The research investigates the existence, nature and magnitude of the preferences of races to voluntarily “self‐segregate” into particular areas of urban housing markets. Housing market theory is employed to develop a model showing how housing price variations within a group can provide unambiguous evidence of their preferences for neighborhood racial composition. The model is operationalized in a multiple regression specification wherein the variations in a given racial group's housing prices become a function of the dwelling's attributes and the attributes of the neighborhood (including quality, status, stability and density) as well as housing submarket location and racial composition. The size and statistical significance of the coefficient of the last attribute provides the evidence sought.The regressions are estimated using two micro‐household data bases from St. Louis (1967) and Wooster, Ohio (1975), and results compared. Results show that St. Louis black owners had an aversion to larger black proportions within black submarket neighborhoods, with .7% lower housing prices associated with a 1% higher percentage black. Racial effects for black owners in preponderantly white areas and for black renters in all areas were statistically insignificant. St. Louis whites of both tenures did not demonstrate aversion to neighborhoods with higher percentages of blacks as long as they remained 25% black or less. In areas 25–50% black, however, white prices were 1.5% lower for owners and 3.2% lower for renters per 1% higher proportion black. Such associations continued in majority‐black areas, although the magnitudes of the price effect became progressively smaller. Wooster whites showed an aversion to living in neighborhoods having even a few percent of blacks, with prices 11% lower for owners and 7% lower for renters in such areas compared to all‐white ones.