Racial Disparities in Home Lending
Blacks and Hispanics pay more for home loans than whites, according to a new study. Unfortunately, the study did not look at the most obvious variables that would have made it useful.
Disparities Found in Sub-Prime Lending (WaPo, p. A2)
About 29 percent of African Americans who bought or refinanced homes last year ended up with high-cost loans, compared with only about 10 percent of white Americans, according to an a consumer advocacy group’s analysis of new data from 15 large national lenders. About 15 percent of Hispanics received such loans offered to borrowers with damaged credit, known as sub-prime loans, according to the analysis conducted by the National Community Reinvestment Coalition, which represents 640 nonprofit organizations. “The good news is that access to credit is more available in the home mortgage market, but the bad news is that it is being offered in a disproportionately unfair way,” said John Taylor, the coalition’s president and chief executive.
Lender groups do not dispute that there are disparities based on race and sex, but say borrowers pay higher rates if they pose a higher credit risk for any one of a variety of reasons, such as low income, poor neighborhood, bad credit history or low down payment.
The coalition says its analysis corroborates its long-held belief that lenders discriminate. Lenders say the new information does not provide a complete picture of the loan-making process. They say borrowers who have poor credit, bad records for paying bills on time, or live in troubled neighborhoods with high foreclosure rates and falling home values are riskier customers than people with good credit who live in neighborhoods where real estate values are rising. Lenders have to charge more money to compensate for the higher risk, they say. Such risk is generally measured by credit scores. In the past, lenders have opposed efforts to add credit scores to the federally required data, saying the obligation would pose an additional regulatory burden and potentially invade the privacy of borrowers.
That blacks and Hispanics pay higher rates on the aggregrate than whites is hardly surprising. No one disputes these groups are poorer, less well educated, have higher unemployment rates, are more likely to have credit problems, and so forth. The question is the causes of these disparities, not their existence.
The report [PDF] is spectacularly unhelpful in shedding light on this, though, since it ignores almost all useful variables. Most of it consists of simple correlational analysis without any cross-tabs. No information on the education, income, employment history, or other economic data save interest rate paid is given. Even the regression analysis is woeful, focusing only on neighborhood:
NCRC conducted regression analysis to determine if the portion of prime and subprime lending increased as segregation increased while controlling for housing affordability. We used a segregation index calculated by the University of Michigan and a homeownership affordability index developed by the National Association of Homebuilders.
NCRC developed a new variable, a ratio, in order to perform the regression analysis. This ratio is the percentage of all subprime loans made to a group of borrowers divided by the percentage of all prime loans made to that same group of borrowers. If the percentage of all subprime loans that are going to a borrower group is greater than the percentage of all prime loans going to the same borrower group then the ratio will be greater than 1. This ratio will be referred to as the fair lending performance ratio in the sense that it measures the performance of lenders at reaching certain types of borrowers. The higher the value of the ratio, the more aggressive subprime lenders are relative to prime lenders in offering higher portions of their loans to groups of underserved borrowers.
Regression analysis was performed for all six borrower categories: African-American borrowers, Hispanic borrowers, LMI borrowers, borrowers in LMI tracts, borrowers in minority tracts, and women borrowers. The regression analysis was used to determine the effects of different variables on the ratio for each of the six borrower categories.
NCRC was concerned with measuring the effects on the performance ratio for the following different variables: White/Black Segregation Index, White/Hispanic Segregation Index, Affordability Index, the percent of African-Americans in an MSA, the percent of Hispanics in an MSA, and the Median Family Income in an MSA.
So, basically, race/ethnicity was a variable that didn’t vary. It’s not particularly interesting that blacks or Hispanics that live in incredibly segregated neighborhoods pay higher lone rates. We all knew that. The questions is why: Because these people are a higher credit risk or because the banks discriminate on the basis of race even when controlling for other factors. If one doesn’t actually control for the other factors, then one doesn’t know.
Kevin Drum agrees up to a point but finds something insidious in the banks’ opposition to releasing information on credit scores:
Providing aggregate credit scores, or even individual credit scores unattached to names, obviously invades no one’s privacy. The fact that lenders are so eager to keep them secret tells you pretty much all you need to know here.
Actually, it doesn’t. Maybe they are hiding this data because they’re racist bastards trying to keep the Little Man down. Or maybe they are just averse to spending a lot of time (and thus, money) compiling data for others’ use.