Social Norms, Collective Efficacy, and Smoking Cessation in Urban Neighborhoods

Abstract

Objectives. We examined the separate and combined relations of neighborhood-level social norms and collective efficacy with individuals’ cigarette smoking cessation.

Methods. We modeled the hazard of quitting over a 5-year period among 863 smokers who participated in the 2005 New York Social Environment Study.

Results. In adjusted Cox proportional hazard models, prohibitive neighborhood smoking norms were significantly associated with higher rates of smoking cessation (second quartile hazard ratio [HR] = 1.17; 95% confidence interval [CI] = 0.59, 2.32; third quartile HR = 2.37; 95% CI = 1.17, 4.78; fourth quartile HR = 1.80; 95% CI = 0.85, 3.81). We did not find a significant association between neighborhood collective efficacy and cessation or significant evidence of a joint relation of collective efficacy and smoking norms with cessation.

Conclusions. Neighborhood social norms may be more relevant than is collective efficacy to smoking cessation. The normative environment may shape health behavior and should be considered as part of public health intervention efforts.

Cigarette smoking is one of the most important causes of preventable morbidity and premature mortality worldwide. Restrictions on permissible locations for smoking have increased in scope and number over the past 2 decades in the United States, and rates of smoking have declined. Nonetheless, an estimated 45.3 million (19.3%) adults in the United States still smoke, and each year, smoking-related diseases claim 443 000 American lives.^1a-b Given the continued high prevalence of smoking and the reduction in risk of disease and premature mortality provided by quitting,^2,3 there is a need to identify factors that support smoking cessation that interventions could target. Past studies have documented the importance of social networks in the cessation of smoking.^4,5 Further study of social determinants of smoking cessation has the potential to inform interventions that promote health by altering the structural context (e.g., changing norms and taxation practices) to complement more traditional individual behavior change approaches.⁶

Substantial variations in the rates of disease by region, state, county, and neighborhood have long been noted.^7–9 There is growing evidence that variations in rates of disease are determined not only by different distributions of individuals between places but also by the social and physical environments in which people live.^10,11 Neighborhoods have long been studied as geographic units wherein residents have distinct social ties, common exposures, and access to resources. Indeed, smoking patterns vary by geographical location, suggesting that aspects of the social and physical environment influence individual smoking behavior.¹² Existing research suggests that characteristics of neighborhoods shape the risk of smoking.^13–15 There is evidence that smoking is associated with neighborhood deprivation and low socioeconomic status.^14–19

The role of neighborhood factors in smoking cessation has been less frequently studied. The only study to examine the relations between the neighborhood environment and smoking cessation found that there was more quitting in areas with higher socioeconomic status.¹⁴ Examining other social environment characteristics in relation to smoking cessation may increase our understanding of how the social environment shapes smoking cessation and suggest avenues for structural intervention.

Social norms, broadly defined as rules that dictate acceptable behavior within a group, are one aspect of the neighborhood social environment that merits study in relation to smoking cessation. Research suggests that social networks may influence smoking cessation through group social smoking norms.^4,5 Similarly, the neighborhood social environment may influence individual smoking cessation by giving rise to social norms that define the boundaries of permissible or desirable behaviors. Situational norms and attitudes have been linked to alcohol consumption,²⁰ dietary intake,¹⁸ and smoking prevalence.^16,18,21

Community social norms are considered a contextual influence in the smoking literature^{16,18,22–25}; however, studies examining the role of norms typically use individual perceptions of norms rather than group-level measures of norms. Of the few studies that have examined social norms and smoking cessation, 1 found that individual perception of social norms surrounding smoking contributed to smoking behavior, including cessation, among adolescents.²⁴ Another study found greater quit ratios in immigrant communities in the United States, pointing to a possible role of social norms influencing behavior.²⁶ The limited existing research suggests that neighborhood social norms are worth examining as a potential factor that shapes smoking cessation.

