How to defend against mortgage fraud in a high-volume market

Following a record-setting year that saw nearly $3.8 trillion in loan originations, industry observers are predicting a smaller, though still substantial, market in 2021 with a significantly larger share of purchases. The latest estimate from the Mortgage Bankers Association predicts that volume will decline by roughly 16% this year to $3.2 trillion as the refinance market cools and purchase volume climbs to a healthy $1.67 trillion level.

The continuing low interest rate environment, work-from-home trends and maturing millennials are all driving purchase demand. This demand, however, is bumping up against low housing inventory, and has driven up home prices to new highs: 9% higher year-over-year with the average loan amount growing to a record $402,200 in December.

If the past is predictive, the coming shift to a purchase market, combined with higher home prices and the enticement of quick appreciation, may result in a resurgence of fraud for housing schemes. The question is: will our industry be ready?

Lulled into complacency
In recent years, the mortgage market has skewed heavily toward refinances. These types of transactions, which involve fewer participants and often occur between a lender and an existing client, tend to be less susceptible to fraud.

Purchase transactions, on the other hand, have more players —
buyers, sellers, real estate agents, appraisers, loan officers and mortgage brokers — resulting in a greater likelihood of fraud.

Many new entrants into the mortgage field — both new fintech lenders and younger staffers at traditional firms — didn’t experience the mortgage crisis firsthand. All they have experienced is the refi-boom of the last few years, in which fraud for profit and fraud for housing have been extremely low. As a result, they may be less attuned to common fraud schemes, and may believe: “We’re doing everything right. Fraud isn’t a problem, and we have a fraud alert tool in place anyway.” This, of course, is exactly when fraud risk grows.

Not a replay of the early 2000s
There are some industry trends that may bear watching: for example, the return of limited document non-QM lending, increased demand for investor and fix-and-flip loans, and the resurgence of third-party originations through wholesale and correspondent lending.

The changing economic landscape, the digital transformation of mortgage origination and real estate settlement, and the pandemic are also creating new challenges in detecting fraud and verifying income and employment. For example, the growth of the gig economy means that nearly 40 million Americans are now their own employers and no longer W-2 borrowers. Similarly, advances in digital lending are reducing in-person interaction during originations and closings. The work-from-home trend, necessitated by the pandemic, is most likely a permanent option at some companies.

New ways of predicting and detecting fraud

Fabricated income and undisclosed debt are the most common forms of fraud for housing. However, other schemes can involve falsified documentation, including inflated or fabricated income and employment verification, synthetic identities and investment income misrepresentation through reverse occupancy schemes.

Proven fraud alert solutions are effective in catching defects that could indicate fraud or a compliance problem as the loan app moves through the workflow process. Likewise, new industry databases are also helping lenders and their tech partners to identify employment fraud.

Current fraud alert systems largely leverage the data on hand (natural intelligence) and work well in standard mortgage origination workflows. Larger originators, however, may need a more targeted workflow solution that helps them gain efficiency by lowering the volume of loans to be reviewed and cleared, so they can focus resources on the highest risk loans. This is where the use of both natural and artificial intelligence can change the landscape of fraud detection.

This next generation of fraud detection is just coming to market. Specifically, solutions that use predictive modeling and pattern recognition scoring to identify the level of fraud and early payment default risk in a specific application or across a portfolio of loans. Larger lenders should consider switching from their current alert-based system and leverage today’s predictive modeling and machine learning techniques that can simultaneously look at models and submodels that consider factors like synthetic identity, fabricated income, fictitious employment, early payment default, undisclosed debt and loan participant risk. First American, for example, has been able to reduce the average review rate of 60-70% to as little as 10% by using a targeted risk score.

With the combined use of both natural and artificial intelligence, lenders can benefit from this newfound operational efficiency, while mitigating EPD’s, buybacks, and lengthy and costly distressed asset care and maintenance. These new analytics will also have practical application in investor loan acquisition and servicing transfers.

Fraud detection is a never-ending journey to find the proverbial needle in the haystack. With the evolution of predictive analytics, machine learning and artificial intelligence, opportunities to get out ahead of the next wave of potential fraud risk is something lenders should explore to proactively protect themselves.