The Impact of Currency Derivative Usage and Earnings ... ANNUAL MEETINGS/2010-Aarhus old...The...
Transcript of The Impact of Currency Derivative Usage and Earnings ... ANNUAL MEETINGS/2010-Aarhus old...The...
The Impact of Currency Derivative Usage and Earnings Management on Firm Value
Feng-Yi Chang,a Chin-Wen Hsin,b and Shin-Rong Shiah-Houc
JEL classification: 180, 420, 610 Keywords: Hedging, Earnings management, Firm value aDepartment of Business Administration, China University of Technology, Taiwan bDepartment of Finance, Yuan Ze University, Taoyuan 32003, Taiwan, Tel.: +886-3-4638800 ext. 2662, Fax: +886-3-4553098, e-mail: [email protected] (corresponding author) cDepartment of Finance, Yuan Ze University, Taiwan.
1
Abstract
Firm managers can use financial hedging (real actions) or earnings management (artificial
techniques) to buffer economic shocks, such as exchange rate fluctuations, to their firm’s operating
performance. This study investigates the potential impact of foreign currency derivative (FCD)
usage and earnings management (EM) on firm value. Using a sample of S&P 500 non-financial
firms from 2001 to 2003, we find that there is a significantly positive impact of FCD usage on firm
market value. The premium from currency hedging is about 11% of firm value. On the other hand,
the contribution of earnings management on firm value only shows for small firms, which tend to
be associated with greater information asymmetry. A further exploration of the value increasing
mechanism indicates that FCD usage indeed increases firm value through its hedging effect on
firm exchange rate exposure. However, the same conclusion does not apply to earnings
management.
2
1. Introduction
In the classic Modigliani and Miller (M&M) world, the efforts in an attempt to enhance firm
value through risk management will come to nothing. Due to the presence of market imperfections,
the benefit of risk management should be reflected from firm market value. Most extant research
in optimal hedging theory concludes that hedging increases firm value by relaxing some M&M
assumptions [e.g., see Mayers and Smith (1982), Smith and Stulz (1985), Leland (1998), Froot,
Scharfstein, and Stein (1993)].1 Recent empirical studies have already documented that hedging is
a value-increasing strategy [Allayannis and Weston (2001), Kim et al. (2006) and Carter et al.
(2006)].2
In addition to hedging, prior research also suggests that firm managers can use their reporting
discretion (earnings management) to buffer economic shocks, such as undesired exchange rate
fluctuations [e.g., see Lambert (1984), DeFond and Park (1997), Barton (2000), Pincus and
Rajgopal (2002) and Leuz et al. (2003)].3 Dechow and Skinner (2000) argue that firm managers
manage earnings to maintain or improve their capital market valuation. Numerous rationales
underlying optimal hedging theory are also applicable to EM, especially those associated with
income smoothing.4 In light of this, we would also expect a positive impact of EM on firm value.
Hedging , categorized as real actions, reduces expected tax liabilities, financial distress costs
1 Prior studies provide mixed empirical supports for optimal hedging theory [e.g., see Nance, Smith, and Smithson (1993), Dolde (1995), Mian (1996), Tufano (1996), G´eczy et al. (1997), Visvanathan (1998), Haushalter (2000), Allayannis and Ofek (2001), Graham and Rogers (2002), and Carter et al. (2006)] 2 Contrary to other studies, Jin and Jorion (2006) do not find a significant impact of hedging on firm value for 119 U.S. oil and gas firms. 3 Lambert (1984) argues that firms have incentives to smooth income by accounting choices, in addition to real actions (hedging). Barton (2000) and Pincus and Rajgopal (2002) suggest that firm managers use financial hedging and EM as partial substitutes for smoothing earnings. 4 For example, Trueman and Titman (1988) argue that firms, by income smoothing, will reduce the probability of financial distress and lower borrowing costs. Similar to hedging, EM may lower the volatility of reported earnings and thus expected tax liabilities. Although EM cannot ensure sufficient internal generated funds for investment opportunities, it does help mitigate the underinvestment problem by building reputations through which firms can create access to external financing or lower borrowing costs.
3
and mitigates underinvestment problems by maintaining a smoother pattern of cash flows. EM
differs from hedging as it only affects reported earnings and is termed as being an artificial
technique [Lambert (1984) and Albrecht and Richardson (1990)]. Although accounting academics
are unwilling to view EM as problematic, investors and regulators often perceive EM as
opportunistic [see Dechow and Skinner (2000) and Beaver (2000)]5 This study investigates the
potential valuation effect of EM and financial hedging. In particular, we examine (1) the possible
impact of FCD usage and EM on firm value; and (2) how FCD usage and EM contribute to value
increasing?
Similar to Allayannis and Weston (2001), our analyses are conducted in the context of
exchange rate risk management.6 Using a sample of S&P 500 non-financial firms from 2001 to
2003, we first show the possible impact of firm size on the use of FCDs and EM. 7 Following
Leuz, Nanda and Wysocki (2003), we develop four measures to capture the different dimensions
of EM at the firm level.8 We find that firms with a larger size exercise lesser EM. As suggested by
Albrecht and Richardson (1990), this may be due to larger firms being subject to greater scrutiny
by the public. Consistent with Barton (2000) and Pincus and Rajgopal (2002), we also find a
partial support for the substitute relation between FCDs usage and EM since firms in the largest
5 Accounting academics generally argue that EM doesn’t matter under the efficient market hypothesis. If market participants have low cost access to the requisite information, they will observe that EM is occurring and make adjustments [see Dechow and Skinner (2000)]. Beaver (2000) also suggest that EM is problematic only if the motivation is opportunistic. 6 We focus on foreign currency management for reasons as follows: (1) U.S. corporations, including those with no foreign operations and no foreign currency assets, liabilities, or transactions, are generally exposed to foreign currency risk [Adler and Dumas (1984)]; (2) foreign currency derivatives are the most widely used derivatives [Allayannis and Weston (2001)]; and (3) we intend to examine whether hedging (or EM) increases firm value by reducing exchange rate exposure. 7 Smith and Stulz (1985) suggest that smaller firms are more likely to hedge since the reduction in expected costs of financial distress is greater for smaller firms. Nance, Smith, and Smithson (1993) offer that a firm’s hedging decision depends on the economies of scale in implementing a risk management program. Mose (1987) and Albrecht and Richardson (1990) present that firm size may affect firm managers’ incentives for EM. 8 Leuz, Nanda and Wysocki (2003) develop four country-level measures to capture various dimensions of earnings management, including income smoothing, discretion in reported earnings and small loss avoidance. We design four measures (EMsmooth, EMcorrelation, EMDAC and EMsmall-loss) at firm level. EMsmooth and EMcorrelation are used to proxy for the degree of income smoothing. EMDAC is used to measure the magnitude of discretionary accruals. EMsmall-loss is used to measure small loss aversion.
4
size quintile use more FCDs and practice lesser EM.
We further examine the possible effect of FCD usage and EM on firm value (Tobin’s Q).
Confirming prior studies [e.g., see Allayannis and Weston (2001) and Kim et al. (2006)], we find
that firms using FCDs have higher Tobin’s Q. The hedging premium is in the range from 11.61%
to 12.62% of firm value, as compared with about 5% in Allayannis and Weston (2001) and about
5% - 10% in Carter et al. (2006).
The four EM measures exhibit different patterns in relation to firm market value. Although
income smoothing functions like hedging activities, we do not find that income smoothing has a
significantly positive impact on firm value. More specifically, we find a mainly negative relation
between Tobin’s Q and the income smoothing measures, suggesting that capital market
distinguishes the valuation effect of EM from that of financial hedging. However, we find a
significantly positive relation between Tobin’s Q and the other two EM measures: EMDAC and
EMsmall-loss. Although firm managers can use discretionary accruals for a variety of purposes, we
would expect that the use of discretionary accruals also increases firm value if EM is more likely
used to maintain or improve capital market valuation.
Since about 70% of our sample firms are diversified across more than one business segment,
we also construct industry-adjusted Q to measure firm value while controlling for the possible
industrial difference. We similarly find a significantly positive impact of FCDs usage on
industry-adjusted Q, while the impact of income smoothing measures is still insignificant.
Although we find that investors tend to distinguish the valuation effect of income smoothing
from that of financial hedging, other EM measures exhibit mixed results. We further investigate
whether the impact of EM differs depending on firm size, a proxy for information asymmetry.
Based on the efficient market hypothesis, accounting academics generally believe that EM doesn’t
matter. Such argument, however, implies that the valuation effect of EM depends on information
5
asymmetry. As suggested by Albrecht and Richardson (1990) whereby firms with a larger size are
subject to greater scrutiny by the public, artificially manipulated earnings are more likely detected
and penalized for larger firms. We further hypothesize that the impact of EM on value-adding is
smaller for larger firms. Although our sample firms are relatively larger, we indeed find that EM
exhibits a different effect on firm value between smaller and larger firms. Income smoothing,
which functions as hedging, only significantly increases firm value for smaller firms. Conversely,
we find a significantly negative impact of income smoothing on firm value for larger firms.
Similar results are found for other EM measures.
Finally, we demonstrate that lower stock return exposure to exchange rate risk leads to higher
firm value. Since prior studies have already documented that both FCD usage and EM mitigate
firm-specific exchange rate exposure, implying that hedging and EM may contribute to increasing
firm value by reducing exchange rate exposure. However, we find that the improvement in firm
value does not apply to the case where the exposure to exchange rate fluctuations is mitigated
through EM.
This study makes the following contributions to existing literature: First, we confirm with
more recent data that currency derivative usage indeed increases firm value. Second, we provide
new findings that EM may also increase firm value, however, the valuation effect of EM depends
on the possible information asymmetry. Third, we provide new findings that lower exchange rate
exposure leads to higher firm value. The evidence explains the possible channel through which
foreign currency hedging or EM may contribute to increasing firm value. Last, this study has
important implication for firm managers that foreign currency risk management may improve firm
value, however, the improvement in capital market valuation is unable to be achieved through EM.
The remainder of this paper is organized as follows. Section 2 reviews the extant literature
and develops research hypotheses. Section 3 describes our methodology to measure firm value and
6
EM. Section 4 reports the summary statistics. Section 5 presents the possible impact of firm size
on FCD usage and EM. Section 6 compares the effect of FCDs and EM on firm value. Section 7
explores whether the impact of EM on firm value depends on firm size, and section 8 concludes.
2. Hypothesis Development
2.1 Hedging and Firm Value
The classic M&M Theorem implies that risk management is irrelevant to firm value when the
financial markets are perfect. The assumptions underlying the theorem include the absence of
income taxes, transaction costs, bankruptcy costs, asset trade restriction, asymmetric information,
and imperfections of financial markets. However, real financial markets are not always frictionless,
and risk management thus may affect the value of a firm. Indeed, recent theories of optimal
hedging generally derive that hedging is optimal by relaxing one or more of the M&M Therom’s
assumptions:
Taxes
Mayers and Smith (1982) and Smith and Stulz (1985) suggest that the structure of the tax
code may create incentives for firms to hedge, especially when corporate taxes are a convex
function of earnings.9 The convexity of the tax code structure implies that a higher volatility of
taxable earnings stream induces higher expected taxes than a smoother earnings stream. If hedging
effectively reduces the volatility of the earnings stream, then the expected corporate tax liability
drops and the value of firm increases, given the assumption that the costs of hedging do not exceed
9 See also Graham and Smith (1999).
7
the hedging benefits.
Bankruptcy Costs and Debt Capacity
Smith and Stulz (1985) argue that the transaction costs of bankruptcy may prompt firms to
hedge. By reducing the volatility of cash flows or earnings, hedging can lower the probability and
expected costs of financial distress. Hedging may reduce the expected costs of bankruptcy in
numerous ways, and in each case firm value increases corresponding to the decrease in bankruptcy
costs. First, firms with volatile cash flows face a greater probability of going bankrupt. Hedging
reduces the direct bankruptcy costs by maintaining a smoother cash flows stream and thus lowers
the probability that debt or interest payments are unable to be repaid [see Smith and Stulz (1985)
and Haushalter (2000)]. Second, as suggested by Smith and Stulz (1985), if a firm heavily relies
on external financing, then hedging helps building a reputation and thus increases the price for its
new debt. Third, hedging reduces the costs of financial distress associated with bond covenants
that may constrain firms’ financial actions [see Smith and Stulz (1985)].
Stulz (1996) and Leland (1998) suggest that hedging permits greater debt capacity by
reducing the volatility of cash flows, and the tax shields generated from the increasing leverage
will add firm value. Consistent with such an argument, Graham and Rogers (2002) find that firms
hedge to increase debt capacity.
Underinvestment
Smith and Stulz (1985) suggest that hedging lessens the influence of binding bond covenants,
which may constrain a firm’s investment decision. Myers (1977) argues that financial distress may
force firms to bypass investments with positive NPV. Such arguments implicitly link hedging and
underinvestment problems. Based on Smith and Stulz’s (1985) bankruptcy cost rationale for risk
8
management, Froot, Scharfstein, and Stein (1993) develop a framework for analyzing the benefits
of corporate hedging. In their model firms with investment opportunities are more likely to
underinvest when external financing is too expensive than internal generated funds. By reducing
the volatility of cash flows, hedging helps ensure a sufficient amount of internal generated funds
for attractive investment opportunities and thus relieves the underinvestment problem. Hedging
enhances firm value to the extent that the underinvestment problem is lessened.
Although the theories of optimal hedging generally conclude that hedging can increase firm
value, Tufano (1998) argues that risk management may also impose important costs on firm value
when agency conflicts are considered. In other words, although hedging ensures sufficient internal
funds for supporting investment opportunities, there is no promise that investment opportunities
are all attractive. When firm mangers with selfish objectives tend to accept poor investment
opportunities, hedging isolates those poor projects from external monitoring by providing
sufficient internal funds. Therefore, hedging may destroy firm value when agency conflicts
between firm managers and shareholders exist.
