mankiw, Romer, and Weil containing countries from Data Sets A and B that overlap; Period 1960-1985

83

E Reduced sample from Penn World Table, v. 6.2, containing countries for which Barro-Lee (2001) data on educational attainment also is available. Period 1960-2000

75

when the OLS model is estimated with the full samples, there is a significant difference in the estimated coefficient of the savings rate, although the period under study is the same. The original MRW estimate of 1.42 is statistically different than the coefficient of 0.88 obtained using the updated data. A possible explanation for this difference is the non-comparability of the samples. Since Data Sets A and B contain only 83 common countries, different coefficients are not unexpected when the full (98 or 97 country) samples are used. However, an examination of Columns 2 and 4 which estimate the OLS model for only the 83 overlapping countries shows that the statistical difference in the coefficients on savings persists even when the samples are identical. The explanation for this difference appears to be a sizable difference in the average savings rate obtained from two different versions of the Penn World Table.

A similar result arises from the OLS estimation of the Augmented Solow Model, Equation 6. Table 3 gives these results. Again, comparing Columns 1 and 2 demonstrates that dropping the non-overlapping countries from the original MRW sample makes little difference in the estimated coefficients. However, there is a significant difference between the estimated coefficients on the log of savings between Data Sets C (the data from the updated version of the Penn World Table) and D (the comparable sample from the original MRW paper) even when looking at an identical period (1960-1985). Comparison of Columns 2 and 3 shows this difference. Coefficients on the population growth rate and the school variable are statistically equal. The coefficients reported in Column 6 are somewhat different from those in Columns 2 and 3, but this difference is expected as Data Set E contains levels rather than growth rates of human capital.

Tables 2 and 3 also contain OLS results when the Textbook and Augmented Solow Models are estimated over updated periods. The OLS results for the period 1960-2000 are similar in nature to the results from the shorter 1960-1985 period, with the coefficient on log savings positive and significant and the coefficient on log population growth negative and significant. This demonstrates that the basic


TABLE 2 OLS ESTIMATION OF THE TEXTBOOK SOLOW MODEL

Dependent Variable: Log GPD per working-age person in terminal year of period

Data MRW PWT62 PWT62 PWT62 Period 1960-1985 1960-1985 1960-2000 1960-2003 (1) (2) (3) (4) (5) (6) (7) (8) Data Set A D B C B C B C

Constant 5.43 6.50 4.98 5.58 2.59 3.43 1.97 2.79 (1.59)a (1.60) (1.55) (1.57) (1.61) (1.58) (1.65) (1.60) ln(I/GDP) 1.42b 1.52b 0.88b 0.95b 0.97b 1.08b 0.95b 1.05b (0.14) (0.15) (0.11) (0.11) (0.13) (0.12) (0.13) (0.13) ln(n+g+δ) -1.99b -1.68b -2.16b -1.99b -3.20b -2.97b -3.42b -3.19b (0.56) (0.56) (0.56) (0.57) (0.57) (0.56) (0.58) (0.56)

2R 0.59 0.64 0.49 0.56 0.59 0.66 0.59 0.67 Countries 98 83 97 83 97 83 97 83 a Standard errors in parentheses. b Significant at the 99% level.

predictions of the Solow model hold for the longer period. These results replicate findings of other authors (Bernanke and Gurkaynak 2001; Karras 2008). Notice that results for the period 1960-2003 (the longest period for which data is available) are nearly identical to results for the period 1960-2000. This can be seen in the comparison of Columns 5 and 6 with 7 and 8 in Table 2, and also in the comparison of Column 4 with Column 5 in Table 3. For this reason, the nonlinear estimation discussed next is limited to the 1960-2000 period; there is no expectation that adding the additional three years of data would change the basic findings.

In summary, with the exception of the estimated coefficient on the log of savings variable, the OLS results originally presented by MRW hold across different versions of the 1960-1985 data. The results also hold over the longer (1960-2003) period. These findings increase confidence in the suitability of the simple linear model; if the model is truly an accurate description of the determinants of steady state income, then it is expected that the basic results would not be sensitive to the time period under study.

The question of interest in this paper, however, is not only whether the original results hold for updated data, but also whether the implicit assumptions of the linear model have an important effect on the empirical results. Use of GAM allows the relaxation of the linearity assumption. Rather than imposing the restriction that the coefficient on each variable is constant across the entire range of that variable, GAM allows the coefficients to vary by applying a data-determined smoothing function to each explanatory variable before estimation. Both the Textbook and the Augmented Solow Models are estimated using GAM, starting with the period 1960-1985, using the original MRW data as well as the updated version of the data (for the same period) from PWT 6.2.

Results of the GAM estimation of the Textbook Solow Model are reported in Table 4. GAM coefficients are reported and can be interpreted as the coefficients on the smoothed explanatory variables. Asymptotic standard errors (corrected for degrees of freedom) are reported in parentheses below the


TABLE 3 OLS ESTIMATION OF THE AUGMENTED SOLOW MODEL


Data MRW PWT62, augmented with

MRW’s SCHOOL variable

PWT62, augmented with Barro-Lee data

Period: 1960-1985 1960-1985 1960-2000 1960-2003 1960-1985 1960-2000 (1) (2) (3) (4) (5) (6) (7) Data Set A D C C C E E Constant 6.84 6.91 6.30 4.14 3.45 5.54 3.70 (1.18)a (1.26) (1.23) (1.25) (1.29) (1.20) (1.25) ln(I/GDP) 0.70b 0.78b 0.30c 0.36c 0.36c 0.27c 0.30c (0.13) (0.16) (0.13) (0.14) (0.14) (0.12) (0.14) ln(n+g+δ) -1.75b -1.71 b -2.00 b -2.93 b -3.16 b -1.21 b -1.88 b (0.42) (0.44) (0.44) (0.44) (0.45) (0.45) (0.46) ln(SCHOOL) 0.65 b 0.58 b 0.66 b 0.64 b 0.62 b 0.73 b 0.86 b (0.07) (0.08) (0.09) (0.09) (0.09) (0.10) (0.12) R2 0.78 0.78 0.73 0.79 0.79 0.71 0.75 Countries 98 83 83 83 83 75 75 a Standard errors in parentheses. b Significant at the 99% level. c Significant at the 95% level.

coefficients. See Hastie and Tibshirani (1990, p. 127) for overview of distribution and inference when using GAM. The sum of squared residuals reported with each explanatory variable in the body of Table 4 is the sum of squared residuals of the model if the current variable is constrained to be linear while all other variables in the model are smoothed. The row labeled ‘sig’ reports a p-value for a significance test of whether the sum of squared residuals when the variable is forced to enter the model linearly is significantly higher than the sum of squared residuals when all variables are smoothed. This is a statistical test, where the null hypothesis is that the sum of squared residuals is equal regardless of whether the variable is constrained to linearity. The alternative hypothesis is that the sum of squared residuals increases when the variable is constrained to linearity. The p-value indicates the confidence level with which the null hypothesis can be rejected.

The sum of squared residuals when all variables are smoothed is reported at the bottom of Table 4 as “GAM e’e.” This value can be qualitatively compared to the OLS sum of squared residuals. For a statistical comparison, however, the significance tests should be consulted as these have been corrected for the loss of degrees of freedom implicit in the GAM estimation. A high p-value for the significance test suggests that imposing the linearity assumption on the current variable has a significant effect on the sum of squared residuals of the estimated model. Similarly, a low p-value indicates that there is no gain from estimating the variable non-linearly.


TABLE 4 GAM ESTIMATION OF THE TEXTBOOK SOLOW MODEL

Dependent Variable:

Log GPD per working-age person in terminal year of period

Period 1960-1985 1960-1985 1960-2000 (1) (2) (3) (4) (5) (6) Data Set A D B C B C Constant 5.31 6.46 5.22 5.82 2.72 3.64 (1.52)a (1.53) (1.49) (1.47) (1.61) (1.57) ln(I/GPD) 1.35 1.47 0.84 0.90 0.95 1.05 (0.14) (0.14) (0.11) (0.11) (0.13) (0.12) e’e 40.62 29.31 48.01 34.14 48.37 34.03 sig 0.51 0.44 0.78 0.95 0.26 0.58 ln(n+g+δ) -1.98 -1.67 -2.03 -1.86 -3.14 -2.87 (0.54) (0.54) (0.54) (0.53) (0.57) (0.55) e’e 44.33 32.37 49.37 33.85 49.21 33.94 sig 0.99 0.98 0.93 0.93 0.59 0.55

2R 0.65 0.70 0.56 0.64 0.61 0.69 Countries 98 83 97 83 97 83 GAM e’e 39.57 28.54 45.75 30.93 47.70 32.80 OLS e’e 45.11 33.01 51.53 36.93 49.81 35.10 a Asymptotic standard errors (corrected for degrees of freedom) in parentheses.

For example, in Column 1, for the variable ln(I/GDP) (the log of average savings), the reported sum

of squared residuals is 40.62. The interpretation is that if this variable is constrained to enter the model linearly (with no smoothing), the sum of squared residuals of the model would jump from 39.57 (the GAM e’e) to 40.62. The p-value (‘sig’) of 0.51 suggests that this increase in the sum of squares is not statistically significant. By contrast, for the variable ln(n+g+δ) (the log of population growth), the reported e’e is 44.33. If this variable was constrained to enter the model linearly, the sum of squared residuals would increase from 39.57 to 44.33. The p-value of 0.99 indicates that this difference is statistically significant at the 0.01 level.

Starting with the period 1960-1985, Columns 1 and 2 of Table 4 report the results of GAM testing on the original MRW data, both the full 98-country sample and the reduced 83 country sample. The savings variable is not significantly non-linear, as there is no significant increase in the sum of squared residuals if linearity is imposed. The results of the significance testing are 0.51 for the 98 country sample and 0.44 for the 83 country sample. However, the population growth variable is significantly non-linear. The results of the significance testing on this variable are 0.99 for the 98 country sample and 0.98 for the 83 country sample. This suggests that imposing a constant coefficient on the population growth variable is not appropriate and that the linear specification is likewise not appropriate.

Columns 3 and 4 of Table 4 show the GAM results for the 1960-1985 period using data from version 6.2 of the Penn World Table. The expectation is that these results should be very similar to the results reported in Columns 1 and 2; however, the results are surprisingly different. Of particular interest is a


comparison of Columns 2 and 4, where the countries included in the sample are identical. Using Data Set D, savings is linear while population is non-linear. Using Data Set C, savings is significantly nonlinear while population growth is only marginally significantly nonlinear. Just as in the OLS results, these differences can best be explained by differences in the measurement of average savings between the two versions of the Penn World Tables.

To provide some insight into whether these differences are substantive or merely statistical, an examination of leverage plots showing the effect of the explanatory variables on income is useful. The slope of the leverage plot gives the effect on income at a given level of the explanatory variable. Figure 1 shows the effect of log savings and log population growth on log income using Data Set A. Figure 2 shows the same effects using Data Set B. Figure 3 shows the effects using Data Set C, and Figure 4 shows the effects using Data Set D. Notice that although there are statistical differences in the GAM results (as reported in Table 4), the graphical analysis indicates the results are qualitatively similar across the various data sets. The effect of log savings is consistently positive, while the effect of population growth is negative to a point, and positive beyond that point.

Table 5 gives the results of the GAM estimation of the Augmented Solow Model. Columns 1 and 2 give results for Data Sets A and D, respectively. Notice that GAM estimation using Data Set A suggests that both population growth and schooling are significantly non-linear, while savings is linear. (P-values

FIGURE 1 TEXTBOOK SOLOW MODEL, DATA SET A, PERIOD 1960-1985

FIGURE 2 TEXTBOOK SOLOW MODEL, DATA SET B, PERIOD 1960-1985

Effect of Variables on Log Income

lny1

985

Log Savings Rate

lns-3.2 -2.8 -2.4 -2.0 -1.6 -1.2 -0.8

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0Log Population Growth Rate

lnpop-2.97 -2.88 -2.79 -2.70 -2.61 -2.52 -2.43 -2.34

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50


solid line shows effect, dashed lines show confidence interv als

lny1

985

Log Savings Rate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-2.4

-1.6

-0.8

-0.0

0.8


lnpop-2.97 -2.88 -2.79 -2.70 -2.61 -2.52 -2.43 -2.34

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50


FIGURE 3 TEXTBOOK SOLOW MODEL, DATA SET C, PERIOD 1960-1985

FIGURE 4 TEXTBOOK SOLOW MODEL, DATA SET D, PERIOD 1960-1985

are 0.99, 0.98, and 0.72 respectively.) In this case, dropping the 15 countries and re-estimating with Data Set D makes a difference in the results. Savings in still linear with p-value 0.66, population growth is still nonlinear with p-value 0.99, but now schooling appears linear, with a p-value of 0.83.

Columns 2, 3, and 5 of Table 5 provide a comparison of GAM estimation for three different data sets for the 1960-1985 period using Data Sets D, C, and E respectively. In all three cases, savings is not significantly nonlinear. Data Sets D and E agree that population growth is significantly nonlinear, while Data Sets C and E agree with Data Set A that human capital accumulation is significantly nonlinear. Although there are some statistical differences across the different data sets, the general result is that log savings affects log income linearly, while log population growth and log human capital accumulation affect log income nonlinearly. Figures 5, 6, 7, and 8 show the leverage plots for the three explanatory variables for Data Sets A, D, C, and E. The leverage plots support the general finding that the behavior of the nonlinearity is consistent across the data sets.

A difference in the GAM results is observed for the 1960-2000 period. In Table 4, Columns 5 and 6 report the results of GAM estimation of the Textbook Solow Model for Data Sets B and C. The results are qualitatively similar for the two data sets; dropping the 14 countries not included in the MRW data does not make a difference in the results of GAM estimation. Importantly, for the 1960-2000 period, neither the log savings nor the log population growth variable is significantly nonlinear. This suggests that the


lny1

985

Log Savings Rate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0


lnpop-2.97 -2.88 -2.79 -2.70 -2.61 -2.52 -2.43 -2.34

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25



lny1

985

Log Savings Rate

lns-3.2 -2.8 -2.4 -2.0 -1.6 -1.2 -0.8

-3.0

-2.4

-1.8

-1.2

-0.6

0.0

0.6

1.2


lnpop-2.97 -2.88 -2.79 -2.70 -2.61 -2.52 -2.43 -2.34

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

1.25

1.50


TABLE 5 GAM ESTIMATION OF THE AUGMENTED SOLOW MODEL


Data MRW PWT62, augmented with

MRW’s SCHOOL Variable

PWT62, augmented with Barro-Lee data

Period 1960-1985 1960-1985 1960-2000 1960-1985 1960-2000 (1) (2) (3) (4) (5) (6) Data Set A D C C E E Constant 7.44 7.66 6.97 5.29 7.18 5.40 (1.09) (1.18) (1.14) (1.21) (1.08) (1.21) ln(I/GDP) 0.63 0.72 0.23 0.30 0.23 0.28 (0.12)a (0.15) (0.12) (0.14) (0.11) (0.14) e’e 20.63 16.92 19.00 19.42 14.74 16.83 sig 0.72 0.66 0.83 0.67 0.80 0.47 ln(n+g+δ) -1.49 -1.39 -1.72 -2.52 -0.55 -1.18 (0.38) (0.41) (0.41) (0.42) (0.41) (0.45) e’e 22.05 18.53 19.26 18.63 15.69 16.38 sig 0.99 0.99 0.89 0.05 0.97 0.07 ln(SCHOOL) 0.66 0.59 0.67 0.71 0.76 0.94 (0.07) (0.08) (0.09) (0.09) (0.09) (0.15) e’e 21.82 17.28 19.66 20.42 16.19 18.47 sig 0.98 0.83 0.95 0.94 0.99 0.97

2R 0.82 0.83 0.79 0.83 0.79 0.79 Countries 98 83 83 83 75 75 GAM e’e 20.07 16.17 17.77 18.54 13.76 16.27 OLS e’e 24.23 20.12 22.46 21.29 18.52 18.85 a Asymptotic standard errors (corrected for degrees of freedom) in parentheses.

OLS specification is appropriate for the Textbook Solow Model for the 1960-2000 period, which is a contradiction of the findings for the 1960-1985 period, where the effect of population growth rate was shown to be nonlinear. Figures 9 and 10 show the leverage plots for the 1960-2000 period, where the effects of the explanatory variables are constant across the entire range of the variables.

The results of GAM estimation of the Augmented Solow Model for the period 1960-2000 are similar in nature. Columns 4 and 6 of Table 5 give GAM results for this period, with human capital accumulation measured in two ways, first as a growth rate (Column 4) and second as a level (Column 6). The results are qualitatively similar: both log savings and log population growth are not significantly non-linear, but log human capital is significantly non-linear. The fact that these results agree with the results of the GAM estimation of the Textbook Model together with the consistency of the results across different samples (83


FIGURE 5 AUGMENTED SOLOW MODEL, DATA SET A, PERIOD 1960-1985

FIGURE 6 AUGMENTED SOLOW MODEL, DATA SET D, PERIOD 1960-1985

FIGURE 7 AUGMENTED SOLOW MODEL, DATA SET C, PERIOD 1960-1985


lny1

985

Log Savings R ate

lns-3.2 -2.8 -2.4 -2.0 -1.6 -1.2 -0.8

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00Log Population Growth R ate

lnpop-2.97 -2.79 -2.61 -2.43

-0.36

-0.18

0.00

0.18

0.36

0.54

0.72

0.90

1.08Log H uman C apital A ccumulation

lnschool-6 -5 -4 -3 -2

-1.6

-1.2

-0.8

-0.4

-0.0

0.4

0.8

1.2


lny1

985

Log Savings R ate

lns-3.2 -2.8 -2.4 -2.0 -1.6 -1.2 -0.8

-1.5

-1.0

-0.5

0.0

0.5

1.0


lnpop-2.97 -2.79 -2.61 -2.43

-0.36

-0.18

0.00

0.18

0.36

0.54

0.72

0.90


lnschool-6 -5 -4 -3 -2

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

g g Effect of Variables on Log Income


lny1

985

Log Savings R ate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75


lnpop-2.97 -2.79 -2.61 -2.43

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00


lnschool-6 -5 -4 -3 -2

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0


FIGURE 8 AUGMENTED SOLOW MODEL, DATA SET E, PERIOD 1960-1985

FIGURE 9 TEXTBOOK SOLOW MODEL, DATA SET B, PERIOD 1960-2000

FIGURE 10 TEXTBOOK SOLOW MODEL, DATA SET C, PERIOD 1960-2000


lny1

985

Log Savings R ate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75


lnpop-2.97 -2.79 -2.61 -2.43

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0Log H uman C apital

lnavsc85-1.6 -0.8 0.0 0.8 1.6 2.4

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5


lny2

000

Log Savings Rate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5


lnpop-3.0 -2.9 -2.8 -2.7 -2.6 -2.5 -2.4 -2.3

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5



lny2

000

Log Savings Rate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5


lnpop-3.0 -2.9 -2.8 -2.7 -2.6 -2.5 -2.4 -2.3

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5


FIGURE 11 AUGMENTED SOLOW MODEL, DATA SET C, PERIOD 1960-2000

FIGURE 12 AUGMENTED SOLOW MODEL, DATA SET E, PERIOD 1960-1985

vs. 75 countries) and measurement strategies for human capital lend strong support to the argument that this result is an accurate description of the relationship of human capital and steady state income.

