ABSTRACT
Estimation of relationships which combine time series and cross-sectional data is a problem frequently encountered in econometrics. The problem is to specify a model which allows for differences in behaviour over such cross-sectional units as firms, households, geographical areas as well as differences in behaviour over time for a given cross-sectional unit. The chapter illustrates the specification and estimation of the temporal cross-section model under simplifying assumptions and discusses the results of a study by Feige and Swamy on the demand for liquid assets in the United States. The majority of temporal cross-section studies of the liquid assets have assumed that behaviour both across geographic units and time periods is identical. The random coefficients approach to aggregation assumes that micro coefficients vary only across micro units and remain stable over time. The chapter points out that if the elements of Xi are assumed to be stochastic and distributed independent of i and ui, there will still be no aggregation bias.
