What Does the New Community Reinvestment Act (CRA) Paper Tell Us?
Dec 11, 2012
There are two major, critical questions that show up in the literature surrounding the 1977 Community Reinvestment Act (CRA).
The first question is how much compliance with the CRA changes the portfolio of lending institutions. Do they lend more often and to riskier people, or do they lend the same but put more effort into finding candidates? The second question is how much did the CRA lead to the expansion of subprime lending during the housing bubble. Did the CRA have a significant role in the financial crisis?
There’s a new paper on the CRA, Did the Community Reinvestment Act (CRA) Lead to Risky Lending?, by Agarwal, Benmelech, Bergman and Seru, h/t Tyler Cowen, with smart commentary already from Noah Smith. (This blog post will use the ungated October 2012 paper for quotes and analysis.) This is already being used as the basis for an “I told you so!” by the conservative press, which has tried to argue that the second question is most relevant. However, it is important to understand that this paper answers the first question, while, if anything, providing evidence against the conservative case for the second.
Where is the literature on these two questions? One starting point is the early 2009 research of two Federal Reserve economists, Neil Bhutta and Glenn B. Canner, also summarized in this Randy Kroszner speech. On the first question Kroszner summarizes research by the Federal Reserve, the latest being from 2000, arguing that “lending to lower-income individuals and communities has been nearly as profitable and performed similarly to other types of lending done by CRA-covered institutions.” The CRA didn’t cause changes to banks’ portfolios, but instead required them to find better opportunities. More on this in a minute.
What about the second question? Here the Bhutta/Canner research notes that only six percent of higher-priced loans (their proxy for subprime loans) were extended by CRA-covered lenders to lower-income borrowers or CRA neighborhoods. 94 percent of these loans were either made by non-traditional banks not covered by the CRA (the “shadow banking system”), or not counted towards CRA credits. As Kroszner noted, “the very small share of all higher-priced loan originations that can reasonably be attributed to the CRA makes it hard to imagine how this law could have contributed in any meaningful way to the current subprime crisis.”
How did those loans do? Here the research compared the performance of subprime and alt-A loans in neighborhoods right above and right below the CRA’s income threshold, and found that there was no difference in how the loans performed. Hence the idea that a CRA-driven subprime bubble isn’t found in the data. (The FCIC’s final report, starting at page 219, has more on this and other research.)
So what does this new research do? It takes banks that were undergoing a normal examination to see if they were in compliance with the CRA, and thus under heightened regulatory scrunity, and compares their loan portfolios with banks that were not undergoing a CRA examination. It finds that the CRA exam increases loans 5 percent every quarter surrounding the event and those loans default 15 percent more often, under the idea that those banks were ramping up their loans to pass the CRA exam.
But this is question 1 territory. 94 percent of higher priced loans came outside CRA firms and outside CRA loans, and this research doesn’t really change that. Since we are talking about regular mortgages – more on that in a second – that higher default isn’t that scary. To put that in perspective, loans made in the quarter following the initiation of a CRA exam in a non-CRA tract are 8.3 percent more likely to be 90 days delinquent. That sounds scary, but it is an increase of 0.1, from 1.2 percent to 1.3 percent. In the CRA tract it is 33 percent more likely to default, going from 1.2 percent to 1.6 percent. FICO scores drop 7 points from 713.9 to 706.9. That’s an increase I wouldn’t want in my portfolio, but it is light-years away from 25%+ default rates, and very low FICO scores, on actual subprime.
This research, if anything, pushes against movement conservative CRA arguments. In light of the evidence in question 2, many conservatives argue that regulators used CRA to push down lending standards, which then impacted other firms. But this paper finds that extra loans aren’t more likely to have higher interest rates, lower loan-to-value, or be balloon/interest-only/jumbo/buy-down mortgages, although there is a slight increase in undocumented loans. And their borrowers aren’t more likely to have risky characteristics themselves. The authors conclude that “this pattern is consistent with banks’ strategic attempts to convince regulators that the loans they extend that meet CRA criteria are not overtly risky.”
Read that again. The authors argue, from their empirical evidence, that regulators were trying to make sure these loans had high standards, and CRA banks tried to comply with that as best they could on the major, visible risks of their loans. This is the opposite argument made by people like John Carney, who believes the CRA “encourag[ed] lenders to adopt loose standards for mortgages.” It also pushes against people like Peter Wallison, who, in his FCIC dissent, argued that CRA loans were more likely to have subprime characteristics or riskier borrowers in ways not captured by a higher-price variable. Not the case.
It also finds that loan volume and risk increases the most during 2004-2006, and points to the private securitization market as an important channel. This, along with characteristics above, pushes back against the idea that the CRA primed a subprime pump in the late 1990s and early 2000s, another favorite of movement conservative finance writers. If anything, banks undergoing CRA exams were caught up in the same mechanisms that were causing the housing bubble itself.
I’m not sure I buy all of the research. If CRA banks take on too many loans during examination, why wouldn’t they just loan less afterwards, balancing out? The paper jumps to argue the opposite, as it is worried that “adjustment costs may cause banks to keep elevated lending rates even after the CRA exam is formally completed.” This is meant to establish their results as a lower-bound, rather than an upper-bound. But really? They managed to ramp up their lending in enough time during this time. Either way it would throw a very different set of interpretations on their research. I’m interested in seeing how other researchers react to these problems. But for now these results don’t change the way we approach the financial crisis.
Originally published at the New New Deal