Who you know, where you live: social capital, neighbourhood and health

ARTICLE IN PRESS

Social Science & Medicine 60 (2005) 2799–2818 www.elsevier.com/locate/socscimed

Who you know, where you live: social capital, neighbourhood and health Gerry Veenstraa,, Isaac Luginaahb, Sarah Wakefieldc, Stephen Birchd, John Eylese, Susan Elliotte a

Department of Anthropology and Sociology, University of British Columbia, 6303 N. W. Marine Dr., Vancouver, Canada, V6T 1Z1 b Department of Geography, University of Western Ontario, London, Ontario, Canada c Department of Geography, University of Toronto, Toronto, Canada d Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontorio, Canada e School of Geography and Geology, McMaster University, Hamilton, Ontorio, Canada Available online 25 December 2004

Abstract This article examines the degree to which relationships between social capital and health are embedded in local geographical contexts and influenced by demographic factors, socio-economic status, health behaviours and coping skills. Using data from a telephone survey of a random sample of adults (N ¼ 1504 respondents, response rate ¼ 60%), the article determines if relationships between involvement in voluntary associations and various measures of individual health are associated with neighbourhood of residence in the mid-sized city of Hamilton, Canada. Associational involvement and overweight status (assessed by body-mass score) were weakly but significantly related after controlling for the other variables; involvement had relationships with self-rated health and emotional distress before but not after controlling for socio-economic status, health behaviours and coping skills. Relationships between neighbourhood of residence and two health outcomes, self-rated health and overweight status, were statistically significant before and after controlling for the other characteristics of respondents; neighbourhood of residence was not a significant predictor of number of chronic conditions and emotional distress in multivariate models. The neighbourhood and associational involvement relationships with health were not dependent upon one another, suggesting that neighbourhood of residence did not help to explain the positive health effects of this particular measure of social capital. r 2004 Elsevier Ltd. All rights reserved. Keywords: Social capital; Social networks; Neighbourhood; Health behaviours; Coping skills; Canada

Introduction The social capital and health discourse, intently focused on certain social networks, i.e., voluntary associations, has generally acknowledged the interconnectedness of the micro-level (individuals participating in such networks), the meso-level (the social networks Corresponding author.

E-mail address: [email protected] (G. Veenstra).

themselves), and the macro contexts that shape both individuals and networks (e.g., political and economic structures). The discourse has not yet seriously grappled with the ways in which associational networks and their health effects are potentially embedded within specific geographical contexts such as the neighbourhood or community. In the context of four neighbourhoods in one mid-sized Canadian city, this article contributes to understanding how social capital influences health and well-being within geo-political contexts by: (i) assessing

0277-9536/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2004.11.013

ARTICLE IN PRESS 2800

G. Veenstra et al. / Social Science & Medicine 60 (2005) 2799–2818

degree and type of involvement in networks of voluntary association for a random sample of adults, (ii) assessing relationships between associational involvement and various measures of physical and emotional health and (iii) determining the degree to which these relationships are dependent upon neighbourhood of residence. Social capital and health Social capital is generally described as a feature of social structure, e.g. a web of cooperative relationships between citizens, high levels of interpersonal trust, and strong norms of reciprocity and mutual aid, that serve to facilitate action for shared benefit (Coleman, 1988; Putnam, Leonardi, & Nanetti, 1993). Such features of social structure—potentially including networks based in voluntary associations—may serve to further the goals of individuals but may also act as direct resources for social groups and communities (Lin, 2001). ‘Social capital’ as a theoretical concept emerged from the sociological and political science literatures (Bourdieu, 1984, 1986; Coleman, 1988; Putnam et al, 1993) and since the mid-1990s has increasingly been incorporated into health research as a way to bring social theory into epidemiological studies, at times as a mechanism to link social or economic inequality and health (Hawe & Shiell, 2000). The social capital and health discourse is not a body of research that identifies a single capital that influences health in an easily identifiable way. Rather, social capital is an element in a theoretically and empirically contentious, broadly defined dialogue. To date, an amorphous group of indicators of social capital (e.g., social networks and support, involvement in associations, measures of trust) have been tied by various theoretical and empirical means (e.g., the character of political governance, economic growth, the quality of health care, stress, social support) to numerous health outcomes (e.g., self-rated health, mortality rates, life expectancy). The breadth of this dialogue makes it difficult to conceptualize as well as investigate empirically how social capital might manifest itself in neighbourhood contexts and subsequently influence health. Social capital in various forms is hypothesized to affect health in three major ways. First, it may influence an individual’s health as a result of its direct and beneficial effects on individual attributes and activities, what are often called the ‘compositional’ health effects of social capital. For example, Berkman, Glass, Brissette, and Seeman (2000) suggest that social networks in general (and, we argue, networks of voluntary association in particular) provide social support, exert social influence, encourage social engagement and facilitate interpersonal bonding for members. These aspects of social networks may then influence the health of members by influencing physiological stress responses,

self esteem and security, health behaviours (e.g., smoking, exercise, high-risk sexual activity, health service utilization) and exposure to infectious disease agents (Berkman et al., 2000). The degree to which such networks, behaviours and exposures are spatially situated and/or their health effects potentially mitigated by spatial context are seldom addressed by public health researchers. In this article we seek to address this gap in the literature by determining if the breadth and depth of associational involvement interacts with neighbourhood of residence as a determinant of individual health in the city of Hamilton, Canada. We also determine if psychological coping skills and health behaviours operate as intervening variables in involvement–health relationships. Lastly, as some kinds of networks may be more likely than others to provide social support, social influence and interpersonal bonding, we explore the salience of participation in different types of associations, e.g., sports, religious, cultural and professional associations, for various measures of health and wellbeing. Second, social capital may influence health indirectly through its effects on the larger social, economic, political and environmental factors that in turn function as determinants of the health of populations. These are usually referred to as the ‘contextual’ health effects of social capital. For example, social capital could affect health by influencing a community’s access to economic resources and material goods (e.g., jobs and economic opportunities, housing, and institutional contacts— Berkman et al., 2000). It may also influence broader aspects of the economy and the polity (Putnam et al., 1993; Helliwell & Putnam, 1995; Rice & Sumberg, 1997; Woolcock, 1998; Fukuyama, 2000) in ways that may have consequences for the health of whole communities, populations and societies (Kawachi, Kennedy, Lochner, & Prothrow-Stith, 1997; Veenstra, 2002). Conversely, social capital may be influenced by other social, economic and political phenomena, with subsequent health implications. For example, social capital is thought by some to mediate relationships between socio-economic factors such as income inequality and population health (Wilkinson, 1996; Kawachi et al., 1997). Given that income inequality may be predictive of health at the level of the neighbourhood (Wilson & Daly, 1997), variation in social capital among neighbourhoods may help to explain the differential effects of the inequality of resources within neighbourhoods on health. A neighbourhood or community with robust social capital may be better able to organize against local environmental hazards as well. In short, health researchers have suggested that social capital can influence the shape and character of the larger social (and geo-political) context in which individuals live their lives, indirectly affecting health (Mohan & Mohan, 2002).

