04 March 2013

FT: Get technical to fix rating agencies

US legislation requires that, before a rating agency can achieve the desired Nationally Recognised Statistical Rating Organisation (NRSRO) status, it must demonstrate a three-year operating record with paying customers. In short, they must show that for three years they managed to survive selling unofficial ratings. Author Arturo Cifuentes says he cannot think of any other market with more stringent entry barriers.

Another uncomfortable fact: the three leading rating agencies rate structured products such as collateralised debt obligations using, allegedly, different methods and proprietary data. However, a casual inspection of their ratings’ records shows a remarkable consistency.

The reputation of the agencies was left in tatters after the subprime crisis. Regulators did not fare any better. After all, they were supposed to oversee the rating agencies and failed miserably at this task. With that as background, it is not surprising that now regulators are trying to appear proactive. But their limited understanding of the technical aspects of this market have forced them to focus on qualitative issues, while missing the more important quantitative issues.

They keep talking about conflicts of interest within the agencies, the problems with the current issuer-pays rating model, and the consequences of deteriorating mortgage-underwriting practices. However, even if these issues were successfully addressed, the credit ratings edifice still has much more fundamental problems.

Fresh thinking is needed to improve the ratings environment. First, a new set of more robust credit risk metrics should be adopted to issue structured product ratings (practitioners and academics have made important progress in this topic). Ratings based on more stable metrics will be less likely to trigger systemic instability.

Second, corporate bonds, municipal and sovereign bonds, and structured products are completely different instruments. Accordingly, their ratings should be based on a different credit risk framework (and perhaps even the rating scale and symbols should be different).

Third, regulators should define, objectively, a credit risk scale. The rating agencies should only be allowed to state, based on whatever methods they see fit, the category in that scale to which an instrument belongs. This is a radical departure from the current situation in which the agencies control two things: first, the ratings scale (for instance, the precise definition of what triple A means), and second, whether a specific instrument meets that definition or not. Finally, regulators should create a centralised web-based data repository where any market participant could download for free all current and historic ratings data. This initiative would promote not only the development of better models, but will also facilitate testing them.

Full article (FT subscription required)

© Financial Times