The latest failed effort to blame the Community Reinvestment Act for Accounting Control Fraud

By William K. Black
(Cross posted at Benzinga.com)

Introduction

The latest effort to blame the Community Reinvestment Act (CRA) for the epidemic of accounting control fraud that drove the crisis is an econometric study by Sumit Agarwal, Efraim Benmelech, Nittai Bergman, and Amit Seru (“the authors”) (“ABBS 2012”). The study does not prove its thesis. The fact that the authors claim it proves causality makes obvious their controlling biases. Their title is “Did the Community Reinvestment Act (CRA) Lead to Risky Lending?” Their abstract answers: “Yes, it did.” They claim that their econometric study proves causality – which is impossible given their methodology. The authors were taught from their freshman years that an econometric study of this nature could not prove causality. Errors this basic and embarrassing demonstrate the crippling grip of the authors’ biases.

The authors’ statement of their thesis is:

“Regulators use this [CRA] compliance information when approving applications for new bank branches as well as for mergers and acquisitions. Banks may therefore have a strong incentive to concentrate their CRA-compliant lending before an exam date to ensure a satisfactory CRA evaluation while minimizing the likelihood that such lending would fail the “safe and sound” criteria at the time of the exam. Moreover, since the exam itself is not an instantaneous event, with a significant time gap between the exam initiation and the submission of the final CRA report, banks also may choose to elevate their lending a few quarters after the CRA exam” (ABBS 2012: 2).

Their thesis statement thesis may seem clear, but it is actually imprecise and arbitrary and reveals disabling errors in their methodology. It fails to exclude alternative explanations for why lending could vary in a time period near a CRA examination. Its imprecision as to when banks supposedly made loans based on the CRA combined with the lack of a convincing causal explanation as to why loans made in this particular time window would have prompted by the CRA while loans outside the window would not be demonstrates fatal flaws in the study’s design.

The study fails to consider alternative explanations

The ABBS (2012) study design also fails to consider a range of other factors relevant to causality that would demonstrate the inability of their thesis to explain the broader pattern of the crisis. I discuss why accounting control fraud provides a superior explanation for the lending pattern that the authors found.

If their thesis were correct then CRA would have produced ever fewer bad loans from 1999 on as the CRA examination frequency was cut sharply for smaller banks by the Gramm Leach Bliley Act from every two years to every five years (ABBS 2012: 11-12). Further, under the Bush administration, CRA supervision became ever weaker compared to the Clinton administration. The assumed “strong” incentive was typically trivial, falling, and overwhelmed by vastly stronger incentives to make bad loans to discourage regulatory takeovers and to increase bank officer compensation through accounting control fraud.

Ignoring accounting control fraud as an alternative explanation

The authors’ failure to consider alternative assumptions that could explain the fact pattern of the crisis and the results of the econometric study is striking. I used the “find” feature to search their article for the words “fraud,” “deceit,” and “misrepresentation.” The article does not contain any of these words. I show below that this omission is particularly curious because the authors implicitly accuse the Nation’s largest banks of being control frauds.

The FBI warned as early as September 2004 that there was an “epidemic” of mortgage fraud and predicted that it would cause an economic crisis if it is not contained – and economists ignore even the possibility that fraud could cause bad loans to be made. White-collar criminologists have testified that the epidemic of mortgage fraud consisted overwhelmingly of “accounting control fraud” and drove the crisis. They also explained what lending behavior optimizes such frauds – the accounting control fraud “recipe” for a lender. The recipe has four ingredients:

Grow extremely rapidly by
Making terrible loans at a premium yield while
Employing extreme leverage and
Providing only grossly inadequate allowances for loan and lease losses (ALLL)

Our literature explains how following the recipe leads to record reported income for lender, great wealth increases for the controlling officers, and catastrophic losses to the lender. The authors do not have a single citation to the criminology literature.

Worse, the authors do not cite George Akerlof and Paul Romer’s classic article entitled “Looting: the Economic Underworld of Bankruptcy for Profit” (Brookings 1993). Akerlof & Romer agreed that what criminologists term “accounting control fraud” (what they called “looting”) is a “sure thing.” It is the norm for economists to fail to have any knowledge of the criminology literature that is essential to understand finance and financial institutions, but they at least have the excuse that they have never studied criminology.

The authors, however, have no excuse for failing to cite Akerlof & Romer and no excuse for failing to consider the alternative that bank officers deliberately make bad loans. Indeed, given that Dr. Seru is the co-author of a contemporaneous study with a far superior research design that demonstrates the presence of “pervasive” control fraud by what the co-authors term our “most reputable” banks, the failure to consider accounting control fraud as driving the making of bad loans is disturbing. The pattern of lending their study identifies is much more likely to be a product of accounting control fraud than the CRA.

