Abstract
This article proposes a method for analyzing data from panels and pooled cross-sections that can separate the effects of exogenous variables on the endogenous variable into cross-unit and longitudinal components. Monte Carlo results suggest that estimation using estimated generalized least squares is close to theoretical expectations for unbiasedness and reasonably reproduces the expected distribution of confidence intervals for a variety of combinations of number of units and time points. To illustrate the substantive utility of the decomposition, an analysis is presented of the effects of unionization and economic conditions on race-gender group industrial earnings inequality that yields dramatically different longitudinal and cross-unit effects.
Get full access to this article
View all access options for this article.
