Policy Brief to CFPB, on Creating Public Mortgage & Foreclosure Databases
The Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 requires the Consumer Financial Protection Bureau (CFPB) to write rules implementing improvements to data collected under the Home Mortgage Disclosure Act (HMDA) and also requires the CFPB to work with the U.S. Department of Housing and Urban Development (HUD) to create a new mortgage default and foreclosure database. The CFPB should act quickly to both commence its HMDA rulemaking and work with HUD to create and make public the sorely-needed national
mortgage default and foreclosure database.
The CFPB should link HMDA, the required foreclosure database, and data on mortgage modifications using the universal loan identifier required by Dodd-Frank.
- Linked databases with publicly available information on the individual loan level will be effective tools to enable the public to hold servicers and secondary market entities accountable for serving all borrowers and communities in a responsible and non- discriminatory manner. Additionally, linked databases would help the public, policymakers, regulators and lenders identify the critical relationship between loan terms and outcomes, and will reduce the burden on and facilitate data collection by servicers and secondary market entities.
- The CFPB should expand HMDA to include loan-level data on all loan modification applications and outcomes, and link these data to HMDA mortgage application data using a universal loan identifier.
The CFPB should expand HMDA data by implementing the changes mandated in the Dodd-Frank Act, as well as additional data enhancements. We recommend that the CFPB:
- Create normalized measures of credit score on a loan-level basis (i.e. for each loan application included in the HMDA data) that will facilitate fair lending analysis;
- Make public vital data on underwriting, and improve disclosure of loan terms and pricing;
- Expand the important and often overlooked multi-family and manufactured loan data in HMDA so that the public can more accurately track loans that support rental housing and manufactured housing; and
- Increase HMDA data’s coverage of the mortgage market by adding data on home equity lines of credit and reverse mortgages. Improve the data for rural areas by expanding the number of banks required to report HMDA data.
- Indicate whether the loans are Qualified Mortgages (QM) or Qualified Residential
The CFPB should improve the public dissemination of HMDA data. The Dodd-Frank Act empowers the CFPB to determine the format of data lenders make public. To allow HMDA data to better serve its statutory purpose, detailed data should be publicly available in easy-to- understand and easy-to-use formats for users at all levels of sophistication. The data made available should include:
- Summary data tables for novice users that are easily accessible and easy to manipulate;
- A website that allows users to run basic data queries and create interactive maps; and
- The “raw” data needed by sophisticated data users, which should include clear instructions for downloading and documentation of data fields.
The HMDA data is currently disseminated about 18 months after the end of the calendar year, compromising the ability of the data to reflect the latest lending trends. The CFPB should disseminate the data more often, on either a quarterly or semi-annual basis. The timeliness of the data will become even more critical since it will include information on loan performance and modifications. The CFPB and its sister agencies should regularly engage the public in discussions about how best to format and disseminate the data to the public.
The Dodd-Frank Wall Street Reform and Consumer Protection Act of 2010 mandates critical improvements to the Home Mortgage Disclosure Act (HMDA), the creation of a mortgage default and foreclosure database, and public disclosure of data on applications for Home Affordable Modification Program (HAMP) loan modifications. All three are important measures of whether or not financial institutions are meeting the housing needs of communities in a non- discriminatory manner.
The CFPB should prioritize its HMDA rulemaking. The new HMDA data is key to transparency in the mortgage market, and, even if the rulemaking is finalized this year, improved HMDA data will not be available to the public until 2016— six years after the passage of the Dodd-Frank Act. The new data’s usefulness in describing the marketplace and detecting risky or worrisome developments early is diminished with each year of delay. The enhanced HMDA data will allow HMDA to better serve its stated purpose, which is to:
- Determine if financial institutions are serving the housing needs of communities;
- Assist public agencies in distributing public-sector investments so as to attract private investment to areas where it is needed; and
- Identify possible discriminatory lending patterns.
The CFPB should work to implement a universal loan identifier to link all mortgage-related public databases, improve HMDA data by implementing the Dodd-Frank Act and our additional recommendations, and work with the Federal Financial Institutions Examination Council (FFIEC) to ensure that HMDA data are easily accessible to the public.
I. Link HMDA, Loan Performance, and Loan Modification Data to Provide a Comprehensive Picture of Borrower Outcomes
The Dodd-Frank Act requires the CFPB to work with HUD to create a mortgage default and foreclosure database and requires lenders to report additional data on their participation in HAMP. In order for these databases to be most useful in describing and understanding the lending marketplace, they must be linked with HMDA data at the loan level.
Universal Loan ID – Linking Databases
There are three data sets included in Dodd-Frank that are critical to shedding more light on mortgage lending and servicing in this country:
(1) Data enhancements to the Home Mortgage Disclosure Act (HMDA) dataset, (2) The establishment of a default and foreclosure database and
(3) The public availability of the Home Affordable Modification Program (HAMP) and proprietary loan modification data. (1)
These datasets are collected at different points in time and often from several institutions or reporting entities, and cannot be easily joined for research purposes. This is why the “universal loan identifier,” or universal loan ID, authorized in Dodd-Frank, is so important.
