6.1. The Incentive Compensation’s Impact on Earnings Smoothing
As earnings volatility changes as a result of earnings management, we additionally verify the above analysis using earnings volatility. Earnings volatility is an important index to indicate risk, and thus, bank managers can decrease recognized risk by decreasing earnings volatility. Therefore, it is known that bank managers smooth earnings to reduce reported earnings’ volatility. Wahlen [
64] and Collins et al. [
60] have found that bank managers use the discretionary loan loss provision for earnings smoothing. If the latent earnings decrease (or increase), the managers can increase (or decrease) reported earnings by decreasing (or increasing) the
LLP, which reduces earnings volatility. Therefore, if earnings are smoothed using
LLP, the
LLP and net income before provisioning will exhibit a positive (+) correlation.
However, variable compensation or equity-linked compensation may weaken earnings smoothing of banks.
Table 4 indicates that variable and equity-linked compensation act as an incentive for upward earnings management. The incentive provided by these types of compensation is expected to indicate a positive (+) relationship with earnings. The variable compensation linked to accounting profit acts as an upward earnings management incentive when latent earnings are within the scope of the bonus payout, and as a downward earnings management incentive when the bonus payout criteria are not met. To maximize equity-based compensation, it is advantageous to manage earnings upward when latent earnings reach the short-term upper threshold. Therefore, as earnings increase, upward earnings management will also increase, which will increase (or decrease) earnings volatility (or earnings smoothing).
Meanwhile, deferred compensation and earnings smoothing may positively correlate, as firms’ aversion to high risk can increase earnings smoothing. As banks’ earnings volatility is recognized as a risk indicator, managers with high risk-aversion tend to make efforts to decrease earnings volatility. Managers with large inside debt holdings are more likely to protect the value of these holdings by undertaking less risky financing and investing activities [
73]. Van Bekkum [
74] analyzed US banks to find that deferred compensation negatively correlated with share price volatility and the risk of failure. Bennett et al. [
75] found that banks’ risk of failure decreased, and its management performance increased, with larger deferred compensation. Further, Tung and Wang [
46] argued that during the financial crisis, a positive correlation existed between inside debt and banks’ management performance. Inside debt restricts operational risks which increase earnings volatility. Empirical analyses reveal that interest rate risk hedging or low-risk syndicated loans could increase if bank managers’ inside debt increased [
76,
77]. Although these results can be attributed to the increased risk aversion caused by inside debt, the decrease in volatility might also partially occur due to earnings smoothing. The larger the deferred compensation is, the greater the earnings smoothing. Therefore, the relationship between the compensation system and earnings management is further verified through earnings smoothing.
Accounting studies of nonfinancial firms measure earnings smoothing as the difference between net profits and operating cash flows [
78]. However, if the loans expand in financial firms, the operating cash flows will decrease compared to the net income, and cash flows will increase if the deposits expand. The difference between operating cash flows and net income is not the result of earnings smoothing, but intrinsic business activity. Thus, the incentive compensation’s impact on banks’ earnings smoothing is verified using the following equation incorporating a model by Ahmed et al. [
62] and DeBoskey and Jiang [
79], which verifies banks’ earnings smoothing by estimating the
LLP’s impact on income volatility:
Banks can decrease reported earnings’ volatility by either decreasing
LLP when latent earnings decrease, or by increasing
LLP when earnings increase. Thus, managers who use discretion to smooth earnings will recognize
LLP more if the
EBP increases, and less if the
EBP decreases, to increase reported earnings [
62]. Therefore, the regression coefficient of
EBP is expected to be positive (
γ12 > 0). The variables of interest in Model (3) include the
EBP*VARIABLE,
EBP*EQUITY_LINKED, and
EBP*DEFERRAL, which are interaction variables between the incentive compensation and
EBP. If
γ1, the coefficient of
EBP*VARIABLE, is statistically significant and negative in Model (3), the higher proportion of variable compensation can result in a smaller earnings smoothing. If
γ2, the coefficient of
EBP*EQUITY_LINKED, is statistically significant and negative, the higher proportion of equity-linked compensation can result in smaller earnings smoothing; if
γ3, the coefficient of
EBP*DEFERRAL, is statistically significant and positive (or negative), the higher proportion of deferred compensation can result in larger (or smaller) earnings smoothing.
