st: IV with robust covariance matrx in case of (H)ACSC standard errors (heteroscedasticity, autocorrelation and spatial corrlelation)

[Garloff Alfred <Alfred.Garloff2@iab.de>]

Hi,

I am trying to implement an estimator for the following problem: Standard errors are allowed to be correlated over time and over space (across regions, which are the observation units), not necessarily to be heteroscedastic. In addition, one RHS variable (basically the only RHS variable besides time and region dummies) is endogenous and I have presumably a good instrument. 

So, here is the question: How to implement this all at a time. In my mind, 

- ivreg2, hac bw() cannot handle the case of spatial correlation
- xtscc, fe cannot handle an endogenous variable

Probably, an approach like doing the first stage by hand and using the projected values in the second stage together with xtscc is feasible. But then, standard errors have to be corrected and I am not able to transfer the correction from a standard "2-stage estimated by hand IV" to the actual case.

Thanks to anyone considering this question!

Alfred Garloff

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/