The closeness of social relationships in a neighborhood is another aspect of the neighborhood social environment that merits study in relation to smoking cessation. Collective efficacy is a neighborhood construct that measures the cohesiveness of a group (social cohesion) and the group's ability to act to achieve goals (informal social control).²⁷ Groups with higher efficacy have agency to produce desired effects and limit undesired ones through their collective action.²⁸ Analyses have shown that higher levels of collective efficacy in a neighborhood protect against negative health outcomes.²⁹ Higher levels of collective efficacy have been associated with lower levels of violent crime and homicide rates,²⁷ obesity in youth,³⁰ and heart disease mortality.³¹

There is limited research on smoking and collective efficacy. Although studies have shown that smoking risk is higher in neighborhoods with lower levels of collective efficacy, these studies did not investigate smoking cessation.^32–34 Related constructs of individual perception of social capital and social participation have been associated with smoking cessation.^35–37

Thus, the literature supports the possibility that neighborhood smoking norms and collective efficacy are important in shaping smoking cessation, but the associations have not yet been examined.

Existing research also suggests a potential convergence of these neighborhood characteristics in shaping smoking cessation. The impact of social norms on smoking cessation may depend on the closeness and influence of social relationships within a neighborhood environment. This notion is supported by a qualitative study in a community in Glasgow, Scotland, that found that high cohesion and strong prosmoking norms combined not only to foster smoking but also to discourage or undermine cessation.³⁸ This convergence was also supported by a quantitative multilevel study that found that in neighborhoods where norms were strongly antismoking, higher collective efficacy protected against smoking and the individual odds of smoking were lower.¹⁶

Informed by the existing literature, we examined relations between the social environment of a neighborhood, as characterized by smoking norms and collective efficacy, and individual smoking cessation. We tested 3 specific hypotheses: (1) neighborhood norms that are less accepting of smoking are positively associated with the incidence of individual smoking cessation, (2) higher levels of neighborhood collective efficacy are positively associated with the incidence of smoking cessation, and (3) neighborhood smoking norms and collective efficacy interact such that smoking cessation is highest where levels of collective efficacy are high and norms are less accepting of smoking.

METHODS

Conducted between June and December of 2005, the New York Social Environment Study is a multilevel study designed to examine the relations between neighborhood environment characteristics and individual substance use outcomes. The 4000 participants were contacted using a random-digit-dial telephone survey of households in the 59 community districts across New York City.

Study Design

In each household, 1 adult aged 18 years or older whose birthday was closest to the date of the survey was interviewed by telephone. Interviews were conducted in either English or Spanish. Of those contacted and eligible, 54% agreed to participate in the study. Respondents were offered $10 in compensation for their participation.

Survey respondents were interviewed with a structured questionnaire that included questions on demographic and socioeconomic characteristics, including age, race/ethnicity, gender, marital status, birthplace, education, income, employment, years lived in the current neighborhood, and interview language. These variables were controlled for as confounders in this analysis.

Independent Variables: Neighborhood-Level Social Environment

Each respondent's address or nearest cross streets were geocoded and linked to a neighborhood unit. We defined the neighborhoods in this analysis as New York City's 59 community districts. Community districts were initially delineated by a resident consultative process organized by the Office of City Planning to reflect residents’ own descriptions of neighborhoods in the 1970s, and consequently these areas represent recognizable neighborhoods with which residents identified, such as the Upper East Side and the South Bronx. The community districts are not arbitrary spatial units; instead, they each share a political and social organization. Previous studies have found relations between aspects of these neighborhood areas and resident health and health behaviors.^{11,16,39–43}

Neighborhood measures of smoking norms and collective efficacy were the averages of the individual responses from the entire study population (i.e., including smokers and nonsmokers) across the 59 community districts. We calculated neighborhood smoking norms by averaging individual responses to questions about the acceptability of smoking, which the National Survey on Drug Use and Health developed.⁴⁴ Respondents were asked their opinion of adults smoking cigarettes regularly and were given the options of “acceptable,” “unacceptable,” and “don't care.” Neighborhood norms were defined as the proportion of residents who believed it was unacceptable for adults to smoke regularly. This construct therefore measured the neighborhood social norm based on reports of all residents, as opposed to individual perceptions of the neighborhood smoking norm. We assessed collective efficacy by averaging individual responses to items capturing social cohesion (5 items) and informal social control (5 items) that Sampson et al. developed.²⁷ The social cohesion subscale included 5 items with Likert responses and assessed residents’ perceptions of the extent to which their neighbors are close-knit, are helpful, get along, share values, and are trustworthy. The informal social control subscale also included 5 items with Likert responses and measured perceptions of the likelihood that neighbors would intervene if children skipped school, sprayed graffiti, or disrespected an adult; if there was a fight; or if the city was closing a firehouse.

