I've been trying to estimate a pVAR using GMM on R using the package panelvar. I'm estimating a dynamic panel VAR with two-step GMM using first-differences.
I have a balanced panel with 378 with a group variable (id) and a time variable (year). I have 14 observations per group (unvaried) and 27 groups in total. In total, I have 120 instruments. I'm a bit concerned about the results of the Hansen J-test and I'm looking for some explanations: I have a Hansen J-test statistic of 0 with a p-value of 1. To my understanding, this would mean that the model is correctly specified. But the fact that the p-value is very high (1.000), it might mean that something deeper is going on.
In my estimation, I have 7 dependent variables, 2 exogenous variables, and I'm using 4 lagged instruments per dependent variable. Why is the p-value of the Hansen test very high?
Thanks in advance!
