Group letter on CFPB’s RFI on HMDA


January 21, 2022

RE: Request for Public Comment, HMDA Assessment, Docket No. CFPB-2021-0018

To Whom It May Concern:

The 54 undersigned organizations extol the value of the new and enhanced Home Mortgage Disclosure Act (HMDA) variables added by the 2015 final rule issued by the Consumer Financial Protection Bureau (CFPB). The Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 (Dodd Frank) required the CFPB to enhance HMDA data by adding a number of variables regarding loan terms and conditions and borrower demographics. The CFPB also used its discretionary authority under Dodd Frank to add more data points such as debt-to-income ratios, reverse mortgages, and characteristics of multifamily and manufactured home lending.

In 2020, the CFPB unwisely raised the reporting threshold from 25 to 100 closed-end loans and from 100 to 200 open-end loans. The undersigned organizations maintain that this rule change violated the statutory and regulatory purposes of HMDA data to assess if lending institutions are meeting needs for housing-related credit and are not discriminating in the provision of credit.

In response to the CFPB’s request for information (RFI), this comment letter will:

  • Affirm the benefits of HMDA data in increasing access to responsible credit and guarding against predatory lending;
  • Describe the harm of the raise in the reporting threshold in terms of subverting HMDA’s statutory purposes;
  • Discuss how the benefits of HMDA data vastly outweigh its costs;
  • Offer suggestions for the design of the CFPB’s assessment including evaluating the impact of the Economic Growth, Regulatory Relief and Consumer Protection Act of 2018 (EGRRCPA) which provided partial exemptions from reporting several new Dodd Frank data points for banks with small loan volumes. We believe this reduces the transparency and equity of the lending marketplace particularly in smaller cities and rural areas where these banks are concentrated.
  • Outline specific recommendations for improving HMDA data points. In particular, the CFPB should contemplate aligning the racial/ethnic and gender categories in HMDA with the proposed categories in the Section 1071 small business/farm data rule. The CFPB should also add the disability status of the applicant to HMDA data.
  • Describe a number of suggested improvements to the data fields including enhancing multifamily lending data, improving the accuracy of loan amount and property value, and adding an Annual Percentage rate (APR) to help compare loan pricing across products and lenders.
  • Improving the rigor of enforcement to ensure accurate data. We describe significant inaccuracies in pricing data and how these can be better identified and rectified.
  • In response to public input, the CFPB improved the accessibility of HMDA data from its website. However, more improvements are needed particularly for data users that are not researchers. The CFPB should make disseminating easy-to-use data a higher priority if HMDA is to more effectively serve its purpose of enabling the public to hold lenders accountable for serving community needs.

Background: HMDA data updated to provide more accurate housing needs assessments and guard against abusive lending

The motivation behind the data improvements required by Dodd Frank was that more transparency in the form of publicly available data was needed to prevent widespread predatory lending that was largely responsible for the financial crisis. By providing information on loan terms and conditions, the data would help regulatory agencies, community groups, and other stakeholders detect increases in abusive lending and take steps to curb such lending before it caused another crisis.

In addition, stakeholders agreed that more information was needed to ensure that lenders were serving the housing and credit needs of vulnerable and/or underserved subgroups within the general population. The CFPB updated multifamily data reporting requirements to ascertain whether lenders were serving the needs of renters in a housing market that was posing challenges for increased homeownership and affordability for renters and homeowners alike. Likewise, lawmakers and the CFPB reacted to the abuses unscrupulous lenders inflicted upon older adults by requiring age to be a variable in HMDA data and information about reverse mortgage lending in HMDA.

Confronted by the financial crisis, the response of Congress in passing Dodd Frank and the CFPB in rigorously implementing Dodd Frank to update HMDA data was not only consistent with the 1975 HMDA law but enhanced the ability of HMDA to achieve its original statutory purposes. HMDA needed to be updated to take into account changes in the lending industry and the demographic and housing market conditions of the country. Only with an update could the statutory purposes of HMDA be upheld.

The statutory purposes are to assess whether lenders are meeting the housing needs of local communities, inform public sector investment decisions, and to detect and prevent discrimination. HMDA data was becoming less informative to assess whether sophisticated loan terms and conditions introduced in the late 1990s and 2000s were too complex to understand or imposed too many costs on vulnerable borrowers. Since complex lending was becoming abusive and deceptive, lenders were not meeting credit and housing needs while the HMDA data, absent an update, could not accurately determine whether needs were being met in a legitimate and responsible manner.

The Need to Update HMDA Data: Surge in Abusive Lending

Abusive and high-cost lending was the largest contributor to the worst recession since the Great Depression. Federal agencies, community organizations, and the public at large were disadvantaged in combating the widespread abusive lending because they lacked detailed data on lending practices in the industry as a whole and for individual lenders, including those violating fair lending and consumer protection laws.

In 2009, the Government Accountability Office (GAO) had to rely upon a private sector database instead of HMDA to document the pervasiveness of risky lending for leading members of Congress. This was a couple of years too late. Some notoriously risky and abusive lenders had already ceased operations after inflicting untold harm on consumers. For example, the FDIC placed Washington Mutual in receivership in 2008 as a result of the thrift’s unsustainable adjustable rate mortgage lending. Ameriquest and New Century, non-bank mortgage companies, offering high volumes of risky and high-cost loans ceased operations around the same time. Although HMDA data was available for these lenders, federal agencies were unable to conduct sufficient fair lending and consumer compliance exams because the HMDA data lacked information on loan terms and conditions, which were necessary for the econometric analyses that could have identified discriminatory, unfair, and deceptive practices.

The GAO report documented a tremendous increase in nonprime (subprime and Alt A lending). Nonprime lending increased from 12 percent of all loans in 2000 to 34 percent in 2006.[1] A large share of subprime loans in these years had characteristics associated with default and foreclosure including adjustable interest rates, less than full documentation of borrower income, high debt-to-income ratios (DTI), and high loan-to-value ratios (LTV).[2] The percentage of subprime loans with DTIs over 41 percent increased from 47.1 percent in 2000 to 59.3 percent in 2007.[3] Over an eight year-time period, the GAO found that 60 percent or more of subprime loans had prepayment penalties.[4]

Subprime lending had multiple risks layered on top of each other. The loans usually did not just have one risky loan term or condition such as adjustable rates but rather several such as adjustable rates, high DTIs, prepayment penalties, and high fees which confronted borrowers with unaffordable and unsustainable loans. The best way to have discovered the extent of risk layering would have been through a publicly available loan level database such as HMDA, which unfortunately did not contain information on loan terms and conditions during those years. As a result of the abuses and risk layering being undetectable, about one quarter of the subprime loans originated between 2000 and 2007 and that were still outstanding in 2008 were in default or had started the foreclosure process.[5]

At the time, HMDA data provided community groups with information suggesting that problematic lending was pervasive in communities of color and even in middle-income and upper-income communities of color. HMDA data, however, was not comprehensive enough to paint a complete picture of the thoroughgoing nature of the abusive lending. As a result, community groups often felt that they were the proverbial canary in the coal mine that the federal agencies and members of Congress were not listening to with enough seriousness.

In July of 2007, NCRC testified before the House Financial Services Committee regarding the effectiveness of HMDA data in rooting out discrimination.[6] The testimony documented racial disparities in high cost lending. It stated that “the new pricing data (in HMDA instituted in the mid-2000s) assists in uncovering discriminatory pricing, but the new pricing data by itself remains incomplete. Because HMDA data do not allow for the observation of fee gouging or dangerous risk layering involving high loan-to-value ratios and reduced documentation lending, unscrupulous lenders can continue to exploit financially vulnerable consumers. Until HMDA data includes more key underwriting variables and loan terms and conditions, the abusive parts of the industry will be one step ahead of the general public in inventing new methods for deceptive and usurious lending.”[7]

In a retrospective account of the financial crisis, Wachter and Levitin stated that investors of private market securities composed of subprime and other non-traditional loans did not have adequate loan level information in order to properly assess the risk of their investments. In addition, collateral backing the loans was over-valued because of a lack of market-wide data on lending characteristics. The Securities and Exchange Commission (SEC) at the time did not mandate consistent data disclosure to investors and did not require loan level disclosure.[8] In short, none of the stakeholders including the regulatory agencies had adequate data, which lead to a woefully inadequate response to the exponential rise of abusive lending.

The data points in HMDA data mandated by Dodd Frank and added by the CFPB per the discretionary authority are a direct response to the financial crisis. The data points include loan terms and conditions that were associated with widespread abuses. These include comprehensive pricing and fee information, DTI, LTV, prepayment penalties, and adjustable rates. These data points remain critical today in order to monitor and prevent abuses in the lending marketplace.

The Need to Update HMDA Data: Protect Subgroups within Communities of Color and Stop Abusive Lending in Underserved Communities

In addition, the HMDA data race and ethnicity categories required updating because community organizations and civil rights organizations had been reporting surges in foreclosures in Asian and Hispanic communities yet HMDA data was not detailed enough to reveal the extent of abusive lending targeted to subgroups within the Asian and Hispanic community. In response, the CFPB updated HMDA data to include disaggregated race and ethnicity categories for Asians and Hispanics. Combined with the enhancements associated with loan terms and conditions, the data will now help stakeholders protect vulnerable populations against a resurgence of predatory lending.

In general the enhanced HMDA data will be invaluable for fair lending enforcement. Journalists Emmanuel Martinez and Aaron Glantz, authored “Kept Out,” an expose highlighting continued mortgage redlining in 61 U.S. cities, using HMDA data, that resulted in city, state, and federal actions to address pervasive racial disparities in bank branching and mortgage lending patterns.[9] Martinez and Glantz, however, did not have access to the enhanced Dodd Frank data so they were unable to fully document the extent in racial disparities in lending. For example, were disparities in access to loans also accompanied by more onerous terms and conditions offered to people and communities of color?

Since the Glantz and Martinez study, a number of recent analyses reported ongoing disparities by race and ethnicity, including disparities revealed by the new disaggregated data. Martinez did a follow-up study across several metropolitan areas. Accounting for the Dodd Frank variables including debt-to-income and loan-to-value ratios:

We found that lenders gave fewer loans to Black applicants than White applicants even when their incomes were high—$100,000 a year or more—and had the same debt ratios. In fact, high-earning Black applicants with less debt were rejected more often than high-earning White applicants who have more debt.[10]

NCRC found disparities using the disaggregated race and ethnicity data. For example, a study showed that one out of three Hispanic borrowers received a government-insured loan but that 52% of Puerto Ricans did. In addition, the report documented disparities in interest rate by disaggregated ethnic category. These disparities need to be further analyzed so that stakeholders can take steps to narrow them. Moreover the new data showed overall disparities by race or ethnicity. Overall, in 2019, Hispanics paid more to borrow than White borrowers. For instance, in closing costs on 30-year home purchase loans made on single-family owner-occupied homes, Hispanic homebuyers paid an average of $6,521, almost twice as much White buyers, who paid $3,187 at the closing of their mortgage in 2019.[11]

Using the new HMDA data, the CFPB found “Black and Hispanic White applicants are on average denied at a higher rate than non-Hispanic White applicants, even if they are within the same credit score range,” However, “for applicants within the same credit score range Black and Hispanic White applicants on average have higher Combined Loan to Value (CLTV) ratios than non-Hispanic White applicants.”[12] In other words, the new HMDA data is needed to further probe disparities more precisely, which the CFPB did not do in its overview report. In order to more fully complete this picture, the CFPB should release credit score information collected by the agency in some form and allow the public to further investigate these disparities and see whether they remain after controlling for CLTV and other variables.

In the wake of the financial crisis, it is vital to identify and take swift action against racial and ethnic disparities that include unfavorable terms and conditions. Communities of color are often among the first to experience abusive lending so curbing such lending in these communities is of paramount importance to prevent the wide-scale extraction of wealth from communities of color, older adults, and other groups subsequently targeted by predatory lenders.

