Work Together, Live Apart? Geographies of Racial and Ethnic Segregation at Home and at Work

Abstract

When scholars map the urban geography of racial and ethnic segregation, they privilege the time when people are at home. When workers commute, however, the tract of residence of one group often becomes the tract of employment of others. It follows that an exclusive focus on the residential geographies of racial groups erases the presence of others who work in those neighborhoods. Not only does this analytical orientation create a false impression of a city's racialized spaces as fixed, but it also misleadingly characterizes neighborhoods as the domain of those who live, rather than work, in them. In addressing this oversight, the study compares levels of residential and work tract segregation for native-born and immigrant groups in a large U.S. metropolitan area, Los Angeles. The analysis reveals that segregation by work tract is considerably lower than by residential tract, suggesting more intergroup interaction takes place during working hours than at home. The difference in segregation between residence and work is very large in the case of native-born whites and Mexican immigrants. These two groups maintain substantially different residential geographies but are quite likely to work in the same tracts. Such work tract complementarities are gender sensitive; they are much more likely between native-born white and Mexican men than between women of these groups. This gendered difference holds across all groups, with men more likely to work in tracts with men from other groups than women with women from other groups. The study offers new perspective on diurnal shifts in urban racial segregation. We conclude by speculating that reduced segregation at workplaces factors into recent increases in rates of interracial partnering, which may, in turn, ultimately leverage change in residential segregation.

Key Words:

• segregation
• race
• gender
• home
• work

Acknowledgments

The National Science Foundation, under grant BCS-9986928, funded this research.

Versions of this article were presented at the Association of American Geographers Annual Meetings in Los Angeles, March 2002, the First Biennial International Population Geography Conference, St. Andrews, Scotland, July, 2002, the Center for Studies in Demography and Ecology at the University of Washington, February 2003, and the California Center for Population Research, UCLA, November 2003. We thank attendees at these sessions for helpful comments and suggestions. We are also very grateful to Audrey Kobayashi and the referees for their suggestions, to Serin Houston for her valuable editorial advice, and to Chase Langford for the maps and charts. Responsibility for the contents rests with the authors.

Notes

These indexes were calculated between all pairs of groups, for tract of residence, tract of work, industry (all census industry categories), and occupation (all census occupation categories). The coefficients report the variance shared between each index series.

Note: Segregation between men and women (MF) is the omitted gender mix category, and native-born whites are the omitted group category.

1. Sermon delivered at Dexter Avenue Baptist Church, Montgomery, Alabama, on 4 November 1956.

2. A quick search of the Web yielded numerous references to this quotation in newspapers and religious magazines. Interestingly, the quote is frequently edited so as to exclude the word “Christian,” thereby generalizing it beyond the specific religion King targeted with his remarks.

3. The analysis was repeated with exposure indexes (see Massey and Denton [1988] for an explanation of the exposure index and a description of its properties relative to other segregation measures, including the dissimilarity index). As they generate the same results with respect to the pattern of differences between residential and work tract segregation, we do not report them here.

4. For this purpose, intergroup indexes of dissimilarity by occupation and industry were calculated using the full range of 1990 census industrial and occupational categories.

5. In this circumstance, workplace segregation will be correlated with the error term in Equation 1, and the OLS estimated coefficient, β₁, would be biased upward.