Abstract
The purpose of this paper is to point out the similarities and differences between the traditional econometric approach used for analysing cross-sectional time series data and the microbased superpopulation approach. The superpopulation approach is applicable when a probability sample of units is actually drawn from a population of units and followed over time. When this situation is at hand the superpopulation approach offers a flexible way of dealing with the choice of estimation technique for analysing macrorelations. In the superpopulation approach the dummy variable technique is used for estimation conditional on the sample. Population parameters which are not identical for each unit in the population are estimated unconditionally using survey sampling principles. Questions regarding the efficiency of the estimators are discussed.