The econometric tools to control composition effects : comparisons and implementation for an analysis of disparities between French departments in average earned incomes of self-employed persons

This document presents the implementation of different econometric tools to control composition effects, for an analysis of disparities between departments in earned incomes of self-employed persons. The main results of this analysis were published in Insee Première n° 1672 "Le revenu d’activité des non-salariés : plus élevé en moyenne dans les départements du nord que dans ceux du sud" ("Earned income in nonemployees : higher on average in France’s northern departments than in the southern").

Mapping earned incomes of self-employed persons by departments shows a strong spatial correlation : the average earned incomes are relatively high in the northern third of France and low in the southern third. These earned income disparities between departments can partly result from structural differences by activity sector and legal category of self-employed. Similarly, other individual characteristics of self-employed as well as the economic environment of the activity contribute to earned income disparities between departments. In order to highlight these potential composition effects, several modeling tools are possible. Some are commonly used in economic geography – descriptive, econometric and geographic structural-residual models –, other are commonly used in labour economics – Oaxaca-Blinder model. In this document, these four models are applied to the self-employed database and their pros and cons are presented. The implementation of an econometric structural-residual model and an Oaxaca-Blinder decomposition model give identical results. On the other hand, the explained parts obtained from these two methods differ slightly from those obtained from a geographic structural model. Finally, some ways to improve the explained part of these models are presented.