An econometric model of ethnic maps and evaluation of the endogenous areal unit problem

Abstract

In this paper, I create an econometric model of ethnic maps and explore how the resulting areal units affect statistical inference. I first estimate several competing models of the Murdock map of African ethnic groups, used widely in applied econometric studies, and do not reject the hypothesis that the map is endogenously determined. I then find that the Murdock map generates more spurious spatial correlations than completely random maps, which in turn generates more spurious spatial correlations than equal-area geocells. I document a distinct problem arises when both conflict outcomes and map units are determined by the same variables.

Keywords: economic tesselation, endogenous unit area, spurious spatial correlation

JEL Classification: C51, C52, C59