What is the Causal Effect of Knowledge on Preferences

We use a novel field experiment which jointly tests two implicit assumptions of updating models in a joint framework: that new information leads to new knowledge and that new knowledge can affect economic decisions. In the experiment, we elicit subjects’ prior knowledge state about a good’s attributes, exogenously vary how much new information about good attributes we provide to subjects, elicit subjects’ valuation for the good, and elicit posterior knowledge states about the same good attributes. Testing for changes in knowledge jointly with changes in preferences allows us to horserace updating models more completely than previous studies since we observe ex ante and ex post knowledge states. Our results are consistent with a model of incomplete learning, fatigue and either confirmation bias or costly search coupled with unbiased priors.