We test the “white backlash” hypothesis—that white racial conservatism grows as black population density increases—using methods and data parallel to those upon which the concept was originally developed. A 1996 Kentucky referendum vote to strike school segregation from the state constitution serves as a rare measure of “old-fashioned racism.” We estimate the white vote for segregation using ecological inference techniques, and show that the county-level pattern of white segregationist support for this referendum actually runs in the opposite direction from that predicted by the backlash hypothesis. Precinct-level Kentucky analysis and county-level analysis of a related South Carolina referendum validate our finding against the white backlash hypothesis. We conclude that the frequent failure of recent research to find white backlash, especially among urban voters, therefore cannot be explained away by the absence of measures of old-fashioned racism. Rather, the geographical pattern of American racial conservatism seems to have changed since the successes of the Civil Rights movement.