Jay Brinkmann, chief economist and senior vice president of research and economics for the Mortgage Bankers Association (MBA), has testified before the Federal Reserve Board of Governors at a hearing entitled, "Potential Revisions to Regulation C—Implementing the Home Mortgage Disclosure Act (HMDA)." Below is Brinkmann's oral statement before the committee, as prepared for delivery.
"My name is Jay Brinkmann and I am the chief economist and head of research at the Mortgage Bankers Association (MBA). I very much appreciate the opportunity to participate in today's hearing of the Federal Reserve Board on potential revisions to its Home Mortgage Disclosure Act (HMDA) requirements.
"I would like to address essentially five questions or areas that need to be addressed. First, what data should be required? Second, how should the data be reported? Third, what should be used as the universal mortgage identifier? Fourth, what data should be made public? Finally, I will address some issues regarding multifamily data.
What data should be required?
Dodd-Frank already requires a significant expansion of the required data elements, although some are left to the discretion of the Bureau of Consumer Financial Protection (CFPB). In addition, we understand the Federal Reserve is looking at some potential additions beyond what is in Dodd-Frank. We have no objection to an expansion of the HMDA data elements as long as that expansion is consistent with the stated purposes of HMDA, the elements are consistent with what is already collected, and the changes would not pose unnecessary burdens on lenders. It should be understood, however, that no matter how many additional data elements are required they will not serve as a reliable proxy for the range of credit models or credit decisions given the sequential nature of the credit decision, variations in decision-making processes among lenders, as well as variations in shopping behavior and self-selection of credit terms by borrowers.
One issue the Fed must keep in mind in determining what data elements to collect is that HMDA requirements should not turn into a safe harbor of allowable credit variables to be considered when making a loan. Freezing credit models into an official sanctioned set of variables would have a deleterious impact on credit availability going forward, limiting the growth of lenders who believe they have a better idea of how to do things. For example, over the years some lenders have come to believe that credit scores are not as important as the number of times a potential borrower has been late with housing-related payments. Some lenders now will simply refuse to make a loan to a borrower who has walked away from a previous mortgage, or appears to be positioning himself or herself for such behavior. None of these considerations are captured in any of the proposed HMDA data elements, nor should they be.
How to report?
In determining definitions and file formats for potential data items, the Fed should use the standard and uniform definitions developed over the last ten years by the Mortgage Industry Standards and Maintenance Organization, Inc. (MISMO). Reliance on MISMO definitions would greatly reduce the regulatory compliance burden by allowing lenders and vendors furnishing HMDA compliance services to pull from existing MISMO-compliant databases to report under HMDA. This would reduce the errors associated with entering data a second time for HMDA purposes and reduce the phase-in period for trying to interpret and then implementing new HMDA definitions. In addition, MISMO standards have already been adopted by Fannie Mae and Freddie Mac.
Reliance on the MISMO dictionary and standards would also help deal with the ambiguity surrounding some of the data elements specified in Dodd-Frank. For example, Dodd-Frank requires that credit scores be reported. MISMO recognizes that there is no such thing as a single credit score, so while it has a field for the score, it also has a field for the credit score vendor (such as Vantage Score or FICO), and the reporting agency. Rather than asking lenders to map multiple fields into a single number to be reported to the Fed, a number that likely would not appear in any credit file nor be used in the credit or loan pricing decision, the Fed could simply ask for the multiple fields dealing with credit scores and do its own mapping depending on whether it is doing a company-level or industry-level analysis.
I cannot stress enough the extent of the regulatory burden that HMDA and other reporting and compliance requirements place on the industry. The largest shares of investments in technology today are going to reporting and compliance needs, with no direct benefit to the companies or their customers. I would hope that the Fed would keep this burden and its costs in mind and minimize future changes in HMDA once these changes are made. Relying on MISMO would not only minimize costs but it would allow minor tweaking of data requirements in the future with less burden.
What to use as the universal mortgage identifier?
The industry already has a uniform mortgage identification number that is issued through the Mortgage Electronic Registration Systems Inc. (MERS). This MERS number is used by a very high percentage of lenders and is integral to numerous origination and secondary market functions. It would cause considerable confusion and unnecessary implementation expense to impose a new mortgage identification protocol on the industry. Reliance on the MERS Mortgage Identification Number (MIN) allows loans to be tracked from origination through sale in the secondary market and subsequent servicing, and is valuable in identifying and preventing mortgage fraud.
For the Fed's purposes, a further advantage of using the MERS MIN is that it would help prevent double counting or the failure to count loans altogether. For example, the current practice of eliminating loans purchased as closed loans from correspondent banks lowers the apparent coverage level of HMDA. In an effort to see what was missing from HMDA, the MBA several years ago did a matched-pair analysis of correspondent loans and found that a large percentage did not have a matching loan in the retail/broker data. Use of the MERS MIN would largely solve the problem of estimating coverage levels because it would permit an explicit matching between retail/broker originations and correspondent originations, it would provide a matching of loans originated in one calendar year and sold in another, and it allow loan data to be double checked against other data sources like Fannie Mae, Freddie Mac and Ginnie Mae.
What to make public?
Federal Reserve staff have developed considerable expertise in the analysis and interpretation of HMDA data. Their annual article in the Federal Reserve Bulletin is the source of information on HMDA for most analysts. In recent years, Fed staff have gone the extra mile to conduct analyses beyond the HMDA data to answer topical policy questions.
However, while it is proper and customary for a firm's regulator to have access to confidential data, care needs to taken before those data are made public. While we see tremendous risk of widespread identity theft if all of the HMDA data elements were to be released in their collected form, particularly when those data are combined with other publicly available data, under Dodd-Frank, decisions on such release now lie with the Board and later the CFPB. The lending industry has poured tremendous resources into safeguarding the private information of our customers, and we have paid large fines for lapses. No doubt we would face the potential of additional fines and public recrimination were we to make the proposed HMDA data elements available to the public at large. That is why any liability associated with the collection and release of these data pursuant to Board rules should lie with the Fed. Moreover, the Board should provide guidance on how lenders should deal with requests that come directly to them for these data.
To a certain degree, we would support a greater release of credit data in some form. While it still would not solve all of the statistical problems associated with trying to mimic credit models with these data, it would go a long way to putting to rest once and for all charges of racism that have been hurled at the industry by various groups over the years that have no basis in fact. The econometric problems of omitted variables, multicolinearity and spurious correlation would still remain, but sufficient data would be available in the public domain to refute most of these charges.
What multifamily data should be reported?
MBA estimates that the 2008 HMDA data contained information on 95 percent of the multifamily loans made that year based on the number of loans, but covered only about 61 percent of their dollar amount. The average multifamily loan in HMDA was about $1.7 million while the average missing loan was about $18.9 million. We question the benefit of expanding the reporting requirements to include a relative small number of high-dollar multifamily projects.
Clearly, the data elements associated with single-family lending are not applicable to any but the smallest multifamily projects. Variables like race and credit score do not apply to limited partnerships, corporations or real estate investment trusts. We suggest that the Fed should examine the usefulness of the multifamily data it collects now with an eye to scaling back the requirement rather than going to large lengths to expand the reporting requirements to cover a small number of large dollar projects.
In conclusion, in making changes to the required data elements of HMDA, the Fed should look carefully at what is needed considering the new data requirements under Dodd-Frank and their costs, integrating the data requirements with what is already being collected, and using data definitions and identifiers that are already in common use. In addition, the Fed should be very concerned with the privacy related issues that would attend a wholesale public release of the new required data elements."
For more information, visit www.mortgagebankers.org.