In everyday decision-making, we weigh the risks and rewards (i.e., probability) associated with competing outcomes in terms of relative payoffs and the potential for negative outcomes (i.e., gains vs. losses). For decades scientists have been investigating decisions under risk and ambiguity and proposed several models that quantify decision-making.

One of the most influential frameworks for researching decision-making under risk is the prospect theory (PT; Kahneman and Tversky, 1979; Tversky and Kahneman, 1991). The theory posits that decisions are affected by how potential outcomes (prospects) are cognitively represented in terms of gains, losses, and their associated probabilities. As such, rational individuals aim to optimize outcomes by weighing the value, probability, and cumulative wealth. According to PT, the value V(x, p) of a simple prospect that pays x with probability p and can be represented as:

where v(p) measures the subjective value of the consequence x, and measures the impact of probability p on the attractiveness of the prospect (Tversky and Kahneman, 1991; Stanton, 2017).

According to prospect theory, the value function V(x, p) exhibits the psychophysics of diminishing sensitivity with the value function being steeper for losses (loss aversion) than for gains. In other words, a loss of €50 is felt more than a gain of €50. Loss aversion reflects a tendency to avoid losses relative to acquiring equivalent gains. The utility of monetary payoffs that depends on previous experiences is what distinguishes loss from risk aversion. Overall, prospect theory argues that decision-makers weigh gains and losses differently, weighing perceived gains more than perceived losses. This shows that humans think in terms of expected utility relative to a reference point (i.e., current financial situation/wealth) rather than in absolute outcomes.

A refinement of the original version of prospect theory, cumulative prospect theory (CPT) applies different weights to the cumulative probability distribution function instead of applying weights to the probabilities of individual outcomes (Tversky and Kahneman, 1992). This ensures that the value of an outcome is multiplied by a decision weight and not by an additive probability (for a model overview, please see Fox and Poldrack, 2009). A recent study across 19 countries and over 4,000 participants replicated the empirical foundations of the prospect theory and showed that none of the prospect theory's theoretical constructs (i.e., isolation effect, framing effects, overweighting of small probabilities, etc.) from the original study were unreliable (Ruggeri et al., 2020). These findings confirm that prospect theory is a robust descriptive model of decision-making under risk and uncertainty.

Unlike risky decision-making, riskless choice relates to prospect theory through the so-called endowment effect (EE) (Thaler, 1980; Kahneman et al., 1991). The endowment effect has been used as evidence for theories of reference-dependent preferences and loss aversion and it shows that the minimum amount of money people are willing to accept (WTA) when selling an item is significantly higher than the minimum amount they are willing to pay (WTP) for the same item if they do not already own it (Kahneman et al., 1991; Ariely et al., 2005; Votinov et al., 2010, 2013; Camerer, 2011; Gächter et al., 2021). Endowing a person with a good (even hypothetically), seems to establish a reference point people move away from reluctantly or if they are paid a large sum (Camerer, 2011). This observation underscores a status quo bias toward own goods. Furthermore, an extreme feeling of possession may lead to an excessive acquisition of items resulting in experiencing difficulties discarding them, which can be linked to hoarding disorder (Pushkarskaya et al., 2020). In another words the EE is a byproduct of loss aversion in a way that the losing things is more painful than enjoyment of gaining them.

The type of goods (utilitarian vs. hedonic) that consumers buy or sell heavily influences the WTA/WTP ratio. Utilitarian goods focus on function, practicality, and are generally more of what people “need” as opposed to hedonic goods which provide more fun, are not as functional, and reflect what people “want” (Hirschman and Holbrook, 1982; Strahilevitz and Myers, 1998). Buying and selling hedonic instead of utilitarian goods has previously been found to increase the WTA/WTP ratio (Chan, 2015). For example, one field survey (Dhar and Wertenbroch, 2000) demonstrated this by collecting data from two hundred seventeen MBA students about the features of the car they owned. Participants rated their car on a scale from one to nine for hedonic dimensions and then separately for utilitarian dimensions. They were then asked how much money they would be willing to sell their car for. As predicted, those with a more hedonic car would demand more money concerning their car's market price than those with more utilitarian cars would. Another study by Cramer and Antonides (2011) tested decision making with respect to hedonic vs. utilitarian food products. They observed strong endowment effect for hedonic compared to utilitarian food products and proposed that status quo bias for hedonic food products may lead to relatively unhealthy food choices. A potential reason for this disparity may be because the owners of hedonic goods may develop a more significant symbolic relationship with those goods than with utilitarian goods (Belk, 1988).

