Some Strange Properties of Panel Data Estimators.

The authors study the biases that are likely to arise in practice with panel data when parameters vary across individuals, but this is not allowed for in estimation. They consider both stationary and non-stationary regressors. They find that biases can be severe for relatively small parameter variation, and that this problem is hard to detect. They study in some detail by Monte-Carlo the performance of the T. Anderson-C. Hsiao estimator in the presence of this particular mis-specification. Copyright 1992 by John Wiley & Sons, Ltd.