Universal loan IDs should first be used by lenders when they report loans under HMDA. Lenders will provide the loan numbers to mortgage servicers and servicers will be able to use the ID when reporting any defaults, foreclosures or modifications related to the loans. This process will relieve the servicer from re-collecting and re-reporting a variety of borrower and loan- related data pieces, including borrower income, race/ethnicity, gender, age and credit score, loan amount, term, APR and mortgage channel. Regulators, policymakers and the public would already have access to this information via the universal loan ID and HMDA. It is the key to increased transparency and accountability, without sacrificing efficiency.
We recommend that the CFPB:
- Require mortgage lenders and servicers to use and track universal loan IDs for all loans reported in HMDA;
- Include the year of origination in the universal loan ID, so data users can easily go back to that year’s HMDA to connect the data. The ID can easily be created from elements already in the HMDA Loan Application Register (LAR) data by joining or concatenating the following: (1) Year, (2) Respondent ID, (3) Agency and (4) Lender Loan Number (which is currently hidden in the public HMDA data); and
- Retroactively create a universal loan ID for loans reported in previous years’ HMDA data, and use it to link to the default and foreclosure database, conduct analysis of the linked data, and share its findings with the public.
- Create and make public a parcel identifier. The CFPB should collect true parcel identifiers, such as Assessor Parcel Numbers (APNs), and use those to create a database assigning all parcels a non-geographic alphanumeric code; that code should continue to be associated with the parcel over time and can be reported in the public HMDA data without creating privacy concerns. HMDA data users will be able to use that code to track, for example, piggyback mortgages and property flipping, but will not be able to pinpoint the location of the property beyond the census tract level.
The CFPB Should Include Loan Modification Data in HMDA
The performance of financial institutions in modifying loans is and will continue to be a major factor in determining whether they are meeting local housing needs. Assisting distressed borrowers and saving homes from foreclosure is integral to reviving the housing market. In this context, the CFPB should require lenders to report loan modification information in HMDA.
Loan modifications are clearly transactions that should be covered by HMDA, and would go to the heart of whether financial institutions are meeting housing needs. The reports on the outcomes of modification applications by housing counselors in California indicate that borrowers of color do worse than white borrowers when trying to obtain loan modifications. (2) This demonstrates both the need for and the value of accurate public information on modification outcomes. The data also meet HMDA’s statutory purpose of helping local government agencies determine how best to allocate resources. Many local governments today are devising Responsible Banking Ordinances that are meant to help them decide which financial institutions deserve their deposits and business. Loan modification data is among the most desired by local governments that seek to know which financial institutions are helping local residents avoid foreclosure and preventing vacant REOs from depressing neighborhood housing values.
There is precedent for collecting such data through the Making Home Affordable (MHA) program with the added requirement that servicers report MHA data by race and ethnicity in the public disclosure. The CFPB should work with Treasury to ensure that:
- All loan modification data, including action taken and outcome (principal reduction, interest rate reduction, forbearance, etc) are reported and made publicly available. This should include the reporting of loan modifications under the Home Affordable Modification Program (HAMP), as well as proprietary modifications; and
- Link this data to HMDA, loan performance data and the foreclosure/default database, through a universal loan identification number.
Loan Performance and Modifications Are Best Reported at Loan Level
The enhanced HMDA data required under Dodd-Frank will provide additional insights into the decision criteria that lenders use to approve or deny mortgage applications. That is a critical function for understanding the mortgage market, but it serves only as a limited snapshot of the performance of the mortgage lending institutions in this country. The HMDA data do not reveal anything about how the loans perform, how borrowers fare in the event of default, or whether borrowers in distress are able to obtain loan modifications to avoid foreclosure.
The additional datasets to be created under Dodd-Frank are meant to complement HMDA with data on loan performance and modifications to fill the existing gap in essential information. In order to maximize the utility of the new data for policymakers and the public, the new databases should be published at the same level of aggregation as HMDA, that is, at the individual loan level with geographic location specified at the census tract level.
The level of aggregation at which the data are published is one of two factors that will determine how useful the data are for policymakers and the public. Dodd-Frank allows, but does not require, the data in the default and foreclosure database to be aggregated at the census tract level. At that level, all the data will show is the number and percentage of outstanding mortgages in the geographic area that went into default within a given period of time.
Dodd-Frank specifies the minimum categories and aggregations of HAMP modification data to be released, but it does not limit the release to those elements. If only the minimum required data are included, the data will show only how many modification requests a servicer has received and whether the servicer agreed or did not agree to modify the loans. As useful as those data would be, they would not provide the information that is essential for policymakers and the public to understand how lenders and servicers are performing in their communities and whether the modifications are offered equitably and in a non-discriminatory fashion.
The purposes of the new databases are to promote transparency and accountability and to provide policymakers and the public with the data they need to make informed decisions. Loan-level data are essential to achieve that purpose, but further enhancements are needed. If the databases are linked with a universal loan ID, policymakers and the public will be able to use the data to better understand the complex relationships among loan and borrower characteristics, loan performance, and loan modification decisions.
