4. Empirical Results
4.3 Correlation Analyses
Table 4 presents Spearman and Pearson correlations among the main variables in our equations. As show in table 4, the correlations between the components of the CEO power index: CEO_HOLD, CEO_FAM and DUALITY are low. The correlations among other variables are also low, except for the correlation between positive earnings (POS_INC) and negative earnings (NEG_INC) is high. The overall results of low inter-correlation among all independent variables indicate that multi-collinearity does not appear to be a problem in the regression model. We subsequently adopt the Variance Inflation Factor (VIF) to check for the potential multi-collinearity problem.
[Insert Table 4 here]
4.4 Empirical Results
4.4.1 Empirical Tests of the Relationship between Asset Write-downs and CEO Compensation
Table 5 reports the results for tests of H1. Panel A presents the coefficient estimates, while Panel B reports results for tests of differences among the equation coefficient estimates. Panel A shows that the coefficient of POS_INC is 0.2000 and is
27
significantly at the 0.01 level. The coefficient of WOTA is -0.055 but is insignificantly.
Panel B shows that the coefficient of POS_INC is significantly higher than the coefficient of WOTA at the 0.05 level (POS_INC-WOTA). The evidence indicates that compensation committees fully shield CEO compensation from the effect of asset write-downs. The result is similar with prior literature that the compensation committee breaks out the components of earnings and intervenes in the compensation process to shield CEO cash compensation from the earnings-decreasing effects of restructuring charges (Dechow et al., 1994; Adut et al., 2003) and strategic expenditures (Duru et al, 2002) to provide proper incentives for CEOs to engage in activities that enhance firm value. Those evidences are from the viewpoint of optimal contracting hypothesis. However, this favorable (to the executive) state of affairs could stem from managerial entrenchment (Gaver and Gaver 1998; Cheng 2003;
Henderson et al. 2010), an alternative explanation for shielding CEO compensation from the effect of losses. Therefore, we will focus on the managerial power hypothesis in hypothesis two.
Panel A of Table 5 shows that the coefficient of BOARD_SIZE is negative but insignificantly because there is mixed results that one member increase in the size of the board is associated with total CEO compensation (Hallock, 1997; Core et al., 1999;
Yermack, 1996; Cyert et al., 2002). The coefficient of BOARD_HOLD is negative but insignificantly. The reason is that the director ownership can help align the interests of directors with those of shareholders. Thus, higher director shareholdings might limit excessive CEO compensation packages leading to a negative relation between director ownership and CEO compensation. However, increased director ownership can also give directors greater voting power and control, which could lead to their entrenchment (Ozkan, 2011). The coefficient of IND_BOARD is significant
28
negative which suggest that more independent outside directors are associated with less CEO compensation (Core et al., 1999). The coefficient of RET is positive but insignificant. Prior literature indicates that there is no significant relation between stock return and CEO compensation (Fu et al., 2002; Lin and Liu, 2003) because a Taiwanese firm’s stock price is affected by much uncertainty unrelated to managerial efforts (Hung and Wang, 2005).
The coefficient of RISK is significant negative which suggest that CEO compensation is less at firms with greater stock return volatility. Finally, the coefficient of LN(TENURE) is significant positive which support that the longer tenure were associated with higher levels of pay (Attaway, 2000; Johnston, 2002). The coefficient of MB is significant positive. This finding is consistent with Smith and Watts’s (1992) argument that firms with greater growth opportunities should have more incentive pay. The coefficient of SIZE is significant positive suggest CEO compensation is positively associated with firm size (Tosi et al., 2000; Dorata and Petra, 2008).
[Insert Table 5 here]
4.4.2 The Effect of CEO Power on the Relationship between Asset Write-downs and CEO Compensation
Table 6 reports the results for tests of H2. Panel A presents the coefficient estimates, while Panel B reports results for tests of differences among the equation coefficient estimates. Hypothesis 2 states that more powerful CEOs experience smaller decreases in compensation by influencing compensation committee to shield CEO compensation from the adverse effect of asset write-downs. The main effect variable WOTA captures the impact of asset write-downs on CEO compensation for the least powerful CEO（HI_POWER=0）. The interaction term HI_POWER×WOTA
29
captures the differential impact for most powerful CEOs.
Panel A shows that the coefficient of POS_INC is 0.199 and significantly at the 0.01 level. The coefficient of WOTA is 0.005 but insignificantly. The coefficient on the interaction term HI_POWER×WOTA is -0.126 and significantly at the 0.01 level.
Taken together with the insignificant coefficient on WOTA and the significantly negative coefficient on HI_POWER×WOTA, this evidence indicates that compensation committees fully shield compensation of CEO with high power from the earnings effect of asset write-downs.
Panel B shows that the coefficient of POS_INC is insignificantly larger than the coefficient of WOTA. Hence, we find no evidence that compensation committees shield compensation of CEO with low power from asset write-downs. The coefficient of POS_INC is significantly larger than the coefficient of (WOTA+HI_POWER×WOTA) at the 0.01 level. The finding is also suggest the managerial power hypothesis that entrenched managers may influence their own pay arrangements to favor themselves (Grinstein and Hribar, 2004; Bebchuk et al., 2007;
Collins et al., 2009; Henderson et al. 2010). Therefore, shielding executive annual pay from losses reflects managerial entrenchment (Gaver and Gaver, 1998; Cheng, 2003).
The results of control variables are similar to the hypothesis one. Therefore, we do not interpret the results again.
[Insert Table 6 here]
4.4.3 The Effect of Individual CEO Power Measure on the Relationship between Asset Write-downs and CEO Compensation
Table 7 presents the estimation results of equation (3). In this specification, we modify equation (2) by including three individual measures of managerial power. As shown in Panels A, the coefficient of POS_INC is 0.197 and the coefficient of WOTA
30
is 0.106. However, Panel B shows that the coefficient of POS_INC is insignificantly larger than the coefficient of WOTA. That is, the compensation committees do not shield compensation from asset write-downs for CEO with low power. Additionally, the coefficients of (WOTA HI CEO HOLD WOTA _ _ ), (WOTACEO_FAMWOTA), and
)
(WOTADUALITYWOTA are all significantly lower than the coefficient of POS_INC at least 0.1 level, respectively. Thus, the evidence suggests that compensation committees fully shield compensation from asset write-downs for CEOs who have higher shareholding, are also one of family members and are also in the position of chairman. Therefore, using the CEO Power Index and individual measures to test our hypotheses get the same results.
[Insert Table 7 here]