GeoJournal 53: 117–124, 2001. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
117
Income segregation, income inequality and mortality in North American metropolitan areas Nancy A. Ross1,2,∗ , Karla Nobrega2 & James Dunn3 1 Department
of Geography, McGill University, 805 Sherbrooke St. West, Montreal, Canada; 2 Health Analysis and Measurement Group, Statistics Canada, R.H. Coats Bldg., 24t h Floor, Ottawa, Canada; 3 Faculty of Medicine, Community Health Sciences Department, Centre for Health and Policy Studies, Health Sciences Centre, G230, 3330 Hospital Drive N.W., Calgary, Canada; (∗ Author for correspondence Tel: 514-398-4307; Fax: 514-398-7437; e-mail:
[email protected])
Received: 12 July 2001; accepted 28 November 2001
Key words: income inequality, income segregation, mortality, North America
Abstract There is a large sociological literature on racial- class-, and economically- based segregation in the United States and there is some evidence that residential segregation by income may have deleterious health consequences for residents of large U.S. cities. The health consequences of segregation in Canadian metropolitan areas, however, remain unknown and the comparison with the U.S. is always compelling. In this paper, we investigate the hypothesis that residential segregation by income may be associated with mortality in Canadian and U.S. metropolitan areas. Given the strong relationship between individual level socioeconomic status and health, it follows that metropolitan areas which isolate individuals economically could produce conditions that severely limit the life chances and therefore the health chances of the most vulnerable. To investigate the association between residential segregation by income and population health, we examined the relationship between working-age mortality and Jargowsky’s (1996) neighbourhood sorting index (NSI) for a large group of North American metropolitan areas. We found a relationship between increased segregation and increased mortality for U.S. metropolitan areas but no such relationship for Canadian metropolitan areas. We also determined that income segregation could not be considered in isolation from income inequality – that, in effect, income inequality provides the propensity for meaningful segregation to occur. We further demonstrated the importance of considering both income inequality and income segregation together, especially when the analysis is intended to compare metropolitan areas. We conclude with a discussion of the need for an improved measure of segregation to better reflect the theoretical arguments for the relationship between concentration of poverty and affluence and population health.
Introduction This principal objective of this paper is to examine the relationship between income segregation, income inequality and mortality in North American metropolitan areas. We hypothesize that mortality is higher in more highly segregated urban areas given the state of empirical and theoretical knowledge in two disparate literatures. The first of these is the recent body of epidemiological research exploring the relationship between income inequality and various summary measures of population health within North America (Ross et al., 2000; Lynch et al., 1998; Kaplan et al., 1996) and internationally (e.g., Wilkinson, 1992). The second may be broadly termed structuralist sociological research on the American urban underclass by, among others, researchers like Wilson (1987), Sampson et al. (1997), Massey (1996, 1990). Geographers and urban sociologists throughout the 20th century long theorized about the importance of neighbourhoods in terms of their effects on the life chances of individ-
uals. We also know from a huge body research in epidemiology that individuals higher up the social ladder in terms of education, income or depth of their social networks have longer life expectancies than individuals further down on the rungs).1 Epidemiological evidence has also demonstrated that social structures, like neighbourhoods, that undermine individual educational attainment, labour market success or social connectedness can influence individual health outcomes, above and beyond the now well-known individual socioeconomic risk (e.g., Cubbin et al. 2000; Robert 1998; Shouls et al., 1996; Jones and Duncan, 1995). The influence of neighbourhood of residence on life chances, furthermore, begins early: neighbourhood poverty is consistently associated with low school readiness and achievement and behaviour and emotional problems in children (see review in Leventhal and Brooks-Gunn, 2000). Thus there is good evidence about individual and neighbourhood socio-economic health determinants. This paper builds on these established relationships by considering the added influence on population health of metropolitan areas
118 in North America that segregate individuals into homogeneous neighbourhoods of wealth and poverty. We extend a previous comparative analysis of the relationship between income inequality and mortality in Canadian and U.S. cities (Ross et al., 2000) to consider the relationship between residential segregation by income and mortality using Jargowsky’s Neighborhood Sorting Index (NSI, see Jargowsky, 1996). While it is admittedly difficult to discuss residential segregation in the U.S. without at least a comment about the interplay between race and class and their co-effects on segregation (Farley et al., 1978), our objective here is to consider segregation by income in North American metropolitan areas. There simply is not the same degree of racial segregation in Canadian cities that would make a racial segregation comparison with the U.S. meaningful. How might segregation be linked to health? There are several reasons to believe that the spatial patterning of income inequality could be an important factor in the relationship between income inequality and population health, but there is a conceptual challenge in developing such explanations. While a highly unequal income distribution within a metropolitan area is suggestive of poverty amidst affluence (which is thought to have direct psychosocial health consequences related to one’s appraisal of place in the social order (Wilkinson, 1996)),2 residential segregation by income means that poverty and affluence are spatially concentrated, and the spatiality of income distribution introduces an added dimension to the income inequality and health relationship. In general, residential segregation by income is a spatial expression of income inequality and it may serve to reinforce any negative effects of inequality, as was first pointed out by Blau in 1977 and reinforced by Massey et al. (1991). Segregation, in effect, sets up the theoretical possibility of ‘triple health jeopardy’ – being poor in a poor neighbourhood that is spatially isolated from lifeenhancing opportunities such as good quality education and employment. Specifically, there are at least three possible explanations for a link between residential segregation by income and population health. The first such explanation appeals to notions of the social and economic consequences of social isolation for lowincome, segregated groups of people. In the first instance, the spatial segregation of lower socio-economic groups can act as an obstacle to economic opportunity, as illustrated by research on the ‘spatial mismatch’ of jobs and housing, a common topic of U.S. urban research. Concentrated poverty, especially in large contiguous areas in inner city or older suburban areas, sets up a ‘spatial mismatch’ between employment opportunities in new high growth suburban areas, public transit patterns and place of residence. Thus the poor become spatially isolated from opportunities to improve their life chances through employment. This effect is compounded by outmigration from central cities of affluent households which further isolates the poor, this time from important job networks, as well as institutions and norms of modern society (Wilson, 1987). The presence of conditions of spatial mismatch, common in U.S. cities, means
that the spatial isolation of lower socio-economic groups serves to compound their disadvantage. There is also evidence that isolation/segregation is difficult to overcome. A study by South and Crowder (1997) showed that it is much more difficult to move out of extremely poor neighbourhoods in metropolitan areas that are highly segregated. A second wrinkle to the social isolation theory concerns the spatial expression of political power. High concentrations of poverty and affluence, in other words, can create large imbalances in power relations within urban areas. There is good evidence that this has helped affluent areas, for example, to consistently and successfully oppose unwanted, annoying, or even toxic land uses in their neighbourhoods (e.g., Greenberg and Schneider, 1996). The second possible explanation for links between residential segregation by income and population health argues that highly segregated metropolitan areas in the U.S. often face a fundamental incapacity to produce health-enhancing public goods. For example, in urban areas where public services (schools, public housing, health care, recreation, etc.) are financed primarily by local property tax revenues, high concentrations of poverty and affluence can create severe inequities in public services. Under such conditions, areas with high social needs typically have fewer fiscal resources (low property tax base) and therefore must charge higher taxes in order to offer only a modest package of social programs (Orfield, 1998). This has the added effect of motivating households living in low tax base municipalities to move, if they can, to higher tax base municipalities where there exists a better tax-benefit equation. This further deprives the departed municipality of much needed tax base. Such patterns and practices are facilitated in many states by relatively liberal regulations concerning municipal incorporation, which can produce a highly fragmented governance structure across metropolitan areas. (Dunn, forthcoming; Orfield, 1998; Massey, 1996). A final theory suggests that residential segregation by income has the capacity to create corrosive social relationships in segregated, deeply impoverished places. It has been hypothesized that residential segregation by income can promote distrust between groups and decline in overall social connection within metropolitan communities. Given what is known about the importance of social relations to health (House et al., 1988; Berkman and Syme, 1979), it is possible that the health of both the poor and the rich could be compromised by deteriorating metropolitan-level social relations (Veenstra, 2001; Kawachi et al., 1997). Despite these compelling theoretical arguments, there has been little large-scale empirical work linking segregation to health outcomes. There has been some recent research on the links between segregation and infectious disease (Acevedo-Garcia, 2001) and the work by Guest et al. (1998) found some direct effects of racial segregation on black mortality in Chicago neighbourhoods. Waitzman and Smith (1998) conducted an 8-year mortality follow-up of individual respondents in 30 large U.S. cities to the U.S. National Health Interview Survey. They found a large and significant relative risk of mortality (for 30 to 60 years olds and for