Existing HMDA data, as well as the Community Investment Impact System data reported to the CDFI Fund, report detailed loan-level data with a tract-level geographic identifier. Those databases show that loan-level data with tract level geographic identifiers can be gathered and reported without undue burden on the reporting entities and are sufficiently anonymous to protect borrower privacy. The new databases to be created under Dodd-Frank must be equally detailed and linkable to enable policymakers and the public to have the information they need to promote policies that will foster a sound and sustainable housing finance system.
Secondary Market Entities in HMDA Data
Currently, HMDA data indicate to whom loans are sold, including Fannie Mae and Freddie Mac. As the CFPB reforms HMDA data, the CFPB must ensure that HMDA data remain effective in monitoring the activities of the secondary market in order to ensure their compliance with fair lending and consumer protection laws. We recommend that the CFPB work with HUD and Treasury to ensure that:
• The universal loan ID is able to capture sales to the secondary market that occur after the year of origination. The current limitation in tracking sales to the secondary market is that HMDA cannot be used to track secondary market activity that occurs after the year of origination, for example, when the lender may hold its loans in portfolio and sell its loans two or more years after they were originated.
II. Scope of HMDA Data Disclosure
The Dodd-Frank Act requires the CFPB to write rules requiring lenders to report additional HMDA data fields, including new borrower demographic information and a measure of credit score, and the CFPB should make all new data fields available to the public at the loan level. In addition, we recommend that the CFPB improve the loan pricing and loan quality information included in HMDA, and increase HMDA’s coverage of loan types and lenders to allow the public to determine if lenders are meeting the housing finance needs of their communities.
The Dodd-Frank amendments to HMDA require the CFPB to collect and make public credit score information, and the CFPB must make that information available on an individual basis for each loan applicant, that is, at the loan level. Individual credit score information, debt-to-income ratios and cumulative loan-to-value ratios are vital tools that will allow HMDA to better serve its purpose, to “help identify possible discriminatory lending patterns and assist regulatory agencies
in enforcing compliance with antidiscrimination statutes.” (3)
In particular, loan-level credit score information will allow regulators and the public to determine whether the lending disparities exposed by current HMDA data are an indicator that homeowners who should qualify for prime loans are systematically denied mortgage loans or steered into higher-cost or FHA loans as a result of their race or the racial composition of their neighborhoods. Providing aggregate credit score information would not increase the public’s ability to identify lending discrimination using HMDA data. Loan-level data can be provided with the applicants’ privacy adequately protected by presenting their credit score information using quantile ranges. In addition, normalizing scores as suggested immediately below protects loan applicants’ privacy by providing approximations of credit scores rather than specific scores.
We recommend that the CFPB make public two measures of credit scores, one that reflects the credit scores assigned to loan applicants by the credit bureaus and another that reflects lenders’ proprietary credit scoring models that are used for underwriting loans.
Credit Bureau Credit Score Reporting
Credit bureau credit scores are important tools for comparing loan decisions across lenders and for identifying lending discrimination across the mortgage market as a whole. The public and regulators will, for example, be able to determine whether African-American borrowers with excellent credit receive higher cost loans more often than white borrowers with similar credit.
Lenders that use credit scores in their mortgage underwriting should report loan applicants’ credit bureau credit scores to the CFPB and should also report the type of score (e.g. FICO or Vantage) and the range of possible scores (e.g. 350-850 or 500-990). Once lenders submit their annual HMDA data to the CFPB, the CFPB should normalize all of the commercial scores to a consistent scale and calculate credit score quantiles using all loan applicant data collected that year. The CFPB should use 20 quantiles, assigning each loan applicant a credit score value of 0-5th percentile, 5-10th…90-95th, 95-100th percentile. If a lender uses more than one credit bureau score when evaluating loan applications, the lender should report the score, type of score and scale for all scores used as well as the weighting the lender applied to the scores, and the CFPB should use appropriately weighted averages of the scores when calculating loan applicants’ credit score quantiles.
Proprietary Credit Score Reporting
Proprietary credit scores are critical tools for comparing loan application decisions made by an individual lender. By making public a measure of proprietary scores, the CFPB will allow the public and regulators to determine if a lender makes different lending decisions based on age, gender, race or other factors when deciding whether to lend to, or what kind of loans to make to, borrowers who have the same risk profile under the bank’s underwriting system, but differing protected class characteristics.
Lenders that create and use proprietary credit scores should report those scores and their meanings to the CFPB. Proprietary credit scores tend to span much smaller ranges than credit bureau scores, for example, they might span A-C or 1-4. The CFPB should examine the range of proprietary scores collected and create appropriate categories to represent lenders’ proprietary scores. The CFPB should, for example, match proprietary scores to a consistent list of values such as “poor,” “fair,” “good,” and “excellent” credit risk based on each lender’s own scoring system.
Categories for “Asian” and Language
The reported HMDA race and ethnicity data for Asian loan applicants obscure the diversity of Asian-Pacific Islander communities and have contributed to a relative dearth of HMDA analysis considering Asian borrowers. In those analyses that have been conducted, Asian borrowers tend to receive loans at rates comparable to white borrowers, though groups working in Asian-Pacific Islander communities know that these findings do not accurately reflect the variety of Asian- Pacific Islander communities or the wide disparities in credit access among these communities.
