Contrary to much of the literature on collective opinion, I find that the low levels and uneven social distribution of political knowledge in the mass public often cause opinion surveys to misrepresent the mix of voices in a society. To assess the bias introduced by information effects, I compare “fully informed” collective preferences simulated from actual survey data to collective preferences revealed in the original data. Analysis of policy questions from the 1988 and 1992 American National Election Studies shows that group differences in knowledge, along with the public's modest average level of political knowledge, can cause significant distortions in measures of collective opinion. The mass public may appear more progressive on some issues and more conservative on others than would be the case if all citizens were equally well informed. To the extent that opinion polls influence democratic politics, this suggests that information effects can impair the responsiveness of governments to their citizens.