Pooling cross-sections with unequal time-series lengths

Abstract

This paper deals with the estimation of economic relationships using N cross-sectional units observed over periods of unequal lengths. A one-way error component model is assumed and the inverse of the variance-covariance matrix is derived. It is shown that GLS can be given a simple tractable WLS interpretation similar to that obtained for the complete panel data case. The basic difference, however, is that the incomplete panel data case has weights that are crucially dependent on the length of time-series available for each cross-section.