The Wall Street Journal recently published an article on the possible impact of testosterone on stock traders. The study is by John M. Coates, a senior research fellow at Cambridge. His general finding is that higher levels of testosterone are correlated with higher levels of risk taking and (during the course of the study anyhow) better trading results. Increased risk taking behavior seems intuitive; however, the one percentage point gain in financial performance is not as clear cut. One could dig into the original study to see how Coates tried to account for this but I do not see a clear way to determine if the trading decisions were good or bad ones based only on the outcome. Taking really risky bets can sometimes yield high returns. However, could those same returns be generated (on average) with more certainty and less risk? Would they have a higher risk adjusted rate of return?
The more interesting study mentioned is that of MIT Sloan’s Andrew Lo. He wired up traders to monitor their psycho-physiological state in real time while executing real trades. It seems he would have much tighter paring of the independent variable to the actual decision making moment. However, this study may suffer the same issues with the outcome measure. Some time ago Professor Lo presented his experimental design in a class I was taking at the MIT Media Lab (Sandy Pentland’s Digital Anthropology Class). I was an MBA student at the time and playing a lot of Texas Hold’em with buddies from the Muddy Charles -- so perhaps that was the inspiration but I suggested he run a similar experiment on Black Jack players. Unlike with the stock market, the expected value of given Black Jack hands can be precisely quantified. On a hand by hand basis you could determine if the subject made the decision that maximized his or her expected value from playing the hand. I would guess that a higher testosterone level and/or greater excited psycho-physiological state would cause subjects to make suboptimal decisions, even if on occasion they got lucky and won.