Risk preferences govern people’s choices under uncertainty, and understanding these preferences is important in many settings. Researchers have identified and estimated different risk preference models, including both expected utility (EU) and non-EU models. The authors provide a survey and discussion of this research, and they find that different models fit the data in different contexts.
How Is This Research Useful to Practitioners?
Risk preferences determine individuals’ choices under uncertainty. Modeling individuals’ risk preferences is essential to business, economics, finance, and policymaking and has popular uses in wealth management, behavioral finance, and consumer finance. For example, how people respond to risk has implications for asset pricing and for structuring health insurance markets in order to achieve a socially desired outcome.
By illustrating with a motivating example, the authors demonstrate the importance of risk preferences in quantitative analyses and how different risk preference models affect out-of-sample predictions and welfare conclusions.
Imagine a hypothetical insurance market in which 100% of the people are willing to pay an $800 premium to buy insurance that covers against a potential $10,000 loss. Assume that when the premium increases from $800 to $1,000, 10% decide not to buy insurance; they are less risk averse than those who stay insured. If the insurance premium keeps increasing, fewer and fewer people will buy the insurance. This simple example illustrates that risk preferences determine the price (the return) a person is willing to pay (accept) for avoiding (taking) the risk.
Different risk preference structures generate different behavior in response to risk. Most of the literature uses the traditional expected utility (EU) model, which assumes that an individual’s utility function fully determines his risk preferences. Researchers have acknowledged (in response to experimental evidence) other factors that influence behavior toward risk and have developed “non-EU” models, such as rank-dependent expected utility (RDEU) and cumulative prospect theory (CPT). RDEU models reflect the fact that people overweight the value of small-probability events and underweight the value of large-probability events. In other words, decision weights do not necessarily equal objective probability weights. CPT extends the probability-weighting features of RDEU models and requires a “reference outcome” as input. It then assesses the value of an uncertain event based on the probability of gains and losses relative to that reference outcome.
How researchers can identify and estimate such models is the theme of this comprehensive survey article. To identify the models, the authors use data from particular settings. In property insurance settings, previous researchers have found support for EU models, but a recent study suggests that non-EU probability distortion models (i.e., subjective probabilities do not equal objective probabilities) outperform EU models in explaining observed risk aversion. In horse races, it is observed that bettors are prone to bet on an actuarially unfair, low-winning-probability horse (i.e., higher risk and lower expected return). Such behavior can be explained either as “risk seeking” (assuming the EU model) or “risk misperception” (assuming a non-EU model). The estimated risk misperception model better explains the bettor’s behavior, suggesting that betting is driven by misperceptions rather than by risk-seeking behavior.
Importantly, the authors point out that risk preferences may not be consistent across contexts. In other words, individuals may not apply the same utility function across all of the various risky decisions they face. The human mind simplifies complex field contexts when conceptualizing possible outcomes, which may also affect the risk preference structure.
How Did the Authors Conduct This Research?
The authors conduct a thorough survey of the literature in the area of risk preferences and then synthesize their findings to generate suggestions. In particular, they concentrate on studies that try to estimate risk preferences using field data, either at the individual or aggregated level.
The authors effectively summarize the development of risk preference studies and provide guidelines on conducting research using field data. This study also helps to frame the direction of future risk preference research.
Abstractor’s Viewpoint
Finance is essentially a study of the risk–return trade-off. Not many studies reveal the risk preference structure in investment markets. Although pioneers in behavioral finance have proposed that investors have tendencies to use heuristics in estimating risk that lead to probabilistic distortion, we have little idea about the size of the error.
As the authors explain, to infer individuals’ risk preference parameters, the objective probabilities of risky events and their corresponding outcomes must be well defined and known. Such information is transparent and available in a casino setting but is much more complicated in many other markets. The field studies assume that the individuals face identical choice sets (e.g., choose one auto insurance policy among three different deductible options), but in practice, different individuals usually face different choice sets based on their level of wealth and other available information.
Sometimes researchers can collect data in a more controlled setting. For example, employees in Hong Kong SAR are required to invest 5% of their monthly salary in the Mandatory Provident Fund (MPF), where the choice set is limited, information is transparent, and the investment amount is at a fixed ratio to salary.
The last step in data collection is to measure investors’ probabilistic expectations, as advocated by the authors. Questions on subjective probabilities can be administered directly.
Undoubtedly, estimating risk preferences in a capital market context would be challenging and complex. But the findings could help investors more accurately price assets and better manage their risks.