
7 minute read
TO REDUCE FRAUD RISK IN 2021, LESS IS MORE, THANKS TO SCIENCE
TO REDUCE FRAUD RISK IN 2021, LESS IS MORE, THANKS TO SCIENCE
By Frank McKenna | Point Predictive
The economic fallout of 2020’s pandemic brought historic unemployment levels that we expect will lead to an increase in mortgage fraud rates in 2021. Experts expect to see more income and employment misrepresentation, higher incidences of credit washing, and an emergence of synthetic identity schemes that could make fraud detection even harder next year.
But as damaging as it is to lenders’ bottom lines, fraud itself may not actually be the industry’s biggest challenge. In a case of the cure being worse than the disease, current methods of detecting fraud may rival the impact of the original threat.
Perhaps the key to stopping more fraud is to do less work, not more. As fraud rises in the next 12 months, perhaps the answer is working smarter, not harder.
THE HIGH COST OF UNTARGETED FRAUD PREVENTION
Preventing mortgage fraud is extremely important to brokers, lenders, and servicers. But over the last 10 years, the process of identifying where to look for it has become increasingly difficult because new technologies are introduced to the masses and the fraudsters, corporate strategic priorities ebb and flow, and no process is ever 100 percent secure. In the effort to identify and prevent fraud in all forms, the industry has progressively layered more and more rigid policies and manual processes on top of a fluid problem space. This result is the industry looking for fraud everywhere, instead of better anticipating where it is likely to be.
Statistically, mortgage fraud is a relatively rare event, affecting only about 1 percent of loans in any given year. Yet current approaches to fraud prevention have buried underwriters under mountains of paperwork and third-party verification data. This ultimately burdens borrowers with lengthy, onerous loan processing and months of duplicative or unnecessary documentation and compliance checkboxes, while the rates of fraud detection are only marginally improved, at best. This disruption comes at a high price.
In 2021, it will be more important than ever to explore reliable ways to accelerate the mortgage loan approval process while also improving our fraud detection abilities. To do this, lenders will need to focus on the top fraud prevention issues plaguing the industry:
• The high rate of reviews and false positives that inundate underwriters and threaten the efficacy of the process of fraud prevention.
• Limited streamlining options for more effective routing and passing lower risk loans more quickly through the system.
Three Steps to Saving Time, Reducing Loss, and Accelerating Low-Risk Lending through Smarter Fraud Prevention
My company, Point Predictive, analyzed more than 80 million loans that have contributed to its anti-fraud lending data consortium. Our analysis demonstrates the opportunity that lenders have in 2021 to improve operating results for their mortgage portfolios.
1. REDUCING FRAUD REVIEW RATES AND FALSE POSITIVES SAVES TIME
The most common types of fraud are misrepresentation (with income misrepresentation still the single most frequent type of mortgage fraud), early payment default, and repurchase risk. They drive far too many foreclosures and burden lenders with enormous financial loss. So, all the “i-dotting” and “t-crossing” that goes into preventing them is understandable.
In extensive data analysis with lenders, Point Predictive data scientists discovered that nearly half of all lenders saw up to 80 percent of their loans subject to some fraud flag or review process that required manual intervention. This high rate of review may be giving lenders a false sense of security.
High review rates result in expensive and time-wasting misdirection of effort. We found that false positives can be as high as 1,000:1 (only one out of thousand actually indicates fraud) and can lead to legitimate risks being overlooked. To compound the issues, in many cases Point Predictive discovered that underwriters began to overlook the red flags when they became too commonplace. In this case, the more legitimate red flags underwriters must clear, the more likely they are to miss the true fraud when it is presented to them.
And lenders are already struggling with volume and speed. One of my colleagues is refinancing a home that he’s owned for 11 years. His package has been waiting on underwriting for nearly four weeks now. Naturally, with issues like that plaguing the process for extremely low-risk loans, manual attention to flagged loans simply can’t scale. Burdensome review makes the underwriting process much more difficult and costly for lenders and consumers than it has to be. Fewer and more accurate alerts earlier in the loan process have shown as much as 65 percent reduction in workload in underwriting and the pre-funding QC, so this is an important area on which to focus.
2. USING DATA AND SCORING MODELS IN ADDITION TO ALERTS AND RULES-BASED MODELS PREVENTS MORE FRAUD MORE EFFICIENTLY
Is it possible to reduce the number of reviews and false positives and still feel confident that we can prevent fraud? With the right information, it’s not only possible, it’s far more efficient. And we can thank science for that, data science to be exact.
A study of alerts and rules, the most used fraud detection tools, showed that in isolation, they could be very poor predictors of risk. They have value, but the complex decisions behind assessing a loan for fraud requires more.
When used in conjunction with alerts and rules, multiple data points and interactions sourced from partners across the industry can more accurately dictate risk levels and provide a more comprehensive and trustworthy basis for how and where to focus your review efforts. Data points can include insights and patterns gleaned from statements on prior credit applications as well as third-party risk, such as geographic clusters of fraud and brokers or other entities associated with suspicious applications in the past.
This type of solution is conceptually easy to understand but very difficult to put into practice. It’s impossible for underwriters to efficiently wade through this volume of data and connect these dots manually. Enter modern risk scoring technology, which can boil down this tsunami of data points into easily consumable scores that efficiently manage down risk.
Since fraud manifests itself in many ways, including borrower risk, property risk, broker risk, and even risk associated with the loan program, it is possible to individually evaluate the risk of those aspects of risk and take a more targeted approach that doesn’t involve subjecting every loan and everything in that loan to the same level of scrutiny. If the broker’s risk score indicates high risk, but the borrower is considered low risk, manual review activities can be focused on the risky aspects of the loan while the less risky aspects can receive a more appropriate and cost-effective treatment.
3. STREAMLINING LOW RISK MORTGAGES SAVES MONEY
With better fraud detection models and targeting of high risks before funding, lenders can save significantly by routing loans that score low for risk through more efficient processing channels.
Our data shows that only 10 percent of loans need to be subjected to the greatest scrutiny. The 30 to 50 percent of loans that have extremely low risk of fraud, early default, or repurchase (and which typically have a lower exposure) can be processed more economically by Level 1 underwriters who focus on compliance-based checks for half the cost of a senior underwriter. It doesn’t make sense to spend the same kind of time on a low-risk loan with an exposure in the hundreds as you would on a high-risk loan with a financial risk in the tens of thousands.
Risk factoring the workload in this way reduces underwriting costs by as much as $500-1,000 per loan. It can significantly improve operational efficiency and free up forensic review and quality control analysts to focus on the higher risk loans.
How significant is the impact of this approach to fraud risk management? For a lender with a monthly volume of 10,000 loans, the pool that can be streamlined would be 3,000 loans. Assuming a savings of $500 for each loan, this represents a monthly savings of $1,500,000 and an annual savings of $18 million.
FINDING BALANCE IN FRAUD PREVENTION
Before 2007, the mortgage industry was customer-focused, but the rise of fraud has driven lending toward an increasingly unwieldy compliance and fraud focus for 13 years now. Now that we’ve become aware of the underlying policy gaps that allowed fraud to flourish and gained new insights about the prevalence of different types of fraud, it’s time to stop treating all borrowers like they are criminals.
2021 is the year to shift to a balanced, sustainable mortgage lending approach that prioritizes customers while clamping down on fraud.
It comes as no surprise that unique approaches to data science will lead the way in transforming how the industry looks at fraud by greatly speeding up underwriting through a more comprehensive and concise understanding of risk.