The Need to Update HMDA Data: Targeting of Older Adults

The National Consumer Law Center (NCLC) and numerous other legal aid societies represented scores of older adults victimized by predatory lenders in the years running up to the financial crisis. NCLC documented how serial refinancing, high costs, high fees, credit insurance added to the loan amount, balloon payments, home improvement scams and other abuses often rendered older adults helpless in defending against foreclosure.[13]

Abusive lenders also exploited older adults with reverse mortgage scams. In a report to Congress, the CFPB documented increasingly risky practices associated with reverse mortgages. For example, the report stated, “Reverse mortgage borrowers are withdrawing more of their money upfront than in the past. In FY2011, 73 percent of borrowers took all or almost all of their available funds upfront at closing. This proportion has increased by 30 percentage points since 2008. Borrowers who withdraw all of their available home equity upfront will have fewer resources to draw upon to pay for everyday and major expenses later in life. Borrowers who take all of their money upfront are also at greater risk of becoming delinquent on taxes and/or insurance and ultimately losing their homes to foreclosure.” Given the rampant abuses in the lending marketplace afflicting older adults, Congress required that age be added to the HMDA data and the CFPB further enhanced the data by adding a reverse mortgage flag. Any diminution of this data will encourage another round of abusive lending as predators will operate under a veil of secrecy.

The Need to Update HMDA Data: Affordable Housing Crisis, Rental Housing, and Manufactured Housing

Some stakeholders have criticized the multifamily data as unrelated to the statutory purposes of HMDA. However, the purpose of HMDA “is to provide the citizens and public officials of the United States with sufficient information to enable them to determine whether depository institutions are filling their obligations to serve the housing needs of the communities and neighborhoods in which they are located and to assist public officials in their determination of the distribution of public sector investments in a manner designed to improve the private investment environment.”[14]

A pressing housing need in communities, given the high cost of housing, is affordable rental housing. Instead of being contrary to the statute, multifamily data in HMDA fulfills a central purpose of the statute. It is also vital in this time period to see if public sector investment in neighborhoods has leveraged multifamily housing lending. Thus, the enhancements to multifamily lending, including the data points on units affordable to lower income tenants, helps stakeholders determine if the public and private sectors, working together, have increased the stock of affordable housing in neighborhoods.

Like HMDA data on multifamily lending, HMDA data on manufactured home lending is vital in order to assess whether lending institutions are responsibly serving the need for manufactured housing. In 2014, the CFPB released a report finding that manufactured housing is disproportionately located in rural areas and residents of manufactured housing are disproportionately older adults with lower incomes and net worth. While manufactured housing holds promise as affordable housing, lenders have engaged in aggressive bait and switch tactics that saddled vulnerable buyers with high cost and unaffordable loans.[15] The CFPB revealed that most manufactured home lending is chattel lending with interest rates 50 to 500 basis points higher than real estate secured loans. Overall, 68 percent of all manufactured home purchase loans were higher cost in 2012.[16] Over time, we hope the HMDA data reduces abusive manufactured home lending by shedding sunlight on this sector.

Raising the Closed End Threshold to 100 Loans Would Enable Hundreds, if Not Thousands, of Lenders to Hide Abuses behind a Veil of Secrecy

The CFPB’s final rule of May 2020 increased the number of loans depository institutions may make before reporting any data under HMDA. Before the rule change, the threshold for reporting HMDA data was 25 closed-end loans, which reduced the number of HMDA reporters by 22 percent when it went into effect after the CFPB’s 2015 HMDA rulemaking. In 2015, the CFPB decided against a higher threshold exempting more lenders stating that, “The Bureau concluded that, if it were to set the closed-end coverage threshold higher than 25, the resulting loss of data at the local level would substantially impede the public’s and public officials’ ability to understand access to credit in their communities.”[17] Inexplicably, however, the CFPB reversed itself and raised the threshold 100 loans.

Raising the threshold eliminated HMDA reporting for thousands of institutions. Using the 2018 data, the CFPB estimated that the HMDA data would lose 107,000 loans in 2018 from the proposed threshold increase to 100 loans.[18] The final 2020 CFPB rule documented that about 40% of depository institutions (1,640 out of 4,120) would be exempted.[19] With the implementation of the 2020 rule, a large number of lenders would no longer be reporting data, thwarting a statutory purpose of HMDA which is assessing whether lenders are meeting housing and credit needs.

In addition, the current regulatory procedure of requiring a depository institution to make at least one single-family loan for home purchase or refinancing before it is even considered as a HMDA reporter must be changed. Even if a lender has not originated a single-family home purchase or refinance loan, it could be a major lender in the geographical areas it serves. For example, if it made 25 or more multifamily loans, it could finance housing for thousands of tenants. In addition, if a lender made several home improvement loans, it is a significant lender in its community but would not be a HMDA reporter due to the threshold counting rules that only consider home purchase or refinance lending. These lenders must publicly report HMDA data so agencies and members of the public can see if they are serving housing needs responsibly and in a non-discriminatory manner.

CRA and Fair Lending Enforcement Weakened by Exempting Lenders from HMDA Reporting

Eliminating HMDA data for several lenders will make fair lending enforcement more difficult and allow predatory lenders to continue their practices. A few years ago, legal aid societies filed a lawsuit against Emigrant Savings Bank. The bank and its affiliates are likely to be exempt from HMDA reporting. For example, Emigrant Funding, an affiliate of the bank reported just 86 applications in its HMDA data in 2017. In 2016, a federal jury found that Emigrant violated the Equal Credit Opportunity Act, the Fair Housing Act, and the New York City Human Rights Law by targeting African-Americans and Hispanics with predatory loans. The bank and its affiliate focused on vulnerable borrowers with credit scores under 600.[20] This case would have been more difficult to pursue in the absence of HMDA data which documented a pattern and practice of targeting people and communities of color.

Raising the thresholds for reporting will also imperil enforcement against unfair and deceptive lending across loan purposes and types. In a report to Congress, the CFPB found that “the reverse mortgage market is increasingly dominated by small originators, most of which are not depository institutions. The changing economic and regulatory landscape faced by these small originators creates new risks for consumers.”[21] Many of these lenders could be exempt from reporting HMDA data in the wake of an increase in the reporting threshold.

Like fair lending and consumer protection enforcement, CRA examination is now more difficult and less able to hold banks accountable because the CFPB increased the reporting thresholds. Frontier State Bank, based in Oklahoma City, Oklahoma, had assets of $607 million. A recent CRA exam recorded 110 HMDA loans but most of these were purchased loans since the bank pursued a strategy of purchasing loans. The bank failed its CRA exam and its lending test, receiving a Needs-to-Improve for the overall rating and on the lending test. It did not make any loans to low- and moderate-income (LMI) borrowers or purchase any loans made to LMI borrowers during 2016, the year of data analyzed by the exam.[22] At more than half a billion dollars, the bank had the capacity to market to and make loans to LMI borrowers but choose against doing so.

If banks like Frontier are exempt from reporting HMDA data in the future, community groups cannot hold them accountable for making good faith efforts to serve LMI borrowers. Examiners will also have a harder time conducting exams since they will now have to go onsite and retrieve data from the bank’s internal files. Now, they can do most of their analysis off-site, using the publicly available HMDA data. CRA will become less effective and more burdensome for small and mid-size banks which remain important sources of credit in large parts of the country.

Reducing Reporting will make it Difficult to Determine if Lenders are Meeting Needs

Raising the thresholds would also imperil HMDA’s statutory purpose of accurately assessing whether housing and credit needs are being met. The CFPB estimated that after raising the threshold to 100 loans, the number of HMDA reported loans would fall by 20 percent or more in 1,200 census tracts. Moreover the percentage of LMI tracts and rural tracts experiencing declines of 20% or more of loans would be double the percentage of all tracts.[23] Members of the public, journalists, academics, and public agencies will not be able to accurately assess whether credit and housing needs are being met in communities across the country in the wake of this significant loss of data.

NCRC also found that a significant number of counties would experience a drop of ten percent or more of applications when the threshold was raised to 100 loans for reporting. Applications are an important measure of impact because fair lending and CRA analysis seeks to uncover whether disparities in denials in addition to originations are present for various geographical areas or groups of borrowers. A significant number of counties (172 or 5.3 percent) experienced a drop of ten percent or more of applications as a result of raising the threshold to 100 loans.

A high number of counties encountered a loss of six percent or more of applications; 334 or 10.4 percent saw a drop of 7 percent or more of applications and 425 or 13.2 percent had a decrease of 6 percent or more. Overall, the ability to accurately assess whether credit needs are being met in about one out of ten counties was substantially compromised by raising the threshold to 100 loans.

Rural counties were disproportionately impacted.[24] Considering counties that experienced a drop of 10 percent or more of applications, 59.3 percent of these counties were rural. For the country as a whole, 40.5 percent of counties were rural. Rural areas generally have less access to banks and credit so disproportionately reducing the robustness of HMDA data in rural counties frustrates HMDA’s statutory purpose to ascertain if credit needs are being met.

Counties in the Midwest and Southwest were disproportionately impacted by large losses in applications as demonstrated via data analysis or visually.[25] Four counties in Texas experienced a decrease of 25 percent or more applications when the reporting threshold was raised to 100 loans. Likewise, three counties in Nebraska and Kansas and two in Oklahoma each saw a drop of 25 percent or more of applications. The interior parts of the country have suffered from either economic contraction or sluggish growth the last couple of decades. These are the regions that need more data on lending and economic conditions, not less.

A recent CFPB analysis concluded that smaller lenders that would be exempt tended to make a higher proportion of their loans in LMI tracts and rural areas. At the same time, more of their loans were to higher income borrowers and for investment properties.[26] While these lenders tended to be more focused on underserved areas, they were not necessarily serving underserved borrowers in these areas and might be neglecting them. Without continued HMDA reporting, monitoring these trends and commenting on CRA exams and fair lending reviews are virtually impossible.

If the CFPB Diminishes Open End Lending Data, a New Round of Abuses and Defaults Will Occur

The CFPB also increased the threshold for reporting open end lines of credit often called Home Equity Lines of Credit (HELOCs). In the years before the financial crisis, HELOC lending was riddled with abuses that resulted in distress and/or foreclosure for large numbers of homeowners. The CFPB stated that investors would purchase homes and would take out open-end loans with high loan-to-value ratios that often ended up in default. In 2008 Congressional testimony, former Comptroller of the Currency John Dugan reported that national bank losses on home equity loans were three times higher in 2007 than 2006.[27] These loans were not visible in the HMDA database before the Dodd Frank enhancements.

According to the CFPB, its permanent increase of the threshold to 200 open end lines of credit resulted in 401 lenders making 69,000 open end lines of credit being exempt from reporting HMDA data. Also, raising the threshold from 100 to 200 loans reduced the number of reporting institutions by 40 percent.[28] This is too many lenders and loans escaping the scrutiny of public and agency review. A repeat of risky practices in vulnerable neighborhoods would thus be too likely to occur.

The CFPB’s review of recent HMDA data illustrated that home equity lines of credit (HELOCs) continued to be a riskier form of lending than closed-end mortgage lending. Multiple risky features were layered on HELOC loans, further increasing the chances of abusive lending and defaults. The CFPB reported that in 2018, 77 percent of HELOC loans were adjustable rate, half feature interest-only payments, and prepayment penalties were present on 28 percent of the loans.[29] These trends continued in 2019 according to the CFPB.[30] In addition, interest rates on HELOCs were high.[31]

Perversely, the CFPB justified its proposal to raise the permanent threshold to 200 open-end loans by stating that open-end lending had increased by 36 percent from 2013 through 2017.[32] However, an increase in open end lending would be a compelling reason to leave the threshold as is so that more lenders and lending would be subject to the visibility that HMDA provides.

CFPB’s Cost Benefit Analysis Showed that Raising the Threshold did not Meaningfully Save Lenders Money

During previous rulemakings, the CFPB conducted various cost-benefit analyses regarding HMDA data collection and reporting. Repeatedly, these cost-benefit analyses revealed that the cost of reporting HMDA data was overwhelmed by the public benefits of the data disclosure. The costs on an industrywide and per lender basis were minimal. In comparison, the public benefits of ensuring that lending institutions were serving credit needs in a responsible manner were immense. For example, Federal Reserve-sponsored research using HMDA data consistently documented that CRA resulted in more lending to people of color and low- and moderate-income borrowers and communities.[33] The CRA impact is most likely due to the public use of data, including HMDA, to examine lenders on their reinvestment performance and to hold them accountable for serving traditionally underserved communities.

