*5.4. Regression Analysis*

Table 7 illustrates the test results of Model (6) and (7). The fixed-effects panel regression model of control year and corporate individual (columns 2 and 3), as well as of control year and industry, are used in regression analysis, and corporate-level cluster-robust standard error is adopted to alleviate possible sequence-related problems (columns 4 and 5). The regression results demonstrate that the regression coefficient between economic policy uncertainty (EPU) and OverPay is always negative, which is significant at the level of 1%, indicating that there is a significant negative correlation between economic policy uncertainty and C-suite overpayment; that is, higher economic policy uncertainty will inhibit the C-suite OverPay policy. Therefore, Hypothesis H1a in this paper is supported. The regression coefficient between economic policy uncertainty (EPU) and UnPerks is always positive and significant, which indicates that there is a significant positive correlation between economic policy uncertainty and C-suite on-the-job consumption. Therefore, higher uncertainty of economic policy will encourage the implicit self-interest behavior of executives. Hypothesis H1b in this paper is thus supported.

The economic significance of the regression in Table 7 is presented as follows. For panel regression 1, OverPay increases 1422% when EPU increases by one standard deviation and UnPerks decreases 717% when EPU increases by one standard deviation. For panel regression 2, OverPay increases 835% when EPU increases by one standard deviation and UnPerks decreases 358% when EPU increases by one standard deviation.

The value for economic significance is very large, this may result from the following two reasons: (1) the values for OverPay and UnPerks are relatively small compared to the real amount of managers compensation and on-the-job consumption and (2) change in OverPay and UnPerks can lead to a reasonable economic significance change for the Pay and Perks.

At the level of the control variables, if CEO also serves as Chairman, he/she will be more incentivized to seek excessive compensation. The reason is that "tyranny of the minority" is more likely to occur when the two powers are concentrated, and Csuite will pursue self-interest by excessive compensation. There is a significant positive correlation between the cash flow generated by corporate business activities and on-the-job consumption of executives. When an enterprise generates larger amounts of cash flow, on-the-job consumption of executives will be "easier" to realize and thus be promoted.

Moreover, the control variable Big4 is negatively correlated with the excess on-the-job consumption of executives, which is also significant. Therefore, when an enterprise chooses "Big 4" accounting firms to perform audits, the excessive consumption of executives will be constrained, which is consistent with the previous research conclusions.


**Table 7.** Analysis of principal regression results.

\* *p* < 0.1, \*\* *p* < 0.05, \*\*\* *p* < 0.01.

### *5.5. Moderating Role of Internal Control*

In order to test Hypothesis H2, this paper constructs Models (7) and (8) and adds the internal control index as a moderating variable to study the role of internal control on the relationship between economic policy uncertainty and executives' self-interest behavior. As illustrated in Table 8, if α3 and β3, the coefficients of the interaction term EPU\*IC, are significantly negative, the effective internal control can strengthen the restraining effect of economic policy uncertainty on OverPay, which is conducive to suppressing executives' explicit self-interest behavior. Meanwhile, effective internal control can also weaken the role of economic policy uncertainty in promoting C-suite on-the-job consumption, and help to restrain executives' invisible self-interest behavior. To summarize, effective internal control is conducive to restraining the self-interest behavior of executives. Therefore, Hypotheses H2a and H2b are verified.


**Table 8.** Moderating effect regression.

\* *p* < 0.1, \*\* *p* < 0.05, \*\*\* *p* < 0.01.

#### **6. Robustness Test and Further Analysis**

#### *6.1. Robustness Test*

In order to ensure the reliability of the research conclusion, we conducted the following robustness tests.

6.1.1. Changing the Measurement Method of Executives' Excessive Compensation and On-The-Job Consumption

Referring to the practices of Luo et al. [43], etc., this paper adopts the total monetary compensation of the Top 3 directors, supervisors, and executives as the absolute compensation of corporate executives, then uses its logarithm as the executive compensation variable (Pay\_r) and calculates the replacement variable (OverPay\_r) of the executives through the model. Drawing from the practices of Luo Jinhui, etc., we adjust the management expenses through the main business income to obtain the on-the-job consumption (Perk\_r) of executives. See Table 9 for the regression results. The test coefficients of the main variables in the regression results have not changed and are still significant, which demonstrates the robustness of the conclusion.


**Table 9.** Variables of executives' self-interest behavior under different methods.

\* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

6.1.2. Changing the Measurement Method of Economic Policy Uncertainty Index

When the above regression analysis was carried out in this paper, the method of processing the economic policy uncertainty index was to take the logarithm of its average. In the robustness test, the index is divided by 100 to obtain EPU2, the replacement variable of the economic policy uncertainty index, and then a regression analysis is carried out once again. The regression results are illustrated in Table 10. As can be seen from Table 10, the uncertainty of economic policy has a negative correlation with OverPay and is still significant, while the uncertainty of economic policy has a positive correlation with UnPerks. The main test coefficient in the regression results has not changed, so the conclusion of this paper is robust.

**Table 10.** Variables of economic policy uncertainty under different methods.



**Table 10.** *Cont.*

\* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.
