Page 1 of 1

No Insignificant Values in GMM Estimation

Posted: Wed Nov 15, 2017 9:48 am
by bfluff
Good evening

I'm trying to run an GMM estimation on my unbalanced panel of 38 countries from 1996 to 2016. My dependent variable is NPF. The six explanatory variables (CC GE PS RL RQ VA) are highly correlated so I ran each one individually. Two of my control variables (ldgdp and lgfcf) are non-stationary so in my specification I ran them as first differences. The other control variables are dcci usint and gdpgpc and are run at level.

If I use all six explanatory variables I get one which is not significant. If I run a standard OLS specification it provides a much more varied list of probabilities. If I run one of the explanatory variables all are significant at 1%.

Do I need to add additional variables to this specification to ensure the GMM can identify the "most" significant variables or am I missing something?

I've attached my workfile. An example of running one explanatory variable is given by CC_EQ. All_indicators_eq is where I ran all six explanatory variables concurrently.

This seems to be a rather simple issue but if anyone is able to help I'd be very appreciative.