We examine the causal link between proximity to fast food and the incidence of childhood obesity among low-income households in New York City. Using individual-level longitudinal data on students living in public housing linked to restaurant location data, we exploit the naturally occurring within-development variation in distance to fast food restaurants to estimate the impact of proximity on obesity. Since the assignment of households to specific buildings is based upon availability at the time of assignment to public housing, the distance between student residence and retail outlets-including fast food restaurants, wait-service restaurants, supermarkets, and corner stores-is plausibly random. Our credibly causal estimates suggest that childhood obesity increases with proximity to fast food, with larger effects for younger children who attend neighborhood schools.