This paper use spatial econometric models to test for racial preferences in a small urban housing market. Identifying racial preferences is difficult when unobserved neighborhood amenities vary systematically with racial composition. We adopt three strategies to redress this problem: (1) we focus on housing price differences across microneighborhoods in the small and relatively homogenous city of Kingston, New York; (2) we introduce GIS-based spatial amenity variables as controls in the hedonic regressions; and (3) we use spatial error and lag models to explicitly account for the spatial dependence of unobserved neighborhood amenities. Our simple OLS estimates agree with the consensus in the literature that black neighborhoods have lower housing prices. However, racial price discounts are no longer significant when we account for the spatial dependence of errors. Our results suggest that price discounts in black neighborhoods are caused not by racial preferences but by the demand for amenities that are typically not found in black neighborhoods.