Are Consumer Protection Regulations Harming the Middle Class?

A new paper by Franceso D'Acunto and Albert Rossi, both at the University of Maryland's Department of Finance, contends that the Dodd-Frank Act resulted in "a substantial redistribution of credit from middle-class households to wealthy households", as lenders reacted to regulations by reducing credit to middle-class households and increasing it to wealthy households.  This conclusion is based on a regression analysis of loan and ZIP-code level HMDA data.  The redistribution point is a serious charge to be leveled at the Dodd-Frank Act, and you can bet that this paper is going to be repeatedly cited by Congressional Republicans in their attempts to repeal Dodd-Frank. 

Unfortunately, the paper is founded on a pair of mistaken factual claims about the legal landscape that are so staggering that I am puzzled how they could have been made in good faith. Once these mistakes are corrected, it becomes apparent that the paper's analysis cannot actually support its claims because it is testing the wrong thing.  The paper is observing changes in the mortgage market that pre-date the implementation of Dodd-Frank.  By definition, then, these changes cannot have been caused by Dodd-Frank.  What the paper shows (without realizing it) is that there has been a redistribution of credit from middle class households to wealthy ones, but that it wasn't caused by Dodd-Frank.  Whoops. 

Here's the problem.  If you want to test the impact of the Dodd-Frank Act on the mortgage market, you need to test the impact of the regulations enacted under Dodd-Frank.  While Dodd-Frank was enacted in late July 2011, the statute is not self-executing.  Instead, it required extensive rulemakings by various government agencies.  In the case of mortgages, the most important rulemakings were those by the CFPB.  The CFPB did not propose its regulations under title XIV of Dodd-Frank (the mortgage title) until 2013, and those regulations did not go into effect until mid-January 2014.  That means that the earliest year for which Dodd-Frank could have affected the mortgage market would be 2014.  The paper, however, looks at market changes around either side of the enactment of Dodd-Frank in 2011.  The paper's data is from 2007-2014, but rather than look at say changes in the market from 2013 to 2014 or 2012 to 2014, the paper looks at changes between 2011 and 2014, which would not seem to be driven by Dodd-Frank regulations.  What gives?  

D'Acunto and Rossi recognize that they aren't actually testing the 2014 Dodd-Frank regulations, but instead, refer readers to an appendix in which they explain that 2011 is nevertheless an appropriate year to test because in that year the major players in the mortgage market became subject to rules substantially similar to Dodd-Frank.  In other words, rather than claiming that the market was anticipating the regulations and changing its business model earlier than necessary, D'Acunto and Rossi argue that there was effective application of Dodd-Frank three years earlier.  Here's the key line from their appendix:  

Even though we do not observe directly the date at which all banks started their compliance process, we do know the exact timing of the compliance process for 14 servicers and the parent companies of 12 of them, which were deemed to be systematically important by the Federal Reserve System in November 2010. In April 2011, the Federal Reserve System disclosed a set of enforcement actions that imposed the immediate compliance to new standards for mortgage originations that were ”substantially similar” to the provisions approved in Dodd-Frank (Braunstein (2011)). These banks,therefore, had no choice but to start complying with the new provisions immediately after April 2011.

The entire test devised by D'Acunto and Rossi depends on this statement being correct.  It isn't.  It isn't even close to correct.  And i's hard to exaggerate what a total failure of research is reflected in D'Acunto and Rossi's claim.  They claim that there are (1) settlements about origination practices that (2) were "substantially similar" to Dodd-Frank.  Both statements are flat out wrong.  Even if they were right, however, the paper's design would still be flawed:

1.  The Fed's April 2011 enforcement actions were about servicing not originations.  The Fed only did one origination-related action, and that was in July 2011 relating to some specific problems with Wells Fargo.  D'Acunto and Rossi even include a hypertext link to the Fed's consent orders, none of which are about originations.  How on earth did they make this mistake?  Did they read the documents they cited?  If so, how did they misunderstand them so badly?  And if not, how did they devise the false story they tell?  