Asset Quality Misrepresentation by Financial Intermediaries: Evidence from RMBS Market. Tomasz Piskorski, Amit Seru & James Witkin (February 2013) (“PSW 2013”).

My comments on that study can be found here.

The authors’ failure to provide a coherent logical explanation for their thesis

Note that the authors claim that the CRA led to “risky lending.” That would be very odd given what the authors admit about the CRA’s provisions.

“The Community Reinvestment Act of 1977 instructs federal financial supervisory agencies to encourage their regulated financial institutions to help meet credit needs of the communities in which they are chartered while also conforming to “safe and sound” lending standards.

By law, such borrowers include any applicant that is located in CRA-target tracts – defined as census tracts with a median income less than 80 percent of the median income at the Metropolitan Statistical Area (MSA) level – as well as low-income and minority borrowers in non-CRA-target tracts” (ABBS 2012: 1, 2).

An obvious problem with the authors’ claim that the CRA caused “risky lending” is that the law calls for “safe and sound” lending standards. If lenders chose to engage in “risky lending” they did so in contravention of the instructions of the CRA and the banking examiners. The authors cannot be claiming that a bank cannot make safe and sound loans to a census tract that is <80% of the median income of the MSA. The authors do not even attempt to make that explicit causal claim. They make the opposite statement – that lenders make unusually low-risk loans due to the CRA.

“We hypothesize that in attempting to comply with the CRA, banks take advantage of this regulatory feature by concentrating lending in CRA tracts to higher-income borrowers, who presumably are less risky. Consistent with this hypothesis, our results show that the effect of a CRA exam in CRA-eligible tracts indeed rises with borrower income” (ABBS 2012: 3).

The discussion of liar’s loans below will explain the unreliability of the authors’ data on borrowers’ incomes.

The authors’ methodological assumptions are unreliable

“Our identification strategy hinges on the assumption that the timing and frequency of the CRA exams themselves are not driven by local credit and economic conditions affecting a given bank’s lending behavior. This is a plausible assumption since in practice both the timing and the frequency of CRA evaluations are driven by a fixed policy that is unrelated to local economic conditions” (ABBS 2012: 2).

The authors admit that one of the many assumptions their study design “hinges on” is that CRA exams are not driven “by local … conditions.” That assumption is implausible “in practice.” Yes, there was a general frequency for CRA examinations, but that frequency is not the same as examination scheduling. Examination scheduling is not known by the banks with any certainty, and this is particularly true at the largest banks, which the authors admit drove their results. This uncertainty is a crippling flaw in the authors’ design, which assumes that the banks adjust their lending because they know a compliance examination is about to start. Examination scheduling is an art that requires constant adjustments driven by the availability of staff, coordination with other examinations by the same agency, coordination with other regulatory agencies, and factors specific to individual banks, cities, or regions. If a senior regulatory official reads a story or watches a documentary alleging a scandal examination schedules can change dramatically and immediately. The length of examinations can also change radically depending on what the examiners find and the priorities of senior supervisors. “In practice,” I intervened with my colleagues who decided such matters to change the timing, scope, and length of many examinations due to facts unique to the S&L.

The authors contradict themselves on the typical time to complete the on-site CRA exam.

“In order to enforce compliance with the CRA, institutions are evaluated periodically through an onsite exam. The exam is conducted by the federal agencies responsible for supervising depository institutions…. [Completing] the onsite exam … takes several months….” (ABBS 2012: 6).

[T]he start and end dates of CRA exams are usually no more than two weeks apart…. (ABBS 2012: 8 n. 9).

The length of examinations is driven overwhelmingly by the size of the institution and how badly run it is. We have data on the on-site time to complete a compliance examination for community banks – a much more complicated process that is far more time consuming than conducting a CRA examination. The FDIC audit of examinations found that the average time to complete a compliance examination on a community bank with the top compliance rating, (“1”), was about 22 days. Averages are pulled strongly by extreme values, so the median number of days to complete the on-site compliance examination was even lower (and the median is what matters for the authors’ study). The authors provide no source for their claim about the on-site length of CRA examinations in measured in “several months” (an ominously vague phrase for an empirical study).

The authors’ claim that the CRA creates a “strong incentive” to make unsafe loans

Three other methodological and logic problems arise from the authors’ vague concept “around CRA exams” and their finding that their results were driven overwhelmingly by the largest banks.

