52nd Congress of the European Regional Science Association: "Regions in Motion - Breaking the Path", 21-25 August 2012, Bratislava, Slovakia
Although there are hundreds of works using PISA microdata to analyse the determinants of educational outcomes, only a few of them have considered the relevance of geography. In this paper, we focus on the analysis of differences in educational outcomes between students in rural and urban schools. Previous studies on this topic started in the United States during the eighties and they have not arrived to a sound conclusion yet. Some authors do not find significant differences between rural and urban students while others find better outcomes for urban ones. It is not clear if this gap is explained by family characteristics or it is related to lower spending in rural schools. The policy debate on how public spending in education should be distributed between rural and urban areas in developing countries is very intense, although academic studies are scarce. However, during the last decades some studies have focused on some Latinamerican countries (Argentina, Peru, Brazil, Chile and Colombia), but again, there is no consensus on the factors behind the rural-urban gap in educational outcomes. Taking into account this background, we use microdata from the 2006 and 2009 PISA waves for Colombia. The Colombian case is particularly interesting from this perspective due to the structural changes suffered by this country during the last years, both in terms of political stability and educational reforms. The descriptive analysis of the data shows that the educational outcomes of urban students are higher than the rural ones in the three subjects covered by the PISA study: Mathematics, Reading and Science in both samples. In order to identify the factors behind this differential, we apply decomposition techniques that have been extensively in Labour Economics but not so much in the context of Economics of Education. In particular, in a first step we use the Oaxaca-Blinder decomposition and, next, we exploit the time variation of the data using the methodology proposed by Juhn-Murphy-Pierce. The results show that most part of the differential is related to family characteristics more than to school characteristics. From a policy perspective, the obtained evidence support actions addressed to improve family conditions and not so much to positive discrimination of rural schools.