Dear Stata users,
I would like to ask you if you deem my reasoning behind my sargan test results sensible?
When I test for validity of instruments after incorporating one lag of the dependant variable (in a dynamic panel estimated using the Bond-Blundell model), the validity is rejected. When I increase the number of lags to four lags, the validity of instruments is not rejected.
Could it be that with one lag included one of the dependant variables in the panel is under-fitted, or exhibits an outlier in the errors, whereas when I increase the number of lags the errors are smaller or the outlier is now explained with the model, and then the sargan test does not reject the validity of instruments hypothesis?
I am afraid that it is simply that the sargan test less often rejects the null when more lags are included... I have tried several other options (max lags for the instruments, subsample results,…).
Thank you for any kind help!
Kind regards,
Vasja
I would like to ask you if you deem my reasoning behind my sargan test results sensible?
When I test for validity of instruments after incorporating one lag of the dependant variable (in a dynamic panel estimated using the Bond-Blundell model), the validity is rejected. When I increase the number of lags to four lags, the validity of instruments is not rejected.
Could it be that with one lag included one of the dependant variables in the panel is under-fitted, or exhibits an outlier in the errors, whereas when I increase the number of lags the errors are smaller or the outlier is now explained with the model, and then the sargan test does not reject the validity of instruments hypothesis?
I am afraid that it is simply that the sargan test less often rejects the null when more lags are included... I have tried several other options (max lags for the instruments, subsample results,…).
Thank you for any kind help!
Kind regards,
Vasja