ARTICLE IN PRESS G. Veenstra et al. / Social Science & Medicine 60 (2005) 2799–2818

There is a third way in which social capital may influence population health: interactions with other health determinants at the individual or group level. At the group level, for example, social capital may interact with neighbourhood wealth as a determinant of population health. The health effects of collective wealth, e.g., less violent crime and more green space, may be more salient in contexts with a dense mapping of networks of association, as residents in such places have resources to better deal with the consequences of crime and networks enabling greater use of green space. At the level of the individual, for example, participation in some social networks may provide a social context within which education is more or less strongly related to health. We suspect that social networks drawn from a broad range of community members may provide contexts within which social status is especially salient to regular interactions, whereas status may not permeate inter-subjective interactions networks drawn from a limited socio-economic stratum. We have seen very little work of this kind in the social capital and health literature. We seek to address this gap in a small way by comparing (in exploratory fashion) the health effects of associational involvement in advantaged neighbourhoods with health effects in disadvantaged neighbourhoods. More empirical research is needed in all three dimensions because the evidence to date for relationships between various conceptions of social capital and health, while promising in some instances, is still equivocal. At the ecological level, using spatially defined communities, several studies have reported a relationship between measures of social capital and overall health measures (Kawachi et al., 1997; Putnam, 2000; Veenstra, 2002; Lochner, Kawachi, Brennan, & Buka, 2003). These relationships are not all large and robust, and could be due to confounding factors. For example, Ellison (1999) found that the prevalence of interpersonal trust was related to aggregate self-rated health at the level of the nation before but not after controlling for societal wealth and income inequality (see also Lynch et al., 2001). At the level of the individual, associational involvement and interpersonal trust have been found to be related to health in some studies (Baum et al., 1999; Rose, 2000) but not in others (Ellaway & Macintyre, 2000; Veenstra, 2000). In general, these studies do not explicitly situate and interpret associational involvement, trust and health relationships within specific geographical contexts. Results from several multilevel studies, a statistically viable way of distinguishing contextual effects from individual (compositional) factors, have identified relationships between poor health and low social capital at levels such as the American state (Kawachi, Kennedy, & Glass, 1999), American community (Subramanian, Kim, & Kawachi, 2002), and Scottish postcode (Ellaway & Macintyre, 2000) after

2801

controlling for individual factors, suggesting that social capital and its health effects may be embedded to some degree within such communities. Some recent qualitative work in disadvantaged neighbourhoods in the UK suggests that neighbourhood characteristics can affect the nature and extent of local social capital, in turn affecting health (Cattell, 2001). This and other research has tended to focus on disadvantaged neighbourhoods, however, meaning that few conclusions can be made about the nature of social capital and health associations in affluent areas and their manifestations in affluent versus deprived neighbourhoods (Forrest & Kearns, 2001). In summary, most social capital and health studies do not provide an in-depth understanding of the local, contingent mechanisms through which social capital may influence health in certain geographical contexts and not in others. During the last decade, a growing body of research has attempted to assess the importance of characteristics of neighbourhoods, and neighbourhood- or local-area deprivation in particular, as determinants of health in and of themselves, above and beyond the health effects of characteristics of the individuals who live in the neighbourhoods (e.g., Duncan, Jones, & Moon, 1993, 1995; Sloggett & Joshi, 1994, 1998; Shouls, Congdon, & Curtis, 1996; Kaplan, 1996; Diez-Roux et al., 1997; Davey-Smith, Hart, Watt, Hole, & Hawthorne, 1998). Even though results from this latter body of work also paint a contradictory picture, as ‘‘there appear to be some area effects on some health outcomes, in some population groups, and in some types of areas’’ (Macintyre, Ellaway, & Cummins, 2002:128), we argue that attributes of neighbourhoods may help to explain the health effects of social capital. Macintyre et al. (2002) note the need for rigorous empirical investigations that explicitly theorize the mechanisms linking neighbourhood and health. In particular, these authors emphasise the importance of collective elements of community life, of ‘‘shared norms, traditions, values, and interests’’ (p. 130), essentially, of social capital, grounded in place, mediating context and composition and potentially linking neighbourhood with health. We attempt to respond to this call via close empirical attention to the neighbourhood-specific health effects of participation in a certain kind of social network in the civil space, the voluntary association. We assembled an original quantitative data set that enables us to explore this line of questioning. Our research utilizes data from a telephone survey administered to a random sample of residents in four neighbourhoods in one Canadian city (in order to compare social capital health effects across neighbourhoods) and a random sample from the remainder of the city (in order to provide baseline comparisons). Specifically, it explores the associations between degree and type of associational involvement and a variety of health



outcomes in both advantaged1 and deprived neighbourhoods, representing our attempt to tease out some of the variable and neighbourhood-contingent compositional effects of social capital for health. Following the arguments raised earlier, we hypothesize that involvement and health will be more strongly related in some neighbourhoods than in others. We also hypothesize that the relationships will be stronger in disadvantaged than in advantaged neighbourhoods: participation in associations may be most beneficial for health in places where other types of resources are not readily available. Noting the multidimensional nature of participation in social networks, we simultaneously assess both breadth and depth of association (number and strength of ties, respectively), hypothesizing that both dimensions have the potential to influence health. We also incorporate consideration of concepts that we believe have the potential to intervene between associational involvement and well-being, namely, psychological coping skills and health behaviours. Finally, we assess participation in various kinds of associations, hypothesizing that some forms of engagement (such as participation in sporting activities) will be more strongly related to some measures of health than others.

Stages of analysis We explored the association between involvement in voluntary associations and health within neighbourhood contexts by addressing the following sequence of analytical stages for each of four measures of health and well-being (self-rated health, emotional distress, number of chronic health conditions and body-mass index score): Analytical Stage One. Is involvement in voluntary associations related to health at the bi-variate level? To address this question, we explored zero-order associations in the full survey sample between overall degree of involvement in voluntary associations, degree of involvement in specific types of associations, and health. We utilized Spearman’s r, a non-parametric measure of association, when both variables were at least ordinal and used Z to assess association and the one-way ANOVA test of significance when one variable was categorical and the other interval or ratio. Analytical Stage Two. Do zero-order involvement–health relationships differ by neighbourhood? We then calculated zero-order associations (Spearman’s r and Z) between overall and type-specific associational 1 ‘Advantage’ in this instance is determined inductively in part by the spatial clustering of latent factors pertaining to various dimensions and types of socio-economic diversity and disadvantage. We describe the identification of spatial clustering among neighbourhoods in Hamilton later in the article.

involvement and health within each of the four neighbourhoods and the remainder of the city as a whole. Analytical Stage Three: Do health effects of associational involvement and neighbourhood of residence interact with one another? We included both involvement and neighbourhood of residence in multivariate logistic regression models on dichotomized versions of the health variables and assessed whether the interaction between involvement and neighbourhood of residence made statistically significant contributions to the models. We also controlled for age and gender (and their potential interaction with one another) at this stage as they may serve to identify spurious involvement–health relationships. Analytical Stage Four: Do personal socio-economic characteristics influence relationships between involvement, neighbourhood of residence and health? We then controlled for socio-economic status, i.e., education and income, in the multivariate models, as they also have the potential to identify spurious involvement–health relationships. We also included the statistically significant interaction terms between age, gender, education and income at this stage in order to control for all of their health effects. Analytical Stage Five: Do health behaviours and/or coping skills influence the relationship between involvement and health? Finally, we added alcohol consumption, smoking, exercise and coping skills variables to the multivariate models, arguing that these variables have the potential to intervene between involvement and health. We also included the statistically significant interaction terms between age, gender, socio-economic status, health behaviours and coping skills. One particular limitation of our analytical inquiry deserves special mention. Our objectives were accomplished using measures of association and logistic regression modelling, not multilevel modelling. While multilevel modelling has become relatively widespread for investigating the effects of social structure and/or neighbourhood on health, it has stringent data requirements, i.e., an absolute minimum of 25 people in 25 places (Paterson & Goldstein, 1992). Nevertheless, concerns have arisen about the appropriateness of multilevel modelling as a tool for investigating contextual effects, particularly at the neighbourhood level. The method assumes a fixed, quantifiable spatial hierarchy that uniformly affects those lodged within it (Duncan, Jones, & Moon, 1998). While this is clearly appropriate in certain settings (e.g., schools, clinics), the validity of these assumptions is less clear in relation to areas and particularly neighbourhoods (which are open, permeably bounded systems, in which local relations of power and association may be contingent and variable). The method also requires that a formal distinction be made between individual-level and


area-level characteristics (Subramanian, Lochner, & Kawachi, 2003), with the effect of reifying the distinction between composition and context and presuming that variables on either side of the context/composition divide play confounding rather than intervening roles (Macintyre et al., 2002). In essence, this leads to models that may ‘throw away’ geography, focusing on the proportion of variance that is ‘left’ after controlling for factors at the individual level rather than elucidating the mechanisms that etch social processes in place (Duncan et al., 1998). The analytical techniques we utilize in this article suit the nature of the data and our research questions.

Sampling and data This study focused on four neighbourhoods in Hamilton, Canada (see Table 1) that were specifically chosen to represent different neighbourhood circumstances, namely, socio-economic and socio-demographic diversity. (See Luginaah et al., 2001 for an in depth description of the selection process). Selection of the neighbourhood areas was conducted using Census Tract-level socio-demographic data from the 1996 Census and utilized three analytical methods: principal component analysis (PCA), local indicators of spatial association (LISA) and geographical information systems (GIS). First, we utilized PCA to reduce seventeen socio-demographic variables to four factors. Factor one was characterized by a low percentage of married persons, high mobility, low education, low dwelling values, low incomes, high poverty rates, a high percentage of lone parent families and low proportions

2803

of the elderly. It was labelled ‘low income, immigrants, single persons households with children’. Factor two was called the ‘blue-collar manufacturing’ component and was characterized by high mobility, fairly high manufacturing employment, very low incomes and high proportions of smokers. Factor three was dominated by recent immigrants who are young, work in manufacturing and are non-English speakers and visible minorities, called the ‘blue-collar immigrants’ component. The final factor contained the highest proportion of elderly persons who have lower incomes and live alone, named the ‘lone, low incomes elders’ component. Next, LISA statistics and an overlay process in Arcview GIS were used to identify spatial clusters of Census Tracts in Hamilton based on these variables (Luginaah et al., 2001). Neighbourhoods were then chosen to represent four spatial clusters based upon the four factors. The Central Downtown Core is a socially diverse and low income neighbourhood, the Northeast Industrial neighbourhood depicts low social diversity and low incomes, the Chedoke-Kirkendall neighbourhood has high incomes and moderate social diversity and the Southwest Mountain neighbourhood shows ethnic diversity but has little economic diversity (see Table 1 and Luginaah et al., 2001). Of the four neighbourhoods, Downtown and Northeast are economically ‘disadvantaged’ relative to one another and the rest of Hamilton while Kirkendall and Mountain represent the advantaged neighbourhoods of the city. We drew a random sample of households from the neighbourhoods and the city as a whole from recent municipal property assessment records and sent contact

Table 1 Demographic characteristics of the neighbourhoods, 1996 Census (adapted from Table 5, Luginaah et al., 2001)

Married (%) Recent immigrants in past five years (%) Persons with 1 year mobility status (%) Persons with high school education or less (%) Persons with post-secondary education (%) Persons in manufacturing employment (%) Average dwelling value ($) Median household income ($) Statistics Canada low income classification (%) Unemployment rate (%) Lone parent families (%) Non-English or French speakers (%) Live alone (%) Aboriginal (%) Visible minority (%) Income inequality, Gini coefficient Elderly population (%)

Kirkendall

Downtown

Northeast

Mountain

City of Hamilton

38.6 3.4 20.0 33.1 49.3 12.7 173,553 40,856 19.5 8.5 16.4 1.1 40.8 1.0 9.0 0.4 14.3

33.0 9.2 24.3 62.6 16.8 20.6 106,780 18,567 53.7 20.6 29.1 8.6 40.0 2.5 23.0 0.4 17.8

43.5 1.9 16.2 60.5 23.5 25.4 96,170 31,835 29.1 12.3 27.3 1.8 26.7 2.5 6.3 0.4 12.0

54.3 2.9 12.4 41.5 36.0 20.0 179,780 55,146 21.1 9.5 16.8 1.1 10.3 0.7 15.0 0.3 8.3

46.2 3.9 15.5 50.2 31.8 20.8 134,788 37,893 23.9 10.9 18.4 2.3 28.0 1.0 10.8 0.4 15.9



letters outlining the nature of the study to potential households. The telephone survey, containing a range of questions designed to capture participation in social and community networks, health status and behaviours, use and access of health services and socio-demographic factors, was administered to respondents between November 2001 and April 2002. The survey, approximately 25–30 min in length, was completed by phone with approximately 300 randomly selected individuals in each neighbourhood along with additional respondents randomly selected from across the city excluding the neighbourhoods, referred to subsequently as the City of Hamilton. The survey was conducted by the Institute for Social Research at York University, Toronto and had a total sample size of N ¼ 1504 and response rate of 60%. Individual weights calculated by the ISR, based on the probability of an individual being selected due to differential household size, were used in all calculations. Responses to selected survey questions are presented in Table 2 and the appendix. Income was not reported by approximately 17% of respondents and there was a higher probability of non-report in females, those with lower self-assessed health status, older persons and those not having completed high school. Due to this systematic bias and to include more cases in the final analysis, we interpolated values using a hot-deck imputation approach (Ford, 1983). This approach stratifies the sample based on gender, age, education and employment status (employed versus unemployed), factors that were associated in our sample with lower reported income. For each missing value, one case with the same characteristics for the stratification variables was chosen randomly and that income value substituted for the missing one. We dichotomized this variable for use in multivariate analysis by distinguishing household incomes below (26.2% of the sample) from those above $30,000. For those income-health relationships that were not roughly linear, i.e., self-rated health and body-mass index, this cut-point represented the point where a superimposed Lowess curve changed direction. We also dichotomized education for use in multivariate analyses by distinguishing respondents who completed high school (79.1% of the sample) from the others. In education-health relationships, this cut-point was at or close to the point where the superimposed Lowess curve changed direction for the self-rated health, number of chronic conditions and body-mass dependent variables. Separate questions were used to measure emotional and physical health status. Emotional health was measured using 20 questions of the General Health Questionnaire, collapsed into a single scale to assess emotional distress. Although originally designed for inperson interviews, the reliability coefficients for this measure of emotional distress are very similar when administered by telephone survey. For use in multivariate analysis we dichotomized this scale by distin-

guishing those who had four or more positive responses (unhappy) from those who had fewer than four positive responses (happy), the standard threshold indicating higher or lower levels of distress (Goldberg, 1972; McDowell & Newell, 1987). Chronic conditions was captured by a single variable representing the number of positive responses to a series of questions regarding physician-diagnosed long term skin conditions, arthritis or rheumatism, asthma, high blood pressure or hypertension, diabetes, urinary tract problems, stomach ulcers, digestive problems and cancer. We dichotomized this variable for use in multivariate models by distinguishing an absence of chronic conditions (representing 44.4% of respondents) from the presence of one or more. Next, we assessed body mass index (BMI) using height and weight, with a score less than 20 representing ‘underweight,’ 20 through 27 representing ‘normal’ and higher than 27 representing ‘overweight’, distinguishing the final category from the first two in multivariate analysis. Finally, self-rated health status was evaluated using a five-part variable based on responses to the question ‘‘In general, compared to other people your age, would you say your health is excellent, very good, good, fair or poor?’’ We dichotomized this variable for use in multivariate analysis by distinguishing fair and poor health from the other responses. To measure social capital we focused specifically on breadth and depth of involvement in voluntary associations. We asked respondents if they belonged to an association and assessed its type, i.e., religious, cultural/ historic, community, social services/health, sports/athletics, pastimes/social/artistic, professional or political (see Table 3 for examples), followed by a question assessing degree of involvement in the association. Respondents were given the opportunity to specify three different associations. Results from these questions were combined into a single index of involvement by assigning 2.0 for each of the groups respondents participate in ‘a great deal’, 1.5 for those with which they are ‘somewhat involved’, 1.0 to those with which they have ‘a little involvement’ and 0.5 to those they are ‘not involved with at all’ but are still members. The minimum score on this index was zero (no groups were mentioned) and the maximum was 6.0. Separate indices of involvement were also created for each of the eight associational types. Each index was markedly skewed, as the proportion of respondents in the entire sample who participated in one or more voluntary associations was only 33.4%. Non-participation was concentrated in the Downtown and Northeast neighbourhoods, where only 29.6% and 24.6% of respondents participated in one or more associations, respectively. To measure health behaviours we asked questions pertaining to smoking, alcohol consumption and frequency of physical exercise. Lastly, to measure coping skills we asked respondents questions pertaining to their

Table 2 Characteristics of the survey sample

Derived item: age

Overall Mean (n)

Kirkendall

Downtown

45.7 (1484)

46.3 (286)

44.9 (249)

Northeast 44.6 (309)

Mountain 47.1 (317)

City of Hamilton 45.5 (324)

Female (%, n)

52.4 (788)

56.6 (162)

49.3 (124)

51.3 (160)

54.0 (174)

50.5 (167)

What is the highest level of education you have completed?

