Review of Predatory Lending Study

A North Carolina-based nonprofit advocacy group, the Center for Responsible Lending (CRL), has released a study claiming to show that North Carolina's 1999 anti-predatory lending law saved consumers $100 million on home mortgages originated during the year 2000 (See report by Keith Ernst, John Farris, and Eric Stein, "North Carolina's Subprime Home Loan Market After Predatory Lending Reform," Center for Responsible Lending, Durham, NC, August 13, 2002). The authors evaluated data reported under the Home Mortgage Disclosure Act (HMDA) from 1998-2000 and concluded that subprime mortgage lending continued to "thrive" in North Carolina after passage of the statute and low-income borrowers continued "to have access to a wide range of choices when selecting a home loan."

North Carolina Governor Michael Easley and Attorney General Roy Cooper hailed the study in a joint press conference in Raleigh. Easley remarked that if other states followed North Carolina's lead in clamping down on predatory lending practices, consumers could save billions of dollars, nationwide. Cooper said the law hasn't hurt the availability of subprime credit in the state, despite the claim of financial services industry lobbyists during the legislative debate in the state's General Assembly.

A careful review of the CRL report reveals quite a different story. The evidence presented does not support, and often contradicts, the report's conclusions. In fact, the data presented are consistent with a decline in subprime lending and a reduction in the availability of mortgage credit to low income borrowers in North Carolina following enactment of the anti-predatory law, despite the author's claims to the contrary. Moreover, the calculations of the dollar value of consumer benefits are largely asserted, and do not derive from the same database used to evaluate the effects of the statute on loan volume. In short, no convincing evidence is offered to show either a reduction in "predatory lending" or net benefits to consumers from restrictions on credit terms.

The following sections take the study's three "key findings" and evaluate the evidence offered by CRL to determine if the facts support the conclusions.

CRL Study Methodology and Data

At the outset it should be noted that all of CRL's conclusions regarding the volume and composition of subprime mortgage lending in North Carolina are based on HMDA reports filed for 1998-2000. Even though this database covers 28 million loans ($3.3 trillion) originated across the U.S. over the three-year period, housing market researchers recognize that HMDA data have serious weaknesses when used to analyze subprime lending. Since CRL relies exclusively on HMDA data in drawing its conclusions it is helpful to review the limitations of the data.

First, HMDA data do not identify a particular mortgage loan as subprime. The Department of Housing and Urban Development (HUD) annually produces a list of lenders believed to be predominately subprime lenders. Studies that examine subprime lending using HMDA data must assume that loans made by lenders on the HUD subprime list are actually subprime loans. However, many lenders designated as subprime lenders also make prime loans. Conversely, many lenders who are not considered predominately subprime lenders nevertheless make large numbers of subprime loans. These loans would be overlooked in any subprime analysis based on the HUD list.

Finally, there are many institutions that are not required to report under HMDA, and their lending activity is not reflected in the HMDA data at all. Currently, a non-depository institution must report under HMDA only if its annual lending for home purchase and refinancing equals 10 percent or more of the dollar value of its loan originations (mortgage and non-mortgage). Many non-depository institutions (e.g., consumer finance companies) have long specialized in making loans of all kinds to "subprime" borrowers. Subprime mortgage loan originations from many of these companies are also missing from the HMDA database.

As a consequence, use of the HUD subprime list in conjunction with the HMDA data produces, at best, a very rough approximation of subprime mortgage lending. Of course, researchers use the HUD/HMDA approximation of subprime lending because there are few alternative sources of data on subprime mortgage activity. Nevertheless, any study that attempts to analyze changes in subprime origination volume using the HUD/HMDA data should temper its findings with a full disclosure of the limitations of the data. No such disclaimer appears in the CRL study, thereby encouraging the reader to place more confidence in the study's conclusions than is warranted.

Moreover, the HMDA data do not include any information on loan pricing or borrower risk characteristics (other than income). Consequently, the HUD/HMDA subprime data cannot be used to identify changes in underwriting standards, e.g., tightening of credit to higher risk borrowers. Of course, such changes are precisely the adjustments that subprime lenders would likely make when confronted with a new statute that imposes restrictions on high-cost loans.

