Earlier this year, a new working paper cast doubt on one of the dominant explanations of the reasons for the 2002-2006 housing bubble in the United States—that growth in mortgage credit and income growth uncoupled as credit flowed to areas to with declining income growth. Instead, economists Manuel Adelino of Duke University, Antoinette Schoar of the Massachusetts Institute of Technology, and Felipe Severino of Dartmouth College, argue that the cause of the increase on household debt was a classic speculative mania.
But a new paper by economists Atif Mian of Princeton University and Amir Sufi of the University of Chicago questions this view of the debt build-up. The seeming flaws in the dominant narrative that an increase in the supply of credit caused the bubble, they say, can be explained by one thing: mortgage fraud.
Mian and Sufi, of course, are not bystanders in the debate over the origins of the housing bubble. Their research pinning the increase in household debt on increased credit growth—the dominant explanation for the housing bubble—has been widely cited and led to their well-regarded book, House of Debt. And the paper by Adelino, Schoar, and Severino was a direct response to a 2009 paper by Mian and Sufi that underpinned their book.
The crux of this debate boils down to this—the relationship between income growth and credit growth at the zip code level is different depending on the source of data on incomes. If the income data comes from the U.S. Internal Revenue Services then the correlation is negative. If the data source is self-reported income from mortgage applications, then the correlation is positive. In other words, the two sets of economists agree that in zip codes where credit flowed, the incomes of homebuyers, measured by incomes on mortgages, were higher than the average income of the zip code, but they disagree on the reasons.
Adelino, Schoar and Severino’s interpretation of this difference is that higher-income individuals were buying houses in low-income zip codes. In essence, these buyers were speculators jumping into the housing market hoping to make money by “flipping” homes.
The new paper by Mian and Sufi takes issue with this interpretation, finding the correlation negative by looking at actual incomes as reported to the IRS. They find the difference between homebuyer income and the zip code income is due to the incomes of borrowers being overstated on mortgages. In other words, the culprit is mortgage fraud. They say that higher-income households weren’t buying houses into the low-credit score zip codes where credit was flowing. In fact, the credit scores of households moving into these zip codes declined over the period. Furthermore, they find income overstatement is higher in low-credit score, lower-income zip codes.
And then, Mian and Sufi show that zip codes with higher overstatement of income on mortgages saw larger increases in the share of mortgages that were not originated in accordance with the so called “conforming loan” standards required by the two giant mortgage insurance giants Fannie Mae and Freddie Mac. These non-conforming mortgages were so-called private-label mortgages, many of which were fraudulent. According to one study, 30 percent of these loans contained some incorrect data. Another shows that private-label mortgages with low documentation made incomes look larger on average by about 29 percent compared to mortgages that conformed to the more strict underwriting standards of Fannie Mae and Freddie Mac.
Mian and Sufi then look at how the overstatement of incomes on mortgage documents varied across zip codes depending on what share of mortgages were originated for Freddie Mac and Fannie Mae, also known as government-sponsored enterprises, or GSEs. What they find is that overstatement is higher in zip codes with fewer mortgages conforming to government standards. As the authors put it, “The decoupling is concentrated when and where fraud was most likely: high non-GSE share zip codes during the mortgage credit boom.”
Adelino, Schoar and Severino do run a test to see if fraud is driving their results and believe the data pass it. But Mian and Sufi argue that this test doesn’t past muster as it doesn’t consider the differences in GSE and non-GSE zip codes.
The evidence complied by Mian and Sufi points strongly to the findings of their original research, showing that income growth and mortgage credit growth did decouple during the housing bubble. Data that shows income growth for borrowers is likely the result of fraud and income overstatement. The two economists suggest that the increase in credit supply itself drove the fraud as lenders looked to make as many loans as possible.
So, perhaps the rumors of death of the disconnect between credit and income growth have been greatly exaggerated.