Figures 11 and 12 show the leverage plots for the Augmented Solow Model in the 1960-2000 period. As in Figures 9 and 10, the effect of log savings is represented by a constantly sloped upward sloping line and the effect of log population growth is represented by a constantly sloped downward sloping line. Of interest here is the shape of the leverage plot for human capital. The effect of human capital on income is positive and increasing with the level of human capital. This suggests that while the effect of human capital accumulation on income is always positive, this effect is more pronounced in economies where there is already a substantial investment rate in and/or stock of human capital. RESULTS OF OUT-OF-SAMPLE FORECASTING

Following Temple (2000), the empirical models developed for the 1960-1985 period can be applied to the 1960-2000 data to test whether the linear or the nonlinear model performs better. Because the use of a nonlinear model leads to a loss of simplicity, and because the results are more difficult to interpret than linear results, it is important to consider whether the nonlinear model provides a substantive improvement over the linear, or whether the gain is mainly statistical. Another concern is the possibility of over-fitting,


lny2

000

Log Savings R ate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00


lnpop-3.0 -2.9 -2.8 -2.7 -2.6 -2.5 -2.4 -2.3

-1.5

-1.0

-0.5

0.0

0.5

1.0


lnschool-6 -5 -4 -3 -2

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5



lny2

000

Log Savings R ate

lns-4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5

-1.05

-0.70

-0.35

0.00

0.35

0.70


lnpop-3.0 -2.9 -2.8 -2.7 -2.6 -2.5 -2.4 -2.3

-1.25

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75Log H uman C apital

lnavsc-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5


since the nonlinear model is developed based on the input data set. One way of addressing these questions is to study whether the model estimated with GAM improves the out-of-sample predictive power of the model of steady state income relative to the model estimated with OLS. If the nonlinear model is over-fit, then the linear model should perform better out-of-sample since the nonlinear model will be particular to the data set that was used to generate it. If the nonlinear model predicts more accurately than the linear model, this would suggest that GAM has uncovered true nonlinear relationships between the variables in the model, and that these relationships persist when additional years of data are included.

It is a simple forecasting exercise to estimate linear and nonlinear models (with OLS and GAM, respectively) from Data Set D, the 83 country sub-set of the MRW data set, for both the Textbook and the Augmented Solow Models. Coefficients are reported in Tables 2 – 5. To test the predictive power of the linear versus the nonlinear model, these coefficients as well as the GAM smoothing functions are applied to Data Set C, the corresponding 83 country sample from Version 6.2 of the Penn World Table for the 1960-2000 period. Equations 10 and 11 in Section 2 describe the procedure used to generate the forecasts.

For the Textbook Solow Model using these two data sets, the nonlinear model outperforms the linear model by about 10% when used to forecast future income based on out-of-sample data. The sum of squared error for the out-of-sample predictions using OLS is 166.34, while the sum of squared error using GAM 148.58. For the Augmented Solow Model the results are similar. The sum of squared error for the out of sample predictions using OLS is 123.75, while the sum of squared error using GAM 110.40. Again, GAM provides an improvement of about 10% in out-of-sample forecasting. These results suggest that the nonlinear model based on the 1960-1985 period is not over-fit but rather has captured true relationships between income and the independent variables (savings, population growth, and education). If a researcher or policy maker was interested in predicting future income based on a current model, this result suggests that the researcher could expect a more accurate forecast using a nonlinear model versus an linear model.

CONCLUSIONS

Results of nonlinear estimation using GAM change depending on what time period is studied.

Although GAM estimation for the period 1960-1985 suggests that the population growth rate affects steady state income in a nonlinear way, this result disappears for the 1960-2000 period. By contrast, the effect of human capital is not constant across the levels of human capital, regardless of the time period under study. The effect of increasing human capital accelerates as the level of the human capital variable increases. This result is found across multiple data sets representing both different time periods and different measurement strategies for human capital. The result is also in keeping with an intuitive notion that while the effect of human capital accumulation will always be positive, the marginal effect of education will be stronger as the existing pool of human capital grows.

In addition, a nonlinear model estimated with GAM performs about 10% better than a linear model in out of sample predictions for both the Textbook and the Augmented Solow Models. This suggests that nonlinear modeling and GAM estimation are useful tools for forecasting. It also demonstrates that the models developed using the 1960-1985 data are relevant outside of this period, suggesting the GAM is uncovering true relationships in between the variables as opposed to over-fitting the initial sample.

There are several questions raised here that could be researched further. Of particular interest is the development of a theoretical explanation for the changing coefficients suggested by GAM. REFERENCES Azariadis, C. & Drazen, A. (1990). Threshold Externalities in Economic Development. Quarterly Journal of Economics, 105, (2), 465-90. Barro, R.J. & Lee, J-W. (2001). International Data on Educational Attainment: Updates and Implications. Oxford Economic Papers, 53, 541-563.


Bernanke, B. & Gurkaynak, R. (2001). Is growth Exogenous? Taking Mankiw, Romer and Weil Seriously. NBER Working Paper No. 8365. Durlauf, S.N. & Johnson, P.A. (1995). Multiple Regimes and Cross-Country Growth Behaviour. Journal of Applied Econometrics, 10, (4), 365-84. Galor, O. & Weil, D.N. (2000). Population, Technology, and Growth: From Malthusian Stagnation to the Demographic Transition and Beyond. The American Economic Review, 90, (4), 806-828. Hansen, B. E. (2000). Sample Splitting and Threshold Regression. Econometrica, 68, (3), 575-603. Hastie, T. & Tibshirani, R. (1990). Generalized Additive Models, New York: Chapman & Hall /CRC. Heston, A., Summers, R. & Aten, B. (2006). Penn World Table Version 6.2, Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania. Karras, G. (2008). Growth and Convergence, 1950-2003: What Can We Learn from the Solow Model? Applied Econometrics and International Development, 8, (1), 5-18. Liu, Z. & Stengos, T. (1999). Non-linearities in Cross-Country Growth Regressions: A Semiparametric Approach. Journal of Applied Econometrics, 14, (5), 527-538. Mankiw, N. G., Romer, D. & Weil, D.N. (1992). A Contribution to the Empirics of Economic Growth. Quarterly Journal of Economics, 107, (2), 407-437. Solow, R. (1956). A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics, 70, 65-94. Stokes, H.H. (1997). Specifying and Diagnostically Testing Econometric Models. Westport, Connecticut: Quorum. (Revised chapters, 2008). Summers, R. & Heston, A. (1988). A New Set of International Comparisons of Real Product and Price Levels Estimates for 130 Countries, 1950-1985. Review of Income and Wealth, 34, 1-25. Temple, J. (2000). Growth Regressions and What the Textbooks Don't Tell You. Bulletin of Economic Research, 52, 3, 181-205. World Bank (2008). World Development Indicators.


The Influence of Legitimacy and Multi-level Environments: A Case Study of

Taiwanese Subsidiary’s Entry Mode Choice in Mainland China

Shaodong Hu Shantou University

Hongxin Yao

Shantou University

Zongling Xu Shantou University

This research studies Taiwanese subsidiaries’ entry-mode choice from the angle of institutional theory. The findings show that Taiwanese subsidiaries in Mainland China imitate each other to gain legitimacy when choosing an entry mode at different levels of institutional environments. Taiwanese subsidiaries prefer to imitate the entry mode of their own group, rather than institutional environments at other levels, and prefer to imitate the entry mode of local industry institutional environment, rather than national industry institutional environment and region institutional environment, which indicates that Taiwanese subsidiaries tend to seek legitimacy in a narrow institutional environment among a multi-level institutional environments in Mainland China. INTRODUCTION When investing directly in foreign countries, the choice of foreign entry mode is a key decision, and a key issue that international companies face. The choice of entry mode has an important impact on the corporation's future performance and survival. Previous research on the entry modes of foreign subsidiary is mainly from an economic perspective with a focus on cost minimization and efficiency maximization to analyze the major factors that affect entry mode of foreign subsidiary. However, it ignores the impact of the institutional environment on the foreign subsidiary. The institutional condition is of vital importance for emerging economies like Mainland China (Peng Weigang, 2009), and is an important factor affects foreign businesses (Meyer, 2001, Contractor, 1990). Therefore, to decide the entry mode a foreign subsidiary should consider not only the economical goals, but also the issue of legitimacy. According to institutional theory, when in a institutional environment, the more foreign subsidiaries that adopt a certain entry mode, the more legitimacy that entry mode is, since new foreign subsidiaries tend to adopt a more legitimacy entry mode (Chan & Makino, 2007; Lu, 2002). This means new foreign subsidiaries imitate the entry modes of established foreign subsidiaries. For a Taiwanese subsidiary investing in Mainland China, it faces constraints in both the external institutional environment, and the internal institutional environment from its parent company.[2]Therefore, this research takes an example of Taiwanese subsidiaries in Mainland China, and attempts to study from the angle of institutional theory,


and from different levels of institutional environments, to find out whether the later entrants have imitated the entry modes of the earlier entrants in Mainland China. LITERATURE REVIEW Theories in relevant literature on foreign entry modes are mainly transaction cost theory, eclectic theory of international production, and the theory of organization ability etc. According to transaction cost theory (Williamson, 1975), the purpose of choosing an entry mode is to reduce transaction cost. It analyzes the organizational boundaries of enterprises from the angle of minimizing transaction costs. According to this theory, asset specificity, the uncertainty of the trade and the frequency of transactions all affect the optimal trading mechanism. According to eclectic theory of international production (Dunning, 1980), the choice of entry modes is due to the advantage of a company’s ownership, advantage of the region and the advantage of internalization. According to theory of organization ability (Kogut & Zander, 1993), the choice of entry modes reflects the most effective way to use previous experience, and explore new ideas. The above theories are mainly from an economic perspective, and based on comparisons between cost, and the benefits of different ownership structures, to take into account economic factors such as minimizing investment venture and trading cost, increasing access to resources, and ensuring the control of assets and operation of subsidiaries. It emphasizes that the cost minimization and efficiency maximization is the major decisive factor in the choice of an entry mode.

Nevertheless, the maximization of efficiency cannot offer a full explanation for the motivation of adopting a foreign entry mode. Foreign subsidiaries investing in a host country should consider both the economic goals and the issue of legitimacy (Dimaggio and Powell, 1983). Especially when investing in developing countries, the foreign capital policy of the host country is regarded as an important factor affecting the equity stake structure (Contractor, 1990). Legitimacy is a generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions (Suchman, 1995). Legitimacy reflects the uniformity of common beliefs between legitimate companies and legitimate actors (includes government, vendor, consumer, professional organization, scientific institute and other relevant persons). Legitimate actors are the observers and constitutors of the companies operating environment. They determine whether the organizational activities are legitimate or not. They offer the resources to enable legitimate companies to operate continuously in a competitive environment, and to incarnate social acceptance for the companies (Suchman, 1995). Based on these rules, legitimate actors have developed their institutional environment, and place the company under the pressure of isomorphism, forcing them to adopt an orthodox organizational form and practice (Dimaggio and Powell, 1983). In the case of more and more organizations adopting a single form or practices, the possibility of imitation by other organization is highly increased (Dimaggio and Powell, 1983). According to institutional theory, because the decision and behavior is adopted by other organizations, it raises the legitimacy of similar decisions or behavior, especially when facing high uncertainty, the imitation is of greater importance since imitation can reduce the uncertainty (Dimaggio and Powell, 1983). Thus, there is imitation between companies. Some recent researchers have emphasized the motive of legitimacy and tested the contribution of imitation in the choice of foreign entry mode (Henisz, 2001; Chan & Makino, 2007; Lu, 2002; Guillen, 2002；2003). Lu (2002) believes that there is an imitation between organizations, and within the organization on the choice of Japanese subsidiaries’ entry mode. Later entrants tended to follow the entry mode patterns established by earlier entrants, which supports the institutional isomorphism. Chan and Makino (2007) argue that when entering the international market, Japanese subsidiaries have faced different institutional pressure from different levels of institutional environment (host country, local industry, parent companies). In order to gain legitimacy, Japanese subsidiary imitates the entry modes of other Japanese subsidiaries. The research by Henisz (2001) about 2705 investment location choices shows that when a foreign subsidiary enters a new market, it will be at risk because of its little experience. At the same time, the decisions and behaviors of other foreign subsidiaries which have entered the market earlier can be referenced to them. Guillen (2002) believes that the experience of investment in Mainland China by Korean business groups


and imitation among the same industry have promoted expansion of Korean business groups in Mainland China. Guillen (2003) studies the entry mode process of Korean investment in Mainland China, from joint ventures to wholly owned subsidiaries, and concludes that the evolution is the outcome of imitating other subsidiaries in the same business group or same industry.

The main industries that Taiwanese business groups invest in Mainland China are the electronic information industry and traditional manufacturing etc. Since these industries have no strict limitation of entry mode for foreign subsidiary, Taiwanese subsidiary can choose their entry mode freely. This research looks at examples of Taiwanese subsidiaries in Mainland China to test whether the later entrants have imitated the entry mode of the earlier entrants in different levels of institutional environments. The innovative contributions are reflected in two points: firstly, no evidence has been found on empirical research for entry modes of Taiwanese subsidiaries in Mainland China. The study about entry modes of foreign subsidiaries focused on economic theory, but ignored legitimacy. This research confirms the imitation of entry modes choice for Taiwanese subsidiaries in Mainland China, microscopically from the perspective of institutional theory; it gives a new explanation for the choice of entry modes for Taiwanese subsidiaries, and compensates for the weakness of the research in this aspect. Secondly, Current literature which studies the impact of different levels of institutional environment (such as host country, industry etc.) is set in the frame of transnational study. There is a limited literature looks at the impact on entry mode for foreign subsidiary entering a country with different institutional environment among regions or industries. This research pays attention to the choice of entry mode for Taiwanese subsidiaries in different levels of institutional environments, the differentiation of institutional pressure Taiwanese subsidiaries face in different regions and industries. Meanwhile, different parent company put different pressures on its subsidiaries. This paper puts emphasis on the impacts of different levels of institutional environments on the choice of entry mode for Taiwanese subsidiaries, as well as supplying a relatively comprehensive explanation for the choice of entry mode for Taiwanese subsidiaries in Mainland China.

INSTITUTIONAL ENVIRONMENT AND RESEARCH HYPOTHESIS

This paper divides the institutional environment that Taiwanese subsidiaries in Mainland China are facing into an external institutional environment and internal institutional environment. External institutional environment is divided into institutional environment at a national industry level, regional level, and local industry level. Internal institutional environment refers to the internal institutional environment of Taiwanese business group. External Institutional Environment and Research Hypothesis National Industry Institutional Environment

In order to develop the economy and optimize the industrial structure, Mainland China has made relevant industrial policies and arrangements for different industries. The foreign funded industrial policy is the major policy tool for guiding the flow of foreign direct investment, and stimulates industrial structure optimization (Pei, 2006; Zhao, 2002). According to the previous industrial policy and investment directory, Mainland China has adopted a lenient policy to electronic, chemical, and metallurgical as well as other technology-intensive industries and capital-intensive industries. It has adopted a stringent policy for clothing, food manufacturing and other labor-intensive industries, except for the industrial policy; the industrial structure is also a very important part of the industrial institutional environment at national level, such as industry concentration ratio、intensity of competition and infrastructure. Industrial structures affect the barrier of entry and profitability of an enterprise, and generate constraints to enterprises. Industrial policy and industrial structure constitute the institutional environment of certain industries, which may have a vital impact on the choice of entry mode for foreign subsidiary. Foreign subsidiary prefer to adopt sole proprietorship mode in industries, which has policy support and the barriers to entry is low, such as in the electronics industry. While, foreign subsidiary


prefers to adopt joint venture in industries that are restricted and the barriers to entry are higher, such as the automotive industry. Region Institutional Environment

Regions in Mainland China are in different development stages. There are significant regional differences in the institutional environment. The main reasons are various. Firstly, the marketization process is different in the gradual reform process, and the economic development is uneven. The marketization and economic level of the southeastern coast is obviously higher than other regions (Fan et al., 2007). Secondly, as an entity with independent interests, the local governments are inconsistent in execution of central laws and policies, which means that where there is a policy, there is also a countermeasure to the different institutional environments in different regions (Yin et al., 2007).

At present, Mainland China is not a unified "big market", but is made up of lot of "small markets" (Meyer, 2008). Under the influence of economic globalization, the local protectionism and market segmentation of Mainland China has been restrained and improved. However, the market of Mainland China still has characteristics of federalism (Meyer, 2008). A research conducted by Taiwan Electrical and Electronic Manufacturer’s Association on the investment environment of Mainland China shows that there is a significant differentiation between legal environments, social environment, economic environment, and operating environment among different regions. For example, the investment environment of cities in Yangtze River Delta region is much better than in the Perl River Delta region. When investing in Mainland China, Taiwanese subsidiaries should not only take into consideration the overall investment environment, but also the local investment environment, especially the local institutional environment. In regions with higher institutional development level, especially in higher marketing regions, foreign subsidiaries tend to adopt sole proprietorship (Pan and Lu, 2006). Local Industry Institutional Environment

There is a unified policy for foreign funded industries; however, there is still differentiation among industrial policies of different regions. We identify two reasons for the variation of the local industry institutional environment. First, the goals of the central government and local government do not match. While the central government focuses on the optimization of industrial structure, the local government pays more attention to the amount of foreign investment, employment promotion and economic growth. Therefore, local government adopts different strategies to execute the policies for foreign funded industries (Yin et al., 2007). Second, the industrial structures differ in regions cross the country, as such, the industrial plans and industrial development strategies are not the same in different regions. For example, Shanghai pays more attention to the development of the modern service industry, especially the financial services; but the western areas tend to develop labor-intensive industries to undertake the shift of industries from developed areas.

