A Bayesian Modeling Approach to Examine the Role of Testosterone Administration on the Endowment Effect and Risk-Taking.

Financial risk-taking and loss aversion are multifaceted phenomena that are the focus of neuroscience, psychology, and economics research. A growing number of studies highlighted the role of hormones (particularly of testosterone) on socio-economic decision-making. However, the effects of testosterone on risk-taking under framing and consumer-based choices and preferences are inconclusive. We investigated the effects of 100 mg testosterone administration on aspects of decision-making within the Prospect Theory framework which is the most used descriptive model of decision-making under risk. We assessed risk-taking under framing and the endowment effect (effect of possession) using Bayesian modeling. Forty men participated in this double-blind placebo-controlled fully-randomized cross-over experiment and performed two tasks. One was a risk-taking task with binary choices under positive and negative framing associated with different probabilities. In the second task participants had to bid money for hedonic and utilitarian items. We observed a significant increase in serum testosterone concentrations after transdermal application. Compared to placebo, testosterone administration increased risk-taking under the positive framing (very large effect size) and decreased under the negative framing (moderate to small). The sensitivity to gain was positive in each framing. Our model showed that decision-making is jointly influenced by testosterone and the trade-off between gains and losses. However, while the endowment effect was more pronounced for hedonic than for utilitarian items, the effect was independent of testosterone. The findings provide novel information on the complex modulatory role of testosterone on risk-taking within the framework of prospect theory and shed light on mechanisms of behavioral economic biases. The proposed models of effects of individual differences in testosterone on risk-taking could be used as predictive models for reference-depended behavior under positive and negative framing with low and high probabilities.