The redistributive politics literature examines how income inequality shapes preferences for redistribution. In this article, I show that inequality has reduced explanatory power in predicting preferences as population density increases. I argue that in cities, socioeconomic segregation - instead of distance - is a stronger predictor of redistributive preferences. To causally identify segregation, I propose a novel instrument that interacts rural-to-urban migration of the poor with the destination city's urban form. I find that spatial integration (reduced segregation) causally increases the incidence of the negative spatial externalities of inequality (e.g., crime, contamination), breeding a level of cross-class aversion. While distance models predict that social proximity generates empathy, actual physical proximity to the poor generates aversion. Spatially integrated cities, thus, have reduced support for redistribution. I argue that beyond the risk of income loss featured in insurance models of redistribution, there are more localized forms of risk from spatial spillovers.