Home   //   DataDriven Insights Blog

Overstated Income & Undisclosed Debt May Drive Fraud in Today’s Market

April 12, 2021  //  BY First American

New Techniques and Technologies Needed to Address Purchase Fraud-for-Housing Risk

First the good news: mortgage fraud, in all its forms, was down in 2020, as it has been much of the last decade. Now the not so good news: market dynamics, changing workforce demographics and more sophisticated fraud schemes, such as synthetic identity, all point to heightened risk for fraud, particularly fraud-for-housing, in the not-too-distant future.

The ingredients are all there: the mortgage market is shifting from refinance (read low fraud risk) to purchase; low rates and tight inventory are pushing up average home prices at a double-digit, year-over-year pace and the size of the average mortgage hit a new high in 2020. So, FOMO, (fear of missing out) is rising for both homebuyers and SFR investors.

Meanwhile, the Mortgage Bankers Association’s 2021 forecast calls for a steep drop in refinances, a steady rise in purchases and an overall market contraction of approximately 25 percent. Having added capacity to handle last year’s volume, many lenders and mortgage brokers will now be chasing share in a smaller overall market. In the past, this has led to looser underwriting and increased income and asset misrepresentation.

Finally, add new factors like the rapid and seismic shift to remote work due to COVID-19, the continued expansion of the gig economy (now 30 million strong) and the growing sophistication of synthetic identity fraudsters and the outlook for fraud changes dramatically.

Types of mortgage fraud and their motivation

Generally speaking, there are two types of mortgage fraud: fraud-for-profit and fraud-for-housing.

Fraud-for-profit usually involves multiple players in the mortgage process—real estate agents, mortgage brokers, appraisers, loan officers, straw buyers, etc. It can also take various forms, such as employment and income, debt, property value and condition to maximize profits on a loan transaction or to ultimately defraud the lender or the borrower. It can be done by a single person, or as a network of individuals. While still a concern for lenders, fraud-for-profit is at a historic low.

On the other hand, fraud-for-housing is more prevalent and more subtle than its fraud-for-profit counterpart. Fraud-for-housing is usually committed by borrowers, sometimes with the knowing assistance of loan officers or others, who misrepresent or omit relevant details about employment and income, debt and credit, or property to obtain a home or an investment property. The most common forms of fraud-for-housing are overstated income and undisclosed debt. Other kinds of fraud-for-housing can include falsified documentation, straw buyers, investment income misrepresentation and out of state and foreign investors.

Industry tools for fraud detection

Today’s in-workflow alert tools, like our industry leading FraudGuard® solution, are effective in detecting various fraud schemes. They’re one of the main reasons traditional fraud is very low today and has been for most of the past decade. These proven solutions are designed to detect fraud, compliance risk and data integrity issues during loan origination. These types of tools use natural intelligence based on triggering factors from millions of past loan applications. They are also customizable to reflect a particular lender’s risk tolerance level. However, they can be overly sensitive to certain variables and end up generating higher levels of false positives which, in turn, add time to clear the alert volume.

One of the more common false positives is due to common name matching. For example, a common name, like Smith, matched an industry ineligible list and upon further review it was found that the loan party was not the same person. Real estate owned (REO) is also a common variable that can trigger alerts. In a number of cases, errors are due to omission of data. For example, uncovering properties that are owned free and clear that were not disclosed on the Schedule of REO. While the borrower may not have a monthly mortgage payment, they are still responsible for taxes, HOA fees and other maintenance expenses.

False positives can create unnecessary reviews of loans that, in reality require a very low level of diligence related to risk. They can ultimately slow the approval process and create trust and adoption issues with users. This is especially a headache for large volume lenders that need a high level of fraud detection but at the same time are trying to drive down cost and accelerate decision making by reducing reviews.

Until recently, larger lenders had only one option when it came to reducing false positives: customization and regular calibration of their traditional fraud tools.

Recently, First American Data & Analytics introduced a second option that uses newly emerging technologies that leverage both natural intelligence and machine learning artificial intelligence (AI) – the AppIntelligence® Score (AI Score).

The AppIntelligence® Score from First American Data & Analytics

AI Score, part of the AppIntelligence Suite of analytics, is designed to overlay workflow alerts and brings to bear a whole new analytics-driven approach. AI Score utilizes First American Data & Analytics’ proprietary predictive fraud indices, employing both natural and artificial intelligence machine learning technologies. It simultaneously runs proprietary models and sub-models to measure risk factors including undisclosed debt, synthetic identity, income, employment, early payment default (EPD), and loan participant risk review.

Recently AI Score was run against a large national lender’s production and was able to identify more than 50 percent of highest potential fraud risk by reviewing as little as 10 percent of loan applications. While results may vary depending on the lender, this targeting enables lenders to focus their reviews on the most at-risk loans, while streamlining loan approvals and reducing operational costs on less risky applications. In addition, AI Score has a retro-scoring capability, so lenders can run a portfolio to identify which loans would have been flagged by the model. This retro-scoring capability will not only catch the known frauds but the frauds that lead to EPDs that were not identified by the lenders current process.

Large-volume lenders can benefit from this new-found operational efficiency while mitigating EPDs, buybacks and costly distressed asset care and maintenance. This may appeal to delegated correspondents that re-underwrite 100% of the registered loans in some capacity. New tools like this will also accelerate portfolio acquisitions and servicing transfers.

Fraud detection is a moving target that must meet present and future challenges. With the prospect of fraud-for-housing on the horizon, lenders can now leverage state of the art predictive analytics, machine learning and AI to get out ahead of the next wave of potential fraud and protect themselves.

Looking for a single source for mortgage analytics, risk, compliance, and valuation? Visit https://dna.firstam.com/appintelligence-suite for more information on the First American Data & Analytics AppIntelligence Suite. To learn more about AI Score or to connect with one of our experts, go to https://dna.firstam.com/mortgage-risk-management.


Home   //   DataDriven Insights Blog