The hindsight bias: A meta-analysis

Abstract

The hindsight bias in probability assessments is one of the most frequently cited judgment biases. A meta-analysis of 122 studies revealed evidence that the bias occurs under some conditions and that its effect can be moderated by a subject’s familiarity with the task and by the type of outcome information presented. The data also suggest that the use of “almanac” questions can generate an unusually large hindsight effect. An observed asymmetry in the effect of the hindsight bias suggests that cognitive and not motivational factors may be the main cause of the bias. Finally, the overall magnitude of the effect of the hindsight bias was found to be small (r = .17). While these results suggest that the bias may not be as worrisome as commonly assumed in the literature, we discuss some situations when an effect this small may be of importance to practitioners. We also show that, depending upon the familiarity of the task and type of outcome information presented, anywhere from a minimum of 0% to a maximum of 7-27% of the population may make different decisions because of the hindsight bias. © 1991.