Working Paper: NBER ID: w25072
Authors: Jonathan Chapman; Erik Snowberg; Stephanie Wang; Colin Camerer
Abstract: We introduce DOSE - Dynamically Optimized Sequential Experimentation - and use it to estimate individual-level loss aversion in a representative sample of the U.S. population (N=2,000). DOSE elicitations are more accurate, more stable across time, and faster to administer than standard methods. We find that around 50% of the U.S. population is loss tolerant. This is counter to earlier findings, which mostly come from lab/student samples, that a strong majority of participants are loss averse. Loss attitudes are correlated with cognitive ability: loss aversion is more prevalent in people with high cognitive ability, and loss tolerance is more common in those with low cognitive ability. We also use DOSE to document facts about risk and time preferences, indicating a high potential for DOSE in future research.
Keywords: loss aversion; cognitive ability; risk preferences; DOSE
JEL Codes: C81; C9; D03; D81
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| DOSE method (C80) | loss tolerance in US population (G52) |
| DOSE method (C80) | loss aversion in US population (G41) |
| higher cognitive abilities (D91) | loss aversion (G41) |
| lower cognitive abilities (D91) | loss tolerance (K13) |
| traditional methods (C90) | measurement error and inconsistency (C83) |
| higher cognitive abilities (D91) | lower risk aversion (D81) |