Canonical Riskless Choice Over Bundles: Aint No Reference Point Here

Prospect Theory (Kahneman and Tversky 1979), one of the most prominent models for valuation of goods and money, presumes that people have convex utility over gains and concave utility over losses; a discontinuity at something like the current wealth level or reference point. This reflects a behavioral pattern confirmed in hundreds of experimental studies where in lottery tasks people show decreasing marginal utility over gains (risk aversion) and increasing marginal utility (risk seeking) over losses relative to this ?reference?. Although it is widely assumed that a reference point is also required to describe riskless choices made over bundles of goods, there is less empirical evidence for this claim. In this paper, using incentive-compatible experimental methods, we challenge the generality of this assumption. We find that in riskless choice over bundles of goods in a canonical budget set experiment, gain-loss asymmetries are not observed even while in interleaved lottery tasks the reference point is observed, in the same subjects. Our results suggest a discontinuity between the value functions inferred from choices over standard lotteries and the utility functions inferred from indifference curves in riskless choice.