Classification and aggregation: An application to industrial classification in cps data

Abstract

Abstract In this paper we offer a method for deciding how to aggregate a set of elementary industries. The method is then applied to the problem of estimating a wage equation that allows for industry‐specific effects. Our method explicitly formalizes the trade‐off between goodness‐of‐fit and parsimony implicit in an aggregation problem. By varying the parameter of the assumed loss function, one obtains a whole sequence of aggregation levels. Further, the resulting sequence is consistent; that is, groupings formed at one level of aggregation will never be undone when one aggregates further.