Panel Data Econometrics

At one time exotic, the use of economic data with both time series and cross-sectional variation has now become common place in modern econometric practice. The term panel data is used for a wide variety of situations in econometrics. It refers to any data set with repeated observations over time for the same individuals. “Individuals” can be workers, households, firms, industries, regions, or countries, to name a few.

If we pool together the national accounts of several countries we obtain a country panel. Data of this kind have been prominent, for example, in recent research on models of growth and convergence. Moreover, much interest has been directed to cross-country or cross-state panels because these data can sometimes provide exogenous variation in institutions or policies that facilitate the identification of parameters of economic interest.

Not surprisingly, some of the econometric issues arising in this context are closely related to time series econometrics. In aggregate panels, the cross-sectional and time series dimensions are often of a similar magnitude. A central modelling issue is how best to accommodate heterogeneity across units. A central statistical issue is the impact of cross-sectional variation for the choice and sampling properties of estimators.

Another class of data sets are household or firm level panels, which are based on surveys, census, administrative records, or company balance accounts. These are usually referred to as “micropanels”. Typically, these panels consist of large cross-sections of individuals observed for short time periods. Examples are the Michigan, Essex, and European Community household income panels, ¹ or the rotating panels on household expenditures conducted in the US and some European countries.

-1-