Abstract
The Black-White difference in infant mortality rates for 1982 through 1986 in 38 large US standard metropolitan statistical areas (SMSAs) varied by a factor of almost seven. In multiple regression analyses the most important predictor of the Black-White difference in the 38 SMSAs was an index of Black-White residential dissimilarity (or "segregation index"), independent of Black-White differences in median family income and poverty prevalence. Certain SMSAs in California had relatively low segregation indexes and small Black-White differences in infant mortality, despite considerable Black-White differences in poverty prevalence. The explanations for the apparent effect of residential segregation should be explored.