Probability judgments in multi-stage problems: Experimental evidence of systematic biases

Abstract

We report empirical evidence that in problems of random walk with positive drift, bounded rationality leads individuals to underestimate the probability of success in the long run. In particular, individuals who were given the stage-by-stage probability distribution failed to aggregate this information in a multi-stage case. Estimations of the long-run probability distribution did not differ much from the given stage-by-stage probability distribution, and were systematically lower than the accurate one. Applications to risk perception in financial markets are considered.