Residential location theory and the measurement of segregation

– Conventional segregation indices indicate complete integration when the proportion of minorities in a metropolitan area is replicated in any subarea. However, this would not occur, even absent discrimination, because majorities and minorities have different distributions of other residence-determining characteristics. Segregation indices derived from predictions of commuting times for blacks, based on white commuting behavior, have a better theoretical foundation, are largely independent of conventional indices, and imply very different segregation rankings for metropolitan areas. La théorie de la localisation géographique et la mesure de la ségrégation RÉSUMÉ. – Les mesures traditionnelles de la ségrégation du logement définissent l'intégration comme se produisant lorsque la proportion de minorités ethniques dans une agglomération est répliquée dans n'importe quelle sous-zone. Ce papier propose des indices de ségregation nouveaux déduits des prédictions des temps de transport domicile-travail des personnes de couleur, sous l'hypothèse qu'ils devraient être identiques aux temps de transport domicile-travail des blancs qui ont les mêmes caractéristiques. Nous montrons que ces nouveaux indices entraînent un classement des agglomerations américaines en termes de ségregation très différents de celui effectué avec les indices standards. * J. S. ZAX: University of Colorado at Boulder, Department of Economics, 256 UCB, Boulder, CO 80309-0256, USA. Telephone: 303-492-8268, FAX: 303-492-8960 e-mail: [email protected] ANNALES D’ÉCONOMIE ET DE STATISTIQUE. – N° 71-72 – 2003 A substantial industry is devoted to measuring and comparing segregation levels across American metropolitan areas. However, it relies upon measures of segregation that define “integration” as occurring when the proportions of minority and majority residents in any part of a metropolitan area are equal to their proportions in the metropolitan area as a whole. This is an extreme assumption. The distributions of many socio-economic and demographic characteristics differ between majority and minority populations. Many of these characteristics may help determine the optimal choice of residential location. Therefore, the geographic distributions of these choices would typically differ between majority and minority populations, even if the latter were not subject to discrimination in housing markets.1 This paper proposes alternative segregation indices. These indices are based on predictions of the behaviors of individual minority workers, under the assumption that they are identical to those of otherwise similar majority workers. Differences between these predictions and the actual behavior of minority workers are, arguably, indications of the costs imposed by segregation. In principle, this methodology is applicable to activity in any market where segregation might affect minority behavior. The housing market is a natural candidate, as the immediate locus of housing discrimination. Labor market behavior might also be affected, if restrictions on residential locations alter employment opportunities. Lastly, segregation is likely to affect commuting patterns, the connection between housing and labor markets. As the initial implementation of this approach, this paper focuses on the comparison of white and black commuting patterns. This comparison is of interest, itself, and has rarely been studied. Moreover, it provides a convenient and tractable opportunity to experiment with the conceptual approach proposed here. Its results constitute a substantial challenge to conventional approaches. Section 1 of this paper briefly describes conventional measures of segregation. Section 2 discusses the effects of segregation in the context of the theory of urban residential location choices. Section 3 introduces the data employed here. Section 4 summarizes regressions specific to each metropolitan area predicting the commuting times of white and black workers. Section 5 analyzes the implications of these regressions for differences between black and white commuting patterns. Section 6 derives metropolitan-level segregation indices from the differences in these patterns. Section 7 concludes. 190 1. BECKER ([1971], 78, footnote 4) makes this same point. 1 Conventional measures of segregation The measurement of residential segregation has a long and multi-disciplinary history. TAEUBER and TAEUBER [1965] is an early contribution that measures segregation for many American cities in 1940, 1950 and 1960. Their work relies upon the index of dissimilarity: