The Geography of Ethnic Residential Segregation: A Comparative Study of Five Countries

Abstract

Few studies have undertaken rigorous comparative analyses of levels of ethnic residential segregation across two or more countries. Using data for the latest available censuses (2000–2001) and a bespoke methodology for such comparative work, this article analyzes levels of segregation across the urban systems of five major immigrant-receiving, English-speaking countries: Australia, Canada, New Zealand, the United Kingdom, and the United States of America. After describing the levels of segregation in each, the article tests a model based on generic factors that should influence segregation levels in all five countries and then evaluates—for the urban population as a whole, for the “charter group” in each society, and for various ethnic minority groups—whether there are also significant country-specific variations in segregation levels. The findings show common factors influencing segregation levels in all five countries: notably the size of the group being considered as a percentage of the urban total, but also urban size and urban ethnic diversity, plus country-specific variations that cannot be attributed to these generic factors. In general there is less segregation in Australia and New Zealand than in the other three countries.

Key Words:

• comparative
• ethnicity
• residential segregation
• urban

Notes

Notes: Min=minimum percentage; Max=maximum percentage; SD=standard deviation.

Notes: Min=minimum percentage; Max=maximum percentage; SD=standard deviation;

N=number of cities.

∗∗∗statistically significant at the 0.001 level or better

∗∗statistically significant at the 0.01–0.001 level

∗statistically significant at the 0.05–0.01 level

Notes: Min=minimum percentage; Max=maximum percentage; SD=standard deviation;

N=number of cities.

∗∗∗statistically significant at the 0.001 level or better

Note: Standard errors are shown in parentheses. Significant regression coefficients at the 0.05 level or better are shown in bold.

Notes: Standard errors are shown in parentheses. Significant regression coefficients at the 0.05 level or better are shown in bold. Key to models (combinations of types): A=V+VI; B=IV+V+VI; C=III+IV+V+VI; D=II+III; E=I+II.

∗∗∗statistically significant at the 0.001 level or better

Note: Standard errors are shown in parentheses. Significant regression coefficients at the 0.05 level or better are shown in bold.

Notes: SD=standard deviation; N=number of cities.

∗∗∗statistically significant at the 0.001 level or better

Note: Standard errors are shown in parentheses. Significant regression coefficients at the 0.05 level or better are shown in bold.

Notes: Standard errors are shown in parentheses. Significant regression coefficients at the 0.05 level or better are shown in bold. Key to models (combinations of types): A=V+VI; B=IV+V+VI; C=III+IV+V+VI; D=II+III; E=I+II.

Notes: SD=standard deviation; N=number of cities.

∗∗∗statistically significant at the 0.001 level or better

∗statistically significant at the 0.05–0.01 level

1. This is the case in work on Canadian cities (see Balakrishnan 2000; Balakrishnan, Maxim, and Jurdi 2005; although see Fong and Wilkes 1999) as well as the United Kingdom (Johnston, Forrest, and Poulsen 2002). In the United States, as Johnston et al. (2004a) point out, most analyses of intermetropolitan differences have similarly lacked rigorous comparative analyses, although see Frey and Farley (1996), Cutler and Glaeser (1997), Cutler, Glaeser, and Vigdor (1999), Iceland, Weinberg, and Steinmetz (2002), Allen and Turner (2005), and Johnston, Burgess, et al. (2006).

2. An official Australian document refers to multiculturalism there as recognizing “Australia's cultural diversity. It accepts and respects the rights of all Australians to express and share their individual cultural heritage within an overriding commitment to Australia and the basic structures of Australian democracy” while seeking to promote “social harmony among the different cultural groups in our society” (DIMIA 2003).

3. These terms are used interchangeably in the text: the urban/metropolitan area and city definitions deployed are those produced by the relevant national census bureaus for data dissemination. The urban areas used are those defined by the various census bureaus: for England and Wales they are the separate built-up areas and not the administrative units (some of which incorporate several urban places) defined by the National Statistics Office; for the United States they are the CMSAs (Combined Metropolitan Statistical Areas) or MSAs (Metropolitan Statistical Areas) defined by the Bureau of Census; for Canada the Metropolitan Areas defined by Statistics Canada; for New Zealand urban places defined by Statistics New Zealand; and for Australia UCLs (Urban Centre Localities) defined by the Australian Bureau of Statistics.

4. There is also a tradition in those countries of academic investigations of segregation and the immigrant experience based on the classic work of the 1920s Chicago School of Urban Sociology.

5. In the U.S. context, for example, Goldberg (1998) contrasts two aspects of segregation because of disadvantage: what he terms the “new segregation,” a process of locking people in to certain types of neighborhood, in contrast to the “old segregation,” where they were locked out of many areas.

