How short is the short period of ‘T’ required in using the Dynamic Panel (GMM) Model? Could T=4 work? My panel dataset is such that N (cross-sectional unit) is more than 300 and T (years) is only 4 (short). Is it possible to consider dynamic panel analysis where, in the specific case, model specification with at least one lagged variable is justifiable?
T=4 works, but you can test for serial correlation in the residuals only for some few lags. Depending on the research question, you can add several lagged explanatory variables, but I generally would use contemporaneous values. As said, it depends on the research question.
Dear Wazza,
I agree with @Philipp. The model stability tests may be good guide.
Technically, such an analysis is possible; but the results can only be interpreted with great caution!
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments