This article presents the results from a study aimed to estimate population on basic geo-statistical areas (census tracts) along three districts of Mexico City through altimetry LIDAR data, high-spatial resolution multispectral imagery, and cadastral data. In a first stage, empirical relationships between population and housing units were developed based on data from the 2000 and 2010 censuses. Indicators derived from remote sensing and cadastral data were used to estimate housing units. Among the population estimation models tested, the one employing relative population increments yielded estimates with 5% of median relative error when the number of housing units was extracted from census. When estimating housing units with remote sensing, the errors increased but with a median value under 15% for the best composite model. The highest population estimation error was observed for Cuauhtémoc district, where not only population seemed to live in more than one land use type, but also the distribution of land uses was rather clustered, which demanded a model with much more information that could not be extracted from remote sensing.