“[W]e find that large lending institutions drive our main findings on the impact of CRA exams on the quantity and quality of extended loans. This is to be expected: federal regulatory agencies consider depository institutions’ CRA scores when considering applications for deposit facilities, including branch openings as well as mergers and acquisitions. To the extent that larger banks are more heavily engaged in mergers and acquisitions activity and expansion through branch openings, they will have a greater incentive to maintain a high CRA score and thus to adjust their lending behavior to satisfy CRA exams”

[T]he effect of CRA examinations, in both CRA-eligible tracts and non-CRA eligible tracts, is concentrated among the largest banks with assets above $50 billion. The coefficients of both the level and interaction-terms of CRA examination variables are smaller in magnitude and, in the case of banks with assets between $1 billion and $50 billion, often not statistically significant” (ABBS 2012: 3,17).

Note that the authors use the phrase “maintain.” One obvious problem with their thesis, particularly under the Bush administration (the relevant administration for their study) is that CRA (and safety and soundness ratings) were so inflated that unsatisfactory ratings virtually disappeared. None of the largest banks were ever at risk of an unsatisfactory CRA rating – and they knew it. They did not have to “adjust their lending behavior” because of CRA examinations because such examinations posed no threat to their CRA ratings. The idea that the largest banks’ mortgage lending was driven by the CRA is so divorced from reality as to be fanciful.

The authors’ purported CRA effect was actually not supported for the vast majority of banks. Their purported “effect of CRA examinations” (another example of falsely-attributed proof of causality) existed at only a tiny number of banks (49) (ABBS 2012: 17 n.18). (The effective number of the largest banks in the sample is also reduced because the study design uses many of them as controls. In any given year, there would be roughly 25 CRA exams of the largest banks.) The effective number is so small (particularly if I am correct in surmising that many of them were merged during the study period) that statistical inferences of association are dubious, particularly where the authors use lead and lag time periods chosen not by theory but by the data plots – and then assert that the data plots prove causality.

The authors’ inconsistent and incomplete analysis of the largest banks as control frauds

But consider the logical inadequacy of the authors’ claim about large banks’ motivations in knowingly making bad loans, not because they were compelled to do so, but because they wanted to grow through acquisitions. Under the authors’ assertions, the banks’ controlling officers are willing to make imprudent loans that they know will cause large losses so they can grow. Why? The CRA does not require them to make bad loans. Banks can get an outstanding CRA rating without making bad loans. The authors report that other banks, the supposed “controls” that are not within the authors’ asserted six quarter CRA examination window, make “safe and sound” loans with materially lower risk in the same CRA census tracks. Why do the largest banks wish to inflict on themselves what the authors describe as “markedly riskier” loans sure to cause the banks much larger losses? The authors will later assert that the lenders realized they could sell their bad loans to the secondary market, but the 2013 study on “pervasive” “misrepresentation” (fraud) by our “most reputable” banks makes the key point – the loans sales were almost always made under reps and warranties that left the seller on the hook for the losses caused by the bad loans. The fraudulent loan sales, therefore, harmed the seller.

The largest banks committed “pervasive” fraud in the sale of the underlying fraudulent mortgages because it made the officers – not the bank – wealthy. This brings us the answer to the question of why the lenders would make “markedly riskier” loans. Following the fraud recipe is a “sure thing” that quickly makes many bank officers wealthy.

The authors actually make a much more radical claim about the nature of the instruction from senior management to the loan officers (and presumably the loan brokers). They assert that the controlling managers delivered a twin directive to their staffs to make many more bad (often fraudulent) loans and to hide that fact from the regulators and many other actors. Those directives were complex and sinister because they ordered a deliberate effort to deceive the examiners into believing that the loans in CRA census tracts were safe when the originators knew they were actually imprudently risky due to “the reduction in loan standards.”

“[W]e find that the reduction in loan standards associated with elevated lending around CRA exams is based primarily on unobservable characteristics. In other words, there is no meaningful change in the observable characteristics of loans made by treatment group banks relative to the control group banks around the CRA exam. This, again, is to be expected under our interpretation of the results, since banks have an incentive to convince regulators that loans extended to meet CRA criteria are not overly risky” (ABBS 2012: 3).

It is important to emphasize that if the authors’ are correct, then the implications of the lenders’ deceit and the requirement to communicate the twin orders from senior management are of immense importance and go far beyond the matters the ABBS study discusses. Those implications falsify the authors’ thesis because they expose an internal logical inconsistency. The authors don’t seem to realize it, but they are describing a “control fraud” and they are claiming that such frauds are sufficiently common among our largest banks to drive their empirical results. I agree with that conclusion and Dr. Seru has agreed with it as I explained in his, and his co-authors’ recent February 2013 study on frauds by our largest banks in the sale of mortgages to the private label secondary market. That study, as I noted, had empirical evidence from the lenders’ own files that they typically knew when there was a second mortgage lien (because they made both loans), yet they frequently misrepresented to the buyer that there was no second lien when they sold their mortgages to “private label” packagers of RMBS. The CRA study, however, has no direct evidence that the bank’s controlling officers knew that the loans to CRA census tracts were unduly risky and that they deliberately deceived the regulators to prevent them from recognizing that fact in order to enhance their CRA ratings.

