- Email: [email protected]
Neighborhood disorder and smoking: Findings of a European urban survey
ARTICLE IN PRESS
Social Science & Medicine 63 (2006) 2464–2475 www.elsevier.com/locate/socscimed
Neighborhood disorder and smoking: Findings of a European urban survey Rebecca Miles Florida State University, Tallahassee, FL, USA Available online 14 August 2006
Abstract Using the Large Analysis and Review of European housing and health Status (LARES) survey, this paper investigates the influence of neighborhood physical disorder on smoking behaviors, and the extent to which it is mediated by perceptions of safety. Indicators of physical disorder: litter, graffiti, and the absence of vegetation on facades, balconies or windows, were directly observed by surveyors. The paper also considers whether the place effects on smoking are similar across the 7 European cities in the study. Results indicate that the odds of smoking are 64% higher for those living in an area rated high on neighborhood disorder compared to low. The effect is substantially greater for men than for women with men in areas rated high on disorder showing odds of smoking that are twice as high as those living in areas rated low. The association does not vary by city of residence. Only a small part of the effect of neighborhood disorder is mediated by perceptions of safety. The finding of a substantial neighborhood physical disorder effect on smoking across a range of cities in Europe adds to the evidence suggesting that environmental interventions are worth pursuing in conjunction with other approaches to smoking prevention. r 2006 Elsevier Ltd. All rights reserved. Keywords: Smoking; Neighborhood disorder; Perceptions of safety; Europe; Neighborhoods and health
Introduction Of all the behaviors that affect health, smoking has the most dire consequences. Smoking is an important risk factor for most of the major causes of death and is responsible for 9% of all cancer deaths in women in European Union countries and 37% in men (Peto, Lopez, Boreham, & Thun, 2004; Peto, Lopez, Boreham, Thun, & Heath, 1992, 1994; Rogers, Hummer, Krueger, & Pampel, 2005). Smoking prevalence is on the decline in the general population but socioeconomic differentials Tel.: +1 850 644 7102.
E-mail address: [email protected]. 0277-9536/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2006.06.011
in tobacco use are widening (Cavelaars et al., 2000; Graham & Der, 1999). A number of cross-sectional studies indicate that at least part of the socioeconomic effect may be due to differences in where people live. In several studies, living in a disadvantaged neighborhood has been found to influence smoking behaviors above and beyond individual characteristics such as age, gender, personal education, employment status, household income, and mental health (Diez Roux, 2003; Duncan, Jones, & Moon, 1999; Kleinschmidt, Hills, & Elliot, 1995; Reijneveld, 1998; Ross, 2000). Other dimensions of neighborhood context, social cohesion, and perceived livability, have also been found to be associated with
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
smoking behavior (Blackman, 2005; Patterson, Eberly, Ding, & Hargreaves, 2004). Health researchers and educators typically consider smoking to be a purely individual choice. Smoking prevention and cessation programs also tend to focus on the individual. However, if living in particular housing or environmental conditions predisposes people to smoke, then efforts to persuade people to quit smoking without addressing the predisposing factors are likely to be ineffective (Byrne & Keithley, 1993, p. 49). In this study, I take advantage of a recent survey carried out by the World Health Organization, the Large Analysis and Review of European housing and health Status (LARES) to investigate whether there are similar place effects on smoking across the various European cities included in the survey. In particular, I analyze the influence of neighborhood physical disorder on smoking behaviors, and the extent to which it is mediated by perceptions of safety. I also seek to ascertain whether these relationships are similar for men and women. Neighborhood physical disorder often reflects a lack of social capital and low levels of both informal and formal social control. It may signal criminals that local residents are unable or unwilling to protect social order (Wilson & Kelling, 1982), while at the same time indicating to residents that it is unsafe and unpleasant to walk or interact socially in the streets (Raudenbush, 2003). A strength of this study is that instead of using aggregate measures of residents’ socioeconomic position to represent neighborhood disadvantage as many studies do, it includes direct observations of neighborhood physical disorder. The LARES also provides measures of the full range of individual-level characteristics that are important to control for in order to specify true place effects, including a widely used measure of mental health. A limitation of the LARES is that it does not include characteristics of the wider community or region in which the households are embedded. In the first part of the analysis, I investigate the relationship between living in areas rated high on physical disorder and other measures of quality-oflife such as residential density, perceived noise annoyance, and perceived safety. In the second part of the analysis, I analyze the relationship between neighborhood disorder, perceptions of safety, and smoking behavior, taking into consideration individual-level risk factors. I find that the effects of neighborhood disorder and perceived safety are
2465
different for men and women but not for the different cities. I include an indicator of city of residence in all of the analyses to account for unmeasured city-level influences on smoking such as social and cultural patterns related to the prevalence and maintenance of smoking behaviors. The findings of this study therefore point to relationships that may be broadly applicable to Europe as a whole. Neighborhood disorder and smoking—the literature The theoretical basis for expecting an association between neighborhood disorder and smoking derives from the broader literature on the role of place in shaping people’s health experiences and from a number of studies investigating neighborhood disadvantage and smoking in particular. A substantial set of studies establishes the existence of an effect of neighborhood deprivation on smoking behavior above and beyond the socioeconomic position of the individual (Diez Roux, 2003; Duncan et al., 1999; Kleinschmidt et al., 1995; Reijneveld, 1998; Ross, 2000). Existing studies suggest there are 4 main plausible mechanisms through which such a place-based effect on smoking might be mediated. These include contagion processes, opportunity structures, social capital, and social inequality. Contagion Place-based contagion processes affect health through people being influenced by others around them and copying their behavior (Jencks & Mayer, 1990). Contagion processes also help shape the cultural and normative standards neighborhoods have and that people use in choosing their own behaviors, including those that influence health (Ross, 2000, p. 267). Given the strong inverse relationship between individual socioeconomic position and smoking, it is likely that disadvantaged neighborhoods have a higher prevalence of smoking than advantaged neighborhoods. Diehr et al. (1993) found that American adults in communities where many people smoke are more likely to smoke themselves (p. 267). Smoking behaviors are highly visible in public places and therefore are more likely to spread through place-based exposure than less obvious health-related behaviors such as wearing a seat belt.
ARTICLE IN PRESS 2466
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
Opportunity structures and institutional resources Opportunity structures are ‘‘socially constructed and socially patterned features of the physical and social environment which may promote or damage health either directly, or indirectly through the possibilities they provide for people to live healthy lives’’ (Macintyre, Ellaway, & Cummins, 2002, p. 132). Studies of place effects on smoking have focused on tobacco advertising and availability, and on the extent to which institutions in the immediate neighborhood environment encourage smoking by permitting it on their premises and selling smokingrelated products, or whether they discourage it by forbidding smoking on their premises and having a sign to that effect, raising awareness about the hazards of smoking, having signs indicating they do not sell to minors and a person responsible for enforcing the policy (Frohlich, Potvin, Chabot, & Corin, 2002). More socially advantaged neighborhoods have more smoking discouraging resources but a study of Canadian youth finds that smoking behaviors are not necessarily lower in these; how people use and interact with opportunity structures explains some of the discrepancies between smoking initiation rates and the structural aspects of the communities investigated (Frohlich et al., 2002). Social capital and beyond Social capital has been defined in a variety of ways in the literature; in general, it refers to collective aspects of neighborhood life such as social cohesion (the presence of strong social bonds and mutual trust), spatial diffusion (exchange of information), support networks (based on reciprocal obligations), and informal social control (perception of lack of control over the immediate environment) (Coleman, 1988; Putnam, 1993). And specifically, it refers to ‘‘the resources or potential inherent in social networks’’ (Sampson, 2001). Social capital tends to be thought of as protective of the public’s health and safety. A small set of studies investigating the link between social cohesion and healthrelated outcomes at the state level in the US find that higher levels of social cohesion are associated with lower mortality rates after controlling for individual risk factors (Kawachi, Kennedy, & Glass, 1999; Kawachi, Kennedy, Lochner, & Prothrow Stith, 1997). In an analysis of smoking in particular, Patterson et al. (2004) finds lower levels of smoking in neighborhoods characterized by social cohesion,
after adjusting for area-based poverty and educational levels. More recent conceptualizations of the social processes that operate in neighborhoods find it useful to distinguish social capital from collective efficacy which is ‘‘a task-specific construct that refers to shared expectations and mutual engagement by residents in local social control’’ (Sampson, 2003, p. 138). Areas characterized by concentrated disadvantage, racial segregation, family disruption, and residential instability tend to have low levels of collective efficacy (Sampson et al., 1997, 1999). Theories of social capital also include an institutional component covering ‘‘the resource stock of neighborhood organizations and their linkages with other organizations, both within and outside the community’’ (Sampson, 2001). These institutions determine the neighborhood’s ability to obtain resources from outside the neighborhood such as police services, health services, economic investment, that in turn help achieve social order. Physical disorder is often an indicator of lack of social capital, as it may signal criminals that local residents are unable or unwilling to protect social order (Wilson & Kelling, 1982), while at the same time indicating to residents that it is unsafe and unpleasant to walk or interact socially in streets and public places (Raudenbush, 2003). Neighborhood disorder is associated with low perceptions of safety (Austin et al., 2002; LaGrange et al., 1992; Skogan, 1990), which in turn are linked to higher crime levels. Neighborhood disorder and crime both play an important role in neighborhood decline (Kelling & Coles, 1996; Skogan, 1990; Wilson & Kelling, 1982). Where the opportunity exists, residents with the means move out, leaving behind those with limited options. Those who stay may purchase means to protect themselves; some avoid public facilities such as public transportation, and others stay inside and try to keep their children inside, only going out when they have to (Kelling & Coles, 1996; Ross, 2000). Blackman (2005) finds less smoking in areas characterized by high-perceived livability. Livability, defined by the London Housing Strategy as requiring ‘ysafe, attractive, clean public spaces, free from the fear of crime and anti-social behaviour’ (London Housing Board, 2003, p. 12, cited in Blackman, 2005), is the converse of neighborhood disorder. One plausible explanation for the effect on smoking posits that the stress of living in an unsafe, unpleasant environment that is poorly served by
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
both public and private sectors, and of the social isolation that residents often experience, makes it easier to start smoking and harder to quit; smoking then becomes a coping mechanism, providing a respite from a stressful environment (Acierno et al., 1996; Anda et al., 1990; Stead, MacAskill, MacKintosh, Reece, & Eadie, 2001). Residential segregation and social inequality Neighborhoods are embedded in larger contexts, and therefore the relationship between neighborhood characteristics and health is likely to be affected by the extent of residential segregation or social inequality in the state or country as a whole. Different forces may drive residential segregation in different places: race/ethnicity, social class, age, sex. Social inequality and residential segregation are likely to exacerbate relative deprivation and low social capital, thereby contributing to the negative effect of disadvantaged places on health. The relationship between residential segregation by race/ethnicity and health outcomes has been investigated only for the United States and primarily in relation to the health status of US blacks; these demonstrate that mortality rates for blacks are higher in urban areas with high levels of residential segregation than for blacks in areas with low levels of segregation (see review in Acevedo-Garcia, Lochner, Osypuk, & Subramanian, 2003). The extent to which income inequality in a society affects population health and mortality has been examined in numerous studies and findings have been mixed (see review in Wen, Browning, & Cagney, 2003). As for smoking behaviors, several ecological studies have investigated their association with social inequality cross-nationally and within nations (Barnett, Pearce, & Moon, 2005; Pampel, 2002). Levels of ethnic inequality between groups were found to affect smoking rates over time in the context of New Zealand (Barnett et al., 2005). The mechanisms mediating the relationship, however, are not limited to material deprivation but include the process of diffusion. Cigarette smoking is taken up first by the higher classes then disseminated to the lower classes, at which point it begins to decline among the higher classes. Pampel (2002) documents the fact that cigarette smoking appeared at different times in the countries of Europe, and that the relationship between social status and smoking varies across nations. His findings indicate that a
