EViews.com

Is is correct to include a Lagged Dependent Variable in a Panel with Fixed Effects using OLS?

Will it produce biassed results? If so, why? And will the size of the time parameter determine if it is biassed or not?

Many Thanks

No. Do an internet search for "dynamic panel data" and "fixed effects".

I have done many internet searches and used my text books. I'm aware of Nickel's (1981) findings.

From what I've gathered, the size of the T parameter (time) determines whether the model is biassed or not?

If this is correct, does anyone know how large T must be in order to overcome issues related to bias?

And are there any tests that can be used to diagnose the existence of bias in an estimation?

You're biased until T goes to infinity. Whether the bias is important in a particular setting is an empirical question. Kiviet describes estimation of the bias term which you can use to evaluate your OLS estimator or to correct for the bias.

Kiviet(1995), "On bias, inconsistency, and efficiency of various estimators in dynamic panel data models," Journal of Econometrics, vol. 68, no. 1, 53-78.

Would it be correct to include a Lagged Dependent Variable in a Panel with Fixed Effects using LS with Cross section SUR Weights?

Correct isn't clearly defined here, but that will certainly biased for finite T. The SUR will make things worse.

Dear Eviews Experts:

Will you please suggest any solutions to correct the bias caused by lagged regressor in OLS panel with fixed effect?

Thanks,

Amber

Arellano-Bond GMM (for Dynamic Panel Data) is supported by EViews.