The variation of local industry institutional environment indicates that the legitimate requirement for an industry in different regions is not the same, and therefore, even in the same industry, the restraints of local industry institutional environment for Taiwanese subsidiaries vary from region to region. Research Hypothesis

Multinational corporations from the same country or same region share the same cognitive domain (Tversky and Kahneman,1974), therefore they would like to imitate each other’s behavior to make a choice, and react to others with similar methods (White,1981;Haveman,1993). Previous research shows that foreign companies tend to concern the behaviors of companies which are from the same country or region (Chan and Makino, 2007; Lu, 2002; Guillen, 2002; 2003). Accordingly, when a Taiwanese subsidiary enters into Mainland China, it will like to imitate the entry mode of subsidiaries of other Taiwanese business groups.

In a certain level of institutional environment, if there are more subsidiaries of other Taiwanese business groups who adopt a certain entry mode in Mainland China, the entry mode is more legitimacy; therefore, new entrants are more likely to adopt this entry mode. While, if there are few subsidiaries from


other Taiwanese business groups choose this entry mode, then this entry mode is less legitimacy, therefore, new entrants is less likely to adopt this entry mode. Therefore we formulate the following hypotheses:

H1: In national industry institutional environment, the later entrants would imitate the entry mode of the earlier entrants of other business group. H2: In regional institutional environment, the later entrants would imitate the entry mode of the earlier entrants of other business group. H3: In local industry institutional environment, the later entrants would imitate the entry mode of the earlier entrants of other business group.

Internal Institutional Environment and Hypothesis

Business groups have provided its subsidiaries with an internal institutional environment, and defined the suitable ways to operate (Kostova and Zaheer, 1999). Subsidiaries must abide the internal institutional demands to gain the internal legitimacy. Some researchers indicate that if the overseas subsidiary is closely bounded up with the parent firm, the parent firm would put lots of pressure on its subsidiary to make its norms and practice consistent with itself. On the contrary, if the overseas subsidiary is relatively independent in operating and management, the parent company would put less pressure on the subsidiary to achieve the internal unity (Davis et al., 2000).

The subsidiaries of the same parent company are restrained by each other in transaction, proprietorship, and controlling aspects. They can share information and experiences, and believe that it is reasonable to adopt similar practice or strategies (Dimaggio and Powell, 1983). Business group can provide subsidiaries with operating experience of other subsidiaries overseas, for example, the earlier entered subsidiaries can provide information to the later entered subsidiaries, and let the latter know which entry mode is best. Organizational theory believes that the practice would be institutionalized with the increase of times adopted. Research by Chan and Makino (2007), Lu(2002) supports this viewpoint, and stresses that isomorphism makes international companies tend to adopt precious entry mode.

According to this logic, we believe that if one entry mode has been adopted more times in the same business group, the higher legitimacy the entry mode is, therefore, the later entrants would tend to adopt the entry mode to keep the internal consistency. Therefore, we make a hypothesis as follow:

H4: When investing in Mainland China, the later entrants would imitate the entry mode of the earlier entrants of the same business group.

METHODOLOGY Sample

The data of the research is from the survey report of Study on Business Groups in Taiwan (2006 version) published by China Credit Information Service Cor. Ltd (Taiwan), which includes business group profile, organizational structure chart, historical evolution, subsidiaries profile and performance information. The historical evolution has included the established time, place, and industries engaged in and establishing background. Subsidiaries profiles provide information that includes the shareholding ratio of the subsidiary when it was established and so on. In this survey report, we choose manufacturing subsidiaries in Mainland China as samples and excluded:

(1) Subsidiaries which lack important information such as shareholding ratio. (2) Business group with only a subsidiary in Mainland China, because a single subsidiary would

confuse the parent companies effects (Bowman and Helfat, 2001), and is hard to measure its internal imitation.


The final sample is made up of 478 subsidiaries in 10 industries (According to two-digit SIC) and 17 provinces. Among which, the number of wholly owned subsidiary is 382, joint ventures 96, the average shareholding ratio of business group in the subsidiaries is 93.6%. Dependent Variable

Dependent Variable refers to the ownership entry mode of the subsidiary. This variable is measured by the shareholding ratio of the Taiwanese business group. We obtained the data of dependent variable from Study on Business Groups in Taiwan (2006 version). Independent Variables

Independent Variables are the institutional variables of different levels, which include national industry entry mode, regional entry mode, local industry entry mode, and business group entry mode. The measurement of entry modes for different levels refers to Lu's (2002) method. When a Taiwanese business group sets up a subsidiary in Mainland China, the percentage that the wholly-owned subsidiaries account for of the total subsidiaries in nationally industrial, regional and locally industrial levels, measures the national industry entry mode, regional entry mode and local industry entry mode respectively. The entry modes of business group is measured by the percentage which the wholly-owned subsidiaries account for the group total subsidiaries in the Mainland China. We obtained the data of independent variables from Study on Business Groups in Taiwan (2006 version). Controlled Variables

We use three local-specific control variables. First is gross domestic product (GDP) per capita in each region to represent the level of economic development of each region. The current studies show that the uncertainty is higher in less developed region, and the transnational enterprises are more likely to adopt joint venture (Shan, 1991). Second is the degree of local marketization. Because the foreign-funded companies grow up in a market economy, they prefer a market economy environment. Therefore, in regions that are highly marketed, they tend to adopt sole proprietorship (Pan and Lu, 2006). Third is the local economic growth rate. Regions have higher economic growth rates are more attractive to foreign-funded enterprises. Foreign-funded enterprises tend to adopt high control level entry modes. We obtained the data of GDP per capita and the local economic growth rate from China Economic Information Web, the local marketization from NERI INDEX of Marketization of China’s Province 2009 Report (Fan et al., 2009).

We control for two types of business group resource: the investment experience in Mainland China and business group size. The experience in investing in Mainland China is measured by the years of operating in Mainland China for the business group. The current studies show that experienced transnational companies prefer high control level entry modes (Delios and Beamish, 1999).Business group size which is measured by the total assets, represents the size of the available pool of resources or capabilities that can be exploited in a new market. We control for its positive effect on the subsidiary's structure of ownership. We also use a relative size of subsidiary as a control variable, measuring by a ratio of the subsidiary's assets to the business group's total assets. We obtained the data on the three variables from Study on Business Groups in Taiwan (2006 version).

The descriptive statistics and correlations of variables are provided in table 1and table 2 respectively.


TABLE 1 THE DESCRIPTIVE STATISTICS OF ALL OF THE VARIABLES

Variables Observed Value Mean Standard

Deviation Minimum

Value Maximum

Value

Ownership Entry Mode 478 0.936 14.875 0.21 1

National Industry Entry Mode 478 0.806 0.122 0.57 0.90

Regional Entry Mode 478 0.797 0.086 0.54 1

Local Industry Entry Mode 478 0.807 0.219 0 1

Business group Entry Mode 478 0.811 0.283 0 1

Business group’s sizea 478 5.223 1.086 3.45 7.68

Subsidiary Relative size 478 0.016 0.044 4.92e-06 0.63 Investment Experience in Mainland China 478 5.469 3.946 0 14

Regional Economic Levelb 478 9.847 0.425 8.69 10.75

Regional Marketization level 478 8.257 0.987 4.11 9.35

Reginal Economic Growth Rate 478 0.170 0.020 0.10 0.23 a Logarithm;in 1000 millions of New Taiwan Currency. b Logarithm;in yuan.

TABLE 2

CORRELATION MATRIX

Variables 1 2 3 4 5 6 7 8 9 10 1 Ownership Entry Mode 2 National Industry Entry Mode 0.26** 3 Regional Entry Mode 0.24** 0.22** 4 Local Industry Entry Mode 0.50** 0.55** 0.38** 5 Business group Entry Mode 0.68** 0.38** 0.30** 0.69** 6 Business group’s sizea 0.08+ -0.06 -0.06 -0.03 0.03 7 Subsidiary Relative size -0.16** 0.07 -0.07 -0.05 -0.12** -0.28** 8 Investment Experience in Mainland China 0.04 0.01 -0.06 0.07 0.06 0.24** -0.19** 9 Regional Economic Levelb 0.05 0.06 0.29** 0.10* 0.05 0.01 0.02 -0.06 10 Regional Marketization level 0.08* 0.20** 0.41** 0.17** 0.10* -0.01 0.04 -0.06 0.67** 11 Reginal Economic Growth Rate 0.09* 0.14** 0.35** 0.18** 0.14** 0.03 -0.10* 0.05 0.20** 0.32**

Note：** p <0.01,* p <0.05,+ p <0.1 ANALYSIS AND RESULTS OLS Estimated Results

We adopt OLS to estimate the models, and test the relationship between entry modes and institutional variables by Stata 10. The results of the analyses are provided in Table 3.


TABLE 3 OLS ESTIMATED RESULTS

Variables Model 1 Model 2 Model 3 Model 4 Model 5

National Industry Entry Mode 0.328***

（5.97）

Regional Entry Mode

0.408*** （4.67）

Local Industry Entry Mode

0.338***

（12.32）

Business group Entry Mode

0.350*** （19.36）

Business group’s size 0.005

（0.83） 0.007

（1.13） 0.008*** （1.22）

0.010** （1.70）

0.006 （1.31）

Subsidiary Relative size -0.490*** （-3.05）

-0.554*** （-3.56）

-0.423 （-2.67）

-0.416*** （-2.96）

-0.239** （-1.98）

Investment Experience in Mainland China

0.0003 （0.19）

-0.0001 （-0.05）

0.001 （0.46）

-0.001 （-0.78）

-0.001 （-0.46）

Regional Economic Level -0.005

（-0.25） 0.006

（0.28） -0.008 （-0.41)

-0.003 （-0.14）

0.001 （0.05）

Regional Marketization level 0.012 （1.29）

0.002 （0.24）

0.001 （0.13）

0.002 （0.18）

0.004 （0.51）

Reginal Economic Growth Rate 0.387

（1.06） 0.207

（0.58） -0.044

（-0.12） -0.088

（-0.27） -0.129

（-0.47）

Constant 0.799*** （4.50）

0.533*** （3.01）

0.653*** （3.69）

0.655*** （4.22）

0.611*** （4.59）

Adj R2 0.025 0.092 0.067 0.262 0.457

F 3.06 (0.006)

7.91 (0.000)

5.86 (0.000)

25.16 (0.000)

58.23 (0.000)

N 478 478 478 478 478 Note: (I) The value in the regression coefficient’s bracket is t, in F value’s bracket is the probability value; (2)* p < 0.10, ** p < 0.05, *** p < 0.01

Table 3 shows model 1 is the basic model that includes all control variables; model 2-5 iterate

successively institutional variables of different levels. According to the regression results, the regression models are very significant as a whole. In model 2-5, the institutional variables from different levels are all positive, which is very significant (p<0.01). It shows that in different levels of the institutional environment, the higher ratio of the wholly-owned, the higher the ratio of its shareholding. According to the regression results, in different levels of institutional environment, the ratio of shareholding of new entrant has direct correlation to entry modes of different levels of the institutional environment, which means the new entrants has imitated the entry mode of different levels of institutional environments. Therefore, it has analyzed and confirmed hypothesis 1-4. In addition, according to the coefficient of determination, model 2-5 has significant improvement compared with model 1, which indicates that any accession of new institutional variable would improve the degree of fitting. Meanwhile, the improvement is ranked as model 5, model 4, model 2, model 3, which indicates that the constructional forces sequence of different levels of institutional environment are the internal institutional environment of business group, local industry institutional environment, national industry institutional environment and regional institutional environment. Instrumental Variables and Two Stage Least Squares (2SLS) Estimated Results

Since there may be internal problems among national industry entry mode, regional entry mode, local industry entry mode, business group entry mode and ownership entry mode, we test the four institutional variables one by one. To solve the estimated error caused by internal problems, we adopt instrumental variables and 2SLS to estimate the models. The instrumental variables should meet the following two requirements: first, the instrumental variables must be exogenous variables, and second, the instrumental


variables must be related with endogenous variables. For the explanation of the instrumental variables of four institutional variables, please refer to table 4.

TABLE 4 INSTRUMENTAL VARIABLES EXPLANATION

Instrumented Variables Instrumental Variables Instrumental Variables Explanation

National Industry Entry Mode

Local Wage Levela To measure the local labor cost.

Industrial variables Dummy Variables which include 10 industries.

Regional Entry Mode

Local Wage Level To measure the local labor cost.

Regional Variables

Dummy variables which are divided into four areas as Pearl River Delta, Yangtze River Delta , Bohai Rim and other area according to the location change of Taiwanese subsidiaries.

The proportion which foreign invest accounts for of local GDP

Represent the ability to attract foreign investment.



Business group entry mode Business group entry mode in Mainland China.

Business group entry mode


Local Industry Entry Mode The Entry modes of Taiwanese subsidiaries in local industry.

a Logarithm;in yuan.

The reasons for choosing these variables as instrumental variables are below: on one hand, from a visual point of view, there is not a direct relation between these variables and the dependent variable ownership entry mode; while on the other hand, a major motive of Taiwanese groups to invest in Mainland China is to use the cheap workforce (Gao and Chen, 1998). Therefore, the local wage level will affect the entry modes of the institutional environment at different levels. Different industries prefer to use different entry modes (Chan and Makino, 2007, Pan Zhen, Lu Minghong, 2006; Guillen, 2003), accordingly, we believe that industrial variables are the main factor to affect the entry modes. The proportion of foreign capital in GDP represents the attractiveness of the region to foreign capital, and it is an important factor affecting the entry modes of foreign-funded enterprises (Delios and Henisz, 2000). The regional property would affect the choice of entry modes ( Pan and Lu, 2006), as such, we believe that the regional variable is an important factor affecting the regional entry modes. Since there is a correlation between the local industry entry mode and the business group entry mode, we set them as each other’s instrumental variables. Table 5 provides the instrumental variables and 2SLS estimated results.

The occurrence of endogeneity and instrumental variables needs stringent tests. For this reason, we adopt Hausman test and Sargan test to check the endogeneity of the model and the availability of the instrumental variables respectively. Models 1-3 in table 5 have passed the test of Hausman and Sargan, which means that models 1-3 have an endogenetic problem, and their instrumental variables are effective. Model 4 has not passed the Hausman test, which means that model 4 has no endogenetic problem.

According to instrumental variables and 2SLS estimated results, the institutional variables coefficients in models 1-4 are obviously positive, and are in accordance with theoretical expectations and OLS estimated results. Therefore, according to empirical analysis, we tested impacts of four institutional variables on new entrants of subsidiaries, and find out that in the institutional environment of different levels, there is a imitation of the new entrants of subsidiaries in their choice of entry modes.


TABLE 5 INSTRUMENTAL VARIABLES 2SLS ESTIMATED RESULTS

Variables Model 1 Model 2 Model 3 Model 4

National Industry Entry Mode 0.336*** (6.11)

Regional Entry Mode 0.280** （2.45）

Local Industry Entry Mode 0.664*** (14.58)

Business group Entry Mode 0.376*** (14.31)


（1.14） 0.007 (1.09)

0.014** (2.12)

0.007 (1.33)

Subsidiary Relative size -0.556*** （-3.57）

-0.444*** (-2.79)

-0.344** (-2.15)

-0.220* (-1.80)


-0.0001 （-0.06）

0.001 (0.38)

-0.003 (-1.53)

-0.001 (-0.52)

Regional Economic Level 0.006

（0.29） -0.007 (-0.36)

-0.0001 (-0.00)

0.001 (0.07)

Regional Marketization level 0.002

（0.21） 0.005

（0.48） -0.009 (-0.93)

0.003 (0.41)

Reginal Economic Growth Rate 0.203 （0.57）

0.091 （0.24）

-0.545 (-1.48)

-0.168 (-0.61)

Constant 0.527*** （2.97）

0.699*** （3.90）

0.515*** (2.90)

0.597*** (4.46)

Adj R2 0.092 0.062 0.040 0.454

F 8.14 (0.000)

3.59 (0.001)

33.02 (0.000)

33.94 (0.000)

N 478 478 478 478

Sargan N*R-sq test 2.761 (0.973)

3.400 (0.493)

0.010 (0.919)

0.565 (0.452)

Wu-Hausman F test 39.38 (0.000)

3.075 (0.080)

171.50 (0.000)

1.981 (0.160)

Note:(1) The value in brackets of regression coefficients are t and f; the value in brackets of Sargan and Hausman are probability value; (2) )* p < 0.10, ** p < 0.05, *** p < 0.01

Robustness Test

In this sample, the subsidiaries whose shareholder ratio is 100% account for 79.9% and the subsidiaries whose shareholder ratio is less than 100% account for 20.1%. This data distribution might affect the regression result; therefore, we have a robustness test for the regressive models.

On the one hand, we defined the dependent variable as a dummy variable, which represents Taiwanese subsidiary’s entry mode, coded 1 for a wholly owned subsidiary and 0 for a joint venture. Logistic regression analyses were conducted to examine. Table 6 provides the logistic regression results. The result shows that the mode 1-5 are significant, the sign and significant of the institutional variables are consisted with the OLS regression.

On the other hand, we use the bootstrap to test the robustness of the estimated results of OLS and instrumental variables and 2SLS. Bootstrap produced new samples from the original samples, which can reduce the statistic deductive variation and provide a more accurate and reliable test, depending on the marginal value produced by the data itself. The results of bootstrap test are in accordance with the estimated results of OLS and instrumental variables and 2SLS. The models that include institutional variables are all significant. Appendix 1 and appendix 2 provide the results. Although in the bootstrap test the standard error of institutional variable increased, it does not affect the significance of the institutional variables, which means that the research results under OLS are robust.