More than six months after it issued its 2020 Rule, the CFPB conceded that, in justifying the threshold increase, it had vastly overstated the cost savings that would result for the depository institutions it had just exempted.  Whereas the agency had initially stated the 100 loan threshold would save institutions $11.2 million per year, the actual savings were far lower—only $6.4 million. The error meant that CFPB assumed a savings of $6,588 per lender when the true savings was only $3,764 per lender.[34] Either of these savings were miniscule but the corrected savings figure makes clear that threshold change does more harm than good. A savings of just a few thousand dollars per lender does not justify the loss in transparency and accountability on the part of thousands of institutions.

The CFPB should estimate the reduction in transparency caused by the EGRRCPA partial exemption in its assessment

In a report issued in May of 2021, the Government Accountability Office (GAO) estimated the impact of EGRRCPA exemptions on HMDA data availability.[35] EGRRCPA exempted depository institutions with less than 500 closed-end or 500 open-end originations from reporting several new Dodd Frank data points including pricing information, loan terms and conditions and enhanced data on loan type. We disagree with the GAO that its findings indicated a minimal impact on data availability due to EGRRCPA exemptions.

The GAO documented that lenders eligible for the partial exemption were about 40 percent of all HMDA reporters in 2018 and 45 percent in 2019. The GAO then stated that their lending activity accounted for just 5 percent of loans in HMDA in these years.[36] However, this conclusion overlooked the impact on fair lending enforcement for 40% to 45% of the lenders in the country. Moreover, these lenders are likely to be important in communities such as smaller cities and rural communities that are relatively underserved. Thus, if they are engaging in discriminatory pricing or steering protected classes to loans with onerous terms and conditions, their HMDA data would be useless for detecting these violations.

Similarly, the GAO’s census tract analysis of the impacts of EGRRCPA exemptions likely underestimated the impact on underserved communities. The GAO concluded that 65,103 of 72,951 census tracts had at least 91% of the data available after the EGRRCPA exemptions.[37] However, the GAO then did not present a detailed analysis indicating which census tracts and the demographic composition of their population would have various percentages of HMDA data available such as less than 50%, 60 to 70%, 70% to 80%, etc. It would seem that a diminution of data by 20 percentage points or more would be problematic for relatively underserved tracts (as measured by loans per capita) with high concentrations of LMI or people of color residents.

A spatial analysis would have shed more light on the impacts of the partial exemption. The GAO stated that there was more of an impact in rural tracts but did not provide any details about where these tracts were.[38] The CFPB should update the analysis by providing heat maps showing in communities across the county where the impacts would be significant.

In addition, the GAO’s analysis was incomplete because the impact of voluntary reporting needs to be assessed over a longer period of time. Over a two year time period (2018 and 2019) covered by the GAO report, about 37 percent of depository institutions eligible for partial exemption for their closed-end loans voluntarily reported the complete data. However, the percentage of voluntary reporters dropped from 2018 to 2019. The GAO calculated that “if it was not for the voluntary reporters, the proportion of exemption codes would have increased from 3 percent of HMDA data to 6 percent.”[39]

The CFPB will likely have at least two years of new HMDA data (2020 and 2021) since the GAO report and will be able to more completely assess the impact of voluntary reporting. It is probable that the number and percentage of voluntary reporters further decreased, perhaps dramatically, further upending the GAO’s minimal impact assessment. Also, if the number and percentage of voluntary reporting did not decline significantly, this would be further evidence that lenders regard HMDA data collection to be of modest cost, while they believe that collection and dissemination provides benefits such as insights into their own practices and lending patterns.

In sum, we call on the CFPB’s assessment of HMDA data to include a more detailed analysis of EGRRCPA’s impact. Such an assessment is likely to find a significant and detrimental impact of EGRRCPA on the statutory purposes of HMDA data including whether lenders are meeting community needs in a responsible and non-discriminatory manner. The assessment should include recommendations to Congress regarding EGRRCPA such as rescission or at least modification. The GAO offered one modification suggesting that lenders required to report HMDA data should be required to report their number of open end loans.[40] Without this reporting, the GAO concluded that the CFPB could not determine if a lender would be exempted from reporting the new Dodd Frank data for open end loans.

Recommendations for Data Points

In this section, we build upon the recommendations in a 2019 comment letter, Comment Letter On Advanced Notice Of Proposed Rulemaking On HMDA Data Points, that a number of us signed and helped develop.[41] Overall, the 2019 letter represents our perspectives. Below, we offer elaborations to the recommendations in the 2019 letter.

Demographic Data Points

Race and ethnicity

The CFPB HMDA rule established important enhancements to the racial and ethnic data points. The disaggregated racial and ethnic data points were vital toward uncovering disparities within the Asian and Hispanic communities as discussed above. In addition, applicants were receptive to the disaggregated categories. The CFPB reported a high incidence of disaggregated categories appearing in the data. For example, about half of applicants reporting Asian as their race in the first data field for race also reported a disaggregated category for Asian in the second data field.[42] By 2020, 61% of applicants that identified as Asian, Hawaiian or Pacific Islander, or Hispanic also chose a disaggregated identifier according to a recent NCRC paper.[43]

In addition, the CFPB recorded disparities within the Asian community that need to be further assessed in future years. The CFPB found:

Hawaiian or Pacific Islander (HoPI) borrowers had lower income and credit scores, higher CLTVs, debt-to-income ratios (DTI), and denial rates than the other AAPI (Asian American and Pacific Islander) subgroups. Furthermore, although Vietnamese, Native Hawaiian, and Other Pacific Islander borrowers had higher average credit scores and incomes, and lower median DTIs and CLTVs than Black and Hispanic White borrowers, their denial rates were similar to those for Black and Hispanic White borrowers.[44]

In a recent report, NCRC recorded the emergence of Asian Indian borrowers as a powerful force in the mortgage market. In 2018, 29% of Asian applicants were Asian Indian, in 2020 that rose to 36% although Asian Indians were only about 24% of the Asian American population.  In that time period almost 654,000 loan applications were recorded with Asian Indian applicants. Not inconsequentially, Asian Indian households in the United States earned a median income of $119,000, about $20,000 more than the Asian median income for that year.[45] As this example makes clear, the disaggregated categories provide important insights into the relative success of various racial and ethnic subgroups in the market and how those that are not as prominent in the market can be assisted.

In addition to disaggregated categories for Asians and Hispanics, the CFPB should consider more detailed categories for African Americans. In its proposed Section 1071 rule for small business loan data, the CFPB proposed the following for African Americans:[46]

Black or African American

  • African American
  • Ethiopian
  • Haitian
  • Jamaican
  • Nigerian
  • Somali
  • Other (print race for example, Ghanaian or South African)

We would also support the addition of a Middle Eastern and North African category for HMDA, similar to that contemplated by the CFPB for Section 1071 data.

The related issue of language access. A related issue is the language in which a loan was negotiated. Advocates have long reported, including at the Federal Reserve’s HOEPA hearings in 2006, the problem of Limited English Proficient (LEP) borrowers being sold mortgages in their primary non-English languages, only to receive English-only documents with significantly worse terms than those promised.[47] Such borrowers have been among the most vulnerable to predatory lending, abusive servicing and loan modification scam practices. The quality of lending and servicing to LEP borrowers is an important indicator of fair lending, and more granular data could help local jurisdictions direct resources to these at-risk communities.

A 2014 Government Accountability Office report found statistically significant disparities in the rate of loan modification denials, cancellations, and re-defaults for LEP borrowers and other protected groups as compared to non-Hispanic white borrowers after analyzing certain loan modification data under the HAMP program. We recommend:

  • Enhance existing reporting by requiring data collection relating to the primary language of the borrower, the language spoken to negotiate the loan and the language of the loan documents.
  • At the very least, add a column in the data indicating if the applicant has limited English proficiency.

Gender Identify and Sexual Identity

The CFPB must consider amending HMDA data to precisely capture gender and sexual identity. We request that the CFPB requires lenders to ask two questions with multiple choice options: one for gender identity and another for sexual orientation. Discrimination unfortunately takes many forms; a particular lender may not discriminate against gay males but may do so against trans individuals or other gender identities or sexual orientations. Therefore, a more accurate method for asking gender identity and sexual orientation is preferred in order to more accurately measure disparities or discrimination.

NCRC suggests that the CFPB consult with the UCLA School of Law, the Williams Institute, which has templates on its website.[48] For gender identity, one possibility is:

  • Male
  • Female
  • Transgender
  • Do not identify as female, male, or transgender

For sexual orientation, one possibility is:

  • Straight
  • Gay or lesbian
  • Bisexual
  • Transsexual, or gender non-conforming

Identify if the borrower is a person with a disability

NCRC supports the recommendation of the National Disability Institute (NDI) that the HMDA data collection requirements include collecting information on the disability status of applicant. As NDI stated in a recent comment letter, about 15 % of the adult population or 40 million people have a disability.[49]

It is likely that the full potential of home ownership is not being realized for persons of disability due to discrimination in the lending marketplace and/or lack of access (either physical or virtual). In order to further understand and ameliorate barriers to access, the CFPB should mandate data collection on the disability status of the applicant. This would help carry out the objectives of the Americans with Disabilities Act (ADA) as well as those of HMDA.


As discussed above, in the years leading up to the financial crisis, older adults, particularly older adults of color, were targeted by abusive lenders. In addition, the CFPB documented worrisome age disparities. Using the 2018 data the agency found that the median credit score for applicants aged 45 to 54 was 744, more than 20 points lower than median score of 767 for applicants aged 65-77.[50] At the same time however, older adults experienced higher denial rates than their younger counterparts.

The bins of 25 to 34; 35 to 44; 45 to 54; and 55 to 64 are useful for HMDA data analysis. Also, we appreciate that the data indicates whether the applicant is 62 years of age or older. Since older adults become eligible for reverse mortgage lending at age 62, it is imperative that the HMDA data enable users to identify if loan applicants qualify for reverse mortgages.

In both 2018 and 2019, the first years in which reverse mortgages were added to the dataset, there were just 61,000 applications reported, and in 2020 this number spiked to 70,114. The origination rate on those applications increased as well, from 54% in 2018 to 61% in 2020 according to a recent NCRC report.[51] Reverse mortgage lending might become even more important during the recovery from the pandemic as cash-poor older adults tap into home equity to meet expenses. Thus, precision in the age bins and the reverse mortgage data point remain important in HMDA data.

Instead of a top code for ages of 74 and higher, we urge the CFPB consider additional bins of 75 to 84 and 84 and older since older adults are living longer. Moreover, the public and federal agencies could assess if patterns of reverse mortgage lending and other types of home lending differ or are similar for the oldest age bins compared to other age bins. The lending patterns for the oldest age bins may also reveal whether there are particular fair lending or affordability concerns specific to the oldest seniors.

HMDA data must have demographic data for purchases

From 2018 – 2020, purchased loans constituted about 12% of all records in HMDA. Current CFPB guidance allows lenders that purchase loans the option of removing demographic data from those records before submission to HMDA each year.[52] A recent NCRC report revealed that some of the country’s largest lenders appeared to exercise this option as they report this data on none of their purchased loans. In 2020 the percentage of purchased loans lacking this data jumped to 75%. Among the 20 lenders that reported the most purchases in 2018-2020, 15 reported no data on 100% of those loans. This suggests they are exercising the option allowing them to strip data from loan records they purchase before reporting that data to HMDA.[53] The secondary market, including purchases by banks, has a critical impact on the ability of traditionally underserved populations to access loans. The CFPB must require demographic data for purchased loans in order to completely assess the fair lending record of banks and other institutions.

