On the Physiology of Investment Biases: The Role of Cortisol and Testosterone John R. Nofsinger a , Fernando M. Patterson b , and Corey A. Shank c a University of Alaska Anchorage, College of Business & Public Policy; b North Carolina Central University; c Dalton State College ABSTRACT The underlying physiological mechanisms of biases are not well understood. As such, we examine the impact of testosterone and cortisol levels on several commonplace investment biases using realistic trading simulations. Cortisol, the biological marker of stress, is posi- tively related to the disposition effect and portfolio turnover, which is consistent with the relation between judgment errors and stress in social settings. Testosterone, the male hor- mone, is also positively related to portfolio turnover, which is consistent with androgen- driven behaviors. Overall, the results show that the endocrine system plays a significant role during financial decision-making, which has important consequences for the finan- cial industry. KEYWORDS Investment biases; Disposition effect; Testosterone; Cortisol; Dual- hormone hypothesis JEL CLASSIFICATION C91; G02; G11 Introduction The stock market is the aggregation of individual investment decisions. At times, market prices can deviate significantly from fundamentals when the sum of those investment decisions is irrational (i.e., have an emotional component), which can result in market bubbles and ensuing crashes (for example, see Barberis and Thaler 2003). Behavioral finance consist- ently shows that irrational decisions stem from cogni- tive errors and behavioral biases. For example, the disposition effect describes the situation in which investors hold on to underperforming assets for too long and sell overperforming assets too soon, resulting in diminished portfolio performance (Shefrin and Statman 1985; Dhar and Zhu 2006). Similarly, invest- ors with high portfolio turnover consistently earn lower portfolio returns (Barber and Odean 1999; Barber and Odean 2000; Barber et al. 2009). Behavioral biases appear to persist over time, even when those biases result in lower returns. One source of those cognitive errors is the individual ’ s underlying biological makeup. The psychobiological mechanisms underlying the disposition effect, portfolio turnover, and the vast majority of documented investment biases have rarely been studied. The related medical literature suggests that steroid hormones – notably testosterone and cortisol – play an important role in risk-reward decisions. This is particularly important in the finance industry, given that it is still very much a male-dominated, androgen-driven field, and given research showing that occupational stress is very high among finance professionals (Kahn and Cooper 1990; Jones et al. 2003; Oberlechner and Nimgade 2005). Additionally, Perman (2012) shows that financial executives use intramuscular testosterone injections or topical testosterone creams as a way to get “ their edge back. ” Exogenous androgen supplementation can lead to increased corticosteroid production, resulting in a situation in which both testosterone and cortisol are highly elevated. Therefore, examining the role of tes- tosterone and cortisol on investment biases is impera- tive to understanding the physiology of financial decisions. One theoretical framework on the joint role of tes- tosterone and cortisol on neural activity is the dual- hormone hypothesis. According to the dual-hormone hypothesis, cortisol modulates the effects of testoster- one on the nucleus accumbens. As a part of the dopa- mine system, the nucleus accumbens is associated with pleasure and irrational risk-seeking behavior (Kuhnen and Knutson 2005). Testosterone and corti- sol cause the release of the neurotransmitter dopamine in the brain, which in turn, modulates cognitive and emotional processes associated with reward-seeking behavior (Frye et al. 2002). Additionally, cortisol acts CONTACT John R. Nofsinger jnofsinger@alaska.edu William H. Seward Chair in International Finance, University of Alaska Anchorage, College of Business & Public Policy, Anchorage AL 99508. ß 2020 The Institute of Behavioral Finance JOURNAL OF BEHAVIORAL FINANCE 2021, VOL. 22, NO. 3, 338 – 349 https://doi.org/10.1080/15427560.2020.1775600 as a regulator of testosterone activity by suppressing testosterone-based cortical and subcortical neural communications that control socially aggressive behavior (for example, see Rubinow et al. 2005 or Van Honk et al. 2010). According to this theoretical framework, testosterone plays a role in risk-seeking behaviors when cortisol levels are low. In support of this view, research shows that the imbalance of high testosterone coupled with low cortisol levels is associ- ated with anger (Hermans, Ramsey, and Van Honk 2008), social aggression (Terburg, Morgan, and Van Honk 2009), social dominance (Mehta and Josephs 2010), and social risk-taking behaviors (Barel, Shahrabani, and Tzischinsky 2017). Overall, it is expected that testosterone and cortisol act individually and in tandem on risk-reward neural systems. Therefore, the present study examines both the indi- vidual and joint roles of testosterone and cortisol on irrational financial decisions. We employ a financial simulation with two multi- point portfolio rebalancing tasks involving long-term investment decision-making to examine the role of testosterone and cortisol on investment biases. Overall, the results show that there are significant individual and joint effects of steroids on investment biases. We find that testosterone is positively related to portfolio turnover. Similarly, cortisol is positively related to the disposition effect and portfolio turnover. Testosterone and cortisol also jointly impact financial choices. Specifically, the testosterone to cortisol ratio has a positive relationship with portfolio turnover, which corroborates the dual-hormone hypothesis dur- ing financial decision-making. These results have sig- nificant implications for financial decisions and performance of all kinds of investors, from the retail investor to Wall Street professionals, especially given the androgen-driven and stressful field of finance. The paper is organized as follows. The next section presents a literature review of the brain regions and hormones involved in financial decisions and dis- cusses our hypotheses. The third section presents the study design. The results are described in the fourth section. We conclude in fifth section. Literature review Testosterone and financial decision-making Transient physiological processes, such as hormonal action, lead to neurochemical changes that underlie the intricate path from molecule to decision-making. Coates, Gurnell, and Sarnyai (2010) assert that testos- terone causes a significant influence on the cognitive processes that involve how individuals comprehend financial information. They argue that testosterone shifts the economic utility functions, confidence levels, and risk preferences through its effect on the brain ’ s nucleus accumbens, which in turn influences financial decisions. Derntl et al. (2009) find that higher testos- terone levels are related to the heightened activity of the amygdala during emotional responses, which is involved in fear-based biased choices (Radke et al. 2015). Brain regions associated with financial deci- sion-making are also directly involved in the process- ing of emotions. It is widely held that cognitive errors, such as investment biases, involve emotions (Yuen and Lee 2003; Lerner, Small, and Loewenstein 2004; Nofsinger 2005). There are a few studies on testosterone and finan- cial choices. Coates and Herbert (2008) examine the influence of testosterone on financial decision-making by measuring the morning testosterone levels of 17 male floor traders for eight consecutive business days under real working conditions. They find that traders achieve higher afternoon returns on days when morn- ing testosterone levels are higher than the median morning testosterone level during the previous eight days. Nofsinger, Patterson, and Shank (2018) find that salivary testosterone is related to excess risk-taking during an investment task. Nadler et al. (2018) use a sample of male traders who are exogenously adminis- tered testosterone to understand the influence of tes- tosterone on asset pricing bubbles, showing that traders who were administered testosterone exogen- ously tend to bid higher amounts for stock prices, which may create mispricing that leads to longer-last- ing bubbles. Overall, the aforementioned studies sug- gest a link between testosterone and financial performance. Coates, Gurnell, and Rustichini (2009) use the ratio of the second to fourth digit length (2 D:4D) of 44 male traders to examine the relationship between in- utero testosterone exposure and financial perform- ance. 1 The authors find that subjects exposed to higher amounts of testosterone in-utero achieve higher returns over the 20 month period of the study. Cronqvist et al. (2015) find that individuals with higher in-utero testosterone exposure hold portfolios with higher amounts of equity and volatility. Additionally, the stock returns of CEOs who have a more masculine face are more likely to have higher volatility and idiosyncratic risk (Ahmed, Sihvonen, and V € ah € amaa 2019; Kamiya, Kim, and Suh 2016). 