PHILADELPHIA Voice mail has become a way for some housing rental agents to discriminate on the basis of race, class and gender.

A just released University of Pennsylvania study offers strong new evidence for the existence of telephone-based discrimination in the rental housing market and concludes that some rental agents discriminate by using linguistic cues to screen callers over the phone.

The study, conducted in Philadelphia and its suburbs by Penn sociologist Douglas Massey, postdoctoral fellow Garvey Lundy and undergraduate students, concluded that:

• African-Americans generally experience less access than whites to rental housing.
• Women experience less access than men.
• Lower-class African-Americans have less access to rental housing than middle-class African-Americans, with poor African-American women experiencing the greatest discrimination.
• Race, gender and class interact to influence rental outcomes.

The study was conducted as part of a Penn undergraduate sociology course in research design.

The class members theorized that rental agents attach different racial and class labels to certain styles of speech. Specifically, when an Africanmerican speaks standard English with a black pronunciation of certain words (black-accented English), rental agents infer that the speaker is African-American of middle-class origins; however, the combination of nonstandard grammar with a black accent (black English vernacular) signals lower-class origins.

To test this theory, the students developed a script to use in calls about rental housing. The multiracial class of males and females modified the script to be read by African-American students in black English vernacular or black-accented English and by white students in white middle-class English.

In nearly half of all cases, callers initially reached some sort of voice mail. When this happened, students, reading from a script, left messages requesting return calls. They left messages at least three times before giving up. If and when a student spoke to a rental agent, he or she read a scripted conversation to gather basic information about the rental terms. The Penn student researchers made 474 calls to inquire about 79 rental units advertised in Philadelphia newspapers or rental guides during March and April 1999.

"The study shows that housing discrimination frequently occurs over the phone before rental agents and African-American clients ever meet," Massey said. "To discriminate, all a landlord needs to say when he or she hears a lack-soundingvoice on the other end of the line is that the unit is already rented. Voice mail and phone tag make it even easier to simply not return messages left by speakers of black English vernacular or black-accented English, thus letting a machine do the racial screening."

The Penn research shows that compared with whites, African-Americans are less likely to get through and speak to a rental agent, less likely to be told of a unit availability, more likely to be charged an application fee and more likely to have credit worthiness mentioned as a potential problem in qualifying for a lease.

And, by far, low-income African-American women are the most disadvantaged. On average, they are assessed $32 more per application than white middle-class men. Moreover, for every call a white man makes to inquire about a rental unit in the Philadelphia area, a low-income African-American woman has to make two calls, roughly doubling her time and effort compared with his.

Beyond simple access, other potential barriers to housing were considered, including instances when a caller credit history was raised as a potential problem. In the Penn study, credit was spontaneously mentioned to only 3 percent of white middle-class men but was brought up as a potential issue for 5 percent of white middle-class women, 10 percent of African-American males regardless of class, 21 percent of middle-class African-American women and 23 percent of lower-class African-American women. Compared with white middle-class men, African-American women were about 16 times more likely to have credit mentioned as an issue, regardless of class.

The research was supported by a Pew Memorial Trust grant to the University of Pennsylvania School of Arts and Sciences. The findings were published in the March 4 issue of Urban Affairs Review.