Information Asymmetry in Decision from Description Versus Decision from Experience

In this paper we investigate the claim that decisions from \textit{experience} (in which the features of lotteries are learned through a sampling process) differ from decisions from \textit{description} (in which features of lotteries are explicitly described). We find that the experience-description gap is not as robust as has been previously assumed. We argue that when this gap appears it is driven to a large extent by asymmetries in information concerning which events are possible and which are certain. First, we find that, when experience-based decision makers sample events without error and then are told what outcomes are associated with each possible event, they are risk seeking for low-probability gains and risk averse for high-probability gains, as in description-based decision making. Second, we find that the experience-description gap for low-probability outcomes appears when rare outcomes are never experienced but disappears when: 1) all distinct outcomes are experienced at least once or 2) never-experienced outcomes are described as possibilities. Third, we find that the experience-description gap for high-probability outcomes is pronounced when decision makers previously experience lotteries that both offered the possibility of a zero outcome (which presumably makes them doubt that an always-experienced outcome is certain), but disappears when they have not previously experienced such lotteries.