Model Failure – By Hugh Young

Published in the South China Morning Post, 28 September 2008

While most industries have made advances over the last fifty years, it seems to me that much of the financial sector has gone backwards. Agriculture, technology and pharmaceuticals have brought huge improvements to our daily lives. The financial industry’s main innovation in recent decades has been the Value-at-Risk model (VAR).

VAR will mean little to non-specialists but it has been at the center of the financial turmoil over the last twelve months. Broadly, VAR is a way of measuring the market risk of a portfolio. This is the maximum loss that may occur given a certain confidence level, over a given period of time. Behind the formulation are some complex equations, but the key point is that in VAR asset price returns are treated as being normally distributed, a pattern that looks like a so-called bell curve.

Unfortunately asset prices are not normally distributed and the extreme risks or ‘fat tails’ on the bell’s everted rim are more common than might be imagined. The practical upshot of this has been evident for some time: as asset prices rose and volatility fell in the years leading up to the current crisis, the models signaled that risk was falling, and thus more of it could be taken.

Call me old fashioned, but I’ve always become more wary when asset prices go up a lot, particularly if volatility falls at the same time. But the VAR boffins didn’t see things that way – and they never have. The use of the bell curve to describe asset prices has got us into trouble before: it was, through the Black and Scholes options pricing model, at the heart of the 1987 market crash known as Black Monday and also the cause of Long Term Capital Management’s demise.

The wonder is it has happened again. In a call to fund investors last year, Goldman Sachs’ CFO said, “We are seeing 25 standard deviation events, several days in a row.” Or in plain English, events were taking place that should only happen every several trillion years! Since this is absurd, because it did happen, the more prosaic explanation is that the model was flawed.

That is a lesson Bank of America’s chief executive Kenneth Lewis might also bear in mind following its takeover of Merrill Lynch last week. He said that the financial industry would be pondering the question of how it got so deeply into such a mess for many years to come. Ken, the computer models would be a good place to start.

Why is VAR so beguiling? I see two forces at work. The study of asset prices has always attracted mathematicians seeking to quantify patterns that might unlock financial advantage – and peer group recognition. Nobel prizes were bestowed on the early pioneers of modern portfolio theory, whose flagship, the capital asset pricing model, assumed that asset prices followed a random (bell curve) walk.

It’s also a human tendency to overcomplicate matters that are simple and simplify, through the use of models, things that are naturally complicated. In this, the rest of us unwittingly conspire in the Nobel quest. Simplification appeals to our desire to make things neat and tidy. But simplification has its ugly side: the desire to cut corners, to get ahead… and can cause serious harm if in the process we sweep away many of the subtle, but important, features of financial market behaviour.

So, despite overwhelming evidence that asset price movements do not follow a random walk and that we don’t behave rationally, and despite the later renunciation of efficient market hypothesis by its original proponent (which is part and parcel of modern portfolio theory), the boffins still have the upper hand.

As usual, the most trenchant comment on efficient markets comes from Warren Buffett who quipped: “If you’re in the shipping business, it’s helpful to have all your potential competitors be taught that the earth is flat.”

At some point, however, it was in everyone’s interests to know that the world was round. This new understanding enabled trade routes to become shorter and safer, cartography to enter a new age, and the prices of many goods shipped by sea to fall.

Alas, in the financial world there has been no such equivalent development. The current crisis may cause us to appreciate that we have been both stupid and greedy, but these are hardly revelations, and certainly not profound ones. More importantly, greed and stupidity are characteristics that we will never eradicate as they are innately human. If anything, in this world of instant gratification, such traits have become all the more dominant – only the muscular world of finance likes to dress them up in different forms, as ‘drive’, ‘ambition’ and so on.

I don’t wish to moralise. But when a great deal of capital is destroyed, as it has been lately (some of it real, much of it illusory), someone has to be accountable. And especially as the earlier rewards for ‘success’ were so tangible. More blame will surely be put on models. But to my mind that would be the most unacceptable excuse of all. If the industry wants to progress, it should start with candour.