CRL Finding #1:
Subprime Lending Continues to Thrive in North Carolina

CRL used the HMDA data to identify several measures of subprime loan activity in North Carolina, relative to the rest of the United States. They employed a comparative approach, and inferred a continued "thriving" subprime market because subprime activity per capita was higher in North Carolina, relative to the U.S. average, in the first full year following the July 1, 1999 passage of the state's anti-predatory lending statute.

The HUD/HMDA data show that "home loan borrowers in North Carolina were 20% more likely than borrowers in the rest of the nation to receive a subprime loan in 2000." In addition, North Carolina had 15% more subprime loans per capita than the rest of the nation in 2000." These observations apparently settle the issue according to CRL.

Of course, a higher incidence of subprime activity in North Carolina relative to the U.S. average during 2000 tells us nothing about the impact of the North Carolina statute that was passed the previous year. There is ample evidence that the size of the subprime market in North Carolina is substantially larger than the U.S. average. For example, impaired credit history is a common characteristic of the subprime borrower. Figure 1 illustrates that the percent of borrowers who are seriously delinquent (90+ days past due) is much higher in North Carolina than the U.S. median across all types of loans.

Figure 1

Printer-Friendly Chart

With a substantially higher proportion of its borrower population having impaired credit histories, North Carolina could continue to exhibit a higher-than-average incidence of subprime lending even if the 1999 statute caused creditors to decrease the supply of subprime loans. Consequently, the question the CRL researchers should have asked is whether subprime lending in North Carolina fell, relative to the U.S. average, following passage of the statute.

Figures 2 and 3 display the data cited in the CRL report and reveal that both of their measures of subprime activity in North Carolina relative to the U.S. did decline. In 1998, the number of subprime loans per 10,000 residents was 35% higher in North Carolina than the U.S. average; by 2000, it was only 15% higher (Figure 2). Similarly, the proportion of all mortgage loans originated in North Carolina that were subprime was 33% higher than the U.S. average in 1998; by 2000 it was only 18% higher (Figure 3).

Neither of the trends in the two figures supports a conclusion of a "thriving" subprime market in North Carolina. In particular, Figure 2 shows that while subprime originations appear to have been declining nationwide during 2000, the decline was larger in North Carolina.

Figure 2

Printer-Friendly Chart

Figure 3

Printer-Friendly Chart

CRL Finding #2:
Choice remains unfettered: North Carolina borrowers continue to have a wide range of choices when selecting a home loan.

Evidence offered to support this conclusion is the observation that "every lender (21 in all) with more than one percent of the 1999 North Carolina subprime market that reported new lending anywhere in the United States reported originating new loans in North Carolina in 2000." (Ernst, Farris, and Stein, 2002, p 5). It is true that all 21 lenders remained active in the North Carolina market, but this tells us little about the borrowers to whom they were lending.

On this point the shortcomings of the HUD/HMDA data are most apparent. As discussed above, not all loans made by the HUD/HMDA subprime lenders are subprime. There is simply no way to tell from the HMDA data the percentage of originations during 2000 that were made by these 21 lenders to subprime borrowers. It is entirely possible that these lenders continued to make mortgage loans in North Carolina but shifted their lending mix toward a higher proportion of prime loans and smaller proportion of subprime loans. Consequently, a wide range of choices may have continued to be available to borrowers in general, but subprime borrowers may have found that some of these lenders no longer welcomed their business. The HMDA provide no information one way or the other.

Further, because the HMDA data do not contain information on borrower risk characteristics, the data can't be used to detect shifts in underwriting guidelines that might have been triggered by the 1999 statute. For example, an economic model of a lender's reaction to higher costs of servicing higher-risk borrowers (consequent to passage of the 1999 North Carolina statute) would predict that lenders would reduce the supply of loans to higher-risk borrowers. A lender could remain active in the subprime market, but set higher acceptance standards, so that borrowers with higher FICO risk scores would be denied loans, even though borrowers with the same risk profile were accepted prior to the 1999 statute. Again, the HMDA data can't be used to confirm or refute this prediction.

CRL Finding #3:
Reductions in predatory lending saved North Carolina homeowners more than $100 million in 2000.

The CRL authors calculate two components of the savings to North Carolina borrowers from a reduction in predatory lending. The 1999 statute banned some loan contractual features, most notably the sale of single-premium credit insurance on all loans and prepayment penalties on loans smaller than $150,000. In addition, the act may have prevented some predatory loans from being made at all, presumably saving the would-be victims from excessive charges. The authors do not provide convincing estimates of net savings for either component.