2.  The "substantially similar" language and claim that the 2011 settlements tracked Dodd-Frank comes from Fed Community Affairs Division Director's Sandra Braunstein's 2011 testimony before the House Financial Services Committee.  The problem here is that Braunstein did not in fact say anything about the April 2011 enforcement actions.  Instead, she was talking about the Fed's 2008 HOEPA rulemaking, which covers certain for higher-priced mortgages, being "substantially similar" to certain Dodd-Frank requirements:

I should note that the underwriting requirements legislated in the 2010 Dodd-Frank Wall Street Reform and Consumer Protection Act (the Dodd-Frank Act), which apply to all mortgage loans, are substantially similar, but not identical to the ability-to-repay requirements adopted by the Board in 2008 for higher-priced mortgage loans.

I'd take some issue with Braunstein describing the 2011 Dodd-Frank act and the HOEPA rules as similar, but reasonable minds could differ on this.  What is very clear, however, is that Braunstein is talking about the 2008 HOEPA rulemaking, not the 2011 settlements.  What is also very clear is that the Fed's 2008 HOEPA rulemaking is not especially similar to the 2014 regulations adopted by the CFPB under title XIV of the Dodd-Frank Act.  For example, there's no HOEPA analog to a qualified mortgage (QM), which gets a statutory safe harbor from the ability-to-repay requirement.  

3.  Even if the 2011 settlements were substantially similar to the Dodd-Frank act itself, wouldn't it be appropriate to actually analyze where there were distinctions, as those distinctions might matter?  D'Acunto and Rossi conduct no such analysis, and indeed, if they had, it would have been impossible to support their claims of substantial similarity, just as it would be to support the claim that the 2011 settlements dealt with origination. 

4.  Even if the the 2011 settlements were substantially similar to the Dodd-Frank Act itself, it's still pretty meaningless because it is the regulations implementing Dodd-Frank, which ultimately governs the mortgage industry, and those regulations weren't even proposed at the time of Braunstein's testimony.  In other words, there's nothing suggesting that the banks were effectively complying with the title XIV regs three years before the regs went into effect and two years before they were proposed.   

Bottom line here is that D'Acunto and Rossi haven't proven any connection whatsoever between the CFPB's Dodd-Frank regulations and changes in the mortgage market.  If anything, their paper indicates that changes in the mortgage market were happening for reasons other than regulation insofar as they show statistically significant changes predating the implementation of the regulations.  What could be driving these changes?  I'm not sure, but let me at least suggest one possibility, namely the GSE's increased enforcement of representations and warranties.  Stricter R&W compliance makes securitization less appealing to originators, and given balance sheet risk, those originators might have chosen to go with a different borrower profile.  

Still, I want to come back to the mistakes on which this paper is founded.  The errors made by D'Acunto and Rossi are so staggering that I cannot understand how they occurred. The mistakes in this paper strike me as altogether different and more serious than those I've seen in other papers I've criticized.  This isn't a matter of relying on data of dubious quality, such as BankRate surveys of free checking or even a tendentious interpretation of sequential relationship as causal. D'Acunto and Rossi's paper is built on an assumption of a legal relationship that is plainly contradicted by the very documents they cite.  I've never seen anything quite like this.  Sadly, this isn't just some technical academic squabble, but the sort of thing that will affect policy debates.  This paper might very well get accepted in a prestigious double-blind peer-reviewed journal as the reviewers are likely to be economists who don't know the legal background well enough to see the problem in the paper's design.  The talisman of double-blind peer-review plus graphs and regression analysis and LaTex formatting can easily bamboozle non-experts into thinking that this is a solid scientific study.  And even if Congressional Republicans know that the paper is flawed, well, is that really going to stop them from waiving it around as a banner for repealing Dodd-Frank?  

So let me say it now:  it's generally malpractice for economists to do legal event studies without a legal expert collaborator, and it's malpractice for economics journals to publish legal event studies without a peer-review by someone with expertise in the legal background.  There are simply too many nuances that economists can miss because they don't understand the legal framework.  

