Hello, I am using panel data with N= 37 (countries) and T= 31 (1990 to 2020). After reading about literature of dynamic panel methods, I found that when T is small and N is large, it's better to use System GMM. And with large T and small N it's better to use PMG-ARDL My questions are how small should be T and N to choose between the two methods ? And in my case (N= 37 ; T= 31) which method is more suitable, GMM or PMG ? NB : Using the PMG, I found more significant results than GMM, in short run. Thank you
There are several methods for analyzing large N and large T panel data. The method you choose depends on the research question you are trying to answer and the assumptions you are willing to make. Some of the methods include fixed effects estimation of a static model with robust or panel corrected standard errors, common correlated effects estimation of a dynamic model, and others
You should go with Two-step GMM because it has a smaller asymptotic variance than one-step GMM, which increases the GMM estimator's effectiveness and the power of associated tests. So far as I know, in addition to being well suited for Large N and T, it is also robust to heteroscedasticity and serial correlation issues. Can refer to this; https://www.sciencedirect.com/science/article/abs/pii/S0304407618301556
However, if you want to use PMG, I recommend that you also run MG and DFE (Dynamic Fixed Effect) and then apply the Hausman Specification test to select the best-fitted model. MG and DFE regression models are not accessible in EViews, although they are in STATA.
Lastly, you can use DOLS and FMOLS to determine how reliable your results are.
- Badges
- Science topic