Monday, October 6, 2014

Crazy is not a hypothesis

One of the criticisms I sometimes hear of behavioral finance, mostly from the rational crowd, is that one is just showing that 'people are crazy' or 'people are stupid'. This is always said dismissively, as if such an observation were trivially true and thus unworthy of observation or elaboration.

The first indication that this is a vastly overblown criticism is given by the fact that, despite the claimed triviality and obviousness of people's stupidity and craziness, these traits don't seem to find their way into that many models - the agents in those models are all rational, you see.

Well, actually, it's a bit subtler than that. Stupid agents have actually been in models for quite a while now, most notably in models that include noise traders, trading on false beliefs or for wholly idiosyncratic reasons.

But agents who could be described as 'crazy' are harder to find - acting in completely counterproductive or irrational ways given a set of preferences and information. So why is that?

The reason, ultimately, is that 'crazy' is usually not a useful hypothesis. It's a blanket name given to a set of behaviors that falls outside of what could be considered rational behavior, or even partially rational (such as kludgy rules of thumb or naive reinforcement learning).

And the reason you know that crazy isn't a useful hypothesis is that it tells you very little about how someone will act, other than to specify what they won't do. How would you go about modeling the behavior of someone who was truly crazy? Maybe you could say they act at random (in which case things look like the noise traders that we labelled as stupid). But are you really sure that their behavior is random? How sure are you that it's not actually predictable in ways you haven't figured out? It seems pretty unlikely that there are large fractions of traders who are in a bona fide need of institutionalisation in a sanitorium, if for no other reason than someone who was really bonkers would (hopefully) struggle to get a job at the JP Morgan trading desk or acquire enough millions of dollars to move financial markets.

The whole point of behavioral economics (and abnormal psychology before it) is to figure out how people are crazy. When someone is doing something you don't understand, you can either view it as mysterious and just say that they went mad, or you can try to figure out what's driving the behavior. But madness is an abdication of explanation.

Good psychiatry reduces the mystery of madness to specific pathologies - bipolar disorder, psychopathy, depression, autism, what have you. 'Madness' functions as the residual claimant, thankfully getting smaller each year.

Good behavioral finance ultimately strives at similar ends - maybe people are overconfident, maybe they use mental accounting, maybe they exhibit the disposition effect. These are things we can model. These things we can understand, and finally cleave the positive from the normative - if rational finance is a great description of what people should do but a lousy description of what they do do, then let's also try to figure out what people are actually doing, while still preaching the lessons we formulated from the rational models.

To say that behavioral finance is just 'people acting crazy' is somewhat like saying that all of economics can be reduced to the statement 'people respond to incentives'.  In a trivial sense, it may not be far from the truth. But that statement alone doesn't tell you very much about what to expect, as the whole science is understanding the how and the why of incentives in different situations - all the hard work is still to be done, in other words.

It's also worth remembering this in real life situations - when someone you know seems to be acting crazily, it's possible they have an unusual form of mental illness as yet unknown to you, but it's also possible that you simply have inadequate models of their preferences and decision-making processes. Usually, I'd bet on the latter.

No comments:

Post a Comment