TABLE 6 ROBUSTNESS TEST: LOGISTIC REGRESSION RESULTS


National Industry Entry Mode

5.497*** （5.78）

Regional Entry Mode 6.179*** （4.06）


5.76*** （8.17）


6.977*** （10.08）

Business group’s size 0.021 (0.18)

0.025 （0.21）

0.061 （0.52）

0.102 （0.75）

-0.050 （-0.28）

Subsidiary Relative size -2.365 (-0.96)

-3.734 （-1.47）

-1.570 （-0.58）

-2.833 （-0.82）

3.210 （0.81）


0.074** (2.36)

0.068** （2.08）

0.084*** （2.61）

0.055 （1.51）

0.081* （1.78）

Regional Economic Level -0.294 (-0.80)

-0.066 （-0.17）

-0.387 （-1.03）

-0.274 （-0.61）

0.069 （0.12）

Regional Marketization level 0.343** (2.15)

0.164 （0.96）

0.173 （1.04）

0.195 （1.01）

0.154 （0.65）

Reginal Economic Growth Rate 9.027 (1.48)

5.141 （0.83）

-1.060 （-0.17）

-0.129 （-0.02）

0.382 （0.04）

Constant term -0.501 (-0.17)

-4.887 （-1.53）

-1.610 （-0.53）

-2.652 （-0.73）

-5.879 （-1.21）

pseudo R2 0.040 0.111 0.076 0.263 0.474 ll_0 -239.7 -239.7 -239.7 -239.7 -239.7 ll -230.3 -213.2 -221.5 -176.6 -126.2 chi2 19.0 53.0 36.5 126.3 227.1 chi2 - 34*** 17.5** 107.3*** 208.1*** N 478 478 478 478 478 Note: (1)The dependent variable is dummy variable, wholly owned subsidiary = 1, joint venture = 0; (2) The bracket in coefficient is z; (3) chi2 equals to chi2 of model 2-5 minus chi2 of mode 1; (4) * p < 0.10, ** p < 0.05, *** p < 0.01.

CONCLUSION AND DISCUSSION

The traditional theories explanation for entry modes focused on economical factors, but ignored the impact of the institutional environment. This research has taken Taiwanese subsidiary ownership entry modes as a social identity and has divided the institutional environment into four levels (national industry, regional, local industry and business group). This research has studied Taiwanese subsidiary choice of entry mode from the perspective of seeking legitimacy. It is a major supplementary explanation for the traditional theory. Study shows that there is imitation among Taiwanese subsidiaries in the choice of entry modes in different levels of the institutional environments in Mainland China. As an important legal mechanism, imitation is a vital approach for Taiwanese subsidiaries to reduce uncertainty and to gain legitimacy in different levels of the institutional environment in Mainland China.

The later entrants of subsidiaries will imitate the entry mode of the earlier entrants in the same business group, which means that the previous entry mode is a "legitimate mode", and the later entrants take it for granted. It also indicates that there is a strong inertia in the business group in the decision of entry mode for investment in Mainland China. Compared with other levels of the institutional environment, the internal institutional environment of the business group has more reasonable interpretation power for entry modes of its subsidiaries. One possible explanation is that the subsidiary is a member of the group, and its greatest pressure is from the parent company's internal institutional


environment, to keep the unity within the whole group, the subsidiary has a strong demand to keep internal legitimacy, thus, the subsidiaries prefer to imitate the entry mode of its own group.

The recent studies about imitation in entry modes (Chan & Makino, 2007; Lu, 2002; Guillen, 2002, 2003; Liu & Makino, 2002) are under the transnational environment, scholars address that the industrial institutional environment of the host country has an important impact on the entry mode. However, it assumes that the institutional environment in different region of the same country is the same and does not consider the impact of the regional institutional environment on the entry mode. In fact, the regional institutional environments are different such as in Mainland China or America (Krugman, 1991, Chan CM, Makino S and Isobe T, 2010). Therefore, the impact on the entry mode of the regional institutional environment should be considered. Our research shows that the institutional environments which Taiwanese subsidiary in Mainland China face are different because of the differentiation of industries and regions. When Taiwanese subsidiary enter Mainland China, they take into consideration the pressure of the industry institutional environment and regional institutional environment, imitating the entry modes of the previously entered Taiwan subsidiaries to gain legitimacy.

The local industry entry mode as the interaction terms between regional entry mode and national industry entry mode, has positive effects on the dependent variables, which indicates that the legitimate requirement of industries in different areas are different. Compared with the regional institutional environment and the nationally industrial environment, the local industry institutional environment has a stronger explanation for entry mode. One possible explanation is that Taiwanese subsidiaries are more likely to pursue legitimacy in a narrow institutional environment. In other words, Taiwanese subsidiaries prefer to seek for legitimacy in local industry institutional environment. This is in accordance with the existing point of view, for example, transnational companies tend to seek legitimacy in a narrow institutional environment (such as a local industry) but not in a wider institutional environment (such as a host country). In research of global and local imitation caused by strategic alliances, companies are more likely to imitate the behavior of the companies who are in the same strategic position (Niche), but not to imitate other companies behavior in the same industry in the world (Garcia-Pont and Nohria, 2002). It indicates that the restrain of the local industry institutional environment to Taiwan subsidiary is more important and direct than that by the national industry institutional environment and regional institutional environment.

Our study has two limitations that suggest some intriguing avenues for future theoretical and empirical refinement. Firstly, the result of research indicates that there is a imitation among Taiwan subsidiaries to pursue legitimacy, however, in addition to the imitative mechanism, there are still regulative and normative mechanisms to pursue legitimacy (Dimaggio and Powell, 1983). Therefore, for future research, the direction is to combine the regulative, normative and imitative mechanism to study entry modes of Taiwanese subsidiaries. Secondly, the samples in this research are limited to Taiwanese subsidiaries. It needs further studies to confirm whether these research results are applicable to other foreign-funded companies. ENDNOTES Author: Shaodong Hu is the corresponding author and can be contacted at :[email protected] [2]Taiwan is part of China, however, the economic system between Mainland China and Taiwan is different, as such, the scholars treat Taiwan direct investment in Mainland China as part of FDI. ACKNOWLEDGMENTS This paper is supported by “Guangdong Province Social Sciences Foundation (08E-15)”, “Major Project of the Key Research Base of Humanities and Social Sciences in Guangdong Province (11JDXM63005)” and “Academic Innovation Team Construction Foundation of Shantou University (ITC10004)”. We are grateful for comments and discussions to Robert Guang Tian.


REFERENCES Agarwal, S., Ramaswami, S.N. (1992). Choice of Foreign Market Entry Mode: Impact of Ownership, Location and Internalization Factors. Journal of International Business Studies，23, (1), 1–28. Bowman, E.H., Helfat, C.E. (2001). Does Corporate Strategy Matter?.Strategic Management Journal, 22, (1), 1–23. Chan, C.M., Makino, S. (2007). Legitimacy and Multi-level Institutional Environments：Implications for Foreign Subsidiary Ownership Structure. Journal of International Business Studies，38, 621-638. Chan, C.M., Makino, S. & Isobe, T. (2010). Does Subnational Region Matter?Foreign Affiliate Performance in the United States and China. Strategic Management Journal, 31, 1226-1243. Contractor, F.J. (1990). Ownership Patterns of U.S. Joint Ventures Abroad and the Liberalization Of Foreign Government Regulations in the 1980s：Evidence from the Benchmark Surveys. Journal of International Business Studies, 21, (1), 55-73. Davis, P.S., Desai, A.B., Francis, J.D. (2000). Mode of International Entry: An Isomorphism Perspective. Journal of International Business Studies, 31, (2), 239–258. Delios, A., Beamish, P.W. (1999). Ownership Strategy of Japanese Firms: Transactional, Institutional, and Experience Influences. Strategic Management Journal, 20, (10), 915–933. Delios, A., Henisz, W.J. (2000). Japanese Firms’ Investment Strategies in Emerging Economies. Academy of Management Journal, 43, (3), 305-323. Dimaggio, P.J., Powell, W.W. (1983). The Iron Cage Revisited：Institutional Isomorphism and Collective Rationality in Organizational Fields. American Sociological Review, 48, April, 147–160. Dunning, J.H. (1980). Toward an Eclectic Theory Of International Production：Some Empirical Tests. Journal of International Business Studies, 11, Spring/Summer, 9-31. Fan, G., Wang, X.L. & Zhun, H.P. (2007). NERI INDEX Of Marketization of China’s Province 2009 Report. Peking: Economic Science Press. Gao, C., Chen, W.R. (1998). An Empirical Study on Ownership Entry Mode of Taiwanese Enterprises in Mainland China. Journal Of Management (In Chinese),15, (3), 393-418. Garcia-Pont C, Nohria N (2002). Local Versus Global Mimetism: The Dynamics of Alliance Formation in the Automobile Industry. Strategic Management Journal, 23, (4), 307-321. Guillen, M.F. (2002). Structural Inertia, Imitation, and Foreign Expansion: South Korean Firms and Business Groups in China, 1987-95. Academy of Management Journal, 45, (3), 509-525. Guillen, M.F. (2003). Experience，Imitation，and the Sequence of Foreign Entry：Wholly Owned and Joint-Venture Manufacturing by South Korean Firms and Business Groups in China，1987-1995. Journal of International Business Studies, 34, 185-198.


Haveman, H.A. (1993). Follow the Leader: Mimetic Isomorphism and Entry into New Markets. Administrative Science Quarterly, 38, (4), 593-627. Henisz, W.J., Delios, A. (2001). Uncertainty, Imitation, and Plant Location: Japanese Multinational Corporations, 1990-1996. Administrative Science Quarterly, 46, (3), 443-475. Kogut, B., Zander, U. (1993). Knowledge of the Firm and the Evolutionary Theory of the Multinational Corporation. Journal of International Business Studies, 19, (3), 411-432. Kostova, T., Zaheer, S. (1999). Organizational Legitimacy under Conditions of Complexity: The Case of the Multinational Enterprise. Academy of Management Review, 24, (1), 64–81. Krugman, P. (1991). Geography and Trade. MIT Press：Cambidge, MA. Lu, J.W. (2002). Intra- and Inter-Organizational Imitative Behavior: Institutional Influences on Japanese Firms’ Entry Mode Choice. Journal of International Business Studies, 33, (1), 19–37. Meyer, M.W. ( 2008). China’s Second Economic Transition：Building National Markets. Management and Organization Review, (4), 13–15. Meyer, K.E. (2001). Institutions, Transaction Costs and Entry Mode Choice in Eastern Europe. Journal of International Business Studies, 32, (2), 357-367. Pan, Z., Lu, M.H. (2006). The Cultural Interpretation to the Entry-Mode Choice of Foreign Direct Investment in China.The Journal of World Economy(In Chinese), (2), 51-61. Peng, M.W. (2009). Perspectives—from China Strategy to Global Strategy, Journal of Strategic Management (In Chinese), 1, (1), 1-13. Pei, C.H. (2006). Absorbing Foreign Direct Investment and Upgrading Industrial Structure--Pondering over the Target of Utilization of Foreign Capital Policy. China Industrial Economy(In Chinese), (1), 33-39. Shan, W. (1991). Environmental Risks and Joint Venture Sharing Arrangements. Journal of International Business Studies, 22, (4), 555-578. Suchman, M.C .(1995). Managing Legitimacy: Strategic and Institutional Approaches. Academy of Management Review, 20, (3), 571–610. Tversky, A., Kahneman, D. (1974). Judgment under Uncertainty: Heuristics and Biases. Science ,185, (4157), 1124–1131. White, H.C. (1981).Where Do Markets Come From?. American Journal of Sociology, 87, (3), 517-547. Williamson, O.E. (1975). Markets and Hierarchies: Analysis and Antitrust Implications. NY: Free Press. Yin, H.F., Pan, Z. & Lu, M.H. (2007). Central - Local Government Relations and Policy Execution: The Example of Foreign Investment Industrial Policies. Management World(In Chinese), (7), 22-36.


Yiu, D., Makino, S. (2002). The Choice between Joint Venture and Wholly Owned Subsidiary：An Institutional Perspective,.Organization Science, 13, (6), 667-683. Zhao, J.P. (2002). The Study of the Industrial Policy for Foreign Investment in China. Management World(In Chinese), (9), 47-52. APPENDIX

APPENDIX 1 ROBUSTNESS TEST: BOOTSTRAP OF OLS


National Industry Entry Mode 0.328***

（4.73）

Regional Entry Mode

0.408*** （4.00）

Local Industry Entry Mode 0.338***

（9.84）


0.350*** （14.36）


（0.78） 0.007

（1.07） 0.008

（1.17） 0.010*

（1.83） 0.006*

（1.62）

Subsidiary Relative size -0.490 （-1.41）

-0.554 （-1.59）

-0.423 （-1.38）

-0.416* （-1.70）

-0.239 （-0.97）


0.0003 （0.22）

-0.0001 （-0.05）

0.001 （0.58）

-0.001 （-0.84）

-0.001 （-0.46）

Regional Economic Level -0.005

（-0.24） 0.006

（0.32） -0.008 （-0.41)

-0.003 （-0.15）

0.001 （0.05）

Regional Marketization level 0.012 （1.32）

0.002 （0.27）

0.001 （0.12）

0.002 （0.18）

0.004 （0.47）

Reginal Economic Growth Rate 0.387

（0.82） 0.207

（0.56） -0.044

（-0.10） -0.088

（-0.26） -0.129

（-0.41）

Constant term 0.799*** （4.13）

0.533*** （3.23）

0.653*** （3.44）

0.655*** （3.95）

0.611*** （4.57）

Adj R2 0.025 0.092 0.067 0.262 0.457

Wald chi2 8.92 (0.178)

32.41 (0.000)

25.19 (0.001)

109.22 (0.000)

236.04 (0.000)

Replications 150 150 150 150 150 N 478 478 478 478 478 Note：(1)The value in the regression coefficient’s bracket is z,the value in brackets of Wald chi2 is pobability value. (2)* p < 0.10, ** p < 0.05, *** p < 0.01。


APPENDIX 2 ROBUSTNESS TEST: BOOTSTRAP OF INSTRUMENTAL VARIABLES AND 2SLS

Variables Model 1 Model 2 Model 3 Model 4

National Industry Entry Mode 0.336*** (4.41)

Regional Entry Mode 0.280** （2.13）

Local Industry Entry Mode 0.664*** (8.61)

Business group Entry Mode 0.376*** (9.49)


（1.21） 0.007 (1.06)

0.014** (2.05)

0.007 (1.47)

Subsidiary Relative size -0.556* （-1.65）

-0.444 (-1.46)

-0.344* (-1.75)

-0.220 (-0.83)


-0.0001 （-0.06）

0.001 (0.41)

-0.003 (-1.59)

-0.001 (-0.56)

Regional Economic Level 0.006

（0.32） -0.007 (-0.39)

-0.0001 (-0.00)

0.001 (0.08)

Regional Marketization level 0.002

（0.20） 0.005

（0.48） -0.009 (-0.92)

0.003 (0.41)

Reginal Economic Growth Rate 0.203 （0.50）

0.091 （0.21）

-0.545 (-1.15)

-0.168 (-0.55)

Constant term 0.527*** （2.88）

0.699*** （3.70）

0.515*** (2.86)

0.597*** (4.58)

Adj R2 0.092 0.062 0.040 0.454

Wald chi2 26.83 (0.000)

13.62 (0.058)

89.95 (0.000)

101.89 (0.000)

Replications 200 300 200 200 N 478 478 478 478 Note：(1) The value in the regression coefficient’s bracket is z, the value in brackets of Wald chi2 is pobability value. (2)* p < 0.10, ** p < 0.05, *** p < 0.01


Great Recession of 2008-2009: Causes and Consequences

Alexander Katkov Johnson & Wales University

The last four years of the turbulent economic performance raised more questions than provided answers about causes of the Great Recession of 2008-2009. The real GDP has declined about 3.7% and the full recovery has been achieved only recently. Most economists are blaming the real estate market collapse and the followed it financial crisis as main causes of the Recession. But the real macroeconomic cause was the change of the national macroeconomic policy from the Demand support to Supply support strategy in the middle of 1980s. This article is discussing the cause of the Recession and its consequences. INTRODUCTION

Most economists are considering the destructive effect of the subprime mortgages and CDOs on the development of the housing market and the followed financial crisis as the major cause of the recession of 2008-2009. This point of view has been summarized in the Report of Financial Crisis Inquiry Commission presented to the U.S. Congress in January of 2011:” We conclude collapsing mortgage-lending standards and the mortgage securitization pipeline lit and spread the flame of contagion and crisis”( The Financial Crisis Inquiry Report, 2011, p. xxiii). The recession that has emerged out of that financial crisis forced the U.S. Government to use both mechanisms of the government regulation: fiscal and monetary policies at the very large scale.

Economic stimulus packages, corporate and banking bailouts, monetary easing at the multitrillion dollars scale has created both supporters and opponents. Some, like James Galbraith (Galbraith, 2008), Thomas Palley (Palley, 2010), Paul Krugman (Krugman, 2009) are criticizing the U.S. Government for the insufficient efforts to use Keynesian economic remedies to ease the recession. Others, like Bruce Yandle (Yandle, 2010), Jeffrey Hummel (Hummel, 2011), Randy Simmons (Simmons, 2011) are strongly against of further governmental regulation attempts and in support of that point of view are discussing some common features between the recent monetary and fiscal policies actions of the U.S. Government and the Soviet command economy practice.

I agree with the opinion that the real estate market collapse and the financial crisis were important causes the recession. But from my point of view they were consequences of the main macroeconomic cause and have played the trigger role. The real macroeconomic cause of the Great Recession was the change of the national macroeconomic policy from the Demand support to Supply support strategy that took place in the middle of 1980s. This change in the economic policy has forced the modification of the structure of the U.S. economy with the substantial diminution of the national manufacturing sector. As the result there were not only big loses in the number of manufacturing jobs but also the strategic change the country’s role in the international division of labor process that make the USA the negative net exporter


of cheap consumer goods from developing countries and expensive energy resources from oil rich countries.

I have expressed this opinion before (Katkov, 2011). The goal of the given article is to provide the broader argumentation in support of the point of view that the cause of the Great Recession was the strategic mistake in changing the core economic policy made by the U.S. Government in 1980-s. The slow recovery from the recession as I will show in this article is the result of the structural changes in the U.S. economy and also the result of some mistakes made by the Obama’s administration in attempts to ease the recessionary pressure and to help the U.S. economy to recover.

In other words, from my point of view the recession was the result of the switch from Keynesian demand support economics to neo-liberal supply support economics, and the slow recovery is the result of the inconsistent application of instruments of fiscal policy offered by Keynesian economics. CHANGE IN ECONOMIC POLICY

The economic strategy before 1980-s has been built on the ideas of the full employment and the wages growth as the result of the growth of productivity. The full employment and the wages growth have stimulated the increase in the demand for goods and services. That growth has influenced the increase in the supply and has stimulated the businesses to invest into new technologies facilitating the further growth of productivity. The “stagflation” of 1970-s seriously damaged this strategy. In 1980-s the Keynesian model of the economic growth which is based on the support of the growth of Aggregate Demand has been replaced by the “neo-liberal” model of economic growth based upon the support of the growth the Aggregate Supply. Increase in demand as it follow from the classical market model supposed to increase the prices of products and services when increase in supply expected to decrease prices. Price decreases are good for consumers because their real incomes as the result of prices decreases will rise. But for producers price decreases means profits decline. To keep profitability at the same level or to increase profitability businesses should decrease costs of production.