119 those over 65) associated with living in a highly segregated metropolitan area (e.g., New Orleans or Chicago) compared with living in a relatively non-segregated metropolitan area (e.g., San Jose or Seattle). These place-level mortality risks, furthermore, were reduced but remained significant even when adjusted for individual characteristics. Still, the present study is the first that we know of to take a cross-national, metropolitan scale approach to investigate the connection between segregation based on income and mortality. Methods Associations between income inequality, income segregation and mortality were studied using 249 U.S. and 53 Canadian metropolitan areas with populations greater than 50,000 in 1990 in the U.S. and 1991 in Canada. The 249 U.S. metropolitan areas were those census metropolitan areas for which correct matches between the mortality and segregation measures could be made.3 Mortality outcomes were limited to three-year average death rates (centred around the census years) for the working-age population (25 to 64 years), given that the principal interest here is in premature mortality and given that the migration patterns of wealthy, healthy retirees to cities in the sunbelt unduly influence the mortality patterns in those cities. The details of the sources of the mortality estimates and the income inequality measure are outlined in our earlier work (Ross et al., 2000). Briefly, income inequality was defined as the proportion of total household income belonging to the less well off 50% of households within a metropolitan area (i.e., the ‘median share’ of income). In a situation of perfect equality, the bottom half of the income distribution for a particular geographic area would receive 50% percent of the income earned in that area. Thus higher scores (those closer to 0.50) on the median share measure indicate more equal income distributions. Income segregation was measured using the Neighbourhood Sorting Index (NSI) developed by Jargowsky (1996). The NSI is defined as the standard deviation of household income between census tracts within a metropolitan area divided by the standard deviation of household income across the entire census metropolitan area and is calculated as: N hn (y¯n − y) ¯ 2 n=1
σN = NSI = σH
H N (yi − y) ¯ 2 i=1
H Where hn is the number of households within a census tract, y is the household income, H is the total number of households and N is the total number of census tracts within a census metropolitan area. When there is no segregation, all neighborhoods have the same mean income, there will be no difference between the total average income and the total neighborhood income, and the NSI will be zero. At the
other extreme, if within a neighborhood there is no income variance, then the variance between the neighborhoods will be equal to the variance across the city and the NSI will be one. This situation represents total income segregation. A twenty-percent representative sample of the 1991 Canadian census was used to calculate the NSI measure. Only private households, not including collective dwellings and households outside of Canada were used. Household income was defined as annual post-transfer and pre-tax total income from all sources. Jargowsky has provided the American NSI figures using census tracts within US metropolitan areas. Jargowsky’s measure has the desirable property of being independent of the mean and variance of income within an area. Also the measure does not depend on any within-city arbitrary thresholds for poverty or affluence. Measures that rely on cutpoints, such as the P-index, for example, can be influenced by changes in the mean incomes within a city even though the level of segregation has not changed (Waitzman and Smith, 1998). Despite these desirable properties, the NSI measure is not independent of the variance in income distributions between cities, a point we will return to later in the paper. The regression modelling proceeded incrementally beginning with U.S. models of first, income inequality, then models of the main effects of income inequality and income segregation, followed by the full model including the interaction between inequality and segregation. All models were weighted by the metropolitan area’s population size. The approach was repeated for the Canadian metropolitan areas. Lastly, combined North American models were calculated with a dummy variable added to take account of mortality differences between Canada and the United States after accounting for median share, NSI and NSI-median share interactions.
Results Income inequality As we have shown previously, Canadian metropolitan areas have more equal income distributions with a higher proportion of total income received by the less well off half of the population than do their American counterparts (Figure 1). The proportion of income accruing to the less well-off half of households ranged from 0.15 (least equal) in Bryan, Texas to 0.25 (most equal) in Jacksonville, North Carolina for the U.S. while the range in Canada was 0.22 (least equal) for Montreal, Quebec to 0.26 (most equal) for Barrie, Ontario. This greater median share for the Canadian cities is generally indicative of less income variation within a city. Canadian metropolitan areas also tend to have lower working-age mortality rates than U.S. metropolitan areas with Canadian cities clustering in the lower right hand corner of Figure 1. There is a significant strong correlation (ρ = −0.43, P < 0.01) between median share and workingage mortality for the U.S. metropolitan areas, but no such relationship exists for the Canadian cities (ρ = −0.05, P = 0.72).
