I have two dependent variables that I will use in two separate models (namely math_score & language_score) and one independent variable (relative_power) that I will use in both models. The two models have the same sample size. I suspect there's endogeneity in relative_power so I add two instruments to address this problem using ivregress 2sls command. When I tried to test the endogeneity with the estat endog command, I get different results for relative_power.
For the model with math_score, the Durbin & Wu-Hausman test statistic is significant, indicating the independent variable relative_power is indeed endogenous, but for language_score the test statistic is not significant, meaning relative_power is exogenous.
My question is that does this often happen? What should I do to conclude the possible endogeneity in relative_power? I have consulted with my Professor about this but they said it's best to look at previous studies and since previous studies claim that the independent variable that I used was endogenous they said I should treat it as such and keep using iv regression for the model with language_score. I am a novice and have not encounter something like this before. Any insight would be appreciated.
For the model with math_score, the Durbin & Wu-Hausman test statistic is significant, indicating the independent variable relative_power is indeed endogenous, but for language_score the test statistic is not significant, meaning relative_power is exogenous.
My question is that does this often happen? What should I do to conclude the possible endogeneity in relative_power? I have consulted with my Professor about this but they said it's best to look at previous studies and since previous studies claim that the independent variable that I used was endogenous they said I should treat it as such and keep using iv regression for the model with language_score. I am a novice and have not encounter something like this before. Any insight would be appreciated.
Comment