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 borrowers being sold mortgages in their native languages, only to receive English-only documents with significantly worse terms than those promised. 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 Limited English Proficient borrowers is an important indicator of fair lending, and more granular data could help local jurisdictions direct resources to these at-risk communities.
We recommend that:
- The category “Asian” for reporting race in HMDA should be broken into subgroups that represent the largest ethnic groups in the nation. California-based community organizations have suggested: Cambodian, Chinese, Filipino, Hmong, Indian, Japanese, Korean, Laotian, Thai or Vietnamese American, and Other Asian (to capture borrowers not of the nationalities listed in any revised HMDA data fields); and HMDA should include fields that indicate the primary language of the applicant(s); whether a loan negotiation was conducted in a language other than English; if so, what that language was; and whether loan and foreclosure documents were translated.
Age of Borrower
The Dodd-Frank Act requires the CFPB to add age of borrower as one of the demographic categories in HMDA data. Tracking the experiences of older adults in the lending marketplace is critical to ensuring fair treatment because predatory lenders targeted older adults for equity- stripping predatory lending in the late 1990s and 2000s, for example, and now abuses associated with reverse mortgage lending are on the rise.
We encourage the CFPB to provide detailed information about age in the raw data, specifically the exact age of the borrower. Summary and aggregate data provided to the public via the FFIEC web page should report age in five year buckets, but the publicly available raw data should have precise information so researchers and members of the public can more precisely track lending trends by age. This information is particularly important for seniors whose eligibility for reverse mortgages and other programs begins at age 62, an age that would likely be obscured in a 60-65 year old age bracket.
The Dodd-Frank Act requires the CFPB to collect many new HMDA data fields on loan terms and pricing as well as credit scores, but does not require the CFPB to collect vital data on loan underwriting. Regulators should work to ensure that lenders are making safe and sound loans in compliance with fair lending laws, but HMDA users are left out of that conversation because HMDA does not include many important underwriting data fields. Debt-to-income (DTI) and cumulative debt-to-income ratios are critical elements of both origination and loan modification decision-making. When coupled with the loan performance and mortgage default database via a unique identification number, these variables will help regulators and the public assess the performance of underwriting choices.
The Financial Stability Board identified five aspects of risky lending: lack of effective verification of income, unreasonable debt service coverage, inappropriate loan-to-value ratios, improper collateral management, and inadequate use of mortgage insurance. (4) DTI and LTV are two of the three top factors in loan decisions (the other is credit score.) Significantly, they are incontrovertible data points which are easily comparable across different lenders.
We recommend that the CFPB collect and make public the following data fields:
- Mortgage Debt-To-Income Ratios: Mortgage DTI should be calculated as the borrower’s final (after any resets) monthly mortgage payment (including principal, interest, taxes and insurance) as a percentage of the borrower’s gross income.
- Cumulative Debt-to-Income Ratios: Cumulative DTI should encompass not just the mortgage but all other elements of the borrower’s debt load. CDTI should be the sum of minimum payments required for all lines of credit and final mortgage payment divided by the income reported to the lender by the borrower.
- Loan-to-Value Ratios: With the decision by Fannie Mae to charge a fee to banks that deliver mortgages with high loan-to-value ratios, this variable has become a significant determinant in access to conventional loans. HMDA data should include the loan to value ratio. In instances where no appraisal was used, the variable should report a separate discrete outcome of Not Available (NA).
- Cumulative Loan-to-Value Ratios: Just as in the case of debt-to-income, lenders were able to escape scrutiny from HMDA users and engage in abusive piggyback lending because HMDA did not included a CLTV field. CLTV should be calculated as the ratio of all outstanding mortgage debt and the appraised value of the property.
- Reason for Denial: the CFPB should require all lenders to provide the top three reasons for a loan denial for loan applications denied.
The Dodd-Frank Act improvements to HMDA will require lenders to report price information for all loans, a significant improvement on the current requirements for reporting limited price information for high cost loans. As several Department of Justice settlements reveal, price discrimination often occurs within the prime market and involves differences of 25 to 50 basis points. This type of disparate treatment is missed by the current HMDA data that only reports price information for high cost loans. In addition, pricing discrimination may be reflected either in points added to the interest rate (typically mislabeled as “discount points” when they are not used to reduce the interest rate) or as a form of a yield-spread-premium or “overage”, or through direct fees, such as loan application, loan processing, or underwriting fees of various kinds.
We recommend that the CFPB:
- Provide detailed price information in the HMDA data and report precise price information for each loan rather than broad buckets that consist of ranges of 50 or more basis points, which would frustrate the ability of the public to identify price disparities; and
- Provide summary and aggregate data that include carefully constructed buckets of price information.
Certain risky loan terms and conditions are highly correlated with poor loan performance, and many of these terms were extremely popular with lenders during the past two decades. The Dodd-Frank Act-mandated improvements to HMDA data combined with additional fields on loan terms will help the public and regulators to identify risky loans.