Other Rationales for Hedging (Irrelevant to Firm Value)
Stulz (1984) and Smith and Stulz (1985) suggest that firm managers’ risk aversion may lead
to corporate hedging. When the wealth of risk-adverse managers is largely concentrated in their
firm’s stock and it is less costly to hedge through the firm, the risk-adverse managers will direct
the firm to hedge. The information content also creates incentives for firm managers to hedge.
DeMarzo and Duffie (1995) offer that financial hedging delivers a signal of managerial ability and
firm managers’ private information on future earnings. They also show that firm managers’
hedging incentive depends on the disclosure criterion of hedging. The economies of scale in
implementing a risk management program may also affect a firm’s hedging decision [e.g., see
9
Nance, Smith and Smithson (1993)]. Prior empirical studies find a significantly positive
association between firm size and hedging [e.g., see Geczy et al. (1997), Mian (1996), Haushalter
(2000), Allayannis and Ofek (2001), Graham and Rogers (2002), and Kim et al. (2006)].
Recent empirical studies in examining the possible impact of hedging generally suggest that
hedging increases firm market value. Allayannis and Weston (2001) first document a significantly
positive relation between the use of FCDs and firm value (Tobin’s Q), and the hedging premium is
on average 4.87% of firm value. Similar to Allayannis and Weston (2001), Kim et al. (2006) show
that hedging, including financial and operational hedging, significantly increases firm value.
Focusing on the U.S. airline industry, Carter et al. (2006) find that jet fuel hedging is positively
associated with firm value and the hedging premium is approximately 5%-10%. Contrary to other
studies, Jin and Jorion (2006) fail to see a positive effect of hedging on firm market value for a
sample of 119 U.S. oil and gas producers. Our first research hypothesis is as follows.
H1: Firm value is increased through FCD usage.
2.2 EM and Firm Value
Although EM may occur in a variety of forms,10 prior research associated with the influence
of EM on firm value generally focuses on income smoothing. Certain rationales underlying
optimal hedging theories also are applicable to income smoothing. For example, Trueman and
Titman (1988) suggest that by smoothing reported earnings, firm managers are able to reduce the
assessment of the firm’s various claimants about the probability of financial distress. EM for
income smoothing purpose thus increases firm value by reducing the firm’s borrowing costs and
by facilitating the transactions between the firm and its stakeholders, e.g., customs, workers, and
10 E.g., income smoothing, big-bath, cookie jar reserve, or income increasing.
10
suppliers, etc. Supporting Trueman and Titman (1988), Francis et al. (2004) find that firms
engaging more in earnings smoothing have a lower cost of capital.
Other rationales underlying the theories of optimal hedging also can be applied to EM. If EM
for income smoothing purpose effectively lowers the volatility of reported earnings, then the
expected tax liabilities are reduced and firm value is increased. Unlike hedging, EM is unable to
ensure sufficient internal generated funds for investment opportunities. However, if the efforts of
income smoothing lead to a observed smoother pattern of earnings, then EM also helps mitigate
the underinvestment problem by building reputations through which firms can create access to
external financing or lower borrowing costs.
Bitner and Dolan (1996) extend the Trueman and Titman (1988) rationale and suggest equity
market valuation as a motivation for income smoothing. In particular, Bitner and Dolan (1996)
argue that the volatility of income streams affects the risk-adjusted discount rate. If smoother
income streams lead investors to apply a lower risk-adjusted discount rate to a firm’s future cash
flows and debt payments, then a higher firm value is expected.11
Without focusing on income smoothing, Dechow and Skinner (2000) suggest that firm
managers have incentives to manage earnings in order to maintain or improve their capital market
valuation. Prior research analyzing the preference of investors or analysts also provides possible
incentives and the possible valuation effect of EM. O’Brien and Bhushan (1990) demonstrate that
analysts prefer to follow firms with a lower volatility of earnings. Michelson, Jordan-Wagner and
Wootton (2000) argue that firms with lower earnings volatility are more attractive to investors than
firms with higher earnings volatility. Firm managers thus have incentives to attract investors and
analysts by EM or hedging. The increase in the following by investors or analysts may enhance
firm market value. Our second hypothesis is as follows.
11 Contrary to Bitner and Dolan (1996), Beidleman (1973), Lev and Kunitzky (1974) and Michelson (1995) suggest that the smoother earnings stream indicates a lower risk and thereby lower firm value.
11
H2: Firm value is increased through EM.
Hedging directly reduces the volatility of cash flows and therefore the volatility of reported
earnings, while EM only alters reported earnings and is termed as being an artificial technique [see
Lambert (1984) and Albrecht and Richardson]. Accounting academics and investors tend to have
different perceptions about EM. Dechow and Skinner (2000) state that “…practitioners and
regulators often see earnings management as pervasive and problematic—and in need of
immediate remedial action. Academics are more sanguine, unwilling to believe that earnings
management is actively practiced by most firms or that the earnings management that does exist
should necessarily concern investors.” Under the efficient market hypothesis, academics may
argue that EM doesn’t matter if the requisite information is fully disclosed and thus investors will
observe that EM is occurring and make reaction. Trueman and Titman (1988) and Bitner and
Dolan (1996) similarly argue that EM is temporal and detectable. In light of this, the argument -
that EM increases firm value by reducing bankruptcy costs, expected tax liabilities, and
underinvestment problem - depends on the extent to which the capital market cannot fully
distinguish the observed smooth earnings stream that is naturally generated from those that are
artificially manipulated with opportunistic motivations.12 Bitner and Dolan (1996) suggest that
artificially manipulated earnings will be detected and discounted if sufficient information is
available. Balsam, Bartov and Marquardt (2002) provide evidence that investors detect EM when
additional information arrives. Therefore, we expect that the valuation effect of EM should be
smaller than that of hedging and that the extent to which EM affects firm value depends on
information asymmetry.
12 Note that EM is not always problematic. Whether EM is problematic depends on firm managers’ motivation.
12
Firm size is a typical proxy for information asymmetry [McLaughlin et al. (1998)]. Albrecht
and Richardson (1990) also suggest that firms with a larger size exhibit a lesser degree of EM for
income smoothing since larger firms are subject to greater scrutiny by the public.13 In light of this,
we would expect that EM does not exhibit a strong impact on firm value for larger firms. Our third
and fourth hypotheses are as follows.
H2A: The increase in firm value is smaller through EM.
H2B: Firm value is increased through EM only for smaller firms.
Most empirical studies examining the possible effect of EM on firm value focus on income
smoothing and generally provide mixed results. Bitner and Dolan (1996) find that the smoothness
of reported earnings significantly increases firm market value (Tobin’s Q). Contrary to Bitner and
Dolan (1996), Allayannis, Rountree and Weston (2005) fail to demonstrate that earnings
smoothness is positively associated with firm value. Bao and Bao (2004) show that whether
income smoothing guarantees higher firm value (price–earnings multiple) depends on earnings
quality.
Numerous studies have already examined the possible relation between EM and stock
performance. For example, Michelson et al. (1995) find that U.S. smoothers have a lower ten-year
annualized return than non-smoothers. Michelson et al. (2000) offer that firms reporting smoother
earnings have significantly higher cumulative average abnormal returns than firms that do not.
Balsam, Bartov and Marquardt (2002) document a negative relation between unexpected
discretionary accruals and cumulated abnormal returns around the 10-Q filing date. In addition, a
rich body of studies have investigated the association between EM and stock performance around
13 Conversely, Moses (1987) suggests that larger firm have greater incentives to smooth earnings.
13
certain corporate events [e.g., see Perry and Williams (1994), Teoh, Welch and Wong (1998a, b),
Rangan (1998) and Kim and Park (2005)]. However, higher stock performance does not
necessarily guarantee higher market value.
3. Methodology
3.1 Measuring Firm Market Value
Following prior works [e.g., see Allayannis and Weston (2001), Jin and Jorion (2006), Carter
et al. (2006), and Kim et al. (2006)], we use Tobin’s Q as a proxy for firm value. Tobin’s Q is
defined as the ratio of the market value of the firm’s assets to the replacement cost of the firm’s
assets [Chung and Pruitt (1994)]:
Tobin’s Q = ( MVE + PS + DEBT ) / TA (1)
where MVE is the market value of a firm’s common equity that is measured as the product of a
firm’s stock price and the number of common shares outstanding in the fiscal year end, PS is the
liquidating value of the firm’s preferred stock, DEBT is calculated as the firm’s short-term
liabilities net of its short-term assets plus the book value of the firm’s long term debt, and TA is the
book value of the firm’s total assets.14
Following Lang and Stulz (1994) and Allayannis and Weston (2001), we also construct a
measure of industry-adjusted Tobin’s Q. Assuming that a firm engages operations in n business
segments, the industry-adjusted Q is defined as the difference between raw Tobin’s Q and
pure-play Q as follows: 14 Chung and Pruitt (1994) suggest that the approximate Tobin’s Q explains 96.6% of the total variability in Lindenberg and Ross (1981) Tobin’s Q
14
⎟⎠
⎞⎜⎝
⎛−⎟⎟
⎠
⎞⎜⎜⎝
⎛=− ∑
=
n
iiiQQsTobinQadjustedIndustry
1' α , (2)
TA
Assetsii =α , (3)
where Tobin’s Q is defined as equation (1), ∑=
n
iiiQ
1α is pure-play Q [see Lang and Stulz (1994)],
Qi is the Tobin’s Q for the firm’s ith business segment and is measured by the median Tobin’s Q of
all single-segment firms with the same three-digit SIC in COMPUSTAT, and iα is the weight of
the assets of firm’s ith business segment to its total assets.
3.2 Measures of EM in Different Dimensions
In this section, we design four EM measures at the firm level: EMsmooth, EMcorrelation, EMDAC
and EMsmall-loss. Following Leuz et al. (2003), the four measures evaluate the degree of EM in
different dimensions, including income smoothing, discretion in reported earnings, and small-loss
aversion. Since certain rationales underlying optimal hedging theory are also applicable to income
smoothing, indicating a potential positive impact on firm value. As suggested by Dechow and
Skinner (2000) that firm managers manage earnings in order to maintain or improve their capital
market valuation, we also hypothesize that EM other than income smoothing has impact on firm
value.
EMsmooth – measures the degree of income smoothing by analyzing the relative volatility of
reported earnings and operating cash flows
15
One of the general forms of EM is income smoothing [Dechev and Skinner (2000)]. Barton
(2001) suggests that firm managers encounter increasing pressure to maintain a smoother earnings
pattern. If a firm’s exposure to exchange rate risk is unhedged (or not well-managed), then the
fluctuations in the exchange rate will increase the earnings volatility and induce firm managers to
engage in earnings smoothing activities. Unlike hedging, the use of EM has no effect on cash
flows. EM associated with an income smoothing purpose can be detected by comparing the
volatility of reported earnings with some benchmark, e.g., cash flows [see Leuz et al. (2003) and
Ecker et al. (2006)]. Following Leuz et al. (2003) and Ecker et al. (2006), we use the ratio of the
standard deviation of operating earnings to the standard deviation of operating cash flows as our
first income smoothing measure. If income smoothing occurs, then reported earnings will become
less volatile than operating cash flows and thus lead to a lower ratio. The ratio is multiplied by -1
to obtain a positive correlation with the EM level. EMsmooth is calculated as follows.
( )1
1,
1, −×⎟⎠⎞
⎜⎝⎛
⎟⎠⎞
⎜⎝⎛
=
−
−
ti
it
ti
it
smooth
TAOANCF
TAIB
EMσ
σ, (4)
where IB is the quarterly earnings before extraordinary items and discontinued operations,
OANCF is quarterly operating cash flows, and TA is quarterly total assets. Operating earnings and
operating cash flows are scaled by lagged total assets to control for the effect of firm size. The
standard deviations of quarterly earnings (IB) and quarterly cash flows (OANCF) are calculated
from a three-year period (t-2 to t) ahead of the estimation of Tobin’s Q in fiscal year-end (year t).
A higher EMsmooth ratio thus implies that firm managers exercise greater income smoothing and is
expected to be positively associated with firm market value.
16
EMcorrelation – measures the degree of income smoothing by calculating the correlation between
changes in accruals and changes in operating cash flows
To mitigate the influence of certain economic shocks, e.g., undesired exchange rate
movements, firm managers may use their accounting discretion to shift income from one period to
another. Such a form of EM is generally achieved by accelerating or delaying the recognition of
revenues or expenses and therefore results in a reversal effect in the next period (or future periods).
Although a negative correlation between accruals and cash flows is a natural result of accrual
accounting, a greater manipulation of accruals to diminish economic shocks to operating cash
flows will make the correlation coefficient more negative. In light of this, our second income
smoothing measure, EMcorrelation, is the correlation coefficient between changes in accruals and
changes in operating cash flows. Similar to EMsmooth, the correlation coefficient is multiplied by -1
to make this measure a positively correlated indicator for EM.
( ) ( )1OANCF ,Accruals −×ΔΔ= ρncorrelatioEM (5)
Quarterly total accruals and operating cash flows are used to calculate EMcorrelation for the
3-year period (t-2 to t) ahead of the estimation of Tobin’s Q in fiscal year-end (year t). Both
changes in accruals and changes in cash flows are scaled by lagged total assets. A larger
EMcorrelation value thus indicates that firm managers exercise greater income smoothing.