Testosterone is one of the major steroid hormones produced in men and women (though in different quantities) that modulates brain activity by binding to intracellular androgen receptors and regulating gene expression via genomic and non-genomic effects (for a comprehensive review, see Brinkmann, 2011). Developments in neuroeconomics and neuroendocrinology highlighted shared biological mechanisms that support decision-making involving economic risk-taking. As such, testosterone is an interesting biomarker for socioeconomic decision-making and for status seeking behaviors.

Engaging in risky behaviors might be an evolutionary strategy to attain higher positions in social hierarchies where status is important. It is therefore likely that testosterone also modulates risk-taking behaviors and several studies reliably associated testosterone with social and economic decision-making under high risk (Eisenegger et al., 2011; Apicella et al., 2014). A recent meta-analysis on studies measuring endogenous levels of testosterone showed a significant albeit small correlation between testosterone and risk-taking which was not influenced by gender nor by outcome measures (Kurath and Mata, 2018). Nevertheless, several other studies reported contrasting see Mehta et al. (2015) or null results (Van der Loos et al., 2013; Derntl et al., 2014; Nadler et al., 2021; Stanton et al., 2021). However, it should be noted that these studies used different approaches with some measuring basal testosterone in saliva (Derntl et al., 2014; Mehta et al., 2015), some in the serum (Van der Loos et al., 2013), while the rest (Nadler et al., 2021; Stanton et al., 2021) administered testosterone exogenously.

Steroid hormones including testosterone preferentially modulate brain activity (Dreher et al., 2007; Votinov et al., 2020) regulating the neural function of areas involved in economic decision-making and emotion regulation. Several studies measured endogenous testosterone levels and found that high testosterone concentrations correlate well with increased risk-taking in both men and women thus indicating a heightened willingness to take financial risks (Garbarino et al., 2011; Stanton et al., 2011a; Chicaiza-Becerra and Garcia-Molina, 2017). Moreover, some studies showed a positive relationship between basal endogenous testosterone levels and decreased risk-aversion thus showing that testosterone may circumstantially promote greater risk neutrality (Apicella et al., 2008; Sapienza et al., 2009; Stanton et al., 2011a). Other studies investigated the effects of exogenous (pharmacologically-elevated) testosterone on risky decision-making. In men, transdermal testosterone shifted investments toward riskier assets (Cueva et al., 2015) and increased risk-taking under conditions of unknown probabilities during strategic decision-making (Goudriaan et al., 2010; Wagels et al., 2017). Exogenous testosterone increased risk-taking in the Iowa Gambling Task (IGT) in a group of women, likely by increasing reward sensitivity while decreasing punishment sensitivity. Similarly, endogenous testosterone accounted for 11% decision-making in the uncertainty phase of the IGT (Singh, 2021).

The literature on testosterone and risk-taking is far from conclusive. For instance, several studies (Zethraeus et al., 2009; Boksem et al., 2013) found no effect of exogenous testosterone (sublingual administration) on women's risk attitudes or ambiguity tolerance. Likewise, in a double-blind within-subjects study investigating risk-taking and loss aversion, Stanton and colleagues (2021) found no consistent relationship between pharmacologically-elevated testosterone and economic decisions (Stanton et al., 2021). Instead, they argue that findings may be explained by several situational moderators.