We recommend that the CFPB write rules requiring lenders to report data fields including:
- Piggy back loan indicators: A categorical variable should be added to HMDA to identify second and third position loans issued in tandem with a first mortgage at origination; and
- Stated-Income: A nominal variable should be added to HMDA indicating if income was
“stated” rather than documented.
In addition, we recommend that the CFPB write rules implementing the Dodd-Frank Act’s loan quality-related HMDA data fields as follows:
- Adjustable Rate Mortgage Reset Date: a number corresponding to the number of years before a loan’s interest rate resets. For example, if a loan resets 18 months after origination, the reset date should be reported as 1.5.
- Balloon Loans: An indication of whether a loan is a balloon loan is important to understand if certain geographical areas, borrower groups, or types of loans are more likely to have balloon payments. The CFPB is also likely to allow QM balloon loans for certain lenders. Understanding how often these loans are offered and how they perform is important for determining if the QM balloon exemption needs to be adjusted in the future.
- Negative Amortization: a ratio of the maximum possible outstanding debt to the appraised value of a home.
- Total Points and Fees Payable at Time of Loan Application or Origination: These variables should be expressed in dollars. We recommend at least two data fields, one for fees that are built into the interest rate as points (discount points, yield-spread-premiums, or overages) and one for direct fees (such as fees for the loan application, loan processing, underwriting, and settlement costs). By capturing both the APR and points and fees, the Dodd-Frank enhancements will further our knowledge of how much fees contribute to total loan costs as well as heighten fair lending enforcement.
- Prepayment Penalty: The Dodd-Frank requirement to include prepayment penalties in HMDA is important. Prepayment penalties add to the probable cost of a loan but are not factored into APR calculations. If a prepayment penalty is applied to the loan, HMDA data should indicate the fee percentage of the penalty and the number of years that the penalty will be in effect.
See Appendix A for a table of suggested formats for all proposed loan quality and pricing variables.
QM and QRM
The implementation on the Qualified Mortgage (QM) and Qualified Residential Mortgage (QRM) rules will have a profound impact on the mortgage market, including impacts on safety and soundness and access to loans for low- and moderate-income people and people of color. In order to assess whether the QM and QRM rules improve safety and soundness without unduly restricting access to credit, the CFPB should include an indicator of whether loans are QM and/or QRM.
We recommend that the CFPB add a new field to HMDA that indicates:
- If loans are QM or QRM. The categorical variable could be 0 if the loan is not QM or QRM, 1 if the loan is QM, and 2 if the loan is QM and QRM.
- If a loan is QM, an additional data field should indicate if it is a safe harbor or rebuttable presumption QM.
Loan Purpose and Type
Home Equity Lines of Credit (HELOCs)
Currently, HMDA allows for optional reporting of HELOCs. The CFPB should require lenders to report HELOCs under HMDA. Many small business owners finance their businesses through HELOCs, especially in immigrant communities. When HELOCs are harmful, they threaten homeownership, business survival, employment of workers, and have serious consequences for our economy.
Since HELOCs tend to remain in the originating lender’s portfolio, including HELOCs in the HMDA data will increase the public’s understanding of mortgage servicers’ loss mitigation and loan modification decisions and will allow regulators and the public to uncover conflicts of interest that prevent lenders from offering otherwise qualified homeowners loan modifications, including principal reduction modifications.
We recommend that the CFPB require lenders to:
- Report HELOCs and home equity loans under HMDA;
- Report the loan purpose for HELOCs and home equity loans, including whether the loan was used to fund a small business, for home improvement, for debt consolidation, or for another purpose. The data will open up new and important areas of research that explore the linkage between housing, business ownership and job creation and maintenance; and
- Indicate whether the HELOC line of credit was initiated as a line of credit or made as a lump sum for the full authorized amount (which typically indicates that a HELOC was actually used as a second lien loan).
The impending retirement of the Baby Boom generation elevates the importance of ensuring fair and equitable treatment of older adults in the lending market. The tremendous loss of wealth due to the foreclosure crisis and the Great Recession contributes to pressing needs for older adults to tap into their home equity by taking out reverse mortgages. Reverse mortgage lending, however, remains ripe for abuses as revealed by the recent Consumer Financial Protection Bureau report. (5)
We recommend that the CFPB:
- Amend HMDA data to include reverse mortgage lending. Reverse mortgages could be added as a category under the existing loan purpose data field.
- Require reverse mortgage lenders to report whether reverse mortgage payments are received as lump sums or line of credits, and require lenders to report key loan terms such as points and fees and interest rates.
Multifamily data (MF) is among the least known, understood and utilized of HMDA data, and there is no other publicly accessible source of data that shows whether lenders are meeting critical rental housing needs. HMDA multifamily data are important to communities trying to rebound from the foreclosure crisis because a number of significant sources of funding for affordable housing, such as redevelopment agency set-aside funds in California, are no longer available. The demand for affordable housing (and affordable housing lending), is increasing as public sector funding for affordable housing is decreasing.