EMDAC - the magnitude of discretionary accruals
Other than income smoothing, firm managers can also use discretionary accruals to alter a
period’s (or certain periods’) reported earnings. We measure discretionary accruals by employing
a cross-sectional modified Jones model [e.g., see Jones (1991) and Dechow, Sloan and Sweeney
17
(1995)]. Discretionary accruals are defined as the difference between total accruals and
non-discretionary accruals.15 To estimate the non-discretionary accruals (NDAC), we use the
following regression:
jtti
ti
ti
titi
titi
ti
TAPPE
TARECREV
TATATAC
εααα ++Δ−Δ
+=−−−−
)()()1(1,
,2
1,
,,1
1,0
1,
, , (6)
where TAi,t-1 is the total assets of firm i in year t – 1, TACi,t is total accruals of firm i in year t and is
calculated as income before extraordinary items and discontinued operations less operating cash
flows, ΔREVi,t is change in sales in year t for firm i, ΔRECi,t is change in accounts receivable in
year t for firm i, and PPEi,t is the gross property, plant, and equipment of firm j in year t. The
firm-specific parameters, α0, α1, and α2, in equation (6) are generated using OLS regression by the
two-digit SIC code. We require a minimum of 15 observations for each industry regression. The
estimated 0α̂ , 1α̂ , and 2α̂ are then used to determine non-discretionary accruals.
The non-discretionary accruals scaled by assets (NDAC) is computed as follows:
)(ˆ)(ˆ)1(ˆ1,
,2
1,
,1
1,0,
−−−
+Δ−Δ
+=ti
ti
ti
tiit
titi TA
PPETA
RECREVTA
NDAC ααα (7)
and the discretionary accruals scaled by assets (DAC) for firm i in year t is measured as:
titi
titi NDAC
TATAC
DAC ,,
,, )( −= , (8)
15 We measure total accruals as earnings before extraordinary items and discontinued operations less operating cash flows. Not all the accruals are derived from earnings management. Non-discretionary accruals represent the normal level of total accruals that are necessary and are associated with sales and investments of fixed assets.
18
We use the 3-year average of the absolute value of DAC to measure the magnitude of
discretionary accruals for the 3-year period ahead of the estimation of Tobin’s Q in the fiscal
year-end (t). EMDAC is calculated as follows:
( )3
2,∑
−==
t
tyyi
DAC
DACEM (9)
As suggested by Kothari et al. (2005), we also include ROAi,t-1 in equation (6) to estimate
EMDAC and obtain similar results. Note that firm managers may use discretionary accruals for
purposes other than income smoothing and this may induce noise into the relation between EMDAC
and firm market value. However, as suggested by Dechow and Skinner (2000), if firm managers
are more likely to exercise discretionary accruals for maintaining or improving their capital market
valuation, then we would expect a positive association between EMDAC and firm value.
EMsmall-loss – degree of small-loss aversion
Prior studies suggest that EM may be exercised to avoid reporting a small loss [e.g., see Hayn
(1995), Degeoreg et al. (1999) and Burgstahler and Dichev (1997)]. Burgstahler and Dichev (1997)
and Leuz, Nanda, and Wysocki (2003) use the ratio of small profit firms to small loss firms to
measure small loss avoidance through EM at the country level.16 In order to adapt this measure to
our firm level assessment of EM, we use the ratio of the number of quarters with small loss
reporting to the total number of quarters with non-missing profits and losses as a measure of a
firm’s tendency to avoid reporting small losses.
16 Dechow et al. (2003) make a caution of this measure as they cannot find evidence that firms utilize discretionary accruals to prevent small-loss reporting.
19
quarters totalof # reporting loss small with quarters of #
=−losssmallEM (10)
To calculate EMsmall-loss, the sample firm is required to have at least ten quarters of data with
non-missing profits or losses within the 3-year (t-2 to t) period ahead of the calculation of firm
value in the fiscal year-end. Since small losses are more likely to be altered by EM than large
losses, firms that do not avoid recognizing small losses in their accounting reports are regarded as
engaging less in EM. A firm is defined as reporting a small quarterly loss if the ratio of quarterly
net income to total assets is in the range [−0.01, 0.00). Similar to EMsmooth, the ratio is multiplied
by -1. A higher score of EMsmall-loss thus implies a higher level of EM. If a continuity of reporting
profits helps build reputations, then we would expect EMsmall-loss to be associated with higher firm
value.
3.3 Regression Framework
We examine the possible impact of FCD usage and EM on firm market value using a multiple
regression model. Following prior studies [e.g., see Allayannis and Weston (2001), Kim et al.
(2006), Carter et al. (2006), and Jin and Jorion (2006)], we use Tobin’s Q as a proxy for firm
market value. In addition to FCD usage and EM, we also include control variables that could have
an impact on firm value: size, access to financial markets, leverage, profitability, growth
opportunity, industrial diversification, geographic diversification, credit rating, and time-effect.
The regression model is as follows:
( )
titittitititi
tijtijtitititi
DCDEMGEOSEGDIV
CAPXRDRLeverageROASIZEDQsTobin
,,10,2,9,8,7,6
,,5,,4,3,2,10,
& ' ln
εγγγγγ
γγγγγγ
++++++
+++++=
−
∑ , (11)
20
where Tobin’s Q is estimated by equation (1). SIZEi,t is the logarithm of a firm’s total assets at
fiscal year-end. ROA is the pre-tax return on total assets. Leverage is measured by total debt as a
percentage of total assets. R&D is the R&D expense as a percentage of annual sales. CAPXR is the
capital expenditure as a percentage of annual sales. DIV is the dividend yield. GEO is the
logarithm of a firm’s total geographic segments. SEG is the logarithm of a firm’s total business
segments. DCD equals 1 if the firm discloses the use of any type of FCDs (e.g., forward contracts,
options, futures or swaps), otherwise 0. EM denotes the EM measure and is calculated for a 3-year
window (t-2 to t) ahead of the estimation of Tobin’s Q in the fiscal year-end (t).
4. Sample Selection and Descriptive Statistics
We conduct our investigation with an initial sample of S&P 500 non-financial firms listed in
2002. Financial firms are excluded from our sample since they may use currency derivatives for
purposes other than hedging, e.g., trading or speculation. Only those firm-year observations with
non-missing Tobin’s Q between 2001 and 2003 are included in our analysis.
We obtain the information about the use of FCDs (e.g., currency forward contracts, options,
futures and swaps) for our sample firms from the 2001 to 2003 10-K reports.17 We gather the
10-K reports from the EDGAR database. Throughout this study, we perform our analyses mainly
17 The information about financial derivatives usage is normally reported in item 7A (quantitative and qualitative disclosures about market risk) and in footnotes associated with financial derivative instruments. In fact, we comprehensively searched the entire content of 10-K reports using all possible keywords, e.g., “hedg”, derivative, forward, options, futures, swap, risk, “currenc”, exchange rate, foreign, etc.
21
based on a binary dummy variable, DCD, which indicates whether a firm uses FCDs during a
fiscal year. DCD equals 1 if a firm discloses the use of any type of FCD in a fiscal year, otherwise
0. Barton (2001) suggests that the notional amount is the better measure to represent a firm’s
financial hedging, we also collect the fiscal year-end notional amount of FCDs when such
information is reported. However, as SFAS No.133 requires firms to measure their derivative
instruments at fair value for all fiscal years beginning after June 15, 2000,18 numerous firms
decide not to disclose the notional amount of their financial instruments.19 Since a derivative’s
fair value is generally much smaller than its notional amount, it is not reasonable to measure a
firm’s financial hedging by the fair value of FCDs held in the fiscal year-end.
[ Insert Table 1 Here ]
Table 1 describes the summary statistics for our sample consisting of 398 non-financial S&P
500 firms. Panel A reports the descriptive statistics for the main firm characteristics. Our sample
firms have a mean (median) value of total assets of $16,845 (7,116) million. The mean (median)
value of total sales is $12,460 (5,611) million. On average, our sample firms conduct operations in
about 4 geographic segments with a range from 2 to 5. Our sample firms also conduct operations
in about 4 business segments with a range from 1 to 10.
In this study Tobin’s Q is used as a proxy for firm market value. Our sample firms have a
mean (median) Tobin’s Q of 1.912 (1.349). As the majority of our sample firms is diversified
across different industrial segments, we also construct industry-adjusted Q to account for the
potential influence from industrial difference. The mean (median) value of industry-adjusted Q is 18 The Financial Accounting Standards Board (FASB) issued SFAS No. 133, Accounting for Derivative Instruments and Hedging Activities, in June 1998. SFAS No. 133, which became effective for all fiscal years beginning after June 15, 2000, establishes accounting and reporting standards of derivative instruments for hedging activities. 19 Only 35.68% of our sample firms report notional amount greater than zero in 2001, as compared with 34.42% in 2002 and 34.01% in 2003.
22
0.805 (0.484).
Panel B of Table 1 reports the information of financial hedging with FCDs. Approximately
66% of the sample firms report the use of FCDs (mean value of DCD is 0.662). The percentage of
firms reporting the use of FCDs is greater than previous studies,20 indicating the increasing
importance of foreign currency hedging in recent years. However, only about 34.9% (414/1186) of
firm-year observations report non-zero notional amount of FCDs and the mean (median) value of
notional amount held in the fiscal year-end is $636.79 (190.33) million.
Table 1, Panel C reports the descriptive statistic of the EM measures used in this study. All
the four EM measures are calculated at firm level. EMsmooth, defined as the volatility of reported
earnings to the volatility of cash flows, has a mean (median) value of -0.366 (-0.213). This ratio
indicates that the reported earnings are much smoother than operating cash flows, and that income
smoothing may exist. The mean (median) value of EMcorrelation is 0.937 (0.988). The mean (median)
value of EMDAC is 0.0618 (0.0404).Consistent with prior studies [e.g., Hayn (1995), Degeoreg et al
(1999) and Burgstahler and Dichev (1997)], most firms never report a quarterly small loss.
5. FCDs or EM: the size effect
Hedging policies as well as EM strategies may be affected by firm size. Smith and Stulz
(1985) suggest that smaller firms tend to hedge since the reduction in expected costs of financial
distress is greater for smaller firms. Nance, Smith and Smithson (1993) argue that a firm’s hedging
decision depends on the economies of scale in implementing a risk management program and that
larger firms are more likely to hedge. Albrecht and Richardson (1990) suggest that firms with a
larger size exhibit a lesser degree of income smoothing since larger firms are subject to greater
20 For example, 42.6% in Allayannis and Ofek (2001), 32~40% in Allayannis and Weston (2001), and 37.4% in Bartram et al. (2004).
23
scrutiny by the public.
In this section we analyze the potential influence of firm size on the use of FCDs and EM.
The sample firms are divided into quintiles based on firm size. We use the fiscal year-end total
assets to measure firm size. Table 2, Panel A shows the percentage of firms reporting the use of
FCDs for each size quintile. We find that 78.6% of firms within the largest size quintile report the
use of FCDs, as compared with 66% for the smallest size quintile. The difference is significant at
the 0.01 level and is consistent with the economies of scales consideration in the hedging decision.
However, we do not find that the percentage of firms using FCDs declines monotonically from
largest size quintile to smallest size quintile. It also seems to not be contrary to Smith and Stulz’s
(1985) suggestion, since the percentage of firms using FCDs for the smallest size quintile is
greater than other quintiles, only with the exception of the largest size quintile.
[ Insert Table 2 Here ]
Table 2, panel B presents the comparisons of the four EM measures across size quintiles. For
the first income smoothing measure, EMsmooth, we find that firms in the smallest size quintile
exercise more income smoothing than firms in the largest size quintile. The mean (median) value
of EMsmooth is -0.301 (-0.219) for the smallest size quintile, which is greater than the mean (median)
of -0.308 (-0.246) for firms in the largest size quintile (the difference is not significant).
Interestingly, neither firms in the smallest quintile nor firms in the largest quintile have the greatest
level of EMsmooth. Instead, the degree of income smoothing is greater for firms in the middle three
size quintiles. As compared with the evidence from Panel A, we find that firms in the largest and
smallest size quintiles have greater percentages of hedging and exercise less income smoothing,
while firms in the middle size quintile use less FCDs and more income smoothing. This evidence
24
thus indicates a possible substitute relation between hedging and EM as suggested by Barton (2000)
and Pincus and Rajgopal (2002). We do not find a similar pattern for the second income
smoothing measure, EMcorrelation. In fact, the mean and median of EMcorrelation do not differ too
much across size quintiles.
Panel B also shows that firms in the smallest size quintile exercise the greatest level of
discretionary accruals (the mean (median) value of EMDAC equals 0.096 (0.063)), while firms in
the largest quintile exercise the lowest level of discretionary accruals (the mean (median) equals
0.044 (0.030)). The difference is significant at the 0.01 level. Similarly, EMDAC declines
monotonically from the smallest size quintile to the largest size quintile. Barton (2000) provides
evidence that firm managers use discretionary accruals for earnings smoothing. Leuz et al. (2003)
find consistent results among their four EM measures. Inconsistent with Barton (2000) and Leuz et
al. (2003), we do not find that firms exercising a greater magnitude of discretionary accruals have
higher income smoothing measures. This may be attributable to the reason that firm managers may
exercise discretionary accruals for purposes other than income smoothing.
The comparison of the small-loss aversion measure (EMsmall-loss) across size quintiles is
shown in the bottom of Panel B. Firms of the largest size quintile significantly report more
quarterly small losses than firms of the smallest size quintile. This indicates that larger firms have
more courage to recognize small losses or, in fact, larger firms suffer more pressure to disclose
true information of firm performance.
We so far find that larger firms significantly exercise lesser discretion to manage earnings.
This may be on account that larger firms rely more on hedging. However, it is also possible that
the information requirement by the public may constrain larger firms’ discretion in earnings
manipulation.
25
6. The Impact of FCD Usage and EM on Firm Market Value
In this section we examine whether the use of FCDs and EM can increase firm market value.
Built on prior theoretical and empirical research, we hypothesize that foreign currency hedging
with FCDs can increase firm market value. As suggested by Dechow and Skinner (2000), firm
managers have incentives to manage earnings in order to maintain or improve their capital market
valuation. On the other hand, certain EM activities also satisfy the rationales underlying the
optimal hedging theory (e.g., reduction in expected taxes, costs of financial distress, or
underinvestment problem), we also hypothesize that EM can increase firm value.
