Using Other People's Estimates: An Experimental Study

Expert opinions are often biased. To test how such bias a ffects the propensity to use opinions, we set up an experiment where subjects estimate the probability of an event that depends on (i) the subject's type, which is observable, and (ii) the unobserved state of the world. Before making their estimate, one group of subjects, the clients, observe the opinion (estimate) of another subject, the expert. The expert has private information about the state, but he may be of a diff erent type than the clients, and therefore biased. Bias is observable and easily corrected. In spite of this, we nd that clients' propensity to use expert opinions is decreasing in the size of the expert's bias. This aversion to use the opinions of biased experts is not explained by computational concerns, ex-post expert informativeness or reluctance to move away from the prior\

By  Jesper Rüdiger

https://dl.dropboxusercontent.com/u/5148008/ExpBIT.pdf