Earlier this week the FDIC and Federal Reserve System jointly sponsored a symposium on Mortgages and the Future of Housing Finance. One look at the agenda, and I knew I had to be there: two full days of panels devoted to what welcoming speaker Federal Reserve Chairman Ben Bernanke described as "policy-oriented research" from a host of PhD economists and a handful of lawyers. The announcement reminded me that in some circles - though not apparently in the policy community - the ever-evolving financial crisis has cast many occupations into disrepute, most especially that of "trained economist." A 1991 report, published in the Journal of Economic Literature, asserts that "graduate programs may be turning out a generation with too many idiot savants skilled in technique but innocent of real economic issues." Today, lament the dissenters, mathematical technique is pursued for its own sake in the journals and departments of economics. The narrowly specialized products of the prevailing educational philosophy have little grasp of psychology, institutional structures or historical precedents. Say the dissenting economists: "This insufficiency has been apparent among those economists giving advice to governments, banks, businesses and policy institutes. Non-quantified warnings about the potential instability of the global financial system should have been given much more attention." Those are my italics. Having stumbled over way too many academic and staff economist working papers while questing for facts and usable numbers, I've come to believe that most of these folks are muddling the discussion. By muddling, I mean hopelessly obscuring the authentic and practical issues. How can option-pricing and logit models pull us out of the foreclosure and servicing quicksand our homes, our household wealth, our futures are sinking into? How could a paper on "ex ante and ex post relations in the mortgage market" prevent the bulldozing of entire neighborhoods? Or discourage once responsible citizens from living rent-free in houses they intend to abandon when the foreclosure machine finally reaches their doors? How can a multivariate regression on lender modification actions assist a homeowner who, though always current on their mortgage, needs to stop the lender from selling the house in a foreclosure auction? The Rubber Meets the Road Happily Laurie Goodman, a long-time mortgage analyst and Senior Managing Director at Amherst Securities, moderated one of the panels at the FDIC/Fed symposium. Goodman managed to make transparent the conflict between mathematical technique and real-world substance. Goodman's panel addressed "Secondary Mortgage Markets," an urgent area of concern if U.S. housing finance is to be reconfigured and revitalized. A PhD-trained economist with demonstrated quantitative chops of her own, Goodman approached the panelists' papers much like a referee providing academic peer review. (The titles of the papers and names of the authors can be found in the agenda posted on the FDIC's site. The FDIC's Office of Public Affairs has assured me that slide presentations and papers presented at the symposium will be available on the website no later than November 5.) Two papers applied statistical modeling techniques to loan level data to achieve conclusions about adverse selection during the mortgage securitization bubble. The results will come as a surprise to actual market participants living in the trenches and bomb craters of this disaster. One paper determined that, during the period 2004-2008, banks sold loans with low risk of default to the GSEs and private issuers and kept loans with higher default risk in their portfolios. The result applied to both prime and subprime loans, though they found the difference in default risks between securitized and portfolio loans were less in subprime than prime. Speaking as a quant, Goodman made recommendations regarding this study's methodology. And speaking as a real-world analyst, she pointed out a problem most real world MBS credit analysts would have immediately caught - the universe of loans on which the results were modeled is incomplete. The problem affected non-GSE loans and was most pronounced in the peak years of originate-to-distribute lending. The underlying data was missing about 38% of the known securitized loans in 2006 private label deals and about 58% in 2007. The subprime representation was also exceedingly low, using information altogether on 22,901 loans sold or retained in years 2004 through 2007. I missed Goodman's count based on the CoreLogic LoanPerformance loan level securities data base, but the OCC/OTS Mortgage Metrics report, easily accessed publicly, indicates more than 2.6 million subprime loans were still surviving as of June 2010. Also, Goodman noted that the reported results were further skewed by lumping loans sold to Fannie and Freddie loans in with FHA/VA loans. On my nickel Frankly, I thought Goodman was being kind to this paper. I can't speak to the mathematical techniques, but reading it I encountered numerous errors in its description of the mortgage market features. I'll mention a few particular especially cringe-worthy.
  • "GSEs offer investors guarantees against default risk, while private issuers often pass the default risk on to parties that are willing to bear it." No, private issuers pass all default risk to the securities. Realized losses are passed up the capital structure from junior most security classes to senior most.
  • GSEs have historically purchased only traditional fixed-rate mortgages and only recently began to purchase alternative mortgages such as adjustable-rate mortgages. No, the GSE's have been securitizing ARMs since the early 1980s.
  • GSE issuance of mortgage-backed securities is subject to detailed SEC filings and public reporting requirements, while private issuers are not. No, the GSEs were exempt from SEC filing requirements, voluntarily registered after the accounting scandals of the early '00s. Private issuers are and always have been subject to SEC securities requirements.
  • The implicit government guarantee might give GSEs an incentive to be more aggressive in risk-taking than private issuers. Goodman's careful dissection of loan level characteristics of Fannie versus prime and subprime PLS gives the lie to this silly assertion.
  • Small lenders were observed to sell higher-default-risk loans and retain lower-risk loans in the last two years of the study. "We attribute this to the fact that small lenders have less to lose from risking their reputations in the secondary market ...." WHAT?
  • " ... the difference for subprime loans becomes insignificant in 2006 and 2007, the height of the real estate bubble." Depending on the index, home prices peaked in 2005 or 2006. The securitization bubble continued to proceed into the first half of 2007.
  • "The difference between subprime and prime loans is likely due to the fact that subprime loans are subject to more scrutiny by investors than prime loans." If only this had been true. We would not be having this discussion today.
Adverse selection bites the hand it feeds The other study made a related assertion in more technical language: when making the decision to originate, a leading national mortgage bank applied lower screening efforts on loans that had higher predicted probability of being securitized. However, once originated, the loans it retained performed worse than the sold loans. The period studied covers the period 2004 to 2008 and the bank sold 89% of its originations over the period. That loosely translates to: 'efforts to adversely select bad loans into securities backfired'. That is, "adverse selection works against the bank once the loans are originated." The authors surmise that investors (by investors, they mean loan buyers, such as the GSEs, Ginnie Mae, investment banks, hedge funds and private investors, but they are stumbling over the fact that a mortgage bank of this description likely securitized non-agency/GSE on its own shelf) had an information advantage over the originating bank due to the time lag between loan origination and loan sale. Specifically, loan buyers had fresher information about current payment status, changes in credit score and balances, and the recent performance of specific loan products, as well as local employment conditions and home price performance. (Again, this wouldn't matter if they sold the private loans to their own security affiliates.) This is a surprising result - and a tasty irony in hindsight. Unfortunately, as Goodman pointed out, the authors lumped data from January 2004 to February 2008 into a single data set. (Doing so would have clarified the supposed role of information that emerged between origination and sale, as conditions worsened more rapidly with successive vintages.) They also did not distinguish loans sold by the channel (GSE, prime private, whole loan, etc.) through which they were sold. Looking at securitization practice and performance across both of these dimensions strongly supports the common sense view that the originate-to-distribute model was inherently an adverse selection process. Proof-in-the-pudding trumps applied math Ok, Goodman conceded, if a player has an originate-to-distribute business model, they might not exhibit adverse selection because the loans that were left, by definition, would be the ones that were difficult to distribute. But you'd have to come in from Mars to think there was no adverse selection across the MBS market. To make her point, Goodman laid out the kind of simple descriptive statistics she has been using with security investors for years. (I recommend tracking down her slides when the FDIC posts them. Her tables and graphs, along with those of a few other presenters at the symposium, are worth keeping on file.) First, she put up a table of origination and securitization volumes by securitization channel (based on widely used data collected by Inside Mortgage Finance Publications). The telling numbers here were securitization rates (percent of originations securitized), which rose steadily in the last decade as home prices went up. Through 2007, the rising securitization rate reflected the increase in the originate-to-securitize sectors, namely subprime and Alt-A. Since 2007, the securitization rate has continued to inch up, reflecting the collapse of origination of jumbo, subprime and Alt-A lending, minimal portfolio origination and the virtually total dependence of the housing finance system on FHA/VA/Ginnie and the GSEs. The shift from agency securitization and bank portfolio to non-agency securitization from 2003 through 2006 was particularly dramatic. Next, she compared risk characteristics and delinquency statistics across Fannie, prime private label and subprime private label securities (PLS) by origination year:
  • Regardless of vintage, Fannies have lower serious delinquencies currently. Contributing to this result, prime PLS have lower concentrations of fixed rate loans and more IO loans than the Fannie vintages. (Adjusting for ARMs, the results are more similar.)
  • Non-prime PLS perform worse. For instance, 47% of the 2006 vintage (the worst) are 90 or more days delinquent, compared to 13.9% prime PLS and 12.5% Fannie originated in the same year. The non-prime PLS are distinguished by a low percentage of fixed rate loans, relatively low FICOs, and high concentrations of loans with IO and negative amortization features.
In particular, risk layered loans were much more likely to packaged in non-prime PLS. With her carefully laid out tables projected on the screens, Goodman concluded, "It is very difficult to think that there was no adverse selection. While an institution may not have explicitly done adverse selection, the entities that specialized in risk layered loans that were channeled into non-Prime PLS ramped up substantially during the 2005-2007 period." Jumping forward to the post-apocalypse period (oops, my term), Goodman provided graphic illustrations of the fact that there is virtually no credit available outside of government and GSE channels. Moreover, the GSEs have tightened their credit standards to the point that FHA/VA loans are the main source of credit for home buyers. This state of affairs has dramatic policy consequences. The taxpayer is now the only source of credit. This situation is made all the more urgent by the fact that a huge number of homes are coming onto the market. "That," Goodman expostulated, "should be the topic of conversation." In other (my) words, the policy and law makers claiming to be resolving this crisis and revitalizing the housing finance system by resolving the GSEs and redefining the government's role had best start with things as they are, not as they are modeled. NOTE: Linda Lowell writes a regular column, called Kitchen Sink, for HousingWire magazine. Editor’s note: Linda Lowell is a 20-year-plus veteran of MBS and ABS research at a handful of Wall Street firms. She is currently principal of OffStreet Research LLC.