Researchers recently set out to find some answers by sending out thousands of fake resumes for jobs around the country, and they turned up a surprising, and little-noticed, answer.

In a nutshell: it’s not that recruiters themselves necessarily have a racial bias; instead, they fear some of their customers do.

Last year, four researchers—John Nunley, economics professor at the University of Wisconsin-La Crosse, Adam Pugh at CUNA Mutual Group in Madison, Wisconsin, Nicholas Romero, an economics professor at the University of Pennsylvania, and Richard Seals, an economics professor at Auburn University in Alabama—conducted an experiment to understand the job market for recent college graduates in the wake of the Great Recession.

They submitted 9,400 fake resumes of nonexistent recent college graduates through online job applications for positions based in Atlanta, Baltimore, Portland, Oregon, Los Angeles, Boston, and Minneapolis. They sent four resumes per job and randomly changed certain variables in the applications, such as college major, work experience, gender, and race.

To signal ethnicity, the researchers gave half the candidates the “typically white” names of Cody Baker, Jake Kelly, Claire Kruger, and Amy Rasmussen. They gave the other half “typically black” names of DeShawn Jefferson, DeAndre Washington, Ebony Booker, and Aaliyah Jackson. (The choice of these names was based on statistical data about name popularity among different ethnic groups.)

No such experiment is perfect, but this is about as good as you can get. It allowed researchers to look at the importance of each variable in isolation.

The results? Young African-Americans still face persistent discrimination in the job market, and it is not tied to socioeconomic status, a lack of a degree, or other factors. Overall, black applicants were invited in for interviews 15.2% of the time, while white applicants received invitations 18% of the time. To put it another way, African-Americans were 16% less likely to get called in for an interview.

The discrimination gap existed in every city. (The best, incidentally, was Baltimore, where it was slight. Second best was Portland.) It existed among those with business-related degrees and those without, for men and women, those from middle-class homes, and so on.

So far, so depressing.

But there’s a twist.

Black applicants faced major discrimination when applying for jobs with a customer focus. Researchers looked for jobs with words like “customer,” “sales,” “advisor,” “representative,” “agent,” and “loan officer” in the description. For jobs such as these, the discrimination gap soared. Instead of facing a 2.8 percentage-point gap between callback rates for whites and blacks, they faced a 4.4-point gap.

For jobs with descriptions that lacked those terms and were instead focused on interaction with coworkers, the level of discrimination collapsed. Descriptions with terms such as “manager,” “administrator,” “coordinator,” “operations,” and so forth, the difference in callback rates was 0.1 to 0.3 percentage points.

In other words, the problem isn’t that Joe Smith doesn’t want to hire young African-Americans, but that he is worried that if he hires a black sales associate, old Mrs. Jones may take her business elsewhere.

This is a small part of a much larger challenge. Some black Americans face high rates of unemployment due to the population group’s disproportionately high levels of poverty and subsequent lack of educational opportunity. But the research findings do seem to hold out some hope for progress.

Meanwhile, if you are a young, black college graduate and struggling to find work, the research does offer some pointers on navigating a job market tarnished with unfairness. The data shows that those with the best chance of getting a callback will have invested in their own education, completed an internship while at college, made the most constructive use of their time after graduation, and have applied, at least initially, for jobs with an internal focus. In Baltimore… or Portland, Oregon.