based/total compensation ratio as the explanatory variable and run the
same QR in Table 4 to check the robustness of the empirical findings.
Following the suggestion in Johnston and DiNardo (1997), we use
the lagged RoE and stock-based/total compensation ratio as the
instrument variables and perform the Hausman test. In the test, we
compare the difference in estimates between OLS and two-stage least
squares (2SLS). The statistics is 2.95 and p-value is .815, and we cannot
reject the null hypothesis that the estimates of OLS and 2SLS are
equivalent. Therefore, based on the test, we conclude the problem of
endogeneity is not a substantial problem in our empirical testing.