6.1 Univariate Tests (Tobin’s Q as a Proxy for Firm Market Value)
We begin with a univariate test to examine the possible impact of FCD usage and EM on firm
market value. The mean of Tobin’s Q is greater than its median, indicating that the distribution of
Tobin’s Q may be skewed. Our hypothesis is tested using both means and medians.
[ Insert Table 3 Here ]
Panel A of Table 3 reports the mean and median of Tobin’s Q for firms using FCDs and firms
not using FCDs. Firms reporting the use of FCDs are classified as FCD users, while firms not
using FCDs are classified as FCD non-users. The mean (median) Tobin’s Q is 1.945 (1.389) for
FCD users, as compared with the mean (median) for FCD non-users of 1.848 (1.284). Supporting
our hypothesis, firms reporting the use of FCDs are rewarded with higher Tobin’s Q (the
difference in median is significant at the 0.01 level).
Table 3, Panel B presents a comparison of Tobin’s Q across EM quintiles. For EMsmooth,
26
EMcorrelation, and EMDAC, the sample firms are partitioned into quintiles based on the scores of each
EM measure. We put firms with the greatest level of EM in Q5 and those with the lowest level of
EM in Q1. Thus, Q1, Q2, Q3, Q4, and Q5 represent the EM level from conservative to aggressive.
The association between EM measures and firm value is examined by comparing the mean and
median values of Tobin’s Q across EM quintiles. For the small-loss aversion measure, EMsmall-loss,
the comparisons are made for firms reporting a quarterly small loss (EMsmall-loss ≠ 0) and firms
never reporting a quarterly small loss (EMsmall-loss=0). Firms having the courage to report small
losses are regarded as practicing a conservative EM than those never reporting a small loss. All the
four EM measures are calculated for a 3-year period ahead of the estimation of Tobin’s Q in a
given fiscal year-end. Under our hypothesis, we expect that firms exercising a more aggressive
level of EM (e.g., firms in the fifth quintile, Q5) will enjoy a higher market value.
The patterns of Tobin’s Q in relation to EM measures generally confirm our hypothesis.
Taking the example of our first income smoothing measure, EMsmooth, firms in the most aggressive
income smoothing quintile (Q5) have the highest mean (median) value of Tobin’s Q of 2.194
(1.710), while firms in the most conservative income smoothing quintile (Q1) have the lowest
mean (median) Tobin’s Q of 1.744 (1.198). The evidence indicates that firms exercising greater
income smoothing are significantly rewarded with a higher Tobin’s Q. The mean (median) of
Tobin’s Q increases monotonically from conservative quintile to aggressive quintile. We obtain
similar patterns for another income smoothing measure, EMcorrelation. The mean of Tobin’s Q is
1.802 for the first EMcorrelation quintile (most conservative), as compared with the mean of 2.346 in
the fifth EMcorrelation quintile (most aggressive). Similarly, the mean (median) of Tobin’s Q
increases (generally) as EMcorrelation becomes aggressive.
The small-loss aversion measure, EMsmall-loss, also exhibits an expected relation to Tobin’s Q.
The mean (median) value of Tobin’s Q is equal to 1.447 (1.019) for firms reporting a small loss
27
(EMsmall-loss ≠ 0), as compared with the mean (median) value of Tobin’s Q of 2.235 (1.654) for
firms never reporting a small loss (EMsmall-loss=0). The Tobin’s Q of firms having the courage to
recognize a small loss is significantly smaller than firms never reporting a small loss.
Differing from other EM measures, the relation between Tobin’s Q and the magnitude of
discretionary accruals, EMDAC, is not so clear. The mean (median) of Tobin’s Q for the fifth
EMDAC quintile (most aggressive) is significantly larger than the first quintile, however, firms in
the fourth EMDAC quintiles (Q4) have a greater Tobin’s Q than other quintiles. Although firms in
the most conservative quintile have the smallest Tobin’s Q, Tobin’s Q does not increase
monotonically when EMDAC becomes aggressive. We have also estimated the component of
discretionary accruals by adding the ROA adjustment into the modified Jones model [see Kothari
et al. (2005)] and obtained similar results (not reported). Since discretionary accruals can be
exercised for purposes other than income smoothing, this probably induces noise into the observed
relation between EMDAC and Tobin’s Q.
6.2 Univariate Tests (Industry-adjusted Q as a Proxy for Firm Market Value)
It is possible that the prior results are due to industrial differences, and not due to the
influence of hedging or EM. In this section we use the industry-adjusted Q as a proxy for firm
market value, instead of a raw Tobin’s Q.
[ Insert Table 4 Here ]
Panel A of Table 4 reports a comparison of industry-adjusted Q for FCD users and FCD
non-users. The mean (median) industry-adjusted Q is 0.897 (0.612) for FCD users, as compared
with a mean (median) for FCD non-users of 0.624 (0.329). Consistent with our hypothesis, FCD
28
users have a significantly higher industry-adjusted Q than FCD non-users (the difference in mean
and median is significant at the 0.01 level).
Table 4, Panel B presents the comparisons of industry-adjusted Q across EM quintiles. The
relationships between EM measures and industry-adjusted Q are mixed as compared with the
results obtained from Tobin’s Q. Taking an example of the income smoothing measures (EMsmooth
and EMcorrelation), although firms in the fifth income smoothing quintile (most aggressive) are
generally larger in industry-adjusted Q than firms in the first income smoothing quintile (most
conservative), however, the difference is mostly not significant. The industry-adjusted Q tends to
increase as EMsmooth and EMcorrelation become aggressive, however, the pattern is not exactly
monotonic. For instance, we do not find that firms in the most conservative income smoothing
quintile have the lowest Q and that firms in the most aggressive income smoothing quintile have
the highest Q. Indeed, we find that firms in the third and fourth EMsmooth quintiles have a higher
industry-adjusted Q than other quintiles and firms in the fourth EMcorrelation quintile have the
highest industry-adjusted Q.
The results of EMDAC and EMsmall-loss are clearer and are consistent with our hypothesis. Firms
in the most aggressive EMDAC and EMsmall-loss quintiles have a significantly higher
industry-adjusted Q than firms in the most conservative quintiles, and the industry-adjusted Q
increases monotonically from the conservative quintile to the aggressive quintile.
6.3 Multivariate Tests (Tobin’s Q as a Proxy for Firm Market Value)
In this subsection we further examine the impact of FCD usage and EM on firm value while
controlling for other related variables, including size, access to financial markets, leverage,
profitability, growth opportunity, industrial diversification, geographic diversification, credit rating,
and time-effect.
29
[ Insert Table 5 Here ]
Table 5 presents the regression results of equation (11) for our sample firms. We estimate
the regression models based on a feasible generalized least squares (FGLS) specification to correct
for both serial correlations across periods and period heteroskedasticity between the residuals for a
given firm. The standard errors are computed to be robust to a general serial correlation. To
examine our hypotheses, we focus on the sign and significance of DCD and the four EM measures.
DCD equals 1 if a firm use FCDs, otherwise 0. Consistent with our hypothesis that the use of
FCDs increases firm market value, we find a significantly positive impact of FCD usage on
Tobin’s Q. The sign and significance of the coefficient for DCD do not change when the EM
measure is included. The hedging premium, indicated by the regression coefficient of DCD, is in
the range from 11.61% to 12.62% and is greater than about 5% in Allayannis and Weston (2001)
and is comparable to 5% - 10% in Carter et al. (2006). The evidence also confirms prior studies
that financial hedging is a value-adding strategy [e.g., see Allayannis and Weston (2001), Kim et
al. (2006) and Carter et al. (2006)].
We also hypothesize that the exercise of EM will enhance firm value, however, investors
should distinguish the valuation effect of EM from that of hedging. The results of our four EM
measures in relation to Tobin’s Q are mixed. Table 5 shows that the two income smoothing
measures (EMsmooth and EMcorrelation) are insignificantly associated with Tobin’s Q. Although we
expect a positive relation between income smoothing and Tobin’s Q, the EMcorrelation even has a
negative impact on firm value. The results do not change when DCD is included. The evidence
from EMsmooth and EMcorrelation implies that the exercise of EM for income smoothing does not
increase firm market value. As discussed before, this may be due to the possible substitute relation
30
between hedging and EM - that is, firms can affect their value by hedging, instead of EM. In
addition, since income smoothing functions as hedging, the insignificance of income smoothing
measures in explaining firm value may be caused by the detectable characteristic of EM.
Contrary to income smoothing measures, the results for EMDAC and EMsmall-loss are consistent
with our hypothesis. Although the different purposes in exercising discretionary accruals may
induce noise into the expected relation between EMDAC and firm value, we find that EMDAC
exhibits a significantly positive association with Tobin’s Q, implying that the greater exercise of
discretionary accruals enhances firm market value. As suggested by Dechow and Skinner (2000),
if the greater use of discretionary accruals is more likely used to maintain or improve their capital
market valuation, then the positive impact of EMDAC on Tobin’s Q is expected. We also find that
firms never reporting a small loss have a significantly higher Tobin’s Q than firms reporting a
small loss. This may be attributable to the fact that avoiding a small loss helps build a reputation
and thus increases firm value.
The control variables in Table 5 generally are statistically significant and exhibit the expected
signs in relation to firm market value. Consistent with prior studies [e.g., see Allayannis and
Weston (2001) and Carter et al. (2006)], we find a significantly negative association between
Tobin’s Q and firm size. We also find that profitability (proxied by ROA) is significantly and
positively related to Tobin’s Q. Leverage (proxied by debt to total assets ratio) is significantly and
negatively associated with Tobin’s Q. R&D expenditure, a proxy for growth opportunity, is
significantly positively associated with Tobin’s Q, indicating that firms with more growth
opportunity are valued with a higher value. Dividend yield, a proxy for access to financial markets,
is significantly negatively related to Tobin’s Q. We find that neither industrial diversification nor
geographic diversification can significantly increase Tobin’s Q. In fact, negative coefficients are
mainly found for industrial diversification and geographic diversification. The adjusted R2s are in
31
the 0.5095 to 0.5968 range and indicate a high level of explanatory power.
6.4 Multivariate Tests (Industry-adjusted Q as a Proxy for Firm Market Value)
We further examine whether our results are robust to alternative measures of firm market
value - the industry-adjusted Q. This is important since a large percentage of our sample firms
conduct operations in more than one business segment.
[ Insert Table 6 Here ]
Table 6 presents the regression results of equation (11) while using industry-adjusted Q as a
proxy for firm value. We estimate the regression models based on a feasible generalized least
squares (FGLS) specification that corrects for both serial correlations across periods and period
heteroskedasticity between the residuals for a given firm. Confirming our hypothesis, we find a
significantly positive relation between the use of FCDs and industry-adjusted Q. The four EM
measures still exhibit mixed results. We find a negative impact of the two income smoothing
measures on industry-adjusted Q. As compared with the significantly positive impact of financial
hedging, the results of income smoothing measures indicate that capital markets do distinguish the
valuation effect of income smoothing from that of hedging. Other control variables generally have
similar signs and significance as those of Table 5, except leverage (becomes insignificant) and
capital expenditure (becomes significant).
7. The Impact of EM on Firm Market Value: Larger Firms vs. Smaller Firms
32
In the previous section we find that investors tend to distinguish the valuation effect of
income smoothing from that of hedging, however, other EM measures (EMDAC and EMsmall-loss)
exhibit a significantly positive impact on firm value. Since information asymmetry may affect
investors’ reaction to EM and firm size is a typical proxy for information asymmetry, we further
compare the impact of EM on firm value between larger and smaller firms. Consistent with Nance,
Smith and Smithson (1993), Albrecht and Richardson (1990), and Barton (2000), our prior results
indicate that larger firms have a higher likelihood to hedge and engage in lesser EM. Since larger
firms rely less on EM and larger firms are subject to greater scrutiny by the public, we hypothesize
that EM will have a lesser impact on firm market value for larger firms.
7.1 Univariate Tests
Table 7 presents a comparison of Tobin’s Q while controlling for size and the degree of EM.
The sample firms are divided into quintiles based on firm size and the four EM measures. Firm
size is measured by total assets. Table 7, Panel A reports the comparison of the mean Tobin’s Q
across size and EMsmooth quintiles. Similar to the prior results of Table 5, Tobin’s Qs are generally
larger in the smaller size quintile. Firms in the smallest size quintile have a mean Q in the range
from 2.556 to 3.508, while the mean value of Tobin’s Q for firms in the largest size quintile is in
the range from 1.289 to 1.953. Generally, the Tobin’s Q increases from the largest size quintile to
the smallest size quintile. Consistent with our hypothesis, we find that the relation between
EMsmooth and Tobin’s Q is stronger for the smallest size quintile. The Tobin’s Q in the smallest
size quintile increases nearly monotonically from the most conservative EM quintile to the most
aggressive EM quintile, however, we do not find a similar pattern for larger size quintiles. As we
observe, the relation between Tobin’s Q and EMsmooth becomes clearer from the largest size
quintile to the smallest size quintile. Similar patterns are found for another income smoothing
33
measure, EMcorrelation.
[ Insert Table 7 Here ]
The small-loss aversion measure (EMsmall-loss) exhibits consistent results for each size
quintiles. We find that the mean value of Tobin’s Q is smaller for firms reporting a small loss than
for firms never reporting a small loss, and the pattern is held for all five size quintiles. In short, the
above evidence supports our hypothesis that smaller firms are more likely to gain benefits from
EM. We suggest that this may be attributable to larger firms being subject to greater scrutiny by
the public and their EM efforts are more detectable.
We also examine the robustness of our results using industry-adjusted Q. Table 8 presents the
comparisons of industry-adjusted Q across size and EM quintiles. We find similarly that whether
EM increases industry-adjusted Q depends on firm size. Consistent with our hypothesis, we find a
more significant pattern between industry-adjusted Q and EM measures in the smaller size
quintile.
