Using subnationally representative data for 679 regions from 64 low- and middleincome countries, we document that geography, especially trade-related features, explains a substantial share of within-country variation in the level of human capital. We further show that a large part of the explanatory power can be attributed to the close interrelation between urbanization and concentration of human capital. These results bring together studies that identify human capital as an important determinant of regional development and the literature on economic geography that emphasizes geography’s role in shaping the spatial distribution of urbanization and economic activity.

Introduction

Geography’s influence on national and subnational economic development is well documented. The relative importance of individual channels through which geography affects development, however, is much less clear. In this study we investigate if and how geography influences the regional level of human capital, a potentially important proximate determinant of economic prosperity (e.g., Acemoglu and Dell (2010); Gennaioli et al. (2013)). For our empirical analysis, we draw on subnationally representative survey data from the Demographic and Health Surveys (DHS) Program and construct various proxies for human capital at the regional level, such as average years of schooling or prevalence of stunting. In total, our dataset encompasses 679 regions from 64 low-and middle income countries. To assess the explanatory power of geographic variables regarding within-country differences in the level of human capital, we employ a cross-sectional OLS regression approach in which we control for country-fixed effects. For parsimony, we include only nine geographic characteristics in our main regression setup. Following Henderson et al. (Forthcoming), this set of features encompasses five agriculture-related variables (caloric suitability, temperature, precipitation, elevation and absolute latitude) and four trade-associated characteristics (accessibility by inland waterways, share of a region’s area that lies with 25 kilometer of an ocean, distance to coastline and terrain ruggedness). Overall, the nine geographic covariates explain a meaningful 23% of the within-country sample variation in average years of schooling. These findings suggest that an important channel through which geography influences regional development is via its effect on human capital.