Back in September
Page, Marianne. 1993. "Racial and Ethnic Discrimination in Urban Housing Markets: Evidence from a Recent Audit Study." PSC Research Report No. 93-290. September 1993.
Over the past twenty years, housing audits have developed into a useful resource for measuring discrimination. During a housing audit, pairs of individuals who differ by race or ethnic background assume identical family and economic characteristics and are trained to exhibit the same behavior when interacting with a housing agent. Each member of the pair visits the same landlord or real estate agent in succession and his or her treatment. This paper makes use data from a series of black/white and Hispanic/Anglo audits conducted by the Urban Institute and Syracuse University during the spring and summer of 1989 and sponsored by the U.S. Department of Housing and Urban Development. The study improves on existing analyses of audit data, in which the dependent variable is a count, by developing a statistical model that explicitly accounts for the variables's discrete, nonnegative nature and its heteroskedasticity.
Results show that blacks are provided with access to about 80% of the units made available to whites, and that Hispanics are shown about 90% of the units shown to Anglos. Hispanics seem to encounter less discrimination than blacks when they search for housing, but their experience is more difficult to explain. While customer prejudice and racial stereotyping significantly influence agent behavior towards blacks looking to purchase homes, the variables used to measure these phenomena have little or no significance in the Hispanic/Anglo audits. The results imply that theories describing white behavior towards blacks may not be relevant for Hispanics in the housing market, and this underlines the need for research on the causes of discrimination that is specific to particular race and ethnic groups. Finally, the paper shows that in the sales market, discimination levels vary significantly across cities, for reasons that remain unclear.