Lagged dependant variables, RE and Pooled OLS

[Anubhi Ghuwalewala]

Dear everyone,

When I include a lagged dependent variable to my RE regression, then sigma_u=0 and the RE is the same as a Pooled OLS.

How do I test whether this model is better than a Fixed effects model? Furthermore, why is it resulting in a Pooled OLS?

[Phil Bromiley]

Please follow the FAQ on asking questions - provide Stata code in code deimiters, Stata output, and sample data using dataex.

You generally cannot use xtreg to estimate models with a lagged dv - the lagged dv generally results in inconsistent parameter estimates. There are many ways to estimate such models including xtivreg, xtivreg2, xtabond, the alternative estimators related to xtabond, xtdpdmt (user written), and SEM.

The lagged dv essentially controls for any stable cross-firm differences making xtreg,re look like ols.
You compare a re and fe model with a hausman test - look it up. I think this is discussed in the xtreg documentation and under hausman.