Not only is testosterone associated with risk-taking and decision-making but it also influences consumer behavior by altering consumer-based choices and preferences. For instance, prenatal testosterone exposure correlates well with men's courtship-related gift-giving (Nepomuceno et al., 2016a,b) in an attempt to attain a certain status. The study investigated the influence of testosterone on men's preferences for positional goods and found that administering testosterone increases men's preference for status brands compared to brands of similar perceived quality but lower perceived status. Testosterone further increases positive attitudes toward positional goods when they are described as status-enhancing (Nave et al., 2018). Consumer choices are thus directly linked to competition and competing interests (Durante and Griskevicius, 2016) which are also subject to endocrine effects (Cherki et al., 2021).

While several studies investigated how people may experience the endowment effect and what role testosterone plays in some aspects of loss aversion, the interaction between testosterone and endowment effect has not yet been thoroughly explored. Similarly, the effects of pharmacologically-elevated testosterone on context-depended risk taking-taking have not been investigated.

As such, we aim to investigate the role of testosterone on risk-taking and loss aversion under positive and negative framing.

First, we examine whether testosterone administration affects decision-making in a series of gambles under positive and negative framing (gain and loss) associated with various probabilities. Under testosterone administration, we expect participants to show increased risk-seeking behaviors regardless of framing type.

Second, we investigate testosterone's effect on loss aversion (endowment effect) during the economic valuation of two goods categories (utilitarian vs hedonic). We hypothesize that the endowment effect (and therefore the WTA/WTP ratio) will be higher for hedonic compared to utilitarian goods. Compared to placebo, we expect testosterone administration to further increase the endowment effect (and the WTA/WTP ratio) for both types of goods.

We used Python (v3.8) for data preprocessing and analyzes. Preprocessing and exploratory data analyzes were performed with pandas (v1.2) and seaborn (v 0.11) libraries. Bayesian statistical analysis and modeling were performed with a custom code based on PyMC3 (v 3.11) (Python package for Bayesian statistical modeling and probabilistic machine learning) and openly accessible through Jupyter notebooks containing detailed research procedures, methods, and results. A detailed description of every step for each analysis with the code and data are available at GitHub and OSF. Bayesian modeling involves making choices concerning prior distributions, likelihood functions and model evaluation, which includes posterior checks and quality assessments (Aczel et al., 2020). Here we provided and briefly described the models we used for data analysis. Full model specification and the description of estimation procedures are available in the Supplementary File.

Using frequentist and Bayesian modeling approaches, this study investigated the effects of testosterone administration on risk-taking decision-making under positive and negative framing and the role of testosterone in the endowment effect. Treatment manipulation was successful and we observed a significant increase in serum testosterone concentrations 1.5 h after transdermal application. Compared to placebo, testosterone administration increased risk-taking under the positive framing and decreased under the negative framing. Our model showed that decision-making is jointly influenced by testosterone and the trade-off between gains and losses. Moreover, we observed the expected endowment effect which was more pronounced for hedonic than for utilitarian goods. Unlike our predictions, however, testosterone did not alter the endowment effect.

There is ongoing debate as to whether testosterone increases risky behavior and choice. Our aim was to examine whether testosterone administration affects risky behavior in a series of games under positive and negative framing (gain and loss) that were associated with different probabilities. We expected participants to show increased risk-seeking behaviors regardless of framing type but we only observed this pattern in the positive framing condition.

Several studies showed a causal or correlational effect of endogenous and exogenous testosterone on increased risk-taking behaviors. For instance, high endogenous testosterone levels correlated with increased risk-taking in both genders and with the willingness to take higher financial risks (Garbarino et al., 2011; Chicaiza-Becerra and Garcia-Molina, 2017). Similarly, baseline endogenous testosterone levels were associated with increased risk-taking (Stanton et al., 2011a) showing that both women and men with higher testosterone levels acted riskier than the groups having lower testosterone levels. Similar results were observed in studies that altered testosterone levels pharmacologically. For instance, exogenous testosterone increased risk-taking in the Iowa Gambling Task (van Honk et al., 2004) by preferentially modulating the punishment and reward sensitivities.

