Non-expected utility and the robustness of the classical insurance paradigm : 074070 (E12) Machina M.J., University of California-San Diego, USA, The Geneva Papers on Risk and Insurance, Vol. 20, nr. 01, 1995, pp. 9–50

• Publisher: Elsevier Science
• Year: 1996
• Tongue: English
• Weight: 228 KB
• Volume: 17
• Category: Article
• ISSN: 0167-6687
• DOI: 10.1016/s0167-6687(96)90144-4

✦ Synopsis

Lute and Fishburn (1991)

derived a general rank-dependent utility model using an operation @ of joint receipt. Their argument rested on an empirically supported property (now) called segregation and on the assumption that utility is additive over CD. This note generalizes that conclusion to the case where utility need not be additive over @, but rather is of a more general form, which they derived but did not use in their article. Tversky and Kahneman (1992), conjecturing that the joint receipt of two sums of money is simply their sum, criticized that original model because CD = + together with additive utility implies the unacceptable conclusion that the utility of money is proportional to money. In the present generalized theory, if CB = f, utility is a negative exponential function of money rather than proportional. Similar results hold for losses. The case of mixed gains and losses is less well understood.

This study examines whether bettors' risk preferences or overconfidence in choosing winners better explains their well documented preference for lowprobability wagers. Although previous studies using racetrack data often suggest that risk-loving behavior explains long-shot preference, such data cannot distinguish between the alternative explanations. We use football betting data to make the comparison and find that overconfidence more closely fits the data. This result complements evidence of overconfidence from behavioral studies as well as stock-market models of overconfident noise traders.