One possible way to decrease costs of production would be in the productivity growth. The increase in productivity will drop the cost per unit of the product manufactured. The backbone of the higher productivity is the new technology. The development of new technology requests high capital investments and an increase of fixed costs. But there is the second possibility: costs of production could be decreased if the costs components, both fixed and variable, will decline. Unfortunately, that option has met very serious constraints in the USA in 1980-s because of the steady wages growth and the increase in the costs of domestic natural resources. At the same time the process of globalization of the world economy has offered another opportunity: to decrease costs manufacturing facilities can be moved closer to sources of less expensive resources, both natural and labor resources. As the result, majority of large and medium size U.S. corporations began the transfer of significant part of their manufacturing facilities abroad, mostly into the developing countries.

This process of off-shoring caused the critical changes in the U.S. economy’s manufacturing sector and wiped out millions of manufacturing jobs. Traditionally, the manufacturing sector comprises business entities engaged in the process of mechanical, physical, or chemical transformation of materials, substances, or components into new products or component parts of manufactured products.

According to North American Industry Classification System (NAICS) manufacturing jobs categories are not well defined. Jobs are considered manufacturing jobs if they are involved with production of new products from raw materials or from components by transforming them into new products. These jobs are creating products not services. The assembling of component parts of manufactured products is considered manufacturing, except in cases where the activity is appropriately classified as construction (U.S. Census, 2007). It is accepted that the establishments in the manufacturing sector are engaged in the transformation of materials into new products. But some jobs performing for example fish processing, water bottling, milk pasteurization and printing and related activities are considered manufacturing jobs. At the same time according to the same classification manufacturing jobs are not involved in book


publishing, logging, mining or construction, even though products are created by these jobs (U.S. Census, 2007).

As the result of the manufacturing jobs outsourcing abroad number of people employed by the manufacturing sector of the national economy during the period of the December of 1979 through December of 2007 (the last year before the Great Recession) decreased from 19.4 million to 13.9 million people – 5.5 million jobs have been lost. At the same time number of people in sales increased during the same period of time from 10.2 to 15.4 million – 5.2 million increase. The number of people employed by the financial sector during the same period almost doubled: increase from 4.8 to 8.4 million (Economic Report of the President, 2008, Table B-46). The national economy that has fewer manufacturers and more sales people should face problems during the economic contraction phase of the business cycle. When there is less produced there is no need in so many sales people.

The decline in number of manufacturing jobs influenced the negative trickledown effect on the process of the jobs creation in other sectors of the national economy and as the result on the process of the creation of the disposable income and the Aggregate Demand.

Table 1 and Table 2 presented bellow shows the changes in some categories of jobs: manufacturing jobs, capital goods producing jobs, service jobs, finances jobs and government jobs since 1961 through 2010. The classification of “Goods Producing Jobs” in Table 1 and Table 2 includes manufacturing jobs described above plus jobs in logging, mining and construction.

TABLE 1 ABSOLUTE CHANGES IN DIFFERENT CATEGORIES OF JOBS IN THE USA 1961-2010

Year Labor Force Manufacturing Goods Prod. Services Finances Government 1961 70,459 15,011 18,647 35,458 2,590 8,706 1970 82,771 17,848 22,179 48,827 3,532 12,687 1980 106,940 18,773 24,263 66,265 5,025 16,375 1990 125,840 17,695 23,723 85,764 6,614 18,415 2000 142,583 17,263 24,649 107,136 7,687 20,790 2007 153,124 13,879 22,233 115,366 8,301 22,218 2010 153,889 11,524 17,755 112,064 7,630 22,482 Source: U.S. Bureau of Labor Statistics, 2011

TABLE 2 CHANGES IN DIFFERENT CATEGORIES OF JOBS AS

PERCENTAGE OF LABOR FORCE

Year Labor Force Manufacturing Goods prod. Services Finances Government 1961 100% 21.3 % 26.5% 50.3% 3.7% 12.4% 1970 100% 21.6% 26.8% 59.0% 4.3% 15.3% 1980 100% 17.5% 22.7% 62.0% 4.7% 15.3% 1990 100% 14.1% 18.9% 68.2% 5.3% 14.6% 2000 100% 12.1% 17.3% 75.1% 5.4% 14.6% 2007 100% 9.1% 14.5% 75.3% 5.4% 14.5% 2010 100% 7.5% 11.5% 72.8% 5.0% 14.6% Source: U.S. Bureau of Labor Statistics, 2011

Table 1 shows the trend of the sharp decreases in numbers of manufacturing and goods producing jobs and Table 2 shows their percentages of the civilian labor force from 1980 through 2007 (- 4.85 million and -2.03 million respectively), and the increases of numbers service jobs total, finances jobs and government jobs(+49.10 million, +3.28 million and +5.84 million respectively). Interestingly to mention


the fact that from 2007 through 2010 (the recession and post-recession years) number of jobs in all above mentioned groups declined except the government jobs. Number of government jobs during this period increased by 264 thousands and became in 2010 almost twice greater than number of manufacturing jobs: 22.48 million versus 11.52 million.

What effect the decline in number of manufacturing jobs had on the entire jobs market? Prior to the introduction of the Supply Side economic strategy in 1980-s all categories of jobs included into Table 1 showed the growth. So, we can conclude that “reaganomics” has motivated or may be even forced American manufacturers to off-shore their manufacturing facilities abroad and to outsource manufacturing jobs to other countries. Manufacturing jobs were always have been the leading factor in other jobs creation. As Josh Bivens has shown in the Economic Policy Institute’s Working Paper #268 “Updated Employment Multipliers for the U.S. Economy (2003)” that 100 jobs in manufacturing sector supports 291 jobs elsewhere in the economy which is more than three times large than retail trade (88), about 2.5 times as large as health services (117), and about twice more than business services (154). In his research Bivens considered the job creation process as the result of three categories of effects: supplier effects, respending effects, and government employment effects. Manufacturing industries are using supplies of materials and parts. Those suppliers are hiring or laying-off people depending of the manufacturers demand for those materials and parts. Employees of manufacturing sector are spending their disposable incomes buying goods and services offered by other industries creating jobs in the apparel, housing, food and etc. industries. Taxes that manufacturing industries’ employees are paying to federal, state and local governments are supporting and creating the government jobs. I think that the manufacturing employment multiplier is even higher than Bivens has calculated because for many manufacturing industries like motor vehicle, machinery manufacturing, electric equipment manufacturing and others machine building industries, because every machine also need except sales people also maintenance and repair specialists. Unfortunately Bivens did not discuss the influence of that existing” maintenance effect” on job creation. Anyway, the fact that 100 jobs in manufacturing sector are responsible for the creation of 291 jobs in other sectors is very impressive.

Another Bivens’s (Bivens, 2003) conclusion is even more important from the perspectives of the analysis of the Graph 1 presented below. In one of his tables he showed the government employment support by different sectors of the national economy. The rough generalization of these data is showing that 100 jobs in manufacturing industries in average supporting approximately the employment of 8.5 employees in the government sector, and 100 jobs in service providing industries are supporting about 4 employees in the government sector. That means that in 2007 before the recession 115.4 million services jobs total could assist in creation of about 4.6 million jobs in the government sector, and 13.9 million manufacturing jobs could support about 1.2 million government jobs. So, total 85% of all labor force in 2007 could support only about 5.8- 6 million jobs in the government sector. It would be logically correct to ask question who is supporting the other 16 million of government jobs?

As Stephen Ezell and Robert Atkinson analysts of the Washington D.C. based think tank “The Information Technology and Innovation Foundation” showed in their research paper “The Case for National Manufacturing Strategy” published in April 2011 that high-tech manufacturing industries have even greater multipliers. Electronic computer manufacturing has a multiplier effect of 16 jobs, meaning 15 other jobs are dependent on one job created in that industry. So the decline in number of manufacturing jobs also means the decline in jobs that would be created if the manufacturing jobs remains in the USA. All of those have been playing the great impact on the U.S. employment situation during the Great Recession. Ezell and Atkinson (Ezell and Atkinson, 2011) also showed in their paper that processes of off-shoring manufacturing enterprises and outsourcing of manufacturing jobs abroad have created the negative affect on the economic growth. They showed that manufacturing growth has lagged overall economic growth, and that the majority of U.S. manufacturing sectors have seen absolute declines in real output over the past decade. They expressed the opinion that the apparent growth in manufacturing output showed by the official statistics is the result of the overinflated estimates of output from two industries—the computer and electronics industry and the petroleum industry.


Manufacturing in fifteen manufacturing sectors that made up 79 percent of U.S. manufacturing according to Ezell and Atkinson (Ezell and Atkinson, 2011) is lagging behind other industries in terms of its rate of economic growth. From 2000 to 2009 when the overall U.S. real GDP increased 15% the total manufacturing sector realized only a 5% growth in real-value-added. It is happened because most of manufacturing sectors have showed during that 10 years period the absolute declines in the output produced ranging from 2% decline in electric equipment manufacturing, 3% in chemicals manufacturing, 14% in machinery manufacturing, 18% in motor vehicles manufacturing, 27% in fabricated metals manufacturing, 28% nonmetallic minerals and primary metals manufacturing to astonishing declines by more than 40 % in manufacturing of apparel, furniture and textile.

Table 1 shows the decline of number of jobs in manufacturing, goods production, in services total and even in financial services during the Great Recession but it shows the growth of government jobs. As we just conclude this growth of government jobs is contradicting the logic of the normal economic development. But it is the only one problem. Another problem of the uncontrolled growth of government jobs when manufacturing jobs are declining is the inevitable decline in the GDP growth as the consequence.

The loss of about the quarter of the labor force in the manufacturing sector probably as official statistics shows (Bureau of Economic Analysis, 2011) has been compensated by the increase in productivity (computer revolution). But the link between the growth of the productivity and the growth of wages and as the result the growth of the consumption has been broken. Growing supply needs demand to grow also. But where are means to finance the growth in consumption if there are less people who are creating new products and more people who are servicing the process of the distribution and inevitably adding into the products cost and the final price? It is a serious problem.

The solution of this problem has been found in the developing of the housing market. The house is the largest and the most valuable asset for most American households. If home’s market value grows the owner can cash out the appreciation of the house value by borrowing this amount from the bank. So, the housing market frenzy became the major source of the finance for the consumption growth in 1990s and the first half of 2000s. That growth in consumption has supported the growth of supply and respectfully the economic growth especially in 2000-2006.

As I showed in the article “An Analysis of the Government Policy to Ease the Recessionary Pressure of 2008-2009” (Katkov, 2011) the development of the housing market as the vehicle of the Aggregate Demand and economic growth faced the problem of the slow growth of net income per capita for most households. This growth of the net income per capita in the USA during the period of 1990-2000 was about 25%. Income has increased (in 2000 dollars) from $20,336 to $25,472. (Census 2000. Demographic Profile Highlights, 2011).

But these data shows the average income growth per capita. Starting the middle of 1980-s there was the emerging trend of the faster income growth among the households belonging to the high income earning group. According to the US Census Bureau, in 2001 40 percent of low income households earned only 12.2% of the total income earned by all households or $21,639 per household. At the same time 20 percent of highest income earning group of households earned 50.2% of the total income. Obviously, this level of income is not sufficient to buy a house in many states. So, the change in the standards of the mortgage issuance according to S. Leibowitz (Leibowitz, 2008) became the Government’s leading strategy.

As the result of the easiness in the mortgage requests approvals the percentage of the home owners in the USA has increased from the year of 2000 to the year of 2004 from 66.2% to 69%. At the same time lower standards made possible to get financing also for people who want to buy a larger house, or to purchase the second home for the vacation purposes, or the house for the future retirement. The quick increase in the demand for houses created their fast price appreciation. During the ten year period from 1995 till 2005 homes prices in constant dollars almost doubled.

But borrowed money sooner or later should be paid back. The households’ debt starts rapidly expand. If in 1981 the household debt was equal about 48 % of the national GDP: debt - $1,507.2 billion versus GDP - $3,128.4 billion, in 2007 these two numbers became practically equal: debt - $13,765.1 billion


versus GDP - $13,841.3 billion. It means that taking the inflation into the consideration the household debt has been grown 5.7 times in 26 years. This debt growth forced the saving rate to drop from 10% in 1980 to 0.6% in 2007 (Economic Report of the President 2008, Table B-30).

So, during the last 30 years the household consumption – the largest component of GDP (about 70%) has been grown as the result of the increase of the rate of debt, not the rate of the income growth. To borrow money you need to provide the collateral. In the situation when for most households their house was their collateral the growing house prices in 1998-2006 allowed households to borrow more and more practically till the moment when the bubble has burst.

So, we can conclude that the support of the Aggregate Demand growth to match the growth of Aggregate Supply through the development and the expansion of the households’ debt financing scheme became the logical result of the acceptance of the Supply Side economic model as the core idea of the new economic policy 30 years ago. This change in the economic policy became from my point of view became the first macroeconomic cause of the Great Recession of 2008-2009. The second cause was the acceptance of the new U.S. role in the global division of labor. THE NEW ROLE IN THE GLOBAL DIVISION OF LABOR

Since World War II the USA became the leading producer and the exporter of capital goods and the leading supplier of the technology and investments abroad. The deficit of trade balance on capital goods and automotive vehicles, parts and engines goods that first time occurred in 1984 ( - 17.5 billion USD) since that time has been grown substantially to -50.0 billion USD in 1995, to -103.7 in 2000 and has reached astonishing number of -161.2 billion USD in 2005.

During last five years this deficit starts to diminish and declined in 2009 to the level of -55.1 billion USD. (Economic Report of the President (2011), Table B-104). Today the USA is the largest global importer of cheap consumer goods: 534.1 billion USD in 2008 and 502.3 billion USD in 2009 (Economic Report of the President (2011), Table B-106.). These imports are not only helped economies of countries like China and Mexico to gain a new economic power to build their economies by using American investments but also created at least three factors which negatively affected the U.S. national economy’s growth during the last 20-25 years: 1) a substantial loss of the internal money flow to other countries for imports; 2) a loss of jobs initially in manufacturing and later also in the services providing sector; and 3) a loss of investments for the domestic economy because of the growing investments abroad. Some statistical data can illustrate above mentioned statements.

TABLE 3 TRADE BALANCES IN GOODS BETWEEN THE USA AND CHINA IN BILLIONS OF USD

1999 2002 2003 2004 2005 2006 2007 2008 2009 2010 --68.7 -103.1 -124.1 -161.9 -201.5 -234.1 -256.2 -268.0 -226.9 -273.1 Source: U.S. Census Bureau.

The deficit in trade between the USA and China, as Table 3 shows, has been grown from $68.7 billion in 1999 to $273.1 billion in 2010. Combined deficit over the period of 12 years on nominal basis equal $2,084.8 billion (U.S. Census Bureau, 2011).

To show how the USA supported the growth of Chinese economy I am offering Table 4. This table is providing the information about the volume of U.S. imports (M) in billions of USD and also as the percent of China’s GDP for the period of 11 years from 2000 through 2010 (U.S. Census Bureau, 2011). American imports of Chinese products became the very important factor of the Chinese economic growth. Consumer expenditures are producers’ incomes. The growing National Income means the growing GDP.


TABLE 4 U.S. IMPORTS OF GOODS FROM CHINA (M) IN BILLIONS OF USD AND AS PERCENTS OF

CHINA’S NOMINAL GDP IN BILLIONS OF USD AT OFFICIAL EXCHANGE RATES

Year 2000 2003 2004 2005 2006 2007 2008 2009 2010 US M 100.0 152.4 196.7 243.5 287.8 321.4 337.8 296.4 364.9 GDP 1,200 1,640 1,920 2,230 2,670 3,380 4,590 5,030 5,710 Percent 8.8% 9.3% 10.3% 10.9% 10.8% 9.5% 7.4% 5.9% 6.4% Source: U.S. Census Bureau.

As this Table shows in average during the 11 years period U.S. Imports from China were equal 8.7% of China’s GDP exceeding during some years the mark of 10%. From my point of view it is the sign of the short-sighted international policy which logically has followed the strategy of offshoring US manufacturing facilities abroad to pursue the necessity of the cost cutting measures under conditions of the “Supply Side” economics. The combined volume of the U.S. imports from China during 11 years period has reached 2,528.4 billion dollars. Big portion of this money can be used to buy American manufactured products and to pay American workers. But money has been used to finance the economic development of the future economic superpower that can become the largest global economy during the next 15-20 years. This fact can be perceived neutrally as the statistical phenomena but it should not be taken easily from the economic perspective. To support its economic growth at the same pace China will need more and more resources. As history is teaching us all political conflicts have the economic foundation: the scarcity of economic resources. So, the strategy chosen in 1980s in the long run practically has weakened the core of the American economy- its manufacturing sector and has support the growth of the new manufacturing giant - China.

Even more alarming situation illustrating the thesis that the “Supply Side” economics practically financed the economic development of other countries instead of the financing the U.S. economy is the situation with the U.S. - Mexico trade. The volume of this trade is less than the volume of the trade with China but the impact of the growing U.S. imports from Mexico on the Mexican economy is even more vivid.

TABLE 5 TRADE BALANCES IN GOODS BETWEEN THE USA AND MEXICO IN BILLIONS OF USD

1994 1995 1996 2000 2005 2006 2007 2008 2009 2010 +1.3 -15.8 -17.5 -24.6 -49.9 -64.5 -74.8 -67.7 -47.8 -66.4 Source: U.S. Census Bureau

As Table 5 shows the deficit in trade with Mexico reached in 2007 $74.8 billion when in 1994 (before NAFTA) it was a surplus of $1.3 billion (U.S. census Bureau, 2011). GDP Mexico in 2010 was $1,034.7 billion, so U.S. imports in 2010 valued of 229.9 billion was equal about 22% of Mexico GDP.

Table 6 indicates that during the 11 years period from 2000 through 2010 the U.S. imports from Mexico have constitute in the average about 20% of Mexico nominal GDP (Source: U.S. Census Bureau, 2011 and International Monetary Fund , 2011).