[ Insert Table 8 Here ]
7.2 Multivariate Tests
In this subsection we further examine the hypothesis that the impact of EM on firm value
depends on firm size while controlling for other variables related to firm value. Firms are divided
into terciles based on firm size (proxied by total assets). The impact of EM on firm market value
for the larger size terciles is expected to be smaller than firms in the smaller size tercile, since
artificial manipulation in earnings is more likely to be detected by the capital market for larger
34
firms.
We develop two dummy variables, D_Large and D_Small, that indicate whether a firm is
classified as larger or smaller in firm size. We then add the interaction of the EM measure and the
two Dummy variables into equation (11). The regression model is as follows:
( )
tititti
ttittitititi
tijtijtitititi
DCDEMSmallDEMeLDEMGEOSEGDIV
CAPXRDRLeverageROASIZEDQsTobin
,,12,2,11
,2,10,2,9,8,7,6
,,5,,4,3,2,10,
*_ *arg_
& 'ln
εγγγγγγγ
γγγγγγ
+++
+++++
+++++=
−
−−
∑, (12)
where D_Large equals 1 if the firm is in the larger size tercile, otherwise 0, and D_Small equals 1
if the firm is in the smaller size quintile. We add two interaction variables, D_Large*EM and
D_Small*EM, to capture the potential difference of EM in explaining firm market value between
larger and smaller firms. Based on our hypothesis, we expect that the estimated coefficient ( 10γ̂ ) of
D_Large*EM will have an insignificant or even negative sign, while the estimated coefficient ( 11γ̂ )
of D_Small*EM will have a more positive sign.
Table 9 presents the FGLS regression results of equation (12) while using Tobin’s Q as a
proxy for firm market value. To examine our hypothesis, we focus on the sign and significance of
the two interaction variables. Consistent with our hypothesis, we find a strong difference in the
impact of EM on firm market value between larger and smaller firms. The coefficients of the
interaction of smaller size dummy and EM are all significant and carry a positive sign, suggesting
that EM significantly increases firm value for smaller firms. Conversely, no coefficient of the
interaction of the larger size dummy and EM is significantly positive. In particular, the interaction
of the larger size dummy and two of the EM measures (EMsmooth and EMsmall-loss) is significant and
has a negative sign, indicating that the greater exercise of EM for income smoothing or small-loss
aversion destroys firm value. Other variables generally have a similar sign and significance as
35
Table 5. We have also used industry-adjusted Q as a proxy for firm market value and obtain
similar results (Table 10).
[ Insert Table 9 Here ]
[ Insert Table 10 Here ]
8. Firm-Specific Exchange Rate Exposure and Firm Value
In the previous section, we have already documented that both the FCD usage and EM
increase firm value, although the valuation effect of EM depends on firm size (a proxy for
information asymmetry). Since prior studies suggest that firms may use either financial hedging or
EM to mitigate exchange rate exposure, in this section we examine whether the reduction in
exchange rate exposure leads to higher firm value, and if so, how investors perceive the two
different mechanisms through which firms may reduce exchange rate exposure to improve firm
value.
Following Adler and Dumas (1984) and Jorion (1990), we estimate firm-specific exchange
rate exposure using the following regression model:
itmtmit
fxiiti RFXR εβββ +++= 0 , (13)
where Rit is the monthly stock return of firm i in period t, FXt is the monthly percentage change of
the trade-weighted exchange rate index, measured as foreign currency per one U.S. dollar in
period t, and Rmt is the monthly return on the market portfolio. We use the Federal Reserve Bank
(FRB) trade-weighted broad dollar index to measure the foreign currency value of the U.S. dollar
36
against 26 selected currencies.21 The time series data of the FRB broad dollar index are obtained
from FRED (Federal Reserve Economic Data) database. Firm-specific exchange rate exposure is
represented as the absolute value of the exposure coefficient, | fxiβ |, estimated by equation (13).
The exposure coefficient is estimated for the three-year window (year t-2 to year t) ahead of the
estimation of Tobin’s Q.
We also develop a dummy variable, EMH, to indicate whether a firm engages aggressively in
EM. Firms are divided into terciles based on the scores of each EM measures. EMH equals 1 if a
firm is in the third EM tercile (most aggressive), otherwise 0. In addition to fxtti ,2,
ˆ−β , we also
include the interaction of fxtti ,2,
ˆ−β and DCD as well as the interaction of fx
tti ,2,ˆ
−β and EMH in the
regression model to examine the possible difference in the valuation effect between FCD usage
and EM. The regression equation is as follows:
( )
tifx
ttiHighfx
ttitifx
ttitititi
tijtijtitititi
EMDCDGEOSEGDIV
CAPXRDRLeverageROASIZEDQsTobin
,,2,11,2,,10,2,9,8,7,6
,,5,,4,3,2,10,
ˆˆˆ
& ' ln
εβγβγβγγγγ
γγγγγγ
+×+×+++++
+++++=
−−−
∑ , (14)
Table 11 reports the regression results. Tobin’s Q is used as a proxy for firm value. Without
controlling for hedging and EM, the impact of exchange rate exposure on Tobin’s Q is examined
by regression (1) of Table 11. The results of regression (1) show that exchange rate exposure is
significantly negatively associated with Tobin’s Q, indicating that lower exchange rate exposure is
valued by investors at a premium. Since previous studies have already demonstrated that either the
FCD usage or EM can mitigate the stock return exposure to exchange rate risk, the evidence thus
21 FRB broad dollar index measures the foreign exchange value of the U.S. dollar against 26 selected currencies of a broad group of major U.S. trading partners- Euro Area, Canada, Japan, Mexico, China, United Kingdom, Taiwan, Korea, Singapore, Hong Kong, Malaysia, Brazil, Switzerland, Thailand, Philippines, Australia, Indonesia, India, Israel, Saudi Arabia, Russia, Sweden, Argentina, Venezuela, Chile and Colombia.
37
implies that firm managers are able to enhance their firm’s value by mitigating exchange rate
exposure through hedging or EM.
Since investors generally view EM as opportunistic and problematic as compared to hedging,
the impact of exchange rate exposure on firm value is further examined while controlling for
hedging and EM (Table 11, regression 2 - 5). The estimated coefficient of fxtti ,2,
ˆ−β now measures
the impact of exchange rate exposure that is naturally generated (through neither FCD usage nor
EM). The estimated coefficients for fxtti ,2,
ˆ−β are significantly negative for all the regressions,
indicating that lower exchange rate exposure that is naturally generated leads to significantly
greater firm value. We find mainly positive coefficients for the interaction of fxtti ,2,
ˆ−β and DCD.
For firms reporting the use of FCDs, the benefit of reducing exchange rate exposure in
value-increasing is smaller, although it is not significant. We suggest that such evidence may be
attributable to that hedging is not costless. We also find that the estimated coefficients of the
interaction variable (EMH × fx
tti ,2,ˆ
−β ) are mainly positive and significant, and that their magnitude
is comparable to the coefficient of fxtti ,2,
ˆ−β , implying that the improvement in firm value does not
apply to the case where the stock return exposure to exchange rate risk is mitigated through EM.
The evidence of EM is consistent with our expectation that EM efforts in mitigating exchange rate
exposure are detectable and are penalized by investors.
[ Insert Table 11 Here ]
9. Conclusions
38
This study compares the possible impact of FCD usage (real actions) and EM (artificial
techniques) on firm value. Using a sample of S&P 500 non-financial firms, we demonstrate that
the use of FCDs significantly increases firm market value (proxied by Tobin’s Q and
industry-adjusted Q). The premium from hedging is about 11% of firm value, which is greater than
prior studies (e.g., about 5% in Allayannis and Weston (2001) and about 5% - 10% in Carter et al.
(2006)).
We find mixed valuation impacts of our four EM measures. In particular, we fail to find a
significantly positive relation between income smoothing and firm value. Since EM for income
smoothing purpose acts much similar to hedging, the evidence thus indicates that investors tend to
distinguish the valuation effect of EM from that of hedging.
We also examine whether the valuation effect of EM depends on firm size, a proxy for
information asymmetry. Since artificially manipulated earnings are more likely to be detected and
penalized by investors for larger firms, we suggest that the mixed impact of EM is attributable to
firm size. Consistent with our hypothesis, we find EM exhibits a greater impact on value-adding
for smaller firms.
We also provide evidence for the possible channel through which currency hedging or EM
can contribute to increasing firm value. We find that lower exchange rate exposure leads to greater
firm value. The evidence implies that firms may improve market value by reducing exposure to
currency risk, however, we show that the improvement in capital market valuation does not apply
to the case where the exchange rate exposure is mitigated through EM.
39
References
Allayannis, G. and E. Ofek (2001), ‘Exchange-Rate Exposure, Hedging, and the Use of Foreign Currency Derivatives’, Journal of International Money and Finance, Vol. 20, pp. 273-296.
Allayannis, G. and J. Weston (2001), ‘The Use of Foreign Currency Derivatives and Firm Market Value’, Review of Financial Studies, Vol. 14, pp. 243-276.
Allayannis, G., B. Rountree, and J.P. Weston (2005), ‘Earnings Volatility, Cash Flow Volatility, and Firm Value’, Working paper.
Albrecht, W.D. and F.M. Richardson (1990), ‘Income Smoothing by Economy Sector’, Journal of Business Finance & Accounting, Vol. 17(5), pp. 713–30.
Balsam, S., E. Bartov and C. Marquardt (2002), ‘Accruals management, investor sophistication and equity valuation: Evidence from 10-Q filings’, Journal of Accounting Research, Vol. 40, pp. 987-1012.
Bao, B-H. and D-H Bao (2004), ‘Income smoothing, earnings quality and firm valuation’, Journal of Business Finance & Accounting, Vol. 31, pp. 1525-1557.
Bartram, S.M., G.W. Brown and F.R. Fehle (2004), ‘International evidence on financial derivatives usage’, SSRN Working Paper.
Barton, J. (2001), ‘Does the use of financial derivatives affect earnings management decisions?’, The Accounting Review, Vol. 76 (1), pp. 1–26.
Beidleman, C.R. (1973), ‘Income smoothing: The role of management’, The Accounting Review, Vol. 48, pp. 653-667.
Bitner, L.N. and R.C. Dolan (1996), ‘Assessing the relationship between income smoothing and the value of the firm’, QJBE, Vol. 35 (1), pp. 16-35.
Burgstahler, D. C. and I. D. Dichev (1997), ‘Earnings Management to Avoid Earnings Decreases and Losses’, Journal of Accounting and Economics, Vol.24, pp. 99–126.
40
Carter, D.A., D.A. Rogers and B.J. Simkins (2006), ‘Does Hedging Affect Firm Value? Evidence from the US Airline Industry’, Financial Management, pp. 53-86.
Chung, K.H. and S.W. Pruitt (1994), ‘A Simple Approximation of Tobin' Q’, Financial Management, Vol. 22, pp. 70-74.
Dechow, P.M., S.A. Richardson and I. Tuna (2003), ‘Why Are Earnings Kinky? An Examination of the Earnings Management Explanation’, Review of Accounting Studies, Vol. 8, pp.355–384.
Dechow, P.M. and D.J. Skinner (2000), ‘Earnings management: Reconciling the views of accounting academics, practitioners, and regulators’, Accounting Horizons, Vol. 14, pp. 235-250.
Dechow, P., R. Sloan and A. Sweeney (1995), ‘Detecting earnings management’, The Accounting Review, Vol. 70 (2), pp. 193–225.
DeFond, M. and C. Park (1997), ‘Smoothing in anticipation of future earnings’, Journal of Accounting and Economics, Vol. 23 (2), pp. 115–139.
Degeorge, F., J. Patel and R. Zeckhauser (1999), ‘Earnings Management to Exceed Thresholds’, The Journal of Business, Vol.72, pp. 1–33.
DeMarzo, P. and D. Duffe (1995), ‘Corporate incentives for hedging and hedge accounting’, The Review of Financial Studies, Vol. 8, pp. 743-771.
Ecker, F., J. Francis, I. Kim, P.M. Olsson and K. Schipper (2006), ‘A Returns-Based Representation of Earnings Quality’, The Accounting Review, Vol. 81 (4), pp. 749-780.
Francis, J., R. LaFond, P.M. Olsson and K. Schipper (2004), ‘Costs of Equity and Earnings Attributes’, The Accounting Review, Vol. 79 (4), pp. 967-1010.
Froot, K., D. Scharfstein, and J. Stein (1993), ‘Risk Management: Coordinating Corporate Investment and Financing Policies’, Journal of Finance, Vol. 48, pp. 1629-1658.
Géczy, C., B.A. Minton, and C. Schrand (1997), ‘Why Firms Use Currency Derivatives?’, Journal of Finance, Vol.52, pp. 1323-1354.
41
Graham, J.R. and D.A. Rogers (2002), ‘Do Firms Hedge in Response to Tax Incentives?’, Journal of Finance, Vol. 57, pp. 815-839.
Graham, J.R. and C.W. Smith (1999), ‘Tax Incentives to Hedge’, Journal of Finance, Vol. 54, pp. 2241-2262.
Haushalter, G.D. (2000), ‘Financing Policy, Basis Risk, and Corporate Hedging: Evidence from Oil and Gas Producers’, Journal of Finance, Vol. 55, pp. 107-152.
Hayn, C. (1995), ‘The Information Content of Losses’, Journal of Accounting and Economics, Vol. 20, pp. 125–153.
Healy, P. and J. Wahlen (1999), ‘A review of the earnings management literature and its implications for standard setting’, Accounting Horizons, Vol. 13, pp. 365–383.
Jin, Y. and P. Jorion (2006), ‘Firm Value and Hedging: Evidence from US Oil and Gas Producers’, Journal of Finance, Vol. LXI(2), pp. 893-919.
