arellano bond estimation
Posted: Thu Dec 08, 2016 4:00 am
Hi
I have a panel data of 34 countries and 24 time periods . I want to estimate a dynamic panel equation of the form
Yi,t = Yi,t-1 + Xi,t + ai +ei,t
the main aim of my estimation is to compute the cross country fixed effects and then explain the difference by institutional factors, but as i understand estimating dynamic panel equation of this form by fixed effects is biased and in the literature it is sugested to use arellano bond technique which asumes diferencing the data in order to remove cross country unobservables (ai) and then use GMM estimation and lags of variables as an instruments. So this is the estimation output
1)Can anybody explain me why the t statistics are so high?
2)what exatly means j-statistic
and the last and most important question for me can i recover from this estimation the country unobservables , I mean can i use the coeficient estimates of differenced variables to compute ai = Yi,t - betta* Yi,t-1 - gamma*Xi,t - residi,t
I have a panel data of 34 countries and 24 time periods . I want to estimate a dynamic panel equation of the form
Yi,t = Yi,t-1 + Xi,t + ai +ei,t
the main aim of my estimation is to compute the cross country fixed effects and then explain the difference by institutional factors, but as i understand estimating dynamic panel equation of this form by fixed effects is biased and in the literature it is sugested to use arellano bond technique which asumes diferencing the data in order to remove cross country unobservables (ai) and then use GMM estimation and lags of variables as an instruments. So this is the estimation output
- Untitled.png (22.01 KiB) Viewed 6485 times
2)what exatly means j-statistic
and the last and most important question for me can i recover from this estimation the country unobservables , I mean can i use the coeficient estimates of differenced variables to compute ai = Yi,t - betta* Yi,t-1 - gamma*Xi,t - residi,t