Does modelling fail fund managers (and others)?
Just seen an article on the Ludwig van Mises website, about Fooled by Randomness by Nassim Taleb. This is the classic contrarian 'anti-statistics' argument, this time made by a financial trader.
Quote from Taleb:
"What has gone wrong with the development of economics as a science? Answer: There was a bunch of intelligent people who felt compelled to use mathematics just to tell themselves that they were rigorous in their thinking, that theirs was a science. Someone in a great rush decided to introduce mathematical modeling techniques (culprits: Leon Walras, Gerard Debreu, Paul Samuelson) without considering the fact that either the class of mathematics they were using was too restrictive for the class of problems they were dealing with, or that perhaps they should be aware that the precision of the language of mathematics could lead people to believe that they had solutions when in fact they had none…. Indeed the mathematics they dealt with did not work in the real world, possibly because we needed richer classes of processes — and they refused to accept the fact that no mathematics at all was probably better."
See my earlier reference to Wigner: scientists have htese doubts, as well as economists and financial traders.
The problem is that sometimes the assumptions do work. Mathematical models and simulations can be used with amazing effectiveness to predict reality, to help us modify or improve it, and so on. At the moment, this is mostly in fields like engineering where the 'physical laws' are well known and the effects can be readily quantified.
Interestingly Taleb's web site has a link to an article he wrote with Benoit Mandelbrot, in which he argues that if we only used fractal or 'power law' statistics, we'd understand financial markets better.
You could say it's a bit like an arms race.
Step one: we develop modelling techniques that will help us to understand a limited range of problems.
Step two: they work well with some of the problems, and we derive benefit from them.
Step three: we apply them more widely. In many cases, we apply them to problems which are so complex that we don't really know if they work or not, but they sound good. (Lots of equations etc.) Sometimes the problem is time-scale: we make predictions, but we don't always hang around to see if they came true or not.
Step four: a cynic points out cases where these techniques aren't working, or probably aren't working, and adds social ('metamodelling') comment on the way that this has become an industry that employs a lot of people - whether it is fund management or EBO modelling. So any chance of objective assessment of the results drowns in career-building or marketing hype. The model is anyway so complex that no-one is really qualfied (or can spare the time) to critique it, except the person who built it and has an interest in selling it.
Step five: back to step one, for a further iteration with better techniques. At certain stages, of course, genuine 'step five' progress may be difficult to distinguish from 'step four' claims. But sooner or later something new emerges that is a genuine benefit to society, and off we go again.
ADD YOUR COMMENT:
COMMENTS SO FAR:
