US Mortgage Market and the Mother of All Databases

The American Banker ran an interesting article on the need for a database containing all the mortgages in the US.

Your humble blogger has been advocating for this since before the credit crisis in conjunction with a loan-level database for structured finance securities.  Just as the loan-level database is needed if market participants are going to be able to analyze and value structure finance securities with all the useful, relevant information, a national mortgage database is needed by mortgage market participants.

Like the structured finance loan-level database, data in the national mortgage database should be standardized.

Like the structured finance loan-level database, data in the national mortgage database should be made available for free to all market participants.

Europe and in particular the ECB and BoE are leading the way in showing how both the structured finance loan-level database and the national mortgage database can be rapidly built.  Specifically, the central banks are making providing data to the loan-level database an eligibility requirement if a structured finance/covered bond security is to be pledged as collateral at the central bank.

There is no reason that the Federal Reserve could not also do this.  There is no reason why the SEC could not make this a requirement of its revision of Regulation AB.  There is no reason that this requirement could not also be added to the Covered Bond legislation moving through Congress.

Why do some mortgage servicers appear to be modifying a great many more loans than others? How many loans are getting principal reduced, rather than interest rate reductions? How many loans in a given neighborhood are more than 30 days late? 

Answering such questions may be critical to craft effective policy responses to the housing crisis. But it's hard to get definitive answers, because data on mortgage performance is incomplete and often expensive. 

Most of the detailed information on mortgage performance is gathered by the two companies, CoreLogic Inc. and Lender Processing Services Inc., and sold in various forms to regulators for sums that can reach hundreds of thousands of dollars for a particular data set. 

Members of Congress, state banking regulators and academics say they have been regularly stymied in attempts to access the data. In many cases they cannot afford it. Sometimes regulators who have the data cannot share it because of proprietary arrangements with data providers. 

... Proponents of the idea say such a database is long overdue. 

"There are no complete data sources and there is over-reliance on third-party vendors," said Richard Neiman, former New York State banking commissioner and former member of the Congressional Oversight Panel for the Troubled Asset Relief Program, who submitted a proposal for such a database in 2009. "We as regulators and stakeholders should not be relying on third party vendors whose data is incomplete and subject to inaccuracies." 

Neiman, who is expected to join PricewaterhouseCoopers as vice chairman of global regulatory practice, said the forthcoming database could be as valuable as the information
collected under the Home Mortgage Disclosure Act, which is used to evaluate possible discrimination in loan approvals. 

"Having a mortgage performance database would provide a similar window on mortgage servicing," he said. "Just like HMDA was used to identify violations in the origination process, a national mortgage database could provide a similar window on servicing
and modifications." 

CoreLogic and LPS depend on voluntary reporting by servicers. And their data has blind spots: CoreLogic covers about 41 million first-lien loans and 7 million second liens,
out of a total of about 60 million mortgages in the country, while LPS has detailed information on about 40 million first liens. 

LPS says building a system like the one called for by Neiman and Dodd-Frank would be unnecessary because currently the mortgage industry, which uses the data to establish benchmarks, is helping to defray the cost of gathering it. 

"Today anybody can license from LPS or from CoreLogic," said Kyle Lundstedt, a managing director at LPS. "We charge the mortgage industry, which benefits, and it is not like the American public is not getting the benefit too, because the regulators who manage the industry on our behalf are also accessing the data. So it leaves me befuddled when someone says, 'Wait a minute, we need to make that data free.' " 

... Having more comprehensive and reliable data could be helpful in assisting distressed homeowners, said Rep. Brad Miller, D-N.C. "There seems to be guesswork on what works and what doesn't work in regard to modifying mortgages," he said. "In the time I have been in Congress, we really have had to rely upon the industry to tell us what is going on and what they have told us has not been reliable. It has not always been accurate and not complete — sometimes plain not truthful — so having an independent source of information would be an enormous benefit." 

