For a couple of years now I’ve had a bad feeling about the field of finance. Though I often inveighed against the mechanical use of models, I didn’t damn the entire endeavor. I still don’t, but the other day, talking to a colleague who was much younger than me, I discovered that we had somewhat similar sentiments. Mine are summarized as follows. I hope it’s just a mood.
Neoclassical finance, the entire structure that includes the efficient market model, Brownian motion, stochastic calculus, the capital asset pricing model, and of course Black-Scholes-Merton, is a beautiful comprehensive far-reaching discipline. It describes a risky system and its consequences elegantly and sparely, and provides lots of insight. I like it. Unfortunately, the real world is more complicated.
One of the responses to that complexity is behavioral finance. This started out with prospect theory as an attempt to provide a comprehensive new foundation for finance and economics. It hasn’t turned out that way. That part of behavioral finance is by and large ignored. Instead, its proponents have streamed off into two directions. The first is neuro/psycho finance, filled with papers discussing brain scans of traders and their risk appetite, which isn’t really finance at all, but belongs (if it belongs anywhere outside of grant applications) in a psychology department. The second is the study of how people react to choice, a collection of mildly interesting statistical studies of CEOs and investors that mostly confirm what everyone knew before, and whose results are most useful in justifying various paternalistic social policies. None of it adds up to a comprehensive discipline; it seems to be mostly strategies for writing papers.
On the sidelines, people with an interest in computational finance work away at more rapid methods of computing the results of various models; they don’t care if it’s right or wrong, they just want it to be accurate. There’s nothing wrong with that and I’ve done it myself, but it’s not central.
There are many things worth doing — studying price formation and the behavior of underlyers, to give just one example. I hold, perhaps sentimentally as a result of my background, some hope for the econophysicists and the builders of agent-based models. They’re most likely not going to find the absolute truth with these cellular automata — these are still models, not theories — but at least they seem to understand what they are doing, and why.
C’est la vie