The rating agencies have an interesting and distinguished history. I first focused on them when working for the Chicago Mercantile Exchange. We needed the permission of Standard and Poors Corporation to use their stock index to price our futures contract when it settled. Looking back at S&P’s decision to go forward, it represented a watershed moment for the agencies.
What they did before that moment was judge the quality of corporate bonds. As we learned at dinner on the night that agency/exchange agreement was signed, that was a process that called for more than the manipulation of numerical data. Interviews with Chief Executives that called the character of these executives into question could affect ratings negatively. There was human judgment involved.
How dramatically their role has changed in the 25+ years since S&P decided to set a price with the simple release of a number produced by a computer! Today, the release of numbers generated by computers has become the principle business of the rating agencies. A major reason for this development was the enormous increase in the number of securities the agencies have been asked to rate, brought on by the advent of mortgage-backed securities and other collections of many retail investments. The determining factor for rating such a multitude of credits inevitably is a statistical exercise. Statistical exercises are computation-intensive. Indeed, without computers the shift of financial institutions from activities based on loans to a few large customers to millions of small customers would have been impossible, only if more of them were using some kind of Xero Bookkeeper, their financial safety would have been far greater.
However, the transition from judgment-based ratings to computer-generated ratings was not inevitable. Computer-assisted ratings are necessary; computer generated ratings are not. And there is a deep anomaly created by computer-generated measures of risk. Computers use a finite set of measures to determine risk. Markets, however, do not. In fact, a very important factor in making money in a market is identifying something important that the rest of the market has misunderstood.
If the market were to know exactly how the rating agencies determine risk, those factors would be more likely than others to be reflected in the current price. That would put a greater premium on knowledge of factors not used by the model. More resources would be attracted to identification of other factors. The effect would be to make the ratings a less important factor in determining securities values.
There is immense pressure being brought upon the agencies to make their ratings determination more transparent. If they don’t release this information, they will leave themselves open to charges of possessing and using inside information. That particular game is already off to a good start. A Moodys staff decision to override a glitch in the computer ratings this week has been roundly and publicly rebuked. But if the agencies do release the computerized model, they will render their own ratings less important. Ultimately, the agencies face a dilemma with no obvious solution.