Options for Measuring Residential Segregation across Small- and Intermediate-Scale Spatial Domains Using Restricted IPUMS Microdata

Mark Fossett, Texas A&M University

I review options for measuring residential segregation using small- and intermediate-scale spatial domains using the restricted IPUMS microdata from the 1940 US decennial census. I give special attention to the “eta” measure developed by Logan and Parman and contrast it with other available alternative measures. Logan and Parman’s eta does implement welcome innovations in segregation measurement; namely, they address two separate sources of bias in segregation measurement by focusing on households instead of persons and assessing segregation in terms of the racial composition of a household’s “neighbors” in an area instead of the “population” of the area in which they reside (which includes the household). Unfortunately, however, eta fails to meet important accepted criteria for segregation measures and consequently cannot be recommended as a general purpose segregation index. Happily, Fossett (2017) has introduced similar innovations in refined computing formulas that yield unbiased versions of more familiar and widely used segregation measures such as the gini index, the dissimilarity index, the Hutchens index, the Theil index, and the separation index, which fare better than eta on technical considerations. Significantly, the refined index formulations introduced by Fossett (2017) also can be applied using the same small spatial domains considered by Logan and Parman. I evaluate claims that eta should be viewed a viable candidate for wider use and recommend against this because eta’s limitations become more severe when it is applied in less specialized circumstances. My central finding is that alternative approaches to measuring segregation at small spatial scales – the use case considered by Logan and Parman – are available and provide superior options for segregation measurement. I review these options and compare results obtained when a wide range of traditional and refined versions of popular indices are applied to measure residential segregation at small and intermediate spatial scales.

See extended abstract

 Presented in Session 249. Urban