Tests of theories of the electoral origins of divided government hinge on the proper measurement of voter preferences for divided government. Deriving preferences for divided government from voters’ ideological positions or responses to the standard American National Election Studies question inflates estimates of the proportion of people who prefer divided government. We present two alternative survey measures of preferences for divided government and evaluate the measures across multiple surveys. We find that the percentage of voters who prefer divided government is smaller than previous studies suggest. Voters who prefer divided government according to the new measures are significantly more likely than other voters to vote in ways that create divided government in both presidential year and midterm congressional elections.