We develop four experimental markets to examine how individuals respond to risk: self-protection and selfinsurance in both private and collective auctions. First, we find evidence that the mechanism used to reduce risk is important. Results indicate that the upper and lower bounds on value were elicited by the private selfprotection and the collective self-insurance markets, respectively. Second, the robustness of these results declined with low-probability lotteries~ We find further evidence that individuals overestimate the impact of low-probability events. Overestimation decreased, however, with repeated market exposure. Third, the four markets induced rapid value formation. Usually only one or two additional market trials were necessary before an individual's perception and valuation of reduced risk stabilized.
Two elements define risk: probability and severity. Ehrlich and Becker (1972) recognized that risk can be reduced by decreasing either element, privately or collectively. They define decreased probability as self-protection, and decreased severity as self-insurance. Recent extensions of self-protection and self-insurance models have illustrated their wide applicability and importance to the theory of individual choice under risk (see, for example, Hiebert, 1983;Centner and Wetzstein, 1987; Shogren and Crocker, 1990a).
Although it is now generally recognized that self-protection and self-insurance exist, minimal attention has been given to systematically evaluating their comparative impact on individual response to risk. Given Tversky and Kahneman's (1981) work on choice under alternative decision frames, one might suspect that how a risk is reduced may be as important as what risk is reduced. The purpose of this article is to examine how individuals respond to risk that is reduced either through private or collective self-protection or self-insurance. We construct an experimental design that incorporates self-protection and *The financial support of the U.S. Environmental Protection Agency and the John A. Walker College of Business, Appalachian State University is gratefully acknowledged. The comments of Tom Crocker, Don Coursey, Cliff Nowell, and Mark Thayer concerning experimental design have been helpful. Joe Kerkvliet, Fred Wallace, Charles Plott, an anonymous referee, and especially W. Kip Viscusi provided useful advice. Kris Etter, Todd Holt, and Kevin Long provided valuable research assistance. The author accepts sole responsibility for all remaining errors.