This paper presents an assessment of the small-sample performance of the three well-known estimators of components variance in random effects model for panel data. The estimators considered are Swamy-Arora, Wansbeek-Kaptayn and Wallace-Hussain. To this end, by simulating a one-way error component model in the form of random effects, small sample performance of three variance estimators is studied. The implications of these results for indentifying the model and its estimation are specified. In these simulations, conditions under which Swamy-Arora estimator is inferior to alternatives are expressed. It is shown that in small samples the estimator thus obtained can give highly wrong guidance. In one-way error component model this small sample size refers to the number of cross-sections.