The dependent variable in Model (3) is
LLP, representing the loan loss provision, and can be used for capital management and signaling effects as well as earnings smoothing [
62]. Therefore, the model includes
TIER1, which represents the ratio of beginning regulatory capital, to control the impact of capital management and signaling effects on
LLP. Additional control variables affecting
LLP include beginning non-performing loans (
BEGNPL) and the change in non-performing loans (
CHNPL). Current
NPL as well as that from the previous year-end impact
LLP [
80]. If managers recognize the loss in a timely manner,
LLP relates to the current
CHNPL; however, if they intend to delay this loss recognition, it relates to the previous
NPL. Therefore,
BEGNPL and current
CHNPL are included as independent variables to control for both previous and current non-performing loans. Based on previous studies, loan size (
LOANS), the bank size (
TALN), profitability (
LOSSNET,
EBP), and the loan charge-off (
LCO)—the variables that might impact
LLP—were included in the control variables [
62,
80,
81]. We also controlled for the loan composition with an impact on the
LLP.
Table 5 displays the results from analyzing incentive compensation’s impact on earnings smoothing using Model (3). Regarding the regression coefficients’ significance, the standard errors for two-way clustering were presented in the same way as in the previous analysis. The earnings before taxes and provisions (
EBP) positively correlated with
LLP; if latent earnings are high, the
LLP increases to decrease reported earnings, and if earnings are low, the
LLP decreases to increase reported earnings, leading to earnings smoothing. The regression coefficient of
EBP*VARIABLE was estimated as significant and negative (−1.6403), and that of
EBP*EQUITY_LINKED also produced a significant, negative value (−0.3763). The regression coefficient of
EBP*DEFERRAL was significant and positive (1.3876), as the positive correlation between
LLP and net income increased with a larger proportion of inside debt. Thus, variable and equity-linked compensation decrease banks’ income smoothing while deferred compensation increases it, as anticipated.
6.2. Endogeneity
Endogeneity can arise from reverse causality or omitted correlated variables in our setting. Prior research claims performance, earnings management, and corporate governance are endogenously determined [
21]. If earnings management is prevalent, then the compensation committee can design compensation structures to provide executives with more inside debt compensation on the expectation that inside debt inhibits earnings management. As the variable or equity-linked compensation amounts vary depending on management performance (profitability) and the share price, it is difficult to determine whether the prior period’s variable or equity-linked compensation was predetermined by considering the level of earnings management. However, the possibility exists that the ratio of deferred compensation is predetermined by reflecting the level of earnings management; therefore, it is necessary to analyze the causal relationship after controlling for earnings management’s impact on deferred compensation to verify the deferred compensation’s impact on earnings management. Thus, this study conducts an additional analysis with two-stage least squares (2SLS) regression to consider the possibility of endogeneity. It is adopted for the estimation of explanatory variables unrelated to the residual to eliminate endogeneity that may occur if there is a correlation between the residual and the explanatory variables.
To estimate the deferred compensation variables in the first-step regression, the following are used as instrumental variables: the marginal tax rate, the change in the future tax rate, the salary increase compared to the previous period, and the rate of salary increase. As the related tax burden is deferred if the compensation is deferred, the economic benefit of deferred compensation increases with a higher personal income tax rate. Therefore, if a higher marginal income tax rate is applied to managers, the preference for deferred compensation will be higher [
82,
83].
Changes in the tax rate also impact managers’ preferences for deferred compensation. If the future tax rate increases, the benefits of deferred compensation decrease; conversely, if the tax rate decreases, the value of deferred compensation increases. Although the personal income tax rate and its change impact deferred compensation, they do not logically relate to earnings management and were used as instrumental variables [
56]. Moreover, Cen [
84] analyzed the determinants of deferred compensation to demonstrate that deferred compensation increased when cash compensation and managers’ wealth were greater. This is because the deferred payment’s benefit of tax savings increases with a decreased need for cash. Therefore, managers’ wealth and the size of cash compensation were used as instrumental variables, as they impact deferred compensation and do not logically relate to earnings management. While Cen’s [
84] study measured wealth based on the change in managers’ equity value of the firm, this study measured it with the average compensation of executives subject to remuneration disclosure, in considering the availability of research data. An over-identification test for the instrumental variables’ validity indicated no correlation between instrumental variables and error terms (
p = 0.3995). Multicollinearity among the regressor variables does not have sever effects on the estimation of parameters with VIF is 1.25~5.02 (mean 2.67). The firm individual fixed effects have taken into account.