CFPB should address growing numbers of originations that lack demographic data on race or ethnicity

More loan records each year are reported by lenders lacking race or ethnicity data. From 2018 to 2020, loans without this data rose from 10.8% to 13.6% of all loan originations. For applications on forward mortgages on owner-occupied, site-built, 1-4 units loans in 2018, there were 12.1% that did not include race or ethnicity. By 2020 that figure rose 14.9% according to a recent NCRC report.[54]

The reporting of this data is voluntary for the applicant. However, in many cases, the lender is required to assign a race or ethnicity based on visual observation or surname. The exception to this is if the loan application was online or through the mail.[55] The public data released under HMDA does not provide detail on whether the application began in person, by phone, or online. However, as mortgage companies, many of which lack a physical presence and rely heavily on fully online applications, have risen to prominence in the mortgage market, they have struggled to collect this data as well as banks or credit unions. The examples of lenders that do well at collecting this data and the high degree of eligible borrowers that provide disaggregated racial data suggests that this may be an issue of how the lender is asking for the information. A working group of lenders and community-based organizations to share best practices might have a positive impact on the data collection rates.

Data Points on Loan Purpose

Multifamily Lending

As stated in our 2019 letter, the purposes of HMDA include assessing if lenders are meeting housing needs and to assist public officials in directing public sector investments in a manner to leverage private lending and capital in disadvantaged communities. A pressing housing need in communities is affordable rental housing. Instead of being contrary to the statute, multifamily data in HMDA fulfills a central purpose of the statute.

We would also argue that it is statutorily required. The HMDA statute refers to the recording of “mortgage loans.” According to the statute, “the term “mortgage loan” means a loan which is secured by residential real property or a home improvement loan.”[56] Further, “The terms “residential real property”…. mean leaseholds, homes …and, combinations of homes or dwelling units and business property, involving only minor or incidental business use, or property to be improved by construction of such structures.”[57] This statutory language indicates that the CFPB does not have the authority to exempt multifamily housing owned by non-person entities/corporations from HMDA reporting.

  • Affordable UnitsWe very much appreciate the new HMDA data field that reports how many units are affordable under a city, state, or federal housing program.  The new data is limited to income-restricted housing and not housing that is “naturally” affordable or kept affordable due to rent-regulation, as is often the case in New York City and a few other areas in the country. That being said, banks are critical to financing income-restricted affordable housing and encouraged to do so through the Community Reinvestment Act (CRA). This data will provide valuable insight into who is financing this source of affordable housing. However, the reporting for the number of units needs to change to better understand the impact of that affordable housing.
  • Number of units on all loans, 1-4 family homes and multifamily buildings. The new HMDA data now reports exact unit counts in 1-4 family loans, which is a huge improvement, but only reports multifamily units in ranges. This data is necessary for the public to understand how many families will be impacted by the loan. While the data is better than in prior years, when it simply reported 1-4 units vs 5+ units, it would be better if the data gave the precise count of units. The public would then have more accurate data on how many units a particular lender or all lenders in the locality were financing. Also, the public could better estimate if supply was rising to accommodate demand for rental units.

Furthermore, the bin reporting does not allow the general public to generate estimates regarding the number of units that are affordable for lower income households. Only crude estimates are possible such as using the median value in a bin and multiplying it by the percentage of units that are affordable. For example, for the bin 25 to 49 units, does the analyst assume that a median number of units of around 37 should be multiplied by an affordable housing percent of say 10 percent to derive the number of units that are affordable? Or is the particular loan actually financing 26 units, which would result in a lower count of affordable units. Thus, inaccurate estimates of affordable housing units for alleviating rental cost burden will be generated. Additionally, the bin >149 units could be 150 units or 1000 or more, which renders this bin essentially useless for unit counts.

Estimating how many units are financed in the aggregate or how many units are affordable is frustrated by this bin reporting. Evaluating the extent to which lenders are responding to housing needs is quite difficult if not impossible. Furthermore, we do not see how bin reporting protects individual privacy since the vast majority of HMDA multifamily loans are for corporations. Bin reporting of units is inconsistent with the statutory purpose of HMDA and without any justification.

  • Occupancy Type: Multifamily properties are almost always investment properties.   However, 1-4 single family homes also provide rental housing in many cases. We often consider non-owner occupied housing to be an investment property, but the new occupancy field, coupled with the number of units, will provide more clarity. As mentioned above in both the spirit and letter of the law, this is not only relevant to HMDA but important and legally required.
  • Debt to Income Ratio is not reported for multifamily loans, but we urge the CFPB to reconsider similar indicators for multifamily housing, such as the Debt Service Coverage Ratio (DSCR), or the Net Operating Income that would allow us to calculate the DSCR and other indicators of overleveraging. The Association for Neighborhood & Housing Development (ANHD) finds that loans with a DSCR below 1.2, particularly on affordable rent-regulated housing, can provide significant incentives to harass and displace tenants – or neglect needed repairs – as the borrower must raise rents in order to pay off the mortgage. For enforcement purposes, it is important to know which lenders are complicit in this behavior, as it can have disastrous effects. One particularly egregious example is that of Raphael Toledano. The NY State Attorney General documented Madison Realty Capital loans to the now notorious landlord, Raphael Toledano, with DSCR well under 1.0X, meaning that the building’s net income was not sufficient to pay the debt. In order to raise the income, Toledano used all sorts of tactics to harass and displace rent-stabilized tenants and bring in higher-paying tenants.[58]
  • Capturing More Bank and Nonbank lenders: This is not a new field, per se, but it must be noted that HMDA is valuable in distinguishing banks from non-bank lenders. The newly expanded data will help us get a better handle on who is lending on multifamily housing – CRA-regulated banks or private companies that are not regulated by the CRA. Trends in bank and non-bank lending will give stakeholders more insights into CRA and fair lending consequences and/or the need to adjust CRA and fair lending enforcement. We recommend eliminating the requirement that a depository institution must originate at least one single-family loan in order to be required to report HMDA data as discussed above. Also, the previous 25 loan threshold for lenders is too high overall, and particularly for multifamily lenders that have lower lending volumes than 1-4 family lenders. The new 100 loan threshold is four times higher than the previous threshold, which we believe was too high for multifamily lending.
  • Loan amount and property value.  Related to DTI and DSCR discussion above, these data points help us to see if loans are being made at an appropriate amount to support the existing tenants, or if they appear too large and could lead to harassment and displacement. Likewise, the CLTV data point will enable stakeholders to spot any increases in high CLTV lending which could also be a red flag for abusive practices.

Manufactured Home Lending

An important part of the nation’s housing stock, manufactured housing accounted for 6.7 million homes in the nation, nearly half of which were in rural communities. Manufactured homes comprised 13 percent of all the occupied housing units in rural and small-town communities and represented a quarter or more of the homes in 264 non-metropolitan area counties.[59] In many of these sparsely populated counties, site-built homes are costly to construct relative to incomes, and manufactured homes provide a more affordable option. While extremely important to the rural housing stock, there are several areas of concern with this type of housing due primarily to financing terms and land ownership.

A review of 2020 HMDA data identified 45 percent (55,185 originations) of all home purchase, first-lien loans involving a manufactured home as secured by the home alone. Using Dodd-Frank added data points, researchers can now ascertain a better understanding of how many borrowers might qualify for a standard, lower-cost, mortgage loan.[60] The HMDA 2020 data included 14,437 borrowers who owned the land under their manufactured home yet secured the loan by the home alone. Ninety-two percent of those borrowers received a high-cost loan. Of those that received a loan secured by both the manufactured home and land, only 41% received a high-cost loan.[61] A closer inspection of these 14,437 borrowers, particularly with the enhanced applicant data, might improve our understanding of whether any of these borrowers could have gotten less costly loans. A 2021 report by the CFPB uses the new HMDA manufactured housing data points to explore such lending patterns and trends[62] and more research is sure to come.

Data Points on Loan Type

Cash-Out Refinances

Details on refinance lending must remain in the HMDA data. In the subprime lending era, refinance lending became a method whereby abusive lenders extracted wealth by convincing borrowers to take out high-cost, cash-out refinance loans as a means of covering non-housing expenses. The cash-outs often left vulnerable borrowers with few sources of wealth and savings after using the lump sum from the refinances. The Financial Crisis Inquiry Commission (FCIC) documented the high use of refinances and cash-out refinances during the subprime lending era. According to the FCIC, 75 percent of the subprime loans were first liens, and of these 82 percent were refinance loans; 59 percent of the refinance loans were cash-out.[63]

As discussed in our 2019 letter, a CFPB study reinforced the importance of maintaining HMDA data on cash-out refinances since the volume of cash-out refinances is significant and these loans are higher cost and taken out by vulnerable borrowers. The CFPB found that a greater percentage of borrowers (15.6 percent vs. 14.1 percent) take out cash-out refinance loans as opposed to non-cash-out refinance loans.[64] In addition, the borrowers of cash-out refinance loans have lower median credit scores than non-cash-out borrowers.[65]

The differences in the patterns of pricing and borrower demographics for the two types of refinance loans need to be monitored to assess whether voluntary or regulatory action in the future is needed to curb a rise of abusive pricing or loan terms and conditions. Therefore, the maintenance of this data point in the HMDA data is necessary.

Data Points on Loan Terms and Conditions

Price Variables, Debt to Income Ratio (DTI) and Combined Loan to Value Ratios (CLTV)

Our 2019 letter discussed the vital importance of pricing variables including interest rates, origination charges, discount points and lender credits in terms of monitoring patterns of pricing by demographic groups and providing an early warning system of any rise in significant pricing disparities. We also discussed how the DTI and CLTV information in the new Dodd Frank data similarly alerted stakeholders to surges of unsustainable lending. Combined with information on the channel of lending, this new data can help determine whether any increase in worrisome lending is confined to one channel (retail or wholesale) or is widespread through the industry including entities regulated at the federal level or those regulated primarily at the state level.

We ask the CFPB to consider adding the Annual Percentage Rate (APR) to the publicly available dataset. While it is possible for advanced data users to calculate an APR, most data users will not be able to do this since the calculations are involved and complex. The APR is an easy-to-understand data point that quickly enables a wide swath of the public to compare pricing across products and lenders. Thus, it will enable the public to more easily hold lenders accountable whose pricing seems much higher than their peers for comparable products to borrowers with similar characteristics.

The CFPB’s recent analysis of HMDA data reaffirms the importance of the pricing data for exploring disparities by race and ethnicity. Using the 2020 data, the CFPB found that the median interest rate for home purchase, first-lien, site-built residences was 3.250% for African Americans and white Hispanic borrowers but just 3.125% for non-Hispanic white borrowers. Likewise, median total loan costs were $6,208, $5,980, $4,344 for white Hispanic, African American and non-Hispanic white borrowers, respectively.[66] These disparities need to be further investigated in future years, with enhanced analyses controlling for demographic, loan and neighborhood characteristics.

In a recent report, NCRC found significant differences by race in loan closing costs, which together with downpayment requirements, can pose significant barriers to homeownership. Closing fees include a number of charges for a variety of different items but primarily they are an origination fee, mortgage insurance charges, and other services related to underwriting of the loan. For home purchase loans, the impact is clearly different across various racial groups, with Asian Indian home buyers, on average, paying $2,896 in closing costs compared with $4,026 paid by non-Hispanic Whites. Black ($5,532), Cuban ($5,932), and Native Hawaiian ($6,666) buyers often pay far more.[67]

NCRC published a white paper illustrating how the new data can shed light on loan pricing overall and pricing received by various demographic groups.[68] The benefits of this disclosure overwhelm the cost to lenders for producing this data, which the CFPB calculated to be a mere $23 per closed-end loan for most reporting institutions.[69] Benefits of this disclosure include enabling analyses to determine how to reduce barriers, including cost, to homeownership and fair lending enforcement.

Debt-to-Income Ratio

The CFPB’s current method for disclosing debt-to-income ratio (DTI) takes into account the critical importance of DTI ratios in determining whether loans are affordable and sustainable, a key component of fair lending and consumer protection analyses. The bins for DTI disclosure provide a clear measurement of debt burden and appropriately balance the utility of the data with protecting the privacy of loan applicants with unusually high or low DTIs. In addition, granular disclosure of ratios between 36 and 50 percent provide key information about relatively high DTIs and allow regulators and the public to determine whether lenders making loans with high DTIs are engaging in abusive, high-cost lending layered with other risks such as adjustable rates and prepayment penalties.

Preliminary analysis of the recently released national HMDA data suggests that DTI ratios provide critical explanatory power when analyzing lending decisions and illuminate potential racial and ethnic disparities in mortgage lending. DTI ratios should therefore remain part of HMDA data collection and continue to be disclosed in the public HMDA data.

