08/06/2026
Today, the Maurice R. Greenberg School of Risk Science (GSRS) welcomed GSRS PhD student Ariane Bonilla to the GSRS Seminar Series to present research co-authored along Professor Glenn Harrison. The presentation explored how economists and actuaries measure risk differently, examining how individuals’ risk preferences shape decision-making and the evaluation of uncertainty. The research also highlighted new approaches to modeling complex risk behaviors, offering insights with implications for actuarial science, economics, and risk management.
Paper title: Risk Measures Derived from Risk Preferences
Abstract: Actuarial science defines risk measures by writing down axioms on a functional
that assigns a scalar to a risk, and then characterizing the functionals that satisfy those axioms. Economics proceeds in the opposite direction: it writes down a model of risk preferences, estimates that model from observed choices, and then asks what scalar summary of a risk is consistent with those preferences and the decisions they rationalize. The natural candidate is the certainty equivalent. We develop the economist’s approach to risk measures, extending the framework of Tsanakas and Desli [2003] in three directions. First, we evaluate risk measures numerically using estimated Rank-Dependent Utility models with flexible functional forms: Expo-Power utility rather than only CARA or CRRA utility, and Power, Inverse-S, or Prelec probability weighting rather than only the Power function. Second, we extend the analysis to “exotic” risk preferences: higher-order risk attitudes such as prudence and temperance, loss aversion and reference dependence, and intertemporal risk aversion. Third, we ask why this approach differs from actuarial practice, and locate the answer in the choice of primitives for axioms, in the insistence that risk measures be consistent with decision rules, and in the recognition that different actors evaluating the same risk need not share risk preferences, or even the same family of risk preference models. Throughout we use discrete probability mass functions and numerical methods, so that no restrictions need to be placed on the risks being evaluated.