 

Comments

[jay westbrook]

With special reference to your last paragraph, right on. Of course, that can cut the other way as well.

[Adam Levitin]

Yes, what's good for the goose is good for the gander. There's lots of lousy empirical work done by attorneys these days. And I've even seen JD/PhDs foul things up on both the law and the empirics. But I still think best practices are to have a collaborator.

[Adam Levitin]

Funny story: when I was on the AALS entry-level market and interviewing with NW, Lee Epstein attacked me for suggesting that I might do empirical work. "What makes you think you can do empirical work without a PhD in statistics," she asked. I was polite and conciliatory in my answer, but I still regret not having responded with, "The same chutzpah that makes you think you can teach at a law school without a JD."

[Bob Lawless]

Adam, I agree with a lot of your general comments. I have not read this specific paper and don't take a position on that either way. What I like to say is that legal scholars tend to have fine-grained knowledge of institutions that helps to ensure reliability and validity in empirical studies. Of course, as Jay points out, it can cut the other way.

The other thing to which I wanted to react was the notion of having a collaborator. Yes. It is good practice to have collaborators period. I have never understood the legal academy's reluctance to embrace scholarly collaboration. In most other fields, working alone is generally a sign that no one will work with you. There should be more collaboration across all forms of legal scholarship.

[FJP]

There is another unexamined issue here. Assume for the moment that certain kinds of regulations of consumer credit may lead lenders to tighten approval standards. Is that necessarily and ipso facto "harmful" to middle class consumers in all cases? I am going to suggest not. After all, one of the problems in the housing crisis was that the availability of no-doc mortgages and the use of various non-transparent subprime products encouraged people to buy more house than they could afford. Also, let's say that regulating fees and capping interest on a kind of loan means that lenders are less likely to approve marginal risks. However, all the consumers who are granted credit benefit from the rate and fee limitations. Focusing just on the marginal change in credit disapprovals does not take in the whole picture.

[Knute Rife]

First, kudos for drop-kicking Lee Epstein. If such things happened more often, law schools would probably be of more use to the profession. I'm amazed you have an academic career, though, after snarking off to a talking head grade demigod like that.

Second, I've read the article. Disclosure: I am skeptical of the use of math in the social studies. This stems largely from excessive exposure to economics and business classes and their combination of false precision and unsupported axioms that would have them laughed out of the sciences. I believe your assessment is correct, namely that the authors begin with assumptions that are unsupported or even wholly invalid and then apply those assumptions to their data with rather less rigor than is called for. I would submit that, at least for D'Acunto, this is not unique. A review of his other research shows a paper titled "Tear Down This Wall Street: The Effect of Anti-market Ideology on Financial Decisions." Its conclusion is, "Contrary to behavioral biases, anti-market ideology makes more sophisticated agents deviate from neoclassical decision-making," but there is no examination of the assumption that the system in any way runs on neoclassical decision-making.

I'm compelled to conclude the authors have a nearly total lack of understanding of how regulations work and how they affect behavior. Consequently, the paper ends up being a study in how not to.

I would add one comment to your alternative causes. When the crisis hit in 2007, the money spigot closed in large part because the securitizers were holding their breath, waiting to see to what extent the IRS would enforce the REIT safe harbor requirements. In about 2012 it became know the IRS would not enforce them at all, and the spigot opened back up. From my seat here in the trenches, bankruptcies have fallen off since then because people can get their hands on financing again and kick the can a little farther down the road. When we see lending tighten back up (as we now see in oil & gas and its dependent industries), it's because the risk is abysmal and the return can't support it, not even the up-front fees.

[Alan White]

Having done the legal coding in a number of coauthored empirical studies of various laws, and collaborated with statisticians, I heartily endorse Adam's point. Here is another example of a seriously flawed study based on misunderstanding the legal environment variables: http://www.creditslips.org/creditslips/2015/04/fast-foreclosures-slow-foreclosures.html.