Skip to content

GeoDevelopment Observatory of Bolivia / Tools

Where do lights predict poverty? Multiscale GWR

Is night-time light linked to poverty in the same way everywhere? A national regression and a multiscale geographically weighted regression (MGWR) of municipal poverty on lights per capita, for 2012 and 2024.

This Colab notebook regresses the multidimensional poverty index of the SDG Atlas of Bolivia 2026 on log night-time lights per capita for the 333 municipalities with lights data, first in 2012 and then with poverty in 2024 on lights in 2020. One national slope explains 29% of the variation in 2012. A multiscale geographically weighted regression (MGWR, with geo4bolivia’s geo.mgwr), which lets the intercept and the slope vary over space, each at its own scale, with bandwidths of 44 neighbours for the intercept and 64 for the slope, raises R² to 0.65. Every local slope is negative, but the link is almost five times stronger in Cochabamba than in Beni. By 2024 the link is a little weaker (national R² 0.25) and the slope’s bandwidth widens to 100 neighbours.

Back to the GeoDevelopment Observatory of Bolivia