Monday, January 28, 2008

More Bogus Numbers from the Mortgage Industry

(Adam Levitin in Credit Slips) The past few months have seen story after story about fraud in the mortgage industry. Now we're seeing a new type of fraud--mortgage lobbying fraud. The Mortgage Bankers Association has been claiming the proposed bankruptcy reform legislation that would significantly roll back the special treatment given to mortgage lenders in chapter 13 bankruptcies would result in residential mortgage interest rates rising 1.5 to 2 percent. (Somehow this number started at 2% and has drifted down to 1.5% without any explanation.) The MBA's number is pure and demonstrable hokum. As Joshua Goodman, a Columbia University economist and I show in a new working paper, permitting bankruptcy modification is likely to have little or no impact on mortgage interest rates or origination volumes. Keep reading below the break for the proof.

The MBA derives its number by looking at the spread in interest rates between mortgages on single-family principal residences, which cannot currently be modified in bankruptcy, and investor properties, which can be modified. The rate spread averages about 38 basis points. The MBA bulks the number up to 150-200bp by amortizing points and the higher down payments typically required on investor properties. (Amortizing down payments is questionable, but that's the least of the problems with the MBA number.)

The problem with the MBA number is that it is derived from a cherry-picked statistic. For conforming mortgages (and some jumbos), most lenders do not vary their rates by property type (holding all other factors, such as credit score, state of property location, loan amount, LTV ratio, and mortgage product constant), with one exception--investor properties. Mortgage rates on single-family homes, which cannot currently be modified in bankruptcy and those on multifamily residences and vacation homes, which can currently be modified in bankruptcy are identical. This means that the mortgage rate variation between single-family principal residences and investor properties is due to factors other than bankruptcy modification risk. It also means that market pricing does not indicate any sensitivity to bankruptcy modification risk. Similar patterns show up in private mortgage insurance premiums and Freddie Mac/Fannie Mae delivery fees. Rather than support the MBA's claim of a 150-200 bp cost for permitting bankruptcy modification, current mortgage pricing indicates that there would be no effect whatsoever.

Joshua Goodman and I also tested the impact of bankruptcy modification using historical mortgage rate data. The most radical type of bankruptcy modification is strip-down (a/k/a cramdown, a/k/a lien-stripping). In strip-down, an undersecured (underwater/upside down) lender's claim is bifurcated into a secured claim for the value of the collateral and an unsecured claim for the deficiency. The secured lender will likely recover the full value of the collateral, but will typically have little or no recovery on the deficiency. Between 1979 (when the Bankruptcy Code took effect) and 1993 (the Nobelman ruling), federal courts varied by district as to whether they would permit strip-down of mortgages on single-family principal residences. This creates a natural experiment for testing the impact of permitting strip-down on mortgage interest rates. The details are in our paper. We found that historically, strip-down likely resulted in an increase in mortgage interest rates of 10-15 basis points, but in some of our specifications, we could not rule out strip-down having no impact whatsoever. In any event, based on our historical findings, there is a zero percent chance the MBA figures are correct.

How the historical figures are to be reconciled with current market pricing is an open question, but at the very least, our study shows that permitting strip-down, which could help hundreds of thousands of families keep their houses and avoid foreclosure, would result in little or no increase in mortgage interest rates.


No comments: