Experiments on the Social Value of Public Information

We present the results of laboratory experiments studying the "CNBC Effect" - situations where the social value of public information is negative. Payoffs depend on two factors: being right (i.e. matching an underlying, but unknown state variable) and coordinating with others, which has no social value. In our baseline treatment, individuals privately receive an informative signal. When we add a second, lower quality private signal, decisions improve. When the lower quality signal is public, however, (a) subjects strategically place inefficiently high weight on the public signal and (b) welfare (aggregate payoff) falls by 12% compared to the baseline. Welfare losses are due both to inefficient information use and greater random variation in choices.