By Karen and Basil Petrou
In the raft of crisis retrospectives released during the ten-year anniversary of the Great Financial Crisis, general consensus continues the conventional wisdom that subprime mortgages were the spark of the subsequent conflagration. A new study from the Federal Reserve Banks of Atlanta and New York mobilizes formidable data to show that hapless subprime purchase-money borrowers were victims, not perpetrators. The borrowers who did the damage that precipitated the debacle were, they find, prime borrowers whipped into a speculative frenzy by the combination of low rates and flagrantly-unwise mortgage lending. Theoretically, post-crisis reforms have solved for this. Actually, maybe not given the exodus of mortgage securitization from regulated entities, sharp rise in cash-out refis, and investment-focused borrowing with house prices well above affordability thresholds in many major markets.
The thesis of this paper is critical to both housing and regulatory policy: if it isn’t credit scores, but borrower purpose that drives probability of default, then GSE underwriting criteria excluding most low-score borrowers are poor risk buffers and all of the bank-capital rules penalizing low-score mortgages miss the safety-and-soundness mark even as they exacerbate economic inequality.
The Last Crisis
Most startling in the study’s findings is that the counties in which lots of subprime purchase-money loans were made are not the ones that experienced big house-price bubbles and subprime-lending booms. House-price growth is found highest in the West, Florida, and the Northeast corridor; highest-growth subprime purchase-money mortgage lending was in the Midwest and Ohio River. Indeed, the regions with the largest share of subprime purchase-money loans are negatively correlated between 2002 and 2006 with areas with the highest price appreciation, with the negative correlation holding when credit scores are considered along with other factors.
Following convention, the paper assigns subprime status to loans with FICOs of 660 or below. Adding subprime refis to the mix does not change this or other key conclusions, the authors say, because refis are enabled by housing-purchase demand. Thought of this way, it’s prime-borrower demand that puts subprime borrowers at risk.
Using this methodology, this paper finds that the crisis perpetrators are individual real-estate investors and speculators, most of whom were high-score, prime borrowers. Frauds such as exaggerated income, inflated valuations, or deceptive loan purpose are not concentrated in subprime purchase-money loans. And, as previous research has shown, foreclosures were more common for prime than subprime borrowers in the early years of the crisis when loan type, not sustained macroeconomic stress, was the main cause of delinquency and then default. In short, a larger credit supply to marginal borrowers (who were increasingly priced out of the market for true home ownership) before the crash doesn’t explain the crash.
The Next Crisis?
Underlying whatever one thinks accelerated the Great Financial Crisis, the conventional wisdom has it that credit supply overloads for subprime borrowers lit the fuse. If subprime lending and house-price appreciation are indeed negatively correlated from 2002 to 2006, then the crisis was not the fault of low-score households looking for a home. Instead, as this paper suggests, credit supply overheating came not from too much lending to risky borrowers, but rather from too much lending for risky products on which high-score borrowers thought they could make a killing.
Since the crisis, the GSEs, VA, and FHA have been credit-score conscious but also product-happy, enabling loans for rapid-fire cash refis, high DTIs, dubious HLTVs without third-party insurance underwriting, and even Airbnb speculation. Bank regulators have similarly squashed lending to low-score borrowers on high-risk grounds, but they have proved far more accepting of high-score borrowers in portfolio loans regardless of product characteristics. As we have frequently noted, CCAR stress-test scenarios have a draconian house-price depreciation scenario. Under it, banks can only make mortgage money from higher-risk loans where pricing takes a large capital cost into account – i.e., the same types of loans that the Fed staff believe fueled the last debacle. Similarly, GSE underwriting is now focused only on high-score borrowers – e.g., Fannie Mae’s weighted average credit score for loans acquired last quarter was 743 – but lots of higher-risk products such as overt investor loans are now part of the mix. VA is way into cash-out refis for high-risk borrowers, perhaps the worst of high-risk products for high-risk borrowers.
And, even as the GSEs are fueling these higher-risk markets – for we know that they are regardless of whether the Fed paper is right also about the relative risklessness of subprime purchase loans – FHFA and Congress are pushing Fannie and Freddie to do still more credit risk transfers. The Hensarling-Delaney GSE-reform discussion draft is so enamored of these risk transfers that it posits a whole new Ginnie Mae full-faith-and-credit guarantee scheme posited on lots more of them. But, what’s a credit risk transfer as practiced by the GSEs? As their websites make clear, these structures are complex tranches with intricate loss layering and uncertain pricing to counterparties with dubious long-term liquidity. As a result, the post-crisis housing-finance system is characterized not only by lots of high-risk products, but also high-risk, opaque securitization structures all too similar to the CDOs and CDO-derived products that did so much harm after 2008.
Could it be that, in the interest of post-crisis purity, we’ve mis-served LMI and first-time borrowers and again over-heated the speculative market? Sure looks like it.