How do individuals perceive and position for Black Swans? In his March 2013 paper entitled “The Psychology of Tail Events: Progress and Challenges”, Nicholas Barberis employs a two-step framework to summarize recent research on the psychology of tail events. He first addresses belief about the probability of a tail event. He then covers actions/decisions based on this belief, with focus on the concept of probability weighting. Based on the available body of research, he finds that:
- A tentative reading of available evidence is that people tend to overestimate the likelihoods of recurrence for tail events with known precedents.
- The dominant view is that people then tend to overweight (as in a portfolio) the expected impacts of such tail events. Related evidence includes:
- As found in several studies, investors price skewness of stock return distributions (stocks with high lagged skewness have low future returns).
- The return distribution of the aggregate stock market is negatively skewed (due to occasional large crashes), with overweighting of these tail events by investors arguably accounting for a high equity premium.
- Individuals tend to overweight the probability of early death, thereby underbuying annuities upon retirement (relative to reward-risk analysis).
- However, some observers argue that people tend to underestimate the probability of tail events with no precedents in readily available/selected samples (events not readily imagined, or intentionally ignored by excluding “outliers”).
In summary, available research suggests that individuals tend to overestimate both the likelihood and impact of “imaginable” extreme events, but they may underestimate the likelihood of the “unimaginable.”
Cautions regarding findings include:
- The paper does not quantify findings.
- An implicit assumption of the author’s framework for summarizing research is that the distributions of experiences (such as investment returns) are tame enough to support “normal” inference. For sufficiently wild distributions, the observed mean and standard deviation are not usefully predictive of the future outcomes. It seems possible that people have evolved in response to somewhat wild distributions, such that we naturally require a “non-normal” margin of safety.
