256 /342
美國產險業
CEO
更迭與再保險需求
256
income after dividends to policyholders, but before federal and foreign income taxes divided
by net admitted assets and growth ratio of net written premiums. Both F-statistics and
J-statistics
18
are used to check instrument relevance and to perform tests of overidentifying
restrictions for exogeneity issue (Stock and Watson, 2007). The F test statistics for
instrument variables relevance test are significant at the 1 percent level, suggesting the
instrument variables are relevant. If J statistics report statistically insignificant, it indicates
the instruments may be valid. In the second stage, the residual of the endogenous variable is
added to the original regression model. If a coefficient of residual of CEO turnover,
organizational structure or corporate governance variables is statistically significant, the
variable is considered to be endogenous; hence the predicted value of the variable replaces
the original value in the regression model. Finally, we do not reject the hypothesis of Durbin-
Wu-Hausman (DWH) test in all models. In other words, CEO turnover, organizational
18 The results are as follows. CEO turnover is an endogenous variable in the Models of change in
reinsurance ratio, reinsurance demand from affiliated reinsurers and reinsurance from non-affiliated
reinsurers, the F-test statistic for the IV relevance test are 4.834, 4.834, and 4.834 significant at the 1%
level, respectively, suggesting the instrument variables are weak. The J-test statistic for the exogeneity
test are 8.446 (0.489), 7.284 (0.608), and 4.906 (0.842) and insignificant, implying that instrument
variables are exogenous. Mutual is an endogenous variable in the Models of change in reinsurance ratio,
reinsurance demand from affiliated reinsurers and reinsurance from non-affiliated reinsurers, the F-test
statistic for the IV relevance test are 24.553, 24.553, and 24.553 significant at the 1% level, respectively,
suggesting the instrument variables are strong. The J-test statistic for the exogeneity test are 12.457
(0.330), 10.909 (0.451), and 3.601 (0.980) and insignificant, implying that instrument variables are
exogenous. Duality variable is an endogenous variable in the Models of change in reinsurance ratio,
reinsurance demand from affiliated reinsurers and reinsurance from non-affiliated reinsurers, the F-test
statistic for the IV relevance test are 2.018, 2.018, and 2.018 significant at the 1% level, respectively,
suggesting the instrument variables are weak. The J-test statistic for the exogeneity test are 12.007
(0.445), 13.507 (0.333), and 10.759 (0.549) and insignificant at the 1% level, implying that instrument
variables are exogenous. Board size is an endogenous variable in the Models of change in reinsurance
ratio, reinsurance demand from affiliated reinsurers and reinsurance from non-affiliated reinsurers, the
F-test statistic for the IV relevance test are 11.659, 11.659, and 11.659 significant at the 1% level,
respectively, suggesting the instrument variables are strong. The J-test statistic for the exogeneity test are
11.252 (0.508), 13.601 (0.327), and 14.281 (0.283) and insignificant at the 1% level, implying that
instrument variables are exogenous. Independent director is an endogenous variable in the Models of
change in reinsurance ratio, reinsurance demand from affiliated reinsurers and reinsurance from non-
affiliated reinsurers, the F-test statistic for the IV relevance test are 6.755, 6.755, and 6.755 significant at
the 1% level, respectively, suggesting the instrument variables are weak. The J-test statistic for the
exogeneity test are 11.043 (0.525), 9.372 (0.671), and 12.530 (0.404) and insignificant at the 1% level,
implying that instrument variables are exogenous. Big 4 auditor is an endogenous variable in the Models
of change in reinsurance ratio, reinsurance demand from affiliated reinsurers and reinsurance from non-
affiliated reinsurers, the F-test statistic for the IV relevance test are 10.748, 10.748, and 10.748 significant
at the 1% level, respectively, suggesting the instrument variables are weak. The J-test statistic for the
exogeneity test are 10.498 (0.572), 13.639 (0.324), and 9.057 (0.698) and insignificant at the 1% level,
implying that instrument variables are exogenous.