Testing for endogeneity: New feature for eteffects in Stata 14
There has been great interest in Stata 14’s eteffects, which obtains treatment effects when unobserved variables affect both treatment assignment and outcomes.
If an unobserved variable affects treatment and outcome, we have an endogeneity problem and cannot obtain accurate estimates of effects using conventional treatment-effects estimators.
We have enhanced eteffects. We can now test if the unobserved variable affects both outcome and treatment.
eteffects estimates the correlation between the unobservables that affect treatment and outcome. If these correlations are 0, we have no endogeneity. We test this hypothesis by typing
. estat endogenous
It is that simple.