EViews.com

Hi
I am trying to estimate multiple equations by GMM using "System" in Eviews7. My problem is that I am to use GMM robust standard errors, but when I check the box "Identity weighting matrix in estimation (2SLS coefs & GMM robust errors)" my coefficients not only go from significant at the 1% level to being highly insignificant. (t-statistics in the magnitude of 0.01 and a p-value of around 0.99) but my coefficients themselves also change in value. This puzzles me because in my book the coefficients should not change due to using GMM robust errors. Any ides as to what I am doing wrong or what the problem is?
Thanks.

As that option says, you are using an identity matrix as the weighting matrix during estimation, which changes the estimation to be 2SLS, not traditional GMM. This will change your coefficient estimates.