Jones, J.J. (1991), ‘Earnings management during import relief investigations’, Journal of Accounting Research, Vol. 29, pp. 193-228.
Kim, Y.S., I. Mathur and J. Nam (2006), ‘Is operational hedging a substitute for or a complement to financial hedging?’, Journal of Corporate Finance, Vol. 12, pp. 834-853.
Kim, Y. and M.S. Park (2005), ‘Pricing of Seasoned Equity Offers and Earnings Management’, Journal of Financial and Quantitative Analysis, Vol. 40 (2), pp. 435-463.
Kothari, S.P., A.J. Leone and C.E. Wasley (2005), ‘Performance matched discretionary accrual measures’, Journal of Accounting and Economics, Vol. 39, pp. 163–197.
Lambert, R. ?(1984), ‘Income smoothing as rational equilibrium behavior’, The Accounting Review, Vol. 59 (4), pp. 604–618.
Lang, L. and R. Stulz (1994), ‘Corporate Diversification and Firm Performance’, Journal of Political Economy, Vol. 102, pp. 142-174.
Leland, H.E. (1998), ‘Agency Costs, Risk Management, and Capital Structure’, Journal of
42
Finance, Vol. 53, pp. 1213-1243.
Leuz, C., Nanda, D., Wysocki, P. (2003), ‘Earnings management and investor protection: an international comparison’, Journal of Financial Economics, Vol. 69, pp. 505-527.
Lev, B. and S. Kunitzky (1974), ‘On the Association Between Smoothing Measures and the Risk of Common Stock’, The Accounting Review, April, pp. 259–70.
Lindenberg, E.B. and S.A. Ross (1981), ‘Tobin's q Ratio and Industrial Organization.’, Journal of Business, January, pp. 1-32
Mayers, D. and C.W. Smith (1982), ‘On the Corporate Demand for Insurance’, Journal of Business, Vol. 55, pp. 281-296.
Mian, S.L. (1996), ‘'Evidence on Corporate Hedging Policy’, Journal of Financial and Quantitative Analysis, Vol. 31, pp. 419-439.
Michelson, S.E., J. Jordan-Wagner and C.W. Wootton (1995), ‘A Market Based Analysis of Income Smoothing’, Journal of Business Finance & Accounting, December, pp. 1179–93.
——— ——— ——— (2000), ‘The Relationship Between the Smoothing of Reported Income and Risk-Adjusted Returns’, Journal of Economics and Finance, Summer, pp. 141–59.
Moses, O.D. (1987), ‘Income Smoothing and Incentives: Empirical Tests Using Accounting Changes’, The Accounting Review (April), pp. 358–77.
Myers, S.C. (1977), ‘Determinants of corporate borrowing’ Journal of Financial Economics, Vol. 5, pp. 147– 175.
Nance, D.R., C.W. Smith, and C.W. Smithson (1993), ‘On the Determinants of Corporate Hedging’, Journal of Finance, Vol. 48, pp. 391-405.
O'Brien, P. and R. Bhushan (1990), ‘Analyst Following and Institutional Ownership’, Journal of Accounting Research, Vol. 28, pp. 55-76.
Perry, S.E. and T.H. Williams (1994), ‘Earnings Management Preceding Management Buyout Offers’, Journal of Accounting and Economics, Vol. 18 (2), pp. 157-179.
43
Pincus, M. and S. Rajgopal (2002), ‘The interaction of accrual management and hedging: evidence from oil and gas firms’, The Accounting Review, Vol. 71, pp. 127-160.
Rangan, S. (1998), ‘Earnings Management and the Performance of Seasoned Equity Offerings’, Journal of Financial Economics, Vol. 50, pp. 101-122.
Smith, C.W. and R.M. Stulz (1985), ‘The Determinants of Firms’ Hedging Policies’, Journal of Financial and Quantitative Analysis, Vol. 20, pp. 391-405.
Stulz, R. (1984), ‘Optimal Hedging Policies’, Journal of Financial and Quantitative Analysis, Vol. 19, pp. 127-140.
Stulz, R.M. (1996), ‘Rethinking Risk Management’, Journal of Applied Corporate Finance, Vol. 9, pp. 8-24.
Teoh, S., I. Welch, and T. Wong, 1998a, Earnings management and the under-performance of seasoned equity offerings, Journal of Financial Economics, Vol. 50, pp. 63-99.
Teoh, S., I. Welch, and T. Wong, 1998b, Earnings management and the long-run market performance of Initial Public Offerings, Journal of Finance, Vol. 53, 63-99.
Trueman, B. and S. Titman. (1988), ‘An explanation for accounting smoothing’, Journal of Accounting Research, Vol. 26, pp. 127–139.
Tufano, P. (1998), ‘Agency Costs of Corporate Risk Management’, Financial Management, Vol. 27, pp. 67-77.
44
Table 1 Summary Statistics This table presents summary statistics for our sample of S&P 500 non-financial firms. The total sample includes firm-year observations with non-missing Tobin’s Q between 2001 and 2003. Panel A reports the descriptive statistics of main firm characteristics. Panel B and Panel C report the descriptive statistics of foreign currency derivative (FCD) usage and earnings management (EM) measures. Tobin’s Q is defined as the ratio of the market value of the firm’s assets to the replacement cost of the firm’s assets in the fiscal year end. Industry-adjusted Q is defined as the difference between raw Tobin’s Q and pure-play Q [Lang and Stulz (1994)]. DCD equals 1 if the firm discloses the use of any type of FCDs (e.g., forward contracts, options, futures or swaps), otherwise 0. NCD denotes the notional amount of FCDs. Only firms reporting non-zero notional amount of FCDs are included. The calculation of notional amount excludes currency swaps. EMsmooth is calculated as a firm’s standard deviation of quarterly operating earnings over the standard deviation of quarterly operating cash flows (both operating earnings and cash flows are divided by lagged total assets). EMcorrelation is the correlation coefficient between accruals changes and cash flows changes (both accruals changes and cash flows changes are divided by lagged total assets). EMDAC is the average absolute value of yearly discretionary accruals scaled by lagged total assets. The discretionary accruals are estimated using Modified Jones model. EMsmall-loss is a ratio of quarters with small-loss reporting to the total quarters. The four EM measures are calculated for a three-year period ahead of the estimation of firm value.
No. of observations Mean Std. Dev. 1st quartile Median 3rd quartile
Panel A: Main Firm Characteristics Total assets (millions) 1186 16845 41413 2899 7116 17757 Total sales (millions) 1186 12460 23132 2366 5611 12923 Dividend yield 1186 1.500 1.888 0 0.915 2.310 R&D/total sales 1186 0.052 0.124 0 0.006 0.055 Capital expenditures/total sales 1176 0.079 0.092 0.030 0.047 0.094 ROA 1186 0.057 0.211 0.022 0.069 0.124 Debt ratio 1185 0.262 0.163 0.147 0.262 0.370 No. of geographic segment 1186 3.799 1.099 3 4 5 No. of business segment 1134 3.543 2.027 2 3 5 Tobin's Q 1186 1.912 1.567 0.918 1.349 2.321 Industry-adjusted Tobin's Q 1181 0.805 1.768 -0.017 0.484 1.306
Panel B: Financial hedging measures
DCD 1186 0.662 0.473 0 1 1 Notional amount (millions) 414 636.79 1241.86 70 190.35 636 Notional amount / total sales 414 0.089 0.122 0.022 0.056 0.111
Panel C: Earnings management measures
EMsmooth 1075 -0.366 0.594 -0.369 -0.213 -0.132 EMcorrelation 1067 0.937 0.150 0.956 0.988 0.996 EMDAC 1151 0.0618 0.0812 0.0239 0.0404 0.0704 EMsmall_loss 1174 -0.076 0.119 -0.083 0 0
45
Table 2 This table presents the comparisons of FCD usage and EM across size quintiles. Firms are divided into quintiles based on firm size (proxied by total assets). DCD equals 1 if the firm discloses the use of any type of FCDs, otherwise 0. The calculations of EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. Panel A reports the mean value of DCD across size quintiles. Panel B reports the mean and median value of EM measures across size quintiles. The difference of the mean (median) value of DCD and EM measures between the smallest size quintile and the largest size quintile, the associated t-statistics/Wilcoxon/Mann-Whitney value, and p-value are reported in the last three columns. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively.
Size quintile
Q1(Small) Q2 Q3 Q4 Q5(Large) Difference (Q1-Q5)
t-statistics/Wilcoxon/Mann-Wh
itney value p-value
Panel A: Currency derivative instruments DCD Mean 0.660 0.644 0.647 0.565 0.786 -0.126 -3.096 0.002
Panel B: Earnings management measures
EMsmooth Mean -0.301 -0.269 -0.274 -0.292 -0.308 0.006 0.236 0.813 Median -0.219 -0.190 -0.211 -0.209 -0.246 0.027 0.353 0.724
EMcorrelation Mean 0.951 0.958 0.959 0.957 0.957 -0.006 -0.724 0.469 Median 0.989 0.988 0.988 0.988 0.986 0.003 0.140 0.889
EMDAC Mean 0.096 0.069 0.054 0.046 0.044 0.053 6.246 0.000 Median 0.063 0.045 0.036 0.033 0.030 0.032 8.916 0.000
EMsmall-loss Mean -0.051 -0.062 -0.081 -0.068 -0.114 0.063 5.275 0.000 Median 0 0 0 0 -0.083 0.083 4.279 0.000
46
Table 3 This table presents the results of univariate tests for the impact of FCD usage and EM on firm value (proxied by Tobin’s Q). The total sample includes firm-year observations with non-missing Tobin’s Q between 2001 and 2003. Panel A reports the mean and median value of Tobin’s Q for FCD users (DCD=1) and FCD nonusers (DCD=0). The difference of Tobin’s Q between FCD users and non-users, the associated t-statistics/Wilcoxon/Mann-Whitney value, and p-value are reported in the last three columns. Panel B reports the mean and median value of Tobin’s Q for EM quartiles. The calculations of EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. For EMsmooth, EMcorrelation, and EMDAC, sample firms are divided into quintiles based on the calculated scores. Q1, Q2, Q3, Q4 and Q5 represent the degree of EM from conservative to aggressive. For EMsmall-loss, the sample firms are divided into two group, EMsmall-loss≠0 and EMsmall-loss =0. “EMsmall-loss =0” denotes firms never reporting quarterly small loss; “EMsmall-loss ≠ 0” denotes firms reporting quarterly small loss. The difference of Tobin’s Q between the most conservative EM quintile and the most aggressive quintile, the associated t-statistics/Wilcoxon/Mann-Whitney value, and p-value are reported in the last three columns. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively. Panel A: Differences of Tobin’ Q: Currency Derivative Users vs. Nonusers
DCD=0 (Nonusers) DCD=1
(Users) Difference (Q1-Q5)
t-statistics/Wilcoxon/Mann-Whitney value p-value
DCD Mean 1.848 1.945 -0.097 -1.009 0.313 Median 1.284 1.389 -0.105 2.119 0.034 Panel B: Differences of Tobin’s Q across Earnings Management Quintiles
Conservative Aggressive
Q1 Q2 Q3 Q4 Q5 Difference (Q1-Q5) t-statistics p-value
EMsmooth Mean 1.744 1.849 2.020 2.169 2.194 -0.450 -3.060 0.002 Median 1.198 1.367 1.414 1.557 1.710 -0.512 4.823 0.000 EMcorrelation Mean 1.802 1.817 1.844 2.197 2.346 -0.544 -3.356 0.001 Median 1.263 1.290 1.392 1.755 1.733 -0.470 4.514 0.000 EMDAC Mean 1.363 1.674 1.917 2.308 2.284 -0.921 -6.583 0.000 Median 1.060 1.305 1.341 1.620 1.559 -0.499 6.721 0.000
EMsmall-loss≠0 EMsmall-loss=0 EMsmall-loss Mean 1.447 2.235 -0.788 -9.144 0.000 Median 1.019 1.654 -0.635 12.518 0.000
47
Table 4 Univariate Tests: The Impact of FCD Usage and EM on Industry-Adjusted Tobin’s Q This table presents the results of univariate tests for the impact of FCD usage and EM on firm value (proxied by industry-adjusted Q). The total sample includes firm-year observations with non-missing industry-adjusted Q between 2001 and 2003. Panel A reports the mean and median value of Tobin’s Q for FCD users (DCD=1) and FCD nonusers (DCD=0). The difference of Tobin’s Q between FCD users and non-users, the associated t-statistics/Wilcoxon/Mann-Whitney value, and p-value are reported in the last three columns. Panel B reports the mean and median value of Tobin’s Q for EM quartiles. The calculations of EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. For EMsmooth, EMcorrelation, and EMDAC, sample firms are divided into quintiles based on the calculated scores. Q1, Q2, Q3, Q4 and Q5 represent the degree of EM from conservative to aggressive. For EMsmall-loss, the sample firms are divided into two group, EMsmall-loss≠0 and EMsmall-loss =0. “EMsmall-loss =0” denotes firms never reporting quarterly small loss; “EMsmall-loss ≠ 0” denotes firms reporting quarterly small loss. The difference of Tobin’s Q between the most conservative EM quintile and the most aggressive quintile, the associated t-statistics/Wilcoxon/Mann-Whitney value, and p-value are reported in the last three columns. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively. Panel A: Differences of Industry-adjusted Q: Currency Derivative Users vs. Nonusers
DCD=0 (Nonusers) DCD=1
(Users) Difference (Q1-Q5)
t-statistics/Wilcoxon/Mann-Whitney value p-value
DCD Mean 0.624 0.897 -0.273 -2.511 0.012 Median 0.329 0.612 -0.282 3.209 0.001 Panel B: Differences of Industry-adjusted Q across Earnings Management Quintiles
Conservative Aggressive
Q1 Q2 Q3 Q4 Q5 Difference (Q1-Q5) t-statistics p-value
EMsmooth Mean 0.779 0.777 0.919 0.964 0.814 -0.035 -0.198 0.843 Median 0.551 0.444 0.629 0.624 0.544 0.007 0.789 0.430 EMcorrelation Mean 0.806 0.719 0.766 1.019 0.992 -0.185 -0.968 0.333 Median 0.518 0.411 0.476 0.674 0.665 -0.147 2.119 0.034 EMDAC Mean 0.235 0.565 0.814 1.136 1.241 -1.007 -6.071 0.000 Median 0.176 0.431 0.584 0.675 0.696 -0.520 6.097 0.000 EMsmall-loss≠0 EMsmall-loss=0 EMsmall-loss Mean 0.458 1.040 -0.582 5.760 0.000 Median 0.249 0.666 -0.418 7.321 0.000
48
Table 5 The Impact of FCD Usage and EM on Tobin’s Q: FGLS Results This table presents the panel regression results for the impact of FCD usage and EM on firm value (proxied by Tobin’s Q). The total sample includes firm-year observations with non-missing Tobin’s Q between 2001 and 2003. The regression model is as follows: ( ) titittititititijtijtitititi DCDEMGEOSEGDIVCAPXRDRLeverageROASIZEDQsTobin ,,10,2,9,8,7,6,,5,,4,3,2,10, & ' ln εγγγγγγγγγγγ +++++++++++= −∑
Tobin’s Q is calculated for 2001, 2002 and 2003, respectively. DCD equals 1 if the firm use FCDs, otherwise 0. EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. SIZE is the logarithm of a firm’s total assets at fiscal year end. ROA is the pre-tax return on total assets. Leverage is measured by total debt as a percentage of total assets. R&D is the R&D expense as a percentage of annual sales. CAPXR is the capital expenditure as a percentage of annual sales. DIV is the dividend yield. GEO is the logarithm of a firm’s total geographic segments. SEG is the logarithm of a firm’s total business segments. All regressions include credit rating and year dummies. The regressions are estimated using a feasible generalized least squares (FGLS) specification that corrects for both serial correlations across periods and period heteroskedasticity between the residuals for a given firm. The standard errors are computed to be robust to general serial correlation. Numbers in the parentheses under the coefficients are the associated t-statistics. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively.
