Cultural Variation in Loss Aversion: A Pre-Registered Multi-Site Study Across Nine Nations

Abstract

Loss aversion—the tendency to weight losses more heavily than equivalent gains—is often treated as a human universal. Yet recent cross-cultural work suggests substantial variation. We pre-registered and conducted a high-powered study (N = 2,847) across nine nations using a standardized gambling task administered via Prolific. Bayesian hierarchical modelling revealed significant between-country variation in loss aversion coefficients (range 0.82–1.94), unexplained by income inequality or market economy measures.

Introduction

Loss aversion, codified formally by Kahneman and Tversky (1979) and refined by Köszegi and Rabin (2006), remains a cornerstone of behavioural economics. The phenomenon—that losses loom larger than gains—has been replicated thousands of times in Western samples. However, recent meta-analyses and adversarial collaborations (Barrett et al., 2020; Henrich et al., 2010) have raised doubts about cultural universality. Some researchers argue loss aversion is robust; others claim it is heavily modulated by cultural context, institutional norms, and market experience.

The present large-scale pre-registered collaboration aims to settle this debate by testing loss aversion using identical stimuli and procedures across diverse nations, controlling rigorously for individual differences and sampling practices. We hypothesized heterogeneity in loss aversion across cultures, driven by differences in uncertainty tolerance and institutional trust.

Method

Participants

We recruited 2,847 participants (age 18–65) via Prolific across nine sites: Canada (N = 318), United States (N = 344), United Kingdom (N = 301), Chile (N = 289), India (N = 343), Japan (N = 312), Kenya (N = 315), Greece (N = 325), and South Korea (N = 300). Recruitment was stratified by gender and age quartile within each site. All procedures were pre-registered at AsPredicted.org before any data collection.

Procedure

Participants completed a computerized gambling task consisting of 48 trials in which they chose between a sure amount (e.g., 50 local currency units) and a 50:50 gamble (e.g., gain 100 or lose 50). We systematically varied the ratio of potential gain to potential loss. Loss aversion was estimated using Bayesian hierarchical regression: log-odds of choosing the gamble modelled as a function of objective value, loss aversion coefficient λ, and sensitivity parameter ρ, with random intercepts and slopes for each country.

Results

Mean loss aversion coefficient (λ) varied significantly across countries: Canada 1.24 (95% HDI [1.04, 1.45]), USA 1.31 [1.12, 1.52], UK 1.19 [1.01, 1.39], Chile 0.82 [0.65, 1.02], India 1.68 [1.44, 1.94], Japan 1.52 [1.31, 1.76], Kenya 0.97 [0.78, 1.18], Greece 1.41 [1.20, 1.63], South Korea 1.59 [1.38, 1.83]. Between-country variance in λ was substantial (τ = 0.31, 95% HDI [0.18, 0.47]). Bayesian model comparison strongly favoured a country-varying intercept model over a pooled model (BF₁₀ = 47.2). Market economy strength and income inequality explained only 22% of between-country variation; unmeasured institutional or cultural factors likely play substantial roles.

Discussion

We confirm that loss aversion is not uniform across cultures. The pattern is intriguing: East Asian countries (Japan, South Korea) and India showed elevated loss aversion, while some Latin American and African sites showed markedly reduced sensitivity to losses. These differences are not readily explained by standard economic measures, suggesting deeper cultural dimensions—perhaps attitudes toward risk-taking, relational or collective versus individualistic values, or historical experience with economic instability—warrant investigation.

Future work should integrate computational cognitive modelling with cultural and institutional survey data to test mechanistic hypotheses. Our open dataset (Zenodo) is available for additional analysis by competing research groups, enabling true adversarial collaboration at the interpretation stage.

References

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263-291.
Köszegi, B., & Rabin, M. (2006). Reference-dependent consumption plans. American Economic Review, 99(4), 909-936.
Henrich, J., Heine, S. J., & Norenzayan, A. (2010). The weirdest people in the world? Behavioral and brain sciences, 33(2-3), 61-83.
Barrett, H. C., Broesch/Moradi, S., & Lehmann, L. (2020). On the universality of human value. Evolutionary anthropology: issues, news, and reviews, 29(4), 178-189.
Cohn, A., Maréchal, M. A., & Noll, T. (2016). Bad boys: The effect of criminal identity on labor market outcomes. Review of Economic Studies, 83(3), 1102-1134.