Calibration without reduction for non-expected utility

Evidence from the lab and the field shows that most people exhibit substantial risk aversion over stakes of hundreds of dollars. Expected utility cannot capture nonnegligible risk aversion over such small stakes without producing implausible risk aversion over large stakes, and under the reduction of compound lotteries axiom, neither can nonexpected utility preferences. Motivated by experimental evidence, this paper assumes that compound lotteries are evaluated recursively and shows that popular nonexpected utility models can be consistent with empirically plausible risk aversion over both small and large stakes. ∗ PhD Candidate, Department of Economics, University of British Columbia. Address: 997 1873 East Mall, Vancouver, BC, V6T 1Z1. E-mail: [email protected]. I would like to especially thank my advisor, Yoram Halevy, for substantial guidance, feedback, and encouragement. For conversations and comments that lead to substantive and expositional improvements, I would also like to thank Nicholas Barberis, Eddie Dekel, David Dillenberger, Eran Hanany, Chad Kendall, Luba Petersen, Matthew Rabin, Zvi Safra, and Lanny Zrill. The author was supported by a SSHRC Canada Graduate Scholarship while this paper was written.