By Bridget Berg
“That which is measured improves.
That which is measured and reported improves exponentially.”
– Karl Pearson
Because mortgage fraud takes time to develop, many institutions, particularly newer ones or shops with new management teams, may be at risk and not realize it. As a veteran in mortgage fraud risk management, I frequently hear comments like these:
- “We have been in business two years and haven’t had any fraud.”
- “Our delinquency rate is low, so I know we don’t have a fraud problem.”
- “My controls are working because we haven’t had a large loss.”
Everything is fine—until it isn’t.
Why is this risk so difficult to pin down?
There are several reasons why mortgage fraud is so difficult to detect and can create a false sense of security.
Mortgage fraud is a relatively rare event. We estimate that only 0.65 percent to 0.70 percent of applications have material misrepresentations. In the post mortgage crisis era, many of these frauds are averted during loan processing. However, the potentially large loss level makes the tolerance for fraud very low.
Many mortgage frauds are undetected. When the economy is doing well and real estate prices are rising, a borrower who committed fraud-for-property is likely able to stay current on the loan. When stressors occur, such as a slow economy or drop in real estate values, the fraud-for-property loans are more likely to default. Even when these loans default, they are generally passed through as credit losses and are not recognized as fraud.
Types of mortgage fraud can be cyclical. In the run-up to the financial crisis, false down payments, inflated income, and straw buyers were some of the more common issues. After the crisis, there was a shift from origination fraud to loss mitigation fraud, such as short sale schemes, REO bid rigging, and loan modification scams.
No one likes to air their dirty laundry. Even when detected, financial institutions are very cautious about disseminating information about fraud issues, especially when they involve internal parties. Investigations involving internal parties may be shielded under attorney-client privilege, so even senior managers may be unaware of these events.
The lag between the fraud activity and its discovery can be lengthy. Consumer fraud that impacts financial institutions tends to be “fast-moving.” Stolen credit cards are used immediately, check kiting is discovered within a day or two. This is not the case with mortgages.
The Discovery Delay Problem
When a known or suspected mortgage fraud is identified, a Suspicious Activity Report (SAR) must be filed with FinCEN within 30 days of discovery. SAR information is available online in the form of reports that track the number of reports filed over time. This has been used as a primary measurement of mortgage fraud levels for many years. However, the report(s) only tracks the date the SAR was filed, and not when the fraudulent activity took place. Thus, the report is easy to misinterpret. In 2013, FinCEN provided a one-time chart to show the timeline between when the activity occurred and when it was reported. The report was quite enlightening, as it showed the difference between the peak of reporting and the peak of activity was about five years (from mid-2006 to mid-2011). Most of the fraud was not identified until long after it occurred.
The delayed findings were heavily influenced by the economy and close examination of defaulted loans. The representations and warranties that are part of most secondary market transactions usually carry a life-of-loan warranty that loans are free from misrepresentation. If a significant market downturn happens where many loans go delinquent and the resulting losses are high, investors are apt to mitigate their losses by identifying fraud in the defaulted loans. They can then ask for a repurchase or reimbursement of the loss from the institution that sold them the loan.
The problem with the delayed discoveries is that by the time a financial institution gets the feedback, it is too late to mitigate the situation. Another issue is that the lack of rapid feedback gives management a false sense of security about their fraud risk. Insufficient fraud detection practices or an undetected scheme could continue for years. Finally, the types of fraud most likely to be discovered early are not representative of overall fraud risk. This can create a skewed picture and could lead to poorly-informed risk management decisions.
When and how is mortgage origination fraud discovered?
There are several times over the life of a loan that fraud can be discovered:
Prior to the loan closing – Of course, this is the best time to discover and prevent it. Verification processes, up-front fraud tools, underwriting, and pre-fund quality control account for most of these discoveries. Unfortunately, few institutions have strong tracking mechanisms to determine how much fraud is being averted, which loans, and what method accounted for the detection. This is a missed opportunity.
Standard post-fund QC reviews by originating lender – Depending on the size of the institution, QC sampling usually looks at about 10% or less of the loan population. This may be supplemented with an adverse or targeted sample. QC reviews are completed the first two to three months after closing. They usually include new verifications including credit reports, income, and asset verifications. Fraud can easily go undetected in a QC review because it is not a fraud-specific review. Credit reports will consistently show most new debts, but asset reverifications may not be returned and may just show the current asset picture versus what it was at the time of the original asset document. Income verifications may not be returned, or if part of a collusive scheme, be returned with false information again. Fraud found in these reviews are most likely to be undisclosed liabilities or job loss that happened prior to closing. Organized schemes usually are sophisticated enough to pass through this type of review.
Investor QC – Investors typically perform some level of QC in the first few months after origination. Again, fraud is not the primary focus of these reviews and they are also most likely to find undisclosed liabilities and job loss prior to closing. Indemnifications or repurchase requests are possible outcomes.
Early Payment Default (EPD) reviews – Loans that become 60 to 90 days delinquent in their first six to 12 months are often subjected to a special QC review. Fraud found in these reviews will be like the earlier post-fund QC reviews, however, a higher percentage of the reviewed population will have issues. Unsophisticated fraud schemes (such as a one-off straw buyer flip) that defaulted immediately are often detected in this type of review. Indemnification or repurchase requests are likely if loan was sold with reps and warrants in the contract.
Severe loss reviews – Many investors will perform QC or root cause analysis on loans with large losses, even those aged more than a year. Egregious frauds and schemes may be detected here, and recourse will be sought from the originator.
Fraud investigations – These take place over the life of the loan, and are usually triggered by a tip about a particular loan, loan officer, appraiser or a fraud scheme. As the investigation continues, the sample will broaden as related loans are identified. This is where most fraud schemes are fully identified. Again, recourse from the originator will be an outcome if it was purchased with reps and warrants in the contract.
Early Feedback Bias
As time progresses, different types of fraud are likely to emerge. Because standard post-fund QC reviews do not provide a good picture of the level of fraud or the type of fraud in loans originated, relying on QC results as the only indicator of fraud risk levels can be misleading.
As we have seen, these reviews tend to skew towards undisclosed liabilities and miss other types of fraud. A good example of how risks shift over time can be seen in Fannie Mae’s monthly reporting. https://www.fanniemae.com/singlefamily/mortgage-fraud-prevention.
The figure below, for example, looks at a 2015 vintage portfolio at two different points and shows the changing types and levels of fraud risk. A difference of 16 months dramatically changed the breakdown of fraud types. As the vintage continues to age, liabilities will likely continue to shrink.
Early feedback from QC reviews may bias risk managers and operations managers to target just the specific issues that were the focus of review. For example, the large share of liabilities-as-a-fraud type provides the most immediate feedback, so lenders may focus on this risk to the exclusion of the broader fraud picture. The issues that emerge later may not garner the same level of attention or root cause analysis and prevention focus. But these later discoveries are those that come from the defaulted loans – the issues that are more likely tied to loss.
The Role of Predictive Analytics
As we have seen, mortgage origination fraud may have a long discovery delay, and different types of fraud are more likely to emerge at different times. For these reasons by the time a problem has been identified, it may have become very large. To prevent this, we need to know which loans are more likely to have fraud, either to target prevention efforts, or to measure risk levels and trend changes.
At CoreLogic, we have the benefit of a consortium-based population of millions of loan applications, and thousands of examples of loans with fraud. We use these populations for predictive modeling. The current model incorporates the most predictive features based on analysis of over 2000 data points regarding loan, borrower, and property elements. The output is a risk ranking score from 1-999. As a point of reference, the score identifies 60 percent of fraud in the top 10 percent of scores. Our clients use the score trends to monitor their fraud risk levels over time and benchmark against the entire consortium.
We also compile the consortium scores into a National Mortgage Application Fraud Index. This is available publicly and is updated quarterly. The index is a tool that lenders can reference to gauge the increasing or decreasing risk levels over time. Because it was modeled from loans that had time to season, and includes both pre-fund and post-fund findings, it provides a complete picture of current mortgage origination fraud risk. (You can learn more about the index, find our quarterly updates, and annual mortgage fraud reports at corelogic.com.) As of the second quarter of 2017, we are seeing the highest risk level since we began our index in 2010, so this is a good time to evaluate whether your institution is focused on a short-term view of mortgage fraud, or a long-term view.
Here are some best practices that our clients are currently using to help them detect and prevent mortgage fraud:
- Within your fraud management tracking system, include structured data fields for: the timing of the suspicious activity; the discovery date; the method of detection; and the type of fraud.
- Report fraud rates by vintage and seasoning age, and report on each origination/application year separately. Compare the years at consistent seasoning ages, for example 2015 vintage as of June 2016 versus 2016 vintage as of June 2017.
- Include as many years of history as possible in your tracking and reporting, as fraud is cyclical over long periods of time, and include pre-fund findings as well as post-fund findings.
- Keep a timeline of changes in controls or policies that are likely to impact fraud risk, such as income verification policies or reduced documentation programs.
- Track leading indicators of fraud risk for trending, including by origination channel, third party originators, loan programs, or loan purpose (purchase or refinance).
It is easy to rely on information that appears to be timely regarding current fraud risk, but be cautious and include a longer-term view and approach to effectively manage the full breadth of mortgage origination fraud, including fraud that may take several years to identify.
Bridget Berg is Senior Director, Fraud Solutions Strategy, for CoreLogic, where she leads the delivery of fraud risk management solutions to the mortgage industry. She can be reached at