The project studies the nature of households' risk preferences. It utilizes a new data set that documents the choices made by a sample of households with respect to auto, home, and umbrella insurance. The data also include the menus from which the households' make their choices, the households' claim histories, and a rich set of demographic information for each household.
The project has two parts. The first part investigates the nature of risk preferences under the assumption that households' subjective beliefs about risk are correct. It initially examines the separate influences of "standard" risk aversion (concave utility over final wealth states), loss aversion, and nonlinear probability weighting on households' deductible choices in auto and home insurance. The first part also will investigate other determinants of households' risky choices, including ambiguity aversion, as well as other policy choices, such as limit choices under liability coverages, which involve large stakes.
The second part of the project will examine households' subjective beliefs about risk. A household's risky choices are jointly determined by their preferences for risk and their subjective beliefs about risk. The project will seek to estimate the joint distribution of risk preferences and subjective beliefs. Although point identification generally will not be possible, the richness of the data will allow for partial identification.
Broader Impact: The project has the potential to advance the understanding of household decision making under uncertainty and to help the scientific community further the development of the theory of choice under uncertainty. The acquisition of the data set presents the opportunity to gain new and important insights into the nature of households' risk preferences and the structure of their beliefs about risk.
The goal of this project was to study the nature of households' risk preferences. It utilized a new data set that documents the choices made by a sample of households with respect to auto, home, and umbrella insurance. Among other things, the data record deductible choices under three types of property damage coverage, limit choices under two types of primary liability coverage, and limit choices under a secondary "umbrella" liability coverage. The data also include the pricing menus from which the households make their choices, the households’ claim histories, and a battery of demographic characteristics for each household. The outcomes of this project are embodied by a series of research articles, some of which have already been published in scholarly journals, while others are undergoing the peer-review process. These articles may help economists and social scientists more broadly to better understand how individuals make their choices in insurance markets (e.g. among deductibles), and can potentially lead to better regulation of these markets. Among the research articles, in this report we highlight two. In 'The Nature of Risk Preferences: Evidence from Insurance Choices', published in the October 2013 issue of the American Economic Review, we have used our data on insurance deductible choices to estimate a structural model of risky choice that permits 'standard' risk aversion, loss aversion, disappointment aversion and probability weighting. We show that loss aversion, disappointment aversion and probability weighting --- though not separately identified without strong parametric assumptions --- each imply a distortion of probabilities, and we demonstrate that such probability distortions are identified. We find that probability distortions --- in the form of substantial overweighting of claim probabilities --- play an important role in explaining the aversion to risk manifested in deductible choices. Once we allow for probability distortions, standard risk aversion is relatively small. This finding is robust to allowing for observed and unobserved heterogeneity in preferences. We demonstrate that neither Koszegi-Rabin loss aversion alone nor Gul disappointment aversion alone can explain our estimated probability distortions, signifying a key role for probability weighting. In 'Distinguishing Probability Weighting from Risk Misperceptions in Field Data', published in the May 2013 issue of the American Economic Review, we develop a strategy for distinguishing rank-dependent probability weighting (RDPW) from risk misperceptions (RM) in field data. Our strategy relies on identifying a field environment with two key properties (which are not satisfied in most existing studies): (i) the objects of choice are money lotteries with more than two outcomes, and (ii) the ranking of outcomes differs across lotteries. In such environments, the ranking of outcomes is irrelevant to agents’ decision weights under RM, which simply correspond to their misperceived probabilities, but it is crucial to agents’ decision weights under RDPW, which are rank dependent. Thus, the models can make distinct predictions and thereby can be distinguished empirically (although exactly how their predictions differ depends on the details of the specific environment).