Application channel

Dodd Frank and its implementing regulations require that, except for purchased covered loans, institutions report the following information about the application channel of the covered loan or application: Whether the applicant or borrower submitted the application directly to the financial institution; and whether the loan was, or would have been, initially payable to the financial institution.[70] This information is currently disclosed without modification in two separate fields.

As with most of the new data points required under Dodd Frank, application channel was added to HMDA as a result of the various predatory practices that contributed to the foreclosure crisis and often disproportionately affected borrowers and communities of color. When combined with many of the other new data elements (i.e. interest rate, points and fees, loan costs, origination charges, discount points, loan term and non-amortization features), application channel is critical to fair lending analyses. It allows regulators and the public to check for abusive or discriminatory conduct by loan channel.

Using Dodd Frank-enhanced HMDA data, the Empire Justice Center found that in Rochester, it was particularly important to monitor riskiness among the 22% of loans in 2018 that were not directly submitted and payable to the reporting institution since a significant number of these were brokered.[71] Some brokers were a source of abusive lending in the years leading up to the financial crisis. The new HMDA data enables the public to better monitor loan terms and conditions associated with brokered loans.

Non-Amortizing Features

Non-amortizing features are another set of data points required by Dodd Frank. The law and its implementing regulations require institutions to report ‘‘the presence of contractual terms or proposed contractual terms that would allow the mortgagor or applicant to make payments other than fully amortizing payments during any portion of the loan term,” and includes whether there is a balloon payment, interest-only payments, negative amortization or any other non-amortizing features.[72] These characteristics were essential elements of loans that drove the foreclosure crisis and future trends in this area are important to monitor, especially in how loans with these features impact vulnerable populations (low-moderate income, borrowers and communities of color, older borrowers) and/or disparately impact protected classes. All of these data elements need to continue to be reported and disclosed without modification so regulators and the public can monitor changes in their use.

The CFPB’s HMDA Data Points article showed that, except for balloon payments, the bulk of the non-amortizing features were attached to HELOC loans, which increased the riskiness of the loans. Of the 243,000 originated loans that included a balloon payment, about 128,000 of them were closed-end loans, and 115,000 were HELOCs.[73] Almost 50% of HELOCs had interest-only payments.

Given that the bulk of non-amortizing features were attached to HELOCs, the Empire Justice Center examined these loans via the New York (NY) State 2018 HMDA data[74] to see how many features were attached to which types of HELOCs. Among the 19,738 HELOC loans originated in NY in 2018, the most common non-amortizing feature was the interest-only payment. Almost 54% of the HELOCs had this. Eight percent of the loans had some other non-amortizing feature.

Non-amortizing terms must remain in the HMDA so monitoring of any surges of risky lending can be undertaken and interventions can be made before these practices spread too widely.

Prepayment Penalties

In the decade before the foreclosure crisis, when subprime lending exploded, abusive prepayment penalties were another predatory feature; 80% of subprime loans had prepayment penalties compared to only 2% of conventional loans.[75] Prepayment penalties kept many consumers in unaffordable, subprime loans, the costs of which often outweighed the interest rate savings.[76] Due to this, Dodd Frank and its implementing regulations require that institutions report the term in months of any prepayment penalty for covered loans or applications, other than reverse mortgages or purchased covered loans.[77]

While the use of prepayment penalties is more limited after the passage of Dodd Frank, it is important to monitor any future increased use or increases in the term, especially by borrower race/ethnicity and in conjunction with other loan features.

The term of any prepayment penalty is already collected by lenders, so the burden of reporting is unlikely to outweigh the benefit of its usefulness to furthering the purposes of HMDA, particularly for helping identify possible discriminatory lending patterns and enforcing fair lending laws. One concern we have with how it is reported and disclosed is the overbroad use of “not applicable” (NA). NA is used for both loan transactions for which reporting is not applicable (i.e. reverse mortgages) as well as for covered loans which have no prepayment penalty. One possible way to distinguish between these two NAs might be to use NA for loan transactions not covered, and to use zero (0) for transactions with no prepayment penalty (for 0 months).

The combination of now knowing which loans are HELOCs and which loans have prepayment penalties (and the term of that penalty) helps regulators and the public better understand the mortgage lending landscape. As seen in the HMDA Data Points article, less than 1% of conventional conforming originations had a prepayment penalty. However, over 28% of HELOC originations had prepayment penalties.[78] Prepayment penalties are an important data point for assuring that any type of lending is not becoming layered with too many risks.

Data Points on Loan Action

Reasons for Denial

The reasons for denial are important data points for fair lending enforcement and homeownership counseling. Differences in the incidence of a reason or reasons for denial by race or gender can be further investigated in fair lending inquiries to ensure that lenders are considering applications in a non-discriminatory manner. Public agencies, regulators and community organizations can use the loan characteristics or loan term and conditions data points to see if reasons for denial are consistent with high DTIs, lower property values, or other data points for particular races or genders.

Housing counseling agencies can use the reasons for denial to assess which barriers are most common for residents in the census tracts that they serve. For example, a high incidence of credit history or DTIs may influence the housing counselor’s curriculum or work sessions with clients. If on the other hand, the primary reasons in the census tracts served by the counseling agency is incomplete application or unverifiable information, the curriculum would emphasize the loan application process.

Data Points on Loan Characteristics

Loan Amount – Midpoint reporting

Loan amount is a pre-Dodd Frank data element that has been disclosed for decades without any apparent risk or widespread instances of facilitating public re-identification of borrowers. Even though loan amount is in county real estate transaction records, there is no indication that adversaries have matched HMDA reported loan amounts to those in county records. Most likely, adversaries simply used county real estate transaction records to identify vulnerable consumers that they thought could be susceptible to abusive marketing. Therefore, our strong preference is that this variable continue to be reported as it was before the Bureau’s disclosure changes in 2018.

We believe modifications to the CFPB’s 2018 mid-point reporting is necessary for smaller loan amounts. For large loan amounts of $100,000 or higher, it does not appear that the accuracy of the data will be significantly diminished if the loan amounts are disclosed as the midpoint of a $10,000 range as the CFPB implemented. However, for home improvement lending, second liens and other lending that consists of smaller dollar amounts, the CFPB’s proposal will result in loan amount data that will be misleading. In those cases, the CFPB should either report the loan amount to the nearest $1,000 as is done now or make its interval smaller.

A few examples indicate the potential for misrepresentation of loan amounts due to the CFPB reporting method. In the Cleveland metropolitan statistical area (MSA) in 2015, there were 1,810 home improvement loan applications with an amount less than $10,000. Of these, 28 percent had values less than $5,000. Furthermore, in markets or communities where housing values are relatively low, the $10,000 midrange data would also be misleading for home purchase and refinance loans. Using the Cleveland example, there were 5,734 home purchase loan applications in 2015 with loan amounts of $50,000 or less.  Thus, a change of $5,000 (to pick the midpoint for a $50,000 loan) would be a 10 percent misrepresentation.

These are not the areas with the lowest property values in the nation, but they indicate the type of problem and misrepresentation that a uniform use of the $10,000 midpoints will create. While the release of the combined loan-to-value (CLTV) ratio may be helpful in general, the CLTV ratio would not deal effectively with these issues for lower loan amounts and property values.  Therefore, at a minimum, the CFPB should consider setting midpoints in smaller ranges for lower loan amounts and lower property values.

Loan amount data is critical for assessing if demand for loans of various amounts in neighborhoods are being met for borrowers of different races/ethnicities and income levels. It is therefore important to strive for as accurate reporting of this data as possible while also being sensitive to privacy risk. The CFPB can modify its reporting method without endangering privacy since this piece of data has not been used to facilitate re-identification or predatory marketing to our knowledge.

Loan Amount – GSE and FHA Limits

An important component of fair lending analysis is assessing the patterns of lending and loan sales for loans within the Government Sponsored Enterprise (GSE) limits. A critical part of this analysis is examining the share of loans made within the GSE limits that are actually sold to the GSEs and to other investors compared to the market as a whole. Presently, this is quite difficult for HMDA users without significant technical and data processing skills and resources.

We are pleased that the CFPB chose to indicate in HMDA data if the loan exceeded the conforming loan limits and is therefore ineligible for sale to the GSEs. In reporting this data, the CFPB needs to ensure that the GSE limit threshold applied to each loan amount is based on the loan amount limits adjusted for the number of units in the subject property, a procedure the public could not do before even with significant data processing resources because data on the number of units in the property was not available. Since the new Dodd-Frank data has the number of units, the CFPB will be in a position to adjust for number of units when calculating if the loan amount exceeds GSE limits.

We ask the CFPB to consider adding an indicator of whether the loan exceeds FHA limits.  As these limits are commonly quite different than the GSE limits, this would provide another valuable data point for HMDA analysis. The number of loans within the FHA limits that are FHA (especially when controlling for applicant income, applicant ethnicity or race, and census tract racial/ethnic composition) has been an important part of fair lending analysis. Indeed, the initial reason for enacting the HMDA was to reveal where conventional and FHA loans were being made in order to identify and combat redlining.

Property Value

Dodd Frank and its implementing regulations require institutions to report the value of the property securing the covered loan or, in the case of an application, proposed to secure the covered loan relied on in making the credit decision [deny, approve but not accepted, originate].[79]

The property value data point is important because it allows HMDA users increased understanding of what went into a lender’s credit decision. On its own, property value helps regulators and the public assess whether financial institutions are serving communities with more modest housing values, one of the purposes of HMDA.

HMDA users can use property value along with the loan amount field to calculate the loan-to-value ratio with respect to the property securing the loan. As the National Consumer Law Center and other advocates noted in their 2014 comments, this information increases understanding of how much equity borrowers have upon origination and helps to identify disparities in how property values affect loan terms. While the expanded HMDA includes a combined loan-to-value ratio, that variable is based on the total amount of debt secured and property that “does not need to be the property identified in §1003.4(a)(9) [the location of the property securing the loan] and may include more than one property and non-real property.”[80] Thus, it is important for the data to allow analysts to calculate loan-to-value ratios in addition to having the combined loan-to-value ratio.

The usefulness of the property value field is reduced by how it is disclosed to the public. The value disclosed is “the midpoint for the $10,000 interval into which the reported value falls, e.g., for a reported value of $117,834, disclose $115,000 as the midpoint between values equal to $110,000 and less than $120,000.”[81] This $10,000 midpoint disclosure significantly reduces the accuracy of any loan-to-value ratio.

The Empire Justice Center found that of the 110,000 home purchase loans reported to HMDA during 2018 in New York State, almost 8,300 or 7.5% were secured by properties of less than $100,000. The reporting of property values as midpoints in intervals of $10,000 could create unacceptable inaccuracies in the value of these 8,300 lower value properties. This is especially true for the loans originated in the communities of color in the Rochester MSA, where the median property value was $85,000; so over half of the 9,850 loans originated here could have large errors in the reported property values. Just as with loan amount, we recommend that the CFPB at least consider smaller intervals for property values below $100,000.

The importance of accuracy in property value and loan amounts is reinforced by how these data points can be used to estimate equity at time of loan origination. Assessing racial or ethnic differences in equity attained can help develop strategies to reduce racial and ethnic differences in wealth. For example, these strategies can address appraisal bias that reduce property values in majority minority neighborhoods. If property is more accurately appraised, people of color could obtain higher property values and thus more equity at loan closing.

Calculating the difference in property value and loan amount is a method for measuring equity at loan closing. In a recent report, NCRC found that Chinese borrowers had the highest amounts of equity, followed by whites, Hispanics and African Americans over the 2018-2020 time period.[82]

Data Excluded from Public Disclosure

The undersigned organizations urge the CFPB to reconsider exclusions from public disclosure for some of the data points and to provide data in other, non-HMDA reports for other data points the CFPB excluded from the public HMDA data. In certain cases, the undersigned organizations believe it is possible to provide useful information derived from the excluded data points so that HMDA’s statutory purposes can be better realized.