2467
nation’s timing in the diffusion process consistently influences the relationship between social status and smoking, whereas the level of inequality in society as a whole does not (Pampel, 2002). Study hypotheses In this study, I focus on the association between neighborhood physical disorder and smoking behaviors. Skogan (1990) identified several forms of physical disorder that were closely related to disorderly behaviors such as public drinking and lounging drunks, loitering youths, and corner gangs, drug use, and noisy neighbors. They include graffiti, damage to public spaces, accumulations of rubbish and refuse, and dilapidated and vacant buildings. The LARES includes measures of 2 of these: graffiti and litter. The LARES also includes several indicators of neighborhood greenery. The one that is consistently associated with a lack of litter or graffiti, and with other indicators of qualityof-life, is the voluntary display by residents of vegetation on facades, windows or balconies; it is also an indicator that residents care about their neighborhood environment. In this study, I investigate the relationship between neighborhood disorder and smoking using a composite index based on the presence of litter or graffiti, and the absence of vegetation on facades, windows or balconies. I also investigate the effect of perceived safety and the extent to which it mediates the neighborhood disorder effect. I investigate interactions between neighborhood physical disorder and city of residence, and between disorder and gender. The hypothesis that city of residence moderates the effect of neighborhood disorder on smoking, i.e. that the relationship between physical disorder and smoking varies depending on the city/country of residence, captures the possible effect of social inequality and residential segregation at the macro-level on how people are ‘‘sorted’’ into neighborhoods. The hypothesis that place effects are different for women than for men derives from a number of studies of smoking behaviors, indicating that gender is an important structural factor that helps create and shape people’s perceptions of constraints and opportunities (Frohlich et al., 2002; Graham, 1987, 1994). There is also some evidence that neighborhood-based friendship networks may enhance behaviors that endanger health, especially for men. In a study of African-American men in the US, Majors
ARTICLE IN PRESS 2468
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
and Billson (1992) suggest that disorderly and unsafe environments attract men whose status is subordinate in the hierarchy of their everyday working world and create a highly visible, public space in which they construct a type of compulsive masculinity. The behaviors defining this type of masculinity—smoking, drug and alcohol abuse, fighting, sexual conquests, dominance and crime— are likely to permeate neighborhood-based friendship networks. The findings of existing empirical studies investigating whether the effect of neighborhood disadvantage and disorder on smoking is different for men and women, however, are mixed (Blackman, 2005; Ross, 2000). Data and methods Data source This study is based on the LARES data collected in 8 European cities during the winters of 2001 and 2002. Further information about sampling and data collection procedures can be found in Bonnefoy, Braubach, Davidson, and Robbel (in press). Participating cities included 3 in Eastern Europe: Budapest, Hungary, Vilnius, Lithuania, and Bratislava, Slovakia, and 5 in Western Europe: Forli, Italy, Bonn, Germany, Geneva, Switzerland, Angers, France, and Ferreira do Alentejo, Portugal. Three survey instruments were used to assess housing conditions and their links to health. First, a questionnaire was administered to an informant from each household in a face-to-face interview. It included among other items, the respondent’s perception of safety. Second, surveyors who received the same training in each of the cities directly observed housing conditions and the quality of the local neighborhood, and recorded their observations on an inspection sheet. The presence or absence of litter, graffiti, and greenery were among the environmental characteristics observed near each sampled household. Third, each member of the household was asked to complete a selfadministered health questionnaire. This included questions about health-related behaviors such as smoking as well as descriptive information about the individual’s health. A total of 3373 households were surveyed. They were randomly selected from population registries, except in Angers where the tax registry of the city was used; this means that households were sampled independently and not within clusters such as census
tracts or block groups. A total of 8519 individuals in these households filled out a health questionnaire. In 5 of the cities, response rates were around or above what would be expected for a survey of this type (43.4%, 50.8%, 56%, 63.3%, and 77.3%). In the other 3, response rates were much lower (27.9%, 32%, and 34.2%); the main reasons given for nonparticipation were: ‘‘no time (23%), no interest (18.4%), fear (10.2%), health reasons (8%), holiday (3.4%), other/unknown (36.9%)’’ (Bonnefoy et al., in press, p. 14). When LARES survey data were compared to official data obtained from the cities, some bias relative to age and housing type was found but it was inconsistent across the cities, and no bias was identified relative to gender; because of the sampling method used, however, average household size in the LARES exceeds that in the surveyed cities (Bonnefoy et al., in press, p. 15). Study population In this paper, I focus on urban households because there are greater health differences due to neighborhood deprivation in urban than rural areas (Mays & Chinn, 1989). All households in Ferreira, Portugal (which is more rural than urban), and all rural households in other study sites were excluded. I also exclude individuals under the age of 15 since the earliest age at which smoking is reported in any of the cities is 15. This leaves 2782 households and 5784 individuals in the analysis. Variables included in the models Smoking behavior: Most existing studies use measures of smoking behavior that reflect relatively moderate levels of smoking, mainly whether or not the individual smoked at least 1 cigarette per day or whether the person currently smokes. In order to facilitate comparison between the results of this study and of prior studies, I use an outcome measure of smoking at least 1 cigarette per day. I also ran the models with a measure of somewhat heavier smoking, i.e. 5 or more cigarettes per day; the results were very similar and so are not reported here. Neighborhood disorder: I create an additive index based on 3 indicators of neighborhood disorder: the presence of graffiti and litter, and the absence of vegetation on facades, balconies or windows. Households that score high on the neighborhood disorder index are located in residential environ-
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
ments with either 2 or 3 of these conditions, those that obtain a moderate score have 1 of the 3 conditions, and those with a low score have residential environments with neither litter nor graffiti, and displaying vegetation on facades, balconies or windows. Most existing studies use aggregated individual data on socioeconomic position from the census to measure neighborhood deprivation and therefore use census geography to define neighborhoods, whether tracts in the US, wards in the UK, or boroughs in Holland. In this study, the indicators of neighborhood disorder refer to the local neighborhood immediately surrounding the housing unit or building, and are therefore more precise measures of the environmental exposure of interest. Feelings of safety: To analyze the extent to which the effect of neighborhood disorder is mediated by perceptions of safety, I include responses to the question whether respondents feel very, somewhat or not at all safe returning home in the dark. This question was asked only of household informants and therefore the analysis investigating feelings of safety is based on a subpopulation including 1 adult per household. I examine men and women separately since women have consistently been found to have lower perceptions of safety than men (Perkins & Taylor, 1996; Skogan & Maxfield, 1981; Toseland, 1982), in particular at night (Taylor & Covington, 1993). Individual-level factors associated with smoking: The LARES allows me to take into consideration a full set of individual characteristics associated with smoking including age, gender, marital status, educational achievement, unemployment, household socioeconomic status (SES), housing tenure, and mental health. A measure of household SES was formed using variables deemed to be comparable across the cities: household size and composition, highest education level of any adult in the household, size of dwelling and number of rooms, number of persons employed, proportion of those aged 18–59 in full-time work, number of full-time equivalent jobs held by people in household, and number of people aged 60 or over in household; for the logistic regressions, the SES score was grouped into 5 categories using cutoffs corresponding to percentiles of the overall distribution. To measure mental health, I created an additive index based on the 5 items from the Medical Outcome Study Short Form 36 (MOS SF-36) identified as good indicators of mental health (Lepie`ge, Ecosse, Pouchot, Coste,
2469
& Perneger, 2001). They are ‘‘During the past month, have you:’’ ‘‘yfelt particularly nervous?’’; ‘‘yfelt so down in the dumps nothing could cheer you up?’’; ‘‘yfelt calm and peaceful?’’; ‘‘yfelt downhearted and miserable?’’; ‘‘ybeen happy?’’ The index ranged from a low of 5 to a high of 30. Because the index was highly skewed, cutpoints for 3 equal groups were used to reduce the index to a 3category variable, high, moderate, and low mental health. Statistical analyses I use logistic regression and the odds ratio (OR) as a measure of effect between smoking behavior and neighborhood disorder, perceived safety, and other individual characteristics. Because the elements of the sample were independently selected, and each local neighborhood environment observed is associated with a single household, hierarchical models are not appropriate. I estimate adjusted ORs and 95% confidence intervals in logistic regression analyses, first including only neighborhood disorder and city of residence and then adding other sets of variables. In the analyses based on data from all respondents, I use the Huber/White/sandwich estimator of variance because the observations are not independent within households. In the analyses including perceived safety that are based on data from 1 informant per household, I use the traditional calculation of the variance estimator. Covariates include city of residence, and the individual characteristics age, age-squared, sex, marital status, educational achievement, unemployment, mental health, household SES, housing tenure. A range of statistical interaction terms were tested, including one investigating whether the relationship between neighborhood disorder and smoking is different for men and women, and one analyzing whether it differs across the 7 cities. Only those that were significant are included here. Findings Of the 5784 individuals in the study population, 75.3% report that they do not smoke daily, 23.8% report smoking on a daily basis, and 0.8% did not respond to the question. There were no significant differences between responders and non-responders by age, gender, or educational achievement. However, a significantly higher proportion of nonresponders were separated, divorced, or widowed
ARTICLE IN PRESS 2470
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
(27% compared to 12% of responders) and a significantly smaller proportion was married (48% compared to 63% of responders). Of the 2782 households in the study population, 471 or 17% live in areas rated high on neighborhood disorder and 820 or 30% live in areas with a low rating. As expected, neighborhood physical disorder is significantly inversely associated both with residents’ perception of their area as evaluated by others and with the measure of household SES (Table 1). The mean perceived area evaluation score of 4.0 for areas rated low on neighborhood disorder is significantly higher than the mean of 3.6 in areas rated high; and the mean household SES score of 21.0 for areas with a low rating is significantly higher than the mean of 18.6 for areas with a high rating. Furthermore, a significantly larger proportion of households in areas rated high on physical disorder are located in neighborhoods comprised mainly of large multi-family buildings compared to households in areas rated low (Table 1); this suggests that areas rated high on neighborhood disorder are more likely to have high residential density. Table 1
suggests neighborhood disorder is also associated with other indicators of quality-of-life. A significantly larger proportion of informants in areas rated high on physical disorder report being strongly bothered or annoyed by general traffic noise and by general neighborhood noise. A significantly higher proportion also report not feeling safe returning home at night compared to informants in areas rated low on physical disorder. To show the association between neighborhood disorder and smoking behavior, I display first the results of the logistic regression adjusting only for city-of-residence (Table 2, Model 1). In the second model, I also adjust for age, sex, and marital status and in the third, I add the interaction terms between neighborhood disorder and sex. And finally, in the fourth model, I include the full set of household and individual-level covariates along with the significant interactions (Table 2, Model 4). In Model 1, neighborhood disorder is significantly associated with smoking: living in an area rated moderate compared to low on physical disorder increases the odds of smoking at least 1 cigarette per day 27%, whereas living in an area rated high compared to
Table 1 Perceived quality of residential area, household socioeconomic position, neighborhood type, noise annoyance, and perceptions of safety by residence in a neighborhood rated high vs. low on physical disorder Variables
Mean perceived area evaluation score (range 1 ¼ very bad to 5 ¼ very good) Mean household socioeconomic score (range 6–35) Neighborhood type Mainly detached or semi-detached houses Mainly apartment blocks up to 4 floors Mixed neighborhood Mainly apartment blocks with 5 or more floors Bothered/annoyed by general traffic noise Strongly Moderately Not at all Bothered/annoyed by general neighborhood noise Strongly Moderately Not at all Feel safe returning home at night No To some extent Yes Significant at po0:01.
High neighborhood disorder ðn ¼ 471Þ
Low neighborhood disorder ðn ¼ 820Þ
Total percent (overall mean)
3.6
4.0
(3.9)
18.6
21.0
(20.1)
3% 12% 24% 62%
16% 26% 27% 31%
11% 26% 22% 41%
24% 40% 36%
17% 42% 41%
19% 40% 41%
21% 41% 38%
10% 35% 55%
13% 37% 55%
37% 25% 38%
16% 24% 60%
22% 24% 54%
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
2471
Table 2 Odds ratios associated with the effect of neighborhood disorder, household, and individual-level factors on smoking at least 1 cigarette per day Model 1
Model 2
Model 3
Model 4
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
1.27 1.48
(1.06, 1.51) (1.18, 1.85)
1.24 1.39 0.49 1.12 0.99 1.60 1.27
(1.03, (1.10, (0.44, (1.09, (0.99, (1.28, (1.01,
1.42 1.90 0.65 1.12 0.99 1.61 1.27 0.75 0.51
(1.14, (1.42, (0.52, (1.09, (0.99, (1.29, (1.01, (0.57, (0.36,
1.30 1.60 0.63 1.13 0.99 1.26 1.15 0.79 0.51
(1.03, (1.18, (0.50, (1.10, (0.99, (1.00, (0.91, (0.59, (0.35,
Primary vs. post-secondary 1st sec. vs. post-sec. 2nd sec. vs. post-sec.
1.41 1.78 1.57
(1.09, 1.83) (1.46, 2.17) (1.30, 1.90)
Unemployed vs. not unemployed
2.73
(1.82, 4.10)
Mental health: Moderate vs. lo High vs. low
0.75 0.63
(0.63, 0.89) (0.53, 0.76)
Socioeconomic postion Lowest vs. group 2 Lowest vs. 3 Lowest vs. 4 Lowest vs. highest Rented vs. owned Unemployment sex
0.87 0.90 0.75 0.66 1.74 0.42
(0.66, (0.68, (0.56, (0.48, (1.40, (0.24,
Neighborhood disorder Moderate vs. low High vs. low Female vs. male Age Age-squared Divorced vs. married Single vs. married Moderate disorder sex High disorder sex
Log pseudo-likelihood
3083.8159
2875.2332
1.50) 1.76) 0.56) 1.15) 0.99) 2.00) 1.59)
2869.402
1.79) 2.54) 0.81) 1.15) 0.99) 2.00) 1.60) 0.99) 0.73)
1.65) 2.17) 0.80) 1.16) 0.99) 1.60) 1.46) 1.06) 0.76)
1.14) 1.19) 0.99) 0.90) 2.16) 0.73)
2614.9043
N.B. all models adjusted for city of residence. Significant at po0:05.
low on physical disorder increases the odds 48%. The size of the effect decreases slightly when demographic covariates are introduced into the model (Table 2, Model 2). Model 3 shows that the interaction between neighborhood disorder and sex is statistically significant; men living in areas rated high on physical disorder are disproportionately more likely to smoke daily than women living in such areas (37% of men compared to 18% of women). A similar series of models was estimated to investigate the mediating effect of perceived safety, and is based on a sample including only household informants (Table 3). I display first the results for neighborhood disorder adjusting only for city-ofresidence (Table 3, Model 1). In the second model, I add perceived safety (Table 3, Model 2). The
significant association between neighborhood disorder and smoking is only slightly reduced, suggesting a small mediating effect of perceived safety. Feelings of safety also display a significant effect of their own: feeling somewhat safe returning home at night compared to not at all safe decreases the odds 32%; feeling very safe compared to not at all decreases the odds of smoking 16% but the result does not reach significance. Adding age, sex, and marital status enhances the effect of perceived safety: feeling very or somewhat safe compared to not at all safe decreases the odds of smoking 63% (Table 3, Model 3). In the final model with the full set of household and individual-level covariates and the interaction terms, the interactions between sex and neighborhood disorder and perceived safety are both
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
2472
Table 3 Odds ratios associated with the effect of neighborhood disorder and perceived safety on smoking at least 1 cigarette per day Model 11
Neighborhood disorder Moderate vs. lo High vs. low
Model 21
Model 43
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
1.39 1.94
(1.11, 1.73) (1.45, 2.58)
1.38 1.92
(1.11, 1.72) (1.43, 2.56)
1.32 1.64
(1.05, 1.67) (1.22, 2.22)
1.74 2.60
(1.17, 2.61) (1.56, 4.33)
0.76 0.86
(0.57, 0.99) (0.68, 1.10)
0.61 0.62
(0.46, 0.82) (0.48, 0.81)
0.46 0.44 0.62 0.35 1.55 2.13
(0.26, (0.27, (0.38, (0.19, (0.79, (1.19,
Feels safe Somewhat vs. not at all Very vs. not at all Moderate disorder sex High disorder sex Feels somewhat safe sex Feels very safe sex Log pseudo-likelihood
Model 32
1441.3461
1418.4274
1323.1319
0.81) 0.73) 1.01) 0.66) 3.04) 3.80)
1210.9901
Significant at po0.05. 1
Adjusted for city of residence. Adjusted for city of residence, age, sex, marital status. 3 Adjusted for city of residence, age, sex, marital status, education, unemployment, mental health, household SES, housing tenure, interactions between sex and neighborhood disorder, sex and perceived safety, sex, and unemployment. 2
statistically significant (Table 3, Model 4). The interaction between perceived safety and sex indicates that like neighborhood physical disorder, the effect of perceived safety on smoking is significantly more important for men than for women; men who do not feel safe returning home at night are disproportionately more likely to smoke daily than women who do not feel safe (36% of men compared to 19% of women). Separate logistic regressions by sex indicate that for men, living in an area rated high on physical disorder increases the odds of smoking 2-fold; for women in similar areas, the odds only increase 7% (data not shown). Perceived safety decreases the odds of smoking daily 75% for men and 23% for women (data not shown). As found in other studies, individual characteristics play a major role in explaining differences in smoking behaviors across these 7 cities. A mental health score of high compared to low reduces the odds of smoking by 59%. Unemployment increases the odds more than 2-fold; the significance of the interaction term, however, indicates the effect is much less for women than for men. Other individual-level effects identified in the full model are for the most part consistent with the literature (Table 2). Women have 59% lower odds of smoking at least 1 cigarette per day than men. The significant age-squared term indicates that both older adults
and young people have lower odds than adults. Divorced or separated adults have significantly higher odds of smoking than those who are married or cohabiting (26% higher), as do those with lower levels of educational achievement compared to those with a post-secondary education (41%, 78%, and 57%). Those in the highest socioeconomic group have 52% lower odds of smoking than those in the lowest group, and renters have 74% higher odds of smoking than residents who own their dwelling. Discussion and conclusions The findings of this analysis of smoking behaviors in 7 European cities indicate that smoking is significantly associated with the physical disorder of local neighborhood environments, and that this relationship is not significantly different in size or direction across the cities in the study. The measure of disorder is based on direct observations by trained surveyors of indicators of neighborhood physical disorder: the presence of graffiti and litter, and the absence of vegetation on facades, balconies, or windows. Across the cities in the study, places rated high on physical disorder are more likely to be located in neighborhoods with high residential density, compared to areas rated low. Furthermore, the residents of such places are more likely to report
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
not feeling safe returning home at night, and being bothered by traffic and general neighborhood noise, suggesting a lower quality-of-life in neighborhoods rated high on physical disorder. The effect of neighborhood physical disorder on smoking behaviors is substantial, ranging from a main effect of 64% to a 2-fold increase when men are analyzed separately. Only a small part of the effect is mediated by perceived safety. The finding that the effect of neighborhood disorder on smoking is significantly less important for women than for men is consistent with a study of a population in Illinois, USA, indicating that neighborhood poverty affects men’s smoking but not women’s (Ross, 2000). It is inconsistent, however, with studies by Duncan et al. (1999) and Kleinschmidt et al. (1995) who do not find a significantly different association between neighborhood deprivation and smoking in UK populations for men and women. Why neighborhood physical disorder might affect men and women’s smoking differently is not investigated here but clues can be found in studies focusing on the behavior of low-status men. These suggest that neighborhood physical disorder signals an inability on the part of residents to maintain social order either informally or through their influence on formal services. Areas with disorderly and unsafe environments attract men whose status is subordinate in the hierarchy of their everyday working world; they provide a public space within which low-status men congregate and display the behaviors such as smoking which define a certain type of masculinity (Majors & Billson, 1992). These behaviors are likely to permeate neighborhoodbased friendship networks and spread through them. The fact that there were consistent links between neighborhood disorder and smoking across the different cities suggests that macro-level forces such as residential segregation and social inequality do not affect the relationship between neighborhood disorder and smoking. This does not mean, however, that they do not have an effect on smoking behaviors independent of local place effects. The finding of a substantial effect of neighborhood physical disorder on smoking across a range of cities in Europe adds to the evidence suggesting that environmental interventions are worth pursuing in conjunction with other approaches to smoking prevention. Preliminary evidence indicates that neighborhood improvement influences smoking in a study based on a combined longitudinal and
2473
cross-sectional survey of the health of residents before and after a neighborhood renewal program in the city of Newcastle, England. Blackman and Harvey (2001) identified a 44% decline over 5 years in the prevalence of smoking following neighborhood improvements. Other studies are needed to corroborate this finding. Improving disadvantaged neighborhoods is worth doing whether or not it reduces smoking prevalence or other health risks. However, if neighborhood improvement programs can be shown to have significant health effects among other quality-of-life benefits, it would strengthen the case for greater resource allocation to such efforts. Building the evidence base for this will require analyses that go beyond the cross-sectional ones featured here. There is enough evidence in the growing body of cross-sectional studies however to indicate that it is well worth the effort of evaluating the health effects of existing neighborhood improvement policies or other housing policies to see if smoking is affected. It may be that how changes in the physical environment are brought about determines whether or not health risks are reduced. Neighborhood improvement programs involving the residents themselves may have a better chance of influencing behaviors like smoking. They yield a sense of collective efficacy in addition to increased satisfaction with the local physical environment and increased perceived safety and decreased stress. Kelling and Coles (1996) tell of several decaying communities in which residents took it upon themselves to improve their situation, in collaboration with law enforcement; among their main efforts were to clean up graffiti before it sends off negative signals to others, and to clean up the litter and refuse in streets and alleys. Crime decreased and quality-of-life increased in those communities following their efforts. Further studies are needed to investigate whether behaviors that threaten health such as smoking, are also affected.