When a bank official, in the course of his employment, knowingly makes a false disclosure to a federal banking regulator with the intent to deceive the regulator he (and the bank) commit a felony. The authors have described a control fraud and asserted that such frauds were common among our largest banks.

Consider how many people would have to get the twin orders from the senior managers under the authors’ assertions. The order has to go to the loan officers, but it also has to go to the loan brokers (because the largest banks that drove the authors’ results frequently lent – particularly to CRA census tracts and lower income borrowers in non-CRA census tracts – through brokers. They typically lent through brokerage firms that, collectively, employed tens of thousands of brokers. The twin orders also have to go to all internal controls at the bank (and that means thousands of employees and officers at the largest banks). If you don’t take the internal controls off line they will block the risky loans. The largest banks would also have to deceive their external controls (the auditors, attorneys, credit rating agencies, the SEC, and the banking regulators). This means that all the bank officers who deal with those external controls must be alerted to the twin orders and instructed to deceive the external controls. (It is a poor scam to lie to the banking regulators and then disclose the true facts about the asset quality in one’s 10K.) Similarly, one must instruct the bank officers who determine the Allowance for Loan and Lease Losses (ALLL) not to boost the ALLL because the bank is making riskier loans. All the largest banks (or their holding companies) are publicly-traded, so we can securities fraud to the list of crimes the authors (implicitly) assert the largest banks committed.

But that is only the first stage of the fraud, for the banks that made imprudent, risky loans frequently sold those loans to the secondary market. They could, of course, disclose that those loans were unduly risky and sell them at a large loss. No one thinks that happened, and the 2013 study on lender misrepresentation confirms an enormous amount of evidence that the lenders engaged in pervasive deceit involving false representations and warranties inflating the safety of the loans they were selling.

The authors of the CRA study claim that the largest banks were superb at deceit. Here is the relevant passage from the longer quotation above. I find that claim dubious. As I explain below, the authors found a significant rise in liar’s loans, which is not only “observable” but among the best indicators of accounting control fraud.

“[T]he reduction in loan standards associated with elevated lending around CRA exams is based primarily on unobservable characteristics. In other words, there is no meaningful change in the observable characteristics of loans made by treatment group banks relative to the control group banks around the CRA exam.”

In plain English, the authors are asserting that the CRA leads to banks making significantly riskier loans, but disguising the fact so cleverly that there are virtually no “observable characteristics” that distinguish the much riskier loans from the much more prudent loans. This claim also leads to the question of how lenders are able to discern the unobservable – particularly when they loan largely through loan brokers at remote sites who have some of the most perverse incentives in financial history to make fraudulent loans. A loan officer who never sees or speaks to the borrower and never verifies the borrower’s income should find it exceptionally difficult to discern that which the authors state cannot be observed from any review of the loan file. Assuming arguendo that the authors are correct, their claim would provide an explanation for why so many of our “most reputable” fraudulently sold fraudulent mortgages to the secondary market – they were uniquely able to identify fraudulent loans even in the absence of “observable characteristics” allowing others to identify bad loans. (One of the typical reps and warranties in such sales to the secondary market was that the mortgages were not fraudulent.)

The authors’ claim about how the twin orders were communicated is simply asserted

The next question is how the authors assert that the controlling officers’ twin directives (make bad loans in CRA census tracts – and deceive the regulators (and myriad others) into believing that the loans are safe – were communicated to everyone that had to receive, understand, and comply with the twin directives. As I have explained, in the case of the large banks this would require communications to tens of thousands of people, most of them loan brokers who were not bank employees. One possibility is that the controlling officers sent everyone a memorandum containing the twin directives. Under the authors’ thesis there would have to be hundreds of thousands of memoranda by the 49 largest banks over the decade-long time period of their study directing loan officers to increase CRA lending during specific time periods “around the CRA exam” – and no such memos in other time periods. Or perhaps the authors believe that the same memorandum gave a third directive – this memorandum is only controlling “around the CRA exam.” Such memoranda would normally be reviewed by the CRA and the safety and soundness examiners, so the memoranda, and the management directives, would have had to been kept secret by over one hundred thousand people and that secret would have to last for over a decade. The authors actually appear to believe that there were such memoranda.

“For example, after management issues an informal companywide directive to increase lending, it may take time to reverse direction, with the result that lending may remain elevated for few more months” (ABBS 2012: 15).

