Diversification is critical to protecting the government in housing finance reform
Last week, Cliff Rossi in the American Banker suggested that our earlier brief on how much private capital should stand in front of a catastrophic government guarantee in a new housing finance system didn’t go far enough in considering the need for lender diversification of the insured portfolio. We agree and disagree.
A key question in the ongoing housing finance reform debate is the amount of private capital needed to take mortgage credit risk ahead of a government guarantee against catastrophe. The Corker-Warner proposal, for example, sets a 10 percent capital requirement. While policymakers and analysts may argue whether this 10 percent is too low or too high, one cannot answer this question without considering both diversification and the form the private capital takes.
As we have shown at length in our recent issue brief, the 2007 Freddie Mac book of business experienced cumulative “defaults” or “credit events” of 10.9 percent, the highest of any vintage year, because of the subsequent nearly 35 percent drop in home prices nationwide. A 10.9 percent default rate translates into a 4.4 percent loss rate.
Applying this worst-year experience to Freddie's entire book of business, it is clear that 5 percent capital would have allowed Freddie to weather the recent economic disaster. The results for Fannie are similar. However, it is critical to recognize that this 5 percent capital estimate reflects the substantial diversification--in size, geography, and, as Rossi correctly points out, multiple originating lenders--in the GSEs' book of business. Without this diversification, it is unlikely that 5 percent would be enough.
Here's why. We looked at 1,000 randomly selected pools of 100, 500, and 2,500 Freddie Mac loans. The default rate ranged from 0.05 percent to 22 percent for the hypothetical 100 loan pools. The distribution is narrower (5 percent to 15 percent) for hypothetical 500 loan pools, and narrows further from 8.25 percent to just over 12 percent for hypothetical 2,500 loan pools. A tighter distribution implies the need for a smaller capital cushion to cover unexpected losses.
Focusing on geographic diversification, we showed that a pool with only Arizona loans had a much higher default rate than a national pool. The average loss rate was in the range of 9.0 percent to 9.5 percent a pool of 2007 vintage Arizona loans, compared to the 4.4 percent loss for the vintage as a whole.
We concluded that both the government and investors would demand large, diversified pools, especially if the government guarantee ran to individual pools, as in the capital markets execution contemplated in Corker-Warner. Rossi's additional point about variability in lender performance raises particular problems for small lenders. How would these get the number of loans needed for a capital market offering without being beholden to an aggregator, such as one of the largest banks? Moreover, even if they did get a critical mass of loans, would those loans represent sufficient diversification? And then should such single-lender pools (from lenders of whatever size) be subject to the additional capital charge Rossi suggests is needed?
Note that while we agree with Rossi's basic point concerning lender diversification, we do not agree the correct conclusion from this is to apply his suggested "seller effect factor," which would increase the capital requirement to account for lack of lender diversification, to the 5 percent capital we determined to be sufficient for a highly diversified insurance entity. That is because when, as in Corker-Warner's bond guarantor execution,the government stands behind an insurer rather than a pool, the government would have the benefit of the diversification of insurer's entire book of business, and moreover, of the insurer's equity capital.
Of course, the insurer itself needs diversification. Ideally, that would come from diverse individual pools, but could also reflect many individually less diverse pools. Without such diversification, for example if the insurer covered loans in only one state, even a 10 percent capital requirement might be insufficient.
In sum, the amount of capital needed to protect the government cannot be divorced from diversification or from the way the capital is structured. This is both highly variable and complex. Ultimately, it is important not to hardwire capital requirements into statute beyond a floor, and to require frequent stress-testing and prompt corrective action to make certain that those who stand ahead of the government always have sufficient capital.