Predicted property values for all buildings in Kampala

Working paper 1 Sep 2026 Cities, Sustainable Growth and Cities that Work

The following study uses KCCA property records, satellite imagery, and geospatial data to estimate property values across Kampala, where official assessments cover only a fraction of the city’s 2.1 million buildings. A LightGBM machine-learning model explains 51% of variation in assessed property rates and correctly ranks buildings in 72% of cases, outperforming a Random Forest benchmark. The model produces the first building-level property value map for the Greater Kampala Metropolitan Area, providing a tool to expand the tax base, improve equity, target infrastructure investment, identify low-value informal settlements, and monitor urban change. Key limitations include regression-to-the-mean and greater uncertainty beyond the KCCA training area.