Less than high school (%, n) High school College/technical school Some university Bachelors degree Post-graduate degree

20.9 36.5 17.2 4.7 13.0 7.6

(310) (543) (256) (69) (192) (114)

13.9 24.5 14.1 7.3 23.8 16.5

(39) (70) (40) (21) (68) (47)

26.2 40.5 12.9 4.8 10.2 5.4

28.3 44.7 15.0 3.8 5.8 2.4

(86) (136) (46) (11) (18) (8)

15.7 34.7 22.3 3.1 16.3 7.7

(50) (110) (71) (10) (51) (25)

21.0 38.3 20.5 4.6 9.1 6.6

(70) (126) (67) (15) (30) (22)

Could you please tell me how much income you and all other members of your household received in the year 2000? Be sure to include income from all sources such as savings, pensions, rent, and unemployment insurance as well as wages.

o $10,000 (%, n) 10–19,999 20–29,999 30–39,999 40–49,999 50–59,999 60–69,999 70–79,999 80–89,999 90–99,999 4$100,000

3.2 10.1 12.9 13.3 12.1 11.4 8.2 7.2 6.3 4.8 10.3

(47) (150) (192) (197) (179) (170) (122) (107) (94) (72) (153)

2.7 5.9 11.5 11.3 9.3 12.4 8.6 7.7 5.5 6.6 18.5

(8) (17) (33) (32) (27) (35) (24) (22) (16) (19) (53)

6.4(16) 18.9 (47) 18.7 (47) 14.6 (36) 10.4 (26) 7.3 (18) 5.6 (14) 5.6 (14) 2.9 (7) 4.6 (11) 5.0 (12)

4.3 9.9 16.4 15.7 14.8 14.8 7.2 4.9 6.3 2.0 3.6

(13) (30) (50) (48) (45) (45) (22) (15) (19) (6) (11)

0.8 10.5 8.7 10.4 10.9 9.6 10.4 6.3 10.0 6.8 15.7

(3) (33) (28) (33) (34) (30) (33) (20) (32) (21) (49)

2.2 7.0 10.6 14.6 14.4 12.4 8.9 10.9 6.2 4.3 8.6

(7) (23) (35) (48) (47) (41) (29) (36) (20) (14) (28)

Calculated scale: emotional distress

Mean (n)

1.05 (1504)

0.95 (287)

1.45 (252)

1.24 (312)

0.84 (322)

0.88 (331)

Calculated item: number of chronic conditions

Mean (n)

1.02 (1504)

0.98 (287)

1.12 (252)

1.08 (312)

0.97 (322)

1.00 (331)

In general, compared to other people your age, would you say your health is excellent, very good, good, fair, or poor?

Excellent (%, n) Very good Good Fair Poor

22.1 35.6 27.6 9.8 4.9

28.8 34.9 26.1 8.4 1.8

19.5 35.1 25.7 10.4 9.3

15.9 30.1 31.5 13.9 8.6

24.4 42.3 23.2 7.4 2.7

22.0 35.4 31.0 8.8 2.8

Mean (sd, n) Not at all (%, n) o once/week About once/week About twice/week About three/week 4three/week

0.794 (1.35, 1504) 10.0 9.3 16.2 14.2 16.4 34.0

(150) (139) (243) (212) (245) (509)

(82) (99) (74) (24) (5)

1.204 (1.66, 287) 5.3 9.4 13.2 15.6 20.7 35.8

(15) (27) (38) (45) (59) (102)

(49) (88) (64) (26) (23)

0.644 (1.20, 252) 15.7 6.4 11.5 13.2 12.4 40.8

(40) (16) (29) (33) (31) (103)

(49) (93) (97) (43) (27)

0.541 (1.14, 312) 10.6 8.6 18.4 11.5 17.9 32.9

(33) (27) (57) (35) (55) (1 0 1)

(79) (136) (75) (24) (9)

0.932 (1.39, 322) 9.9 11.4 16.9 17.4 13.8 30.7

(32) (36) (54) (56) (44) (98)

(72) (116) (102) (29) (9)

0.657 (1.22, 331) 9.3 10.1 19.7 13.1 16.7 31.3

(31) (33) (65) (43) (55) (103)

2805

Calculated index: overall associational involvement In the past three months, how often have you engaged in any physical exercises or activities?

(330) (532) (412) (146) (73)

(65) (102) (32) (12) (25) (13)

ARTICLE IN PRESS

Derived item: respondent’s gender


Question/variable

2806

Table 2 (continued ) Question/variable

Overall

Kirkendall

Downtown

Northeast

Mountain

City of Hamilton

During the past 12 months, how often did you drink alcoholic beverages?

Not at all (%, n) o once/month About once/month 2–3 times/month About once/week 2–3 times/week 4–6 times/week Every day

21.7 19.2 12.3 12.5 12.4 13.7 3.2 5.0

At the present time do you smoke cigarettes daily, occasionally, or not at all?

Daily (%, n) Occasionally Not at all

21.9 (330) 6.4 (97) 71.6 (1076)

16.9 (48) 7.3 (21) 75.9 (217)

28.5 (72) 4.5 (11) 67.0 (169)

29.1 (90) 6.7 (21) 64.2 (199)

16.6 (54) 5.6 (18) 77.7 (251)

19.8 (66) 7.7 (25) 72.5 (240)

Now, I’d like to ask you about how you cope with the daily demands in your life. How would you rate your ability to handle the day-to-day demands in your life, like work or school or family responsibilities?


26.7 38.9 24.6 7.1 2.8

(397) (579) (366) (106) (41)

27.1 43.0 22.9 6.2 0.7

(77) (122) (65) (18) (2)

22.3 34.9 26.1 11.6 5.1

(55) (86) (64) (29) (12)

26.0 35.2 27.8 8.2 2.8

(81) (109) (86) (25) (9)

29.2 37.8 23.8 5.7 3.4

(93) (121) (76) (18) (11)

27.7 42.8 22.5 4.9 2.1

(91) (140) (74) (16) (7)

How would you rate your ability to handle unexpected and difficult problems, like a family or personal crisis?


18.7 37.7 31.9 9.0 2.8

(277) (559) (472) (133) (42)

20.5 34.6 34.2 8.2 2.6

(58) (98) (97) (23) (7)

15.4 32.1 34.9 11.6 5.9

(38) (79) (86) (29) (15)

19.0 34.5 34.2 10.1 2.2

(59) (107) (106) (31) (7)

17.5 42.2 28.3 8.8 3.1

(56) (134) (90) (28) (10)

20.3 43.2 28.8 6.7 1.0

(66) (140) (94) (22) (3)

(324) (288) (184) (187) (185) (204) (48) (75)

12.6 12.6 12.8 13.9 13.9 19.4 7.1 7.7

(36) (36) (36) (40) (40) (55) (20) (22)

26.0 19.3 14.5 11.3 9.9 9.9 2.7 6.5

(64) (48) (36) (28) (24) (24) (7) (16)

24.4 21.6 12.9 11.2 12.9 12.2 2.2 2.7

(76) (67) (40) (35) (40) (38) (7) (8)

23.7 20.0 13.9 11.8 11.8 11.5 1.9 5.3

(76) (64) (45) (38) (38) (37) (6) (17)

21.5 21.9 8.2 14.2 13.1 15.1 2.5 3.6

(71) (72) (27) (47) (43) (50) (8) (12)

ARTICLE IN PRESS


Table 3 Associational categories and examples

Church (most of the responses) Anglican Business Women Buddhist association Catholic Women’s League Jehovah’s Witnesses

Jewish Women Organization Knights Columbus Olive Branch Ministries drop-in centre Synagogue

Cultural/Historic: cultural (i.e., ethnically based) and historical organizations

Canadian Warplane Heritage Museum

Member of a Korean choir

Francophone Community Centre Lithuanian Canadian Community

Mennonite council

Community: community organizations, seniors organizations, volunteering with seniors, associations for schools or youth, associations pertaining to environmental and nature concerns

YMCA/YWCA Literacy group Ladies auxiliary Volunteer Hamilton Neighbourhood watch Rotary Club Neighbourhood association Optimist club Charity work through Hamilton Tiger Cats Kiwanis Hamilton Association for Community Living The Legion Volunteer at seniors home Seniors club

The Hamilton senior games Big Sisters Association Girl Guides Parent council Big Brothers Board of directors at day-care Volunteer for Hamilton District School Scouts Air Cadets Healthy Moms, Healthy Babies Nutrition and activities for children at school board Green Peace Royal Botanical Gardens (RBG) SPCA

Social Services/Health: healthy lifestyles organizations, health-care associations and institutions, associations for helping the homeless and disadvantaged

Volunteer at hospital (most of the responses) Alcoholics Anonymous Women for Sobriety Burn support group in hospital Victoria Order of Nurses Brain Injuries Services in Hamilton Canadian Diabetes Association Ronald McDonald House McMaster Children’s Centre

Canadian National Institute for the Blind Cancer assistance program The Good Shepherd United Way Hamilton Out of the Cold Meals on Wheels Sexual assault centre Red Cross emergency response team

ARTICLE IN PRESS

Religious: religious organizations, institutions and associations

2807

Examples G. Veenstra et al. / Social Science & Medicine 60 (2005) 2799–2818

Category and descriptor

2808

Table 3 (continued )

Sports/Athletics: organized sports and athletics leagues and associations (either as player or coach)

Swimming club Basketball association Ball hockey team Bowling league Hamilton rep hockey Hamilton ringette association Hamilton old sports association

Hamilton youth soccer association Minor Hockey Association Figure skating Fitness club Tai-chi Volunteer coaching as teacher

Pastimes/Social/Artistic: big artistic institutions, small-scale artistic groups, social hobbies, artistic hobbies

Hamilton Art Gallery Theatre Aquarius Hamilton Orchestra Singing choir Hamilton poetry centre Hamilton quilters guild Art club

Clarinet choir Yacht club Men’s club Women’s club Horticultural Society Model railway club Friday night social group

Professional: professional organizations and associations

Teacher’s federation of Ontario Union Medical specialist foundation Canadian Medical Association Librarian Association College of Nurses of Ontario

Hamilton Lawyers Club Ontario Dental Association McMaster Alumni Association McMaster Medical Students Council Writers Union

Political: political parties and committees, activism associations

Social Awareness Society Status of Women-City of Hamilton Chamber of Commerce

Canadian Alliance Green Party

ARTICLE IN PRESS

Examples


Category and descriptor


ability to handle the day-to-day demands in life, such as work, school, or family responsibilities; and their ability to handle unexpected and difficult problems, like a family or personal crisis. We dichotomized both coping variables for use in multivariate analysis by distinguishing ‘fair’ and ‘poor’ responses from the rest.

Results We note that body-mass index (Cramer’s V ¼ :076; p ¼ :036), emotional distress (Z¼ :102; p ¼ :004), selfrated health (Cramer’s V ¼ :103; po:001) and overall associational involvement (Z¼ :177; po:001) differed significantly by neighbourhood. (The remainder of the city of Hamilton was included in these comparisons as a fifth category.) Good health measured in these ways and high associational involvement were highest in Kirkendall and Mountain, the ‘advantaged’ neighbourhoods. In light of this provocative finding, are involvement and health related to one another at the level of the individual and do these relationships differ by neighbourhood? To answer these questions, we explored the five analytical stages in sequence for each of the dependent variables, beginning with self-rated health. Involvement, neighbourhood and self-rated health Although more involvement corresponded with better self-rated health in the overall sample (Table 4, addressing analytical stage one), the association was only statistically significant in Kirkendall and Northeast (Table 4, addressing stage two), a relationship that was apparently explained in large part by involvement in sports/athletics in particular. After controlling for gender, age and neighbourhood of residence in Model I of Table 5, overall involvement retained a modest but non-significant effect on self-rated health (addressing stage three). The interaction term between involvement and neighbourhood of residence was not statistically significant and so was not retained in the model. Upon controlling for income and education (Model II, addressing stage four), involvement lost strength as a predictor of self-rated health. Model III addresses the final stage of analysis: after additionally controlling for smoking, alcohol consumption, exercise habits and coping skills, associational involvement was mostly unrelated to self-rated health. In summary, although zero-order relationships suggested that the effect of involvement on self-rated health may have been dependent upon the neighbourhood context, multivariate modelling suggested that this seeming dependence was influenced in part by age and a direct effect of neighbourhood of residence on health. Also, the (potentially causal but weak) relationship between overall involvement and self-rated health that

2809

persisted after controlling for age, gender and neighbourhood of residence may have been explained in large part by income and education (as antecedent variables that influence involvement and then health, or as variables making the involvement and health relationship partially spurious). When it comes to sensitive exploration of causal pathways, however, regression models such as these are not adequate to the task. At this point we can only state with confidence that the relationship between involvement and self-rated health weakened considerably after controlling for these demographic, socio-economic, health behaviour and coping variables. Involvement, neighbourhood and chronic conditions The index of involvement was significantly and positively related to the number of chronic conditions (r ¼ :074; p ¼ :002), such that more involved respondents tended to have more chronic conditions on average. After controlling for age, gender and neighbourhood of residence, the involvement coefficient was small and non-significant (Table 6). It appears that older people and women were more likely to report chronic conditions and to be involved in associations, creating a spurious zero-order relationship between involvement and chronic conditions and making the remaining analytical stages irrelevant. In short, neither associational involvement or neighbourhood of residence was a significant predictor of number of chronic conditions after controlling for age and gender. Involvement, neighbourhood and emotional distress More involvement corresponded with better scores for this measure of emotional well-being in the Mountain neighbourhood, a relationship that was apparently explained by participation in religious and professional groups in particular (Table 4). Overall involvement and the dichotomized version of emotional distress remained significantly related after controlling for age, gender and neighbourhood of residence (Model I in Table 7). The interaction term between involvement and neighbourhood of residence was non-significant. Controlling for income and education had no effect on the involvement–health relationship but did remove most of the effect from neighbourhood of residence (Model II). Upon inclusion of health behaviours and coping skills, involvement lost strength as a predictor of emotional distress to the point of non-significance (Model III). In summary, the relationship between involvement and emotional distress held before and after controlling for neighbourhood of residence, age, gender and socioeconomic status. The relationship did not hold in a meaningful way after controlling for health behaviour and coping variables, suggesting that the health

2810

Overall

Kirkendall

Downtown

Northeast

Mountain

City of Hamilton

.102 (o.001, 1780) .111 (.041, 341) .054 (.023, 1795) .026 (.636, 345)

.065 (.237, 333) .109 (.038, 366) .007 (.901, 366) .122 (.018, 374) .009 (.865, 336) .021 (.682, 372) .150 (.004, 366) .100 (.052, 376)

Religious involvement & self-rated health Cultural/historic involvement & self-rated health Community involvement & self-rated health Social services/health involvement & self-rated health Sports/athletics involvement & self-rated health Pastimes/social/artistic involvement & self-rated health Professional involvement & self-rated health Political involvement & self-rated health

.044 (.067, 1780) .010 (.687, 1780) .053 (.026, 1780) .023 (.336, 1780) .104 (o.001, 1780) .075 (.002, 1780) .026 (.274, 1780) .036 (.133, 1780)

.010 (.851, 341) o.001 (.999, 341) .103 (.057, 341) .018 (.736, 341) .146 (.007, 341) .075 (.169, 341) .006 (.918, 341) .107 (.048, 341)

.108 (.049, 333) .108 (.750, 333) .049 (.376, 333) .008 (.886, 333) .077 (.161, 333) .069 (.211, 333) .035 (.521, 333) .021 (.700, 333)

.062 (.237, 366) .060 (.250, 366) .004 (.939, 366) .043 (.410, 366) .178 (.001, 366) .102 (.051, 366) .026 (.623, 366) .062 (.233, 366)

.005 (.929, 366) .060 (.250, 366) .056 (.286, 366) .035 (.507, 366) .076 (.145, 366) .041 (.439, 366) .057 (.280, 366) .080 (.124, 366)

.010 (.842, 374) .010 (.850, 374) .116 (.025, 374) .053 (.305, 374) .028 (.586, 374) .030 (.559, 374) .038 (.468, 374) .006 (.907, 374)

Religious involvement & emotional distress Cultural/historic involvement & emotional distress Community involvement & emotional distress Social services/health involvement & emotional distress Sports/athletics involvement & emotional distress Pastimes/social/artistic involvement & emotional distress Professional involvement & emotional distress Political involvement & emotional distress

.057 (.015, 1795) .005 (.829, 1795) .022 (.348, 1795) .045 (.055, 1795) .046 (.050, 1795) .075 (.002, 1795) .021 (.378, 1795) .065 (.006, 1795)

.071 (.190, 345) .100 (.063, 345) .022 (.688, 345) .069 (.204, 345) .014 (.795, 345) .065 (.228, 345) .039 (.465, 345) .136 (.012, 345)

.037 (.505, 336) .007 (.901, 336) .038 (.486, 336) .055 (.315, 336) .023 (.678, 336) .110 (.043, 336) .008 (.883, 336) .062 (.257, 336)

.012 (.822, 372) .015 (.769, 372) .073 (.163, 372) .054 (.295, 372) .049 (.350, 372) .068 (.192, 372) .036 (.491, 372) .088 (.090, 372)

.144 (.006, 366) .098 (.061, 366) .017 (.745, 366) .054 (.301, 366) .088 (.092, 366) .085 (.103, 366) .115 (.028, 366) .081 (.124, 366)

.057 (.267, 376) .023 (.656, 376) .040 (.440, 376) .003 (.958, 376) .016 (.757, 376) .040 (.436, 376) .073 (.160, 376) .023 (.656, 376)

ARTICLE IN PRESS

Overall involvement & self-rated health Overall involvement & emotional distress


Table 4 Relationships between involvement in associations and health; Spearman’s r (p, N)

Table 5 Logistic regression models, probability of fair or poor self-rated health Model I B

SIG

Exp(B)

B

SIG

Exp(B)

B

SIG

Exp(B)

0.123

0.051 o.001 0.648 0.007 o.001 0.432 o.001 0.521

0.885

0.082

0.922

0.028

0.094 0.429 0.655 0.246 0.008 0.462 0.895 0.634

0.911 1.536 1.924 0.782 1.009 0.63 0.409 0.53

0.135 0.279 0.697 0.331 0.014 0.287 0.815 0.094

0.679 0.002 0.639 0.303 0.005 0.249 0.178 0.418 0.003 0.743 0.019 0.005 0.38 0.003 0.021 o.001 0.418 0.012 0.348 0.019 0.039 0.004

0.972

0.887 1.891 2.176 0.817 1.02 1.102

0.202 0.001 0.725 0.078 0.004 0.352 0.081 0.094 o.001 o.001

0.12 0.637 0.778 0.202 0.02 0.097

2.904

0

0.055 1471 0.04 85.3 0 85.3 52.98 (o.001) 0.035 0.062

0.694

0.037

2.001

1.243

0.001

0.288 1451 0.07 85.5 50 85.5 79.91 (o.001) 0.053 0.095

1.615 0.829 0.133 0.11 1.769 0.198 0.031 0.021 0.882 0.762 1.792

1.144 1.322 2.008 0.718 0.986 0.751 0.443 0.91 0.199 0.437 0.875 0.896 0.171 0.821 1.032 1.021 2.417 0.467 6 1416 0.16 87.8 61.7 86.7 181.99 (o.001) 0.12 0.215

ARTICLE IN PRESS

N D2 Sensitivity % Specificity % Percent correctly classified % Model w2 (sig.) Cox & Snell R2 Nagelkerke R2

Model III


Involvement Neighbourhood Kirkendall Downtown Northeast Mountain Age Gender (Female) Education (less than high school) Income (less than $30,000) Smoking Not at all Occasionally Alcohol Exercise Coping with daily demands Coping unexpected difficulties Age Smoking Occasionally Age Smoking Daily Gender Education Gender Income Constant

Model II

2811



Table 6 Logistic regression model, probability of one or more chronic conditions Model I B Involvement Neighbourhood Kirkendall Downtown Northeast Mountain Age Gender (Female) Constant

0.061 0.242 0.122 0.028 0.239 0.048 0.368 2.008


behaviours and/or coping variables intervened between involvement and emotional well-being. Involvement, neighbourhood and body-mass index Overweight and then underweight respondents were the least likely to be involved in voluntary associations (Z¼ :069; p ¼ :033) at the bi-variate level. Overall involvement was a significant predictor of overweight status after controlling for age, gender and neighbourhood of residence (Model I in Table 8). The interaction term between involvement and neighbourhood of residence was not statistically significant, and neither health behaviours and coping skills or income and education influenced the involvement–health relationship in a meaningful way. In contrast with the other health variables, involvement remained a moderately strong and statistically significant predictor of overweight status after controlling for all the other variable. In this instance we can state with some confidence that more involvement in networks of associations corresponded with a lower likelihood of being overweight before and after controlling for a wide variety of other predictors. Summary of results With respect to our research questions, we found that depth and breadth of involvement in voluntary associations was significantly related to emotional distress and over-weight status and almost significantly related to

SIG

Exp(B)

0.16 0.505 0.172 0.542 0.87 0.168 o.001 0.001 o.001

1.063 0.785 0.894 0.972 0.788 1.049 1.445 0.134 1484 0.1 62.3 68.2 65.8 212.13 (o.001) 0.133 0.178

self-rated health before and after controlling for age, gender and neighbourhood of residence. More participation in voluntary associations apparently had a positive relationship with these measures of health. Socio-economic status, measured by income and education, was potentially antecedent to the relationship between involvement and self-rated health (but not to the relationships between involvement and the other health measures). The degree to which respondents engaged in healthful behaviours and the degree to which they felt equipped to deal with the problematic parts of life potentially intervened in relationships between involvement and self-rated health and emotional distress (but not body-mass and number of chronic conditions). Finally, relationships between involvement and health were not obviously embedded within specific neighbourhoods. To summarize the secondary findings, neighbourhood of residence had a statistically significant relationship with several measures of well-being—residents of the Northeast neighbourhood in particular were the most likely to report fair/poor self-rated health and to be overweight after controlling for the other variables. We also found that older respondents were more likely to have chronic conditions and be overweight, but were less likely to report emotional distress. Women were more likely to report chronic conditions and emotional distress, smoking was negatively related to self-rated health and emotional distress, a higher consumption of alcohol was related to better self-rated health, more exercise corresponded with better self-rated health and

Table 7 Logistic regression models, probability of emotional distress Model I B

SIG

Exp(B)

B

SIG

Exp(B)

B

SIG

Exp(B)

0.184

0.022 0.021 0.417 0.009 0.041 0.941 0.002 0.001

0.832

0.186

0.83

0.114

0.245 0.568 0.477 0.004 0.019 0.535 0.19 0.665

1.278 1.765 1.611 1.004 0.981 1.707 1.209 0.514

0.092 0.372 0.316 0.116 0.001 0.693 0.583 0.351

0.238

0.701 0.506 0.033 0.445 1.559 0.212 0.007 0.298 1.285

0.174 0.434 0.768 0.211 0.266 0.709 0.921 0.001 0.026 0.395 0.003 0.001 0.184 0.495 0.02 o.001 0.419 0.044 0.009 0.106

0.892

1.267 2.041 1.723 0.978 0.983 1.828

0.021 0.14 0.404 0.042 0.077 0.989 0.001 0.003 0.424 0.001

0.237 0.713 0.544 0.022 0.017 0.603

1.932

0

0.145 1484 0.04 89.6 0 89.5 42.233(o.001) 0.028 0.052

1.437

0.001

1464 0.05 89.4 0 89.4 52.27 (o.001) 0.036 0.073

1.096 1.451 1.371 0.891 1.001 2 1.792 1.42 0.496 0.603 1.034 1.561 0.21 0.809 0.993 0.742 0.277 1424 0.13 90.1 68.8 89.9 125.22 (o.001) 0.084 0.171

ARTICLE IN PRESS


Model III


Involvement Neighbourhood Kirkendall Downtown Northeast Mountain Age Gender (Female) Education (less than high school) Income (less than $30,000) Smoking Not at all Occasionally Alcohol Exercise Coping with daily demands Coping unexpected difficulties Age Exercise Exercise Income Constant

Model II

2813

2814

Table 8 Logistic regression models, probability of being overweight Model I B

SIG

Exp(B)

B

SIG

Exp(B)

B

SIG

Exp(B)

