Sources of bias in the prediction of future events

Abstract

The pattern of results of three studies indicates possible qualifications on the availability heuristic for probability estimation proposed by Tversky and Kahneman (Cognitive Psychology, 1973, 5, 207–232). Several lists of possible events were prepared which were either personal or external (national/ international) in nature and which had either positive or negative outcomes. Subjects were asked to estimate the probability of each event occurring during each of four decades in the future. The results indicated a highly significant interaction (p < .001) between type of outcome (positive or negative) and time. Negative events were seen initially as more likely than positive events, but as decreasingly likely further in the future; conversely, positive events were predicted as increasingly likely over time, being seen as more likely than negative events at the final time period. This finding was replicated with a different group of subjects rating a different set of items, also at the (p < .001) level. Socioeconomic background appeared to make no difference in the pattern of results. A third experiment asking for only a single estimate for each event failed to produce the “optimism” effect and confirmed the original prediction made from the availability heuristic. The implications of this “optimism” effect for policy planners should be considered.