Working Paper: NBER ID: w23579
Authors: Benjamin L. Collier; Daniel Schwartz; Howard C. Kunreuther; Erwann O. Michel-Kerjan
Abstract: We examine risk preferences using the flood insurance decisions of over 100,000 households. In each contract, households make a small stakes decision, the deductible, and a large stakes one, the coverage limit. Over 94 percent of household choose one of the two lowest deductibles out of six options, and 77 percent fully insure, select a coverage limit of at least their home’s replacement cost. Households must be extremely risk averse to explain each of these choices with standard expected utility models. Households’ deductible choices imply a median relative risk aversion of 108. Households’ coverage limit choices require a median relative risk aversion of at least 112. Their substantial risk aversion over large stakes is due to households’ tendency to fully insure despite paying premiums well above their contracts’ expected value. Allowing for probability distortions improves our models and explains the small and large decisions of most households in our data. Assessing rank dependent utility models, we find that households follow two tenets of prospect theory: overestimation of small probabilities and diminishing sensitivity to losses.
Keywords: Risk Preferences; Insurance Decisions; Flood Insurance
JEL Codes: D12; D81; H42; Q54
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Probability distortions (C46) | Model's explanatory power for insurance choices (G52) |
| Cumulative prospect theory (D81) | Understanding of insurance decisions (G52) |
| Household deductible choices (G52) | Median relative risk aversion of 1.08 (D11) |
| Household coverage limit choices (G52) | Median relative risk aversion of at least 1.12 (D11) |