Developing spatial measures of residential segregation using kernel density estimation

In the USA residential segregation has been shown to reinforce racial inequality and promote labour market segregation through income polarisation and differential rates of unemployment. However the available population data can compromise the measurement of residential segregation due to the MAUP and the checkerboard problem. Also debate continues over the desirable criteria of residential segregation indexes. This paper has two objectives: i) to investigate the use of kernel density estimation as a means of spatially smoothing population data prior to the measurement of residential segregation ; and ii) to explore the merits of different criteria underpinning spatial measures of residential segregation, given the arguments of Reardon and O'Sullivan and other contemporary literature. These objectives will be illustrated by the analysis of residential segregation in the Sydney Commuting Area in 2001, using different parameter values for the underlying kernel density estimation.