120
Figure 1. Working-age mortality by proportion of income belonging to the less well off half of households, U.S. (1990) and Canadian metropolitan areas (1991).
Figure 2. Working-age mortality rate and the Neighbourhood Sorting Index, U.S. (1990) and Canadian metropolitan areas (1991).
121
Figure 3. Working-Age Mortality by Proportion of Income Belonging to the Less Well Off Half of Households and Neighourhood Sorting Index, U.S. Metropolitan Areas (1990) and Canadian Metropolitan Areas (1991).
Income segregation
Regression analyses
While there was a great deal of separation between Canadian and American metropolitan areas on income inequality, there was little Canada-U.S. separation on the NSI measure (Figure 2). The median NSI value in the U.S. was 0.33 and in Canada the value was 0.30 (a significant difference (P = 0.04), but the scatterplot of the relationship between mortality and segregation shows that Canadian and U.S. metropolitan areas look remarkably similar. The NSI ranged from least segregated, 0.19 in Joplin, MO to most segregated, 0.70 in Naples Florida in the U.S. and from the least segregated 0.13 in Moncton, New Brunswick to the most segregated 0.52 in Montreal, Quebec in Canada. There is a modest significant positive correlation between NSI and working-age mortality in the US (ρ = 0.18, P < 0.01). The correlation between mortality and NSI in Canada is actually negative (ρ = −0.28, P = 0.04), which is contrary to the hypothesized direction of the relationship. NSI and median share are correlated (ρ = −0.19, P < 0.01) for the 302 North American metropolitan areas, as well as for the 249 U.S. metropolitan areas (ρ = −0.14, P = 0.03) and for the 53 Canadian metropolitan areas (ρ = −0.28, P = 0.05). Thus inequality (lower median share scores) is consistently associated with segregation (higher NSI scores).
U.S. models. Model 1a indicates a strong negative relationship between income inequality as measured by the median share and working-age mortality in U.S. metropolitan areas (Table 1). The interpretation of the coefficient here is that a hypothetical 1% increase in the share of income to the bottom half of the income distribution would result in a decrease in the mortality rate of the working-age population of about 22 deaths per 100,000. The magnitude of this effect is perhaps best appreciated by the following: the hypothetical decline in mortality from income redistribution is about the equivalent of eliminating all deaths from motor vehicle accidents and breast cancer in the U.S., combined. When segregation as measured by the NSI is added (Model 2a), the coefficient for income inequality is reduced slightly and there is a significant effect of segregation on mortality in U.S. cities (P < 0.01). The interpretation of the segregation effect here is that a one point increase in the NSI measure is associated with an increase in the mortality rate of 1.5 deaths per 100,000, while holding constant the effects of income inequality. This may be an oversimplified interpretation, however, given that income inequality and segregation are correlated. In other words, it is unrealistic to imagine holding either income inequality or income segregation constant given their interrelationship (see discussion of multicollinearity in Neter et al., 1990). The full model including the interaction between inequality and segregation (Model 3a) indicates that the effect of inequality remains significant while the other effects no longer maintain statistical significance. The model, however, containing the main effect of inequality and the interaction between inequality and segregation (Model 4a)
122 Table 1. Regression results.
U.S. Model 1a Median share Model 2a Median share NSI Model 3a Median share NSI Interaction:Median share-NSI Model 4a Median share Interaction: Median share-NSI CANADA Model 1b Median share Model 2b NSI NORTH AMERICA Model 1c Median share Country Model 2c Median share NSI Country Model 3c Median share NSI Interaction: Median share-NSI Country Model 4c Median share Interaction: Median share-NSI Country
Coefficient
P -value
Adj R 2
−21.71
<0.01
0.33
−17.51 1.53
<0.01 <0.01
0.23
−29.90 − 5.62 33.95
0.02 0.44 0.32
0.23
terpreted as the incremental mortality gap between Canada and the United States, after accounting for the other variables in the models. For example, in Model 2c, the mortality rate in Canadian metropolitan areas is lower than the U.S. metropolitan area mortality rate by 39 deaths per 100,000, even after accounting for the influences of income inequality and income segregation. As was the case for the U.S. model (4a), the main effect for income inequality and the interaction effect between inequality and segregation are both significant predictors and they account for 45% of the variation in working-age mortality in North American metropolitan areas.