Credit Score

The CFPB should reconsider its exclusion of credit score data from the publicly available HMDA data. Credit score data is essential for fair lending analysis in order to determine whether similarly situated applicants are treated differently solely due to their race, ethnicity, gender, or age.

Although the CFPB states that credit score data is not useful to identify applicants, the Bureau suggests that credit score data of applicants identified via non-credit score data fields could be a source of embarrassment or help adversaries engage in abusive marketing. To substantially reduce the risk of reputational harm, the CFPB could use a normalized credit score reporting format. It would be harder for the general public to understand, for example, what someone’s credit score expressed as a z-score means than a precise reporting of a FICO score or other credit score.

The CFPB should normalize the credit score data reported each year and report loan applicants’ credit scores either as z-scores, a measure of a credit score’s place in the overall distribution of credit scores for loan applicants that year, or in percentile ranges based on the distribution of loan applicants’ credit scores. Z-scores have the advantage of being useful for statistical analysis.

If the CFPB opts to retain its proposed exclusion of credit scores, it should consider summary reporting of credit scores by census tract for the aggregate (all lenders) and for each lender. For each census tract, the CFPB could report in one of two ways:

1)    The number and percentage of applicants denied loans and the number and percentage of applicants approved for loans in each quintile of normalized credit scores.

2)    The 25th, 50th and 75th percentiles of normalized credit scores for applicants denied loans and the same percentiles of normalized scores for applicants approved for loans.

Although not as comprehensive as loan-level credit score data, summary data at the census tract level would be nevertheless valuable for fair lending analysis to assess if the industry as a whole or individual lenders are treating similarly situated neighborhoods differently due to the racial, ethnic, income or age composition of the neighborhood.

National Mortgage Licensing System and Registry (NMLSR)

The NMLSR data field opens a whole new and important level of analysis for the HMDA data.  Today, many lenders, especially national banks that make loans across the country operate through mortgage brokers. Even though the lender is liable for the final loan decision, a borrower’s contact is through the broker – and that broker is also liable. Brokers often steer people to certain products and/or work with real estate sales agents that may steer people to particular properties or areas (as brokers often work in a relatively small geographic area).

Including some form of the loan originator’s ID in the HMDA data represents a critical opportunity to make transparent a previously hidden part of the mortgage lending process – one that is particularly important for issues of discrimination and reinvestment. After all, a lender’s decision to work with particular brokers can open up critical markets as well as close off opportunities.

The CFPB is correct that releasing the NMLSR ID for a particular “individual loan originator” (as defined in the Mortgage Licensing Act – 12 CFR § 1026.36) might make it possible to link legal documents (where that ID number is required) with an individual’s HMDA data. On the other hand, the Resource Center for the NMLS Website states that:

The NMLS Unique Identifier is the number permanently assigned by the Nationwide Mortgage Licensing System & Registry (NMLS) for each company, branch, and individual that maintains a single account on NMLS. The NMLS Unique Identifier (“NMLS ID”) improves supervision and transparency in the residential mortgage markets by providing regulators, the industry and the public with a tool that tracks companies and individuals across state lines and over time (emphasis added).

It continues explaining the NMLS Unique Identification Number Specifications:

NMLS assigns a unique identifier (NMLS ID) to each entity that has a record in the system. An NMLS ID is assigned to each company (Form MU1), branch (Form MU3), and natural person (Form MU2 or Form MU4) when the entity first creates its record in NMLS.

Using the NMLS ID for a company or branch rather than each individual would eliminate the ease of re-identification (as the individual ID required on several legal documents would not be disclosed). Having the ID for the mortgage company – and branch – would provide valuable information that would give the public access to this previously hidden and critical part of the mortgage market. Therefore, the CFPB must reconsider the weight of the public benefit – especially as it is included in Dodd-Frank and noted as an important factor by the NMLS itself – and consider the option of using the company and branch ID.

Property Address

While property address cannot be disclosed in the publicly available HMDA data, we encourage the CFPB to develop a hashed value to include in the publicly available data. One purpose of a location variable for a unique residential unit is to determine whether loan flipping is occurring. Loan flipping is a predatory tactic in which abusive lenders target borrowers for a series of refinancings that only increase debt and strip equity. Since no data is proposed to be disclosed that will assist the public in tracing loan flipping, the CFPB should consider reporting an indicator of loan flipping on a census tract level. Perhaps the Bureau could calculate the median for the number of times loans are secured by a given property over a multiple year time period and then indicate census tracts with a threshold of properties above and below the median. This would identify those tracts with potential flipping as well as other tracts that may be underserved by lenders. In general for excluded variables, the CFPB should be open to developing supplemental data that answer critical fair lending, CRA and consumer protection queries.

 More Rigorous Enforcement to Eliminate Outliers

Among the new data points due to the Dodd Frank enhancements are the interest rate and rate spread, which identify the interest rate at the time of closing on a home purchase and the difference between that rate and the average published rate for other loans that closed on that same day. Both of these data fields have a number of outlier records that can complicate the calculation of averages if the outliers are included in an analysis. For HMDA researchers, these outliers are a latent threat to the accuracy of their analysis and should be approached with caution.

Outliers in a dataset can change the average figure and misrepresent the true nature of the data being analyzed. NCRC’s initial analysis with the 2018 data indicated the presence of outlier data, as the average interest rate and rate spreads were far larger than our experience suggested was accurate.[83] Among interest rates reported on originated loans, there were 1,365 reported rates over 15%. One reported an interest rate of 260,000%. This is clearly an error, and while it is the most extreme example there are several hundred records with rates that are substantially higher than what would seem to be realistic.

For rate spreads, the lender reports the percentage above or below the daily average as a positive or negative decimal. For applications that did not result in an origination, the rate spread and interest rate data is either missing or of questionable value so we limited our outlier analysis to originated loans. In 2018, among originated loans, there were 51,489 loans with rate spreads over 5% from the average on the day of closing. A number of these rate spreads were over 100% higher or lower than the average. Twelve were reported to be 9999997% below average! Clearly, these are errors and although the extreme examples are rare, if one falls into the dataset a researcher has selected for analysis, it can skew results in ways that are not immediately apparent.

After identifying outliers we were able to eliminate .03% of originations based on the interest rate and 1.6% of originations based on rate spread. Results of our analysis after screening out these outliers show a substantial difference, indicating the previous results were skewed by the presence of these outliers.

Deleting up to 1.6% of originations can diminish the accuracy of the data in representing actual pricing trends. The CFPB should take steps to reduce errors so researchers do not have to eliminate significant amounts of data.

We urge the CFPB to establish outlier thresholds for price data, based on historical analysis, which would require additional CFPB scrutiny if a lender exceeds these thresholds. In the NCRC analysis, we identified interest rate and rate spread figures that were two standard deviations from the mean. The CFPB could use a metric like this to identify outliers and could also decide upon a number or percentage of such outliers, which indicates a probability of erroneous data from a lender. Finally, if the CFPB finds that a lender’s submissions are in error, a monetary penalty should be certain and significant.

NCRC also recommends that the CFPB consider developing resources to help users learn how to work with HMDA data more effectively, including how to identify possible errors. This could include periodic webinars, a dedicated Slack channel, or other method determined by the CFPB. Giving users easier access to this knowledge is crucial in developing a cohort of experts to assist in quality control and the sharing of ideas.

Missing and Exemption Codes Need to be Re-Worked

When the CFPB restructured HMDA data in 2018, the agency categorized ‘Exempt’ and Not Applicable (NA) values numerically for several variables. ‘Exempt’ occurs when a lender does not report a data point because a lender is not required to report the data point such as those lenders qualified for the partial EGRRCPA exemptions that pertain to certain data required by Dodd Frank discussed above. Not applicable applies when a data point is not necessary. For example, a lender would not supply a “reason for denial” when the lender originated a loan to a borrower.

Certain fields in the data now include coded 4-digit numbers which identify the record as either ‘Exempt’ or ‘Not Available’ (e.g. 1111, 7777, 8888, etc.). This occurs primarily in numeric fields like interest rate and rate spread, but also appears in a variety of other fields. These improper values can appear to be part of the data as extreme outliers to analysts who are not familiar with HMDA data when in fact they should function as indicators, rather than as data values. When taking a large “slice” of data, these indicators are easily misread as data values, substantially shifting totals and averages in an analysis. These records are not immediately apparent, and can substantially shift any average figures derived from those fields.

The CFPB should replace four digit numbers in cases of Exempt or NA data. For some variables, it appears that the numbers can be replaced by Exempt or NA as is the case for the debt-to-income ratio.[84] In cases when the numbers cannot be replaced, the data dictionary and other documentation should clearly inform the user about potential issues when analyzing the data and suggest steps to take to remove possible errors.

Dissemination of HMDA Data Must be improved

Although the CFPB has made improvements to its dissemination of data over the last few years, we believe that more resources could be devoted to making the data easier to use for the public that seek to measure lender CRA and fair lending performance. The CFPB’s emphasis now seems to be on helping technology developers create data tools for analysis that they can then sell or otherwise disseminate. However, most members of the public will not have access to these data tools. Instead, they seek summary tables or the ability to download the data quickly into excel tables. We suggest that the CFPB can significantly improve the usability of the data for the public at a modest cost. This must be a high priority because more easily accessible data bolsters HMDA’s purpose to hold lenders accountable for serving community needs in a responsible manner.

In our, Letter To CFPB Director Requesting Significant Improvement In The Public Dissemination Of HMDA Data, we described how the Federal Financial Institutions Examinations Council (FFIEC) previously provided easy-to-use preformatted tables to the general public that described lending trends by demographics of borrower and neighborhood by loan type and purpose for individual lenders and all lenders, as a group, on a metropolitan level.[85] The FFIEC tables and data downloads enabled data users at all levels from novice to experienced to retrieve and use the data.

The CFPB’s method of dissemination made it more difficult for beginner and occasional uses. These users no longer have access to preformatted, easy-to-use tables. As described in our previous letter, the CFPB’s current method of dissemination is not optimal for achieving the purposes of HMDA, which is enabling the public to determine if lenders are serving credit needs in a responsible and non-discriminatory manner. If many users, particularly beginner and occasional users, cannot access the data in a meaningful manner, they cannot use the data to determine if lenders are meeting needs.

The tables in the appendix to our previous letter revealed a significant level of detail in the previous FFIEC tables. For example, using data from 2016 on refinance lending on a metropolitan level, a FFIEC table showed applications received, loans made, applications denied and other actions on applications by ethnicity, gender and income of borrowers.[86] The table had both loan counts and dollar amounts. The table was available in either PDF or Excel format.

Another FFIEC table described loan pricing for home purchase, conventional, first lien loans, showing both prime loans and loans that are generally considered high-cost that are 1.5 percentage points or greater than the average prime offer rate.[87] The data was available by race, gender, income level of borrower and by census tract categories. This table was also available as a PDF or in Excel

In contrast, the newer CFPB summary tables provide truncated options for cross-tabulations. An analyst can tabulate loan action by race or ethnicity for an individual lender or all lenders, as a group, on a state, metropolitan or county level.[88] However, in contrast to the FFIEC summary tables, the CFPB query engine cannot produce demographic break downs for combinations of loan actions, loan purposes or loan types, which significantly limits the utility of the summary tables. In general, the user cannot tabulate the number and percentage of conventional home purchase outcomes or other loan types and purposes for various demographic groups or census tract categories.[89] The aggregate reports for metropolitan areas have the same limitations.[90] Also, none of the newer Dodd Frank variables regarding loan prices or terms or conditions is available for the summary tables.

The new CFPB map feature enables an analyst to assess the extent to which geographical areas are served or underserved by displaying loans or other action categories on a per capita basis on a national map.[91] While providing useful context, this mapping function is constrained by lack of accompanying summary tables that do not provide the level of detail that the previous FFIEC tables did for geographical areas.

The CFPB may want to consider replacing population in the denominator of ratio displayed on the map with the number of owner-occupied housing units, a comparator used in CRA exams. The number of owner-occupied housing units is the more appropriate comparator because it serves as a hard constraint on lending; lenders cannot make more home loans than the number of units.

