We use the German bombing of London during the Second World War as an exogenous source of variation to provide evidence on neighborhood effects. We construct a newly-digitized dataset at the level of individual buildings on wartime destruction, property values, and socioeconomic composition in London before and after the Second World War. We develop a quantitative spatial model, in which heterogeneous groups of individuals endogenously sort across locations in response to differences in natural advantages, wartime destruction and neighborhood effects. We find substantial and highly localized neighborhood effects, which magnify the direct impact of wartime destruction, and make a substantial contribution to observed patterns of spatial sorting across locations.