−20.21 7.41
<0.01 <0.01
0.23
Interpreting the Interaction Effect: Income Inequality as Propensity for Segregation
− 1.11
0.78
0.00
−43.89
0.36
0.02
−20.87 −0.33
<0.01 <0.01
0.51
−15.97 0.02 −38.85
<0.01 0.04 <0.01
0.45
4.68 −7.15 −24.50 −38.82
0.03 0.48 <0.01 0.30
0.46
−17.39 3.92 −38.82
<0.01 0.04 <0.01
0.45
has both predictor variables attaining statistical significance with a reasonable proportion in the variation in working-age mortality explained (R-sq = 0.23). Canadian models. For the US metropolitan areas, median share explains 33% of the mortality variance, whereas in Canada, the median share explains a trivial amount of the variance in mortality rates. Neither inequality nor segregation is an important predictor of working-age mortality in Canadian metropolitan areas (Models 1b and 2b). North American models. The North American models, not surprisingly, follow a similar pattern to the U.S. models. Both income inequality and income segregation are significant predictors of workingage mortality (Models 1c and 2c). The country dummy variable in each of the North American models may be in-
Interaction effects are best visualized graphically. Figure 3 is a 3-D scatterplot of the relationship between income inequality, income segregation and working-age mortality for the North American cities. We saw earlier (Figure 2) that the Canadian cities do not appear to be strikingly different from U.S. cities on the NSI segregation measure even though Canadian cities have more equal income distributions, and conventional wisdom would suggest that U.S. cities are more highly segregated. The NSI measure alone, however, does not take into account whether a city is more or less equal than another city in terms of the overall distribution of household income, instead it is indicative of relative differences in the variation of income within a neighbourhood relative to variation across that individual city. Therefore, one should be cautious when using it to compare cities without also considering their income distributions. Figure 3 demonstrates that income inequality provides, effectively, the propensity for segregation to occur within any given city and the increases in risks of mortality, as shown in the regression analyses, are linked to the interaction between inequality and segregation. In other words, there are additional deleterious population health effects resulting from being in an unequal metropolitan area which is also highly segregated.
Discussion The objective of this paper was to examine the relationships between income inequality, income segregation and mortality for North American metropolitan areas. As we have shown in previous work, Canadian metropolitan areas have both more equal income distributions and lower mortality rates for the working-age population than their U.S. counterparts. Canadian metropolitan areas are not as segregated by income, although the distinction between the metropolitan areas within the two countries was not as stark as was anticipated. Segregated U.S. cities tend to be characterized by a large contiguous area, either the inner city or the near suburbs, of extreme poverty, absence of services and jobs, and a large proportion of black residents. Canadian metropolitan areas have not tended to follow this pattern. Indeed, a major difference in the social geography of Canadian versus American cities has been the endurance, and in some instances
123 even growth, of wealthy neighbourhoods in large Canadian inner cities, supported by restrictive zoning regulations and preserved by proximity to institutions like prestigious private schools (Ley, 1993). To resolve this apparent inconsistency, we showed that income segregation could not be considered in isolation from income inequality – that, in effect, income inequality provides the propensity for meaningful segregation to occur. As regards the population health relationships, there was a strong relationship between income inequality and working-age mortality for U.S. metropolitan areas and a modest, but significant relationship between income segregation and working-age mortality. Neither of these hypothesized relationships held for Canadian metropolitan areas. The U.S. findings lend some plausibility to the hypothesis that the spatial patterning of income inequality is important for understanding the broader socio-economic determinants of population health. That said, the segregation measure used in this analysis, the Neighbourhood Sorting Index (NSI), does not allow us to evaluate the plausibility of the hypothetical explanations we enumerated at the outset of this paper: isolation, public goods, and decline in social relations. Indeed, a major shortcoming of the NSI is that it is a-spatial – it treats all spatial units as independent observations and cannot take account of spatial contiguity, and