Abstract: How do partisan identities affect the facts people believe? Can opposing partisans learn effectively from one another? Or do identities interfere with learning by increasing in-group biases (overweighting of in-group and underweighting of out-group messages)? While the existing literature tends to distinguish theoretically between unbiased Bayesian updating and biased motivated reasoning, these competing theories can be observationally equivalent when source accuracy cannot be measured. We therefore investigate the influence of partisan identities and political knowledge on source accuracy and belief updating. In terms of source accuracy, we hypothesize there will be a gap between perceived accuracy of in-group partisan sources and out-group partisan sources, that this gap will be increasing in identity strength and the magnitude of affect toward each group, and decreasing in group-specific knowledge. We rely on an incentivized belief elicitation procedure (the Binarized Scoring Rule) to measure source accuracy directly in terms of second order beliefs (beliefs about the accuracy of others’ knowledge). We find strong support for our identity-based and affective hypotheses, but mixed support for our knowledge hypotheses. We also propose and sketch an experimental design to measure source accuracy indirectly from a belief updating task in which we manipulate the partisanship of information sources and infer source accuracy from the change between posterior and prior beliefs.
For more information, please contact the co-organizers:
Cathy Hafer (catherine.hafer@nyu.edu) and Congyi Zhou (zhoucongyi@nyu.edu).