An experimental test of some economic theories of optimism

Economic theories of optimism provide different rationales for the phenomenon of motivated reasoning, and a recent empirical literature has tested some of them, with mixed results. We contribute to this literature with a novel experimental test of two mechanisms, according to which optimism is respectively predicted when (1) the potential material losses due to the bias are relatively small or (2) the cognitive costs of the bias are small enough. In our design, these two accounts predict inflated expectations regarding some future payoff. Contrary to that, the average subject tends to (slightly) underestimate that financial prospect. Although a minority of the subjects overestimate systematically, the size of their errors is rather reduced, and they hardly differ in their personal characteristics from the rest of the subjects. In fact, optimism in our experiment is correlated with the sample observed, in that it is more likely when a subject observes relatively few good signals. This is again at odds with (1) and (2). These mechanisms, we conclude, do not appear to fully capture under which circumstances people fail into a positivity bias. Yet (1) seems to be empirically less relevant, in that we observe a similarly limited level of bias irrespectively of its monetary cost.