In the USA Today article "Fed: Minorities pay more for housing loans" (9/14/5), Sue Kirchhoff reports that "Minorities are far more likely than whites to take out higher-priced loans to buy or refinance a home and are denied loans more often." This conclusion is drawn from a Federal Reserve review of 2004 Home Mortgage Disclosure Act (HMDA) information. The differences presented are stark. The incidence of higher-priced mortgages among African-Americans was 32.4%, 20.3% among Hispanics, and only 8.7% for non-Hispanic whites. The Fed also stated that these differences could be largely explained by differences in income. We shall examine the information presented in this article using the eight criteria presented in Mathematics: A Quantitative Reasoning Approach.

The goal of the study was to determine bias in lending towards minorities in mortgage loans. The population being studied is the set of all mortgage applicants - and the lenders with whom they worked. This was an observational study, reviewing data already collected by government requirement. This was appropriate. A controlled experiment could attempt to recreate the collected data, but would introduce additional complications.

The Federal Reserve is a nonpartisan economic part of the government. This type of data is fundamental to the work it does and regulates. It, however, does not lend to individuals, so it is not part of the population being studied.

The sample set being studied is identical to the population being studied. Therefore, there can be no selection bias.

The variables of interest are reasonably well defined. The prime rate is the interest rate charged to a lender's most credit-worthy customers. Subprime rates are 0.1% to 0.6% higher than the prime rate, and are offered to higher-risk customers. The interest rate given to mortgage customers is a matter of record, and easily collected - or at least the junk mail I receive would indicate so.

There is one acknowledged confounding variable: credit scores. The credit scores are not part of the data collected by HMDA, and so were unavailable to the researchers. This may appear to be a major flaw, but it is not necessarily so. Many of the items that are factored into credit scores were available. Income, location and loan amount are specifically mentioned; all but the last directly affect one's credit score. Adjusting for these factors approximately halved the percentage of minorities offered sub prime loans, accounting for the Fed's claim that factors such as income account for the bulk of the difference.

There were no survey questions used, so the setting and wording is not applicable.

The results presented are factual. Caveats, disclaimers, and the like pepper the article, quickly watering down the initial and somewhat sensational headline, despite its apparent accuracy.

The conclusions presented are all encompassed within the first paragraph, and are not expanded on - though other information from the report indicated that approximately 2% of lenders (though not necessarily 2% of loans) showed a statistically significant difference in lending to minorities. The conclusions presented are very mild.

This study appears to be valid, and meaningful, though the researchers only hint at the cause behind the correlation uncovered.

Debt-to-income ratio is a key financial indicator of an individual's financial health. A thought experiment using average statistics for male income, segregated by race, along with average American credit card debt gives some insight into the situation.

We find that the average credit card debt in the United States is $8,562. In 2003, the average salary for white males was approximately $41K, Black males made about $32K, and Hispanic males made approximately $26K. This alone gives respective debt ratios of 0.21, 0.26, and 0.33, which could alone account for the results of the Fed's study, and strongly suggests that even a small racial wage disparity further harms those who can least afford higher interest rates.

This thought experiment simply shows that such an explanation is possible. A further course of study would ideally include more detailed information, such as credit scores at the time of the loan. That additional information could provide a stronger causal link than the back-of-the-envelope estimates above. Further analysis of the debt burden of lower-income families may also help demonstrate the cause of the noted disparity.