This is an extract from our submission to the Agencies (Fed/FDIC/OCC and OTS) in respect of their deliberations into ratings alternatives and ideal ratings guidelines. This particular excerpt describes a feature we think is important but nevertheless overlooked — the decision as to whether a security is inherently ratable.
“Ratability”: If the Agencies are to rely on an external rating provider, we would recommend that each rating relied upon be accompanied by a signed letter that states that the provider believes the rated security is inherently “ratable” and that the provider has the systems in place to rate and monitor the rating of the security. The ratability condition should, in turn, be fulfilled in two parts:
- “Supportability”: Not only should the assumptions used be transparent (e.g., “a default rate of 3% was assumed”) but they should be supportable (e.g., “a default rate of 3% was assumed based on…”). The necessity to support each assumption made encourages the assessor to perform an additional level of due diligence before providing each assumption; and from a liability perspective it advantageously serves to heighten the likelihood that the assessor will revisit her assessment immediately a material change occurs in respect of the information or data supporting any assumption. Last, the need to support assumptions serves as a material impediment to rating agencies massaging data or creating “magic numbers” to inflate their ratings.
- Predictive Content: For a rating to be reliable and meaningful, it ought to hold some predictive content. Irrespective of the quality of a ratings model itself, the threshold to meet could be this: is there reason to believe that the outputs of the model adequately capture reality? For example, to the extent the accuracy of a rating depends on numerous assumptions — which are perhaps based on sparse, incomplete or unreliable data — and whose inter-dependencies themselves are imperfectly understood, we would consider the rating to hold little predictive content as the outputs of the model are as likely to hold true as they are to be false. This is not identical to the “monkey-throws-dart” problem, but essentially the test is whether the rater is sufficiently well positioned to make quantitative estimates as to future reality to command the fee charged for its service.
Posted on August 31, 2010
0