Abstract
Individuals often must learn about a state of the world when both the state and the credibility of information sources (experts) are uncertain. We argue that learning in these "rank-deficient" environments may be subject to a bias that leads agents to over-infer expert quality. Agents who encounter information or experts in different order disagree about substance because they endogenously disagree about the credibility of each others' experts, as first impressions about experts have long-lived influences on beliefs about the state. This arises even though agents share common priors, information, and biases, providing a theory for the origins of disagreement. Our theory helps explain why disagreement about substance and expert credibility often go hand-in-hand and is hard to resolve in a wide-range of issues where agents share common information, including economics, climate change, and medicine.