That is the title of a recent paper by Schneider and Sutter (2026). While many people are familiar with the concept of risk aversion, the dispersion of outcomes (i.e., risk) is not the only thing that matters. The skewness of distribution (prudence) and risk in the tails of the distribution (temperance) matters. For instance, prudent individuals may be more likely to play the lottery since downside risk is moderate, but there is a very high upside (risk are right skewed).
Key questions in the literature are how prudent/temperate are people and does prudent/temperate preference matter for real-world behaviors. The Schneider and Sutter paper aims to answer these questions. I summarize the paper below.
Methodology
The authors surveyed 658 children and adolescents, aged 10 to 21 years. Their methodology relied on estimating respondents certainty equivalents. Certainty equivalent is the sure amount of money that makes a subject indifferent between a lottery and a sure amount of money. The specific approach the authors use is a bisection approach where participants choose between a sure payoff and a lottery in an iterative manner. Specifically,
The choice in the first decision task was between 70 taler for sure and an equal chance of getting zero or 140 taler…If a subject chose the sure payoff, the amount of the sure payoff would
be decreased in the next iteration, whereas if they chose the lottery, the sure payoff would be increased. From three such iterations, we approximate indifference amounts for a specific lottery, i.e., the certainty equivalents, by taking the mid-point of the interval in which it must lie
The authors then elicit certainty equivalents between the other points in the 0 to 140 taler choice space based on what the certainty equivalent was for the original paper. The authors validate the approach by having the respondents choose between left- and right-skewed lotteries–specifically (30, 70, 100, 100) vs. (50, 50, 80, 120) where each of the 4 outcomes has a 25% chance. Note that two paired lotteries have the same expected value.
Results
The authors found that individuals were risk averse (Arrow-Pratt coefficient of risk aversion,
expressed in standard deviations, of r = 0.46 (median: 0.35)). Females were more risk-averse than males. The authors also found that 67% of respondents were prudent as well. “The mean (median) value of the simple prudence measure, expressed in standard deviations, is pCE = 0.80 (0.36) , with 0 indicating prudence neutrality.” Fifty seven percent of respondents were also temperate. The mean (median) Denuit-Eeckhoudt measure of temperance was = 0.30 (0.02).
Relevance to health care
Do these findings matter for health care? It turns out the answer is ‘yes’. The authors look at unhealthy behaviors (e.g., high BMI, lack of physical activity, smoking and drinking), as well as addictive smart phone use and relate these behaviors to measures of risk aversion, prudence and temperance.
Most importantly, prudence is strongly related to health-related behavior, but risk
aversion is not. This holds particularly true for addictive behavior like the obsessive
use of smartphones, but also extends to including smoking and drinking, or being
overweight
Moreover, risk preferences matter wen using generalized and risk-adjusted cost effectiveness analysis (GRACE). While GRACE relies on risk preferences over quality of life outcomes (rather than financial outcomes), these findings–and the bisection methodology–is highly relevant to that literature as well.
You can read the full paper here. Note that the online appendix has a helpful literature review examining risk aversion and field behavior, which is a helpful resource.