In addition to diminishing the usefulness of data for novice or occasional users, the CFPB website is difficult for intermediate users to access. Intermediate users are most likely using excel and not the more expensive and sophisticated software programs. They need to be able to download the raw HMDA data on a county, metropolitan and state level into excel. We are pleased that the CFPB has a section of its HMDA website that facilitates these downloads.[92] However, it is not easy to find and can be confused with another part of the website that suggests that downloads at these geographical levels are only available for the years 2007-2017.[93] The CFPB should improve the labeling and clarity on its website. In addition, it appears that there are some multi-state metropolitan areas such as Chicago that cannot be downloaded from the website.

Intermediate and advanced users can also benefit from a much improved data dictionary.[94] All of the data dictionaries and the formats (schemas) for the HMDA data fields should be in documents that have a feature to print them and download them into spreadsheets.  This is not the case at present.  One option would be to provide a multi-page spreadsheet so that each data dictionary and format has its own page.

The Loan Application (LAR) data is not easy to use regarding the name of lenders, which appear in another database called the Reporter Panel. The CFPB could either include a manual regarding how to more easily append the lender name to the data in the LAR or could add the lender name when a user downloads the LAR on a state, metropolitan area or county level.

We also recommend adding the “other_lender_code from the Reporter Panel. The codes identify whether the filing institution is a bank, a mortgage subsidiary of a bank or bank holding company, an affiliate of a bank, or an independent mortgage company. This allows users to separate mortgage companies from banks and to separate the mortgage companies that are subsidiaries or affiliates of banks from those that are independent. Even more helpful would be if the CFPB provided directions, a function, or an Excel macro for adding the name of the lender, the “other_lender_code,” the top holder RSSD number, top holder name and other variables in the public panel to the end of the HMDA data fields.[95] This would make it easier for data users to identify if lenders have parent companies.

Large volume lenders are required to report HMDA data to the CFPB on a quarterly basis. We recommend that the CFPB then should immediately make this data available to the public so that trends in access and pricing for a significant segment of the market can be analyzed on a more real time basis. Action can be taken more quickly to address any worrisome trends regarding access and affordability for traditionally underserved populations.

The fixes recommended by this letter appear to be readily achievable for the CFPB. For beginner and occasional users, the summary tables and/or query engines can be improved to produce tables that are as detailed and useful as the previous FFIEC tables. The CFPB should also consider adding user manuals that describe how the data can be analyzed.

We also recommend that the CFPB creates a consumer and community advisory committee that would meet at least on a quarterly basis to consider the functionality and improvements to the website. An advisory committee is ideally suited to providing input from the public about the ability of the data and the CFPB’s website to achieve HMDA’s purpose as a law empowering the public to assess financial institutions’ responsiveness to needs.


The undersigned organizations oppose any diminution of the enhancements the CFPB made to HMDA data as it was implementing Dodd Frank. The CFPB enhanced the data carefully over several years after multiple requests for public comment and in response to widespread lending abuses. Moreover, as the lending industry, housing markets, and the nation’s population underwent significant changes, it was necessary to augment HMDA data so that the data could be used to assess whether housing and credit needs associated with older adults, subgroups within the Asian and Hispanic communities, and residents of multifamily and manufactured homes were being met responsibly and in a non-discriminatory manner.

In addition, the CFPB must rescind its 2020 rule setting reporting thresholds at 100 for closed-end loans and 200 for open-end loans. These thresholds exempt thousands of lending institutions from HMDA reporting and thus subvert the purposes of HMDA to ensure that lenders are responding to credit needs in a non-discriminatory manner. Also, the CFPB’s assessment should carefully examine the loss in the ability of HMDA data to measure fair and responsible lending in smaller cities and rural areas caused by the partial exemptions of EGRRCPA.

The CFPB should propose changes to make HMDA data more accurate in reporting the experiences of various racial and ethnic groups as well as including disability status of the applicant in the data. More precision is likewise needed in the reporting of gender identity and sexual orientation.

While the new Dodd Frank data created a rich database, improvements can nevertheless be made to demographic variables such as age of borrower and those concerning loan type such as multifamily lending and loan characteristics including loan amount and property value.

Lastly and importantly, the CFPB needs to continue to make improvements in the public dissemination of the data. HMDA must be thought of as the “people’s data.” Only when the data becomes easily accessible for the public and users as various skill levels, will the data fully realize its potential to hold lending institutions accountable for equitable and responsible lending.

We hope that the CFPB adopts a number of recommendations in this letter and we believe it would be beneficial to stakeholders if the CFPB held meetings to further explore the issues to be tackled in its assessment. We would be eager to engage the CFPB in such a meeting.

If you have any questions, please contact Josh Silver, Senior Advisor at the National Community Reinvestment Coalition (NCRC), on jsilver@ncrc.org.

The following organizations support the views expressed in this letter.


Americans for Financial Reform Education Fund

Better Markets

Consumer Action

National CAPACD- National Coalition for Asian Pacific American Community Development

National Community Reinvestment Coalition

National Consumer Law Center (on behalf of its low-income clients)

National Council of Asian Pacific Americans

National Disability Institute

National Fair Housing Alliance

National NeighborWorks Association

National Association of American Veterans, Inc.

Prosperity Now



Historic District Developers / Ensley District Developers

NAACP Economic Programs



Pima County Community Land Trust



California Coalition for Rural Housing

California Reinvestment Coalition




Community Reinvestment Alliance of South Florida

Goldenrule Housing & Community Developme

Metro North Community Development Corp.



Neighborhood Improvement Association



Brighton Park Neighborhood Council

Chicago Community Loan Fund

Housing Action Illinois

Illinois People’s Action

Neighborhood Housing Services of Chicago

The Resurrection Project

Universal Housing Solutions CDC




Urban Coalition of Appraisal Professionals



Massachusetts Affordable Housing Alliance

Springfield NHS



Maryland Consumer Rights Coalition



Southwest Economic Solutions



Housing Justice Center

Jewish Community Action



Metropolitan St. Louis Equal Housing and Opportunity Council

R.A.A. – Ready, Aim, Advocate



Housing Education and Economic development (HEED)

MS Communities United for Prosperity (MCUP)


New Jersey

New Jersey Citizen Action


North Carolina

Henderson and Company/ National Trust for Historic Preservation


New York

Association for Neighborhood and Housing Development (ANHD)

Devotion USA, Inc.

Empire Justice Center

Long Island Housing Services, Inc.



Housing Research & Advocacy Center dba Fair Housing Center for Rights & Research

Ohio Fair lending Coalition



CASA of Oregon

Housing Oregon



Renewable Manufacturing Gateway dba Community Growth Fund


Rhode Island

HousingWorks RI



Southern Dallas Progress Community Development Corporation

TCH Development Inc.



[1] Government Accountability Office (GAO), Characteristics and Performance of Nonprime Mortgages, p. 1, https://www.gao.gov/assets/100/96332.pdf

[2] GAO, Characteristics and Performance, pp. 4 and 9.

[3] Ibid., p. 10.

[4] Ibid., p. 13.

[5] Ibid, p. 4.

[6] Testimony of John Taylor, President and CEO of NCRC, before the Oversight and Investigations Subcommittee of the House Financial Services Committee, Rooting Out Discrimination in Mortgage Lending: Using HMDA as a Tool for Fair Lending Enforcement, July 25, 2007, https://ncrc.org/wp-content/uploads/2007/07/ncrc_test_reg_oversight_hearing_finsvs_july_07(2).pdf

[7] Testimony of John Taylor, p. 14.

[8] Adam J. Levitin and Susan M. Wachter, The Great American Housing Bubble, Harvard University Press, 2020, pp. 189-190.

[9] Emmanuel Martinez and Aaron Glantz, Kept Out, For people of color, banks are shutting the door to homeownership, February 2018, https://www.revealnews.org/article/for-people-of-color-banks-are-shutting-the-door-to-homeownership/

[10] Emmanuel Martinez and Lauren Kirchner, The Secret Bias Hidden in Mortgage-Approval Algorithms, the Markup, August 2021, https://themarkup.org/denied/2021/08/25/the-secret-bias-hidden-in-mortgage-approval-algorithms

[11] Agatha So, Unidos US and Jason Richardson, NCRC, Hispanic Mortgage Lending: 2019 HMDA Analysis, https://www.ncrc.org/hispanic-mortgage-lending-2019-analysis/

[12] CFPB, An Updated Review of the New and Revised Data Points in HMDA, Further Observations using the 2019 HMDA Data August 2020, pp. 53-54, https://files.consumerfinance.gov/f/documents/cfpb_data-points_updated-review-hmda_report.pdf

[13] NCLC, Consumer Concerns: Information for Advocates Representing Older Adults, Helping Elderly Homeowners Victimized by Predatory Mortgage Loans, https://www.nclc.org/images/pdf/older_consumers/consumer_concerns/cc_elderly_victimized_predatory_mortgage.pdf

[14] HMDA purpose, 12 USC 29 Section 2901, https://www.law.cornell.edu/uscode/text/12/2801

[15] Seattle Times, The mobile-home trap: How a Warren Buffett empire preys on the poor, February 2016,  https://www.seattletimes.com/business/real-estate/the-mobile-home-trap-how-a-warren-buffett-empire-preys-on-the-poor/

[16] CFPB, Manufactured housing consumer finance in the United States, September 2014, pp. 5-6, https://files.consumerfinance.gov/f/201409_cfpb_report_manufactured-housing.pdf

[17] CFPB, Proposed Rule, Docket No. CFPB-2019-0021, pp. 17-18, original double-spaced version posted on agency website, https://files.consumerfinance.gov/f/documents/cfpb_nprm-hmda-regulation-c.pdf

[18] CFPB, Data Point: 2018 Mortgage Market Activity and Trends, August 2019, p. 60, https://files.consumerfinance.gov/f/documents/cfpb_2018-mortgage-market-activity-trends_report.pdf

[19] CFPB, 12 CFR Part 1003 [Docket No. CFPB–2019–0021] RIN 3170–AA76 Home Mortgage Disclosure (Regulation C)  Federal Register, Vol. 85, No. 92, Tuesday, May 12, 2020, p. 28371.

[20] Legal Services NYC, Jury Finds Emigrant Bank Liable for Discrimination in First Reverse Redlining Case to be Tried in Federal Court,  https://www.legalservicesnyc.org/news-and-events/press-room/1033-jury-finds-emigrant-bank-liable-for-discrimination-in-first-reverse-redlining-case-to-be-tried-in-federal-court

[21] CFPB, Reverse Mortgages: Report to Congress, June 28, 2012, p. 9, https://files.consumerfinance.gov/a/assets/documents/201206_cfpb_Reverse_Mortgage_Report.pdf

[22] FDIC, CRA Exam of Frontier Sate Bank, November 2017, https://www5.fdic.gov/CRAPES/2017/21978_171113.PDF

[23] CFPB Final Rule, May 2020, p. 28373.

[24] County designations are rural, central, or outlying and are obtained from the Census, see https://www.census.gov/geographies/reference-files/time-series/demo/metro-micro/delineation-files.html

[25] See maps produced by Jason Richardson, NCRC, September 2019 at  https://public.tableau.com/profile/jason.richardson#!/vizhome/HMDALoanThresholdImpactMap/Dashboard1 for a map of the impact of various thresholds.

[26] CFPB, A Brief Note on General Lending Patterns of Small to Medium Size Closed-end HMDA Reporters, June 2021, p. 4, https://files.consumerfinance.gov/f/documents/cfpb_general-lending-patterns-hmda-reporters_report_2021-06.pdf

[27] Testimony of Comptroller of the Currency John Dugan before the United States Senate Committee on Banking, Housing and Urban Affairs, March 2008, p. 12, https://www.occ.gov/news-issuances/congressional-testimony/2008/pub-test-2008-28-written.pdf

[28] CFPB Proposed Rule, pp. 130-131, and CFPB Final Rule, p. 28379.

[29] CFPB, Introducing New and Revised Data Points in HMDA: Initial Observations from New and Revised Data Points in 2018 HMDA, August 2019, pp. 33-37, https://files.consumerfinance.gov/f/documents/cfpb_new-revised-data-points-in-hmda_report.pdf

[30] CFPB, An Updated Review of the New and Revised Data Points in HMDA Further Observations using the 2019 HMDA Data, August 2020, pp. 35-36, https://files.consumerfinance.gov/f/documents/cfpb_data-points_updated-review-hmda_report.pdf

[31] CFPB, Ibid, p. 67.