Dependent Variable: Smoking Cessation

We assessed smoking behavior and cessation using the World Mental Health Comprehensive International Diagnostic Interview tobacco module.⁴⁵ Self-report of smoking is comparable to biomarkers of smoking,⁴⁶ and self-assessments collected in person and by phone have been found to be equivalent.¹² We used measures in the tobacco module, including retrospectively recalled ages of smoking initiation and cessation, to define a study population of smokers and the cessation outcome, effectively reconstructing the longitudinal course of smoking and quitting before the study. We examined time to incidence of quitting among the subpopulation of those at risk (those who smoked more than 1 pack a week for more than 2 months) during any portion of the 5-year interval preceding the survey. We selected the 5-year interval by balancing the need to have an adequate number of smokers and quitting events so that there would be sufficient power against the need to only reconstruct smoking and quitting history as far back as necessary because neighborhood measures were not available for years before the 2005 survey. We did not want to assume neighborhood measures were constant over an unreasonably long period. Based on power calculations, the 5-year interval was the shortest period that provided reasonable power; it provided 76% power to detect a hazard ratio (HR) of 1.75 and 91% power to detect an HR of 2.00. We used information on current age, age first smoked, age when quit smoking, and age when moved to the neighborhood to construct the study population and outcome events. We classified individuals as part of the analysis population of smokers if they ever smoked at least once a week for at least 2 months during the 5-year interval before the survey, and we considered only the time that they were smoking and residing in the neighborhood of residence as of the 2005 survey as time at risk. If a person initiated smoking after 2000 or moved to the neighborhood after 2000, we included only their years as a smoker and residing in the neighborhood. As these variables provide precision to the level of year intervals, we included individuals who had resided in the neighborhood for at least 1 year by definition. The outcome was incidence of quitting smoking, which we assessed by reported quit age. If a person quit more than once, we included only the first quitting event.

Analysis

We conducted all statistical analyses with Stata 11.0 (StataCorp LP, College Station, TX). To account for the probability of selection for interview, we weighted analyses by the number of persons in the household divided by the number of telephone lines. All analyses also accounted for nonindependence of residents of the same community district with robust variance estimates. To examine incidence of quitting across the 5-year analysis period, we modeled the hazard of quitting with Cox proportional hazard models. The model estimates the relative hazard, relating the yearly hazard of quitting smoking to a change in the exposures of interest. We conducted additional analyses for a range of analysis time periods from 6 to 10 years to assess the sensitivity of results to the 5-year time window selected.

Our analysis involved 3 steps. (1) We examined the relations of smoking norms and collective efficacy to smoking cessation in descriptive analysis. (2) We analyzed the relations from step 1 adjusted for individual level confounders, including gender, age, race/ethnicity, income, marital status, education, place of birth, employment, and years lived in the current neighborhood, to assess the adjusted relations of smoking norms and collective efficacy to quitting. (3) We included an interaction between neighborhood smoking norms and collective efficacy to assess the potential combined association with smoking cessation.

RESULTS

There were 4000 study participants; the mean number of residents per community district was 68, with a range of 19 to 144. Of the 1755 ever smokers in the New York Social Environment Study population, 863 smokers were eligible for this analysis because they were at risk for quitting smoking during some portion of the 5-year period from 2000 to 2005. Thus we conducted the analysis with the subsample of 863 smokers at risk for quitting, and we calculated the neighborhood measures based on the entire study population of 4000. The analysis population contributed 3317 person-years of time at risk for quitting. There were 127 events of smoking cessation over the study period.

The study population characteristics are listed in Table 1, with frequencies and weighted percentages. The cumulative incidence of quitting over the 5-year study period was 15.5%, and the incidence density was 3.8 per 100 person-years; 43.7% of the sample was White, 27.9% was African American, 22.9% was Hispanic, and 5.5% was another race/ethnicity. A large portion (41.6%) of the sample reported incomes of < $40 000 a year. Additionally, 56.5% of the sample was male.