Acknowledgments I thank Xavier Bonnefoy, Matthias Braubach and members of the WHO LARES study team for the survey data and their assistance in many ways, Donald Lloyd for his statistical advice, and Richard Rogers, Charles Nam, and anonymous reviewers for helpful comments on earlier drafts.
ARTICLE IN PRESS 2474
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
References Acevedo-Garcia, D., Lochner, K. A., Osypuk, T. L., & Subramanian, S. V. (2003). Future directions in residential segregation and health research: A multilevel approach. American Journal of Public Health, 93(2), 215–221. Acierno, R. A., Kilpatrick, D. G., Resnick, H. S., Saunders, B. E., & Best, C. L. (1996). Violent assault, posttraumatic stress disorder, and depression: Risk factors for cigarette use among adult women. Behavior Modification, 20(4), 363–384. Anda, R. F., Williamson, D. F., Escobedo, L. G., Mast, E. E., Giovino, G. A., & Remington, P. L. (1990). Depression and the dynamics of smoking. A national perspective. Journal of the American Medical Association, 264(12), 1541–1545. Austin, D. M., Furr, L. A., & Spine, M. (2002). The effects of neighborhood conditions on perceptions of safety. Journal of Criminal Justice, 30(5), 417. Barnett, R., Pearce, J., & Moon, G. (2005). Does social inequality matter? Changing ethnic socio-economic disparities and Maori smoking in New Zealand, 1981–1996. Social Science & Medicine, 60(7), 1515–1526. Blackman, T. (2005). Smoking, worklessness, low income and liveability: joining up smoking cessation and neighbourhood renewal programmes. School of Applied Social Sciences, Durham University, unpublished paper. Blackman, T., & Harvey, J. (2001). Housing renewal and mental health: A case study. Journal of Mental Health, 19(5), 571–580. Bonnefoy, X., Braubach, M., Davidson, M., & Robbel, N. (in press). A pan-European housing and health survey: Description and evaluation of methods and approaches. International Journal of Environment and Pollution. Byrne, D., & Keithley, J. (1993). Housing and the health of the community. In R. Burridge, & D. Ormandy (Eds.), Unhealthy housing: Research, remedies and reform (pp. 41–66). London, Glasgow, New York, Tokyo, Melbourne, Madras: E & FN SPON. Cavelaars, A. E. J. M., Kunst, A. E., Geurts, J. J. M., Crialesi, R., Grotvedt, L., Helmert, U., et al. (2000). Educational differences in smoking: International comparison. British Medical Journal, 320(7242), 1102–1107. Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95–S120. Diehr, P., Koepsell, T., Cheadle, A., Psaty, B. M., Wagner, E., & Curry, S. (1993). Do communities differ in health behaviors? Journal of Clinical Epidemiology, 46, 1141–1149. Diez Roux, A. V. (2003). Area characteristics, individual-level socioeconomic indicators, and smoking in young adults: The CARDIA study. American Journal of Epidemiology, 157(4), 315–326. Duncan, C., Jones, K., & Moon, G. (1999). Smoking and deprivation: Are there neighbourhood effects? Social Science & Medicine, 48, 497–505. Frohlich, K. L., Potvin, L., Chabot, P., & Corin, E. (2002). A theoretical and empirical analysis of context: Neighbourhoods, smoking and youth. Social Science & Medicine, 54(9), 1401–1417. Graham, H. (1987). Women’s smoking and family health. Social Science & Medicine, 25(1), 47–56. Graham, H. (1994). Gender and class as dimensions of smoking behaviour in Britain: Insights from a survey of mothers. Social Science & Medicine, 38(5), 691–698.
Graham, H., & Der, G. (1999). Influences of women’s smoking status: The contribution of socioeconomic status in adolescence and adulthood. European Journal of Public Health, 9, 137–141. Jencks, C., & Mayer, S. E. (1990). The social consequences of growing up in a poor neighborhood. In L. E. Lynn, & M. F. H. McGeary (Eds.), Inner-city poverty in the United States (pp. 111–186). Washington, DC: National Academy Press. Kawachi, I., Kennedy, B. P. E., & Glass, R. (1999). Social capital and self-rated health: A contextual analysis. American Journal of Public Health, 89(8), 1187–1193. Kawachi, I., Kennedy, B. P. E., Lochner, K. S. M., & Prothrow Stith, D. (1997). Social capital, income inequality, and mortality. American Journal of Public Health, 87(9), 1491–1498. Kelling, G. L., & Coles, C. M. (1996). Fixing broken windows: Restoring order and reducing crime in our communities. New York, NY: The Free Press. Kleinschmidt, I., Hills, M., & Elliot, P. (1995). Smoking behaviour can be predicted by neighbourhood deprivation measures. Journal of Epidemiology and Community Health, 49(Suppl. 2), S72–S77. LaGrange, R. L., Ferraro, K. F., & Supancic, M. (1992). Perceived risk and fear of crime: Role of social and physical incivilities. Journal of Research in Crime and Delinquency, 29, 311–334. Lepie`ge, A., Ecosse, E., Pouchot, J., Coste, J., Perneger, T. (2001). Le Questionnaire MOS SF-36: Manuel de l’utilisateur et guide d’interpre´tation des scores. Editions ESTEM. Macintyre, S., Ellaway, A., & Cummins, S. (2002). Place effects on health: How can we conceptualise, operationalise and measure them? Social Science & Medicine, 55(1), 125–139. Majors, R., & Billson, J. M. (1992). Cool pose: The dilemmas of black manhood in America. New York, NY: Lexington Books. Mays, N., & Chinn, S. (1989). Relation between all-cause standardized mortality ratios and two indices of deprivation at regional and district level in England. Journal of Epidemiology and Community Health, 43, 191–199. Pampel, F. C. (2002). Inequality, diffusion, and the status gradient in smoking. Social Problems, 49(1), 35–57. Patterson, J. M., Eberly, L. E., Ding, Y., & Hargreaves, M. (2004). Associations of smoking prevalence with individual and area level social cohesion. Journal of Epidemiology and Community Health, 58(8), 692–697. Perkins, D. G., & Taylor, R. B. (1996). Ecological assessments of community disorder: Their relationship to fear of crime and theoretical implications. American Journal of Community Psychology, 24, 63–107. Peto, R., Lopez, A. D., Boreham, J., & Thun, M. (2004). Mortality from smoking in developed countries 1950– 2000 (2nd ed.). Oxford: Oxford University. Available online at /http:// www.ctsu.ox.ac.uk/tobaccoS, accessed 9/24/05. Peto, R., Lopez, A. D., Boreham, J., Thun, M., & Heath, C., Jr. (1992). Mortality from tobacco in developed countries: Indirect estimation from national vital statistics. Lancet(339), 1268–1278. Peto, R., Lopez, A. D., Boreham, J., Thun, M., & Heath, C., Jr. (1994). Mortality from smoking in developed countries 1950– 2000. Oxford: Oxford University Press. Putnam, R. (1993). The prosperous community: Social capital and community life. American Prospect, 35, 42.
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475 Raudenbush, S. W. (2003). The quantitative assessment of neighborhood social environments. In I. Kawachi, & L. F. Berkman (Eds.), Neighborhoods and health (pp. 112–131). New York, NY: Oxford University Press. Reijneveld, S. A. (1998). The impact of individual and area characteristics on urban socioeconomic differences in health and smoking. International Journal of Epidemiology, 27, 33–40. Rogers, R., Hummer, R. A., Krueger, P. M., & Pampel, F. C. (2005). Mortality attributable to cigarette smoking in the United States. Population and Development Review, 31(2), 259–292. Ross, C. E. (2000). Walking, exercising, and smoking: Does neighborhood matter? Social Science & Medicine, 51, 265–274. Sampson, R. J. (2001). How do communities undergird or undermine human development? Relevant contexts and social mechanisms. In A. Booth, & N. Crouter (Eds.), Does it take a village? Community effects on children, adolescents, and families. Mahwah, NJ: Lawrence Erlbaum. Sampson, R. J. (2003). Neighborhood-level context and health: Lessons from sociology. In I. Kawachi, & L. F. Berkman (Eds.), Neighborhoods and health (pp. 132–146). New York, NY: Oxford University Press.
2475
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277(5328), 918–924. Sampson, R. J., Morenoff, J., & Earls, F. (1999). Beyond social capital: Spatial dynamics of collective efficacy for children. American Sociological Review, 64, 633–660. Skogan, W. G. (1990). Disorder and decline. New York: Free Press. Skogan, W. G., & Maxfield, M. G. (1981). Coping with crime. Newbury Park, CA: Sage. Stead, M., MacAskill, S., MacKintosh, A., Reece, J., & Eadie, D. (2001). It’s as if you’re locked in: Qualitative explanations for area effects on smoking in disadvantaged communities. Health & Place, 7(4), 333–343. Taylor, R., & Covington, J. (1993). Community structural change and fear of crime. Social Problems, 40, 374–395. Toseland, R. W. (1982). Fear of crime: Who is most vulnerable? Journal of Criminal Justice, 10, 199–209. Wen, M., Browning, C. R., & Cagney, K. A. (2003). Poverty, affluence, and income inequality: Neighborhood economic structure and its implications for health. Social Science & Medicine, 57(5), 843–860. Wilson, J. Q., & Kelling, G. (1982). Broken windows: The police and neighborhood safety. Atlantic Monthly.
