Hispanics delinquent on mortgages 4.7 times as often as whites

Going on a half dozen years after the mortgage meltdown that began in 2007, the evidence continues to trickle in about the key role of diversity in the disaster. Granted, there's very little demand for hard-headed analyses. Here, for example, is a paper finished in 2011 that has, according to Google, been cited once:

Mortgage Default by 2009: Effects of Race, Ethnicity and Economic Standing During the Boom Years  

Heather Luea
Vanderbilt University  

Adam Reichenberger
Bureau of Labor Statistics  

Tracy Turner
Kansas State University 

Abstract: This paper examines the determinants of 2009 mortgage delinquency by race and ethnicity using new household-level data on mortgage distress from the Panel Study of Income Dynamics. Controlling for homeowner and loan characteristics as well as residence in a nonrecourse state, we find startling differences in mortgage delinquency rates that cannot be explained by observables. The unexplained black/white gap corresponds to a 44% higher likelihood that black homeowners will be delinquent on their mortgages relative to non-Hispanic white homeowners. The unexplained difference in Hispanic mortgage delinquency relative to non-Hispanic white homeowners is even greater, at double the black/white delinquency gap.

... The economic decline that began in 2007 was preceded by nearly two decades of government-aided, rapidly rising homeownership rates among minority households (Bostic and Lee, 2007). Given this and the severity of the recent economic crisis, it is important to understand the extent to which minority households have weathered the crisis as well as non-Hispanic white households, all else equal. Indeed, the recent and historical role played by the US government and nonprofit agencies in boosting access to homeownership by underrepresented groups makes understanding these groups’ outcomes particularly relevant.4 

Footnote 4: As recently as June 2002, President Bush announced a goal of closing the homeownership gap for minority households by 5.5 million households by the end of 2010 through innovatiosn such as zero-down-payment loans. That administration's efforts followed more than a decade of housing market interventions, including President Clinton’s National Homeownership Strategy, a trillion dollar commitment by Fannie Mae, the Campaign for Homeownership of the Neighborhood Reinvestment Corporation, and expanded lending to low-income and minority households in part as a result of the implications of the Community Reinvestment Act (Turner and Smith, 2009). 

... As a preview of our findings, we find that black and Hispanic households that own their housing in 2005 are significantly more likely to become delinquent on their home loans by 2009 than non-Hispanic white homeowners. We find an unconditional, weighted likelihood of delinquency of 11.3%, 16% and 3.4% for black, Hispanic, and non-Hispanic white homeowners, respectively, making black homeowners 7.9 and Hispanic homeowners 12.6 percentage points more likely to be delinquent than non-Hispanic white homeowners. 

Let's break those delinquency-by-2009 rates out:

Whites: 3.4%
Blacks: 11.3% (3.3X the white rate)
Hispanics: 16.0% (4.7X the white rate)

The sample sizes of 2005 homeowers in the PSID are not huge: 2344 for whites, 810 for blacks and 263 for Hispanics (the number of Hispanics in a long-running longitudinal study naturally lags behind their number in the population). Also this study design excludes the 2006-07 vintage of new mortgages, which were the bottom of the barrel. However:

While the sample size of the PSID may be considered small compared to loan-based samples (for example, in the 2009 survey, there are roughly 8,000 households), the PSID has a number of advantages over larger samples that are either not household-based or not longitudinal. First, importantly, the unit of observation is the household, and the PSID collects extensive household-level data on employment, income, wealth, and housing costs and characteristics. In 2009, for the first time in the history of the PSID, survey respondents were also asked questions regarding mortgage delinquency and foreclosure, making this a dataset well suited for our study. Second, the PSID is a longitudinal dataset following families from as early as 1968 to present. Using the PSID, we have borrower and loan characteristics overtime, which to our knowledge are data not available in any other single dataset. Third, once sample weights are applied, the PSID is a nationally representative sample of the US population.

Statistically adjusting for all the info in the PSID, it turns out that there are still substantial racial gaps in staying current on mortgages:

Conditioning on extensive borrower and particularly loan characteristics reduces the race and ethnicity gaps in mortgage sustainability considerably, but does not entirely eliminate these gaps. In the full specification, we find that black and Hispanic homeowners remain 1.5 and 3 percentage points more likely to be delinquent than non Hispanic white homeowners, respectively. These unexplained effects are  large relative to the underlying mortgage delinquency rate of 3.4% for non-Hispanic white households.

In other words, statistically adjusting for everything they can come up with (e.g., income), there are still unexplained racial gaps:

Whites: 3.4%
Blacks: 4.9% (1.44X the white rate even after adjustment)
Hispanics: 6.4% (1.88X the white rate after adjustment)

In many ways, the first set of numbers is the more important. As the population shifts from whites to Hispanics, the delinquency rate would tend to get worse. 

But the second table can help explain why money-hungry but politically true-believing lenders like Angelo Mozilo of Countrywide could mess up so badly. You are not allowed to use race/ethnicity in credit modeling, but it turns out that race/ethnicity still matters a lot even in cases where the facts you are allowed to look at are all the same. During the 1990s, Mozilo became convinced that it was sheer racism to worry that Hispanics could default at higher rates than the model predicts.

We find startling differences by race and ethnicity in mortgage delinquency rates that cannot be fully explained by observables ...

The homeownership rates of black and Hispanic households have been and remain substantially below that of non-Hispanic white households. That certain groups experience low homeownership rates is cause for concern particularly to the extent that these gaps are involuntary and in light of the possibility that homeownership generates private and community benefits (i.e., Turner and Luea, 2009; Haurin, Dietz, and Weinberg, 2002). Belief in the positive externalities of homeownership has motivated substantial efforts in the past two decades to boost the homeownership attainment of underrepresented groups, and these efforts have generated relative gains in minority homeownership (Bostic and Lee, 2007). Evidence is mounting that the Great Recession has adversely impacted minorities to a greater extent than non-Hispanic white households. It is likely that the economically disadvantaged households that are losing their homes are some of the same households propelled into homeownership through federal assistance to begin with. If there is a silver lining, it may be that, according to recent work by Molloy and Shan (2011), post-foreclosure households on average do not end up in either less desirable neighborhoods or more crowded living conditions than what they experienced as homeowners. Determining the extent to which housing policy may have fueled the 2009 differential delinquency rates by minority status and why, and whether these households are nonetheless better off for their homeownership stint, would be valuable information for future policy design.

This might also be valuable information for current immigration policy design.