Multifamily HMDA reporting is inconsistent and suspect, perhaps because commercial lenders that are HMDA reporters are not clear on what is required to be reported under HMDA. Also, in states like New York and Texas where consolidation, extension and modification agreements (CEMAs) and modification, extension and consolidation agreements (MECAs) are prevalent, HMDA excludes a large number of loans that would otherwise be counted as refinance loans and indeed do count as such when lenders undergo Community Reinvestment Act (CRA) exams.
The CFPB should take into account the quality of multifamily loans and their impact on the stability of the affordable housing stock, and ensure that HMDA data include indicators of affordability for multifamily lending. If housing projects were not designed to be affordable for the long-term, or worse, were speculatively underwritten and thus likely to have a destabilizing effect on the community, the loans financing those units should not be classified as supporting affordable housing. There are already too many examples of banks claiming community development credit on CRA exams for affordable housing that ultimately is of poor quality or destabilizes communities. Just a few examples:
- Washington Mutual used to report as part of its CRA and community relations activities a large number of multi-family loans for buildings in low- and moderate- income neighborhoods where the market-rate rent was affordable to the community. Yet, these same properties were of such poor quality that advocates deemed them “slum housing.”
- The problem of “predatory equity” is well-documented in New York City where banks have underwritten loans in rent-regulated, lower-rent apartments for much more than the building is worth, such that the only way to make a profit is to evict rent-regulated tenants (often through harassment or unscrupulous eviction practices and/or neglect needed repairs and maintenance in the building. Similar practices have been documented in California where owners financed rent-controlled buildings with the intent to evict long-standing, lower-rent paying tenants and convert the buildings to condos or tenancy in common.
We recommend that the CFBP require lenders to:
- Report construction loans, in addition to the three types of permanent loans currently reported (home purchase, refinance, and home improvement);
- Report MECA/CEMA agreements in HMDA as a distinct type of permanent loan (in addition to refinance, home purchase, and home improvement); and
- Identify loans for affordable housing, and add a field describing why the housing is affordable so communities can evaluate both the quantity of affordable housing being financed and the potential long-term impact on the community. Four indicators that will help communities understand the quantity and quality of affordable housing that lenders are financing are:
- Percentage of units that are affordable: HMDA data should indicate the percentage of units in the development that are or will be affordable to residents of the following income levels: low (<50% AMI), moderate (50%-80% AMI); middle (80%-120% AMI) and upper (120%+ AMI); (6)
- The number of housing units related to the loan.
- If the building is deed-restricted, which would indicate that it receives public subsidies (such as LIHTC or other financing mechanisms), or is otherwise required to maintain particular levels of affordability over a specific period of years.
- The Debt Service Coverage Ratio at the time of origination. For example, a Debt Service Coverage Ratio that is below 1.2 suggests that the loan is speculative and possibly based on the expected removal of many of the existing tenants paying moderate rents. Such loans deserve particular scrutiny as they could undermine the affordable housing or cause financial pressure that could result in physical deterioration due to neglect of needed repairs.
Manufactured housing now houses approximately 8% of American households, and most of those households are made up of seniors or low-income families. Military families are also a sizable part of the population that lives in manufactured housing. Manufactured housing finance is unusually heterogeneous: loan terms can range between 5 and 30 years, interest rates fluctuate between prime and 20%, and collateralization is inconsistent.
The existing HMDA data do not provide much clarity about these loans in spite of their significant role in providing mortgage credit for traditionally underserved populations. Absent improvements, users of the data can at best be skeptical of their findings.
We recommend that the CFPB require lenders to report:
- Separate property type statuses for loans on manufactured homes secured as real property versus those classified as personal property; and
- Only loans made on manufactured homes that are intended for use as a habitation. Many manufactured homes will not be sold for this purpose. Manufacturers sell many manufactured homes to construction companies and schools. However, given that a buyer must stipulate design features for a home, there can be no doubt about the ultimate purpose of the home.
Coverage of Lenders
The CFPB has an important opportunity to revisit exemptions from HMDA reporting. Current technology facilitates data collection and reporting. Exemptions based on reduction of burden that seemed justified several years ago are now obsolete due to technological improvements. The current exemption for banks with assets of $42 million or less misses a segment of banks that are vitally important for rural and underserved markets. The Housing Assistance Council (HAC) estimated that the asset exemption for the year 2009 exempted 989 banks and that 70 percent of these banks were headquartered in rural communities. Banks that are headquartered in non- metropolitan areas and have their branches only in non-metropolitan areas are exempt from reporting requirements. HAC calculated that over 3,000 banks were headquartered in non- metropolitan counties and therefore likely qualified for the HMDA exemption. Non-bank institutions can be exempt if they meet various criteria and originated up to 100 loans; such an institution that made 99 loans could be important in a local market and missed by the current HMDA rules.
We recommend that the CFPB eliminate or significantly narrow these exemptions and capture more lending in rural counties and smaller metropolitan areas, which is imperative for ensuring non-discriminatory and equitable treatment.