In explaining their first two major findings, the CRL authors failed to note that the HMDA data signal a decline in subprime lending in North Carolina during 2000. Possibly this failure was because such an observation did not support their claim of a healthy subprime market in North Carolina. However, to be able to show savings to North Carolina borrowers resulting from a decline in predatory lending, they need to observe a reduction in subprime loans. Only in support of Finding #3 do they acknowledge the decline in originations per capita. By comparing ratios of North Carolina subprime borrowing relative to the U.S. before and after the 1999 statute, the authors concluded "the data show that the North Carolina subprime market declined 6.6% beyond that predicted solely by the relative decline in other markets."

The next step in the authors' calculation of borrower savings is pure assertion. The authors state "this 6.6% net real decline is consistent with a significant reduction in predatory lending." They translate the 6.6% reduction into a calculated decline of 2,700 loans, all deemed predatory in the subsequent calculations of dollar savings. They provide no further evidence in support of this conclusion. To repeat, the authors assert that all of the reduction in subprime lending observed in North Carolina in the year following passage of the 1999 statute was a decline in predatory lending. Of course, this assertion totally ignores a variety of other possibilities, including the possibility that subprime lenders may have raised underwriting standards so that some previously creditworthy borrowers (possibly 2,700 such borrowers) were denied loans.

As for the calculated dollar savings from banned loan features, these do not hold up well under closer scrutiny. None of the data used to support the savings from banned contractual features derives from the HMDA data. Consequently, the calculations are not based on a careful study of the loans on which the previous conclusions are based.

If they do not inspect a representative sample of loans (originated in North Carolina or elsewhere), then how do the authors calculate savings to borrowers? One example from their report illustrates their methodology.

Savings from the ban on Single Premium Credit Insurance: The 1999 North Carolina statute bans the sale of single-premium credit insurance (SPCI) on all mortgage loans (prime and subprime). However, the authors focus only on the resulting savings to the borrowers holding the 31,500 subprime loans made in North Carolina during 2000. Based on conversations with credit insurance industry representatives regarding SPCI product penetration rates and average premiums, the authors projected that 20% of the 31,500 subprime loans originated in North Carolina during 2000 (i.e., 6,300 loans) would have had SPCI in the absence of the 1999 statute. Further, they estimated a total premium cost of $6,600 per loan with SPCI. Multiplying $6,600 per loan by 6,300 loans yields a savings of $41.6 million to North Carolina borrowers.

Set aside for the moment the issue of whether the assumed penetration rate and premium cost per loan are reasonable estimates. The author's calculation makes a fundamental and unstated assumption. For the calculation to be correct, all of the premium paid for SPCI must be a waste—no net benefit or value to the borrower whatsoever. In essence, the authors assume that any payment for credit insurance is equivalent to paying something for nothing. Of course, this assumption is inconsistent with the observation that tens of millions of borrowers have voluntarily purchased the single-premium credit insurance policies that are in effect across many types of consumer loans in the U.S. It is also contradicted by the observation that borrowers who are offered credit insurance on a monthly premium basis continue to purchase the product.

The ban on Single Premium Credit Insurance provides nearly 42% of the authors projected savings to North Carolina borrowers from reducing predatory lending. The remaining savings derive from applying similar methods to estimate incidence and dollar cost of 1) flipped refinances with "no net benefit to the borrower", 2) cutting "excessive" fees, 3) prohibiting prepayment penalties on loans < $150,000.¹ Closer inspection of each of these calculations reveals a pattern of basing dollar savings estimates on incidence rates inferred from small "convenience" samples. Even more troubling from a methodological standpoint, the authors repeatedly rely on the (unstated) assertion that a ban or limitation placed on a particular contractual feature triggers no additional cost or loss of benefit to the borrower.

Because their methods do not rely on sound statistical procedures, there is little or no scientific justification for the authors' calculations. In the end, the estimates of consumer savings are merely assertions.

¹For a more detailed discussion of the source of the estimates underlying the dollar savings calculations see Eric Stein, "Quantifying the Economic Cost of Predatory Lending," Center for Responsible Lending, Durham, NC, October 30, 2001.