Dynamic panel with T=17 and N = 5

[Ana Guerrero]

Hi there,

I have a panel data set with N=5 and T=17, and I would like to include lags of the dependent variable, i.e., create a dynamic panel model. Is T=17 sufficient to avoid biased estimates? I’ve read Professor Wooldridge's recommendation to use xtscc, but in my case, I would like to specify a dynamic panel model. What would be the best option for this scenario?

[Maxence Morlet]

The Nickell bias is of order, in the limit, -((1+rho)/(T-1)) I believe, where rho is the coefficient on the lagged dep. var.

I do not think you have the sufficient amount of observations to try GMM.

xtscc is when you have a very long panel and you want to allow for cross-cluster correlation.

[Sebastian Kripfganz]

You could try a bias-corrected estimation with my xtdpdbc command (which requires that all regressors besides the lagged dependent variable are strictly exogenous). But statistical inference is generally challenging with those data dimensions.