Abstract
There has been hot debates on the appropriateness of using fixed- versus random effects models in the panel data analysis. Although much has been written on the theoretical properties of both approaches, recommendations for applied researchers are often confusing. In this paper, we discuss four combined fixed- and random-effects estimators, including leave-one-out, inverse-variance weighted, Stein and optimal weights combination methods. We compare the performance of these estimators using a series of Monte Carlo experiments that vary the sample sizes, degrees of endogeneity and degrees of heterogeneity. We then provide the guidance to help researchers to choose among these estimators.