The Buyback Epidemic -- What’s Fraud Got To Do With It?

Register now

Buybacks (repurchases) have devastated the subprime sector while showing signs of invading the alt-A and prime markets as well. A quick check of several websites reveals lists of over 50 lenders that have gone out of business or been forced into undesirable mergers or acquisitions over the past four months. This challenging environment has forced many investors, lenders, brokers, insurers and even regulatory agencies to search for the root causes that are having such a negative affect in loan quality and performance.

One of the key-contributing suspects is fraud. But how do we quantify the effect that fraud has had in this wave of buybacks and how do we develop effective solutions to mitigate the impact of fraud in our lending markets? As the pioneer of the automated fraud detection market in 1997, I would like to share with you some of my observations of the different fraud categories and their impact on the recent surge in buyback activity, as well as some suggested solutions to each individual fraud issue.There are several different database analyses available that attempt to quantify the dozens of different mortgage fraud categories. Certainly one the most prevalent, based on the largest database of known fraud defaults, is Fannie Mae’s. This information is also very current including updates through February of 2007 and happens to parallel our automated fraud findings. Income, the largest category at 25%, includes both income and employment misrepresentations. For the past few years, "stated-income" loans, along with "no income no asset" loans became an addiction for lenders and investors who became satisfied with an AVM and a neighborhood flip analysis as their only fraud detection tools. During the past year, dramatic market changes exposed this myopic approach creating a strong demand for income/employment verification tools. Recent studies indicate that as much as 70% of stated-income loans contain falsified income or employment causing them to be nicknamed "liar loans." In many circles NINA loans are now referred to as NINJA loans (no income no job). There are currently three income ranges, one payroll register income and several IRS tax income verification tools. Intelligent fraud platforms rap rules and scoring around these data reports and can actually set up intelligent cascades to maximize the effectiveness and efficiency of the overall income verification process. There are at least three employment verification databases that can identify such things as self-employment, nonexistent businesses, hidden liens and bankruptcies.

Property (16%) indicates that "a specific material fact about the property and/or the comparable sales was misrepresented." There are several neighborhood analyses that measure flip and foreclosure comparable activity within a close proximity to the subject property. Three or so also analyze the subject’s sales history for flips, foreclosures and non-arms length transactions. Only two rap intelligent rules and scoring around all of these quality measurements.

Assets (15%) "indicate that the borrower funds information was inflated or fabricated." This is one of the more elusive automated verifications, one platform verifies several different types of assets, but borrower funds verification is still on the horizon.

Only one fraud detection platform provides liability (13%) verification coverage. Specific areas include foreclosures, bankruptcies, liens and judgments and other assets owned.

Occupancy (11%) is one of the most important frauds due to its loss severity with a 30% higher average loss than other types of fraud due to the property damage risk associated with non-owner-occupied foreclosed and REO properties. Two fraud platforms contain multiple occupancy verification tools and both provide applicable rules and scoring.

Value (7%) is lower than our platform’s value category at 10%, but still surprisingly low to most lenders particularly in comparison to the heavy reliance on AVM products over the past few years mentioned above. There are several AVM cascade models available. Only two fraud platforms have AVM information imbedded in their platform rules and scoring configuration.

SSN (7%, "a significant discrepancy in the SSNs used to qualify the borrowers") and credit (6%, "the borrower’s identity and/or credit history was/were misrepresented") definitions appear to overlap somewhat, which would also align their combined affect (13%) much closer to our ID theft category (12%), which includes SSN variances, ID theft and other credit issues. ID theft is one of the fastest growing crimes in America and there are a few ID verification tools out there. Also, a few vendors specialize in SSN verifications with the Social Security Administration and at least two platforms include SSA verifications. There are dozens of tools that verify OFAC and SSN death lists, but only one platform matches ID information to driver’s license records.

Three of the aggregated fraud platforms provide various watch list and third party performance report cards, while one of them provides loan level broker and appraiser license verification.

In summary, today’s mortgage fraud landscape is made up of a complex myriad of different fraud schemes and types, each with an array of automated fraud tools decked against them. When shopping for these tools lenders and investors have the option of beau tic shopping for multiple standalone solutions or they can visit a Wal-Mart Super Center (aggregated fraud platform) that has everything under one roof. In either case, quick tests against actual repurchases, or other forms of credit losses, or known fraud cases (blind sample tests) will indicate a pre-funding fraud identification rate ranging from 40% to 70% or more. An extrapolation of these results can quickly measure the ROI and loss mitigation effectiveness of each solution.

Steve Halper is the founder and CEO of DataVerify.

For reprint and licensing requests for this article, click here.