Ever since the subprime mortgage market began to implode, people have been trying to exonerate the government policies that contributed to the mess.
In some sense, the attempt is just absurd. It just is not plausible to claim that decades of government intervention in the housing market—especially the campaign to expand mortgage lending and home ownership that began in earnest in the mid-1990s—had nothing to do with the housing boom and bust.
But the attempts just keep coming. The latest is a deeply flawed paper from the St. Louis Federal Reserve Bank that purports to prove that the push for “affordable housing” had nothing to do the origination or price of subprime mortgages . You can read a quick—and much too approving— summary of the paper on the Washington Post’s Wonkblog .
“In contrast to studies of the effect of affordable housing goals in the prime market, we find no evidence that affordable housing legislation affected the subprime market during the subprime crisis,” the authors of the paper write.
They’ve done an admirable amount of work on a dataset of subprime mortgages and made sure to include securitized subprime, something earlier studies had missed. But they manage to miss their own silent assumption—and by missing it, they bungle the entire study.
What the study sets out to do is discover whether subprime mortgages that qualify as meeting affordable housing policy goals were differently priced or performed differently than other subprime mortgages. The intimidating sounding name they give this is “regression discontinuity analysis.”
Spoiler alert:They discover that there isn’t much of a difference. They were priced about the same, and they defaulted at about the same rate.
The hidden assumption that leads them astray, however, is that financial institutions are capable of finely tuning their lending operations so that they can have one set of lending standards for policy-qualifying mortgages and another for non-qualifying mortgages. While this is theoretically possible, it is enormously difficult and costly to implement—especially for very large, complex financial institutions.
The potential negative consequences for not meeting affordable housing policy goals embedded in the Community Reinvestment Act were taken extremely seriously by banks. Failure to satisfy regulators that your bank was doing its all to meet those goals could result in severe penalties and restrictions, including the inability to acquire other banks or open new branches.
To assure that they didn’t miss the goals, the banks changed the way they made home loans. They automated, they promoted true-believers in the new mortgage products, they fired and demoted anyone who attempted to stick by more traditional lending standards.
As I reported at Business Insider in June 2009 :
Banks actively upended their entire mortgage lending structures in pursuit of CRA loans. At one point, Bank of America took away the power of any of its local branch offices to decline loans. It then sent senior managers from Charlotte to supervise every single mortgage application. This occurred as early as 1988, long before securitization took off. Soft knowledge was downgraded in favor of hard knowledge, all in pursuit of CRA compliance.
In short, in order to make enough affordable housing policy loans to ward off regulators, banks had to lower their standards on all loans.
The Fed study assumes this fact out of existence. It simply asserts—without any argument or evidence—that if mortgages were being made because of policy or if lending standards were being changed to make policy-friendly mortgages, the effects would be confined to the small subset of mortgages that actually qualified.
Imagine a bar that wants to attract pretty women to drink at its bar. It declares that Thursday nights will be “ladies night” with half-priced drinks for women. But ladies night attracts not just pretty women—it attracts women of all sorts. If you ran a regression discontinuity analysis on the prices paid by women, you’d conclude that the attempt to attract pretty women had nothing to do with the price of drinks. The dataset would show you that all the women drank at the same price.
That’s exactly what the fellas from the St. Louis Fed have done here.
Questions? Comments? Tips? Email us atNetNet@cnbc.com or send a text message to: 917-740-8477.Call us at 201-735-4638.