Abstract
<jats:p>This chapter presents a methodology to estimate municipal gross domestic product (GDP) in Mexico by integrating nighttime light (NTL) satellite imagery with econometric reasoning and machine learning. This chapter addresses a practical and analytical limitation: in Mexico, GDP is systematically available at the national and state levels, but not as a regular official series at the municipal level. This gap complicates local economic analysis, regional planning, and the evaluation of territorially focused public policy. To address this limitation, the study combines state-level GDP data, population-based municipal disaggregation, and multiple indicators extracted from VIIRS nighttime lights. Three model families are assessed: ordinary least squares (OLS), Random Forest, and a hybrid model in which Random Forest is used to learn the nonlinear structure that remains in the OLS residuals. The chapter argues that this combined strategy is particularly useful in settings characterized by territorial heterogeneity, uneven statistical infrastructure, and a large informal sector.</jats:p>