III. Improvements to Public Disclosure of HMDA Data
The development and implementation of formats for the public release of the HMDA data can and should be carried out through an extensive and intensive process of interaction with the full range of public users. HMDA was the result of public pressure to allow the public to analyze lending patterns. Historically it has been a powerful resource for education, enforcement, and development resulting in billions of dollars in reinvestment and affordable housing programs. Over time, however, due to the difficulties with the formats for public release of the data, the data are now mainly used by technically-skilled organizations and academics. We want to make the data accessible so that it once again is used by scores of local organizations every year.
Frequency of Dissemination: The HMDA data is now disseminated about 18 months after the end of the calendar year, compromising the ability of the data to reflect the latest lending trends. The CFPB should disseminate the data more often, on either a quarterly or semi-annual basis. The timeliness of the data will become even more critical since it will include information on loan performance and modifications. Loan performance can change significantly within a three or six month time period due to variations in lending practices, economic conditions, or interest rate fluctuations. If data on loan performance is reported on an annual basis, important changes during the year will be missed, as well as the ability to relate those changes to lending practices or economic factors. Likewise, trends in modifications can change significantly within a year due to changes in public or private modification programs. Annual data will not allow for the tracking of the impact of these changes.
Public Access: The current format of online release and data has actually limited the ability of all but the technically savvy and academic groups to use the data. The web site is organized in a confusing manner. For example:
- The web site could benefit from better organization and from more explanations of the content of the different HMDA, Census, and HUD estimates of metropolitan statistical area (MSA) median family incomes (MFI). The different data sets need to have non- technical names that indicate what is actually contained in each data set. For example, sets of actual raw HMDA, census, and MFI data should not be called “reports”, but should be labeled as actual data. The preformatted tables of HMDA data for each lender in each MSA should be labeled as something like “Lender MSA Tables” and not “Disclosure Tables”.
- The website should provide a PDF version of a guidebook that contains the formats for the “Aggregate” and “Disclosure” tables, the format for the LAR raw data (i.e., the data dictionary), the formats for the census and MFI data, directions for downloading the LAR program and data as well as the other available data sets, and directions on the use of the LAR software. At present, one needs to figure out how to locate and download these data sets and the LAR software before one can see what the data look like. Much of this information on the software data sets is now essentially hidden in the “help” function which one can access only after successfully downloading the program.
- Another improvement would be allowing users to operate the LAR software program on the site rather than requiring users to download the program to their own computers. The online guide should also include instructions on how to do this.
- Where the site contains links to various data sets, the links need to be clearer. For example, from the main FFIEC site, if the user clicks on “Census Reports,” the user has to search to find the actual census data; and these “reports” are unclear. In fact, if the user clicks on the “HMDA” link rather than the census link, the user gets to the page that links directly to the Census Data Products.
The LAR Software: The Windows software for the use of the raw LAR data has some significant problems and is unstable. It also fails to provide tables for valuable information that links reporting lenders and parent companies, which is essential to understand how lenders split their marketing and products across different applicant and community characteristics. For example:
- The program seems to freeze up or abort, or fail to load. This seems to be an especially serious problem with the 2010 data. Testing needs to be done to understand how the software downloads and operates on a variety of computers and operating systems.
- The software and downloading instructions need to be more clear and should not use jargon that is not clear, such as “populate the fields” rather than “fill in the blanks” for a search.
- The “Institutions” table of lenders can only be printed for the whole nation (not exported) and is organized by lender ID rather than lender name. For the 2011 data, the list is 139 pages long (with no page numbers). The data for a lender’s parent company is essentially hidden in the utility called “Export All Institution Records” – which, itself, is a confusing title. Moreover, once exported, there are no field names and there is no guide to identify what data elements are in each field. There should be a summary table of lenders for each MSA that would provide the total applications, loans, and parent company data in a
- Lending institutions, particularly bank holding companies, are complex and can have a number of affiliates and subsidiaries that report HMDA data. Over the years, identifying all of the affiliates and subsidiaries for a parent institution has been difficult. The FFIEC website provides information in PDF format about parents and affiliates, but the PDFs are difficult to use and can be incomplete. HMDA data need to be significantly improved in terms of easily identifying parents and their affiliates and subsidiaries.
- Because the exported data do not contain field names, this makes it necessary for each user to create a separate field name spreadsheet into which they can paste the exported data.
- While data for an individual lender can be extracted by either the lender ID or name, users need to understand that the lender IDs are only unique when both the agency and lender numbers are combined. Having a field that combines these codes into a single ID would create a truly unique code for each lender.
- The table for the LAR Code Sheet found in the “Information” tab has no page numbers – and the codes for some fields are split between these unnumbered pages.
- The software does not provide data in a format that can be used for mapping. Being able to export data aggregated by census tract would provide data that could be used for mapping or provide online maps of MSAs and/or counties showing lending by census tract.
Format Issues: As the CFPB adds new data fields to HMDA and creates the loan performance database, the CFPB and the FFIEC need to be committed to the principle that the data should be easily available to and useable by the general public. In this context, every effort needs to be made to release the data on a loan-level basis. Beyond this principle, there are some issues with the present formats for the public release of the data that severely restrict its broad public use. Aside from the issues discussed above for the software formats, there are issues with the formats for the preformatted tables. Examples of these issues include:
- A critical deficiency is the limitation on exporting the “Aggregate” and “Disclosure” tables only as PDFs and not as a TXT or CSV file that can be used in spreadsheet (Excel) programs. This is a step backwards from the CD released for the 2004 data that provided for exporting these tables in an excel format, though it still lacked table names and field headings in many cases.