TABLE 6 U.S. IMPORTS OF GOODS FROM MEXICO (M) IN BILLIONS OF USD AND AS PERCENTS

OFMEXICO’S NOMINAL GDP IN BILLIONS OF USD AT OFFICIAL EXCHANGE RATES

Year 2000 2003 2004 2005 2006 2007 2008 2009 2010 US M 135.9 138.1 155.9 170.1 198.3 210.7 215.9 176.7 229.9 GDP 671.9 700.2 759.6 848.5 951.7 1,035 1,094 879.2 1,035 Percent 20.2% 19.8% 20.5% 20.0% 20.8% 20.4% 19.7% 20.1% 22.2% Sources: U.S. Census (U.S. Imports) and International Monetary Fund (Mexico’s GDP)

The combined volume of the U.S. Imports from Mexico during 11 years period has reached $1,897.4 billion. Again we could expect that the sufficient portion of money spent on imports could be spent on American made goods and could be paid American workers.

Next table (Table 7) shows that the trade deficit in twelve years period from 1997 through 2008 has more than quadrupled (Economic Report of the President. 2011, Table B-105). It is the illustration of the core idea of the economic strategy implemented by the U.S. Government since 1980-s: the increase consumption of imported goods paid by money borrowed by household from banks who have borrowed money from the Federal Reserve System who has borrowed the substantial part of its funds from countries who exported their products to the USA: China, Japan, Saudi Arabia, Brazil. But after the housing bubble has burst the economic model based on the premise that the growth of the consumption can be built on the real estate value appreciation and that it will work as the engine of the economic growth over the very long period of time has shown its inefficiency.

TABLE 7 U.S. BALANCES OF TRADE (TRADE IN GOODS) IN BILLIONS OF USD 1997-2009

1997 1999 2001 2003 2004 2005 2006 2007 2008 2009 -198.1 -346.0 -427.2 -541.5 -665.6 -783.8 -839.6 -823.2 -834.7 -506.9

Source: Economic Report of the President 2008, Table B-103 and Economic Report of the President 2011, Table B-105).

It is obvious that consumption based upon the increasing debt which has been supported by growing housing prices cannot be a factor of economic growth for the long time. When housing bubble has burst households lost the ability to borrow money from banks and many of them now owe to the banks more than the market prices of their houses. Consumption has dropped, economy has contracted.

Table 8 which based upon data of Bureau of Economic Analysis (Bureau of Economic Analysis. 2011) shows changes in exports and imports of capital goods from 1985 through 2010. Table 9 shows changes in exports and imports of automotive vehicles, engines and parts. If imports of automotive vehicles, engines and parts during this 16 years period was always more than twice higher than exports, imports of capital goods has reached exports of capital goods around of 2005. That means that around year of 2005 the USA lost their status of the manufacturing leader of the world.

TABLE 8 U.S. REAL EXPORTS AND IMPORTS OF CAPITAL GOODS EXCEPT AUTOMOTIVE IN

BILLIONS OF USD, SEASONALLY ADJUSTED AT ANNUAL RATES (1985-2010)

Year 1985 1990 1995 2000 2005 2010 Exports 79.2 153.0 247.6 367.9 376.1 466.1 Imports 64.5 119.5 233.0 359.3 393.1 479.5 Net Exports +14.7 +33.5 +14.6 +8.6 -15.0 -13.2 Source: U.S. Bureau of Economic Analysis


TABLE 9 U.S. REAL EXPORTS AND IMPORTS OF AUTOMOTIVE VECHICLES,

ENGINES AND PARTS IN BILLIONS OF USD, SEASONALLY ADJUSTED AT ANNUAL RATES (1985-2010)

Year 1985 1990 1995 2000 2005 2010 Exports 25.2 34.8 65.3 74.9 110.2 115.8 Imports 70.3 87.3 118.4 193.5 251.4 234.3 Net Exports -45.1 -52.5 -53.1 -118.6 -141.2 -118.7 Source: U.S. Bureau of Economic Analysis

The next table (Table 10) that also has been built on the basis of the data the Bureau of Economic Analysis of the U.S. Department of Commerce (Bureau of Economic Analysis, 2011) shows how over 30 year period from 1981 through 2010 values added by three selected industries: manufacturing, finance and insurance, and government as the percent of the U.S. GDP have drastically changed. The manufacturing industry input into GDP declined from almost 20% to about 11.7%. The government industry input during this period of time has slightly fluctuated between 12 and 14%. The finance and insurance industry input has increased from 5% to about 8.5%. That means that the input of manufacturing industry into the domestic output in the USA during the last 30 years declined by about 40%, but the input of the finance and insurance industry increased by 70%. As the result the value added by the finance and insurance industry has increased from the about 25% of the value added by the manufacturing industry to almost 73%. Interestingly to see that in terms of the percentage of GDP value added by the Government was the same in 1981 and 2010 – 13.6%. But because the real GDP in chained 2005 dollars has increased from 1981 to 2010 about two times the Government’s input into GDP in terms of the dollar value also doubled.

TABLE 10 VALUE ADDED BY INDUSTRY AS A PERCENTAGE OF U.S. GROSS DOMESTIC PRODUCT

Year 1981 1985 1990 1995 2000 2005 2010 Manufacturing 19.8 17.8 16.7 15.9 14.2 12.4 11.7 Finance 5.0 5.5 6.0 6.6 7.7 8.1 8.5 Government 13.6 13.8 13.9 13.4 12.2 12.6 13.6 Source: U.S. Bureau of Economic Analysis

All tables have illustrated the decline of the role and the input of the U.S. manufacturing sector into the economic development of the USA: as the percentage of GDP, as the net exports factor of the economic growth, as the source of manufacturing jobs, and as the creator of the jobs in other sectors of the national economy.

So, when the U.S. economy has contracted in 2008 the weaknesses of the manufacturing sector just made the recession deeper. During the recession usually durable goods manufacturers are cutting the production and employment, but nondurable goods manufacturers are keeping both at the same level. Because most of the nondurable goods are imported now from developing countries the ability of this sector to play the role of the savior of the economy and employment was very limited.

We can conclude that the change in the economic policy in 1980s and the change in the U.S. role in the global division of labor were two macroeconomic causes of the recent recession. The recession was inevitable and it must be substantial because the housing market collapse has eliminated the ability of this sector to make another miracle and help economy quickly recover as it happened during the recession 0f 2001. At the same time some mistakes that have been done during the recession made the recovery process long and painful and elevated the status of the recent recession to the level of the “Great Recession”.


HOW THIS RECESSION HAS AFFECTED THE U.S. ECONOMY AND WHAT METHODS HAVE BEEN USED TO FIGHT THE RECESSION

To fight this recession the U.S. Government used both fiscal and monetary policies mechanisms. First of all the U.S. Congress passed in October of 2008 The Emergency Economic Stabilization Act of 2008 which adopted the Troubled Asset Relief Program (TARP). This program gave rights to the U.S. Treasury to buy mortgages and some other financial instruments for the amount of 700 billion dollars. But TARP has not been able to recover lending activities of banks which have received monies from the Federal Government. The recession has deepened and could grow into the depression. During two quarters after Stabilization Act the growth rate of the national economy dropped a big time: the fourth quarter of 2008 – 5.4%, and the first quarter of 2009 drop was -6.4%. Responding to that the U.S. Congress has passed in February of 2009 The American Recovery and Reinvestment Act of 2009. According to this Act 787 billion dollars should be spent to help economy to get out of the crisis, including spending on health care, unemployment, objects of infrastructure and alternative sources of energy. The stimulus package was intended to create jobs and to promote the investment and consumer spending during the recession.

In addition to 1.5 trillion dollars that Government used to stimulate economic growth the Federal Reserve System increased the money supply by about 2.25 trillion dollars buying securities from banks and providing funds to fight possible defaults in payments of the owners of student loans, automotive loans and credit cards. No doubt, that Government took the leading role in helping economy to overcome the recession.

Two years after the Recovery Act was implemented we can observe some positive consequences of its stimulus package:

1. GDP is not falling anymore. Starting the last two quarters of 2009 and after it shows the growth. (Table 11 and Table 12).

2. The rate of unemployment dropped in November 2011 compare with December 2009 from 10% to 8.6%. This is a positive but not the substantial improvement. New jobs have been created but mostly in the government sector of the national economy.

3. The extension of the terms of unemployment benefits payments up to 99 weeks in many states helps to support families of almost 15 million Americans who lost their jobs as a result of the recession.

4. The Federal Government has provided support to state and municipal governments not only in a form of funds for the infrastructure repairs (roads, bridges) and to support jobs of teachers, police officers and firefighters but also in a form of subsidies of interest payments on municipal bonds. As result of these subsidies the interest on municipal bonds increased from 4.5% to 7%, so municipal governments have been able to obtain about 50 billion dollars of investors’ money to finance local projects.

TABLE 11

U.S. REAL GDP QUARTERLY CHANGES IN 2008-2010 (IN CHAINED 2005 DOLLARS)

08/1 08/2 08/3 08/4 09/1 09/2 09/3 09/4 10/1 10/2 -1.8 1.3 -3.7 -8.9 -6.7 -0.7 1.7 3.8 2.5 2.3 Source: Bureau of Economic Analysis.

Table 11 shows (Bureau of Economic Analysis, 2011) the sharp decline of GDP in 2008-2009 and slow and anemic growth in 2009-2010. Why the recovery was slow and weak? Why the economy needed 9 quarters to reach the level of GDP of the fourth quarter of 2007 - the last quarter prior the contraction that happened during the first quarter of 2008 which served the role of the introductory stage of the recession of 2008-2009?

Answer to those questions we can get from the analysis of Table 12. This table based upon the data from the Bureau of Economic Analysis (Bureau of Economic Analysis, 2011) includes information about


quarterly changes in the U.S. GDP and its components: “Personal Consumption Expenditures” – (C), “Gross Private Domestic Investment” - (IG), “Government Expenditures” – (G) and “Net Exports “- (X-M) during 17 quarters starting with the 3d quarter of 2007 and finishing with the 3d quarter of 2011. The base quarter in this analysis is the 4th quarter of 2007, the last quarter when GDP has shown the growth. After that quarter the U.S. GDP has started to contract and the contraction has lasted through the 3d quarter of 2009. The GDP has recovered to the level of the 3d quarter of 2007 only by the 3d quarter of 2011. In other words the U.S. economy did not show an economic expansion during the long period of 4 years.

Table 12 shows that GDP itself and its components of ”Personal Consumption Expenditures – (C), “Government Expenditures” – (G) and “Net Exports “- (X-M) today by the end 0f 2011 in the Third Quarter of the year 2011 have exceeded its values of the Fourth Quarter of the year 2007 when the Great Recession practically has started (GDP: $13,352/ $13,326; C: $9,449.5/ $9, 312.6; G: $2,508.2 / $2,455.3; X-M: -$409.4 /-$564.6). But the main GDP component reflecting the economic growth - “Gross Private Domestic Investment” - (IG) despite of the stimulus program still did not have reached the level that it had before the recession has started (IG: $1,796.6/ $2,123.6). The second important component of economic growth that stimulates an economic growth through the increase in the Aggregate Demand – the component of “Personal Consumption Expenditures” – (C) has reached the level of the 4th quarter of 2007 only during the 4th quarter of 2010. (C: $9,328.4 / $9,312.6).

TABLE 12 U.S GDP AND ITS COMPONENTS QUARTERLY CHANGES 1N 2007-2011 (IN 2005 USD)

Year GDP C IG G X-M 2007 Q3 13,269.80 9,285.20 2,176.30 2,447.90 -638.1 2007 Q4 13,326.00 9,312.60 2,123.60 2,455.30 -564.6 2008 Q1 13,266.80 9,289.10 2,055.70 2,473.90 -550.2 2008 Q2 13,310.50 9,285.80 2,024.00 2,484.50 -486.2 2008 Q3 13,186.90 9,196.00 1,934.70 2,510.70 -484.6 2008 Q4 12,883.50 9,076.00 1,744.60 2,520.50 -478.0 2009 Q1 12,663.20 9,040.90 1,490.40 2 509.60 -404.2 2009 Q2 12,641.30 8.998.50 1,397.20 2,546.00 -331.8 2009 Q3 12,694.50 9,053.30 1,407.30 2,554.20 -352.4 2009 Q4 12,813.50 9,060.20 1,522.00 2,548.50 -346.9 2010 Q1 12,937.70 9,121.20 1,630.00 2,540.60 -376.8 2010 Q2 13,058.50 9,186.90 1,766.80 2,564.00 -437.4 2010 Q3 13,139.60 9,247.10 1,766.80 2,570.30 -458.7 2010 Q4 13,216.10 9,328.40 1,734.50 2,552.30 -414.2 2011 Q1 13,227.90 9,376.70 1,750.90 2,513.90 -424.4 2011 Q2 13,271.80 9,392.70 1,778.40 2,508.20 -416.4 2011 Q3 13,352.80 9,449.50 1,796.60 2,508.20 -409.4 Source: Bureau of Economic Analysis.

Because the U.S. GDP itself has reached the pre-recessionary level only during the 3d quarter of 2011 we can conclude that the U.S. economy just probably barely recovered from the recession and that the lower than in 2007 level of Gross Private Domestic Investment cannot provide enough stimuli for the economic growth. From my point of view this is another consequence of the diminishing role of manufacturing in the USA. In this context the fact that the official unemployment rate only recently dropped below 9.0% and was kept at the level between 9 and 10 percent during the years of 2010 and 2011 is corresponding to slow recovery of the business investments. It is becoming especially alarming because very often the trend of GDP changes is following the trend of changes in IG.


So, it not a surprise that to recover from such deep recession economy took more time than three years. The reason why it took so long for the economy to recover was the ill implementation of the Keynesian advise to increase the government spending when economy in the contraction faze to offset the decline in the aggregate demand created by households and businesses. When economy needs the stimulus package from the government the government should be able to provide this package at the scale that economy needs. Traditionally economy will need the stimulus package that will help it return the level of the so called “potential” or “full employment” GDP.

The American economist Arthur Okun who had served as the Chairman of the Council of Economic Advisers in 1968-69 has formulated the empirical law that has been called the Okun’s Law. According to this Law every additional percent of actual official unemployment rate above the natural rate of unemployment will generate two percent of the loss of the “potential” GDP or GDP gap. The GDP gap according to Okun is the difference between the “potential” GDP and actual (nominal) GDP.

The official rate of unemployment in 2008-2009 was around 10%. It means that the actual official unemployment rate has exceeded the so called “natural rate” of unemployment, which today is considered to be around of 5 %, by exactly 5 %. So, according to Okun’s Law the U.S. potential GDP has declined during the of the first year of the recession by 10% (5% x 2) or about 1.45 trillion dollars. The Nominal GDP in the 2d quarter of 2008 was 14,415.5 billion dollars. Respectfully 10% of that amount was 1,441.6 billion dollars. So, if the investment (fiscal) multiplier of the government spending in the USA is about 1.5 to bridge this gap between potential and real GDP the U.S. Government should invest into the national economy about 960 billion dollars. In other words, it looks that the economy has needed larger stimulus package in 2009 than the package that the government has been able to offer. It also means that economy may be still needs the additional stimulus package of about 180 - 200billion dollars to achieve the full and speedy recovery. The additional government’s stimulus package may be one of the possible options that can be considered by the President Obama’s administration and the U.S. Congress. But after Democrats lost the majority in the Congress it became very difficult for the Administration to get the support of this idea from the law makers. But this option was valuable and could work, from my opinion, if it had been implemented in 2009-2010.

But any increase in government spending without the increase in taxes means that the national debt will grow even more. In the last ten years the deficit growth became the self-feeding monster because the guaranteed interests’ payments often can be obtained only from the next borrowing efforts. In addition to this problem there are some others that are affecting the debt growth. The changes in the demographic structure of the population are increasing the government obligations to the growing group of elderly citizens in the form of social security payments and Medicare. Plus the military-industrial complex is constantly lobbing its interests forcing the government to increase the military expenditures. As the result, the federal government expenditures for the last 80 years since 1929 through 2009 have been grown from 7.7% of GDP to 36.2% (Bureau of Economic Analysis, 2011). In the dollar amount this total government spending increased from $8.0 billion in 1929 to $5,261.8 billion in 2010 in current dollars (Bureau of Economic Analysis, 2011). The deficit of the Federal budget jumps to $1,554.9 billion dollars in 2009 (Bureau of Economic Analysis, 2011). The last surplus of $121.0 billion dollars the U.S. Federal budget had in 2000 (Bureau of Economic Analysis, 2011). The national debt has reached the astonishing level of $15.13 trillion dollars in September of 2011(U.S. National Debt Clock, 2011). This amount is about equal the U.S GDP in current dollars. The U.S. nominal GDP in 2010 was $14,755.0 billion dollars (Bureau of Economic Analysis, 2011) and by the end of 2011 probably would be about $15.2 trillion.

The Keynesian economic model based upon the idea that the aggregate demand should be stimulated through increase in spending to achieve the economic growth. To stimulate consumption of households, which is by far the largest component of the expenditures formula of GDP, taxes should be decreased. But lower taxes means less government receipts. So, to keep government spending high to support their growth during the recessions the government must increase the borrowing or increase taxes.

During the last 40 years the attitude towards the fiscal policy has been changed few times. Probably it was the effect of the ability of the mechanism of the government regulation to adapt towards changing economic environment. In 1960-s most economists and politicians have considered the fiscal policy as the


very efficient tool of the national economy regulation. But high levels of the government spending during the Vietnam War have increased the federal budget deficit and have provoked the growth of the inflation. Taxes’ increases as the government’s response to the situation have negatively affected the consumption as the largest component of the Aggregate Demand. The decline in consumption caused two recessions of 1970 and 1974-1975 and the long period of stagflation of 1970s when low rate of the economic growth have been complemented by high rates of inflation.

The so called “neo-liberal” economic model that has been introduced in 1980-s has used as the weapon against the stagflation the expansionary fiscal policy. The major argument used then was against the high tax rates. High tax rates would negatively affect the consumption and business investments but would not be capable in slow growing economy to generate enough tax revenues for the government. As the result, the highest marginal personal income tax rate that has reached during Eisenhower and Kennedy terms 91 % has been lowered down during the Reagan term to 28% in 1988 (Tax Foundation, 2011). The tax rate for highest income earners have been decreased more than 3 times but in 2004 the half percent of people with highest income have contributed 26.1 % of the personal income tax receipts collected by the federal government. In 1960 the taxes paid by that half percent of the tax payers with highest incomes have contributed only 14% of the federal government personal income tax receipts. The decrease in tax rates did worked. During 8 years when Reagan was in office the tax rates have been decreased about 2.5 times but the federal government receipts from the personal income tax almost doubled: from 308.7 billion in 1980 to $549.0 billion in 1989 (Bureau of Economic Analysis, 2011). When military spending is rising, as it was during Reagan presidential term, the deficit would have the tendency to grow and this rise in government spending can stimulated the economic growth. When military spending is declining, as it was during Clinton presidential term, the economic growth would stimulate the budget surplus.