The authors do not quote any such directive, assert the actual existence of any such directive, or quote anyone as confirming the existence of such a directive. The authors are also, at best, incomplete about the hypothetical “informal companywide directive” for if their thesis were correct the direction could not be “increase lending.” The directive would have to be the twin directive: because we wish to enhance our CRA rating, you are directed to increase bad, often fraudulent, lending and hide that fact from the regulators and many other parties. Such a directive would be a “smoking gun” in fraud actions and would be cited in many complaints seeking to impose liability. The authors cite no such directive.

The reality, which the authors would probably agree with if they focused on the issue, is that these messages are sent through compensation packages and the suborning of internal and external controls. The message they sent, however, was the usual accounting control fraud recipe, typically compounded by the subsequent fraudulent sale of the mortgage to the secondary market. The CRA had nothing to do with the message. Predatory lending, however, did have an important role. Consider the most basic level of predatory lending that is typically ignored. By 2006, roughly 40% of the loans made that year were liar’s loans. A liar’s loan was a terrible deal for any honest borrower because it carried a significantly higher interest rate. (The self-employed can use their tax returns to secure a cheaper loan.) Only fraudulent lenders made substantial amounts of liar’s loans. It was typically the lender who put the lies in the liar’s loans. The optimal loan was one that appeared far safer (because the lender or its agents inflated the borrower’s income and the house appraisal) and charged a premium interest rate (because it was a liar’s loan and the borrower was less financially sophisticated or could not read English well). It makes a great deal of sense for a fraudulent lender to target census tracts with less wealthy borrowers because they are less likely to be financially sophisticated and read English well. What the authors are observing is not the impact of the CRA. It is likely that the study shows one of the impacts of accounting control fraud.

Other methodological weaknesses

We examine and regulate small banks much differently than we do huge banks, and the examination is far shorter for smaller banks. We have “resident” examiners located at our largest banks. Examination is constant at such banks and regulators’ scheduling of specific examinations is likely to track idiosyncratic concerns about the each of the largest banks. Regulatory agencies differ – dramatically – so there can be no assumption that the regulators (even though subject to the same examination frequency guidance) act uniformly – particularly with regard to the largest banks. The regulators are not randomly distributed – any result largely driven by the largest banks is actually dealing disproportionately with a single regulator (the Office of the Comptroller of the Currency (OCC)) or with the combination of two regulators (the OCC and the Federal Reserve), which introduces even more varied scheduling “in practice.”

The FDIC audit of its examination process of “community banks” (< $1B in assets) found so few unsatisfactory compliance ratings (“few, if any”) that they did not bother to present data. Even the weakest “Satisfactory” rating (3) was rare (“ranging from 19 to 73” banks) (p. 13 & n 11).

“Moreover, since the exam itself is not an instantaneous event, with a significant time gap between the exam initiation and the submission of the final CRA report, banks also may choose to elevate their lending a few quarters after the CRA exam” ABBS (2012: 2).

One aspect of FDIC compliance examination timing is regular, and brief. The FDIC audit of compliance examinations (which included CRA in that era) of community banks found that the “elapsed days between onsite examination work and the issuance of the final report did not vary much according to ratings from 2009 forward, averaging about 1 month.” Unfortunately for the ABBS authors’ research design, other timing aspects “vary widely.”

“Compliance Examinations. For purposes of presenting meaningful information pertaining to the timeframes for compliance examinations, we excluded timeframes for CRA performance evaluations (which are rarely conducted as stand-alone assignments) and instances where compliance examinations and CRA performance evaluations were conducted together. In addition, we found that the statistics for compliance examinations reflected few, if any, institutions rated 4 or 5 during 2007-2011.10 This caused the turnaround days associated with those ratings to vary widely and did not provide sufficient information for comparison and analysis purposes. As a result, we omitted this data from our analysis.

Figure 4 shows wide variability in the average onsite turnaround days for institutions with a compliance rating of 3, which we attribute to the low level of institutions receiving that rating” FDIC Exam Audit: 13.

The authors’ lack of a prior theoretical basis for their study’s time window

“The identifying assumption is that around the time of CRA exams, because of their increased incentive to comply with CRA standards, banks will shift their lending behavior toward borrowers that improve their CRA rating. By law, such borrowers include any applicant that is located in CRA-target tracts – defined as census tracts with a median income less than 80 percent of the median income at the Metropolitan Statistical Area (MSA) level – as well as low-income and minority borrowers in non-CRA-target tracts. Importantly, in comparing lending behavior of banks that underwent CRA exams to those that did not, our analysis always includes tract-by month fixed effects. This battery of fixed effects implies that our regressions are identified through variation between treatment group and control group banks within a given tract in a given month. As such, the fixed effects control for potentially important demand-side variables affecting the lending behavior of all banks within a given tract-month” ABBS (2012: 2).

