John Dizard of the Financial Times looks at the role of financial models in the recent market perturbations and argues that most observers are making the problem seem more complicated than it actually is. First, some firms had systems that were designed unduly for the convenience of the traders, and it proved way too easy to game them. Second, models are serving as scapegoats, obscuring the role of “lazy greed” (see the real causes of the poor performance at Goldman’s flagship, Global Alpha, for an illustration).
Dizard is right to point to the dangers of letting traders have undue influence on how their positions are marked, which is what some risk control systems unwittingly enable. Indeed, controlling traders has always been a basic problem for the securities industry. Readers of Richard Bookstaber’s “A Demon of Our Own Design” may recall his discussion of some of the types of trader chicanery that he uncovered.
From the Financial Times:
Even if the National Rifle Association says it, it is still true: Guns do not kill people, people kill people. And financial models do not destroy value. For that you can blame greed, laziness and stupidity on the part of the people who put the wrong data or the wrong parameters in their financial analytics. Drive a car without brakes and, guess what, you are going to get into an accident. Allow a trader to put on positions, using a self-serving pricing model that does not match what the back office or the risk managers have, and you will also have an accident. Or, rather, a predictable disaster.
There are some fundamental limits on the probabilistic models used to value securities, and derivative securities in particular. The models assume liquid, continuous markets. Illiquidity, or systemic freezes during which market participants cannot get bids for securities (or offers, for those in a short squeeze), would appear to be hard to model. And yet, illiquidity happens.
That was not, however, a secret before the summer crunch. And even with the known limits to existing probabilistic models, there was a lot that could have been done that was not done by the buy and sell sides in the credit markets.
Back in July, I had an interesting conversation with the people at Numerix, a leading provider of pricing and risk analytics in New York. Numerix is active in the forex, equities, inflation derivatives, and hybrid markets, but it is particularly dominant in producing the key software for valuation and risk modelling in the credit markets. I had wondered what use the Federal Reserve and other regulators had made of their software in monitoring what was then a developing crisis. The answer: none. The Feds just did not subscribe to the service. They still do not, by the way. And they wonder why they did not see the crunch coming.
I went back to Numerix to get their after-action report. Steve O’Hanlon, Numerix’s president, told me: “In my opinion [credit market people] did two things wrong. One, if you are going to manage positions in hard to value securities, you have to do it consistently across trading, operations, and risk management. They weren’t doing that. Two, they were making stupid bets because of lazy greed.”
The second problem is a universal and permanent aspect of human nature, which can only be countered with self-control and a strong institutional culture of honesty and risk recognition. The first one is more technical, and can actually be fixed. For example, as Mr O’Hanlon points out: “Traders like to use Excel-based models, because Excel is so flexible. But from an operations point of view, Excel is the worst, because it is so free form. For operations, you want to lock everything down,” so accounting and management can be consistent. “For risk control, you want to use the same model configuration from one day to the next, so when you do profit and loss, and mark to market, you don’t get drift because [traders] are setting up their models differently every day.”
If the trading models are not integrated with the risk models, you can be certain that the traders will game the risk management models to do things they should not. They will do that either for the desks’ P&L ( and their bonuses), or for favoured customers…
Over the next several weeks, we are going to have several entertaining public excoriations of the rating agencies and other Wall Street institutions. The rating agencies have proven they do not deserve their monopoly privileges. And if any industry deserves to be downsized it is Wall Street. There have been too many incompetents and people consumed by that “lazy greed”.
It will not, however, be possible to go back to the old system of banks and insurers directly allocating credit through committee systems. And yes, analytics need to be tweaked, and their parameters examined more carefully by the designers and users. A bigger issue, though, is the defective culture of many of the firms using them. The main problem in the recent crisis was not the complexity of credit analytics, but the intentional obscuring of who benefited from the mispricings of risk.