A Comparison of Aspatial and Spatial Measures of Segregation

Abstract The modifiable areal unit problem arises when the boundaries that define neighborhoods affect perceived levels of segregation. Scholars postulate that this problem is exacerbated when one uses a definition of neighborhoods that is based on administrative units; doing so leads to an aspatial measure of segregation, which may or may not adequately account for the spatial relationships among residential locations. In this article, we assess whether aspatial and spatial definitions of neighborhoods produce different perceived levels of income segregation. Using an original individual‐level dataset on income in San Mateo County, California, we define each individual's neighborhood in three ways – two aspatial and one spatial. On the basis of these definitions of neighborhoods, we then estimate residential income segregation using the local Moran's I statistic. We report two primary findings. First, the three measures generate different perceived levels of income segregation. Specifically, we observe less income segregation when using the aspatial measures as compared with the spatial one. Second, the inconsistencies between these measures are systematic in such a way as to lead to different inferences when used to predict individual voter turnout.