Dear Mr. Pollard:
Thank you for the opportunity to submit these comments on the Proposed Rule on Credit Score Models:

  1. In my opinion, the Proposed Rule overall is sensible and well-considered, and consistent with sound housing finance.
  2. Since credit scores are part of the analysis and management of credit risk, the principal decisions about their use should rest with those who take the credit risk—in this case, with Fannie Mae and Freddie Mac. The process as defined by the Proposed Rule thus puts the primary responsibility for analysis and decisions in the right place, with Fannie and Freddie, with review by the FHFA as regulator.
  3. It certainly makes sense for Fannie and Freddie to consider various available alternative credit score models, as provided in the Proposed Rule, but the primary decision criterion should always be each model’s contribution to accurately predicting future loan credit performance. The Proposed Rule reasonably suggests consideration of each model’s accuracy and reliability on its own, as well as when used within Fannie and Freddie’s credit management systems, but the latter is clearly the more important question.
  4. As the Proposed Rule importantly observes, “Credit scores are only one factor considered by [Fannie and Freddie] in determining whether to purchase a loan.”
  5. It is essential, as reflected in the Proposed Rule, for considerations of credit score models to take into account the time, effort, complexity, uncertainty, and costs (direct and indirect) to the mortgage industry of alternative decisions. In particular, the effects on smaller mortgage lenders should be addressed.
  6. It is a good idea to have the possibility of small-scale experiments or “pilot programs,” if appropriate, as the Proposed Rule provides.
  7. The Proposed Rule suggests using the standard definition of default with a time horizon of two years from loan origination. Consistent with the very long term of mortgage loans, I believe longer time horizons should also be tested for the extent of continuing predictive power of credit score models.

These are my personal views. It would be a pleasure to discuss any of them further.

Respectfully submitted,

Alex J. Pollock

Distinguished Senior Fellow
R Street Institute
1212 New York Ave. NW, Suite 900 Washington, DC 20005 202-900-8260 apollock@rstreet.org