The risk of putting risk at the heart of banking

Citation

Cookson, R. (2000), "The risk of putting risk at the heart of banking", Balance Sheet, Vol. 8 No. 6. https://doi.org/10.1108/bs.2000.26508faf.002

The risk of putting risk at the heart of banking

Richard Cookson

A friend, recently retired from running the capital-markets operation of a big investment bank, at the grand old age of 37, puts it thus: "Value-at-risk models are a hocus-pocus of science and art". That, it must be said, was one of his more printable comments. Big banks have spent large fortunes trying to get a grip on the risks that they run, the better to control them. The evidence suggests, however, that the tools that banks and academics have developed to try and control the risks that they run in financial markets, might actually be making them riskier.

If this seems an observation of uncommon stupidity, consider the following: financial shocks seem now to happen with a regularity bordering on the monotonous. There was, first, the stockmarket crash of 1987; then the collapse of the exchange-rate mechanism in 1992-1993; the bond-market crash in 1994; the Mexican crisis later that year; East Asia's turmoil starting in 1997; Russia's default and the near-meltdown in the capital markets in 1998; and, most recently, the virtual collapse of the junk-bond market.

A litany that long, you might agree, suggests that something is going badly wrong. You might not, however, agree that that something is banks' risk-management systems. After all, these crises had myriad causes. Indeed they did, but the way in which they spread may well have had something in common. That something may have something to do with the way in which banks control their risks.

Why might this be so? It all starts with the famous option pricing model developed by Fischer Black, Myron Scholes and Robert Merton, in 1973. Although options are centuries old, pricing them was little more than guesswork. The three financial economists developed an algorithm to do the job. That, it transpires, was both a blessing and a curse.

The most important of the inputs into the algorithm is an estimate of how volatile an asset will be over the lifetime of the option. If he knows this, the option seller can create equal and offsetting positions in the option, thus protecting himself from loss, a process known as delta hedging. The first indications that this might be a problem came in 1987. An advisory firm, Leyland, O'Brien and Rubinstein, had twigged that rather than expensively buy options from other people, firms or investors could simply delta hedge themselves. Unfortunately, the huge numbers of fund managers that tried to sell shares to protect themselves in October 1987, found that the market was both less liquid and more volatile than they had assumed. What was worse, they further pushed the market against themselves.

In recent years, investors have bought proper insurance from banks (among others). Banks, so they claim, are more expert in hedging these exposures. But unless there are enough "natural sellers" of volatility ­ fund managers, say, who are happy to sell puts to gain a premium and buy an asset at a targeted rate ­ dynamic hedging always has the potential to turn a problem into a crisis. There are, however, few enough natural sellers of options in developed markets and almost none in developing ones.

And so to VAR models. These, too, rely on measuring volatility, and thus also rely on dynamic hedging. VAR models rely on two things: the volatility of individual assets, and the correlation between them. They are, in effect, an uncertain relationship between lots of uncertainties. Crucially, they have also, partly by dint of regulatory strictures, which force banks to tie the amount of capital to their VAR, become prescriptive rather than merely descriptive. That is, if their VAR climbs, banks must either put more capital aside or cut their risk positions (don't forget that banks are leveraged institutions).

And VAR models are hugely affected by changes in volatility. If markets become more volatile, VAR goes up by at least a proportional amount (and more if the bank is short options). So you can see why it is at least arguable why financial crises spread so widely and rapidly: VAR links markets together.

This does indeed seem to have happened in several recent crises. It is certainly what happened in the autumn of 1998. Faced with a VAR that was climbing rapidly, banks had to cut risky positions wherever they could. And as in 1987, the effect was to push volatility up and the markets against them. The effects were made much worse by the fact many investment banks had very similar positions. This should, perhaps, come as no great surprise: they all have the same Bloombergs and similar education. If things get out of kilter, then they put on the same sorts of trades. When a crisis erupts, however, they all have to head for the exit at the same time. Sound familiar?

In the aftermath of that particular crisis, bank bosses said that they would place far less reliance on VAR models, cut the amount of punting they did, and use a bit more common sense. But none of this really helps in the long run. Banks are, after all, under huge pressure to generate juicy returns for shareholders. If bog-standard markets, such as bonds and foreign exchange, thanks to advances in information technology, become much less profitable, from where else can banks earn a decent return? There is always M&A and advisory work, and leading equity issues, but this business is increasingly going to a handful of big firms, and it is unlikely to last if the stockmarket continues in its present jittery mood (and don't forget that Japan's bear market has lasted for more than a decade). So it is not unreasonable to suppose that many investment banks will have to continue to punt for their living.

Which brings us back to square one. Options, in effect, measure future uncertainty. The measure of that future uncertainty is volatility. But future volatility is, by definition, unknowable. Putting uncertainty at the centre of an algorithm and forcing banks to use it to measure their risks, is thus a recipe for trouble.

Richard Cookson is banking and markets editor for The Economist.