Measuring Segregation and Outgroup Exposure: The k-Nearest-Neighbors Metric

• In Event: Geography and Spatial Analysis

Abstract

In this paper, I develop two new metrics – one of individual outgroup exposure, and one of area-level aggregate segregation – that are based on the geocoded locations of individuals and thereby bypass the problems of scale and ecological inference that limit the potential of existing exposure and segregation measures. I use an original network census to confirm that my exposure metric, the k- nearest-neighbors score, accurately reflects social contact, which affects intergroup relations. Moreover, I demonstrate that my exposure metric is a better proxy for social contact than two current measures. Next, I demonstrate that my segregation metric, the mean of the k-nearest-neighbors score, is an accurate measure of aggregate residential segregation. I use a simulation approach to demonstrate that the proposed aggregate segregation metric can more accurately discriminate between segregated and integrated areas than two commonly-used existing metrics that rely on dividing the region into areal units. Moreover, in contrast to existing segregation measures, I show that the standard deviation of the k-nearest-neighbors score is able to distinguish between different patterns of segregation, i.e. whether the two groups are divided at a single border, or are divided into multiple interlocking clusters. These measures could be of broad applicability to scholars studying ethnic politics, especially in the Global South.

Author

• Jeremy Spater, Princeton University