- The table names themselves are confusing. “Disclosure Tables” should be called “Individual Lender MSA Tables” and the “Aggregate Tables” should indicate that they are the “MSA Aggregate Tables,” for example.
- Table #1 (for both the Aggregate and Disclosure formats) provides some summary data by census tract. This format could be revised so that it could be used easily in mapping programs.
- When members of the public request an institution’s data directly from the lender either prior to the full HMDA release in September or because the lender had to correct its data, the lender should be required to provide the data in an electronic format that is essentially the same as the LAR records produced by the HMDA Raw LAR data software. Dodd- Frank empowers the CFPB to require lenders to provide data in a format the CFPB requires and is user-friendly, not simply formats that are convenient for the lender. We have experienced instances of lenders providing raw data in hundreds of pages of PDFs rather than in a data file.
- As a corollary to this, when a lender is required to make significant corrections to its data, there should be a public posting link on the HMDA web site so that the public knows which lenders have faulty data and that they can request the corrected data (though it would be even more helpful if the corrected data were posted in the website).
- Finally, when a lender has made significant corrections to its data in any MSA, the lender should pay to have a new set of Disclosure Tables produced for each of those MSAs.
A Program to Support Public Use of HMDA Data: Because the process for creating access to and formats for the HMDA data is intended to be a dynamic process, not a regulatory process, it should be implemented with full and regular interaction with public users. Some recommendations for the scope of this program include:
- The CFPB and FFIEC should create an ongoing program to survey community groups’ problems using the data, technical capacity, and support needs.
- The FFIEC and CFPB should work with community organizations and members of the public to make a web page that is highly accessible for beginners and still useful for advanced users of the data. Beginners should be able to access easy-to-understand tables in excel format that contain summary information on lenders for metropolitan areas, counties, and groupings of census tracts. Beginners should be able to ask the web page basic queries. For example, the web page should have the capacity to show how many home purchase loans African-Americans or other demographic groups receive in various geographical areas. The web page should also display data in interactive maps. Advanced users should have easier access to properly labeled and formatted raw data.
- Other federal agencies such as HUD require local jurisdictions to use data and maps. The CFPB should coordinate with these agencies to incorporate HMDA and foreclosure/loan performance data in the various agency data and mapping tools. (7)
- The CFPB should maintain a website for the public to share uses, templates, public domain software, etc.
The CFPB should ensure that HMDA, loan modification, and mortgage default and foreclosure data are made public at the loan level and are linked with a universal loan identifier. These data are key to determining if financial institutions are meeting local housing needs. To find out if lenders are meeting mortgage credit needs in a non-discriminatory manner, HMDA data must be expanded to include loan-level measures of credit score and other vital underwriting factors and loan terms. Finally, the improved data must be made broadly accessible to the public. The CFPB should meet regularly with HMDA users, including community groups, to ensure that the HMDA website, prepared data tables and raw data are accessible and useful to the public.
Thank you for the opportunity to provide input on these critical changes to HMDA. If you have any questions or would like to discuss these recommendations further, please contact:
List of Contributors
Adam Rust, Reinvestment Partners
Alexis Iwanisziw, Neighborhood Economic Development Advocacy Project
Barbara Van Kerkhove, Empire Justice Center
Calvin Bradford, National People’s Action
Chip Bromley, Organize!Ohio/Ohio Fair Lending Coalition
Debby Goldberg, National Fair Housing Alliance
Jaime Weisberg, Association for Neighborhood and Housing Development
Jim Campen, Massachusetts Affordable Housing Alliance
Josh Silver, National Community Reinvestment Coalition
Kevin Stein, California Reinvestment Coalition
Lance George, Housing Assistance Council
Lisa Donner, Americans for Financial Reform
Spencer Cowan, Woodstock Institute
(1) See Sections 1094, 1447 and 1483 of the Dodd-Frank Wall Street Reform and Consumer Protection Act for the language pertaining to these datasets.
(2) California Reinvestment Coalition, Chasm Between Words and Deeds VIII: Lack of Bank Accountability Plagues
Californians, April 2012, available via http://www.calreinvest.org.
(4) Financial Stability Board. “FSB Principles for Sound Residential Mortgage Underwriting Practices.” April 2012. http://www.financialstabilityboard.org/publications/r_120418.pdf
(5) Consumer Financial Protection Bureau, Reverse Mortgages: Report to Congress, June 28, 2012, available via http://files.consumerfinance.gov/a/assets/documents/
(6) The income categories presented here are CRA categories. These may make sense since a large segment of lending institutions reporting HMDA are covered by CRA. However, the CFPB should consider income categories used by the Low Income Housing Tax Credit program, Section 8, and other programs to determine which income categories make most sense in this proposed data field for multifamily lending.
(7) See (http://egis.hud.gov/cpdmaps/ as an example of a federal agency mapping tool.