While several studies showed a positive association between basal testosterone or pharmacologically elevated testosterone and risk-taking (van Honk et al., 2004; White et al., 2006; Goudriaan et al., 2010; Apicella et al., 2014; Evans and Hampson, 2014), there is also evidence of null results (Zethraeus et al., 2009; Boksem et al., 2013; Ortner et al., 2013; Van der Loos et al., 2013; Derntl et al., 2014) or even negative correlations (Stanton et al., 2011b; van Anders et al., 2012). However, direct causal comparisons between various studies investigating risk-taking and testosterone are limited by methodological constraints that may hamper interpretations. For instance, the doses used to pharmacologically elevate testosterone levels are heterogeneous within and between sexes (i.e., in men, exogenous applications vary from 50 to 150 mg transdermal doses vs. the standardized 0.5 mg sublingual administration in women). Likewise, endogenous measures of testosterone levels are computed using 2D:4D ratios, saliva, blood, or hair sampling that all have inherent limitations.

It should be considered that other factors may play a role as well. The study using the Iowa Gambling Task showed a joint role of cortisol and testosterone in decision-making in the Singh (2021). Specifically, cortisol impaired decision-making in trials with high uncertainty, while testosterone improved decision-making in high risk trials. Several other studies examined the effect of testosterone using the Balloon Analog Risk Task (BART). Goudriaan et al. (2010) examined the effects of testosterone after treating participants with an aromatase inhibitor (letrozole 2.5 mg) for a week and found increased risk-taking in the high testosterone group in the BART but not in the Iowa Gambling Task (Goudriaan et al., 2010). Moreover, Wagels and colleagues (2017) found an interaction between MAOA polymorphism and testosterone and showed that MAOA-S carriers played riskier than the MAOA-L carriers (Wagels et al., 2017).

Only a handful of studies used a similar approach to ours in testing risk-taking behaviors using binary decision games. Stanton et al. (2011b), for instance, measured testosterone levels in participants who performed a risk preference, loss aversion, and ambiguity task and found a U-shaped effect for risk and ambiguity. In other words, participants with low and high testosterone levels were risk- and ambiguity-neutral, whereas individuals with intermediate testosterone levels were risk- and ambiguity-averse. These findings contradict our results where we observed increased risk-taking behavior under positive framing. Our model showed that the shift to risky behavior depends on the individual level of testosterone fluctuation, the gain at stake, and the probability of obtaining this gain. We observed a reverse behavioral pattern under negative framing. Contrary to our predictions, testosterone administration tuned participants' behavior to be more risk-averse. Regardless of treatment (placebo or testosterone), however, participants were more risk-taking when the prospected financial gains were higher. As such, we suggest that participants follow an individual outcome optimization strategy and reduce potential loss when inevitable while increasing potential gains when there is a chance for it. This optimization strategy, however, is less susceptible to hormonal influence and more in line with contextual factors. Nadler et al. (2021) presented participants with two gambles under positive framing. Participants could select either a sure option to get $1 or gamble with obtaining $3 with a varying probability between 0.1 and 1 or getting nothing (Nadler et al., 2021). They excluded from analysis participants who showed intransitive preferences and classified the remaining participants into three risk-taking levels: risk-averse, slightly risk-averse, and risk-loving. Their analyzes revealed similar risk-taking profiles and no treatment effects. Another recent study by Stanton et al. (2021) similarly found no consistent relationship between testosterone and economic decision-making. Since testosterone does not work in isolation, it is likely that potential contextual moderators may alter the circumstances under which testosterone affects economic decision-making.

In the endowment task participants state their monetary preference for buying and selling various items falling into hedonic or utilitarian categories. Results showed that the WTA/WTP ratio was significantly higher and thus we observed the predicted endowment effect. The finding aligns with previous literature (Thaler, 1980; Knetsch, 1989; Kahneman et al., 1991; Van Dijk and Van Knippenberg, 1998; Carmon and Ariely, 2000; Votinov et al., 2013; Gächter et al., 2021) that demonstrated an endowment effect across various types of goods (real owned and imaginary) as well for different types of transactions (goods-to-money, goods-to-goods). Moreover, as predicted, we found a larger endowment effect for hedonic compared to utilitarian items. Previous findings argued that owning hedonic items might facilitate the development of a symbolic relationship to the items compared to items serving a clear purpose (i.e., utilitarian; Belk, 1988). This emotional attachment toward an item can be associated with more positive emotions and increased aesthetic value which might resemble the feelings hedonic items yield (Ariely et al., 2005).