Social Science & Medicine 63 (2006) 2464–2475 www.elsevier.com/locate/socscimed
Neighborhood disorder and smoking: Findings of a European urban survey Rebecca Miles Florida State University, Tallahassee, FL, USA Available online 14 August 2006
Abstract Using the Large Analysis and Review of European housing and health Status (LARES) survey, this paper investigates the influence of neighborhood physical disorder on smoking behaviors, and the extent to which it is mediated by perceptions of safety. Indicators of physical disorder: litter, graffiti, and the absence of vegetation on facades, balconies or windows, were directly observed by surveyors. The paper also considers whether the place effects on smoking are similar across the 7 European cities in the study. Results indicate that the odds of smoking are 64% higher for those living in an area rated high on neighborhood disorder compared to low. The effect is substantially greater for men than for women with men in areas rated high on disorder showing odds of smoking that are twice as high as those living in areas rated low. The association does not vary by city of residence. Only a small part of the effect of neighborhood disorder is mediated by perceptions of safety. The finding of a substantial neighborhood physical disorder effect on smoking across a range of cities in Europe adds to the evidence suggesting that environmental interventions are worth pursuing in conjunction with other approaches to smoking prevention. r 2006 Elsevier Ltd. All rights reserved. Keywords: Smoking; Neighborhood disorder; Perceptions of safety; Europe; Neighborhoods and health
Introduction Of all the behaviors that affect health, smoking has the most dire consequences. Smoking is an important risk factor for most of the major causes of death and is responsible for 9% of all cancer deaths in women in European Union countries and 37% in men (Peto, Lopez, Boreham, & Thun, 2004; Peto, Lopez, Boreham, Thun, & Heath, 1992, 1994; Rogers, Hummer, Krueger, & Pampel, 2005). Smoking prevalence is on the decline in the general population but socioeconomic differentials Tel.: +1 850 644 7102.
E-mail address: [email protected]. 0277-9536/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2006.06.011
in tobacco use are widening (Cavelaars et al., 2000; Graham & Der, 1999). A number of cross-sectional studies indicate that at least part of the socioeconomic effect may be due to differences in where people live. In several studies, living in a disadvantaged neighborhood has been found to influence smoking behaviors above and beyond individual characteristics such as age, gender, personal education, employment status, household income, and mental health (Diez Roux, 2003; Duncan, Jones, & Moon, 1999; Kleinschmidt, Hills, & Elliot, 1995; Reijneveld, 1998; Ross, 2000). Other dimensions of neighborhood context, social cohesion, and perceived livability, have also been found to be associated with
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
smoking behavior (Blackman, 2005; Patterson, Eberly, Ding, & Hargreaves, 2004). Health researchers and educators typically consider smoking to be a purely individual choice. Smoking prevention and cessation programs also tend to focus on the individual. However, if living in particular housing or environmental conditions predisposes people to smoke, then efforts to persuade people to quit smoking without addressing the predisposing factors are likely to be ineffective (Byrne & Keithley, 1993, p. 49). In this study, I take advantage of a recent survey carried out by the World Health Organization, the Large Analysis and Review of European housing and health Status (LARES) to investigate whether there are similar place effects on smoking across the various European cities included in the survey. In particular, I analyze the influence of neighborhood physical disorder on smoking behaviors, and the extent to which it is mediated by perceptions of safety. I also seek to ascertain whether these relationships are similar for men and women. Neighborhood physical disorder often reflects a lack of social capital and low levels of both informal and formal social control. It may signal criminals that local residents are unable or unwilling to protect social order (Wilson & Kelling, 1982), while at the same time indicating to residents that it is unsafe and unpleasant to walk or interact socially in the streets (Raudenbush, 2003). A strength of this study is that instead of using aggregate measures of residents’ socioeconomic position to represent neighborhood disadvantage as many studies do, it includes direct observations of neighborhood physical disorder. The LARES also provides measures of the full range of individual-level characteristics that are important to control for in order to specify true place effects, including a widely used measure of mental health. A limitation of the LARES is that it does not include characteristics of the wider community or region in which the households are embedded. In the first part of the analysis, I investigate the relationship between living in areas rated high on physical disorder and other measures of quality-oflife such as residential density, perceived noise annoyance, and perceived safety. In the second part of the analysis, I analyze the relationship between neighborhood disorder, perceptions of safety, and smoking behavior, taking into consideration individual-level risk factors. I find that the effects of neighborhood disorder and perceived safety are
2465
different for men and women but not for the different cities. I include an indicator of city of residence in all of the analyses to account for unmeasured city-level influences on smoking such as social and cultural patterns related to the prevalence and maintenance of smoking behaviors. The findings of this study therefore point to relationships that may be broadly applicable to Europe as a whole. Neighborhood disorder and smoking—the literature The theoretical basis for expecting an association between neighborhood disorder and smoking derives from the broader literature on the role of place in shaping people’s health experiences and from a number of studies investigating neighborhood disadvantage and smoking in particular. A substantial set of studies establishes the existence of an effect of neighborhood deprivation on smoking behavior above and beyond the socioeconomic position of the individual (Diez Roux, 2003; Duncan et al., 1999; Kleinschmidt et al., 1995; Reijneveld, 1998; Ross, 2000). Existing studies suggest there are 4 main plausible mechanisms through which such a place-based effect on smoking might be mediated. These include contagion processes, opportunity structures, social capital, and social inequality. Contagion Place-based contagion processes affect health through people being influenced by others around them and copying their behavior (Jencks & Mayer, 1990). Contagion processes also help shape the cultural and normative standards neighborhoods have and that people use in choosing their own behaviors, including those that influence health (Ross, 2000, p. 267). Given the strong inverse relationship between individual socioeconomic position and smoking, it is likely that disadvantaged neighborhoods have a higher prevalence of smoking than advantaged neighborhoods. Diehr et al. (1993) found that American adults in communities where many people smoke are more likely to smoke themselves (p. 267). Smoking behaviors are highly visible in public places and therefore are more likely to spread through place-based exposure than less obvious health-related behaviors such as wearing a seat belt.
ARTICLE IN PRESS 2466
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
Opportunity structures and institutional resources Opportunity structures are ‘‘socially constructed and socially patterned features of the physical and social environment which may promote or damage health either directly, or indirectly through the possibilities they provide for people to live healthy lives’’ (Macintyre, Ellaway, & Cummins, 2002, p. 132). Studies of place effects on smoking have focused on tobacco advertising and availability, and on the extent to which institutions in the immediate neighborhood environment encourage smoking by permitting it on their premises and selling smokingrelated products, or whether they discourage it by forbidding smoking on their premises and having a sign to that effect, raising awareness about the hazards of smoking, having signs indicating they do not sell to minors and a person responsible for enforcing the policy (Frohlich, Potvin, Chabot, & Corin, 2002). More socially advantaged neighborhoods have more smoking discouraging resources but a study of Canadian youth finds that smoking behaviors are not necessarily lower in these; how people use and interact with opportunity structures explains some of the discrepancies between smoking initiation rates and the structural aspects of the communities investigated (Frohlich et al., 2002). Social capital and beyond Social capital has been defined in a variety of ways in the literature; in general, it refers to collective aspects of neighborhood life such as social cohesion (the presence of strong social bonds and mutual trust), spatial diffusion (exchange of information), support networks (based on reciprocal obligations), and informal social control (perception of lack of control over the immediate environment) (Coleman, 1988; Putnam, 1993). And specifically, it refers to ‘‘the resources or potential inherent in social networks’’ (Sampson, 2001). Social capital tends to be thought of as protective of the public’s health and safety. A small set of studies investigating the link between social cohesion and healthrelated outcomes at the state level in the US find that higher levels of social cohesion are associated with lower mortality rates after controlling for individual risk factors (Kawachi, Kennedy, & Glass, 1999; Kawachi, Kennedy, Lochner, & Prothrow Stith, 1997). In an analysis of smoking in particular, Patterson et al. (2004) finds lower levels of smoking in neighborhoods characterized by social cohesion,
after adjusting for area-based poverty and educational levels. More recent conceptualizations of the social processes that operate in neighborhoods find it useful to distinguish social capital from collective efficacy which is ‘‘a task-specific construct that refers to shared expectations and mutual engagement by residents in local social control’’ (Sampson, 2003, p. 138). Areas characterized by concentrated disadvantage, racial segregation, family disruption, and residential instability tend to have low levels of collective efficacy (Sampson et al., 1997, 1999). Theories of social capital also include an institutional component covering ‘‘the resource stock of neighborhood organizations and their linkages with other organizations, both within and outside the community’’ (Sampson, 2001). These institutions determine the neighborhood’s ability to obtain resources from outside the neighborhood such as police services, health services, economic investment, that in turn help achieve social order. Physical disorder is often an indicator of lack of social capital, as it may signal criminals that local residents are unable or unwilling to protect social order (Wilson & Kelling, 1982), while at the same time indicating to residents that it is unsafe and unpleasant to walk or interact socially in streets and public places (Raudenbush, 2003). Neighborhood disorder is associated with low perceptions of safety (Austin et al., 2002; LaGrange et al., 1992; Skogan, 1990), which in turn are linked to higher crime levels. Neighborhood disorder and crime both play an important role in neighborhood decline (Kelling & Coles, 1996; Skogan, 1990; Wilson & Kelling, 1982). Where the opportunity exists, residents with the means move out, leaving behind those with limited options. Those who stay may purchase means to protect themselves; some avoid public facilities such as public transportation, and others stay inside and try to keep their children inside, only going out when they have to (Kelling & Coles, 1996; Ross, 2000). Blackman (2005) finds less smoking in areas characterized by high-perceived livability. Livability, defined by the London Housing Strategy as requiring ‘ysafe, attractive, clean public spaces, free from the fear of crime and anti-social behaviour’ (London Housing Board, 2003, p. 12, cited in Blackman, 2005), is the converse of neighborhood disorder. One plausible explanation for the effect on smoking posits that the stress of living in an unsafe, unpleasant environment that is poorly served by
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
both public and private sectors, and of the social isolation that residents often experience, makes it easier to start smoking and harder to quit; smoking then becomes a coping mechanism, providing a respite from a stressful environment (Acierno et al., 1996; Anda et al., 1990; Stead, MacAskill, MacKintosh, Reece, & Eadie, 2001). Residential segregation and social inequality Neighborhoods are embedded in larger contexts, and therefore the relationship between neighborhood characteristics and health is likely to be affected by the extent of residential segregation or social inequality in the state or country as a whole. Different forces may drive residential segregation in different places: race/ethnicity, social class, age, sex. Social inequality and residential segregation are likely to exacerbate relative deprivation and low social capital, thereby contributing to the negative effect of disadvantaged places on health. The relationship between residential segregation by race/ethnicity and health outcomes has been investigated only for the United States and primarily in relation to the health status of US blacks; these demonstrate that mortality rates for blacks are higher in urban areas with high levels of residential segregation than for blacks in areas with low levels of segregation (see review in Acevedo-Garcia, Lochner, Osypuk, & Subramanian, 2003). The extent to which income inequality in a society affects population health and mortality has been examined in numerous studies and findings have been mixed (see review in Wen, Browning, & Cagney, 2003). As for smoking behaviors, several ecological studies have investigated their association with social inequality cross-nationally and within nations (Barnett, Pearce, & Moon, 2005; Pampel, 2002). Levels of ethnic inequality between groups were found to affect smoking rates over time in the context of New Zealand (Barnett et al., 2005). The mechanisms mediating the relationship, however, are not limited to material deprivation but include the process of diffusion. Cigarette smoking is taken up first by the higher classes then disseminated to the lower classes, at which point it begins to decline among the higher classes. Pampel (2002) documents the fact that cigarette smoking appeared at different times in the countries of Europe, and that the relationship between social status and smoking varies across nations. His findings indicate that a