In XXI century Bush and Obama administrations both used the expansionary fiscal policy. During the recession of 2001 this policy has helped to recover the economic growth very quickly. But later on the growing military expenditures caused by wars in Iraq and Afghanistan have intensified negative effects of other crises that emerged from decisions and actions made by the previous administrations: deregulation of financial markets during the Reagan presidential term, and the deregulation of the housing market during Clinton presidential term. As the result the uncontrolled growth of the federal deficit and the national debt and the very deep financial crisis that was caused by the collapse of the real estate market have amalgamated with negative effects of the changes in the macroeconomic policy and strengthen the recessionary prospects. As the result the national economy has been pushed into very deep and painful recession of 2008-2009.

Most economists would agree with the importance of government spending during the recession. But governments spending cannot substitute in the long run the business investments as the government jobs paid from taxes cannot substitute manufacturing job to stimulate the economic growth. We should look for the complementation not substitution. Economy still needs more help from the government. But this help should be provided not only in the form the second financial stimulus package. It looks that the structure of the U.S. economy needed to be changed and more manufacturing should be returned from the overseas to the USA.

How it can be done is the different question and the topic of the independent research. Among possible strategies that can help are tax incentives to companies that would return manufacturing jobs into the USA. Government grants to finance the R&D in manufacturing sector can help companies to develop new technologies. New technologies can ensure the U.S. competitive advantage in the global division of labor. Even the creation of joint ventures between private companies and government entities for example in the energy sector of the national economy can help to increase the role and the impact of the manufacturing sector on economic growth. The spectrum of possible government incentives is very broad. The main problem is not the lack of ideas and forms but the lack of the cooperation between politicians, business leaders and general public in discussion and implementation of these ideas in the form of the new economic policy.


REFERENCES Bivens, J. (2003). Updated Employment Multipliers for the U.S. Economy. Working Paper #268. Economic Policy Institute. http://www.epi.org/page/-/old/workingpapers/epi_wp_268.pdf Bureau of Economic Analysis.(2011). Full-Time Employment Employees by Industry 2003-2007. http://www.bea.gov/iTable/iTable.cfm?ReqID=9&step=1 Bureau of Economic Analysis (2011). Real Exports and Imports of Goods and Services in 1995-2011 in Chained Dollars. http://www.bea.gov/iTable/iTable.cfm?ReqID=9&step=1 Bureau of Economic Analysis (2011). Table 1.1.5. Gross Domestic Product. http://www.bea.gov/national/index.htm Bureau of Economic Analysis (2011). Table 3.2. Federal Government Current Receipts and Expenditures. http://www.bea.gov/iTable/iTable.cfm?ReqID=9&step=1 Bureau of Economic Analysis (2011). U.S. GDP Quarterly Changes in 2008-2011 (Percent Change From Preceding Period Based on Chained 2005 Dollars). http://www.bea.gov/national/index.htm Bureau of Economic Analysis (2011). U.S. GDP and Its Components Quarterly Changes in 2007-2011 (in Billions of Chained 2005 Dollars). Revised data on 09/27/2011. http://www.bea.gov/iTable/iTable.cfm?ReqID=9&step=1 Bureau of Economic Analysis (2011. Value Added by Industry as a Percentage of GDP in 1981-2010. http://www.bea.gov/iTable/iTable.cfm?ReqID=9&step=1 Bureau of Labor Statistics (2011). Employees on nonfarm payrolls by major industry sector, 1961 to date. ftp://ftp.bls.gov/pub/suppl/empsit.ceseeb1.txt Economic Report of the President (2008). Table B-30. Disposition of Personal Income, 1959-2007. http://www.gpoacess.gov Economic Report of the President (2011). Table B-35. Civilian Population and Labor Force, 1929-2010. http://www.gpo.gov/fdsys/pkg/ERP-2011/pdf/ERP-2011-table35.pdf Economic Report of the President (2008), Table B-46. Employees on Nonagricultural Payrolls by Major Industries, 1959-2007, Tables: B-30, B-46, B-103. http://www.gpoacess.gov Economic Report of the President (2011). Table B-104. - U.S. International Trade in Goods by Principal End-Use Category, 1965-2010. http://www.gpo.gov/fdsys/pkg/ERP-2011/pdf/ERP-2011-table104.pdf Economic Report of the President (2011). Table B-105. U.S. International Trade in Goods by Area 2002-2010. http://www.gpo.gov/fdsys/pkg/ERP-2011/pdf/ERP-2011-table105.pdf Economic Report of the President (2011). Table B-106. - U.S. International Trade in Goods on Balance of Payments (BOP) and Census Basis and Trade in Services on BOP basis 1981-2010. http://www.gpo.gov/fdsys/pkg/ERP-2011/pdf/ERP-2011-table106.pdf


Ezel, S. J., Atkinson R.D., (2011). The Case for National Manufacturing Strategy. The Information Technology and Innovation Foundation. http://www.itif.org/files/2011-national-manufacturing-strategy.pdf The Financial Crisis Inquiry Report. Final Report of the National Commission on the Causes of the Financial and Economic Crisis in the United States. January 2011. Official Government Edition. http://www.gpo.gov/fdsys/pkg/GPO-FCIC/pdf/GPO-FCIC.pdf Galbraith, J. K. (2008). The Collapse of Monetarism and the Irrelevance of the New Monetary Consensus. 25th Annual Milton Friedman Distinguished Lecture at Marietta College, Marietta, Ohio. March 31, 2008. http://utip.gov.utexas.edu/papers/CollapseofMonetarismdelivered.pdf Hummel, J.R. (2011). Ben Bernanke versus Milton Friedman: The Federal Reserve’s Emergence as the U.S. Economy’s Central Planner. The Independent Review, 15, (4), 485-518. International Monetary Fund (2011). World Outlook Databases. http://www.imf.org/external/pubs/ft/weo/2011/02/weodata/weorept.aspx?sy=1980&ey=2016&sc Katkov, A. (2011). An Analysis of the Government Policy to Ease the References: Recessionary Pressure of 2008-2009. Journal of Business and Behavioral Sciences, 21, (2), 145-153. Krugman, P. (2009). How Did Economists Get It So Wrong? New York Times. September 2, 2009. http://www.nytimes.com/2009/09/06/magazine/06Economic-t.html?pagewanted=1&em Liebowitz, S. (2008). Anatomy of a Train Wreck. Causes of the Mortgage Meltdown. The Independent Institute. The Independent Policy Report. October 3, 6-10. Palley, T. (2010). Plan B for Obama. New America Foundation. September 6, 2010. http://newamerica.net/publications/policy/plan_b_for_obama Simmons, R.T. (2011). Beyond Politics (Revised and updated edition). The Roots of Government Failure. The Independent Institute. Oakland, California. ISBN 978-1-59813-042-3. Tax Foundation (2011). Federal Individual Income Tax Rates. 1913-2011. http://www.taxfoundation.org/files/fed_individual_rate_history_nominal&adjusted-20110909.pdf U.S. Census Bureau (2011). Census 2000 Demographic Profile Highlights. http://www.factfinder.census.gov/servlet/ U.S. Census Bureau (2011). Trade in Goods with China. http://www.census.gov/foreign-trade/balance/c5700.html U. S. Census Bureau (2011). Trade in Goods with Mexico. http://www.census.gov/foreign-trade/balance/c2010.html U.S. Census, NAICS. http://www.census.gov/cgi-bin/sssd/naics/naicsrch?code=31&search=2007 NAICS Search U.S. National Debt Clock: Real Time. http://www.usdebtclock.org/


Yandle, B. (2010) Lost Trust. The Real Cause of the Financial Meltdown. The Independent Review, 14, (3), 341-361.


The Effect of Fringe Benefits on the Paid Overtime Hours in Japan

Hui-Yu Chiang Osaka University

In this paper, we adapt Bell and Hart’s model (1999) to examine paid overtime work by using Japanese Survey of Company Fringe Benefits data which includes information on the employer’s provision of fringe benefits as well as paid overtime hours for individuals. By including a crucial labor demand variable- the quasi-fixed cost which is omitted from Bell and Hart’s (1999) report, the present study provides more complete documentation of the structure of labor costs with paid overtime work. INTRODUCTION

In this paper, we argue that higher fringe benefit costs may have an effect on paid overtime hours. This result occurs because a fringe benefit is a quasi-fixed employment cost that rises with the number of workers rather than hours worked. As the fringe benefit becomes more expensive to provide, firms have an incentive to substitute hours per workers for employment.

Employer-provided fringe benefits represent a large and growing share of compensation paid to Japanese workers. According to fringe benefit costs data for 2006 from the Japan Business Federation (Nippon Keidanren), fringe benefits represented 15.1% of the average Japanese worker’s total compensation (See Appendix-Table A1). Over the past 40 years, employer expenditures on fringe benefits have grown over 30% (See Appendix-Figure A1). Given these figures, one is not surprised that the role of fringe benefits in the labor market has attracted a great deal of attention from academic economists and policy-makers.

Figure A2-1 reveals the declining trend in hours worked per month for those who worked at least 40 weeks in the previous year in Japan. Average hours worked per month fell by over 15 hours from 1981 to 2007. The declining trend is the net consequence of declining scheduled working hours more than compensating for the relative constancy of overtime hours (Figure A2-2). At the same time, as shown in Figure A1, employer expenditures on fringe benefits have drifted upward over time. The increasing costs of fringe benefits would make employers less willing to reduce the working hours of workers. Shorter working hours will increase these fixed costs per worker hour and the overall hourly cost. This explanation might partly explain the relative constancy of the overtime working hours.

Bell and Hart (1999) used the UK Labour Force Survey to examine paid and unpaid overtime work for both males and females. However, their study contained no information on a key component of the labor demand model: the employer’s quasi-fixed employment costs, including statutory and non-statutory fringe benefits that do not depend on hours worked.

In this paper, we adapt Bell and Hart’s model (1999) to analyze the Survey of Company Fringe Benefits data in Japan. These data include information on the employer’s provision of fringe benefit as well as paid overtime hours for individuals. By including a crucial labor demand variable - the quasi-fixed


cost which is omitted from Bell and Hart’s (1999) report, the present study provides more complete documentation of the structure of labor costs with paid overtime work.

The structure of the paper is as follow. In the next section, we outline theory and findings from the international literature on the quasi-fixed employment cost and overtime work. In section 3, we introduce the data set and the model used in this study. In section 4, we present the key results of the empirical analysis. In the conclusion we briefly review the findings of our study. THEORY AND LITERATURE REVIEW

Oi (1962) proposed a short-run theory of employment that rests on the premise that labor is a quasi-fixed factor. Employers can change their labor input by changing the number of employees, the hours per employee, or both. The way in which they adjust will depend on the relative costs of the different options. Quasi-fixed costs that they do not increase proportionately with hours worked drives a wedge between the marginal cost of hiring an additional worker and working an existing worker more hours. This generates an important distinction between the number of workers and hours worked per worker in yielding a given labor input. Some of these quasi-fixed costs will bias firms toward working their existing employees more intensively instead of hiring additional employees. Increasing the number of employees would be more costly given the quasi-fixed costs.

Ehrenberg and Schumann (1983) provided evidence that quasi-fixed employment costs influence employer overtime choices. Their study used establishment-level data from various years of the Employer Expenditure for Employee Compensation surveys. The basic empirical methodology was to regress annual overtime hours per employee on control variables and the ratio of quasi-fixed labor costs to the overtime wage. Typical findings indicated a statistically significant positive association across establishments between this ratio and the use of overtime.

A few studies have applied the quasi-fixed costs theory to working hours. For example, the quasi-fixed costs theory has been applied to show that an increase in the cost of providing health insurance has a significant effect on work hours. Cutler and Madrian (1998), using data from the Current Population Survey (CPS) and the Survey of Income and Program Participation (SIPP), showed that rising health insurance costs during the 1980s increased the hours worked by those with health insurance by up to 3%. Katestner and Simon (2002) also examined data from the 1989–1998 March Current Population Surveys and found that the number and type of state-mandated health insurance benefits were unrelated to weeks of work, wages, and the prevalence of private insurance coverage, but positively associated with weekly work hours. Dolfin (2006) used the 1982 Employer Opportunity Pilot Project (EOPP) cross-sectional firm-level US data to examine the size of firms’ quasi-fixed employment costs and their importance in affecting hours of work. The measures of quasi-fixed employment costs used relate to recruiting, search, hiring, training, and firing. The results show that higher costs are associated with longer hours.

Regarding the study of overtime hours, Bell and Hunt (1999) used the UK Labour Force Survey for 1993/94 to examine the determinants of paid and unpaid overtime work for both males and females. In their study, working paid hours were associated with manager status, age, being married, high standard hourly wages, and union coverage.

Our analysis differs from previous studies in some other important ways. First, we use cross-sectional data on individuals, whereas most previous studies used only firm- or industry-level data. The use of data at the level of individuals allows us to control for a variety of supply-side factors affecting individual work decisions that cannot be accounted for with firm-level data. Second, most studies have focused on the health insurance, which is only part of total nonwage compensation, as a quasi-fixed cost effect. The quasi-fixed costs considered in our study included fringe benefits such as health insurance, a pension, and employment insurance as a quasi-fixed cost effect. Third, although Dolfin (2006) examined recruiting and training costs as quasi-fixed costs, they used the total hours worked per week as a dependent variable and did not distinguish among standard working hours, unpaid overtime hours, and paid overtime hours, although the overtime premium wage rate is usually larger and different from the straight-time hourly wage. Finally, we adapted Bell and Hart’s (1999) model to analyze the Survey of Company Fringe


Benefits data for Japan, which include information on an employer’s provision of fringe benefits as well as paid overtime hours for individuals. With this framework, we can document more completely the structure of labor costs with paid overtime work. RESEARCH METHOD Data

The data used in our analysis were taken from the Survey of Company Fringe Benefits in 2006, collected during December 2006. This survey is a triennial survey of company fringe benefits at the establishment level in Japan and has been conducted by the Life Insurance Culture Centre from 1980 to 2006. Companies were selected according to a stratified sampling method by industry and by firm size. An interview-administered questionnaire was completed at the respondents’ workplaces. The final response rate was usually between 45% and 52%, which is comparable to other company surveys in Japan. This study analyzed only the 2006 survey, for which both an employer survey and an employee survey were conducted. The survey region includes the Tokyo Metropolitan District and 12 ordinance-designated cities, and the survey targets were private companies in Japan with five or more regular employees. The samples consisted of 1504 companies in the employer data set and 2972 workers in the employee data set (including 2052 full-time employees and 920 part-time employees). Respondents reporting that they were part-time workers were excluded.

Although the employer data set and employee data set could not be matched in this survey, use of the employee data set was sufficient for our analysis because it contained detailed information on fringe benefits and paid overtime hours. The survey listed the types of fringe benefits within a company including housing, health care, living expenses, bereavement benefits, and leisure benefits and asked whether the respondent was offered each of these fringe benefits. Descriptions of these fringe benefits are given in Table A2. In addition, the survey asked the following about the respondent’s overtime hours: How many paid overtime hours were worked per week on average last month?

Further information was collected on the employee’s age, gender, marital status, presence of children, job tenure, union coverage, and occupation, along with the employing establishment’s industrial classification and regional location. Estimation Methods

Approximately 50% of the full-time workers in our sample did not engage in paid overtime work, as measured weekly. If we use ordinary least-squares to estimate a regression to censored observations, the estimates are inconsistent. Therefore, the Tobit model is necessary. For the Tobit model, the structure equation is:

𝑦𝑖∗ = 𝑋𝑖𝛽 + ∈𝑖 where 𝜖𝑖 ~𝑁 (0,𝜎2). 𝑦∗ is a latent variable that is observed for values greater than τ and censored otherwise. The observed y is defined by the following measurement equation:

𝑦𝑖 = 𝑦∗ 𝑖𝑓 𝑦∗ > 𝜏 𝑦𝑖 = 𝜏𝑦 𝑖𝑓 𝑦∗ ≤ 𝜏

y here denotes the overtime hours a week by one full-time worker. We assume that τ = 0 because the data are censored at 0 in our sample. Thus, we have

𝑦𝑖 = 𝑦∗ 𝑖𝑓 𝑦∗ > 0 𝑦𝑖 = 0 𝑖𝑓 𝑦∗ ≤ 0


The likelihood function for the censored normal distribution is:

L = [ 1𝜎 ∅

𝑦− 𝜇𝜎 ] 𝑑𝑖 [1− 𝛷

𝜇 − 𝜏𝜎 ] 1−𝑑𝑖

𝑁

𝑖

where τ is the censoring point. In our tobit model, we set τ = 0 and parameterize µ as 𝑋𝑖𝛽 . This gives us the likelihood function for the tobit model:

L = [ 1𝜎

∅ 𝑦𝑖 − 𝑋𝑖𝛽

𝜎 ] 𝑑𝑖[ 1 −𝛷 𝑋𝑖𝛽𝜎 ] 1−𝑑𝑖

𝑁

𝑖

The log-likelihood function for the tobit model is:

lnL = 𝑑𝑖 −𝑙𝑛𝜎 + 𝑙𝑛∅ 𝑦𝑖 − 𝑋𝑖𝛽

𝜎 + (1− 𝑑𝑖) 𝑙𝑛1 −𝛷

𝑋𝑖𝛽𝜎

𝑁

𝑖=1

The first part of the overall log-likelihood corresponds to the classical regression for the uncensored

observations, while the second part corresponds to the relevant probabilities that an observation is censored. Measures

The variables used in our study followed those of Bell and Hart (1999). The difference was that we added fringe benefit cost as the main independent variable. The means and standard deviations of the variables used in the analysis are presented in Table 1.

TABLE 1 SAMPLE MEANS AND STANDARD DEVIATIONS

Whole sample Male Female Variable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. paid overtime 3.19 5.31 3.89 5.96 2.42 4.37 statutory fringe benefits 49612.00 11651.68 50468.81 11280.13 48659.97 11985.23 non-statutory fringe benefits 3633.82 4096.22 4031.90 4272.20 3248.59 3882.46 total fringe benefits 52258.80 14028.21 53442.33 14057.04 51113.44 13913.68 male =1 0.53 0.50 age 38.66 10.28 39.83 10.41 37.36 9.99 married =1 0.53 0.50 0.65 0.48 0.39 0.49

child =1 0.48 0.50 0.34 0.47 0.19 0.39 union =1 0.33 0.47 0.40 0.49 0.26 0.44 income 455.24 222.75 548.58 233.36 349.18 151.67 tenure 121.10 107.57 141.34 120.25 98.63 86.11 experience 18.26 10.72 19.07 10.77 17.36 10.60

Sample size 2052 1080 972 *total fringe benefit is calculated as statutory plus non-statutory fringe benefits.


