People Believe If 90% Prefer A over B, A Must Be Much Better than B. Are They Wrong?

Abstract

We show that consumers confuse consensus information in polls—such as 90% prefer product A over product B—with differences in liking—the extent to which poll respondents like A better than B. Consequently, they interpret a 90% consensus in favor of A as the average liking of A being considerably higher than the average liking of B. We demonstrate empirically and with simulations that—while this can be true—it is more probable that the average liking of A is only slightly higher than that of B. This regularity is robust to the sign and size of the correlation between ratings for A and B, and across most distributions for A and B’s liking. Consumers are not aware of this regularity and believe that 90% consensus implies A being much better than B. Communicators (marketers, managers, public policy makers, etc.) can capitalize on these erroneous inferences and strategically display preference information as consensus or as liking ratings, leading to dramatic shifts in choices. Consumers’ erroneous inferences can be corrected by educating them about the shape of the distribution of liking differences. We discuss theoretical and managerial implications for the understanding and usage of polls.