2 Given the previously mentioned literature about JOURNAL OF BEHAVIORAL FINANCE 339 testosterone and risk seeking behavior, we postulate our first hypothesiss: H1: Investors with higher testosterone have a higher turnover ratio and will trade more frequently than investors with lower testosterone levels. Furthermore, Lu and Teo (2018) use the facial mas- culinity, which measures testosterone exposure pre- natal and postnatal, and find that hedge fund managers have a negative risk adjusted return due to lottery seeking behavior. Following Lu and Teo (2018) ’ s findings, we posit our second hypothesis: H2: Investors with higher testosterone levels will exhibit more lottery seeking behavior and be more likely to invest in funds that have higher previous returns because the higher returns suggest higher risk. However, one criticism of using proxies for testos- terone exposure is that they do not always agree with the results of studies that use adult testosterone levels. For example, Nave et al. (2017) find that exogenous testosterone levels are related to a greater number of incorrect answers during the Cognitive Reflection Test. In contrast, Bosch-Dom enech et al. (2014) show that individuals with higher in-utero testosterone exposure exhibit a greater number of correct answers during the Cognitive Reflection Test. In the current study, we use salivary measures of testosterone because of its unintru- sive nature of sample collection and the superior prac- tical applicability of the results. Our measure is dynamic because an individual ’ s hormones slightly change throughout the day and can drastically change over the decades. In contrast, measures of in-utero exposure are static and do not change after birth. Cortisol and financial decision-making Cortisol receptors are located throughout the brain; however, they are significantly concentrated in areas linked to irrational decision-making, including the frontal cortex, amygdala, and hippocampus (McEwen and Sapolsky 1995; Sapolsky 1996; de Quervain et al. 1998; Vedhara et al. 2000; MacLullich et al. 2005; Lupien et al. 2009; Margittai et al. 2016). Specifically, cortisol exerts a stimulating role in the aforemen- tioned brain areas. For example, higher cortisol levels are related to increased amygdala activation (Van Stegeren et al. 2007), and it is positively correlated with the anatomical volume of the amygdala and hippocampus (Pagliaccio et al. 2014). Moreover, research shows that increased cortisol levels are asso- ciated with avoidance tendencies, which is likely due to its impact on the amygdala (Roelofs, Elzinga, and Rotteveel 2005; Van Peer et al. 2007). Overall, the role of cortisol on cognitive function, judgment error, and emotional response strongly suggests that it asserts an important role during investment decisions and investment biases. Few studies have examined the relationship between cortisol and financial decision-making. Kandasamy et al. (2014) find that exogenous cortisol is related to greater risk aversion when participants are provided with alternative lottery gambles. Coates and Herbert (2008) find that salivary cortisol levels increase with market volatility in a sample of 17 male floor traders, suggesting that as market volatility increases, traders become more stressed. Nofsinger, Patterson, and Shank (2018) conclude that salivary cortisol levels are related to lower amounts of port- folio diversification and portfolio returns, which is consistent with risk-averse behavior. Given the afore- mentioned research on cortisol and risk aversion and avoidance behavior, we posit our third hypothesis: H3: Investors with higher cortisol levels will be more likely to commit the disposition effect. Cueva et al. (2015) use a double auction market to examine the impact of cortisol and testosterone on market bubbles and instability, finding that both corti- sol and testosterone increase trading activity that cre- ates market instability. Therefore, we postulate our final hypothesis: H4: Investors with higher cortisol levels will trade excessively. Dual-Hormone hypothesis and financial decision-making Few studies have examined the dual-hormone hypoth- esis during financial decision-making. For example, Nofsinger, Patterson, and Shank (2018) find that indi- viduals who have a higher testosterone to cortisol ratio are more likely to sell poorly performing stocks, which suggests that this ratio is related to the disposition effect. Mehta et al. (2015) use the Balloon Analog Risk Task to measure economic risk-taking and find that testosterone is positively related to increased risk-tak- ing, but only when cortisol levels are low. Therefore, it is essential to examine testosterone in tandem with cor- tisol in the context of investment biases. 3 Method Sample We recruit 41 students (28 men and 13 women) from a Master of Science in Finance (MSF) program who were enrolled in the program ’ s Financial Software 340 J. R. NOFSINGER ET AL. course following Institutional Review Board (IRB) approval. 