The Office of the Comptroller of the Currency, which, together with the soon-to-be-phased-out Office of Thrift Supervision, gathers data on the servicing operations of nine banks accounting for 34 million loans, has made strides in tracking loan performance. Among other advances, the agency's quarterly Mortgage Metrics Report has helped establish common definitions for such categories as subprime and prime. Before the establishment of Mortgage Metrics, these definitions differed from bank to bank. 

However, the report only covers 63% of the mortgages in the country and it does not cover the substantial holdings of second lien loans on bank balance sheets. 

Alan White, a law professor at Valparaiso University in Indiana who has studied various data on mortgage servicing, said the OCC data that is available publicly does not allow
researchers to get a comprehensive picture. 

"The OCC data is a biased sample," he said. "What you get are summaries. The loan-level data is not available anywhere to the public or to researchers. You can get summary statistics from the Mortgage Bankers [Association] once a quarter and then the OCC and OTS metrics comes out, but they give you the numbers they want to give you." 

Another issue is the need to protect borrowers' privacy. 

"We don't share that data because it is supervisory data and the data at loan level includes privacy information," said an OCC official, who did not want to be identified. 

However, White said it should be possible to have a system that releases detailed loan information without compromising borrowers' personal information. "You wouldn't want public releases that included the names and addresses of borrowers," he said. "But I think that there is also certainly with the bank regulators a lot of concern about protecting bank proprietary information. If you were able to look at what kinds of loss severities resulted from foreclosures and modifications, you might get an idea of which servicers are doing a better or worse job for investors and the timing and severity of losses." 

Your humble blogger developed and patented the information system that Professor White would like to see.  The system protects borrower privacy with the standards that Congress approved for protecting patient privacy under HIPAA.

Researchers can gain access to some databases at the Treasury Department, but that comes with restrictions, according to White. 

"Treasury has been very secretive about their data," he said. Although the agency releases periodic report cards on the Home Affordable Modification Program, "and it is possible to
gain access to some of the proprietary databases, but they have licensing agreements and most of the time the licensing agreements prevent researchers from disclosing information that would hurt the reputations of the subscribers to these services" — the subscribers being the banks. 

Andrea Risotto, a Treasury spokeswoman, said she had no knowledge of special access agreements with academics. "I can only answer for Hamp," she said. "We have a database online that literally has every data element for every homeowner who has ever participated or even submitted one piece of paper to apply for the program." 

Lundstedt at LPS argued that a national database would be expensive.... However, government officials say that licensing agreements with LPS and CoreLogic actually end up costing government agencies unnecessary amounts of money and that the new mortgage database could help resolve what has been an enduring issue. 

"We are buying both LPS and CoreLogic mortgage performance data, and other agencies are doing this as well, so those companies are collecting a lot of money," said a senior HUD
official, who also requested anonymity....  

Some agencies cannot even afford all of the data they say they need to do their job. "A lot of this data is very expensive," Neiman said. When he was New York's banking commissioner, he said, "we had some budget restrictions, so our first thought was can we reach out to some of our federal  counterparts, and they are under contractual restrictions for sharing it." 

And when the department could afford to buy private data, Neiman said, it found inaccuracies. "On a quarterly basis we were having to go back and correct data with regard to foreclosure filings," he said. "It is for that very reason that we established a mandatory requirement within the state that institutions have to file pre-foreclosure information with us as well as the filing of lis pendens" — pending foreclosure actions — "because we were continuously finding inaccuracies and lack of completeness from the private sources of that information." 

There are formidable challenges to establishing the national mortgage performance database.
HUD officials say that it could take several years to build it and that no money has been appropriated for the project. (Dodd-Frank does not set a deadline.) Another question is how
the agency would build the database. Would HUD (which already publishes loan-level performance data for Federal Housing Administration loans on its Neighborhood Watch website) gather the data itself or could it buy the information from vendors like CoreLogic and LPS? 

"A potential barrier is will they agree to sell it to us given what we are supposed to do with the data," the senior HUD official said. "Would we render their products obsolete by
giving away the information at the census tract level Congress asked for?"