Dependent variable: Tobin’s Q Earnings management measures in different dimentions
EMsmooth EMcorrelation EMDAC EMsmall-loss c 1.8500*** 1.5752*** 1.5865*** 1.9359*** 1.9497*** 1.4390*** 1.4532*** 1.7514*** 1.7666*** ( 7.72) ( 7.50) ( 7.58) ( 7.41) ( 7.55) ( 7.06) ( 7.14) ( 7.99) ( 8.13)
Size -0.1315*** -0.1030*** -0.1049*** -0.1203*** -0.1217*** -0.1043*** -0.1072*** -0.1105*** -0.1132*** (-5.45) (-4.91) (-5.06) (-5.08) (-5.23) (-4.92) (-5.08) (-4.79) (-4.96)
ROA 1.7216*** 2.4524*** 2.4391*** 1.7660*** 1.7650*** 2.5867*** 2.5750*** 1.5830*** 1.5771*** ( 3.25) ( 5.91) ( 5.84) ( 3.07) ( 3.10) ( 7.24) ( 7.17) ( 3.16) ( 3.19)
Leverage -0.2735* -0.1934 -0.1834 -0.2969* -0.2812* -0.2446 -0.2366 -0.2979* -0.2850* (-1.66) (-1.26) (-1.21) (-1.78) (-1.70) (-1.55) (-1.52) (-1.78) (-1.72)
R&D 1.2487** 1.6222*** 1.5880*** 1.1390** 1.1012** 1.4980*** 1.4625*** 1.1987** 1.1687** ( 2.46) ( 5.16) ( 5.29) ( 2.33) ( 2.34) ( 3.39) ( 3.42) ( 2.32) ( 2.34)
CAPXR 0.0490 0.0185 0.0956 -0.0604 0.0238 0.0377 0.1271 -0.0239 0.0724 ( 0.28) ( 0.10) ( 0.51) (-0.33) ( 0.13) ( 0.22) ( 0.72) (-0.14) ( 0.41)
DIV -0.0677*** -0.0764*** -0.0796*** -0.0868*** -0.0900*** -0.0653*** -0.0661*** -0.0690*** -0.0699*** (-5.89) (-5.54) (-5.63) (-5.47) (-5.56) (-5.90) (-5.93) (-6.00) (-6.06)
SEG -0.0761** -0.0404 -0.0574* -0.0472 -0.0660* -0.0439 -0.0590* -0.0448 -0.0630* (-2.18) (-1.18) (-1.71) (-1.30) (-1.85) (-1.32) (-1.78) (-1.31) (-1.86)
GEO -0.0348 -0.0073 -0.0511 0.0201 -0.0292 0.0016 -0.0382 0.0060 -0.0414 (-0.78) (-0.17) (-1.12) ( 0.43) (-0.61) ( 0.04) (-0.88) ( 0.14) (-0.92)
EM 0.0112 0.0131 -0.1290 -0.1383 0.8679*** 0.8870*** 0.8129*** 0.8171*** ( 0.11) ( 0.13) (-0.73) (-0.78) ( 3.55) ( 3.67) ( 3.81) ( 3.82)
DCD 0.1164** 0.1161** 0.1262** 0.1033** 0.1200** ( 2.38) ( 2.33) ( 2.37) ( 2.27) ( 2.45)
Adj. R2 0.5348 0.5611 0.5661 0.5095 0.5154 0.5931 0.5968 0.5421 0.5475 n obs. 1095 987 987 979 979 1062 1062 1085 1085
49
Table 6 The Impact of FCD Usage and EM on Industry-Adjusted Q: FGLS Results This table presents the panel regression results for the impact of FCD usage and EM on firm value (proxied by industry-adjusted Q). The total sample includes firm-year observations with non-missing industry-adjusted Q between 2001 and 2003. The regression model is as follows: ( ) titittititititijtijtitititi DCDEMGEOSEGDIVCAPXRDRLeverageROASIZEDdjusted Qindustry-a ,,10,2,9,8,7,6,,5,,4,3,2,10, & ln εγγγγγγγγγγγ +++++++++++= −∑
Industry-adjusted Q is calculated for 2001, 2002 and 2003, respectively. DCD equals 1 if the firm use FCDs, otherwise 0. EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. SIZE is the logarithm of a firm’s total assets at fiscal year end. ROA is the pre-tax return on total assets. Leverage is measured by total debt as a percentage of total assets. R&D is the R&D expense as a percentage of annual sales. CAPXR is the capital expenditure as a percentage of annual sales. DIV is the dividend yield. GEO is the logarithm of a firm’s total geographic segments. SEG is the logarithm of a firm’s total business segments. All regressions include credit rating and year dummies. The regressions are estimated using a feasible generalized least squares (FGLS) specification that corrects for both serial correlations across periods and period heteroskedasticity between the residuals for a given firm. The standard errors are computed to be robust to general serial correlation. Numbers in the parentheses under the coefficients are the associated t-statistics. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively.
Dependent variable: Industry adjusted Q Earnings management measures in different dimensions
EMsmooth EMcorrelation EMDAC EMsmall-loss c 4.0039*** 3.2119*** 3.2634*** 4.7914*** 4.8718*** 3.0504*** 3.1031*** 3.6733*** 3.7303*** ( 6.31) ( 5.22) ( 5.28) ( 5.63) ( 5.73) ( 5.09) ( 5.12) ( 6.42) ( 6.52)
Size -0.2706*** -0.2051*** -0.2129*** -0.2502*** -0.2565*** -0.1977*** -0.2082*** -0.2160*** -0.2260*** (-4.05) (-3.21) (-3.33) (-3.76) (-3.89) (-3.21) (-3.33) (-3.50) (-3.64)
ROA 2.8724** 4.5067*** 4.4474*** 3.0531** 3.0558*** 4.8740*** 4.8347*** 2.5084** 2.4990** ( 2.48) ( 4.14) ( 4.02) ( 2.56) ( 2.60) ( 4.79) ( 4.66) ( 2.45) ( 2.49)
Leverage -0.6236 -0.7734 -0.7307 -1.0054 -0.9418 -0.6921 -0.6572 -0.8742 -0.8265 (-0.92) (-1.27) (-1.21) (-1.57) (-1.49) (-1.07) (-1.02) (-1.37) (-1.31)
R&D 1.8702* 2.5230*** 2.4226*** 1.3118 1.1845 2.6784*** 2.5701*** 1.7504* 1.6723* ( 1.87) ( 3.59) ( 3.62) ( 1.39) ( 1.33) ( 3.12) ( 3.12) ( 1.79) ( 1.78)
CAPXR 1.1354** 0.8399* 1.0963** 0.6653 0.9453* 1.0388** 1.3208*** 0.9022** 1.1854** ( 2.28) ( 1.74) ( 2.22) ( 1.36) ( 1.89) ( 2.17) ( 2.63) ( 1.99) ( 2.52)
DIV -0.0511* -0.0973** -0.1073*** -0.1123** -0.1232*** -0.0493* -0.0518* -0.0475* -0.0505* (-1.88) (-2.38) (-2.67) (-2.40) (-2.71) (-1.82) (-1.92) (-1.73) (-1.86)
SEG -0.0946 0.0205 -0.0331 -0.0038 -0.0634 -0.0193 -0.0645 -0.0034 -0.0547 (-0.91) ( 0.20) (-0.32) (-0.04) (-0.60) (-0.19) (-0.63) (-0.03) (-0.53)
GEO -0.0700 0.1072 -0.0375 0.1979 0.0351 0.0279 -0.0974 0.0536 -0.0870 (-0.51) ( 0.80) (-0.25) ( 1.45) ( 0.23) ( 0.23) (-0.69) ( 0.44) (-0.63)
EM -0.1347 -0.1226 -1.0515 -1.1052 1.5406 1.6152 1.1342** 1.1597** (-0.46) (-0.43) (-1.39) (-1.44) ( 1.48) ( 1.58) ( 2.13) ( 2.20)
DCD 0.3612** 0.3822** 0.4155** 0.3245** 0.3553** ( 2.23) ( 2.28) ( 2.40) ( 2.04) ( 2.23)
Adj. R2 0.2798 0.3020 0.3089 0.2797 0.2880 0.3223 0.3274 0.2687 0.2755
n obs. 1090 983 983 975 975 1057 1057 1080 1080
50
Table 7 Univariate Comparisons of Tobin’s Q across Size and EM Quintiles This table presents the comparisons of average Tobin’s Q across size and EM quintiles. The total sample includes firm-year observations with non-missing Tobin’s Q between 2001 and 2003. The sample firms are divided into quintiles based on firm size and the four EM measures. Firm size is measured by total assets. EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. Q1, Q2, Q3, Q4 and Q5 represent the degree of EM from conservative to aggressive. For EMsmall-loss, the sample firms are divided into two group, EMsmall-loss≠0 and EMsmall-loss =0. “EMsmall-loss =0” denotes firms never reporting quarterly small loss; “EMsmall-loss ≠ 0” denotes firms reporting quarterly small loss reporting. The four EM measures are calculated for a three-year period ahead of the estimation of firm value. Panel A: EMsmooth Quintiles
Conservative Aggressive Q1 Q2 Q3 Q4 Q5 Large size (Q1) 1.296 1.830 1.572 1.289 1.953
Q2 1.351 1.219 1.515 1.939 1.388 Q3 1.688 1.403 1.897 1.860 1.810 Q4 1.700 1.942 1.855 1.846 2.497
Small size (Q5) 2.556 2.663 3.203 3.508 3.290 Panel B: EMcorrelation Quintiles
Conservative Aggressive Q1 Q2 Q3 Q4 Q5 Large size (Q1) 1.622 1.536 1.326 1.901 1.751
Q2 1.258 1.452 1.552 1.630 1.564 Q3 1.466 1.645 1.567 2.127 1.823 Q4 1.873 1.694 1.867 1.913 2.480
Small size (Q5) 2.565 2.736 2.976 3.041 3.939 Panel C: EMDAC quintiles
Conservative Aggressive Q1 Q2 Q3 Q4 Q5 Large size (Q1) 1.153 1.470 1.666 1.437 1.819
Q2 1.150 1.338 1.317 1.780 1.525 Q3 1.552 1.614 1.549 1.994 1.947 Q4 1.687 1.680 2.194 2.097 2.041
Small size (Q5) 2.161 2.553 3.377 3.331 3.089 Panel D: EMsmall-loss quintiles
Conservative Aggressive EMsmall-loss≠0 EMsmall-loss=0 Large size (Q1) 1.162 1.781
Q2 1.067 1.653 Q3 1.426 1.895 Q4 1.426 2.282
Small size (Q5) 2.474 3.310
51
Table 8 Univariate Comparisons of Industry-Adjusted Q across Size and EM Quintiles This table presents the comparisons of average industry-adjusted Q across size and EM quintiles. The total sample includes firm-year observations with non-missing industry-adjusted Q between 2001 and 2003. The sample firms are divided into quintiles based on firm size and the four EM measures. Firm size is measured by total assets. EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. Q1, Q2, Q3, Q4 and Q5 represent the degree of EM from conservative to aggressive. For EMsmall-loss, the sample firms are divided into two group, EMsmall-loss≠0 and EMsmall-loss =0. “EMsmall-loss =0” denotes firms never reporting quarterly small loss; “EMsmall-loss≠0” denotes firms reporting quarterly small loss reporting. The four EM measures are calculated for a three-year period ahead of the estimation of firm value. Panel A: EMsmooth Quintiles
Conservative Aggressive Q1 Q2 Q3 Q4 Q5 Large size (Q1) 0.268 0.647 0.529 -0.025 0.600
Q2 0.339 0.344 0.259 0.803 -0.061 Q3 0.597 0.094 0.712 0.746 0.471 Q4 0.932 1.051 0.824 0.599 1.293
Small size (Q5) 1.642 1.632 2.181 2.275 1.715 Panel B: EMcorrelation Quintiles
Conservative Aggressive Q1 Q2 Q3 Q4 Q5 Large size (Q1) 0.491 0.135 0.392 0.846 0.382
Q2 0.212 0.458 0.597 0.132 0.153 Q3 0.438 0.538 0.390 0.891 0.363 Q4 1.099 0.787 0.730 0.822 1.319
Small size (Q5) 1.571 1.707 1.769 1.931 2.537 Panel C: EMDAC quintiles
Conservative Aggressive Q1 Q2 Q3 Q4 Q5 Large size (Q1) 0.142 0.209 0.822 0.208 0.628
Q2 0.242 0.395 0.174 0.524 0.123 Q3 0.162 0.452 0.310 0.827 1.001 Q4 0.187 0.524 1.215 1.209 1.170
Small size (Q5) 1.046 1.579 2.038 2.002 2.097 Panel D: EMsmall-loss quintiles
Conservative Aggressive EMsmall-loss≠0 EMsmall-loss=0 Large size (Q1) 0.285 0.459
Q2 0.047 0.506 Q3 0.314 0.651 Q4 0.422 1.235
Small size (Q5) 1.518 2.075
52
Table 9 The Impact of EM on Firm Value: Larger Firms vs. Smaller Firms This table presents the panel regression results for the impact of FCD usage and EM on Tobin’s Q. The regression model is as follows: ( ) titittittittititititijtijtitititi DCDEMSmallDEMeLDEMGEOSEGDIVCAPXRDRLeverageROASIZEDQsTobin ,,12,2,11,2,10,2,9,8,7,6,,5,,4,3,2,10, *_*arg_& 'ln εγγγγγγγγγγγγγ +++++++++++++= −−−∑
The total sample includes firm-year observations with non-missing Tobin’s Q between 2001 and 2003. DCD, EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. SIZE, ROA, Leverage, R&D, CAPXR, DIV, GEO, SEG are defined as Table 5. D_Large equals 1 if the firm is in the largest size tercile, otherwise 0. D_Small equals 1 if the firm is in the smallest size tercile, otherwise 0. All regressions include credit rating and year dummies. Numbers in the parentheses under the coefficients are the associated t-statistics. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively.
Dependent variable: Tobin’s Q Earnings management measures in different dimensions EMsmooth EMcorrelation EMDAC EMsmall-loss c 1.8501*** 1.8468*** 1.7673*** 1.8513*** 1.7400*** 1.7624*** 2.0915*** 2.1173*** ( 7.72) ( 7.82) ( 5.68) ( 5.93) ( 7.08) ( 7.23) ( 8.42) ( 8.61)
Size -0.1375*** -0.1375*** -0.1007*** -0.1107*** -0.1173*** -0.1207*** -0.1487*** -0.1526*** (-5.49) (-5.59) (-3.16) (-3.42) (-4.64) (-4.83) (-5.63) (-5.83)
ROA 2.4804*** 2.4677*** 1.7359*** 1.7376*** 1.6979*** 1.6864*** 1.5425*** 1.5361*** ( 6.49) ( 6.39) ( 3.03) ( 3.07) ( 3.22) ( 3.23) ( 3.18) ( 3.22)
Leverage -0.1902 -0.1818 -0.2627 -0.2540 -0.3888** -0.3938** -0.3559** -0.3434** (-1.28) (-1.23) (-1.57) (-1.53) (-2.21) (-2.24) (-2.18) (-2.13)
R&D 1.8156*** 1.7763*** 1.1241** 1.0872** 1.2305** 1.2090** 1.1708** 1.1389** ( 6.49) ( 6.54) ( 2.33) ( 2.33) ( 2.37) ( 2.40) ( 2.27) ( 2.29)
CAPXR -0.0037 0.0659 -0.0690 0.0131 -0.1172 -0.0084 0.0078 0.1099 (-0.02) ( 0.37) (-0.39) ( 0.07) (-0.69) (-0.05) ( 0.05) ( 0.63)
DIV -0.0746*** -0.0774*** -0.0876*** -0.0905*** -0.0728*** -0.0735*** -0.0697*** -0.0707*** (-5.64) (-5.74) (-5.58) (-5.62) (-6.13) (-6.15) (-6.30) (-6.34)
SEG -0.0408 -0.0559* -0.0501 -0.0671* -0.0660* -0.0830** -0.0373 -0.0562* (-1.20) (-1.67) (-1.38) (-1.86) (-1.79) (-2.27) (-1.10) (-1.67)
GEO -0.0076 -0.0466 0.0239 -0.0224 -0.0041 -0.0480 0.0043 -0.0459 (-0.18) (-1.07) ( 0.51) (-0.47) (-0.09) (-1.03) ( 0.10) (-1.04)
EM -0.0206 -0.0174 -0.2037 -0.2018 -0.0008 -0.0101 0.9990*** 0.9947*** (-0.20) (-0.17) (-1.17) (-1.15) (-0.01) (-0.17) ( 3.23) ( 3.19)
D Large*EM -0.2049* -0.1954* 0.0394 0.0498 0.1558 0.1530 -0.8318*** -0.8346*** (-1.86) (-1.84) ( 0.54) ( 0.69) ( 1.60) ( 1.60) (-2.60) (-2.62)
D Small*EM 0.1286* 0.1221* 0.1180** 0.0972* 0.1713* 0.1595 0.7303** 0.7725** ( 1.82) ( 1.72) ( 2.01) ( 1.67) ( 1.73) ( 1.63) ( 2.14) ( 2.24)
DCD 0.1033** 0.1185** 0.1141** 0.1266*** ( 2.21) ( 2.19) ( 2.22) ( 2.61)
Adj. R2 0.5719 0.5757 0.5120 0.5170 0.5410 0.5454 0.5528 0.5589n obs. 987 987 979 979 1002 1002 1085 1085
53
Table 10 The Impact of EM on Firm Value: Larger Firms vs. Smaller Firms This table presents the panel regression results for the impact of FCD usage and EM on industry-adjusted Q. The regression model is as follows:
( ) titittittittititititijtijtitititi DCDEMSmallDEMeLDEMGEOSEGDIVCAPXRDRLeverageROASIZEDQdjustedindustry-a ,,12,2,11,2,10,2,9,8,7,6,,5,,4,3,2,10, *_*arg_& ln εγγγγγγγγγγγγγ +++++++++++++= −−−∑
The total sample includes firm-year observations with non-missing industry-adjusted Q between 2001 and 2003. DCD, EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. SIZE, ROA, Leverage, R&D, CAPXR, DIV, GEO, SEG are defined as Table 5. D_Large equals 1 if the firm is in the largest size tercile, otherwise 0. D_Small equals 1 if the firm is in the smallest size tercile, otherwise 0. All regressions include credit rating and year dummies. Numbers in the parentheses under the coefficients are the associated t-statistics. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively.
Dependent variable: Industry adjusted Tobin’s Q Earnings management measures in different dimensions EMsmooth EMcorrelation EMDAC EMsmall-loss c 4.0760*** 4.0705*** 4.6250*** 4.9420*** 3.1832*** 3.1555*** 4.4486*** 4.5312*** ( 5.88) ( 5.88) ( 3.88) ( 4.20) ( 4.39) ( 4.37) ( 6.68) ( 6.81)
Size -0.3110*** -0.3116*** -0.2354** -0.2703** -0.2146*** -0.2153*** -0.3040*** -0.3169*** (-4.24) (-4.27) (-1.97) (-2.30) (-2.79) (-2.80) (-4.36) (-4.51)
ROA 4.5326*** 4.4809*** 2.9951** 3.0026** 4.8493*** 4.8232*** 2.4595** 2.4492** ( 4.68) ( 4.51) ( 2.53) ( 2.58) ( 4.70) ( 4.60) ( 2.50) ( 2.54)
Leverage -0.7745 -0.7369 -0.9288 -0.8856 -0.6793 -0.6524 -1.0180 -0.9709 (-1.33) (-1.27) (-1.45) (-1.40) (-1.06) (-1.02) (-1.61) (-1.55)
R&D 3.1563*** 3.0319*** 1.2617 1.1305 2.6395*** 2.5321*** 1.7122* 1.6317* ( 4.74) ( 4.69) ( 1.35) ( 1.28) ( 3.06) ( 3.05) ( 1.78) ( 1.77)
CAPXR 0.7755* 1.0049** 0.6380 0.9180* 1.0211** 1.3011*** 0.9677** 1.2619*** ( 1.69) ( 2.13) ( 1.35) ( 1.89) ( 2.17) ( 2.62) ( 2.18) ( 2.75)
DIV -0.0931** -0.1021*** -0.1147** -0.1249*** -0.0503* -0.0530** -0.0500* -0.0532** (-2.36) (-2.65) (-2.51) (-2.80) (-1.91) (-2.02) (-1.88) (-2.01)
SEG 0.0230 -0.0243 -0.0088 -0.0650 -0.0200 -0.0647 0.0164 -0.0360 ( 0.23) (-0.24) (-0.08) (-0.61) (-0.20) (-0.63) ( 0.16) (-0.35)
GEO 0.1020 -0.0249 0.2118 0.0532 0.0318 -0.0921 0.0511 -0.0949 ( 0.79) (-0.17) ( 1.57) ( 0.34) ( 0.26) (-0.66) ( 0.43) (-0.71)
EM -0.2794 -0.2600 -1.2218 -1.2456 1.0803 1.0957 1.6325*** 1.6384*** (-0.96) (-0.89) (-1.63) (-1.64) ( 0.52) ( 0.54) ( 2.60) ( 2.62)
D Large*EM -0.4937* -0.4655 0.1936 0.2337 1.5428 1.1130 -1.9889** -1.9980** (-1.67) (-1.63) ( 0.80) ( 0.99) ( 0.56) ( 0.41) (-2.52) (-2.57)
D Small*EM 0.4976** 0.4681* 0.2694 0.2063 0.4720 0.6146 1.6404* 1.7492* ( 2.10) ( 1.96) ( 1.43) ( 1.12) ( 0.26) ( 0.35) ( 1.81) ( 1.90)
DCD 0.3357** 0.4048** 0.3221** 0.3671** ( 2.07) ( 2.35) ( 2.01) ( 2.33)
Adj. R2 0.3164 0.3212 0.2821 0.2899 0.3215 0.3265 0.2779 0.2852 n obs. 983 983 975 975 1057 1057 1080 1080
54
Table 11 Exchange Rate Exposure and Firm Value This table presents the panel regression results for the impact of exchange rate exposure on Tobin’s Q. The total sample includes firm-year observations with non-missing Tobin’s Q between 2001 and 2003. Firm-specific exchange rate exposure is represented as the absolute value of the exposure coefficient, | fx
iβ |, estimated by equation (13). DCD, EMsmooth, EMcorrelation, EMDAC and EMsmall-loss are defined as Table 1. SIZE, ROA, Leverage, R&D, CAPXR, DIV, GEO, SEG are defined as Table 5. EMH equals 1 if the firm is in the most aggressive EM tercile, otherwise 0. All regressions include credit rating and year dummies. Numbers in the parentheses under the coefficients are the associated t-statistics. ‘***’, ‘**’ and ‘*’ indicate statistical significance at 1%, 5% and 10%, respectively.
Dependent variable: Tobin’s Q
Earnings management measures in different dimensions
EMsmooth EMcorrelation EMDAC EMsmall-loss (1)
(2) (3) (4) (5)
c 1.8611***
1.8790*** 1.8581*** 1.8689*** 1.8510***
( 7.78)
( 7.67) ( 7.67) ( 7.42) ( 8.01)
Size -0.1239***
-0.1248*** -0.1225*** -0.1230*** -0.1216***
(-5.09)
(-5.11) (-5.07) (-4.96) (-5.07)
ROA 1.6562***
1.6347*** 1.6080*** 1.6574*** 1.5810***
( 3.19)
( 3.17) ( 3.14) ( 3.16) ( 3.26)
Leverage -0.3884**
-0.3923** -0.3858** -0.3714** -0.3600**
(-2.31)
(-2.33) (-2.30) (-2.20) (-2.21)
R&D 1.2476**
1.2594** 1.2686** 1.2209** 1.1759**
( 2.38)
( 2.39) ( 2.39) ( 2.37) ( 2.31)
CAPXR -0.0534
-0.0241 -0.0301 -0.0552 0.0103
(-0.32)
(-0.15) (-0.18) (-0.34) ( 0.06)
DIV -0.0676***
-0.0668*** -0.0655*** -0.0672*** -0.0676***
(-5.78)
(-5.74) (-5.70) (-5.69) (-5.82)
SEG -0.0545
-0.0549 -0.0585* -0.0604* -0.0543
(-1.55)
(-1.56) (-1.67) (-1.72) (-1.55)
GEO 0.0166
0.0053 0.0054 0.0076 0.0061
( 0.38)
( 0.12) ( 0.13) ( 0.18) ( 0.14) fx
tti ,2,ˆ
−β -0.0224* -0.0410** -0.0487*** -0.0501** -0.0776***
(-1.81) (-2.17) (-2.62) (-2.55) (-3.03)
DCD× fxtti ,2,
ˆ−β
0.0185 0.0194 0.0212 0.0332
( 0.86) ( 0.91) ( 0.99) ( 1.40)
EMH× fx
tti ,2,ˆ
−β
0.0352* 0.0639*** 0.0282 0.0784***
( 1.76) ( 3.09) ( 1.45) ( 3.01)
Adj. R2 0.5338
0.5351 0.5391 0.5353 0.5446
n obs. 1067
1067 1067 1067 1067