7. Sensitivity test
Because a company’s inclusion in the SSE Corporate Governance Sector is the result of a self-selection (voluntary
application) process,14 it is possible that some companies in the Non-governance Sector have good corporate
governance, but simply have not submitted an application. If this is the case, the results will be biased.
To determine whether our conclusions are robust, we perform a sensitivity test using the two-stage procedure
developed by Heckman (1976).
In the first stage, we estimate a Probit choice equation and obtain inverse Mills ratios. In the second stage,
we include the inverse Mills ratios as an explanatory variable in the primary model to control for the potential
endogeneity induced by self-selection. A number of strict constraints are necessary in implementing the Heckman
(1976) procedure successfully. For example, at least one exogenous independent variable that has no
direct effect on the dependent variable in the second-stage regression should be included in the first-stage
choice model. Lennox et al. (2012) find that many accounting studies fail to select proper variables when using
selection models and thus obtain inconsistent results.15 Hence, we carefully select the explanatory variables in
the first stage and include 11 factors considered to have an effect on corporate governance, such as Auditcomm
(establishment of an audit committee), Dual (CEO duality), DirScale (board of director scale), Fnctl (final controller
type) and M_Stockholder (frequency of stockholder meetings) in the choice model. Of these factors, at
least Fnctl and M_Stockholder have no significant effect on audit fees16 and can thus play the role of an exogenous
independent variable excluded in the second-stage regression. The first-stage choice equation is as follows
(Model 3).
7. Sensitivity test
Because a company’s inclusion in the SSE Corporate Governance Sector is the result of a self-selection (voluntary
application) process,14 it is possible that some companies in the Non-governance Sector have good corporate
governance, but simply have not submitted an application. If this is the case, the results will be biased.
To determine whether our conclusions are robust, we perform a sensitivity test using the two-stage procedure
developed by Heckman (1976).
In the first stage, we estimate a Probit choice equation and obtain inverse Mills ratios. In the second stage,
we include the inverse Mills ratios as an explanatory variable in the primary model to control for the potential
endogeneity induced by self-selection. A number of strict constraints are necessary in implementing the Heckman
(1976) procedure successfully. For example, at least one exogenous independent variable that has no
direct effect on the dependent variable in the second-stage regression should be included in the first-stage
choice model. Lennox et al. (2012) find that many accounting studies fail to select proper variables when using
selection models and thus obtain inconsistent results.15 Hence, we carefully select the explanatory variables in
the first stage and include 11 factors considered to have an effect on corporate governance, such as Auditcomm
(establishment of an audit committee), Dual (CEO duality), DirScale (board of director scale), Fnctl (final controller
type) and M_Stockholder (frequency of stockholder meetings) in the choice model. Of these factors, at
least Fnctl and M_Stockholder have no significant effect on audit fees16 and can thus play the role of an exogenous
independent variable excluded in the second-stage regression. The first-stage choice equation is as follows
(Model 3).
การแปล กรุณารอสักครู่..