Demystifying Mortgage Origination Fraud Statistics

Sometimes, mortgage origination fraud is easy to detect and other times it is frustratingly difficult. When a borrower deliberately lies about income, employment or identity, the fraud is clearly visible. But what about the case of an owner-occupied loan where the borrower never moves in, but still makes the payments as agreed? Even more intriguingly, what about instances of first-payment or early-payment default where more than one in four are actually origination fraud in disguise?

Regardless of how challenging mortgage origination fraud is to detect, it also turns out to be one of the most difficult financial fraud types to analyze because lenders generally do not know which loans will be determined fraudulent until 24 to 36 months after funding. As a result, most origination fraud analyses are looking in the rear view mirror and commenting on market conditions that are no longer current.

The widely available, regularly produced reports detailing the state of U.S. mortgage origination fraud fall into three general categories (the first two of which are looking, primarily, in the “rear view mirror”):

1. Reports based on confirmed frauds reported to governmental or quasi-governmental bodies to fulfill legal or regulatory obligations. 

2. Reports based on criminal and civil mortgage fraud cases reported in the public record.

3. Reports based on predictive indices (including models and alerts) derived from current loan applications, where outcomes are generally not yet known.

On the surface, the content of these three report types can appear conflicting and may even present opposite conclusions. But similar to the ancient parable of six blind men touching an elephant and disagreeing about the nature of the beast, each of the major industry fraud reports is accurate while using different data sets to deliver conclusions from a particular point of view.

Reports Based on Suspicious Activity Reports

The Mortgage Loan Fraud Update by the federal government's Financial Crimes Enforcement Network details statistics on suspicious activity reports related to mortgage loans and is a well-regarded example of the first report type. The report provides a solid, retrospective view of mortgage fraud that occurred three to five years ago, but has just now come to light. In their latest report, Second Quarter 2011 Analysis, SAR filings were up nearly 100% on a year-over-year basis, but according to the report, that is indicative of problems from “the height of the real estate bubble” and “is directly attributable to mortgage repurchase demands and special filings generated by several institutions.”

This FinCEN report also analyzes new quarterly filings with recent suspicious activity (90 days prior to filing) to assess current fraud trends. Given the short time window between activity and filing, these are generally a combination of prefunding confirmed fraud (income, occupancy, assets, liabilities, or identity), settlement/closing fraud (funds disbursement fraud, forged lien releases, etc.) and distressed property fraud (short-sale misrepresentations, appraisal fraud, etc.).

Reports Based on Civil and Criminal Cases

A well-known example of the second report type is the Mortgage Fraud Index from FraudBlogger.com. The report uses an index calculated on active criminal and civil mortgage fraud cases. In general, this is another good trailing indicator of the state of mortgage fraud from past years. The lag time between the commission of the fraud and a fraudster's day in court can be as long as (or longer than) the lag in SAR filings. Many active mortgage fraud cases today relate to events that occurred five or more years ago. For example, in a recent major win for the mortgage industry, a former real estate agent was convicted of defrauding 14 mortgage lenders and 34 homeowners out of more than six million dollars between 2006 and 2008. (For details, see Real Estate Agent Convicted of Defrauding Lenders.)

Reports Based on Predictive Indices

Mortgage industry solutions providers generally produce the third type of mortgage fraud report. The two best-known examples are the CoreLogic Mortgage Fraud Trends Report and Interthinx Mortgage Fraud Risk Report. Both vendors offer a predictive index that attempts to capture the level of fraud found in current originations, based on each vendor's proprietary fraud-risk evaluation model, which is trained on lender-reported fraud data. This report type offers a barometer of overall fraud risk and includes both prefunding confirmed frauds (that are stopped by the lender and generally do not result in financial loss) and likely (but not confirmed) post-funding frauds.

Some of the conclusions about the current fraud risk level reached by the vendors are similar. For example, the latest CoreLogic report indicates overall fraud risk appears to have stabilized as “after a 20% increase in 2009, the CoreLogic Fraud Index…remained relatively flat throughout 2010 and the early part of 2011.” Similarly, the latest Interthinx report indicates that its “Mortgage Fraud Risk Index…has remained essentially unchanged for the past six quarters.”

In other instances the vendors may reach quite different conclusions, such as which geographic areas pose the greatest fraud risk. The CoreLogic report indicates that Chicago (606XX) is the riskiest area for fraud (measured at the three-digit ZIP-code level). The Interthinx report offers three choices: Stockton, Calif., as the riskiest metropolitan statistical area; Nevada as the riskiest state; and Bakersfield, Calif., (93304) as the riskiest five-digit ZIP code.

Making use of this seemingly contradictory vendor-supplied information

Return to the parable of the elephant and realize each vendor is describing the world as it sees it. Neither vendor sees all loan applications from the entire mortgage industry, unlike the government's FinCEN repository for all SAR filings.

That means each vendor must extrapolate from incomplete information. Each vendor is also making a prediction about fraud, based on statistical models of how fraud was perpetrated in the past. There is always a measure of uncertainty in any statistically based analysis.

Finally, neither vendor measures in a way that allows apples-to-apples comparisons. We won't know if either vendor has predicted correctly until three or more years have passed, when we can look back at the confirmed fraud reports on all of today's loan applications.

Perhaps the most comprehensive report is the FBI's annual Mortgage Fraud Report. The agency's latest 2010 Mortgage Fraud Report references all of the sources listed above, along with additional reports of each type. The FBI report also does a commendable job of pointing out the differences in measurement parameters, reporting windows, and definitions across the various reports.

Conclusion

The best advice about reconciling mortgage origination fraud reports and statistics is not to try. Inconsistencies and discontinuity are inherent in the nature of the documents because each has a particular point of view based on a specific set of assumptions.

Recognize instead that each offers unique value when taking into account whether the report is retrospective or predictive. Use retrospective reports to interpret the past and use predictive reports to adjust your current and future risk-management policies and procedures. Incorporating both forward-looking and backward-looking report conclusions as part of your ongoing prevention systems and policies is your best defense against mortgage origination fraud.

David Johnson is the vice president of fraud and consortium solutions for CoreLogic.