Testosterone administration did not alter the endowment effect in general nor did alter the preference for goods. Investigating items individually, we found an increase in WTA/WTP ratio only for a few of the items in both groups. We believe this effect could be item specific and bound to the subjective value individuals ascribed to them or, alternatively, could be random noise. Previous studies found that the effect of testosterone on consumer behavior was associated with status or owning luxury goods that served to attract mates, intimidate potential rivals (Nepomuceno et al., 2016a,b) or change product preferences (Aspara and Van Den Bergh, 2014). However, Wu and colleagues (2017) tested the association between competition, testosterone fluctuation, and conspicuous consumption and found no evidence for a hormonal regulation of these effects (Wu et al., 2017). Nevertheless, winners in the competition showed increased self-reported WTP and an implicit bias toward high-status products. These studies, however, measure testosterone levels predominantly using prenatal markers (i.e., 2D:4D ratio) that have limited ecological validity (Berenbaum et al., 2009; Nadler et al., 2021). Thus, one cannot rule out the situational effects of decision-making that may be additive to increased endogenous testosterone concentrations. Future research should address the association between the type of consumer goods and test real items that participants own while controlling for their emotional attachment to a given item (e.g., childhood toy or present from a beloved one).

Our results should be interpreted in the light of several limitations. Only young, healthy men took part in this study. Moreover, due to inherently different hormonal profiles, the results cannot be generalized to women. Nevertheless, prior animal model and human research suggests that acute testosterone effects are expected to manifest predominantly in men due to the distinct hormonal profiles. Last, we did not have an endogenous measure of testosterone concentration prior to the testing sessions (2D:4D measurement) to correlate with baseline intake testosterone levels. There is evidence that prenatal testosterone plays a strong role in the activational effects of testosterone for higher-order social cognition (Terburg et al., 2016). Nevertheless, 2D:4D ratio results regarding testosterone effects are rather heterogenous and susceptible to measurement errors/bias. Future studies should consider other factors which may influence testosterone concentrations, such as levels of enzyme aromatase, sex hormone binding globulin (SHBG) and expression of androgen receptors (AR).

This study showed the effect of testosterone administration on risk-taking behavior in young healthy men. Testosterone increased risk-taking behaviors under positive framing and risk aversion under negative framing. However, the sensitivity to gain remained positive in each framing. We found evidence for the endowment effect which was increased for hedonic com- pared to utilitarian items. This effect was independent of testosterone. Our results underline the modulatory role of testosterone on risk-taking within different framing. The models we tested can be used as predictive models for risk-taking behavior and may be applied in larger and more diverse samples. Further studies should examine the gender-specific aspects of risk preference and the link to the modulatory effects of testosterone among males and females. Additionally, the effect of different age groups on risk preference warrants further research considering that the testosterone concentrations vary over the lifespan, specifically in men. This could further be extended beyond financial risk-taking to other risky behaviors (i.e., exploration, promiscuity, etc.).

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

The studies involving human participants were reviewed and approved by Ethics Committee of the Medical Faculty, RWTH Uniklinik. The patients/participants provided their written informed consent to participate in this study.

MV and AP designed the experiment. AP performed the experiment. IK and MV analyzed the data. MV, IK, and AP wrote the manuscript. MV, IK, UH, KK, and AP revised the manuscript and provided critical feedbacks. All authors contributed to the article and approved the submitted version.

AP was supported by a scholarship from the International Research Training Group. The Neuroscience of Modulating Aggression and Impulsivity in Psychopathology (IRTG-2150) of the German Research Foundation (DF'Projektnummer 269953372/GRK2150).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.