0.098

0.029 0.01 0.134 0.212 0.068 0.554 0.001 o.001

0.907

0.098

0.906

0.115

0.243 0.139 0.336 0.087 0.012 0.438 0.402 0.461

0.784 0.87 1.399 0.917 1.012 0.645 0.669 1.586

0.215 0.174 0.351 0.073 0.012 0.509 0.482 0.229 0.211 0.098 0.011 0.056 0.043 0.325 0.134 0.838

0.015 0.022 0.256 0.376 0.049 0.681 0.003 0.182 0.002 0.413 0.742 0.442 0.864 0.776 0.112 0.846 0.247 0.027 0.024

0.891

0.763 0.794 1.364 0.903 1.012 0.618

0.031 0.019 0.182 0.463 0.052 0.615 0.002 o.001 0.007 0.002

0.932 0.496 0.751

0.004 0.443 0.093

0.271 0.231 0.311 0.102 0.011 0.481

0.787

o.001

0.455 1403 0.03 65.5 46.2 64.5 46.74 (o.001) 0.033 0.045

0.884

0.003

0.413 1389 0.03 66.2 49.2 64.7 62.05 (o.001) 0.043 0.06

0.807 0.84 1.42 0.929 1.012 1.663 0.617 0.795 0.81 0.907 0.989 0.946 1.044 1.383 0.874 0.433 2.541 1.642 0.472 1356 0.06 67.4 53.8 65.5 101.986 (o.001) 0.072 0.099

ARTICLE IN PRESS


Model III


Involvement Neighbourhood Kirkendall Downtown Northeast Mountain Age Gender (Female) Education (less than high school) Income (less than $30,000) Smoking Not at all Occasionally Alcohol Exercise Coping with daily demands Coping unexpected difficulties Gender Alcohol Gender Coping with unexpected difficulties Smoking Occasionally Income Smoking Daily Income Constant

Model II


more emotional distress, and day-to-day coping skills corresponded with better self-rated health scores and a lower incidence of emotional distress, in all cases after controlling for the other variables of interest.

Conclusion In this article we sought to determine the degree to which the health effects of associational involvement were neighbourhood-specific, bringing together the literature that describes neighbourhood effects on health (but seldom with explicit explanatory mechanisms factored into the empirical analysis) and the literature that describes the health effects of participation in voluntary associations in the civil space (but seldom exploring the degree to which networks and their health effects are geographically situated). We found that overall involvement in voluntary associations had a positive relationship with well-being, especially with self-rated health, emotional distress and body-mass index score, after controlling for age, gender and neighbourhood of residence, although the relationships were weak at best (and, excepting the body mass index score, did not persist after controlling for all of the demographic, socio-economic, health behaviour and coping variables examined in this article). The relationships between involvement and both self-rated health and emotional distress were seemingly explicated in part by health behaviours and coping skills. We speculate that these social networks may facilitate certain behaviours and skills that have a positive effect on wellbeing. This may be because of the presence of the resources inherent to such networks in the face of difficult circumstance: knowledge of such resources may work to promote a sense of mastery or personal locus of control and reduce stress, and ‘‘being assured of one’s worthiness as an individual and a member of a social group sharing similar interests and resources [y] provides emotional support’’ (Lin, 2001). On the other hand, the behavioural codes prescribed by certain social groups may mitigate health-damaging behaviours and encourage health-promoting ones. We emphasize, however, that our statistical models are based upon cross-sectional data and cannot assess causal pathways. We found a neighbourhood effect on health that persisted even after controlling for several demographic and socio-economic characteristics of respondents, especially for the self-rated health and body-mass dependent variables. Still, the neighbourhood of residence and associational involvement relationships with health did not interact with one another, suggesting that the healthrelevant aspects of associational participation were not embedded within these neighbourhoods. In short, our hypotheses pertaining to the variable health effects of

2815

this measure of social capital in (advantaged and disadvantaged) neighbourhoods were not supported by the data. We speculate that either: (i) the associations our respondents belong to span many neighbourhoods, making neighbourhood-specific health effects of associational involvement weak to nonexistent, and/or (ii) associational networks manifest themselves similarly in all neighbourhoods, at least as far as their healthrelevant characteristics are concerned. Our data set did not allow us to discern the degree to which these associations span neighbourhoods. Further research into the spatially situated nature of such networks of association is needed in order to clarify this (non-) finding. This article makes a contribution to the social capital and health literature in part due to the variety of measures of health and the unique nature of our measure of associational involvement. Most studies of this type are limited to one measure of health. We utilized measures of both physical and emotional wellbeing and found that associational involvement and neighbourhood of residence were indeed related to some but not all measures of health. Also, unlike most other studies of social capital, focused solely on breadth of associational involvement (and weighing equally all levels of involvement, ranging from active involvement to merely paying dues), we have combined breadth of associational involvement with depth of involvement in one measure. Future research in this area should seek to disentangle the health effects of breadth and depth of associational involvement. Finally, we comment on other directions for future research. Although social capital is thought by some health researchers to have the potential to explain neighbourhood effects on health, other scholars express doubt in its potential (e.g., Macintyre et al., 2002), the latter preferring that health researchers focus on other attributes of the social, economic and political environment with health implications. While we believe that it is premature to discard social capital from public health’s arsenal of conceptual tools for exploring place effects, we find ourselves attracted to non-Colemanian and nonPutnamian theoretical renderings of social capital. In particular, an approach to social and economic inequalities that treats social capital as a form of power—the ability for persons and groups to carry out their will even when opposed by others—rather than as a means for reaching consensus and accomplishing goals shared by everyone might prove particularly beneficial in public health research. Empirically, Earls’ work on collective efficacy (locals using local social relations to effect change) may be relevant (see Earls & Carlson, 2001; Sampson, Raudenbusch, & Earls, 1997). Theoretically, Bourdieu’s (1984) approach to social and especially economic, educational and cultural capital as means of identifying cultural distinctions and then identifying



social class groupings may be promising, especially given the intriguing similarities between his notion of ‘social space’ and notions of geo-political space utilized in the ‘place and health’ literature.

Acknowledgments The Deconstructing the Local-level Determinants of Health project was funded for 2000–3 by the Social Sciences and Humanities Research Council of Canada through the ‘Society, Culture and the Health of Canadians’ Strategic Theme.

Appendix A Emotional distress: general health questionnaire (GHQ) 1. Now I’d like to know how you’ve been feeling over the past two weeks. First, over the past two weeks have you lost much sleep over worry? /yes, noS [If yes,] would you say more than usual or the same as usual for you? /same as usual, more than usualS 2. Second, over the past two weeks have you felt constantly under stress? 3. Next, over the past two weeks have you felt you couldn’t overcome your difficulties? 4. Over the past two weeks have you been feeling unhappy and depressed? 5. Over the past two weeks have you been losing confidence in yourself? 6. Over the past two weeks have you been thinking of yourself as a worthless person? 7. Over the past two weeks have you been taking things hard? 8. Over the past two weeks have you found everything getting on top of you? 9. Over the past two weeks have you been feeling nervous and tense all the time? 10. Over the past two weeks have you found at times you couldn’t do anything because your nerves were too bad? 11. Over the past two weeks have you felt that you are playing a useful part in things? 12. Over the past two weeks have you felt capable of making decisions about things? 13. Over the past two weeks have you been able to enjoy your normal day-to-day activities? 14. Over the past two weeks have you been able to face up to your problems? 15. Over the past two weeks have you been reasonably happy, all things considered? 16. Over the past two weeks have you been managing to keep yourself busy and occupied?

17. Over the past two weeks have you been getting out of the house as much as usual? 18. Over the past two weeks have you been satisfied with the way you’ve carried out your tasks? 19. Over the past two weeks have you been able to concentrate on whatever you’re doing? 20. Over the past two weeks have you felt on the whole you are doing things well? Chronic conditions 1. Now I’d like to ask about certain health conditions you may have. First, do you have any long-term skin conditions, for example, eczema, that have been diagnosed by a health professional? /yes, noS 2. Do you have hay fever or other allergies that a have been diagnosed by a health professional? 3. Do you have arthritis or rheumatism? 4. Have you been woken by an attack of shortness of breath at any time in the last 12 months? 5. Have you had an asthmatic attack in the last 12 months? 6. Are you currently taking any medication for asthma, including inhalers, aerosols or tablets? 7. Do you have high blood pressure or hypertension? 8. Do you have heart disease? 9. What about diabetes? 10. Urinary problems or kidney disease? 11. A stomach ulcer or ulcers? 12. Do you have other digestive problems? 13. Do you have any type of cancer?