TABLE 1—

Population Characteristics: New York Social Environment Study, New York City, 2005

Characteristic	Unweighted No.(Weighted %)
Smoking
Current smoker	736 (84.5)
Quit smoking	127 (15.5)
Sex
Female	392 (43.5)
Male	471 (56.5)
Age, y
< 25	100 (15.7)
25–35	169 (19.8)
36–45	198 (20.1)
46–55	200 (22.7)
56–65	116 (13.2)
> 65	80 (7.9)
Race/Ethnicity
White	382 (43.7)
Black or African American	247 (27.9)
Asian	21 (2.8)
Hispanic	180 (22.9)
Other	20 (2.7)
Income, $
> 80 000	153 (19.1)
40 000–80 000	266 (31.3)
< 40 000	375 (41.6)
Missing	69 (8.0)
Marital status
Married	278 (37.9)
Divorced	126 (11.3)
Separated	56 (6.0)
Widowed	64 (6.6)
Never married	336 (38.2)
Education
Graduate work	125 (13.2)
College graduate	188 (20.1)
Some college	230 (27.9)
High school graduate	198 (24.0)
< high school graduate	117 (14.8)
Birthplace
New York City	513 (59.9)
Other US city	179 (18.3)
Another country	169 (21.2)
Employment
Working full time	435 (49.5)
Working part time	88 (11.2)
Looking for work or unemployed	111 (13.4)
Retired	106 (11.2)
Homemaker	30 (4.0)
Student	25 (3.6)
On leave	67 (7.2)
Years lived in neighborhood
0–7	301 (33.5)
8–21	284 (35.2)
≥ 22	278 (31.3)

Initial graphical examination of the shape of the relations between neighborhood exposures and smoking cessation suggested an upward trend and eventual plateau for the relation between neighborhood smoking norms and quitting. Therefore, we categorized the neighborhood exposures by quartiles to capture this shape. We have also reported associations for continuous neighborhood measures. The continuous measures were centered and standardized with a change of 2 standard deviations.

Bivariable analysis of the cumulative incidence of smoking cessation by neighborhood collective efficacy, smoking norms, and individual demographic characteristics is presented in Table 2 along with unadjusted HRs. The cumulative incidence of smoking cessation increased across the first 3 quartiles of smoking norms that were less accepting of smoking and decreased slightly in the fourth quartile. Quitting did not appear to differ substantially by the quartile of collective efficacy.

TABLE 2—

Bivariable Analysis of the Cumulative Incidence of Smoking Cessation by Covariates: New York Social Environment Study, New York City, 2005

Characteristic	Quit, Unweighted No. (Weighted %)	Cox Proportional HR	F-test P
Smoking norms (quarters)
First quartile (least unacceptable)	31 (12.0)	1.00	.07
Second quartile	35 (13.8)	1.60
Third quartile	39 (20.5)	2.03
Fourth quartile (most unacceptable)	22 (17.1)	1.45
Collective efficacy (quarters)
Lowest	27 (17.3)	1.00	.46
Lower	41 (18.9)	1.14
Higher	31 (13.1)	0.70
Highest	28 (13.9)	0.89
Sex
Female	53 (14.5)	1.00	.83
Male	74 (16.2)	1.05
Age, y
< 25	14 (15.6)	1.30	.53
25–35	24 (17.3)	0.80
36–45	26 (13.1)	0.70
46–55	30 (14.9)	1.00
56–65	18 (15.0)	1.08
> 65	15 (19.8)	1.30
Race/Ethnicity
White	51 (14.7)	1.00	.44
Black or African American	34 (14.4)	1.04
Asian	3 (15.2)	0.52
Hispanic	35 (18.7)	1.30
Other	2 (13.5)	0.23
Income, $
> 80 000	23 (18.2)	1.00	.35
40 000–80 000	37 (14.3)	0.66
< 40 000	53 (13.7)	0.73
Missing	14 (22.8)
Marital status
Married	51 (18.5)	1.00	.09
Divorced	16 (11.7)	0.61
Separated	12 (22.2)	1.86
Widowed	6 (10.1)	0.47
Never married	42 (13.6)	0.85
Education
Graduate work	20 (17.2)	1.00	.98
College graduate	23 (14.8)	1.02
Some college	32 (14.3)	1.10
High school graduate	33 (16.4)	1.04
<high school graduate	19 (16.2)	0.94
Birthplace
New York City	69 (14.3)	1.00	.22
Other US city	20 (11.1)	1.00
Another country	38 (22.6)	1.56
Employment
Employed, student, or retired	114 (16.2)	1.00	.7
Looking for work or unemployed	13 (10.8)	0.87
Years lived in neighborhood
1–7	35 (13.0)	1.00	.74
8–22	48 (15.7)	1.00
> 22	44 (18.0)	1.21

Note. HR = hazard ratio.

