Many argue that government policies aimed at increasing first-time homebuyers caused the housing crisis. This narrative persists despite considerable evidence to the contrary. Our Housing Credit Availability Index suggests the presence of risky products, not increased lending to riskier borrowers, was a significant contributor to the crisis.

New data from the government-sponsored enterprises (GSEs) now suggest a further culprit: mortgage refinance activity. At the height of the boom, mortgage refinances (refis) were more likely to default than mortgages taken out to purchase a home, mostly because many people were treating their homes as ATMs through cash-out refinances. Eighty-four percent of GSE refinances in 2006 and 2007 were cash-out refinances. These refinanced loans suffered from sloppier underwriting, so for any set of observable risk characteristics, these refinance loans defaulted more often than purchase loans.

During the housing boom, refis defaulted more often than purchase loans

Our data include 30-year GSE mortgages that were fully documented and fully amortizing and originated between 1999 and 2015. Fifty-six percent of these mortgages were refis, and 44 percent were purchase loans. These data do not include any nontraditional products, such as interest-only, 40-year, or negatively amortization loans. These nontraditional products, which were more heavily refis than our data, performed worse than comparable loans in our database.

Between 1999 and 2015, 4.48 percent of refis and 3.3 percent of purchase loans went more than 180 days delinquent. In the peak year of 2007, 15.78 percent of refis and 9.95 percent of purchase loans went more than 180 days delinquent.

Since 2009, delinquency rates have been low for all loans. The largest differences in performance between purchase and refi loans were loans originated in the boom period of 2004–08.

Conventional wisdom suggests that refis should be less risky than purchase loans and default less because the borrowers have a known history of payment. So these results are surprising, especially given the stronger credit characteristics of refis, such as lower loan-to-value (LTV) and debt-to-income ratios.

From 1999 to 2015, refis had an LTV ratio of 68.4, compared with 79.2 for purchase loans. Purchase borrowers also had marginally higher debt-to-income ratios—34.9 versus 33.4. These higher risks were partially offset, however, by the marginally higher FICO scores of purchase borrowers—730 versus 736.

Lenders lose more when refis default than when purchase loans default

To understand whether lenders lose as much when refi and purchase loans default, we need to estimate the probability that loans that go 180 days delinquent will liquidate, and then calculate the severity if they do liquidate.

We find that 25 to 30 percent of delinquent refis become current or prepay, 60 percent liquidate, and 10 to 15 percent remain persistently delinquent (neither current nor liquidated). Purchase loans are marginally less likely to become current and are more likely to be liquidated than their refinance counterparts.

Purchase loan severity is about 10 points lower than refi severity. Lenders lose more money with a delinquent refi than with a delinquent purchase loan.

This is partly because many purchase loans have mortgage insurance, which lowers losses. But looking separately at loans with and without insurance, purchase loans consistently have lower severities than refi loans.

The figure below shows that the loss rates for refinance mortgages are higher than those for purchase mortgages, with the largest differences between 2005 and 2008.

The bottom line

These new GSE data reveal that for GSE 30-year, full documentation loans, defaults were more common and the losses were higher on refinance mortgages than on purchase mortgages. Although there is a lot of blame to go around for the poor quality of loans before the housing crisis, these data reaffirm that purchase borrowers were not the primary culprits.