2467
nation’s timing in the diffusion process consistently influences the relationship between social status and smoking, whereas the level of inequality in society as a whole does not (Pampel, 2002). Study hypotheses In this study, I focus on the association between neighborhood physical disorder and smoking behaviors. Skogan (1990) identified several forms of physical disorder that were closely related to disorderly behaviors such as public drinking and lounging drunks, loitering youths, and corner gangs, drug use, and noisy neighbors. They include graffiti, damage to public spaces, accumulations of rubbish and refuse, and dilapidated and vacant buildings. The LARES includes measures of 2 of these: graffiti and litter. The LARES also includes several indicators of neighborhood greenery. The one that is consistently associated with a lack of litter or graffiti, and with other indicators of qualityof-life, is the voluntary display by residents of vegetation on facades, windows or balconies; it is also an indicator that residents care about their neighborhood environment. In this study, I investigate the relationship between neighborhood disorder and smoking using a composite index based on the presence of litter or graffiti, and the absence of vegetation on facades, windows or balconies. I also investigate the effect of perceived safety and the extent to which it mediates the neighborhood disorder effect. I investigate interactions between neighborhood physical disorder and city of residence, and between disorder and gender. The hypothesis that city of residence moderates the effect of neighborhood disorder on smoking, i.e. that the relationship between physical disorder and smoking varies depending on the city/country of residence, captures the possible effect of social inequality and residential segregation at the macro-level on how people are ‘‘sorted’’ into neighborhoods. The hypothesis that place effects are different for women than for men derives from a number of studies of smoking behaviors, indicating that gender is an important structural factor that helps create and shape people’s perceptions of constraints and opportunities (Frohlich et al., 2002; Graham, 1987, 1994). There is also some evidence that neighborhood-based friendship networks may enhance behaviors that endanger health, especially for men. In a study of African-American men in the US, Majors
ARTICLE IN PRESS 2468
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
and Billson (1992) suggest that disorderly and unsafe environments attract men whose status is subordinate in the hierarchy of their everyday working world and create a highly visible, public space in which they construct a type of compulsive masculinity. The behaviors defining this type of masculinity—smoking, drug and alcohol abuse, fighting, sexual conquests, dominance and crime— are likely to permeate neighborhood-based friendship networks. The findings of existing empirical studies investigating whether the effect of neighborhood disadvantage and disorder on smoking is different for men and women, however, are mixed (Blackman, 2005; Ross, 2000). Data and methods Data source This study is based on the LARES data collected in 8 European cities during the winters of 2001 and 2002. Further information about sampling and data collection procedures can be found in Bonnefoy, Braubach, Davidson, and Robbel (in press). Participating cities included 3 in Eastern Europe: Budapest, Hungary, Vilnius, Lithuania, and Bratislava, Slovakia, and 5 in Western Europe: Forli, Italy, Bonn, Germany, Geneva, Switzerland, Angers, France, and Ferreira do Alentejo, Portugal. Three survey instruments were used to assess housing conditions and their links to health. First, a questionnaire was administered to an informant from each household in a face-to-face interview. It included among other items, the respondent’s perception of safety. Second, surveyors who received the same training in each of the cities directly observed housing conditions and the quality of the local neighborhood, and recorded their observations on an inspection sheet. The presence or absence of litter, graffiti, and greenery were among the environmental characteristics observed near each sampled household. Third, each member of the household was asked to complete a selfadministered health questionnaire. This included questions about health-related behaviors such as smoking as well as descriptive information about the individual’s health. A total of 3373 households were surveyed. They were randomly selected from population registries, except in Angers where the tax registry of the city was used; this means that households were sampled independently and not within clusters such as census
tracts or block groups. A total of 8519 individuals in these households filled out a health questionnaire. In 5 of the cities, response rates were around or above what would be expected for a survey of this type (43.4%, 50.8%, 56%, 63.3%, and 77.3%). In the other 3, response rates were much lower (27.9%, 32%, and 34.2%); the main reasons given for nonparticipation were: ‘‘no time (23%), no interest (18.4%), fear (10.2%), health reasons (8%), holiday (3.4%), other/unknown (36.9%)’’ (Bonnefoy et al., in press, p. 14). When LARES survey data were compared to official data obtained from the cities, some bias relative to age and housing type was found but it was inconsistent across the cities, and no bias was identified relative to gender; because of the sampling method used, however, average household size in the LARES exceeds that in the surveyed cities (Bonnefoy et al., in press, p. 15). Study population In this paper, I focus on urban households because there are greater health differences due to neighborhood deprivation in urban than rural areas (Mays & Chinn, 1989). All households in Ferreira, Portugal (which is more rural than urban), and all rural households in other study sites were excluded. I also exclude individuals under the age of 15 since the earliest age at which smoking is reported in any of the cities is 15. This leaves 2782 households and 5784 individuals in the analysis. Variables included in the models Smoking behavior: Most existing studies use measures of smoking behavior that reflect relatively moderate levels of smoking, mainly whether or not the individual smoked at least 1 cigarette per day or whether the person currently smokes. In order to facilitate comparison between the results of this study and of prior studies, I use an outcome measure of smoking at least 1 cigarette per day. I also ran the models with a measure of somewhat heavier smoking, i.e. 5 or more cigarettes per day; the results were very similar and so are not reported here. Neighborhood disorder: I create an additive index based on 3 indicators of neighborhood disorder: the presence of graffiti and litter, and the absence of vegetation on facades, balconies or windows. Households that score high on the neighborhood disorder index are located in residential environ-
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
ments with either 2 or 3 of these conditions, those that obtain a moderate score have 1 of the 3 conditions, and those with a low score have residential environments with neither litter nor graffiti, and displaying vegetation on facades, balconies or windows. Most existing studies use aggregated individual data on socioeconomic position from the census to measure neighborhood deprivation and therefore use census geography to define neighborhoods, whether tracts in the US, wards in the UK, or boroughs in Holland. In this study, the indicators of neighborhood disorder refer to the local neighborhood immediately surrounding the housing unit or building, and are therefore more precise measures of the environmental exposure of interest. Feelings of safety: To analyze the extent to which the effect of neighborhood disorder is mediated by perceptions of safety, I include responses to the question whether respondents feel very, somewhat or not at all safe returning home in the dark. This question was asked only of household informants and therefore the analysis investigating feelings of safety is based on a subpopulation including 1 adult per household. I examine men and women separately since women have consistently been found to have lower perceptions of safety than men (Perkins & Taylor, 1996; Skogan & Maxfield, 1981; Toseland, 1982), in particular at night (Taylor & Covington, 1993). Individual-level factors associated with smoking: The LARES allows me to take into consideration a full set of individual characteristics associated with smoking including age, gender, marital status, educational achievement, unemployment, household socioeconomic status (SES), housing tenure, and mental health. A measure of household SES was formed using variables deemed to be comparable across the cities: household size and composition, highest education level of any adult in the household, size of dwelling and number of rooms, number of persons employed, proportion of those aged 18–59 in full-time work, number of full-time equivalent jobs held by people in household, and number of people aged 60 or over in household; for the logistic regressions, the SES score was grouped into 5 categories using cutoffs corresponding to percentiles of the overall distribution. To measure mental health, I created an additive index based on the 5 items from the Medical Outcome Study Short Form 36 (MOS SF-36) identified as good indicators of mental health (Lepie`ge, Ecosse, Pouchot, Coste,
2469
& Perneger, 2001). They are ‘‘During the past month, have you:’’ ‘‘yfelt particularly nervous?’’; ‘‘yfelt so down in the dumps nothing could cheer you up?’’; ‘‘yfelt calm and peaceful?’’; ‘‘yfelt downhearted and miserable?’’; ‘‘ybeen happy?’’ The index ranged from a low of 5 to a high of 30. Because the index was highly skewed, cutpoints for 3 equal groups were used to reduce the index to a 3category variable, high, moderate, and low mental health. Statistical analyses I use logistic regression and the odds ratio (OR) as a measure of effect between smoking behavior and neighborhood disorder, perceived safety, and other individual characteristics. Because the elements of the sample were independently selected, and each local neighborhood environment observed is associated with a single household, hierarchical models are not appropriate. I estimate adjusted ORs and 95% confidence intervals in logistic regression analyses, first including only neighborhood disorder and city of residence and then adding other sets of variables. In the analyses based on data from all respondents, I use the Huber/White/sandwich estimator of variance because the observations are not independent within households. In the analyses including perceived safety that are based on data from 1 informant per household, I use the traditional calculation of the variance estimator. Covariates include city of residence, and the individual characteristics age, age-squared, sex, marital status, educational achievement, unemployment, mental health, household SES, housing tenure. A range of statistical interaction terms were tested, including one investigating whether the relationship between neighborhood disorder and smoking is different for men and women, and one analyzing whether it differs across the 7 cities. Only those that were significant are included here. Findings Of the 5784 individuals in the study population, 75.3% report that they do not smoke daily, 23.8% report smoking on a daily basis, and 0.8% did not respond to the question. There were no significant differences between responders and non-responders by age, gender, or educational achievement. However, a significantly higher proportion of nonresponders were separated, divorced, or widowed
ARTICLE IN PRESS 2470
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
(27% compared to 12% of responders) and a significantly smaller proportion was married (48% compared to 63% of responders). Of the 2782 households in the study population, 471 or 17% live in areas rated high on neighborhood disorder and 820 or 30% live in areas with a low rating. As expected, neighborhood physical disorder is significantly inversely associated both with residents’ perception of their area as evaluated by others and with the measure of household SES (Table 1). The mean perceived area evaluation score of 4.0 for areas rated low on neighborhood disorder is significantly higher than the mean of 3.6 in areas rated high; and the mean household SES score of 21.0 for areas with a low rating is significantly higher than the mean of 18.6 for areas with a high rating. Furthermore, a significantly larger proportion of households in areas rated high on physical disorder are located in neighborhoods comprised mainly of large multi-family buildings compared to households in areas rated low (Table 1); this suggests that areas rated high on neighborhood disorder are more likely to have high residential density. Table 1
suggests neighborhood disorder is also associated with other indicators of quality-of-life. A significantly larger proportion of informants in areas rated high on physical disorder report being strongly bothered or annoyed by general traffic noise and by general neighborhood noise. A significantly higher proportion also report not feeling safe returning home at night compared to informants in areas rated low on physical disorder. To show the association between neighborhood disorder and smoking behavior, I display first the results of the logistic regression adjusting only for city-of-residence (Table 2, Model 1). In the second model, I also adjust for age, sex, and marital status and in the third, I add the interaction terms between neighborhood disorder and sex. And finally, in the fourth model, I include the full set of household and individual-level covariates along with the significant interactions (Table 2, Model 4). In Model 1, neighborhood disorder is significantly associated with smoking: living in an area rated moderate compared to low on physical disorder increases the odds of smoking at least 1 cigarette per day 27%, whereas living in an area rated high compared to
Table 1 Perceived quality of residential area, household socioeconomic position, neighborhood type, noise annoyance, and perceptions of safety by residence in a neighborhood rated high vs. low on physical disorder Variables
Mean perceived area evaluation score (range 1 ¼ very bad to 5 ¼ very good) Mean household socioeconomic score (range 6–35) Neighborhood type Mainly detached or semi-detached houses Mainly apartment blocks up to 4 floors Mixed neighborhood Mainly apartment blocks with 5 or more floors Bothered/annoyed by general traffic noise Strongly Moderately Not at all Bothered/annoyed by general neighborhood noise Strongly Moderately Not at all Feel safe returning home at night No To some extent Yes Significant at po0:01.