The results of the adjusted models examining relations of neighborhood smoking norms and collective efficacy with smoking cessation are presented in Tables 3 and 4. Consistent with other analyses of these data, we excluded 2 neighborhoods with outlier values for norms from the norms analyses (note that results were the same whether individuals in outlier neighborhoods (n = 36) were excluded or included).¹⁶ Neighborhood smoking norms were significantly associated with individual smoking cessation, controlling for confounders. Across the quartiles of smoking norms, the hazard of quitting smoking was higher in neighborhoods in which smoking was less acceptable, with a plateau in the association in the neighborhoods with the most prohibitive norms (second quartile HR = 1.17; 95% confidence interval [CI] = 0.59, 2.32; third quartile HR = 2.37; 95% CI = 1.17, 4.78; fourth quartile HR = 1.80; 95% CI = 0.85, 3.81). The model with a continuous measure of smoking norms resulted in a measure of association in the same direction; however, the CI crossed the null value, likely because the continuous measure did not capture the shape of the relation (HR = 1.95; 95% CI = 0.97, 3.08). The relation between collective efficacy and smoking cessation was essentially null (Table 4). There was no interaction between collective efficacy and neighborhood smoking norms in this analysis (P = .55); this last model is not presented.

TABLE 3—

Cox Proportional Hazard Model for Social Norms and Smoking Cessation: New York Social Environment Study, New York City, 2005

Characteristic	HR (95% CI)
Smoking norms (quarters)
First quartile (least unacceptable) (Ref)	1.00
Second quartile	1.17 (0.59, 2.32)
Third quartile	2.37 (1.17, 4.78)
Fourth quartile (most unacceptable)	1.80 (0.85, 3.81)
Sex
Female (Ref)	1.00
Male	0.87 (0.53, 1.41)
Age, y
< 25	1.38 (0.58, 3.30)
25–35	0.88 (0.37, 2.06)
36–45	0.66 (0.32, 1.34)
46–55 (Ref)	1.00
56–65	0.97 (0.45, 2.11)
> 65	0.97 (0.35, 2.66)
Race/ethnicity
White (Ref)	1.00
Black or African American	0.79 (0.44, 1.41)
Hispanic	0.96 (0.51, 1.82)
Other and Asian	0.24 (0.05, 1.14)
Income, $
> 80 000 (Ref)	1.00
40 000–80 000	0.67 (0.33, 1.34)
< 40 000	0.63 (0.34, 1.16)
Marital status
Married (Ref)	1.00
Divorced	0.62 (0.27, 1.38)
Separated	1.94 (0.84, 4.46)
Widowed	0.40 (0.12, 1.32)
Never married	0.82 (0.44, 1.52)
Education
College or graduate school (Ref)	1.00
Some college	1.23 (0.54, 2.83)
High school graduate	1.35 (0.61, 2.76)
<high school graduate	0.95 (0.50, 1.81)
Birthplace
New York City (Ref)	1.00
Other US city	1.01 (0.51, 1.99)
Another country	1.87 (1.04, 3.37)
Employment
Employed, student, or retired (Ref)	1.00
Looking for work or unemployed	0.88 (0.46, 1.68)
Years lived in neighborhood
1–7 (Ref)	1.00
8–22	0.95 (0.51, 1.76)
> 22	1.37 (0.68, 2.74)

Note. CI = confidence interval; HR = hazard ratio.

