This paper provides a new dataset of regional income inequalities within countries based on satellite nighttime light data. First, we empirically study the relationship between luminosity data and regional incomes for those countries for which regional income data are available. Second, we use our estimation results for an out-of-sample prediction of regional incomes based on the luminosity data. These results enable us to investigate regional income differentials in developing countries that lack official income data. Third, we calculate commonly used measures of regional inequality within countries based on predicted incomes. An investigation of changes in the dispersion of regional incomes over time reveals that approximately 67–70% of all countries experience sigma-convergence. Forth, we study different major determinants of within-country changes in inequality, i.e., the determinants of the convergence process. We find evidence for an N-shaped relationship between development and regional inequality. Resources, mobility, trade openness, aid, federalism and human capital are also very important.