Do you really believe that? The effect of economic incentives on the acceptance of real-world data in a polarized context

Attitudes and expectations towards others are major drivers of political polarization, while little is known about actual differences in beliefs and behaviors between partisan groups. We designed an experiment where self-reported attitudes were contrasted with economically point estimates of official data where participants received an economic benefit for correct answers. Our design offers three key contributions: 1) when measuring attitudes, a small partisan sub-group with extreme attitudes is the main reason for the observed partisan gap, while this group disappears when measuring incentivized data estimates; 2) economically-incentivized and unincentivized measures within individuals hardly correlate; 3) we provide a novel measure of perceived polarization, where individuals guess data estimates of those with opposing party preferences and receive an economic compensation for correct guesses. This novel perceived polarization measure correlates with attitudes but not with data estimates, supporting models linking polarization more to expectations towards others than to actual behavioral differences. This casts further doubt on standard surveys measuring attitudes and points towards strategies to lower perceived polarization within contested issues.