The direct benefits of infrastructure in developing countries can be large, but if new infrastructure induces in-migration, congestion of other local publicly provided goods may offset the direct benefits. Using the example of rural household electrification in South Africa, we demonstrate the importance of accounting for migration when evaluating welfare gains of spatial programs. We also provide a practical approach to computing welfare gains that does not rely on land prices. We develop a location choice model that incorporates missing land markets and allows for congestion in local land. Using this model, we construct welfare bounds as a function of the income and population effects of the new electricity infrastructure. A novel prediction from the model is that migration elasticities and congestion effects are especially large when land markets are missing. We empirically estimate these welfare bounds for rural electrification in South Africa, and show that congestion externalities from program-induced migration reduced local welfare gains by about 40%.