4 This specialty course is taken toward the end of the program after students have completed core courses, including Securities Analysis, Portfolio Management, Financial Futures, and Fixed Income Investments, and Financial Risk Management. Thus, participants possess superior knowledge of finance compared to average individuals. The Financial Software course heavily emphasizes the use of the Rotman Interactive Trader 2.0 (RIT 2.0), a financial trading simulation application developed at the BMO Financial Group Finance Research and Trading Lab at the University of Toronto. We employ the RIT 2.0 simulation software to collect data pertaining to finan- cial choices and outcomes during a series of invest- ment tasks. The data used in this study is collected toward the end of the course to allow participants enough time to master the use of RIT 2.0. Participation incentives include the potential to earn bonus points toward the course final grade and a monetary reward. The exact nature of incentives is revealed just prior to data collection to maximize anticipatory arousal, stress, and the competitive nature of professional trading. Specifically, participants are informed that their overall risk-adjusted performance during the investment simulations could influence their final course grade by up to one-third of a letter grade (e.g., from A- to A). Additionally, the top three risk-adjusted performers (measured via the Sharpe ratio) receive monetary rewards of $75, $50, and $25, respectively, in the form of gift cards. James and Isaac (2000) argue that it is problematic to use tournament incentives. However, in this study, we are concerned with the psychophysiological underpinnings of finan- cial decision-making in individuals, and not with the effect of participation incentives on overall market performance. Furthermore, Kempf and Ruenzi (2008) show that tournament incentives are used widely in the financial industry, and it is the aim of our study to simulate real-life conditions associated with invest- ment decisions as much as possible. Data Physiological (i.e., testosterone and cortisol levels) and financial data are collected toward the end of the course, early in the morning. This time of day is chosen to better capture the circadian nature of testos- terone and cortisol levels, which are highest during the morning hours and gradually decline throughout the day. Additionally, this time of day is when trading hours of major U.S. stock exchanges begin, and the average daily trading volume and market volatility tend to be high (Foster and Viswanathan, 1993). After students arrive at the facility, they are given 2 salivary test tubes (for before and after comparison), from which the levels of salivary testosterone and cor- tisol are measured via mass spectrometry, which is the standard method of measuring salivary hormone levels in the medical field. In collecting the salivary samples, we follow the recommendations of Granger et al. (2004) regarding the validity of salivary measures, including the use of Salivette test tubes and sugar-free chewing gum. 5 Overall, the aforementioned steps are taken to ensure maximum comparability with real-life investor behavior and consistency with related studies, such as Coates and Herbert (2008). For further details on the saliva collection instructions, see Appendix A. As this study investigates how hormones impact investment biases, it is appropriate to use the data from the first saliva test. The post-trial sample is appropriate for examining how investment perform- ance impacts hormone levels. In our sample, the pre- trial average testosterone level is 30.16 pg/mL (SD ¼ 17.2) for men and 5.29 pg/mL (SD ¼ 8.01) for women, while the mean pretrial cortisol level for the entire sample is 5.57 nmol/L (SD ¼ 3.34). 6 We con- vert raw testosterone levels to z-scores based on the sample gender distribution as in other endocrine stud- ies using a mixed gender sample (e.g., Mehta, Jones, and Josephs 2008; Nofsinger, Patterson, and Shank 2018). Cortisol levels do not exhibit gender differen- ces, and as such, raw levels do not need to be standar- dized by gender. In addition to exploring the individual effects of testosterone and cortisol on investment biases, we examine their joined effect as predicted by the dual-hormone hypothesis by means of the testosterone to cortisol ratio, or T/C ratio. 7 Immediately following the collection of the first sal- ivary sample, the participants engaged in RIT 2.0 investment simulations. 8 During each simulation, par- ticipants are asked to create a portfolio of assets and to rebalance it at the end of years 5, 10, and 15 of a simulated 20-year period. There is an initial endow- ment of $500,000 with the goal of achieving a port- folio value of $1,500,000 at the end of the 20-year period (an annualized return of 5.65%). They have five ETF ’ s to choose from with known historical return and volatility distributions, as shown in Appendix B. Funds not invested in an ETF are kept in a “ CASH ” account with zero return and volatility. Additionally, participants know whether the correl- ation between ETF ’ s is high, medium, or low. Each ETF evolves as a random walk with positive drift JOURNAL OF BEHAVIORAL FINANCE 341 based on its historical return and volatility. Therefore, the price path followed during a given period has no impact on future