Dependent Variable To analyze the effect of fringe benefit costs, we used paid overtime working hours for the dependent

variables. Independent Variables

The main independent variable was fringe benefit costs including statutory fringe benefit cost and non-statutory fringe benefit cost. The employee data set in the Survey of Company Fringe Benefits in 2006 provided information on the types of fringe benefits within a company and whether the employees could use each fringe benefit. However, the survey had no information on firms’ expenditures for each fringe benefit. The best source of data on the employer cost for various fringe benefits is the General Survey on Working Conditions.

The General Survey on Working Conditions, an annual survey of the wage and working hours system at the industry level and firm-size level, has been conducted by the Ministry of Health, Labour and Welfare in Japan since 1984. In the 2006 survey, approximately 5341 enterprises were selected according to a random sampling method from among private enterprises that employed more than 30 regular employees.

The final response rate was 82.7%, making the final sample 4416. An advantage of using this survey is the rich information on the mean labor cost for statutory fringe benefit costs such as health insurance, pensions, and employment insurance, and for non-statutory fringe benefit costs such as for housing, health care, living expenses, bereavement allowance, and leisure benefits. To create a new variable of benefit costs, the information on the mean employer cost for certain fringe benefits by industry and by firm size make the combination of two different data sets possible.

To construct a measure of benefit costs, we followed Montgomery and Cosgrove (1993) and Buchmueller (1999) in combining the fringe benefit dummy variables obtained from the Survey of Company Fringe Benefits in 2006, with cost-based weights, which are the mean costs for various fringe benefits taken from the General Survey on Working Conditions in 2006 to create the new benefit costs variable. Control Variables

The control variables are based on those used in Bell and Hart’s (1999) study on Tobit estimates of paid overtime hours. They include worker characteristic variables such as annual income, age, marital status, child, tenure, experience, education, occupation, and manager and firm characteristic variables such as union, firm size, and area. However, Bell and Hart’s (1999) analysis contained no information on fringe benefit costs. The present study added the variable of fringe benefit costs to investigate how it affects overtime work. RESULTS

The main analysis purpose was to examine if fringe benefit cost as a quasi-fixed employment cost has an effect on paid overtime hours. If fringe benefit cost is statistically significant in our model, this might suggest a problem with omitted variables in Bell and Hart’s (1999) study.

Findings for predicting the effect of fringe benefit costs on paid overtime hours by total workers are reported in Table 2. We estimated the effect of statutory fringe benefits on paid overtime hours in model 1 and non-statutory fringe benefits in model 2. We also accounted for statutory fringe benefits and non-statutory fringe benefits in the total fringe benefits in models 3. In Table 2, the coefficient for fringe benefits showed strong statistical significance at the 1% level in models 1 through 3. This indicated that accepting more fringe benefits from firms meant that workers worked longer paid overtime hours. In addition, non-statutory employer-provided fringe benefits have more effect than statutory ones on the paid overtime hours.


TABLE 2 TOBIT ESTIMATES OF PAID OVERTIME HOURS

Total workers Non-manager Variables (1) (2) (3) (4) (5) (6) statutory fringe 0.676*** 0.750*** benefits (0.203) (0.208) non-statutory fringe 2.434*** 2.397*** benefits (0.620) (0.634) total fringe benefits 0.691*** 0.740***

(0.175) (0.178) male 1.637*** 1.710*** 1.656*** (0.532) (0.532) (0.532) manager –9.259*** –9.243*** –9.243*** (0.941) (0.942) (0.940) age –0.460** –0.471** –0.459** –0.539** –0.550*** –0.539**

(0.207) (0.207) (0.206) (0.213) (0.213) (0.213) age square 0.003 0.003 0.003 0.004 0.004 0.004

(0.003) (0.003) (0.003) (0.003) (0.003) (0.003) married –0.299 –0.312 –0.297 –0.041 –0.036 –0.034

(0.548) (0.548) (0.547) (0.555) (0.556) (0.554) youngest child –0.862 –0.758 –0.854 –0.903 –0.794 –0.886 under 11 (0.586) (0.585) (0.585) (0.600) (0.601) (0.599) union 2.066*** 1.816*** 1.902*** 2.082*** 1.883*** 1.918***

(0.596) (0.606) (0.600) (0.615) (0.625) (0.619) log of income 4.335*** 3.942*** 4.182*** 5.343*** 5.002*** 5.199***

(0.762) (0.773) (0.763) (0.711) (0.722) (0.712) firm size 100-500 3.010*** 2.642*** 2.957*** 3.406*** 3.019*** 3.343***

(0.692) (0.690) (0.689) (0.714) (0.712) (0.711) firm size 500-1000 2.629*** 1.753** 2.397*** 3.007*** 2.134*** 2.752***

(0.747) (0.774) (0.747) (0.767) (0.796) (0.767) firm size 1000- 1.627** –0.294 1.082 1.717** –0.174 1.136

(0.738) (0.887) (0.749) (0.752) (0.904) (0.763) area control yes yes yes yes yes yes constant –18.823*** –13.457*** –18.112*** –22.732*** –17.279*** –21.924***

(5.001) (4.975) (4.952) (5.035) (5.002) (4.985) Sample size 1903 1903 1903 1603 1603 1603 Robust standard errors are in parentheses. ***indicates significance at 1% level; **indicates significance at 5% level; *indicates significance at 10% level.

Altogether, control variables such as male, income, and union were positively associated with paid overtime hours. Control variables such as manager and age were negatively related to paid overtime hours. Males worked more paid overtime hours than females. Considering that sex may have different effects on overtime hours, we later conducted another regression by sex.

We also calculated the same equation excluding managers because this group worked little paid overtime. In our sample (as Table 3 shows), 82.23% of male managers and 87.80% of female managers reported working unpaid overtime; therefore, we excluded the variable of manager in the next estimate by sex.


TABLE 3 PAID WEEKLY OVERTIME BY OCCUPATION

Weekly paid overtime hours (%) Occupation 0 1–6 7–12 13–20 21–40 obs male manager 82.23% 10.15% 6.09% 1.52% 197 sales 62.50% 20.45% 13.64% 3.41% 176 technician 38.42% 26.43% 22.62% 8.99% 3.54% 367 laborer 38.19% 31.25% 20.83% 9.72% 144 clerical 44.37% 35.76% 11.92% 5.30% 2.65% 151

female manager 87.80% 7.32% 2.44% 2.44% 41 sales 56.10% 26.83% 10.98% 4.88% 1.22% 82 technician 47.92% 36.81% 11.11% 3.47% 2.63% 144 laborer 57.89% 23.68% 15.79% 2.63% 38 clerical 55.11% 32.74% 8.10% 3.57% 0.49% 617

As shown in Table 4, fringe benefits variables were very significant by sex. A non- statutory fringe benefit had more of an effect on overtime paid hours than a statutory fringe benefit in the male model. Age was statistically significant for women but not men. Married status was not statistically significant for both men and women. Having a child aged less than 11 years was associated with low paid overtime hours for women. Union membership meant long paid overtime hours for men. Bell and Hart (1999) argued that hours agreements are based on work scheduling over specific tasks, they would expect that unpaid hours would be linked to an absence of union collective bargaining agreements. For union workers, collective bargaining arrangements covering working time would supersede individual task assignments. Therefore, they expect unpaid overtime work to be negatively associated with union coverage. In contrast, they also expect paid overtime work to be positively associated with union coverage. Log income was statistically significant and positively related to the paid overtime work for both men and women in our model. This outcome contrasts with findings of Bell and Hart (1999) using UK Labour Force Survey but corroborates other findings of Trejo (1991) using the Current Population Surveys (CPS) data.

Is there a relationship between paid overtime work and job tenure? To explore the issue further, we recalculated the equations using two human capital-related Mincer variables, job tenure (length of stay in the current job) and work experience (length of labor market experience since completing full-time education). Table 5 shows the results. Because age had no effect on paid overtime hours in the male equation, we used job tenure and work experience (including quadratics in tenure and experience) instead of age. We found that the probability of paid overtime rose in job tenure in the male equation although it declined in work experience. Hart and Ma (2010) also provided evidence of a relationship between paid overtime hours and job tenure that emphasized the role of specific human capital investment. They also found that paid overtime was related positively to job tenure and negatively to work experience.



-By Sex (Without Manager Sample)

Male Female

Variables (1) (2) Bell and

Hunt (1999)

(3) (4) Bell and

Hunt (1999)

statutory fringe 0.716** 0.795*** benefits (0.322) (0.255) non-statutory fringe 2.563*** 2.068** benefits (0.931) (0.825) age –0.148 –0.197 0.890*** –0.719*** –0.688*** 0.425**

(0.348) (0.347) (5.170) (0.251) (0.252) (2.500) age square –0.002 –0.001 -0.011*** 0.007** 0.007** -0.007***

(0.004) (0.004) (-5.45) (0.003) (0.003) (-3.03) married –0.297 –0.300 -1.902 *** 0.049 0.019 0.452

(0.903) (0.903) (-2.75) (0.654) (0.657) (0.680) youngest child –0.625 –0.524 -0.568 –1.988** –1.908** -1.368 under 11 (0.875) (0.874) (-0.54) (0.826) (0.829) (-1.28) union 3.351*** 3.031*** 6.705*** 1.033 1.005 3.624 ***

(0.905) (0.922) (11.390) (0.796) (0.808) (5.400) log of income 3.691*** 3.433*** -1.519*** 4.924*** 4.399*** -0.370***

(1.269) (1.279) (-11.55) (0.921) (0.930) ( -2.61) firm size 100-500 3.640*** 3.243*** 3.055*** 2.645*** (1.112) (1.112) (0.868) (0.868) firm size 500-1000 3.262*** 2.279* 2.457** 1.667 (1.128) (1.186) (0.996) (1.024) firm size 1000- 1.605 –0.564 1.859** 0.355 (1.144) (1.384) (0.944) (1.132) firm size control yes yes yes yes yes yes area control yes yes yes yes yes yes constant –19.949** –14.457* -14.347 –17.529*** –11.562* -21.382

(8.261) (8.315) ( -3.49) (6.157) (6.029) (-4.38) Sample size 804 804 6144 827 827 6045 Robust standard errors are in parentheses. ***indicates significance at 1% level; **indicates significance at 5% level; *indicates significance at 10% level. In Bell and Hunt’s model, they used t-value in parentheses. In addition, they used hourly wage instead of income.



-Male, Tenure and Experience Instead of Age

Variables (1) (2) (3) statutory fringe benefits 0.687** (0.323) non-statutory fringe benefits 2.540*** (0.937) total fringe benefits 0.715***

(0.275) tenure 0.024* 0.023* 0.023*

(0.013) (0.013) (0.013) tenure square –0.0001** –0.0001* –0.0001*

0.000 0.000 0.000 experience –0.582*** –0.569*** –0.569***

(0.175) (0.175) (0.175) experience square 0.009** 0.009** 0.009**

(0.004) (0.004) (0.004) married –0.516 –0.529 –0.499

(0.899) (0.899) (0.898) youngest child under 11 0.025 0.103 0.004

(0.874) (0.872) (0.872) union 3.339*** 3.014*** 3.171***

(0.908) (0.924) (0.912) log of income 3.594*** 3.329** 3.423***

(1.298) (1.308) (1.301) firm size 100-500 3.191*** 2.839** 3.139***

(1.118) (1.116) (1.114) firm size 500-1000 2.768** 1.836 2.505**

(1.146) (1.194) (1.148) firm size 1000- 0.885 –1.216 0.316

(1.163) (1.386) (1.177) area control yes yes yes constant –22.134*** –17.565** –21.380***

(7.388) (7.418) (7.343) Sample size 799 799 799 Robust standard errors are in parentheses. ***indicates significance at 1% level; **indicates significance at 5% level; *indicates significance at 10% level.

CONCLUSION

The purpose of this study was to obtain more insight into the effect of fringe benefits on paid overtime hours. To investigate the relationship, hypotheses were derived from quasi-fixed employment cost theory. Higher quasi-fixed employment costs such as fringe benefits were expected to have a positive impact on the paid overtime hours. In section 2, we discussed the motive for paid overtime work on the basis of quasi-fixed cost theory. No generally accepted theory exists with regard to paid overtime hours, and thus our empirical analysis of Japanese micro-data from the Fringe Benefits Survey provides useful and meaningful results.


Based on our analysis, it is believed that this article contributes to the quasi-fixed employment cost and overtime work literature in several ways. First, we use cross-sectional data on individuals, whereas most previous studies used only firm- or industry-level data. The use of data at the level of individuals allows us to control for a variety of supply-side factors affecting individual work decisions that cannot be accounted for with firm-level data. Second, most studies have focused on the health insurance, which is only part of total nonwage compensation, as a quasi-fixed cost effect. The quasi-fixed costs considered in our study included fringe benefits such as health insurance, a pension, employment insurance, housing, health care, living expenses, bereavement benefits, and leisure benefits as a quasi-fixed cost effect. Finally, we adapted Bell and Hart’s (1999) model to analyze the Survey of Company Fringe Benefits data for Japan, which include information on an employer’s provision of fringe benefits as well as paid overtime hours for individuals. With this framework, we can document more completely the structure of labor costs with paid overtime work. REFERENCES Baicker, K. & Chandra, A. (2006). The Labor Market Effects of Rising Health Insurance Premiums. Journal of Labor Economics, 24, (3), 609-634. Bell, D.N.F. & Hart, R. A. (1999). Unpaid Work. Economica, 66, 271-290. Cutler, D. & Madrian, B. (1998). Labour Market Responses to Rising Health Insurance Costs: Evidence on Hours Worked. RAND Journal of Economics, 29, (3), 509-530. Dolfin, S. (2006). An Examination of Firm’s Employment Costs. Applied Economics, 38, 861- 878. Ehrenberg, R.G. & Schumann, P.L. (1983). Compliance with the Overtime Pay Provisions of the Fair Labor Standards Act. Journal of Law & Economics, 25, (1), 159-181. Hart, R.A. & Ma, Y. (2010). Wage-Hours Contracts, Overtime Working and Premium Pay. Labour Economics, 17, (1), 170-179. Kaestner, R. & Simon, K. l. (2002). Labour Market Consequences of State Health Insurance Regulation. Industrial & Labor Relations Review, 56,(1), 136-159 . Oi, W. (1962). Labor as a Quasi-Fixed Factor. Journal of Political Economy, 70, 538-555. Thurston, K. (1997). Labor Market Effects of Hawaii’s Mandatory Employer-Provided Health Insurance. Industrial and Labor Relations Review, 51, (1), 117-136. Trejo, S. J. (1991). Compensating Differentials and Overtime Pay Regulations. American Economic Review, 81, 719–70. —— (1993). Overtime Pay, Overtime Hours, and Labour Unions. Journal of Labor Economics, 11, 253–278.


APPENDIX

TABLE A1 FRINGE BENEFITS AS A PROPORTION OF TOTAL COMPENSATION IN 2006

Total compensation

Wages and salaries 587658 84.87%

Nonwage benefits

Statutory fringe benefits

total 76437 11.04% health insurance 26031

pension 40657 employment insurance 9208

others 534

Non-statutory fringe benefits

total 28350 4.09% housing 13496 health care 3296 living expense 6301 bereavement allowance 924

leisure 2240 others 2098

Data source: The Survey of Company Fringe Benefits from Japan Business Federation in 2006


TABLE A2 QUESTIONS ON FRINGE BENEFITS IN THE 2006 SURVEY

What kind of fringe benefits does your company have? Mean 1 company housing 23.88% 2 bachelors’ dormitory 24.90% 3 housing allowance 41.28% 4 loan for purchasing your own house 15.59% 5 medical care subsidy in addition to health insurance 7.99% 6 subsidy for complete medical checkup 30.56% 7 disease screenings 26.61% 8 mental health (stress) consultation 21.00% 9 long-term disability benefits 10.19% 10 subsidy for childcare/babysitter (including night childcare center) 3.51% 11 day-care center 2.68% 12 childcare leave/short-time work 28.07% 13 support for nursing care helper (including subsidy) 1.41% 14 special payment for disaster/death 62.33% 15 retirement allowance if death occurs 31.43% 16 bereaved family pension 13.40% 17 asset-building savings/in-house savings deposits 43.03% 18 stock ownership 28.65% 19 stock option 5.51% 20 subsidy for club activities 13.84% 21 subsidy for or usage of resort house/fitness facilities 24.81% 22 life planning course 7.65% 23 asset management course 5.56% 24 retirement preparation education 8.72% 25 study abroad (or in domestic university) 5.60% 26 support for public qualification/subsidy for distance learning 18.27% 27 long leave for self-betterment 14.67% 28 company cafeteria 23.10%


TABLE A3 VARIABLE DEFINITIONS

Variable Description paid overtime number of paid overtime hours statutory fringe benefits cost of statutory fringe benefits (unit- Japanese Yen)

non-statutory fringe benefits cost of non-statutory fringe benefits (unit- Japanese Yen)

total fringe benefits cost of statutory fringe benefits and non-statutory fringe benefits (unit-Yen) male 1 if male, 0 if female

union 1 if the company has a union, 0 otherwise

income annual income (tax included,unit-10000Yen) age age

married 1 if married, 0 otherwise

child 1 if one has a child under 11 years old, 0 otherwise tenure months with current employer

experience age of individual minus age when completed full-time education

occupation 1=manager, 2=clerical, 3=sales, 4=laborer, 5=technical

FIGURE A1 GROWTH OF FRINGE BENEFITS

Data source: The Survey of Company Fringe Benefits in 2006


FIGURE A2-1 TOTAL WORKING HOURS IN JAPAN

(Average per Month)

Data source: Basic Survey of Wage Structure

FIGURE A2-2 OVERTIME HOURS WORKED IN JAPAN

(Average per Month)

Data source: Basic Survey of Wage Structure


m.€¦ · Journal of Applied Business and Economics Editors Dr. Adam Davidson Dr. William Johnson...

Documents

Transcript of m.€¦ · Journal of Applied Business and Economics Editors Dr. Adam Davidson Dr. William Johnson...