identify larger concentrations of poverty and affluence. The shortcoming of this measure is underscored by Massey and Denton’s (1988) analysis showing five distinct dimensions of segregation from 20 commonly used segregation measures: evenness, exposure, concentration, centralization and clustering. They describe these characteristics of segregation in the following manner: ‘At a general level, residential segregation is the degree to which two or more groups live separately from one another, in different parts of the urban environment. This general understanding masks considerable underlying complexity, however, for groups may live apart from one another and be ‘segregated’ in a variety of ways. Minority members may be distributed so that they are overrepresented in some areas and underrepresented in others, varying on the characteristic of evenness. They may be distributed so that their exposure to majority members is limited by virtue of rarely sharing a neighborhood with them. They may be spatially concentrated within a very small area, occupying less physical space than majority members. They may be spatially centralized, congregating around the urban core, and occupying a more central location than the majority. Finally, areas of minority settlement may be tightly clustered to form one large contiguous enclave, or be scattered widely around the urban area.’ (pp. 282–283) Their 1988 analysis was based on segregation measures derived from the 1980 U.S. census, and was conducted using a subset of U.S. metropolitan areas. They found that although correlated, the five dimensions were reasonably empirically distinct.4 From a health point of view, then, it worth reflecting on the theoretical links between the various dimensions of segregation and health outcomes. It may well be that a
subset of the dimensions identified by Massey and Denton (1988) and Massey et al. (1996) take on theoretical salience when the objective is to link segregation to health. In addition to incorporating spatial contiguity, as argued above, a more robust segregation index would allow for Massey and Denton’s five different dimensions of segregation to be distinguished empirically, and their association with population health ascertained. From the above, it could be argued that the concepts of isolation, concentration and centralization should be focused on in studies linking segregation to health. In any event, it is clear from the foregoing that further research on the relationship between residential segregation by income and population health is needed, and that an empirical strategy that can help to distinguish between competing hypotheses about the pathways between residential segregation, life chances, and health chances, would be an influential next step in advancing our understanding of the socio-economic determinants of population health.
Acknowledgements We gratefully acknowledge funding support from the Canadian Population Health Initiative.
Endnotes 1 See
a recent issue of the International Journal of Epidemiology (2001:30) for eight papers and several commentaries on socioeconomic health differentials, 2 For a fuller discussion of the theorectical links between inequality and health see Ross et al. (forthcoming). 3 This is a smaller number than the 282 U.S. metropolitan areas that were used in the Lynch et al. (1998) and the Ross et al. (2000) analyses. The 249 were the subset for which the geography for the segregation and mortality measures was consistent. The effect of losing metropolitan areas is likely to diminish somewhat the strength of the relationship between segregation and mortality in the U.S. given that there were several large metro areas (e.g., New York City) excluded. The 53 Canadian metropolitan areas in this analysis are identical to those that appear in Ross et al. (2000). 4 A subsequent analysis (Massey et al., 1996) using 1990 U.S. census data for a much larger set of U.S. metropolitan areas reached a different conclusion. The dimensions of evenness and exposure were very apparent again in 1990, but the dimensions of concentration, centralization and clustering did not share the same degree of emperical distinctiveness.
References Acevedo-Garcia D., 2001: Zip code-level risk factors for tuberculosis: neighbourhood environment and residential segregation in New Jersey, 1985–1992. Am. J. Publ. Health, 91: 734–741.
