Empirical studies of the economic effects of climate change (CC) largely rely on climate anomalies for causal identification purposes. Slow and permanent changes in climate-driven geographical conditions, i.e. CC as defined by the IPCC (2013), have been studied relatively less, especially in Africa which remains the most vulnerable continent to CC. We focus on Lake Chad, which used to be the 11th-largest lake in the world. This African lake the size of El Salvador, Israel, or Massachusetts slowly shrunk by 90% for exogenous reasons between 1963 and 1990. While water supply decreased, land supply increased, generating a priori ambiguous effects. These effects make the increasing global disappearance of lakes a critical trend to study. For Cameroon, Chad, Nigeria, and Niger – 25% of sub-Saharan Africa's population –, we construct a novel data set tracking population patterns at a fine spatial level from the 1940s to the 2010s. Difference-in-differences show much slower growth in the proximity of the lake, but only after the lake started shrinking. These effects persist two decades after the lake stopped shrinking, implying limited adaptation. Additionally, the negative water supply effects on fishing, farming, and herding outweighed the growth in land supply and other positive effects. A quantitative spatial model used to rationalize these results and estimate aggregate welfare losses taking into account adaptation shows overall losses of about 6%. The model also allows us to study the aggregate and spatial effects of policies related to migration, land use, trade, roads, and cities.