Teaching finance the hard way

I was recently asked for my views on whether and how we should change the way we teach finance after all that we have seen in 2007 and 2008. Some of my thoughts are as follows.

• Quantitative models based on non normal fat tailed distributions with non linear dependence structures (copulas) are hard, but we must not shirk hard mathematics. It is much easier to talk about 2.33 standard deviations and simple correlations, but much of this is a delusion when applied to financial markets. It is even easier to adopt the viewpoint of some of Taleb’s followers that models are useless, but this is the path of nihilism. I think there is no place in finance teaching either for delusions or for nihilism. It is the creativity and subtlety of Mandelbrot that attracts me.
• Fat tailed distributions also required us to re-examine the use of historical data in financial modeling and simulation. Five or even ten years of historical data tell us very little about the true distribution if it is fat tailed. A lot of what happened during 2008 would have appeared ex ante impossible to anyone looking at several years or even a few decades of history. But one can see many parallels if one is prepared to go back several decades or a few centuries. Since we do not usually have high quality data going that far back, the implication is that historical simulation should be de-emphasized in favour of robust models that are qualitatively consistent with decades if not centuries of historical experience and extrapolate far beyond recent experience. I have long been fond of saying that one must approach the study of finance with Ito’s lemma in one hand and Kindleberger’s book in the other. Needless to say, the version of Ito’s lemma that I like to have in one hand is the one for semimartingales and not the one for Brownian motion alone!
• There is a need to shift from behavioural traits to hard nosed rational models. It is amazing but true that so much of what happened during 2007 and 2008 can be explained as the rational response of economic actors to altered fundamentals. I think that the crisis has taught us that finance is not a branch of psychology. For example, reliance on credit history (FICO scores) during credit appraisal assumes that default is a behavioural trait that can be measured using past track record. Rational models (Merton style models) assume that people default when it is rational to do so and focuses attention on modeling the fundamentals (for example home prices). Clearly lenders would have been much better relying on rational models rather than presumed behavioural traits. Unfortunately, during the lending boom, behavioural models held sway and these were supported by the short historical time series data that was then available.
• Much of modern finance deviated too far from its micro foundations in terms of well defined fundamentals. Derivative models allow us to compute implied volatility and implied correlations (and if necessary the entire implied risk neutral distribution) and start valuing anything without any regard to fundamentals at all. Models then become over calibrated to markets and under grounded in fundamentals. For example, quite often derivative textbooks and courses do not encourage us to ask questions like: what is the fair value of an option if we assume that the underlying is 10% overvalued in the marketplace. Just as finance is not a branch of psychology, it is not a branch of mathematics either.
• We need to teach more about the limits to arbitrage not in terms of behavioural finance, but in terms of well specified market micro structure with proper attention paid to transaction costs, leverage, and collateral requirements. The important stream of literature linking funding liquidity and market liquidity needs to be part of the core courses in financial markets.
• Perhaps we teach too much of ephemeral institutional detail. A lot of the details which we taught to our students during the last 3-5 years has been obsoleted by changes in the market structure. Investment banks are gone, the Libor market is barely recognizable and risk free government paper is no longer risk free. When we are preparing students for a career and not for their first job, we must emphasize functions and not institutions; concepts and not context.

In short, I believe finance teaching particularly in MBA courses during the early and mid 2000s became too soft and easy to cater to the needs of an ever growing body of students who sought a career in finance without any real aptitude for the subject. We dumbed finance down for the mass market. The time has come to go back to teaching finance the hard way – and perhaps there will be fewer students in the classroom.

Posted at 2:39 pm IST on Mon, 29 Dec 2008         permanent link

Categories: Bayesian probability, post crisis finance

Comments