I attended an interesting talk today – John Coates from Cambride (a former Wall Street trader and trading desk manager) presented his recent work on naturally produced steroids and financial market instability (abstract here).
Here is what I took away from the talk:
– As Greenspan famously noted, the behaviour of real people in markets involves a lot more seemingly irrational sentiment than economic theory accounts for. Hence, our models are incomplete.
– In the neuroscience of decision making, people know about things like the winner effect: the phenomenon whereby two animals participating in a duel go away with altered levels of testosterone (winner has more of it and loser less) so that subsequent encounters are altered in a positive feedback cycle.
– This works well in small regions but a large enough growth in testosterone levels makes this winning animal over-confident. In nature, this means the animal takes risks, leaves its young ones exposed and forages recklessly, hence eventually dying.
Dr. Coates asks, do traders do the same things in financial markets? He presented some evidence to suggest that they do – that testosterone is correlated with performance (loosely, risk levels) and cortisol seems to measure external risk in terms of price volatility, etc. He bases these claims on salivary samples obtained from real life noise traders.
It was interesting to hear this. I am not entirely convinced about details – e.g., how much potential was there for data snooping? how much confidence can we place on an 8-day trial (even though it involved 400 traders), etc. In any case, I learned something new and the speaker may well be onto something useful. I also found various tidbits interesting – most traders are young males (highly testosterone dependent) and maybe if there were more women or older men (less susceptible to the above effect) then ensemble behavior might be different. Dr. Coates didn’t actually substantiate such claims but these are intriguing suggestions.
I was somewhat disappointed that one potentially very interesting point wasn’t really adressed – whether we could also approach these issues in a computational setting where we have more leverage over experiments. He motivated his talk with the statement that traditional economic theory, and even behavioral economics, does not really provide a satisfactory explanation of things like bubbles and crashes. So, I thought he is setting out to do that – but, in fact, he is quite far from that and based on the way he answered questions it is unclear if he is even attempting that. Nonetheless, it seems like a great direction – a simple start, without getting into the messy neurobiology, would be to define agents in an artificial market who have these testosterone/cortisol responses and then see if bubbles/crashes and other phenomena emerge. That would say something cool!