6. On the role of voluntary segregation, see the results of a major simulation exercise by Fossett (2006), and the subsequent discussion in the same issue of that journal, which suggests that high levels of segregation are entirely consistent with high levels of voluntary self-segregation, without either discrimination or disadvantage in the housing market. Disproportionate “in-group preferences” for living close to one's co-ethnics and distancing oneself from both the charter group and other minority groups are “segregation-promoting not integration-promoting,” according to Fossett (2006, 258). See also Trudeau (2006).

7. Ellis, Wright, and Parks (2006) have suggested that part of this assimilation process can be related to interethnic group partnering: in Los Angeles, immigrants with partners from a different ethnic group than themselves are less likely than those whose partner is from the same ethnic group to live in segregated enclaves. Somewhat similarly, Johnston, Poulsen, and Forrest (2006b) have shown that people who claim a mixed ethnicity in the United Kingdom are less likely to live in relatively segregated areas than those who identify with one of the country's main ethnic groups.

8. For example, such difficulties have led to some reconsideration of the bases of multiculturalism. In a major speech in December 2006 the U.K. Prime Minister, Tony Blair, not only claimed that “the right to be in a multicultural society was always, always [sic] implicitly balanced by a duty to integrate, to be part of Britain” but also concluded that “Our tolerance is part of what makes Britain, Britain. So conform to it; or don't come here.” The full text is http://www.number10.gov.uk/output/Page10563.asp (last accessed 31 July 2007).

9. There is no evidence of any spatial segregation of the Aboriginal population (defined in the census as those of Aboriginal and Torres Island descent) in Australian cities, so they are not considered further here.

10. Bangladesh became a separate country after the initial waves of Bengali immigrants from East Pakistan (Dench, Gavron, and Young 2006). On the situation of Muslim migrants in the United Kingdom, many of whom are from Pakistan and Bangladesh, see Peach (2006b).

11. The other reasons include comparability over space and time. Most of the indices are not standardized—indeed, with the exception of the index of isolation, standardization procedures have not been developed (see Cutler, Glaeser, and Vigdor 1999; Noden 2000)—and are very much a function of a group's size relative to the number of unit areas. Thus comparing the degree of segregation between two groups in the same city at the same census, one of which forms 15 percent of the total population and the other 2 percent, is not straightforward; nor is a comparison of the distribution of one group at two dates if it forms 5 percent of the total at the first and 12 percent at the second, even if the unit areas stay the same. (For an example of these problems, see Clark 2007.)

12. There is room for further exploration of this claim, but we have no reason to suspect that our findings are particular to the chosen boundaries.

13. Scotland had to be omitted because it has a separate census reporting architecture.

14. The Australian, Canadian, and U.S. first nation groups are not studied here because of their small numbers and the absence of visible segregation within the country's cities.

15. The entropy index is defined as ENTROPY _j =−Σ_i [PERCENT _ij ∗, LOG _e (PERCENT _ij )], where PERCENT _ij is the size of ethnic group i in metropolitan area (MA) j as a percentage of the total MA population; and ENTROPY _j is the measure of entropy for MA j.

16. It has been suggested by one referee that it is improper to use ordinary least squares (OLS) regression when testing such a model because the spatial clustering of the observations could have an impact on the error terms. The nature of the model (comparing five countries) means that in one sense spatial clustering is built in to the entire research program and is taken into account by the dummy variables introduced at stage 2 of the model. Regarding spatial clustering within the individual countries (if there is any), this would only be relevant if there are spatial processes operating so that near places are more alike than more distant places in their segregation levels. We do not specify any such processes, and we treat each observation (city) as an independent entity, hence the use of OLS.

17. That set of interaction terms, rather than others linking country to urban size and/or entropy, was selected after initial exploration of a range of models. Charter group size is the dominant determinant of segregation levels among the three generic variables, as indicated in the results discussed below.

18. For comparisons between the situation of the Maori in New Zealand and Blacks in the United States, see Johnston, Poulsen, and Forrest (2004b, 2006c).

19. Comparisons of the situations of Blacks in the United States and Maori in New Zealand are feasible, though difficult because of the small size of most New Zealand urban areas; they are reported in Johnston, Poulsen, and Forrest (2004, 2006c). Similarly it is possible to conduct formal analyses within individual countries, notably the United States (on which see the comparisons of Black and Hispanic segregation levels in Johnston, Poulsen, and Forrest 2006a).

20. We have omitted Australians classified as of European ancestry, virtually all of whom (the exceptions are in Adelaide, Melbourne, and Sydney) live in Type I and II areas, and for whom there are no direct comparisons in other countries.

21. It is, of course, possible to use birthplace data as Peach (2006a) does, but these have major problems as the Australian data deployed here indicate (see also Poulsen, Johnston, and Forrest 2004).

22. For a comparable argument in the United Kingdom, see the article in The Times by columnist Matthew Parris at http://www.timesonline.co.uk/article/0,1065-2490974.html (last accessed 31 July 2007).

23. For a comparison of the situations for Blacks and Hispanics in the United States, see Johnston, Poulsen, and Forrest (2006a).