Resumen de Multigroup and Multilevel Residential Segregation:: The U.S. case, 1989-2005
Ricardo Mora Villarrubia, Javier Ruiz-Castillo Ucelay
This paper presents a multigroup and multilevel approach to the study of residential segregation for the public school population in U.S. urban areas in 1989 and 2005. The multigroup approach includes Native Americans, Hispanics, Asians, blacks and whites. The multilevel approach distinguishes neighborhoods, cities, and regions, leading to notions of within-cities, between-cities, and between-regions residential segregation. Measures of these notions are computed using the Mutual Information, or M segregation index. The decomposability properties of the M index are exploited for two purposes: first, to identify the smallest set of regions in the U.S. which include similar states in terms of racial mix; and second, to establish the precise relationship between our measurements and the classic residential segregation literature that focuses on pairwise comparisons. Among the empirical results, three findings should be highlighted. First, the set of regions and racial groups for whom between-regions and between-cities segregation is important differs from the set for whom the within-cities component dominates. Second, minorities segregation declines but their population share goes up, while whites population share declines but their segregation index increases. Third, the portion of segregation disregarded by the traditional pairwise approach is large in all cases and increasing in the case of within cities segregation between whites and blacks.