High neighborhood disorder ðn ¼ 471Þ
Low neighborhood disorder ðn ¼ 820Þ
Total percent (overall mean)
3.6
4.0
(3.9)
18.6
21.0
(20.1)
3% 12% 24% 62%
16% 26% 27% 31%
11% 26% 22% 41%
24% 40% 36%
17% 42% 41%
19% 40% 41%
21% 41% 38%
10% 35% 55%
13% 37% 55%
37% 25% 38%
16% 24% 60%
22% 24% 54%
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
2471
Table 2 Odds ratios associated with the effect of neighborhood disorder, household, and individual-level factors on smoking at least 1 cigarette per day Model 1
Model 2
Model 3
Model 4
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
1.27 1.48
(1.06, 1.51) (1.18, 1.85)
1.24 1.39 0.49 1.12 0.99 1.60 1.27
(1.03, (1.10, (0.44, (1.09, (0.99, (1.28, (1.01,
1.42 1.90 0.65 1.12 0.99 1.61 1.27 0.75 0.51
(1.14, (1.42, (0.52, (1.09, (0.99, (1.29, (1.01, (0.57, (0.36,
1.30 1.60 0.63 1.13 0.99 1.26 1.15 0.79 0.51
(1.03, (1.18, (0.50, (1.10, (0.99, (1.00, (0.91, (0.59, (0.35,
Primary vs. post-secondary 1st sec. vs. post-sec. 2nd sec. vs. post-sec.
1.41 1.78 1.57
(1.09, 1.83) (1.46, 2.17) (1.30, 1.90)
Unemployed vs. not unemployed
2.73
(1.82, 4.10)
Mental health: Moderate vs. lo High vs. low
0.75 0.63
(0.63, 0.89) (0.53, 0.76)
Socioeconomic postion Lowest vs. group 2 Lowest vs. 3 Lowest vs. 4 Lowest vs. highest Rented vs. owned Unemployment sex
0.87 0.90 0.75 0.66 1.74 0.42
(0.66, (0.68, (0.56, (0.48, (1.40, (0.24,
Neighborhood disorder Moderate vs. low High vs. low Female vs. male Age Age-squared Divorced vs. married Single vs. married Moderate disorder sex High disorder sex
Log pseudo-likelihood
3083.8159
2875.2332
1.50) 1.76) 0.56) 1.15) 0.99) 2.00) 1.59)
2869.402
1.79) 2.54) 0.81) 1.15) 0.99) 2.00) 1.60) 0.99) 0.73)
1.65) 2.17) 0.80) 1.16) 0.99) 1.60) 1.46) 1.06) 0.76)
1.14) 1.19) 0.99) 0.90) 2.16) 0.73)
2614.9043
N.B. all models adjusted for city of residence. Significant at po0:05.
low on physical disorder increases the odds 48%. The size of the effect decreases slightly when demographic covariates are introduced into the model (Table 2, Model 2). Model 3 shows that the interaction between neighborhood disorder and sex is statistically significant; men living in areas rated high on physical disorder are disproportionately more likely to smoke daily than women living in such areas (37% of men compared to 18% of women). A similar series of models was estimated to investigate the mediating effect of perceived safety, and is based on a sample including only household informants (Table 3). I display first the results for neighborhood disorder adjusting only for city-ofresidence (Table 3, Model 1). In the second model, I add perceived safety (Table 3, Model 2). The
significant association between neighborhood disorder and smoking is only slightly reduced, suggesting a small mediating effect of perceived safety. Feelings of safety also display a significant effect of their own: feeling somewhat safe returning home at night compared to not at all safe decreases the odds 32%; feeling very safe compared to not at all decreases the odds of smoking 16% but the result does not reach significance. Adding age, sex, and marital status enhances the effect of perceived safety: feeling very or somewhat safe compared to not at all safe decreases the odds of smoking 63% (Table 3, Model 3). In the final model with the full set of household and individual-level covariates and the interaction terms, the interactions between sex and neighborhood disorder and perceived safety are both
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
2472
Table 3 Odds ratios associated with the effect of neighborhood disorder and perceived safety on smoking at least 1 cigarette per day Model 11
Neighborhood disorder Moderate vs. lo High vs. low
Model 21
Model 43
OR
95% CI
OR
95% CI
OR
95% CI
OR
95% CI
1.39 1.94
(1.11, 1.73) (1.45, 2.58)
1.38 1.92
(1.11, 1.72) (1.43, 2.56)
1.32 1.64
(1.05, 1.67) (1.22, 2.22)
1.74 2.60
(1.17, 2.61) (1.56, 4.33)
0.76 0.86
(0.57, 0.99) (0.68, 1.10)
0.61 0.62
(0.46, 0.82) (0.48, 0.81)
0.46 0.44 0.62 0.35 1.55 2.13
(0.26, (0.27, (0.38, (0.19, (0.79, (1.19,
Feels safe Somewhat vs. not at all Very vs. not at all Moderate disorder sex High disorder sex Feels somewhat safe sex Feels very safe sex Log pseudo-likelihood
Model 32
1441.3461
1418.4274
1323.1319
0.81) 0.73) 1.01) 0.66) 3.04) 3.80)
1210.9901
Significant at po0.05. 1
Adjusted for city of residence. Adjusted for city of residence, age, sex, marital status. 3 Adjusted for city of residence, age, sex, marital status, education, unemployment, mental health, household SES, housing tenure, interactions between sex and neighborhood disorder, sex and perceived safety, sex, and unemployment. 2
statistically significant (Table 3, Model 4). The interaction between perceived safety and sex indicates that like neighborhood physical disorder, the effect of perceived safety on smoking is significantly more important for men than for women; men who do not feel safe returning home at night are disproportionately more likely to smoke daily than women who do not feel safe (36% of men compared to 19% of women). Separate logistic regressions by sex indicate that for men, living in an area rated high on physical disorder increases the odds of smoking 2-fold; for women in similar areas, the odds only increase 7% (data not shown). Perceived safety decreases the odds of smoking daily 75% for men and 23% for women (data not shown). As found in other studies, individual characteristics play a major role in explaining differences in smoking behaviors across these 7 cities. A mental health score of high compared to low reduces the odds of smoking by 59%. Unemployment increases the odds more than 2-fold; the significance of the interaction term, however, indicates the effect is much less for women than for men. Other individual-level effects identified in the full model are for the most part consistent with the literature (Table 2). Women have 59% lower odds of smoking at least 1 cigarette per day than men. The significant age-squared term indicates that both older adults
and young people have lower odds than adults. Divorced or separated adults have significantly higher odds of smoking than those who are married or cohabiting (26% higher), as do those with lower levels of educational achievement compared to those with a post-secondary education (41%, 78%, and 57%). Those in the highest socioeconomic group have 52% lower odds of smoking than those in the lowest group, and renters have 74% higher odds of smoking than residents who own their dwelling. Discussion and conclusions The findings of this analysis of smoking behaviors in 7 European cities indicate that smoking is significantly associated with the physical disorder of local neighborhood environments, and that this relationship is not significantly different in size or direction across the cities in the study. The measure of disorder is based on direct observations by trained surveyors of indicators of neighborhood physical disorder: the presence of graffiti and litter, and the absence of vegetation on facades, balconies, or windows. Across the cities in the study, places rated high on physical disorder are more likely to be located in neighborhoods with high residential density, compared to areas rated low. Furthermore, the residents of such places are more likely to report
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475
not feeling safe returning home at night, and being bothered by traffic and general neighborhood noise, suggesting a lower quality-of-life in neighborhoods rated high on physical disorder. The effect of neighborhood physical disorder on smoking behaviors is substantial, ranging from a main effect of 64% to a 2-fold increase when men are analyzed separately. Only a small part of the effect is mediated by perceived safety. The finding that the effect of neighborhood disorder on smoking is significantly less important for women than for men is consistent with a study of a population in Illinois, USA, indicating that neighborhood poverty affects men’s smoking but not women’s (Ross, 2000). It is inconsistent, however, with studies by Duncan et al. (1999) and Kleinschmidt et al. (1995) who do not find a significantly different association between neighborhood deprivation and smoking in UK populations for men and women. Why neighborhood physical disorder might affect men and women’s smoking differently is not investigated here but clues can be found in studies focusing on the behavior of low-status men. These suggest that neighborhood physical disorder signals an inability on the part of residents to maintain social order either informally or through their influence on formal services. Areas with disorderly and unsafe environments attract men whose status is subordinate in the hierarchy of their everyday working world; they provide a public space within which low-status men congregate and display the behaviors such as smoking which define a certain type of masculinity (Majors & Billson, 1992). These behaviors are likely to permeate neighborhoodbased friendship networks and spread through them. The fact that there were consistent links between neighborhood disorder and smoking across the different cities suggests that macro-level forces such as residential segregation and social inequality do not affect the relationship between neighborhood disorder and smoking. This does not mean, however, that they do not have an effect on smoking behaviors independent of local place effects. The finding of a substantial effect of neighborhood physical disorder on smoking across a range of cities in Europe adds to the evidence suggesting that environmental interventions are worth pursuing in conjunction with other approaches to smoking prevention. Preliminary evidence indicates that neighborhood improvement influences smoking in a study based on a combined longitudinal and
2473
cross-sectional survey of the health of residents before and after a neighborhood renewal program in the city of Newcastle, England. Blackman and Harvey (2001) identified a 44% decline over 5 years in the prevalence of smoking following neighborhood improvements. Other studies are needed to corroborate this finding. Improving disadvantaged neighborhoods is worth doing whether or not it reduces smoking prevalence or other health risks. However, if neighborhood improvement programs can be shown to have significant health effects among other quality-of-life benefits, it would strengthen the case for greater resource allocation to such efforts. Building the evidence base for this will require analyses that go beyond the cross-sectional ones featured here. There is enough evidence in the growing body of cross-sectional studies however to indicate that it is well worth the effort of evaluating the health effects of existing neighborhood improvement policies or other housing policies to see if smoking is affected. It may be that how changes in the physical environment are brought about determines whether or not health risks are reduced. Neighborhood improvement programs involving the residents themselves may have a better chance of influencing behaviors like smoking. They yield a sense of collective efficacy in addition to increased satisfaction with the local physical environment and increased perceived safety and decreased stress. Kelling and Coles (1996) tell of several decaying communities in which residents took it upon themselves to improve their situation, in collaboration with law enforcement; among their main efforts were to clean up graffiti before it sends off negative signals to others, and to clean up the litter and refuse in streets and alleys. Crime decreased and quality-of-life increased in those communities following their efforts. Further studies are needed to investigate whether behaviors that threaten health such as smoking, are also affected.
Acknowledgments I thank Xavier Bonnefoy, Matthias Braubach and members of the WHO LARES study team for the survey data and their assistance in many ways, Donald Lloyd for his statistical advice, and Richard Rogers, Charles Nam, and anonymous reviewers for helpful comments on earlier drafts.
ARTICLE IN PRESS 2474
R. Miles / Social Science & Medicine 63 (2006) 2464–2475
References Acevedo-Garcia, D., Lochner, K. A., Osypuk, T. L., & Subramanian, S. V. (2003). Future directions in residential segregation and health research: A multilevel approach. American Journal of Public Health, 93(2), 215–221. Acierno, R. A., Kilpatrick, D. G., Resnick, H. S., Saunders, B. E., & Best, C. L. (1996). Violent assault, posttraumatic stress disorder, and depression: Risk factors for cigarette use among adult women. Behavior Modification, 20(4), 363–384. Anda, R. F., Williamson, D. F., Escobedo, L. G., Mast, E. E., Giovino, G. A., & Remington, P. L. (1990). Depression and the dynamics of smoking. A national perspective. Journal of the American Medical Association, 264(12), 1541–1545. Austin, D. M., Furr, L. A., & Spine, M. (2002). The effects of neighborhood conditions on perceptions of safety. Journal of Criminal Justice, 30(5), 417. Barnett, R., Pearce, J., & Moon, G. (2005). Does social inequality matter? Changing ethnic socio-economic disparities and Maori smoking in New Zealand, 1981–1996. Social Science & Medicine, 60(7), 1515–1526. Blackman, T. (2005). Smoking, worklessness, low income and liveability: joining up smoking cessation and neighbourhood renewal programmes. School of Applied Social Sciences, Durham University, unpublished paper. Blackman, T., & Harvey, J. (2001). Housing renewal and mental health: A case study. Journal of Mental Health, 19(5), 571–580. Bonnefoy, X., Braubach, M., Davidson, M., & Robbel, N. (in press). A pan-European housing and health survey: Description and evaluation of methods and approaches. International Journal of Environment and Pollution. Byrne, D., & Keithley, J. (1993). Housing and the health of the community. In R. Burridge, & D. Ormandy (Eds.), Unhealthy housing: Research, remedies and reform (pp. 41–66). London, Glasgow, New York, Tokyo, Melbourne, Madras: E & FN SPON. Cavelaars, A. E. J. M., Kunst, A. E., Geurts, J. J. M., Crialesi, R., Grotvedt, L., Helmert, U., et al. (2000). Educational differences in smoking: International comparison. British Medical Journal, 320(7242), 1102–1107. Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95–S120. Diehr, P., Koepsell, T., Cheadle, A., Psaty, B. M., Wagner, E., & Curry, S. (1993). Do communities differ in health behaviors? Journal of Clinical Epidemiology, 46, 1141–1149. Diez Roux, A. V. (2003). Area characteristics, individual-level socioeconomic indicators, and smoking in young adults: The CARDIA study. American Journal of Epidemiology, 157(4), 315–326. Duncan, C., Jones, K., & Moon, G. (1999). Smoking and deprivation: Are there neighbourhood effects? Social Science & Medicine, 48, 497–505. Frohlich, K. L., Potvin, L., Chabot, P., & Corin, E. (2002). A theoretical and empirical analysis of context: Neighbourhoods, smoking and youth. Social Science & Medicine, 54(9), 1401–1417. Graham, H. (1987). Women’s smoking and family health. Social Science & Medicine, 25(1), 47–56. Graham, H. (1994). Gender and class as dimensions of smoking behaviour in Britain: Insights from a survey of mothers. Social Science & Medicine, 38(5), 691–698.
Graham, H., & Der, G. (1999). Influences of women’s smoking status: The contribution of socioeconomic status in adolescence and adulthood. European Journal of Public Health, 9, 137–141. Jencks, C., & Mayer, S. E. (1990). The social consequences of growing up in a poor neighborhood. In L. E. Lynn, & M. F. H. McGeary (Eds.), Inner-city poverty in the United States (pp. 111–186). Washington, DC: National Academy Press. Kawachi, I., Kennedy, B. P. E., & Glass, R. (1999). Social capital and self-rated health: A contextual analysis. American Journal of Public Health, 89(8), 1187–1193. Kawachi, I., Kennedy, B. P. E., Lochner, K. S. M., & Prothrow Stith, D. (1997). Social capital, income inequality, and mortality. American Journal of Public Health, 87(9), 1491–1498. Kelling, G. L., & Coles, C. M. (1996). Fixing broken windows: Restoring order and reducing crime in our communities. New York, NY: The Free Press. Kleinschmidt, I., Hills, M., & Elliot, P. (1995). Smoking behaviour can be predicted by neighbourhood deprivation measures. Journal of Epidemiology and Community Health, 49(Suppl. 2), S72–S77. LaGrange, R. L., Ferraro, K. F., & Supancic, M. (1992). Perceived risk and fear of crime: Role of social and physical incivilities. Journal of Research in Crime and Delinquency, 29, 311–334. Lepie`ge, A., Ecosse, E., Pouchot, J., Coste, J., Perneger, T. (2001). Le Questionnaire MOS SF-36: Manuel de l’utilisateur et guide d’interpre´tation des scores. Editions ESTEM. Macintyre, S., Ellaway, A., & Cummins, S. (2002). Place effects on health: How can we conceptualise, operationalise and measure them? Social Science & Medicine, 55(1), 125–139. Majors, R., & Billson, J. M. (1992). Cool pose: The dilemmas of black manhood in America. New York, NY: Lexington Books. Mays, N., & Chinn, S. (1989). Relation between all-cause standardized mortality ratios and two indices of deprivation at regional and district level in England. Journal of Epidemiology and Community Health, 43, 191–199. Pampel, F. C. (2002). Inequality, diffusion, and the status gradient in smoking. Social Problems, 49(1), 35–57. Patterson, J. M., Eberly, L. E., Ding, Y., & Hargreaves, M. (2004). Associations of smoking prevalence with individual and area level social cohesion. Journal of Epidemiology and Community Health, 58(8), 692–697. Perkins, D. G., & Taylor, R. B. (1996). Ecological assessments of community disorder: Their relationship to fear of crime and theoretical implications. American Journal of Community Psychology, 24, 63–107. Peto, R., Lopez, A. D., Boreham, J., & Thun, M. (2004). Mortality from smoking in developed countries 1950– 2000 (2nd ed.). Oxford: Oxford University. Available online at /http:// www.ctsu.ox.ac.uk/tobaccoS, accessed 9/24/05. Peto, R., Lopez, A. D., Boreham, J., Thun, M., & Heath, C., Jr. (1992). Mortality from tobacco in developed countries: Indirect estimation from national vital statistics. Lancet(339), 1268–1278. Peto, R., Lopez, A. D., Boreham, J., Thun, M., & Heath, C., Jr. (1994). Mortality from smoking in developed countries 1950– 2000. Oxford: Oxford University Press. Putnam, R. (1993). The prosperous community: Social capital and community life. American Prospect, 35, 42.
ARTICLE IN PRESS R. Miles / Social Science & Medicine 63 (2006) 2464–2475 Raudenbush, S. W. (2003). The quantitative assessment of neighborhood social environments. In I. Kawachi, & L. F. Berkman (Eds.), Neighborhoods and health (pp. 112–131). New York, NY: Oxford University Press. Reijneveld, S. A. (1998). The impact of individual and area characteristics on urban socioeconomic differences in health and smoking. International Journal of Epidemiology, 27, 33–40. Rogers, R., Hummer, R. A., Krueger, P. M., & Pampel, F. C. (2005). Mortality attributable to cigarette smoking in the United States. Population and Development Review, 31(2), 259–292. Ross, C. E. (2000). Walking, exercising, and smoking: Does neighborhood matter? Social Science & Medicine, 51, 265–274. Sampson, R. J. (2001). How do communities undergird or undermine human development? Relevant contexts and social mechanisms. In A. Booth, & N. Crouter (Eds.), Does it take a village? Community effects on children, adolescents, and families. Mahwah, NJ: Lawrence Erlbaum. Sampson, R. J. (2003). Neighborhood-level context and health: Lessons from sociology. In I. Kawachi, & L. F. Berkman (Eds.), Neighborhoods and health (pp. 132–146). New York, NY: Oxford University Press.
2475
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277(5328), 918–924. Sampson, R. J., Morenoff, J., & Earls, F. (1999). Beyond social capital: Spatial dynamics of collective efficacy for children. American Sociological Review, 64, 633–660. Skogan, W. G. (1990). Disorder and decline. New York: Free Press. Skogan, W. G., & Maxfield, M. G. (1981). Coping with crime. Newbury Park, CA: Sage. Stead, M., MacAskill, S., MacKintosh, A., Reece, J., & Eadie, D. (2001). It’s as if you’re locked in: Qualitative explanations for area effects on smoking in disadvantaged communities. Health & Place, 7(4), 333–343. Taylor, R., & Covington, J. (1993). Community structural change and fear of crime. Social Problems, 40, 374–395. Toseland, R. W. (1982). Fear of crime: Who is most vulnerable? Journal of Criminal Justice, 10, 199–209. Wen, M., Browning, C. R., & Cagney, K. A. (2003). Poverty, affluence, and income inequality: Neighborhood economic structure and its implications for health. Social Science & Medicine, 57(5), 843–860. Wilson, J. Q., & Kelling, G. (1982). Broken windows: The police and neighborhood safety. Atlantic Monthly.