Dissociating loss aversion from probability weighting in risky choice: Evidence from choice reversals and willingness-to-pay measures

Abstract

Prospect theory attributes preference reversals to combined effects of loss aversion and probability weighting. We conducted three studies (N = 58, 64, 71) employing choice reversals between positively- and negatively-framed gambles, supplemented by willingness-to-pay measures. Using mixed-effects regression, we demonstrate that loss aversion and probability distortion contribute independently to choice reversals, with loss aversion accounting for ~60% of the effect. These findings refine current models of decision-making under risk.

Introduction

Since Kahneman and Tversky's seminal work on prospect theory, framing effects in risky choice have been robustly documented and attributed to two main psychological mechanisms: loss aversion (heightened sensitivity to potential losses relative to equivalent gains) and probability weighting (non-linear transformations of stated probabilities). Despite decades of research, the relative contribution of these two processes remains contested. Most previous work has relied exclusively on choice data, which provides limited insight into the underlying psychological mechanisms because both processes pull choices in the same direction under certain conditions.

The present work employs multiple methodological approaches—choice reversals combined with monetary valuation tasks—to disentangle the contributions of loss aversion and probability weighting. We hypothesize that by comparing choice patterns across different response modes, we can more precisely estimate the relative strength of each mechanism. This approach should yield insights applicable to real-world contexts in which decision-makers must commit resources based on both choices and valuations.

Method

Participants

Study 1 enrolled 58 university undergraduates (M age = 19.8 years, 32 female). Study 2 included 64 participants recruited from the community (M age = 34.2 years, 38 female). Study 3 had 71 undergraduates (M age = 20.1 years, 41 female). All participants reported normal or corrected vision and were compensated with course credit or $8 CAD.

Procedure

In each study, participants made decisions about monetary gambles presented on a computer screen. Gambles were framed in positive terms (e.g., "win X with probability p") or negative terms (e.g., "lose X with probability p"). Across studies, participants completed 20–30 choice trials using a two-alternative forced-choice format, followed by willingness-to-pay (WTP) tasks on a subset of gambles using a multiple price list procedure. Study 1 and 2 employed a within-subjects design crossing frame (gain vs. loss) and probability (10%, 50%, 90%). Study 3 added a between-subjects manipulation of reference point to test dissociation.

Results

In Study 1, choice data revealed significant framing effects: 76% of participants chose the risky option when the frame was positive and certain outcomes framed as losses (χ²(1) = 42.3, p < 0.001). The same gambles presented in gain-frame format elicited risky choice in only 31% of participants. WTP measures, however, showed smaller framing effects (Mpositive = $4.32, Mnegative = $3.87), t(57) = 2.1, p = 0.041.

A mixed-effects logistic regression model predicting binary choice (risky vs. certain) from frame, probability, and participant random slopes revealed a main effect of frame (β = 1.84, SE = 0.31, p < 0.001) and an interaction between frame and probability (β = 0.028, SE = 0.009, p = 0.002). Individual differences in the WTP ratio (negative-frame WTP divided by positive-frame WTP) were significantly predictive of choice reversal magnitude (r = 0.58, p < 0.001), suggesting loss aversion explains approximately 34% of variance in choice reversals.

Studies 2 and 3 replicated and extended these findings. Across all three studies, bootstrap resampling (10,000 iterations) estimated loss aversion coefficients ranging from 1.52 to 1.89, with 95% CIs narrowly excluding 1.0, indicating robust effects. Probability weighting distortions (estimated using power-law functions) accounted for an additional 25–35% of choice variance, independent of loss aversion effects.

Discussion

Our results indicate that loss aversion and probability weighting are dissociable mechanisms contributing independently to choice reversals. The finding that loss aversion explains approximately 60% of framing effects (based on relative variance explained) aligns with recent theorizing emphasizing the centrality of loss sensitivity in decision-making. However, the persistent influence of probability distortion underscores that simple models treating these as a single consolidated effect are inadequate.

The convergence of results across three independent samples and multiple response modes strengthens confidence in these estimates. The use of willingness-to-pay, in particular, offered a complementary window into valuation that partially mitigated response-mode effects known to inflate apparent loss aversion in choice-only paradigms. Future work should extend this approach to higher-stakes decisions and to investigate whether loss aversion and probability weighting predict real-world financial and medical decisions in similar proportions.

References

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291.
Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5(4), 297–323.
Prelec, D. (1998). The probability weighting function. Econometrica, 66(3), 497–527.
Camerer, C. F., & Ho, T.-H. (1994). Violations of the betweenness axiom and nonlinearity in probability. Journal of Risk and Uncertainty, 8(2), 167–196.
Wakker, P. P., & Tversky, A. (1993). An axiomatization of cumulative prospect theory. Journal of Risk and Uncertainty, 7(2), 147–175.
Gonzalez, R., & Wu, G. (1999). On the shape of the probability weighting function. Cognitive Psychology, 38(1), 129–166.