price paths, similar to the real stock market. Based on the chosen portfolio adjustments during each 5-year period, we measure three of the best studied investment biases for each participant, as sum- marized in Cronqvist and Siegel (2014): the dispos- ition effect, performance chasing, and portfolio turnover. The disposition effect is measured as the difference between the ratio of the realized to unreal- ized gains and the ratio of the realized to unrealized losses, as in Odean (1998), Dhar and Zhu (2006), and Cronqvist and Siegel (2014). Performance chasing is defined as the net cash flow difference in the amount bought and sold of the asset with the highest return during the previous 5-year period, similar to Cronqvist and Siegel (2014). Finally, turnover is defined as the sales volume traded divided by the value of the portfolio, as in Barber and Odean (1999, 2000, 2001) and Cronqvist and Siegel (2014). As with any research, there are always limitations to what hypothesis we can test. We would like to explore as many common investment biases as possible. However, we are limited by the nature of the experi- ment and, in some cases, a lack of theory leading to testable hypotheses. For example, one common invest- ment bias in the literature is investors preference for skewness. However, since the RIT uses diversified ETFs whose returns are normally distributed based on the mean and standard deviation, it is not possible to meas- ure investor skewness preference. Additionally, we do have the data to examine investors preference for domestic stocks. However, there is a lack of evidence in the literature to support a hypothesis on why hormones would impact a locality preference. Moreover, this study was conducted with a large portion of international stu- dents, which makes it difficult to determine which domestic ETF they would be more likely to purchase. For example, we would have to make a judgment call on whether foreign exchange students from China would consider the U.S. domestic equity or the emerg- ing market equity as their “ home ” asset. As such, we elected to study these three biases because we found suf- ficient evidence in the literature to support our hypoth- esis, and they are testable with our experimental design. Results Testosterone, cortisol, and investment biases Table 1 presents the sample summary statistics, start- ing with the number of male and female participants in Panel A. Panel B shows that most participants are young professionals in their mid-twenties, as expected in a sample of graduate business students. Overall, partici- pants indicate a very high level of comfort with the RIT 2.0 platform, and they report very high levels of pretrial preparation. However, participants have a low level of professional experience in investments. Therefore, the sample is most representative of naïve investors with sig- nificant academic knowledge in finance. Panel C of Table 1 displays sample scores on investment biases. On average, subjects show a posi- tive disposition effect, which is consistent with a pro- pensity to sell more winning ETFs than losing ETFs. The negative performance chasing score indicates that subjects sold the best performing asset more often than they chased the highest performing ETF. Finally, the subjects had an average portfolio turnover of 17%. Table 2 presents the results of OLS regressions in which behavioral biases (i.e., disposition effect, per- formance chasing, and turnover) are regressed on the independent physiological variables of interest (Testosterone z-score and Cortisol) and several control variables (i.e., pretrial preparation, comfort level with RIT 2.0, professional trading/investment experience, gender, and age). 9 We examine the hormones separ- ately with the impact of testosterone alone in model 1, cortisol alone in model 2, and testosterone, cortisol, and the control variables jointly in model 3. The effects of hormones on the disposition effect are pre- sented in Panel A. We find that cortisol has a positive relationship to the disposition effect. This result is consistent with the various studies showing the role of cortisol in cognitive errors (McEwen and Sapolsky 1995; Sapolsky 1996; de Quervain et al. 1998; Vedhara et al. 2000; MacLullich et al. 2005; Lupien et al. 2009; Margittai et al. 2016). Additionally, the relationship between cortisol and the disposition effect may also be a result of the role of cortisol in the approach-avoid- ance neural system (Roelofs, Elzinga, and Rotteveel 2005; Van Peer et al. 2007). It has been widely docu- mented that the mammalian stress response, which is modulated by cortisol, is associated with prolonged freezing reactions and avoidance behavior, which may explain the lower propensity to realize losing ETFs. Furthermore, the results show that trading experience and age are both positively related to the disposition effect, suggesting that experience (in both trading and