I am investigating conditional convergence across Indian states using panel data. I include the state name, year, SDP per capita, and a number of conditioning variables such as Public Expenditure, Literacy, Rural Banks per Capita. I hypothesize that I will find weak evidence of conditional convergence when controlling for some of the production fundamentals of each state.

I wanted to check that I have done the correct robustness checks for my model. I am using R, and the plm package in particular. The conditional convergence regression will include a specification with a pooled OLS and controlling variables, a specification with state fixed effects and controlling variables, and a specification with state fixed effects as well as time fixed effects and controlling variables.

  • I attempted to do a Hausman test for fixed vs. random effects, but it was unsuccessful as I believe my regressors are too dependent.

  • I did a Breusch-Pagan test for heteroskedasticity

  • I also used the Ramsey RESET test for functional form.

  • I will be using SEs clustered at the state level.

Are there any R tests I am missing for pooled OLS and state/time fixed effects? What assumptions have I left out?

I also was interested in adding a more sophisticated specification: the GMM model. I found in the literature that OLS estimates are inconsistent (biased upwards) in the presence of a lagged dependent variable and fixed effects. GMM was used in a number of the papers I read, but I dismissed it initially. Then I started to think that including a lagged growth rate variable would be prudent. The plm package appears to support GMM models.

I understand that the GMM estimator allows past realizations of the dependent variable to affects its current value, and I think it would add a lot of value, but I am unsure of the assumptions/robustness tests I need to perform. If it is relatively straightforward I would like to move forward with it, but I would appreciate any thoughts.

Also please let me know if any more detail is needed!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.