The results of the 2SLS analysis in
Table 6 indicate that a significant, negative correlation exists regarding the variable and equity-linked compensation’s relationships with discretionary
LLPs, but the deferred compensation’s regression coefficient exhibited no statistical significance. Therefore, it is suggested that the analysis result of
Table 4 can be attributed to deferred compensation’s impact on earnings management, rather than endogeneity.
Although the variables affecting accounting choices were controlled across all possible ranges, the possibility still exists that the omitted variables correlate with both incentive compensation and accounting choices. For example, firms with high growth expectations are likely to have low inside debt incentives [
45,
85], and these firms are also likely to have strong incentives to engage in earnings management to avoid missing earnings benchmarks [
86]. Considering this, we verify the correlation between the current incentive compensation and accounting choices. As current incentive compensation is decided in the next period, the manager cannot know the details of such compensation while preparing the current financial statements. Therefore, this implies that the relationship between current incentive compensation and accounting choices involves the impact of a variable that simultaneously influences both incentive compensation and accounting choices.
As
Table 7 illustrates, current incentive compensation does not significantly impact earnings management. Therefore, the result from
Table 4 has not occurred from an endogenous variable that simultaneously impacts both the compensation and earnings management. Additionally, the analysis of the relationship between the current compensation and earnings smoothing indicated no significant relationship between variable compensation and earnings smoothing, which contrasts the observation that the prior year’s variable compensation significantly reduces earnings smoothing. The degree to which the current variable compensation weakened earnings smoothing was statistically insignificant at −0.4971, but the prior year’s variable compensation (−1.6403) was statistically significant at the 1% level. Although some significant relationship exists between the current deferred compensation and current earnings smoothing, the sizes of regression coefficients and significance were smaller than those from the prior year’s deferred compensation. The regression coefficient to estimate the current deferred compensation’s impact was 0.9671, which was smaller than that of the prior periods’ deferred compensation (1.3876), and the statistical significance (at 5% level) was also lower than that of the prior year’s compensation at the 1% level. However, the impact of current equity-linked compensation on earnings smoothing did not significantly differ from the prior year’s compensation.
Additionally, we directly test whether earnings management affects compensation by analyzing earnings management’s impact on variable compensation. We find that the earnings management from both (t − 1) and t have no significant effect on variable compensation; the results are not reported for brevity.
6.4. Additional Analysis
When
ALLP is a continuous variable, outliers may affect the results. Further, omitted variables or the non-linear relationships between variables may also have an impact. Therefore, we transform the dependent variable in Model (2) into a binomial variable to perform a logit analysis, the results of which are similar to those in
Table 4.
Foreign ownership of banks has recently increased, and thus, foreign investors seeking short-term stock investment profits in addition to the FSB compensation standard may impact managerial decision making on the earnings management. Therefore, a foreign ownership variable (foreign) is added to Model (2) to analyze foreign ownership’s effect on banks’ earnings management, which does not reveal a significant relationship. An additional analysis was conducted using foreign ownership as a mediating variable, and consequently, foreign ownership does not significantly affect the relationship between the compensation system and earnings management.
A variety of corporate governance variables affect the compensation systems and are a significant indicator of managerial sustainability. The diversity of the board (gender, professionality, etc.) in relation to corporate governance can affect the relationship between compensation and transparency of accounting. Therefore, it is necessary to estimate the impact of these variables on compensation and sustainability. However, since the data such as the gender of executives cannot be obtained, this study does not include these variables. Other research suggests that the variables affecting accounting transparency vary depending on the corporate environment. Therefore, social context should be considered when expanding this study to verify the relationship between compensation and earnings management in other social contexts.