age) does not prevent investors from making irrational decisions. Panel B shows the results for the relation- ship between testosterone and cortisol and perform- ance chasing. However, we do not find results to 342 J. R. NOFSINGER ET AL. support our hypothesis of a relationship between tes- tosterone and performance chasing. Panel C presents the relationship between the hor- mones and portfolio turnover. The results show that testosterone is significantly and positively related to portfolio turnover, whereas cortisol is significantly and positively related to the disposition effect when looking at model 2. However, cortisol becomes insig- nificant once testosterone and the control variables are included. Additionally, we find that preparation and the male dummy variable are related to lower lev- els of portfolio turnover. Given that the expected returns of the ETFs don ’ t change based on the previ- ous 5-year simulation, better prepared participants may have less of a reaction to how the ETF has per- formed in the past. Barber and Odean (2001) show that men exhibit higher portfolio turnover than women. While we find that men exhibit a lower turn- over ratio, our results suggest that this relationship may be due to testosterone levels rather than gender itself as gender is modulated by testosterone, with men naturally exhibiting significantly higher levels of testosterone than women. Table 3 displays the relationship between the dual hormone hypothesis and investment biases. We find that the effects of testosterone on portfolio turnover is modulated by cortisol. When testosterone levels are high, relative to cortisol levels, there is a positive rela- tionship with portfolio turnover. That is, testosterone levels are only positively related to portfolio turnover when cortisol levels are low. These results are consist- ent with studies showing a significant relationship between the testosterone to cortisol ratio and a variety of risky social behaviors (Terburg, Morgan, and Van Honk 2009; Mehta and Prasad 2015). Our analysis so far assumes that our continuous hormone variables are the correct measure for pre- dicting behavior. However, there could be much noise in these variables or individuals with extremely low or high levels distorting the linear relationship. Table 1. Summary Statistics. Panel A: Gender Total Male Female 41 28 13 Panel B: Age, comfort, investment experience, and preparation Mean Median Std. dev. Age 27.8 25.65 6.52 Comfort with RIT 2.0 3.77 4.00 0.81 Investment experience 0.73 0.00 1.53 Preparation 3.82 4.00 0.72 Panel C: Summary statistics of investment biases Mean Std. Dev. Disposition effect 0.16 0.62 Performance chasing 0.01 0.12 Turnover 0.17 0.21 This table displays the subject sample statistics. Panel A shows the gender breakdown of the subjects. Panel B shows the subject ’ s age, number of years of investment experience preparation and comfort with RIT 2.0. Panel C shows the investment biases for the subjects. Panel D shows that summary statistics for the subjects ’ hormones. Table 2. Hormones and investment biases. Variables (1) (2) (3) Panel A: Disposition effect Testosterone 6.71 2.04 (1.334) (-0.349) Cortisol 8.87 14.72 (1.809) (2.387) Preparation 5.68 (0.740) Comfort 1.30 (0.199) Trading experience 10.40 (2.982) Male 22.92 (-1.927) Age 1.84 (2.037) Observations 156 156 156 R-squared 0.011 0.021 0.102 Panel B: Performance chasing Testosterone 0.97 0.23 ( 1.192) ( 0.240) Cortisol 1.10 1.02 ( 1.370) ( 0.986) Preparation 1.23 ( 0.963) Comfort 0.75 (0.688) Trading Experience 0.86 ( 1.458) Male 0.63 (0.320) Age 0.06 (0.422) Observations 246 246 246 R squared 0.006 0.008 0.026 Panel C: Portfolio Turnover Testosterone 4.20 4.17 (3.075) (2.621) Cortisol 2.58 0.34 (1.894) ( 0.200) Preparation 5.76 ( 2.729) Comfort 0.07 (0.037) Trading Experience 0.98 (1.000) Male 7.74 ( 2.380) Age 0.25 (0.990) Observations 246 246 246 R-squared 0.037 0.014 0.086 This table displays OLS regression coefficients for the relationship between hormones and investment biases with t-statistics in paren- thesis. The Disposition Effect is presented in Panel A, Performance Chasing in Panel B, and Portfolio Turnover in Panel C. Controls are sum- marized in Table 1. Male is a dummy variable equal to 1 for males and 0 otherwise. Significant results are displayed at the 10% ( ), 5% ( ), and 1% ( ) levels. JOURNAL OF BEHAVIORAL FINANCE 343 As a robustness test, we sort and group people by their hormone level measurement and examine their relationship to investment biases by categories. As such, we split the sample into three terciles based on each of their hormone levels individually. Results are presented in Table 4 with an illustration of the results displayed in Figure 1. We include a dummy variable that equals 1 if the subjects are in the top tercile for each hormone (highest amount of the hor- mone) to examine how individuals high in testoster- one and high in cortisol trade compared to those with lower levels. The results in Table 4, Panel A confirm our pre- vious findings that testosterone is positively related to portfolio turnover and that cortisol is positively related to the disposition effect. We also find that those who are the most stressed are more likely to have a higher portfolio turnover compared to indi- viduals with lower levels of cortisol, which is similar to model 2 of Table 2. This result is consistent with Cueva et al. (2015), who find that cortisol levels are positively related to trading activity during bubbles. However, we fail to find support for our hypothesis that testosterone will be positively related to per- formance chasing, as previous literature suggests that testosterone is positively related to lottery seek- ing investments (Lu and Teo 2018). We believe that this lack of results is due to the choices for which the subjects had to invest. Our sample of well diver- sified ETFs may not have ETFs that display enough skewness and lottery type incentives to produced results to where we would expect to see this type of behavior. Conclusion This study examines the individual and joint effects of testosterone and cortisol on investment biases using a sample of graduate finance students and realistic financial simulation software. The results show that hormonal activity is an important underlying element of investment biases. Specifically, testosterone and cor- tisol are both related to irrational financial behavior. For example, both testosterone and cortisol are posi- tively associated with portfolio turnover, and cortisol is positively related to the disposition effect. Overall, our results show that both cortisol and testosterone are related to judgment errors resulting in invest- ment biases. Our results have several important applications because they underscore the relevance of physiological Table 3. The dual hormone hypothesis and invest- ment biases. Variables Disposition effect Performance chasing Turnover Testosterone / Cortisol 4.50 0.09 4.66 ( 0.794) (0.098) (3.139) Preparation 2.21 0.99 5.19 (0.289) ( 0.790) ( 2.510) Comfort 4.53 1.18 0.28 ( 0.690) (1.101) (0.160) Trading Experience 9.24 0.76 0.52 (2.619) ( 1.295) (0.540) Male 24.92 0.71 6.33 ( 2.051) (0.355) ( 1.937) Age 1.61 0.09 0.04 (1.713) (0.583) (0.159) Observations 156 246 246 R-squared 0.067 0.019 0.092 The Dual Hormone Hypothesis and Investment Biases. This table displays OLS regression coefficients and t-statistics in paren- thesis. The Testosterone/Cortisol Ratio is the dependent variables of interest. Controls are summarized in Table 1. Male is a dummy variable equal to 1 for males and 0 otherwise. Significant results are displayed at the 10% ( ), 5% ( ), and 1% ( ) levels. Table 4. Hormones and investment biases by group. Variables Disposition effect Performance chasing Turnover Panel A: Testosterone and cortisol Testosterone top tercile 0.69 1.51 6.81 (0.063) ( 0.823) (2.260) Cortisol top tercile 26.96 0.03 7.33 (2.338) (0.013) (2.303) Preparation 5.72 1.02 4.28 (0.746) ( 0.791) ( 2.019) Comfort 0.77 1.00 0.66 (0.120) (0.924) (0.371) Trading experience 9.92 0.74 1.59 (2.896) ( 1.269) (1.669) Male 26.84 1.02 9.57 ( 2.166) (0.506) ( 2.902) Age 1.79 0.10 0.39 (1.999) (0.641) (1.607) Observations 156 246 246 R-squared 0.097 0.022 0.099 Variables Disposition Effect Performance Chasing Turnover Panel B: Dual-hormone Testosterone/cortisol top tercile 14.73 1.18 4.89 ( 1.257) (0.622) (1.541) Preparation 2.08 0.95 5.40 (0.274) ( 0.758) ( 2.573) Comfort 6.22 1.39 0.06 ( 0.919) (1.256) ( 0.030) Trading experience 8.72 0.76 1.17 (2.557) ( 1.330) (1.227) Age 25.87 0.86 6.96 ( 2.131) (0.434) ( 2.095) Male 1.70 0.07 0.18 (1.837) (0.434) (0.693) Observations 156 246 246 R-squared 0.072 0.021 0.064 This table displays OLS regression coefficients and t-statistics in paren- thesis. Testosterone and cortisol (Panel A), and the testosterone to corti- sol ratio or T/C Ratio (Panel B) are the dependent variables of interest. Controls are summarized in Table 1. Male is a dummy variable equal to 1 for males and 0 otherwise. Significant results are displayed at the 10% ( ), 5% ( ), and 1% ( ) levels. 344 J. R. NOFSINGER ET AL. processes during financial decision-making. From an academic standpoint, our attempts to better under- stand market behavior would be enhanced by the inclusion of physiological processes that affect investor behavior. As long as we fail to acknowledge realistic human behavior, asset pricing models will continue to break down during empirical tests. From a practical standpoint, market participants would benefit from understanding how their physiological makeup influ- ences their financial choices and resulting perform- ance. This is particularly important in the finance profession, which is highly male-dominated and among the most stressful. Based on our results, we strongly advocate the use of awareness, reflection, and monitoring of endocrine functions that may affect judgment during financial decision-making. There are several ways in which this study may be extended. First, it would be interesting to repeat the study addressing the weaknesses, including (1) partici- pants with significant investment experience, (2) real- life working conditions, (3) strong monetary incentives, (4) multiple hormonal sampling throughout the day, and (5) following participants during several days until proper endocrine baselines are established. Second, there is a multitude of hormones and chemicals that act on the brain areas associated with financial deci- sion-making. It would be interesting to determine the relative composition of these molecules during eco- nomic tasks and to better understand their joint role (i.e., a multi-hormone hypothesis). Finally, we recog- nize that financial decisions are frequently undertaken in a group setting. It would be interesting to examine the role of hormonal activity on financial decision- making when group dynamics come into play, and in a fast globalizing world, groups with diverse sociodemo- graphic profiles. Notes 1. The ratio between the length of the second and fourth fingers (2D:4D) is smaller in males exposed to higher prenatal (in-utero) testosterone. 2. The ratio between the bizygomatic width divided by the upper-face height is larger in individuals with higher exposure to testosterone in prenatal and postnatal. 3. For a greater discussion on the impact of neural activity or hormones on financial behavior see Nofsinger and Shank (2020). 4. Prior to conducting this study, Institutional Review Board (IRB) approval was obtained from the university ’ s Office of Research (Protocol Approval #IRB-13-003). 5. Salivary samples avoid the invasiveness of serum samples, preserving realistic conditions. However, a known disadvantage of salivary samples is the difficulty in comparing raw hormone levels across studies. Benchmarking is constrained by the specific procedure used to determine the hormone concentration. Therefore, we take a conservative approach when comparing our findings to similar studies. 6. Unlike testosterone, which is markedly different in men and women, cortisol levels are comparable in both sexes. 7. Nonlinear effects of testosterone and cortisol are not examined in this study due to multicollinearity concerns. 8. Saliva samples are sealed and refrigerated at -20oC immediately after collection. Salivary testosterone and cortisol levels are measured via mass spectrometric analysis at a University toxicology lab facility. 9. Investment outcomes were determined through random selection from each asset class return distribution. There were two simulated round outcomes in which no ETF experienced a loss. After those rounds, participants had no losers in their portfolios, thus were not subject to the disposition effect scenario. Therefore, the disposition effect only has data for decisions after four rounds. Performance chasing and turnover have data for all six rounds following initial investment simulations. Figure 1. Cortisol, testosterone and investment biases. Note: This figure presents a graphical representation of how low and high groups of testosterone and cortisol relate to the dispos- ition effect and portfolio turnover. JOURNAL OF BEHAVIORAL FINANCE 345 Acknowledgement We would like to thank the seminar participants at the 2018 Financial Management Association Conference, 2018 Midwest Finance Association Conference, Oklahoma State University, Washington State University at Vancouver, Eastern Michigan University, and Dalton State College for helpful comments. ORCID John R. Nofsinger http://orcid.org/0000-0003-0048-4458 Corey A. Shank http://orcid.org/0000-0002-1698-875X References Ahmed, S., J. Sihvonen, and S. V € ah € amaa. 2019. “ CEO Facial Masculinity and Bank Risk-Taking. ” Personality and Individual Differences 138:133 – 9. doi:10.1016/j.paid.2018. 09.029 Barber, B., and T. Odean. 1999. “ Do Investors Trade Too Much? ” American Economic Review 89 (5):1279 – 98. doi: 10.1257/aer.89.5.1279 Barber, B., and T. Odean. 2000. “ Trading is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors. ” The Journal of Finance 55 (2):773 – 806. doi:10.1111/0022-1082.00226 Barber, B., and T. Odean. 2001. “ Boys Will Be Boys: Gender, Overconfidence, and Common Stock Investment. ” The Quarterly Journal of Economics 116 (1): 261 – 92. doi:10.1162/003355301556400 Barber, B.,. Y. Lee, Y. Liu, and T. Odean. 2009. “ Just How Much Do Individual Investors Lose by Trading? ” Review of Financial Studies 22 (2):609 – 32. doi:10.1093/rfs/hhn046 Barberis, N., and R. Thaler. 2003. “ A Survey of Behavioral Finance. ” In Handbook of the Economics of Finance , vol- ume 1, part 2, edited by G. M. Constantinides, M. Harris, and R. Stulz, 1053 – 128. Amsterdam: Elsevier Science B.V. Barel, E., S. Shahrabani, and O. Tzischinsky. 2017. “ Sex Hormone/Cortisol Ratios Differentially Modulate Risk- Taking i