6.1.3. Instrumental Variable Regression

In this paper, instrumental variables are used for regression to reduce the interference of endogeneity problem on the regression conclusion. drawing on the research of existing literature [45], the uncertainty of global economic policy will affect the uncertainty of China's economic policy, which accords with the correlation characteristics of the instrumental variables. However, the uncertainty of global economic policy will not directly affect the choice of C-suite self-interest behavior in China, which accords with the exogenous characteristics of the instrumental variables. Therefore, this paper uses the global economic policy uncertainty index as an instrumental variable to alleviate the endogeneity problems that may exist in the original regression model. Its calculation method is the same as the measurement method of China's economic policy uncertainty in the main regression, obtained by calculating the logarithm of the annual arithmetic average. Table 11 illustrates the regression results for the instrumental variables. From the results of Table 11, we can find that the regression coefficients of economic policy uncertainty and executive excessive compensation are always negative and significant, while the regression coefficients of economic policy uncertainty and executive excessive on-the-job consumption are always positive and significant, which is consistent with the original regression results. Therefore, the model in this paper can still be used to draw consistent research conclusions after using instrumental variables to control endogeneity problems.




**Table 11.** *Cont.*

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

#### 6.1.4. Adding Macro-Level Control Variables

The conclusions of this paper may be subject to endogeneity problems due to the lack of macro-level variables. To alleviate this issue and prevent endogenous errors caused by changes in macro-economic factors, this paper refers to the research of Li and Yang [46], Gulen [4], etc., and adds the year-on-year GDP growth variable to the control variables of the original regression model. The regression results are illustrated in the Table 12. After controlling for the macro-factors, the regression results are still stable, alleviating the endogenous factors caused by the lack of macro-level variables.

**Table 12.** Adding macro-control variables.



**Table 12.** *Cont.*

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

#### 6.1.5. Controlling the Fixed Effects of Provinces

Since the uncertainty levels of different regions can be different, the influence of provincial factors can only be overlooked if controlling for the influence from industries. Therefore, the fixed effects of provinces were added to the robustness test, so as to reduce endogeneity problems. As can be seen from Table 13, the regression coefficient between economic policy uncertainty and OverPay remains negative and significant at the level of 1%. The regression coefficient between economic policy uncertainty and executives' excessive on-the-job consumption is still positive and significant, which is consistent with the main regression results.

**Table 13.** Controlling provinces as fixed effects.



**Table 13.** *Cont.*

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