Choosing the Correct Panel Data Model

[Travis Mitchell]

Hi there, I am estimating a fiscal reaction function in dynamic form (lagged dependent variable) and trying to choose the right model. My dataset is N=13 and T=18. A previous study has used system GMM but I am doubtful about the method because from Roodman (2009) system GMM is dubious for N<20. Also the system GMM method was developed under the assumptions of N large and T small. My data sets shows signs of endogeneity, cross-sectional correlation, heteroscedasticity and autocorrelation. Another study on the same region use Panel FE IV regressions but this does only takes care of endogeneity and not the other issues. I am thinking that a panel model with 2 lags of the endogenous variable and panel corrected standard errors should suffice? May also with FGLS?

Newt to Statalist.

[Carlo Lazzaro]

Travis:
welcome to this forum.
1) -xtivreg- id for static panel data regressions;
2) can't you go -xtabond- in the dynamic setting?

[Travis Mitchell]

Thank you Carlo. My concerns are: (1) xtireg does not allow the researcher to control for the other ills - autocorrelation and heteroscedasticity unless I am missing someting? (2) Ideally I'd like to go dynamc but the diff-GMM and sys-GMM are set up for large N and small T. My data has small N and small T, and more precisely N<20.

[Carlo Lazzaro]

Travis:
1) -xtivreg- provides researches with -vce(cluster clusterid)- option for non-default standard errors;
2) Quoting -xtabond- Description from its entry in Stata .pdf manual:

xtabond fits a linear dynamic panel-data model where the unobserved panel-level effects are correlated with the lags of the dependent variable, known as the Arellano–Bond estimator. This estimator is designed for datasets with many panels and few periods, and it requires that there be no autocorrelation in the idiosyncratic errors.

[Travis Mitchell]

Thank you Carlo noted on xtivreg. On 2. it supports my point that xtabond is not suitable because my panels are 13. This does not qualify as "many panels". Thanks again.