The continuing investigation into the U.S. housing bubble

As a general rule, the larger and more important an event, the longer it takes to sort out its causes. It’s been close to nine years since the bursting of the U.S. housing bubble and researchers are still sorting out the exact dynamics that lead to such a drastic increase in private debt. Obviously, a large amount of household debt was at the heart of the Great Recession of 2007-2009, either directly by affecting household consumption or more indirectly as debt instruments at the heart of the global financial system. But who borrowed this money? And what drove the increase in lending?

Economists Atif Mian of Princeton University and Amir Sufi of the University of Chicago are doing some of the best research on the causes of the housing bubble and the consequences on the broader economy once it popped. In a new working paper published by the National Bureau of Economic Research, Mian and Sufi build on their previous research, focusing on local area dynamics, to look specifically at individual borrowers using data from credit bureaus.

Their results support their earlier findings: debt grew quickest among borrowers with low credit scores. The ratio of debt to income grew rapidly for lower-score borrowers. The ratio grew by 80 percentage points from 2000 to 2006 for the bottom 60 percent of the credit-score distribution and doubled for those in the bottom 20 percent. And low-credit borrowers increased their debt the most in areas that had the strongest housing price growth.

Mian and Sufi interpret these results as supporting their hypothesis that rising house prices lead low-credit borrowers to borrow against the bubble. This behavior, according to the two economists, was the reason for a large part of the increase in household debt. Furthermore, borrowers from the bottom 40 percent of the credit score distribution were primarily responsible for the defaults from 2007 to 2010.

But what does about other research on the housing bubble say? Manuel Adelino of Duke University, Felipe Severino of Dartmouth College, and Antoinette Schoar of the Massachusetts Institute of Technology challenged Mian and Sufi’s earlier interpretation of the debt loads by ZIP code-level data. Adelino, Schoar, and Severino use individual level data on mortgage borrowing and incomes from mortgage applications and find that middle- and high-income borrowers were the main source of borrowing during the bubble and delinquencies during its bursting. In an earlier paper, Mian and Sufi pointed out these data are flawed due to fraud because of overstated incomes.

In this new paper, Mian and Sufi also show other reasons for the discrepancy. First, the mortgage data used by Adelino, Schoar, and Severino misses many borrowers with low credit scores, which makes borrowing appear more skewed upward in the credit distribution. Secondly, the three economists sort borrowers in their analysis by their 2006 credit scores, which makes many borrowers appear higher in the credit distribution as borrowing during the bubble pushed up credit scores. Mian and Sufi sort by 1997 credit scores to avoid the effects of the bubble.

But what’s the broader importance of the results Mian and Sufi present? As the authors note, several economists are working on modeling the dynamics of the housing bubble. There’s a disagreement about whether the increase in debt was the result of a loosening of a borrowing constraints, in which lenders let borrowers take out loans with less collateral, or of a lending constraint, in which lenders give out more loans. One way to test these competing hypotheses is to see if the debt-to-income ratio moves more than the ratio of debt to the value of housing.

Mian and Sufi find that debt-to-income ratios rose steeply for most households and mostly at the bottom of the credit-score distribution while debt-to-home value ratios actually declined for all groups above the bottom 20 percent of the distribution. This is a pretty good sign that lending constraints were loosened, perhaps by “advances” in lending technology such as securitization, which enabled mortgage lenders to sell their mortgages to institutional investors and then originate more mortgages to sell into the securitization machine.

So while the full history of the debt dynamics of the housing bubble isn’t finished, we are getting closer. The image is still fuzzy, but it’s increasingly coming into focus.