The treatment of neglected heterogeneity in cross-sectional data sets.

Abstract

This study is to show how one can apply classification analysis to pure cross-sectional data in economics to build up models that incorporate behavioral differences and that, therefore, allow for more accurate and efficient policy applications. The study looks into both the problems that typically arise when behavioral heterogeneity is neglected in the empirical estimation process and a possible solution to the problem of identifying viable groups of economic actors with homogeneous behavioral response patterns.

The study is not aimed at providing a general method that is applicable for every possible cross-sectional data set and/or every possible heterogeneity pattern. Rather, it demonstrates, for a particular economic example of interest in regional economics, how one can take advantage of a variety of multivariate procedures in the general area of classification analysis to help reveal the nature of heterogeneity and to explore possible avenues to detect homogeneity.

In this study, the testing methodology (Zietz, 2000) for the problem of neglected heterogeneity is successfully implemented in practice. The test results confirm the existence of the problem of neglected heterogeneity for the original unclassified cross-section data used in this study.

This study is able to develop an objective and effective classification methodology for a given specific data set to discover homogeneous subgroups with similar individual characteristics that are related to their economic behavior.

As demonstrated by this study, one can establish a close and economically meaningful relationship between group-specific economic behavior and the individual characteristics of the economic agents in each group. Such relationships should be of significant economic value. For example, economic policies that try to target specific groups need exactly this type of group-specific information. It is also valuable for such issues as deriving forecasts for the aggregate of all observations. Specifically, the results provide not only useful weights that can be used to aggregate the subgroups but also the different behavioral patterns that need to be aggregated.

The procedures developed in this study apply to many research areas outside of economics. Specifically, they apply whenever one is facing the problem of neglected heterogeneity. The application of the methodology in the area of economic education is illustrated as an example in the study.

The study can be used by the readers who are interested in homogeneity-heterogeneity topics as a learning tool because all steps are carefully laid out and discussed.

Readers interested in potential applications of the techniques in the field of educational research are provided with a number of suggestions on how the methodology can be of potential use.