Note the vagueness of the phrase “around the time.” The authors operationalized the phrase as “the six quarters surrounding a CRA exam,” but there is no apparent logic to that specification. Using arbitrary lags and lead times in such an econometric study is very risky because the temptation is to play around with the specifications until something “works.” That is precisely what the authors did. They admit that they were not operationalizing and hypothesis testing any theoretical basis for their six-quarter window.

“We do not have a prior on how many quarters before and after the lending stays elevated. We experiment with more than three quarters on either side of the CRA exam and find that the elevated lending occurs only from three quarters before the CRA exam until three quarters after the exam” (ABBS 2012: 16 n. 17).

Nevertheless, the authors purport to prove causality and quantify the lower-bound cost of what they purport to be the CRA’s perverse incentive to make bad loans.

“First, we find that in the six quarters surrounding a CRA exam, lending by treatment group banks (i.e., those undergoing the CRA exam) is elevated on average by 5 percent as compared to lending by control group banks. Second, using data that track loan performance, we show that loans originated by treatment group banks around CRA exams are 15 percent more likely to be delinquent one year after origination as compared to those originated by control group banks. The evidence therefore shows that around CRA examinations, when incentives to conform to CRA standards are particularly high, banks not only increase lending rates but appear to originate loans that are markedly riskier” (ABBS 2012: 3).

“We find evidence of elevated lending by treatment group banks around the CRA exam during the 2004–2006 period. Moreover, our results show that the differential performance of loans originated by treatment and control group banks around CRA exams is particularly strong during the 2004–2006 period. This coincides with the time period when private securitization boomed and might therefore reflect an unexplored channel through which the private securitization market induced risky lending in the economy” (ABBS 2012: 4).

The authors’ unwilling to explore alternative explanations and their dubious time window

Or, they could be finding that liar’s loans were pervasively fraudulent, expanded massively between 2003 and 2006, and that accounting control fraud expanded greatly during that period. Note that the authors do not explore alternative explanations, but latch on to their desired story and interpret everything they find as supporting their thesis without discussing alternative explanations. They did not even mention the alternatives, much less analyze them.

But consider as well the author’s semi-alternative explanation – lenders made far more fraudulent loans because they learned that they could fraudulently, and profitably, sell such loans to the secondary market by making false reps and warranties. Because the CRA’s enforcement start at a weak base level and became progressively weaker during the time period covered by the authors’ study, the rise of fraudulent sales by mortgage originators to the secondary market would be a superior explanation for why bad loans became more common and bad, fraudulent loans increasingly became terrible, fraudulent loans during 2004-2006. The epidemic of fraudulent lending also hyper-inflated the bubble and led to a host of loan characteristics (e.g., “pick-a-pay” mortgages) designed to delay defaults on the fraudulent loans. Collectively, all of these factors pushed towards greatly increased losses.

Note the strength of the authors’ adjectives and conclusions and the weakness of their logic and reliance on conclusions in this key passage that explains their overall theory.

“Regulators use this compliance information when approving applications for new bank branches as well as for mergers and acquisitions. Banks therefore have a strong incentive to concentrate their CRA-compliant lending before an exam date to ensure a satisfactory CRA evaluation while minimizing the likelihood that such lending would fail the “safe and sound” criteria at the time of the exam. Moreover, since the exam itself is not an instantaneous event – there is a time gap between the exam initiation and the submission of the final CRA report – banks may also choose to elevate their lending a few quarters after the CRA exam” (ABBS 2012: 9).

The “therefore” in the second sentence asserts that the first sentence logically demonstrates the validity of the second sentence. It does not. First, they have no basis for the use of the word “strong.” Second, the fact that compliance is considered during applications for acquisitions does not mean that there is any incentive (much less a “strong” one) for the bank to make bad loans because a CRA examination might occur or has occurred recently. Under the Bush administration, which is essentially the authors’ study period, CRA ratings were virtually always mildly positive, acquisitions (much less branches) were almost invariably approved, and very little happened even when banks were openly hostile to the CRA. Third, as I have explained there was no incentive for an honest bank to make bad loans. There was a strong incentive for control frauds to make bad loans, but that had nothing to do with the CRA. Fourth, the six quarter CRA window has no logic. They authors do not explain why they chose the window on either side of the exam, beyond “banks may also choose to elevate their lending a few quarters after the CRA exam.” Yes, and they “may” not, and they “may” choose to do so seven months before and 45 days after. A bank examination uses an “as of” date – a date prior to the day the examination begins. It will typically be a month-end date sufficiently prior to the exam start date that the data will be available for that month. The “as of” date can often be roughly 45 days prior to the exam start date. Normally, the examiners do not consider data that relate to events after the “as of” date (unless it deals with a major loss). This suggests that the authors’ guess that banks “may” increase bad loans four-to-eleven months after the “as of” date of the CRA exam in order to try to improve their CRA rating would rarely be true. This is a fatal problem for the authors’ methodology and conclusions. Consider this prediction: “we predict that in response to a CRA exam, banks will shift lending toward these tracts” (ABBS 2012: 12). The key phrase is “in response.” That phrase purports that the study can demonstrate causality, which this study design inherently cannot do. Further, there is no valid basis for assuming that loans made after the CRA exam’s “as of” date were made “in response” to the CRA exam.

