Decision-making: Why do we underestimate rare events?

Mohammed Abdellaoui, Professor and Director of Research at CNRS - October 15th, 2014
The black swan theory was developed by the scholar Nassim Nicholas Taleb to describe unforeseeable, low probability events with significant consequences.

Uncertainty accompanies most individual and collective decisions. Given the unknown probabilities of available options, people make choices based on experience. But this way of functioning is biased: it leads to underestimating or ignoring rare events. This could explain, for example, why in 2005 only half of the colossal damage caused by the North Atlantic hurricanes in the U.S. was covered by insurance. Mohammed Abdellaoui and his co-authors set out to understand this surprising phenomenon.

Mohamed Abdelaoui ©HEC Paris

Holder of a Ph.D. in mathematical economics and econometrics, Mohammed Abdellaoui is Professor of decision science at HEC Paris and Director of Research at CNRS (GREGHEC). He is (...)

See CV

SUBJECTIVE PERCEPTION OF “DESCRIBED” UNCERTAINTY

In the study, when decision makers chose based on descriptions of probability distributions, they tended to feel an increase in the probability of a somewhat likely loss (50% chance of losing €1000) less strongly than the same increase in the probability of a very unlikely loss (5% risk of losing €1000 euros) or an almost certain loss (95% risk of losing €1000). This behavior is also observed for gains. In the language of prospect theory, this means that decision makers facing uncertainty overestimate small probabilities and underestimate moderate and high probabilities. The study of Abdellaoui and his co-authors confirms this result.


SUBJECTIVE PERCEPTION OF “DESCRIBED” UNCERTAINTY

In the study, when decision makers chose based on descriptions of probability distributions, they tended to feel an increase in the probability of a somewhat likely loss (50% chance of losing €1000) less strongly than the same increase in the probability of a very unlikely loss (5% risk of losing €1000 euros) or an almost certain loss (95% risk of losing €1000). This behavior is also observed for gains. In the language of prospect theory, this means that decision makers facing uncertainty overestimate small probabilities and underestimate moderate and high probabilities. The study of Abdellaoui and his co-authors confirms this result.

guillemet
How the information is communicated influences people’s perceptions.



SMALL SAMPLES AND SHORT MEMORY

On the one hand, in an uncertain context, policy makers tend to trust conclusions based on very small samples: they need few examples to form an opinion. In addition, decision makers tend to draw lessons from more recent situations. For example, after a natural disaster, the inhabitants of the devastated region are often inclined to get insurance. But a significant portion of those who sign up in the immediate aftermath of the disaster will terminate their contract a few years later.


PRESENTATION OF THE INFORMATION

Finally, how the information is communicated influences people’s perceptions. A dull description using figures presented in terms of probabilities does not have the same impact on behavior as information gained through experience. Road safety campaigns focused on communicating accident rates have little success, whereas signs indicating that fatalities have occurred at that particular location have greater impact. Besides the fact that statistics have little capacity to provoke uncertainty, they are sometimes misunderstood and misused, even by the most sophisticated decision makers. In an uncertain context, these factors contribute to incoherent decisions that may have costly negative, economic consequences.


Based on an interview with Mohammed Abdellaoui and the article “Experienced vs. Described Uncertainty: Do We Need Two Prospect Theory Specifications?” by Mohammed Abdellaoui, Olivier L’Haridon, and Corina Paraschiv published in Management Science in June 2011.

Practical Applications
Practical Applications

In the two contexts of uncertainty studied by Abdellaoui and his co-authors, individual behavior diverges from the predictions of the standard model of rationality, or the expected utility model. When individuals make decisions based on experience (rather than known options), their behavior also differs from predictions of prospect theory, causing them to underestimate or ignore rare events. Awareness of these phenomena and understanding their causes can help improve and modernize practices in risk management. In particular, it would be possible to take into account the fact that even well-trained decision makers are not immune to statistical or behavioral bias.

Methodology
Methodology

Abdellaoui and his co-authors compared the choices observed in laboratory experiments with those predicted by the prospect theory of Tversky and Kahneman (1992) in two contexts. In the first, individuals chose between alternatives described in the form of probability distributions (“described” risk). In the second, the decision maker discovered each alternative through a sampling process of unknown probability distribution (“experienced” risk).