124 Berkman L.F. and Syme L., 1979: Social networks, host resistance, and mortality: A nine-year follow-up study of Alameda County residents. Am. J. Epidemiol., 109: 186–204. Blau P.M., 1977: Heterogeneity and Inequality: A Primitive Theory of Social Structure. Free Press, New York, N.Y. Cubbin C., LeClere F.B., Smith G.S., 2000: Socioeconomic status and injury mortality: Individual and neighbourhood determinants. J. Epidemiol. Comm. Health, 54: 517–524. Dunn, J.R., forthcoming, ‘Public Goods’, Metropolitan Inequality, and the Social Production of Health. In: Drache D. and Sullivan T. (eds), Health Reform: Public Success, Private Failure, 2nd ed. McGill-Queen’s University Press. Farley R., Schuman H., Bianchi S., Colasanto D. and Hatchett S., 1978: Chocolate city, vanilla suburbs: Will the trend toward racially separate communities continue? Soc. Sci. Res., 7: 319–344. Greenberg M. and Schneider D., 1996: Environmentally Devasted Neighbourhoods: Perceptions, Realities and Policies. Rutgers University Press, New Brunswick, NJ. Guest A.M., Almgren G., and Hussey J.M., 1998: The ecology of race and socioeconomic distress: Infant and working age mortality in Chicago. Demography, 35: 23–34. House J.S., Landis K.R., and Umberson D., 1988: Social relationships and health. Science 214: 540–545. Jargowsky, P.A., 1996: Take the money and run: economic segregation in U.S. metropolitan areas. Am. Soc. Rev., 61: 984–998. Jones K. and Duncan C., 1995: Individuals and their ecologies: Analysing the geography of chronic illness within a multilevel modelling framework. Health and Place, 1: 27–40. Kaplan G.A., Pamuk E., Lynch J.W., Cohen R.D., and Balfour J.L., 1996: Income inequality and mortality in the United States: analysis of mortality and potential pathways. Brit. Med. J., 312: 999–1003. Kawachi I., Kennedy B.P., Lochner S.M., and Prothrow-Stith D., 1997: Social capital, income inequality and mortality. Am. J. Publ. Health, 87: 1491–1498. Leventhal T. and Brooks-Gunn J., 2000: The neighbourhoods they live in: The effects of neighbourhood residence on child and adolescent outcomes. Psychol. Bull., 126: 309–337. Ley D.F., 1993: Past elites and present gentry: Neighbourhoods of privilege in the inner city. In: Bourne L.S. and Ley D.F. (eds), The Changing Social Geography of Canadian Cities. McGill-Queen’s University Press, Montreal. Lynch J.W., Kaplan G.A., Pamuk E.R., Cohen R.D., Heck K.E., Balfour J.L. et al., 1998: Income inequality and mortality in metropolitan areas of the United States. Am. J. Publ. Health, 1998: 1074–1080. Massey D.S., 1996: The age of extremes: Concentrated poverty and affluence in the twenty-first century. Demography, 33: 395–412. Massey D.S., 1990: American apartheid: Segregation and the making of the underclass. Am. J. Sociol., 95: 1153–1188.
Massey D.S. and Denton N.A., 1988: The dimensions of residential segregation. Soc. Forces, 67: 281–315. Massey D.S., Gross A.B., and Eggers M.L., 1991: Segregation, the concentration of poverty, and the life chances of individuals. Soc. Sci. Res. 20: 397–420. Massey D.S., White M.J., and Phua V.-C., 1996: The dimensions of segregation revisited. Soc. Meth. Res. 25: 172–206. Myles J., Picot G. and Pyper W., 2000: Neighbourhood Inequality in Canadian Cities. Analytical Studies Branch Research Paper Series, Statistics Canada, Ottawa. Neter J., Wasserman W., and Kutner M.H., 1990: Applied Linear Statistical Models. Irwin, Boston, MA. Orfield M., 1997: Metropolitics: A Regional Agenda for Community and Stability. Brookings Institution Press and the Lincoln Institute of Land Policy, Washington, DC and Cambridge, MA. Robert S.A., 1998: Community-level socioeconomic status effects on adult health. J. Health Soc. Behav., 39: 18–37. Ross N.A. , Wolfson M.C. and Dunn J.R., forthcoming: Why is mortality higher in unequal societies? Interpreting income inequality and mortality in Canada and the United States. In: Boyle, Curtis, Gatrell, Graham and Moore (eds), The Geography of Health Inequalities in the Developed World. Ross N.A., Wolfson M.C., Dunn J.R., Berthelot J.-M., Kaplan G.A., and Lynch J.W., 2000: Relation between income inequality and mortality in Canada and the United States: Cross sectional assessment using census data and vital statistics. Brit. Med. J., 320: 898–902. Sampson R.J., Raudenbush S.W., and Earls F., 1997: Neighbourhoods and violent crime: A multilevel study of collective efficacy. Science, 277: 918–924. Shouls S., Congdon P., Curtis S., 1996: Modelling inequality in reported long term illness in the U.K.: combining individual and area characteristics. J. Epidemiol. Comm. Health, 50: 366–376. South S.J. and Crowder K.D., 1997: Escaping distressed neighborhoods: Individual, community, and metropolitan influences. Am. J. Soc., 102: 1040–1084. Veenstra G., 2001: Social capital and health. ISUMA Can. J. Pol. Res., 2: 72–81. Waitzman N.J. and Smith K.R., 1998: Separate but lethal: The effects of economic segregation on mortality in metropolitan America. Milbank Mem. Fund. Q., 76: 341–373. Wilkinson R.G., 1992: Income distribution and life expectancy. Brit. Med. J., 301: 165–168. Wilkinson R.G., 1996: Unhealthy Societies: The Afflictions of Inequality. Routledge, London. Wilson W.J., 1987: The Truly Disadvantaged. University of Chicago Press, Chicago.