[32] CFPB Proposed Rule, p. 32.

[33] Lei Ding and Leonard Nakamura, Don’t Know What You Got Till It’s Gone: The Effects of the Community Reinvestment Act (CRA) on Mortgage Lending in the Philadelphia Market, Working Paper No. 17-15, June 19, 2017, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2991557

[34] CFPB’s “Correction Notice,” 85 Fed. Reg. 69119, https://www.federalregister.gov/documents/2020/11/02/2020-22891/home-mortgage-disclosure-regulation-c-correction-of-supplementary-information

[35] Government Accountability Office (GAO), Home Mortgage Disclosure Act: Reporting Exemptions Had a Minimal Impact on Data Availability, but Additional Information Would Enhance Oversight, May 2021, https://www.gao.gov/products/gao-21-350.

[36] GAO, May 2021, p. 18.

[37] GAO, May 2021, p. 20.

[38] GAO, May 2021, p. 21.

[39] GAO, May 2021, p. 23.

[40] GAO, May 2021, p. 36.

[41] See https://www.ncrc.org/comment-letter-on-advanced-notice-of-proposed-rulemaking-on-hmda-data-points/

[42] CFPB, Introducing New and Revised Data Points in HMDA: Initial Observations from New and Revised Data Points in 2018 HMDA, August 2019, pp. 20 and 21, and Table 3.2.2, pp. 106-111 https://files.consumerfinance.gov/f/documents/cfpb_new-revised-data-points-in-hmda_report.pdf

[43] Jason Richardson, Joshua Devine, and Jamie Buell, NCRC 2020 Home Mortgage Report: Examining Shifts During COVID-19, NCRC, January 2022, p. 16, https://www.ncrc.org/ncrc-2020-home-mortgage-report-examining-shifts-during-covid/

[44] CFPB, Data Point: Asian American and Pacific Islanders in the Mortgage Market – Using the 2020 HMDA Data, July 2021, https://files.consumerfinance.gov/f/documents/cfpb_aapi-mortgage-market_report_2021-07.pdf

[45] Jason Richardson, Joshua Devine, and Jamie Buell, NCRC 2020 Home Mortgage Report, p. 22.

[46] Bureau of Consumer Financial Protection, 12 CFR Part 1002, Docket No. CFPB-2021-0015, RIN 3170-AA09

Small Business Lending Data Collection under the Equal Credit Opportunity Act (Regulation B), Notice of Proposed Rule (NPR), Request for Public Comment, p. 803, https://files.consumerfinance.gov/f/documents/cfpb_section-1071_nprm_2021-09.pdf

[47] See Building Sustainable Homeownership: Responsible Lending and Informed Consumer Choice—Public Hearing on the Home Equity Lending Market, Federal Reserve Board (June 16, 2006), Tr. at 85, 12 238-39, 13 238-40, 14 242-45, 15 250-5216 (available at http://www.federalreserve.gov/events/publichearings/hoepa/2006/20060616/transcript.pdf).

[48] UCLA School of Law, the Williams institute, Best Practices for Asking Questions to Identify Transgender and Other Gender Minority Respondents on Population-Based Surveys (GenIUSS), September 2014, https://williamsinstitute.law.ucla.edu/publications/geniuss-trans-pop-based-survey/

[49] National Disability Institute comment letter to CFPB, https://www.regulations.gov/comment/CFPB-2021-0015-0081.

[50] CFPB, Introducing New and Revised Data Points in HMDA: Initial Observations from New and Revised Data Points in 2018 HMDA, August 2019, https://files.consumerfinance.gov/f/documents/cfpb_new-revised-data-points-in-hmda_report.pdf, Table 6.4.4, p. 195; the statistic is for conventional conforming loans.

[51] Jason Richardson, Joshua Devine, and Jamie Buell, NCRC 2020 Home Mortgage Report: Examining Shifts During COVID-19, p 25.

[52] See https://www.consumerfinance.gov/rules-policy/regulations/1003/b/#5, See #6 which informs lending institutions, “When you purchase a covered loan and you choose not to report the applicant’s or co-applicant’s ethnicity, race, and sex, you must report that the requirement is not applicable.”

[53] Jason Richardson, Joshua Devine, and Jamie Buell, NCRC 2020 Home Mortgage Report: Examining Shifts During COVID-19, p. 14.

[54] Ibid, p. 9.

[55] CFPB, Appendix B to Part 1003 — Form and Instructions for Data Collection on Ethnicity, Race, and Sex, https://www.consumerfinance.gov/rules-policy/regulations/1003/b/#1

[56] https://www.law.cornell.edu/uscode/text/12/2802

[57] https://www.law.cornell.edu/uscode/text/12/1464#c_6_A

[58] Jaime Weisberg, The “Bad Boy” Carveout, ANHD, May 2017, https://anhd.org/blog/bad-boy-carveout

[59] Housing Assistance Council (HAC) tabulation of American Community Survey 2019, five-year estimates. Rural and small town refers to HAC’s census tract classification based on housing density and commuting patterns. For specifics on this definition see the following report (page 113): http://www.ruralhome.org/storage/documents/ts2010/ts_full_report.pdf and https://www.minnpost.com/politics-policy/2018/05/trailer-parks-may-be-twin-cities-most-endangered-form-affordable-housing/

[60] See the work done in these brief and note how improved HMDA data might expand upon such work. Laurie Goodman and Bhargavi Ganesh. 2018. Challenges to Obtaining Manufactured Home Financing. Urban Institute Brief. This brief, as of 6/4/19, can be found at the following url: https://www.urban.org/research/publication/challenges-obtaining-manufactured-home-financing

[61] This data refers to home purchase, first lien loans. The threshold for first lien loans to be considered high costs is 1.5 percentage points over the interest rate charged on a similar prime rate loan.

[62] CFPB. 2021. Manufactured Housing Finance: New Insights from the Home Mortgage Disclosure Act, Report date 5/27/21 located at the following url: https://www.consumerfinance.gov/data-research/research-reports/manufactured-housing-finance-new-insights-hmda/

[63] The Financial Crisis Inquiry Report, Final Report of the National Commission on the Causes of the Financial and Economic Crisis in the United States, 2011, PublicAffairs, New York, p. 80.

[64] CFPB, New and Revised Data Points, p. 27.

[65] CFPB, New and Revised Data Points, p. 49.

[66] CFPB, Data Point: 2020 Mortgage Market Activity and Trends, August 2021, pp. 21-22, https://files.consumerfinance.gov/f/documents/cfpb_2020-mortgage-market-activity-trends_report_2021-08.pdf

[67] Jason Richardson, Joshua Devine, and Jamie Buell, NCRC 2020 Home Mortgage Report: Examining Shifts During COVID-19, p. 31.

[68] National Community Reinvestment Coalition, NCRC’s HMDA 2018 Methodology: How To Calculate Loan Price Jason Richardson – https://ncrc.org/ncrcs-hmda-2018-methodology-how-to-calculate-loan-price/

[69] Final Rule, CFPB, https://www.federalregister.gov/d/2015-26607/p-1530

[70] Dodd Frank, p. 724, as found at: https://www.govinfo.gov/content/pkg/PLAW-111publ203/pdf/PLAW-111publ203.pdf and CFPB Interpretation, as found at: https://www.consumerfinance.gov/policy-compliance/rulemaking/regulations/1003/4/#a-33.

[71] See the 2019 letter referenced above.

[72] Dodd Frank, p. 723, as found at: https://www.govinfo.gov/content/pkg/PLAW-111publ203/pdf/PLAW-111publ203.pdf and CFPB Interpretation, as found at: https://www.consumerfinance.gov/policy-compliance/rulemaking/regulations/1003/4/#a-27.

[73] CFPB, New and Revised Data Points, p. 34.

[74] Includes all originated first-lien open-ended conforming conventional loans on 1-4 family site-built principle residences that had an MSA identifier (including 9999, not in an MSA).

[75] https://www.responsiblelending.org/sites/default/files/nodes/files/research-publication/ib008-PPP_in_Subprime_Loans-0604.pdf

[76] Ibid.

[77] Dodd Frank, p. 723, as found at: https://www.govinfo.gov/content/pkg/PLAW-111publ203/pdf/PLAW-111publ203.pdf and CFPB Interpretation, as found at: https://www.consumerfinance.gov/policy-compliance/rulemaking/regulations/1003/4/#a-20.

[78] CFPB, New and Revised Data Points, p. 37.

[79] From PUBLIC LAW 111–203—JULY 21, 2010, DODD-FRANK WALL STREET REFORM AND CONSUMER PROTECTION ACT, 124 STAT. 1376, p. 723, as found at: https://www.govinfo.gov/content/pkg/PLAW-111publ203/pdf/PLAW-111publ203.pdf and 12 CFR Part 1003 (Regulation C), and CFPB Interpretation, as found at: https://www.consumerfinance.gov/policy-compliance/rulemaking/regulations/1003/4/#a-28.

[80] https://www.consumerfinance.gov/policy-compliance/rulemaking/regulations/1003/4/#a-24.

[81] CFPB, “Executive Summary of the HMDA Data Disclosure Policy Guidance,” December 21, 2018, as found at: https://files.consumerfinance.gov/f/documents/HMDA_Data_Disclosure_Policy_Guidance.Executive_Summary.FINAL.12212018.pdf.

[82] Jason Richardson, Joshua Devine, and Jamie Buell, NCRC 2020 Home Mortgage Report: Examining Shifts During COVID-19, p. 45.

[83] Jason Richardson and Jad Edlebi, Limiting Interest Rate And Rate Spread Outliers In HMDA Data, NCRC, May 2020, https://ncrc.org/limiting-interest-rate-and-rate-spread-outliers-in-hmda-data/

[84] See CFPB data dictionary, https://ffiec.cfpb.gov/documentation/2018/lar-data-fields/

[85] See NCRC letter, https://www.ncrc.org/ncrc-letter-to-cfpb-director-requesting-significant-improvement-in-the-public-dissemination-of-hmda-data/

[86] See https://www.ffiec.gov/hmdaadwebreport/AggTableList.aspx

[87] See https://www.ffiec.gov/hmdaadwebreport/AggTableList.aspx

[88] See https://ffiec.cfpb.gov/data-browser/data/2020?category=states

[89] It appears the user can tabulate the number and percentage of loan originations by loan purpose but not the other action categories. See https://ffiec.cfpb.gov/data-browser/data/2020?category=states.

[90] https://ffiec.cfpb.gov/data-publication/aggregate-reports/2020/AK/11260

[91] See https://ffiec.cfpb.gov/data-browser/maps/2020?geography=state

[92] See https://ffiec.cfpb.gov/data-browser/data/2020?category=states

[93] See https://www.consumerfinance.gov/data-research/hmda/historic-data/

[94] https://ffiec.cfpb.gov/documentation/2018/lar-data-fields/

[95] The documentation for the public panel is available via https://ffiec.cfpb.gov/documentation/2020/panel-data-fields/. It is also not easy to access the public panel. One has to remember it is in the national snapshot section of the HMDA data landing page. It should be more visible.

Print Friendly, PDF & Email

Redlining and Neighborhood Health

Before the pandemic devastated minority communities, banks and government officials starved them of capital.

Lower-income and minority neighborhoods that were intentionally cut off from lending and investment decades ago today suffer not only from reduced wealth and greater poverty, but from lower life expectancy and higher prevalence of chronic diseases that are risk factors for poor outcomes from COVID-19, a new study shows.

The new study, from the National Community Reinvestment Coalition (NCRC) with researchers from the University of Wisconsin–Milwaukee Joseph J. Zilber School of Public Health and the University of Richmond’s Digital Scholarship Lab, compared 1930’s maps of government-sanctioned lending discrimination zones with current census and public health data.

Table of Content

  • Executive Summary
  • Introduction
  • Redlining, the HOLC Maps and Segregation
  • Segregation, Public Health and COVID-19
  • Methods
  • Results
  • Discussion
  • Conclusion and Policy Recommendations
  • Citations
  • Appendix

Complete the form to download the full report: