Ideological Bias and Trust in Social Networks PRELIMINARY: PLEASE DO NOT CITE OR CIRCULATE

Two key features robustly describe ideological differences in society: (i) individuals persistently disagree about objective facts; (ii) individuals also disagree about which sources can be trusted to provide reliable information about these facts. We develop a model in which these patterns arise endogenously as the result of small deviations from Bayesian information processing. Individuals receive information from direct observation, subject to an ideological bias of which they are unaware. They also receive information through social networks and media. Sources differ in their reliability and individuals must learn which sources they can trust. We show that the entry of partisan information sources could generate large ideological disagreements, even when individuals have arbitrarily small biases and observe the same set of sources. Individuals may also come to trust the most biased like-minded sources, leading to ideological segregation in social networks and extreme bias by strategic media outlets.