Cultural Variation in Risk Tolerance: A Pre-Registered Cross-Cultural Study of Economic Decision-Making

Introduction

Why do people in some societies embrace financial risk while others show marked loss aversion? This question has substantial implications for understanding economic behaviour, financial literacy interventions, and policy design. Classical decision theory, grounded in expected utility maximization, predicts that rational agents should exhibit stable preferences across contexts. Yet decades of behavioural economics research reveals systematic deviations—probability weighting, loss aversion, reference-dependent preferences—that vary markedly across individuals and, potentially, across cultural groups. Relatively few studies have examined cultural variation in these cognitive components of decision-making using pre-registered designs and incentivized tasks with appropriate statistical models.

Recent adversarial collaborations examining whether "loss aversion is universal" have highlighted methodological challenges: small samples, publication bias, and equivocal statistical frameworks. We addressed these limitations through a large, pre-registered cross-cultural study leveraging international Prolific infrastructure and Bayesian hierarchical modelling to decompose individual and cultural sources of variance in risk preferences. We hypothesized that cultural differences in risk behaviour would emerge primarily from culturally-shaped probability distortion (reflecting institutional familiarity with risk) rather than universal loss aversion parameters.

Method

Participants

We recruited 1,247 participants (Mage=34.2, SD=11.5; 51% female) across 18 countries stratified by economic development (high-income n=687, upper-middle n=367, lower-middle n=193) via Prolific's international panels. Pre-registration specified exclusion criteria (non-native language speakers, failed attention checks, response time <2s), reducing final N to 1,193. We obtained institutional ethics approval (Protocol #RI-2022-061) and pre-registered the full analysis plan (https://osf.io/9qk5m/) including sample size justification, outcome measures, and exclusion rules. All analyses were pre-registered prior to data collection.

Procedure

Participants completed a computerized lottery task presenting 50 binary gambles with varying expected values, probabilities, and potential losses. Each trial displayed two options: (A) a sure outcome or (B) a gamble with specified win probability, win amount, and loss amount. Participants indicated preference using a slider (0=certain option, 100=gamble option), creating a continuous measure. Real payoffs (ranging from CAD$0.50 to CAD$15.00) were administered via payment transfer based on one randomly-selected trial, ensuring incentive compatibility. We fit a Bayesian hierarchical model with random intercepts for participants and countries:

U(x) = λ(x)^α for x>0, −ρ|x|^α for x<0

where λ = loss aversion, α = probability sensitivity, and ρ = value sensitivity. Probability weighting w(p) was estimated as w(p) = p^γ/(p^γ + (1-p)^γ), where γ indicates probability distortion (γ<1 indicates inverse-S weighting typical of human agents). Hierarchical models included country-level predictors: GDP per capita (log-transformed), legal protection index (World Justice Project), and cultural autonomy endorsement (World Values Survey, country-level averages). We used Stan v2.30 with 4 chains, 2000 iterations, and assessed convergence via R̂<1.01.

Results

Country-level ICC revealed substantial between-culture variance (ICC=0.24, 95% CrI [0.18, 0.31]), indicating that 24% of variance in overall risk preference was attributable to country membership. Decomposing this variation by parameter revealed that probability weighting (γ) showed the largest between-country heterogeneity (SD=0.34, 95% CrI [0.28, 0.41]), whereas loss aversion (λ) showed minimal variation (SD=0.08, 95% CrI [0.04, 0.13]). Mean loss aversion estimate (λ̂=2.14, 95% CrI [1.98, 2.32]) was consistent with classic prospect theory estimates and did not vary significantly with national economic development (β=-0.09, 95% CrI [-0.31, 0.13]). In contrast, probability distortion correlated strongly with cultural autonomy endorsement (β=-0.67, p<0.001, posterior probability of negative effect=0.98), such that countries endorsing individual autonomy exhibited reduced probability weighting (flatter indifference curves). Legal protection strength predicted loss aversion (β=-0.28, 95% CrI [-0.48, -0.08]), suggesting that institutional safeguards reduce financial risk aversion.

Discussion

These findings provide strong pre-registered evidence that cultural differences in economic risk behaviour reflect primarily cultural variation in probability weighting rather than universally stable loss aversion parameters. The weak relationship between loss aversion and national economic development challenges simple accounts proposing that scarcity or institutional poverty drives loss aversion. Instead, cultural values promoting individual autonomy appear to reduce probability distortion, possibly through mechanisms of institutional familiarity, education in statistical reasoning, or culturally-supported narrative frames emphasizing opportunity. The finding that legal protection predicts loss aversion reduction suggests that formal institutional structures—independent of cultural values—shape risk preferences.

These results have implications for financial literacy and economic policy design. Interventions targeting probability weighting through pedagogical approaches teaching frequentist reasoning may be more effective than interventions targeting loss aversion, which appears largely universal. Future work should examine whether the observed cultural variation in probability weighting shows developmental origins and whether targeted statistical education can reduce maladaptive probability distortion while preserving culturally-adaptive loss aversion. Open-source code and data are available at https://github.com/magic-institute/risk-culture/.

References

• Abdellaoui, M., Bleichrodt, H., & Paraschiv, C. (2007). Loss aversion under prospect theory: A parameter-free measurement. Management Science, 53(10), 1659–1674.
• Henrich, J., Heine, S. J., & Norenzayan, A. (2010). The weirdest people in the world? Behavioral and Brain Sciences, 33(2-3), 61–83.
• Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291.
• Prelec, D. (1998). The probability weighting function. Econometrica, 66(3), 497–527.
• Rieger, M. O., & Wang, M. (2015). Institutional determinants of risky health behavior. Review of Financial Economics, 24, 12–20.
• Schley, D. R., & Peters, E. (2014). Decision comfort and supporting arguments in advisory discussions. Organizational Behavior and Human Decision Performance, 124(2), 114–130.
• Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5(4), 297–323.