The value of noise for informational cascades

Informational cascades are said to occur when rational agents ignore their own private information and blindly follow the actions of other agents. Models for such cascades have been well studied for Bayesian agents, who observe perfectly the actions of other agents. In this paper, we investigate the impact of errors in these observations; the errors are modelled via a binary symmetric channel (BSC). Using a Markov chain model, we analyze the net payoff of each agent as a function of his signal quality and the crossover error probability in the channel. Our main result is that a lower error level does not always lead to a higher payoff when the number of agents is large.