The other side of the window is at least conceivable, but it too is unsupported. Why nine months before the CRA exam instead of one month or 18 months? There is no theoretical or practical basis for the authors’ window. Note that the authors are claiming that the officers that control the bank are engaged in a deceitfully clever form of fraud. They know that the loans they make to enhance their CRA status will are “markedly riskier” and will inflict markedly greater losses on the bank. They know that these markedly greater defaults will show up with a lag, so they start making the “markedly riskier” loans only nine months before the CRA examination in the hopes that the defaults will not occur during the CRA exam, which would violate the “safe and sound” lending concept that is one of the CRA’s pillars.

There are multiple problems with this analytical approach. First, it appears to be made up. The authors present no direct evidence that the banks’ controlling officers developed such a strategy, directed that their personnel and loan brokers comply with the strategy, and communicated the strategy to the tens of thousands of people that would have to comply with it. Instead, they discuss a hypothetical “informal companywide directive” (ABBS 2012: 15).

Second, the authors do not explain why the banks’ controlling officers viewed making bad loans as desirable. (The accounting control fraud recipe explains why they do so, but the authors do not present the relevant economics and criminology literature.) The lenders could make sound loans in CRA census tracts. The authors do not argue to the contrary, but they appear to implicitly assume that banks could not do so.

Third, the safety and soundness exam is vastly more important to the regulators and the banks than is the compliance exam and safety and soundness examinations are considerably more frequent than compliance exams. By deliberately directing that loan officers and loan brokers make “markedly riskier” loans, the controlling officers are exposing themselves to far more threatening regulatory responses that could place the bank in receivership, result in enforcement and civil actions against the controlling officers, and even prosecution if the officers engaged in deceit. The absence of a sound theoretical basis for the authors’ pre-CRA exam window is a further reason why one cannot conclude that any lending changes were made “in response” to the CRA exam.

The authors double-down on their excessive claims for their study – it not only proves causality; it is only the “lower bound” quantification of the harm induced by the CRA.

“Because the empirical strategy nets out the baseline effect, our estimates of the effect of CRA evaluations provide a lower bound to the actual impact of the CRA” (ABBS 2012: 4).

The authors did find accounting control fraud

Accounting control fraud is what the authors actually found. Recall the passage I quoted from their study in which they said that the differences in lending that produced “markedly riskier” loans and default rates were “unobservable.” The authors note one important exception to that finding.

“One exception in our results is the proportion of low-documentation mortgages originated around the CRA exam. The increase in low-documentation loans is economically large with a coefficient of 0.048 (significant at the 5 percent level), representing an increase in low-documentation loans of 23.4 percent relative to the unconditional mean. The change in low-documentation loans seems consistent with the lending institution’s elevating lending primarily on unobservable dimensions to convince regulators that loans extended to meet CRA criteria are not overtly risky” (ABBS 2012: 19).

Liar’s loans went up – considerably. This effect was likely to have been particularly strong in 2003-2006. Liar’s loans were endemically fraudulent, with a fraud incidence of 90% and investigations have confirmed that it was overwhelmingly lenders and their agents (loan brokers) who put the lies in liar’s loans. The same study that found the 90% fraud incidence in liar’ loans noted two facts of special relevance here. In 60% of the loans the borrower’s income was inflated by more than 50% – making the loan appear to be far lower risk, but also not a good way to appear to be making CRA loans to less wealthy homeowners in non-CRA census tracts. The 2006 MARI study also emphasized to the industry that federal financial regulators were warning the industry against making liar’s loans. One test the authors could run is whether the largest banks undergoing CRA exams did not increase their liar’s loans substantially in non-CRA census tracts. The authors found LTV to be mostly unchanged, but the sharp rise in the frequency of appraisal fraud is likely masked by the authors’ reliance on data that is the fruit of appraisal fraud. Similarly, the authors rely on debt-to-income ratios that are also the product of fraud by lenders and their agents acting through fraudulent liar’s loans.

