Monday, August 03, 2009

In Defense of Financial Theories

I recently read a ridiculous critique of Value at Risk (VaR) by Pablo Triana in BusinessWeek (“The Risk Mirage at Goldman,” Aug 10, 2009). His review of this advanced financial technique is scathing:
VaR-based analysis of any firm's riskiness is useless. VaR lies. Big time. As a predictor of risk, it's an impostor. It should be consigned to the dustbin. Firms should stop reporting it. Analysts and regulators should stop using it.
Mr Triana bases his assertion on the observation that VaR is “a mathematical tool that simply reflects what happened to a portfolio of assets during a certain past period,” and that “the person supplying the data to the model can essentially select any dates.” My response to his argument is simply to ask, “Isn’t that true of any model or theory…?” Mr Triana goes on to argue that:
VaR models also tend to plug in weird assumptions that typically deliver unrealistically low risk numbers: the assumption, for instance, that markets follow a normal probability distribution, thus ruling out extreme events. Or that diversification in the portfolio will offset risk exposure.
In essence, Mr Triana seems to be saying that normally distributed results have bounds, and that portfolio diversification does not offset risk. Neither of his assertions are supported by probability theory or the empirical evidence. Yet, Mr Triana goes on to conclude, “it’s time to give up analytics so that real risk can be revealed.”

Mr Triana does a disservice to the financial services industry and public at large with his dramatic commentary. Yes, the discipline of finance has much to learn from the ongoing economic crisis, and of course, financial theory in general will evolve based on these recent lessons. However, just because one gets a bad meal in one restaurant does not mean that one should quit going to restaurants.

Financial theories such as VaR stand as state-of-the-art tools in the business of finance and risk management. These techniques are grounded in the same stochastic methodologies that are used by engineers in virtually every industry. To dismiss VaR so completely without considering its utility for supporting effective financial decisions is tantamount to sending financial theory back to the dark ages. Our knowledge of finance needs to advance as a result of what is happening in the economy, not go backwards.


Mihai said...

There's a HUGE difference between usage of stochastic techniques in engineering and usage of same in economic / financial models.
It has been proven by a vast body of experience that many engineering problems CAN be modelled by systems of equations that are (many times) linear and of low order (major exceptions may be found in complex systems with chaotic behaviour).
On top of the fundamental linear systems simple stochastic models do the estimation job quite nicely for engineering (physical) problems.

Can financial models (for VaR or other measures) claim similar linear simplicity and same proven reliability? One would think not...
There is just too much poorly understood interdependency between too many variables .

Leonid said...

What seems to be the case is that the assumption of normal distribution of portfolio returns in VaR analysis is indeed an over-simplification. It might still be useful as an initial approximation, but in general, the actual distributions constructed by means of the Forward Kolmogorov (Fokker-Planck) equation would be much preferred. This is what can be considered a modern approach to VaR analysis.

Jonathan Jacobs said...

If you estimate a 95% VAR, your results should be worse than that with 1-in-20 probability. That doesn't eliminate risk, it measures it. This year negative excursions probably generally exceeded the 95% VAR, in the biggest market meltdown since 1987 (over 20 years ago). If negative outcomes had otherwise not exceed the 95% VAR since then, I'd say linear VAR worked better than anyone had a right to expect. The problem is not in the models but the modelers, who appear to had convinced people that to quantify risk is to eliminate it.

Post a Comment