TABLE 4—

Cox Proportional Hazard Model for Collective Efficacy and Smoking Cessation: New York Social Environment Study, New York City, 2005

Characteristic	HR (95% CI)
Collective efficacy (quarters)
Lowest (Ref)	1.00
Lower	0.94 (0.49, 1.82)
Higher	0.60 (0.29, 1.25)
Highest	0.75 (0.36, 1.57)
Sex
Female (Ref)	1.00
Male	0.90 (0.57, 1.44)
Age, y
< 25	1.34 (0.54, 3.31)
25–35	0.75 (0.32, 1.75)
36–45	0.67 (0.34, 1.32)
46–55 (Ref)	1.00
56–65	0.97 (0.47, 2.03)
> 65	1.60 (0.62, 3.94)
Race/Ethnicity
White (Ref)	1.00
Black or African American	0.97 (0.53, 1.81)
Hispanic	1.11 (0.61, 2.03)
Other and Asian	0.26 (0.05, 1.29)
Income, $
> 80 000 (Ref)	1.00
40 000–80 000	0.64 (0.34, 1.21)
< 40 000	0.65 (0.36, 1.17)
Marital status
Married (Ref)	1.00
Divorced	0.61 (0.28, 1.32)
Separated	1.81 (0.78, 4.22)
Widowed	0.33 (0.09, 1.16)
Never married	0.85 (0.47, 1.55)
Education
College or graduate school (Ref)	1.00
Some college	1.02 (0.45, 2.82)
High school graduate	1.12 (0.57, 2.19)
<high school graduate	0.91 (0.49, 1.72)
Birthplace
New York City (Ref)	1.00
Other US city	0.91 (0.46, 1.79)
Another country	1.74 (0.95, 3.20)
Employment
Employed, student, or retired (Ref)	1.00
Looking for work or unemployed	0.89 (0.45, 1.74)
Years lived in neighborhood
1–7 (Ref)	1.00
8–22	0.94 (0.51, 1.73)
> 22	1.22 (0.61, 2.45)

Note. CI = confidence interval; HR = hazard ratio.

Sensitivity analyses conducted for periods of 6 to 10 years suggested consistent magnitude and significance of relations for the quartile norms measure, with the exception of the fourth quartile, which increased in magnitude and was statistically significant for the periods of 7 to 10 years. For the continuous norms measure, the relation was stronger and statistically significant for the periods of 7 to 10 years.

DISCUSSION

In a representative urban population, we found a relation between neighborhood-level smoking norms and individual smoking cessation. The results supported our first hypothesis, finding increased smoking cessation in neighborhoods where smoking was less acceptable. However, the results did not support our second hypothesis that more neighborhood collective efficacy would be associated with increased smoking cessation. Finally, the results did not support our third hypothesis of combined effects of neighborhood smoking norms and collective efficacy on smoking cessation.

These results support and advance the previous literature linking situational norms to incidence of quitting. Although previous studies found that individual perception of social norms were associated with cessation among adolescents, to our knowledge no previous studies have examined group-level norms and adult populations.²⁴ This analysis suggests that social norms may shape smoking cessation for adults in the urban neighborhood setting.

In addition, our findings are congruent with the documented success of environmental tobacco policies to promote smoking cessation.^47–49 One study comparing Massachusetts towns with strict antismoking restaurant regulations with those without found that stronger regulations increased quit attempts among smokers who had previously tried to quit. The authors suggested that regulation seemed to reinforce antisocial smoking norms among smokers who already viewed smoking in bars as socially unacceptable.⁴⁷ Our results support this association between social norms and cessation, suggesting that implementation of environmental or workplace tobacco policy may operate by changing social norms and thereby fostering a social environment more conducive to quitting.

In this analysis, the relation between smoking norms and smoking cessation had a plateauing shape, such that the strength of the relation between norms and quitting smoking flattened as neighborhood norms became increasingly prohibitive. This suggests that there may be something unique about people who remain smokers in neighborhoods with strong norms against smoking. They may be more resilient to community norms, perhaps because of being more isolated from the social processes of the neighborhood. Supporting this possibility, a network analysis of smoking behavior found that smokers are more likely to be on the periphery of social groups.⁵ Although the causal direction of this relationship between smoking and isolation is unknown, it is plausible that in unaccepting environments, smoking stigmatization may induce smokers to form stronger relationships with one another that reinforce continued smoking. Alternatively, there may be a subgroup of smokers for whom the addiction is so strong that behavior is unaffected by social pressure.