Accounting control fraud was particularly common among the largest banks and exams could trigger increased fraud by those banks

The very limited associations that the authors found at the Nation’s largest banks could be explained by the incentive accounting control frauds face when they are examined. Lenders engaged in accounting control fraud have an incentive to increase their fraudulent and predatory lending when they are facing an examination. As Akerlof and Romer (1993) and criminologists have long recognized, accounting control fraud produces several “sure things.” It produces record (albeit fictional) short-term reported bank income and great wealth to the officers. But high reported bank income also provided something of immense importance in terms of examinations and potential regulatory actions based on adverse examination findings – effective immunity. Citigroup was banned during part of the authors’ study period from merging – because its operations and internal controls were so pathetic that it was in an unsafe and unsound condition. Senior bank officers were well aware of the fact that high reported bank income was the “trump card” they could play to fend off the unusual vigorous regulator.

“Richard Spillenkothen, the Fed’s director of Banking Supervision and Regulation from 1991 to 2006, discussed banking supervision in a memorandum submitted to the FCIC: ‘Supervisors understood that forceful and proactive supervision, especially early intervention before management weaknesses were reflected in poor financial performance, might be viewed as i) overly-intrusive, burdensome, and heavy-handed, ii) an undesirable constraint on credit availability, or iii) inconsistent with the Fed’s public posture’ (FCIC 2010: 54).

Indeed, in the same June 2005 Federal Open Market Committee meeting described earlier, [a] participant noted the risks in mortgage securities, the rapid growth of subprime lending, and the fact that many lenders had “inadequate information on borrowers,¨ adding, however, that record profits and high capital levels allayed those concerns (FCIC 2010: 171).

In [the case of WaMu] … and others that the Commission studied, regulators either failed or were late to identify the mistakes and problems of commercial banks and thrifts or did not react strongly enough when they were identified. In part, this failure reflects the nature of bank examinations conducted during periods of apparent financial calm when institutions were reporting profits (FCIC 2010: 307).

Senior supervisors told the FCIC it was difficult to express their concerns forcefully when financial institutions were generating record-level profits. The Fed’s Roger Cole told the FCIC that supervisors did discuss issues such as whether banks were growing too fast and taking too much risk, but ran into pushback. ‘Frankly a lot of that pushback was given credence on the part of the firms by the fact that – like a Citigroup was earning $4 to $5 billion a quarter. And that is really hard for a supervisor to successfully challenge. When that kind of money is flowing out quarter after quarter after quarter, and their capital ratios are way above the minimums, it’s very hard to challenge’ (FCIC 2010: 307).

The Commission concludes that the banking supervisors failed to adequately and proactively identify and police the weaknesses of the banks and thrifts or their poor corporate governance and risk management, often maintaining satisfactory ratings on institutions until just before their collapse. This failure was caused by many factors, including beliefs that regulation was unduly burdensome, that financial institutions were capable of self-regulation, and that regulators should not interfere with activities reported as profitable” (FCIC 2010: 308).

Conclusion:

The authors assert that the Nation’s largest banks sought to make “markedly riskier” loans while deceiving their regulators and external controls and secondary market customers into believing that the loans were “safe and sound.” The authors’ study design cannot prove causality. The mechanism they posit – the more frequent CRA examinations become the more bad loans they make – must mean that bad loans declined during the ’00 decade because the CRA exam frequency and rigor fell sharply after 1999. The authors fail to explain why the controlling bank officers chose to make “markedly riskier” CRA loans rather than “safe and sound” CRA loans in accordance with the statute. The authors’, unexplained, assertion is that CRA examinations caused the senior leadership of the Nation’s largest banks to send “guidance” throughout the enterprise and the tens of thousands of loan brokers who work with the largest banks – for over a decade. The guidance contained three directives: (1) increase materially the number of loans eligible for CRA compliance, (2) make those loans “markedly riskier”, and (3) hide the increased risks from the regulators, investment markets, and customers. The authors argue that the Nation’s largest banks are control frauds and that they successfully deceived the regulators and other external controls. The authors assert that up to the 45 largest banks issued such guidance memoranda to tens of thousands of people every two years. The authors, however, are unable to cite such memoranda or find any source among the hundreds of thousands of bankers who worked for the control frauds or there brokers who says that they were ever given such guidance. The authors do not even attempt to explain why the senior bank officers made bad loans and engaged in deception rather than make the “safe and sound” loans that the CRA encourages.

Contrary to the authors’ claims, they do not prove causality and their model is inherently incapable of proving causality. They have picked lags and leads not pursuant to any theory, but only after seeing the observed data. A better explanation for the association they have shown in the lending by the largest banks is that they were accounting control frauds who wished to report stellar profits to help fend off any regulatory reforms.