I am working on a project that has 4680 observations (18 years, 265 regions), in which one variable is endogenous. We tried various combinations of two IVs, but always end up with a significant Sargan-Hansen J Test. Recently I have found this quote:

"Testing of over-identifying assumptions is less important in longitudinal applications because realizations of time varying explanatory variables in different time periods are potential instruments, i.e., over-identifying restrictions are automatically built into models estimated using longitudinal data."

Does it mean that for panel data that contains large number of observations the Sargan-Hansen J Test is irrelevant as it almost always ends up significant?

The precondition of the test is that at least one of the IVs is a relevant IV. Should we therefore restrict ourselves to using one IV?


I disagree with the quote. For example, in dynamic panel data models $$ y_{it}= \alpha_i+\rho y_{it-1}+u_{it} $$ higher lags of $y_{it}$ are invalid instruments if $u_{it}$ is itself autocorrelated, i.e., if the dynamics of the system have not been specified sufficiently richly.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.