Several commenters on my last post make the perfectly valid point that strong public support for referendum questions banning racial "preferences" in the generally liberal states of California, Washington, and Michigan is an indicator of public opposition to affirmative action programs. True enough. However, all three of these referenda were worded as banning "preferences." In a 1997 Houston referendum, opponents succeeded in getting the city government to reword an otherwise similar ballot question as banning "affirmative action" rather than "preferences" (see here for an account critical of the City's decision). Sure enough, the initiative was defeated by a 55-45 majority, even in relatively conservative Houston. In referenda as in polls, whether the public supports affirmative action policies depends on how the question is worded.