Ecological inference with entropy econometrics : using the Mexican census as a benchmark

Most regional empirical analyses are limited by the lack of data. Researchers have to use information that is structured in administrative or political regions which are not always economically meaningful. The non-availability of geographically disaggregated information prevents to obtain empirical evidence in order to answer some relevant questions in the field of urban and regional economics. The objective of this paper is to suggest an estimation procedure, based on entropy econometrics, which allows for inferring disaggregated information on local income from more aggregated data. In addition to a description of the main characteristics of the proposed technique, the paper illustrates how the procedure works taking as an empirical application the estimation of income for different classes of Mexican municipalities. It would be desirable to apply the suggested technique to a study case where some observable data are available and confront the estimates with the actual observations. For this purpose, we have taken the information contained in the Mexican census as a benchmark for our estimation technique. Assuming that the only available data are the income aggregates per type of municipality and State, we make an exercise of ecological inference and disaggregate these margins to recover individual (local) data.