Body mass index 1. How tall are you? /centimetresS 2. How much do you weigh? /kilogramsS

Participation in voluntary associations 1. Please tell me the name of any group or voluntary organization you belong to. 2. How involved are you in the activities and affairs of this group or organization you belong to? /very involved, somewhat involved, not very involved, not at all involvedS 3. Is there any other group or voluntary organization you belong to?

References Baum, F., Bush, R., Modra, C., Murray, C., Cox, E., Alexander, K., & Potter, R. (1999). Epidemiology of participation: an Australian community study. Journal of Epidemiology and Community Health, 54, 414–423.

ARTICLE IN PRESS G. Veenstra et al. / Social Science & Medicine 60 (2005) 2799–2818 Berkman, L., Glass, T., Brissette, I., & Seeman, T. (2000). From social integration to health: Durkheim in the new millennium. Social Science & Medicine, 51, 843–857. Bourdieu, P. (1984). Distinction: a social critique of the judgement of taste. London: Routledge Trans: R. Nice [1979]. Bourdieu, P. (1986). The Forms of Capital. In G. R. John (Ed.), Handbook of theory and research for the sociology of education. New York: Greenwood Press. Cattell, V. (2001). Poor people, poor places, poor health: the mediating role of social networks and social capital. Social Science & Medicine, 52, 1501–1516. Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95–S121. Davey-Smith, G., Hart, C., Watt, G., Hole, D., & Hawthorne, V. (1998). Individual social class, area-based deprivation, cardiovascular disease risk factors, and mortality: the Renfrew and Paisley study. Journal of Epidemiology and Community Health, 52, 399–405. Diez-Roux, A., Nieto, F. J., Muntaner, C., Tyroler, H. A., Comstock, G. W., Shahar, E., Cooper, L. S., Watson, R. L., & Szklo, M. (1997). Neighbourhood Environments and coronary heart disease: a multi-level analysis. American Journal of Epidemiology, 146(1), 48–63. Duncan, C., Jones, K., & Moon, G. (1993). Do places matter? A multi-level analysis of regional variations in healthrelated behaviour in Britain. Social Science & Medicine, 37(6), 725–733. Duncan, C., Jones, K., & Moon, G. (1995). Psychiatric morbidity: a multi-level approach to regional variations in the UK. Journal of Epidemiology and Community Health, 49, 290–295. Duncan, C., Jones, K., & Moon, G. (1998). Context, composition and heterogeneity: using multilevel models in health research. Social Science & Medicine, 46(1), 97–117. Earls, F., & Carlson, M. (2001). The social ecology of child health and well-being. Annual Review of Public Health, 22, 143–166. Ellaway, A., & Macintyre, S. (2000). Social capital and selfrated health: support for a contextual mechanism. Social Science & Medicine, 90, 988. Ellison, G. (1999). Income inequality, social trust, and selfreported health status in high-income countries. Annals New York Academy of Sciences, 896, 325–328. Ford, B. L. (1983). An overview of hot-deck procedures. In W. G. Madow, I. Olkin, & D. B. Rubin (Eds.), Incomplete data in sample surveys (pp. 185–207). New York: Academic Press. Forrest, R., & Kearns, A. (2001). Social cohesion, social capital, and the neighbourhood. Urban Studies, 38(12), 2125–2143. Fukuyama, F. (2000). Social capital. In L. E. Harrison, & P. Samuel (Eds.), Culture matters: how human values shape human progress (pp. 98–111). New York: Basic Books. Goldberg, D. (1972). The detection of psychiatric illness by questionnaire. London: Oxford University Press. Hawe, P., & Shiell, A. (2000). Social capital and health promotion: a review. Social Science & Medicine, 51, 871–885. Helliwell, J., & Putnam, R. (1995). Economic growth and social capital in Italy. Eastern Economic Journal, 21, 295–307.

2817

Kaplan, G. A. (1996). People and places: contrasting perspectives on the association between social class and health. International Journal of Health Services Research, 26(3), 507–519. Kawachi, I., Kennedy, B., Lochner, K., & Prothrow-Stith, D. (1997). Social capital, income inequality, and mortality. American Journal of Public Health, 87, 1491–1498. Kawachi, I., Kennedy, B., & Glass, R. (1999). Social capital and self-rated health: a contextual analysis. American Journal of Public Health, 89(8), 1187–1193. Lin, N. (2001). Building a network theory of social capital. In N. Lin, K. Cook, & R. Burt (Eds.), Social capital: theory and research (pp. 3–30). New York: Aldine de Gruyter. Lochner, K., Kawachi, I., Brennan, R., & Buka, S. (2003). Social capital and neighborhood mortality rates in Chicago. Social Science & Medicine, 56(8), 1797–1805. Luginaah, I., Jerrett, M., Elliott, S., Eyles, J., Parizeau, K., Birch, S., Abernathy, T., Veenstra, G., Hutchinson, B., & Giovis, C. (2001). Health profiles of Hamilton: spatial characterisation of neighbourhoods for health investigations. Geojournal, 53(2), 135–147. Lynch, J., Davey Smith, G., Hillemeier, M., Shaw, M., Raghunathan, T., & Kaplan, G. (2001). Income inequality, the psycho-social environment and health: comparisons of wealthy nations. Lancet, 358(9277), 194–200. Macintyre, S., Ellaway, A., & Cummins, S. (2002). Place effects on health: how can we conceptualise, operationalise, and measure them? Social Science & Medicine, 55, 125–139. Mohan, G., & Mohan, J. (2002). Placing social capital. Progress in Human Geography, 26(2), 191–210. McDowell, I., & Newell, C. (1987). Measuring health: a guide to ratings scales and questionnaires. London: Oxford University Press. Paterson, L., & Goldstein, H. (1992). New statistical methods for analyzing social structures: an introduction to multilevel models. British Educational Research Journal, 17, 387–393. Putnam, R., Leonardi, R., & Nanetti, R. Y. (1993). Making democracy work: civic traditions in modern Italy. New Jersey: Princeton University Press. Putnam, R. (2000). Bowling Alone. The collapse and revival of American community. New York: Simon and Schuster. Rose, R. (2000). How much does social capital add to individual health? A survey study of Russians. Social Science & Medicine, 51, 1421–1435. Rice, T., & Sumberg, A. (1997). Civic culture and government performance in the American states. Publius, 27, 99–114. Sampson, R., Raudenbusch, S., & Earls, F. (1997). Neighbourhoods and 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 UK: combining individual and area characteristics. Journal of Epidemiology and Community Health, 50, 366–376. Sloggett, A., & Joshi, H. (1994). Higher mortality in deprived areas: community or personal disadvantage? British Medical Journal, 309, 1470–1474. Sloggett, A., & Joshi, H. (1998). Deprivation indicators as predictors of life events in 1981–1992 based on the UK ONS longitudinal study. Journal of Epidemiology and Community Health, 52, 228–233.



Subramanian, S., Kim, D., & Kawachi, I. (2002). Social trust and self-rated health in US communities: a multilevel analysis. Journal of Urban Health: Bulletin of the New York Academy of Medicine, 79(4, Supplement 1), S21–S34. Subramanian, S., Lochner, K., & Kawachi, I. (2003). Neighborhood differences in social capital: a compositional artifact or a contextual construct? Health and Place, 9(1), 33–44. Veenstra, G. (2000). Social capital and health: an individuallevel analysis. Social Science & Medicine, 50, 619–629.

Veenstra, G. (2002). Social capital and health (plus wealth, income inequality and regional health governance). Social Science & Medicine, 54, 849–858. Wilkinson, R. (1996). Unhealthy Societies. The afflictions of inequality. London: Routledge. Wilson, M., & Daly, M. (1997). Life expectancy, economic inequality, homicide, and reproductive timing in Chicago neighbourhoods. British Medical Journal, 314, 1271–1274. Woolcock, M. (1998). Social capital and economic development: toward a theoretical synthesis and policy framework. Theory and Society, 27, 151–208.