Unbiased Indices of Uneven Distribution and Exposure: New Alternatives for Segregation Analysis
Mark Fossett, Texas A&M University
Wenquan C. Zhang, Texas A&M University
This paper introduces new versions of widely used segregation indices that measure uneven distribution and exposure. The new versions have the desirable characteristic of being unbiased; under random distribution the new versions of indices of uneven distribution have expected values of zero and the new versions of indices of exposure have expected values that track group representation in the city population. In contrast, the expected values of conventional versions of indices of uneven distribution and exposure are subject to bias that can be substantial in some circumstances, sometimes rendering them untrustworthy altogether. To deal with this, segregation studies routinely restrict study designs. The new versions extend the feasible scope of segregation studies; they are easy to compute, substantively appealing, and easy to comprehend; and they replicate conventional indices when bias is tolerable and provide reliable results when conventional versions of indices break down.