The association between testosterone and risk-taking behavior has been widely investigated across behavioral economics, neuroendocrinology, and social neuroscience, but empirical results remain inconsistent. To clarify this relationship, we conducted a multilevel random-effects meta-analysis of 52 empirical studies (94 independent effect sizes; total N = 17,340), the most comprehensive so far, examining correlations between testosterone levels or manipulations and risk preferences across diverse paradigms. The aggregated effect was statistically null (r = -0.0021, 95% CI -0.0431, 0.0389, p = .919), indicating no reliable link between testosterone and risk-taking. Publication bias diagnostics (trim-and-fill and fail-safe N) suggested that this null effect is not driven by selective reporting. Meta-regressions revealed significant heterogeneity across testosterone measurement type. Moreover, only lottery-based economic tasks showed a modest positive association, whereas other paradigms (e.g., BART, IGT, self-report) did not. A separate meta-analysis of sex differences found no moderating effect, suggesting that testosterone-risk correlations are not reliably stronger in males than females. Overall, the evidence challenges the notion that testosterone provides a general hormonal basis for human risk preferences. Instead, findings support a biopsychosocial framework in which "risk taking" reflects the interaction of task demands, cognitive-affective processes, and situational context, with endocrine effects appearing narrow, context-dependent, and method-specific. Future work should employ preregistered, multi-measure designs and direct endocrine assays to test mechanistic pathways more precisely.
Rodríguez et al. (Mon,) studied this question.