Our finding of no significant relation between collective efficacy and smoking cessation is not consistent with previous research suggesting that constructs related to collective efficacy (perception of social capital and social participation) are related to increased smoking cessation.^35,50 To our knowledge, this is the first study to directly measure group-level collective efficacy, thus other measures of social context even when closely related may measure different social processes. Our results suggest that collective efficacy may not be relevant to smoking cessation.

Our results did not support our hypothesis of a combined relation between collective efficacy and smoking norms. An interaction between collective efficacy and smoking norms was found in an analysis of smoking prevalence in this data set¹⁶; together, these results suggest that neighborhood processes may operate differently for smoking cessation than for overall smoking. Perhaps neighborhoods with high collective efficacy and norms against smoking are effective in preventing smoking onset but individuals who are already smokers feel more isolated in these settings. Future analyses should consider these questions explicitly.

Although this analysis was not intended to identify demographic predictors of smoking cessation, it is worth noting that birthplace was the only individual variable associated with a difference in cessation incidence in our multivariate model. Foreign birth was associated with higher quitting incidence compared with those born in the United States. This result supports previous literature that reports higher cessation in immigrant communities and may indicate that immigrants are influenced by the general antismoking norms in the United States that may be quite different from countries of origin.²⁶

This analysis examined first cessation within the 5-year time frame. Although we chose this relatively short interval to limit the temporal difference between neighborhood measures assessed in 2005 and retrospective reports of cessation, social norms around tobacco may have changed in this interval. Individual demographics, such as employment, also may have changed during this period. Further prospective studies of this association would be informative. Additionally, given that relapse is common among people who quit smoking, it would be useful to consider the role of the social context in quitting maintenance in future research. Our analysis did not include environmental variables that may shape smoking norms such as tobacco availability and density of advertising, which may both affect smoking norms and shape smoking behavior.⁴⁹ Future research considering the role of these factors in addition to norms would be informative in terms of their potential role in the generation of social norms in a community. Further, contacts in other social groups (e.g., workplace, social media) could provide alternative communities beyond the geographic neighborhoods that influence smoking; it would be useful for future work to examine norms and other aspects of the many environments that affect smoking behavior. Individuals residing near the borders of neighborhoods may be affected by adjacent neighborhoods as well as their residential neighborhood. It would be interesting for future research to examine influences of adjacent neighborhoods. The 54% cooperation rate for the survey is also a limitation if participants and nonparticipants were substantially different in terms of the relations of interest in this analysis. The demographic profile of the study population is similar to 2000 US Census measures of New York City overall, thus increasing confidence that the smokers in our population represented the general population of smokers. However, the participants may still differ from those in the city overall in ways that we were unable to capture.

Despite these limitations, this study has several important strengths. The World Mental Health Comprehensive International Diagnostic Interview tobacco module allowed us to reconstruct a longitudinal history of smoking from the cross-sectional survey responses. The neighborhood-level social environment variables were not individual perceptions of community characteristics but measures based on responses of the general population (both smokers and nonsmokers) in each neighborhood. Additionally, we used a multilevel design examining group-level norms linked to individual smoking cessation in contrast to an ecological analysis of smoking cessation rates.^15,51

Although many public health campaigns target group norms as a method of changing health behavior (e.g., the National Youth Anti-Drug Media Campaign,⁵² the antitobacco “truth” campaign,⁵³ and the 5-a-Day for Better Health Program⁵⁴), group norms have rarely been studied in epidemiology. Our results support the use of these types of interventions along with antismoking policy or social media campaigns targeted at neighborhood units. Our research demonstrates the importance of explicitly studying how group norms may shape complex health behavior. Our findings suggest that social norms may operate differently for the onset of smoking and for smoking cessation. Additionally, some smokers may be more resistant to norms' effects and should therefore be addressed with distinct and complementary interventions. Smokers who are unaffected by neighborhood social norms may be more isolated from health services in general and may benefit from more primary public health outreach. Additional study to identify distinct social networks of smokers would be a useful step toward targeting interventions. Overall, this study advances our understanding of the role of the social context in shaping smoking cessation, suggesting that norms in the general community are related to incident quitting among smokers. The results lend support to the notion that the normative environment shapes health behavior and should be considered as part of public health intervention efforts.⁶