Measuring Segregation Patterns and Change: a Co-Location Quotient Approach

There are many segregation measures introduced and utilized in geographic research up to this date. Because residential segregation can be defined in more than one way the measure’s formulation is dependent on the particular definition the researcher is trying to reflect. Another distinctive feature of the quantitative exploration of segregation is the role of geographic scale. In contrast, global indices focus on overall level of spatial separation of population in the urban area while local indices assume that the index magnitude varies from place to place across the city. The main purpose of this study is to introduce a new measure of segregation that focuses on the lack of interactions of the population groups and to explore its properties. The proposed measure is a modified co-location quotient (CLQ) that was originally applied to point data as a measure of spatial association between two categorical variables. The first part of this dissertation introduces two versions of modified CLQ that are applicable to categories of areally aggregated population. One is the global measure that captures the overall exposure of one population group given the presence of another group. The local version of the measure describes levels of exposure for every single spatial unit. Both, global and local quotients have two basic specifications – two-group CLQ and same-group CLQ. Each variant of the measure allows the option to include the neighborhood size in computation, which theoretically defines the space within which people have the possibility for interaction. The use of CLQ in the proposed mathematical configuration expands the discussion of dimensions of segregation by suggesting the connection between different dimensions that are covered by co-location measure. Using publicly available data from U.S. Census Bureau on racial composition of population CLQs were computed for thirty urban areas, where twenty nine are metro areas and one is Washington D.C. The basic units of analysis are census tracts and block groups that contain aggregated population counts. Three decennial releases are used: 1990, 2000 and 2010. The results suggest an overall, but uneven, increase in the exposure of white people in given urban areas. Patterns of concentration for white people remained stable over the time span. But the concentration of black people shows a substantial decrease indicating an increasing exposure of blacks in the global sense. Conversely, same-group CLQs for whites and for blacks indicate unequal experiences for these two population groups in America. Additionally, various visualization techniques related to co-location measure were explored. The pointillist approach, suggested in this study, is found to be particularly effective technique for displaying CLQ results compared to widely utilized choropleth mapping