In the wake of the 2007–2009 financial crisis (and the accompanying recession), researchers examined many factors that precipitated the downturn. In "Owner-Occupancy Fraud and Mortgage Performance," Ronel Elul, Aaron Payne, and Sebastian Tilson explain the role of one particular factor: mortgage fraud, specifically, fraud where mortgage applicants identify themselves as owner-occupants despite being investors. Such mortgage fraud was widespread, their research shows, and it contributed to the housing boom and bust that ushered in the financial crisis.
They rely on a dataset1 that is a match between loan-level mortgage data and credit bureau data, making it possible to isolate borrowers who describe themselves as owner-occupants. By including information about where each borrower lived both before and after originating a mortgage, the data also show if each borrower moved to a new address after securing a mortgage (as one would expect if the borrower intended to live in the newly purchased property).2 In addition, the data also show if the borrower had a first mortgage, likely on another property the borrower actually lived in. The authors use these two features to identify fraudulent borrowers. The final sample consists of approximately 150,000 loans made to purchase homes, originated between January 2005 and December 2007.
The study's design clearly distinguishes between honest and dishonest mortgage applicants. What's more, the data capture the extent of fraud across a broad landscape of mortgage types (as well as the ways in which they are packaged and brought to market).3 Looking across all mortgages in the sample, the authors find that the share of fraudulent borrowers peaked at 5.2 percent in the first half of 2006, then returned to below 2 percent in 2008. Across the U.S., the areas with the highest fraud rates were California and Washington, D.C., where more than 13 percent of loans were fraudulent. The authors also find high fraud rates in states particularly affected by the housing bubble, such as Nevada and Florida.
Why did these borrowers engage in fraud? The authors show that one important motivation was to obtain a lower interest rate. (Honest homeowners tend to qualify for lower rates. By posing as honest homeowners, dishonest applicants were borrowing at lower rates than they would normally qualify for had they divulged that they were investors.) This was particularly the case for riskier loans, such as those with smaller down payments or to borrowers with lower credit scores. These riskier loans are normally unavailable to investors.
Dishonest borrowers were notably riskier than those borrowers who declared their status as investors. First, they were more likely to pursue mortgages that were big relative to the market value of the property they were buying. (In real-estate parlance, these mortgages had questionably high loan-to-value ratios.)
Second, they were more likely to have low credit scores than were investors. (To generate a credit score, a model uses credit bureau data on patterns of borrowing and repayment to predict the likelihood that a borrower will repay a loan: The lower the score, the higher the predicted risk of default.)
Third, they were substantially more likely to have low-documentation loans, interest-only mortgages, adjustable-rate mortgages, and brokered mortgages. (These types of loans are riskier than conventional mortgages.)
And fourth, they pursued big mortgages — so-called jumbo loans with values greater than $417,000 — more often than did investors; this meant that when their loans eventually defaulted, losses were larger.
Having shown that fraudulent borrowers were riskier at the time of loan origination, the authors then demonstrate that many of these borrowers failed to meet their mortgage obligations. As of December 2008, 25 percent of fraudulent borrowers in the sample group were either delinquent on their loans or in outright default, whereas less than 10 percent of investors found their loans in similar distress. Furthermore, despite making up less than 5 percent of total borrowers, dishonest borrowers accounted for 15.9 percent of the dollar share and 11.7 percent of the number of delinquent or defaulted loans that originated during the period.
To pinpoint the degree to which identity fraud contributed to default risk, the economists use a model that accounts for a variety of factors known to affect the likelihood of default, including the estimated change in loan-to-value ratios (measured from the loan origination date through December 2008), credit scores, changes in the unemployment rate, whether the loan had low documentation, and whether the borrower had multiple first liens. The model shows each factor's percentage contribution to default risk. (For example, the presence of multiple liens accounts for about 40 percent of the additional risk borne by fraudulent loans.)
Elul, Payne, and Tilson conclude not only that dishonest borrowers are more likely to default than are investors, but also that they are more likely to default "strategically." That is, these borrowers' default decisions were driven more by declines in the value of the property than by factors like unemployment. The researchers arrive at this insight by looking at how the relationship between default rates and loan-to-value ratios varies by borrower type: They find that fraudulent borrowers are more likely to default when loan-to-value ratios rise. Defaulting dishonest borrowers are also more likely to preserve their bankcard credit. By contrast, defaulting honest homeowners and investors are more likely to draw down their credit limits, which suggests that either they are trying to keep their home or they have weathered severe shocks before defaulting.
The issuance of mortgages to fraudulent borrowers aggravated the economic downturn of 2007–2009. The wave of mortgages issued to dishonest borrowers led to widespread dislocations in the pool of outstanding mortgage debt. As Elul, Payne, and Tilson demonstrate, the level of risk embedded in these loans was higher than widely believed, largely because dishonest borrowers obtained mortgages they did not qualify for.
1 The raw data are maintained in a repository known as Equifax Credit Risks Insight Servicing and Black Knight McDash Data.
2 To protect privacy, the data provider scrambled the addresses.
3 This includes mortgages guaranteed by government-sponsored enterprises (GSEs), held in bank portfolios, and folded into packages of mortgage-backed securities. The authors show evidence of occupancy fraud across all types, notably in the large market for loans guaranteed by GSEs, which have not been examined before.