Discovered Preferences for Risky and Non-Risky Goods

We develop an axiomatic theory that integrates the discovered preference hypothesis into neoclassical microeconomic choice theory. A theory in which preferences must be discovered through experience can explain patterns observed in choice data, including preference reversals, evolution of or instability in risky choice, and errors that decline with repetition as seen in contingent valuation data. With reasonable assumptions, we show that preferences for common, high-ranked, and non-stochastic choice items are learned quickly and thus should appear stable. However, initially low-ranked choice items may remain persistently mis-ranked. Preferences for choice items with stochastic outcomes are difficult to learn, so choice under uncertainty is subject to error. At finite time, a choice item is more likely to be mis-ranked if it has stochastic outcomes, if it is initially low-ranked, or if it appears rarely in choice sets. The existence of a default option may or may not render correct ranking more difficult. Undiscovered preferences can lead to real welfare loss as agents make choices not congruent with their true preferences. This theory is amenable to tests using laboratory experiments. Preference discovery has implications for policy, and the process of discovery may contaminate choice data in a variety of contexts.