Article Text

Download PDFPDF

The added value of accounting for activity space when examining the association between tobacco retailer availability and smoking among young adults
  1. Martine Shareck1,2,3,4,
  2. Yan Kestens2,3,4,
  3. Julie Vallée2,3,4,5,
  4. Geetanjali Datta2,4,
  5. Katherine L Frohlich2,3
  1. 1London School of Hygiene and Tropical Medicine, London, UK
  2. 2Département de médecine sociale et préventive, Université de Montréal, Montréal, Quebec, Canada
  3. 3Institut de recherche en santé publique de l'Université de Montréal (IRSPUM), Montréal, Quebec, Canada
  4. 4Centre de recherche du centre hospitalier de l'Université de Montréal (CRCHUM), Montréal, Quebec, Canada
  5. 5Centre National de la Recherche Scientifique (CNRS), UMR Géographie-Cités, Paris, France
  1. Correspondence to Dr Martine Shareck, London School of Hygiene and Tropical Medicine, 15–17 Tavistock Place, London WC1H 9SH, UK; martine.shareck{at}


Background Despite a declining prevalence in many countries, smoking rates remain consistently high among young adults. Targeting contextual influences on smoking, such as the availability of tobacco retailers, is one promising avenue of intervention. Most studies have focused on residential or school neighbourhoods, without accounting for other settings where individuals spend time, that is, their activity space. We investigated the association between tobacco retailer availability in the residential neighbourhood and in the activity space, and smoking status.

Methods Cross-sectional baseline data from 1994 young adults (aged 18–25) participating in the Interdisciplinary Study of Inequalities in Smoking (Montreal, Canada, 2011–2012) were analysed. Residential and activity locations served to derive two measures of tobacco retailer availability: counts within 500 m buffers and proximity to the nearest retailer. Prevalence ratios for the association between each tobacco retailer measure and smoking status were estimated using log-binomial regression.

Results Participants encountering high numbers of tobacco retailers in their residential neighbourhood, and both medium and high retailer counts in their activity space, were more likely to smoke compared to those exposed to fewer retailers. While residential proximity was not associated with smoking, we found 36% and 42% higher smoking prevalence among participants conducting activities within medium and high proximity to tobacco retailers compared to those conducting activities further from such outlets.

Conclusions This study adds to the sparse literature on contextual correlates of smoking among young adults, and illustrates the added value of considering individuals’ activity space in contextual studies of smoking.

  • Disparities
  • Environment
  • Prevention

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Despite an overall decline in many developed countries,1–3 smoking remains disproportionally prevalent in young adults aged 18–25 years.1 ,2 Since smoking is the leading preventable cause of premature death,4 the concentration of this behaviour among young adults is of great public health concern. Early smoking initiation is associated with less success with quitting, a longer smoking duration5 and, consequently, a higher risk of suffering from smoking-related health consequences including several cancers and cardiovascular diseases.

One avenue of promising population-level tobacco control intervention lies in targeting contextual features such as tobacco retailer availability,6 defined as the degree of convenience in obtaining tobacco products, which is affected by the number and location of retailers.6 ,7 The density of tobacco retailers around home8–10 or school11 ,12 has been associated with a higher likelihood of smoking in youth. In one study, retailer density in the residential but not the school neighbourhood was associated with youth smoking.13 Among adults, high tobacco retailer density14–18 and closer proximity15 ,17 ,19 from home have been associated with a number of smoking behaviours including status,14 ,19 initiation,16 ,17 higher nicotine addiction and lower readiness to quit,18 as well as lower smoking abstinence.15 In longitudinal studies, tobacco retailer proximity, but not density, in the residential area was associated with adults’ lower success at smoking abstinence20 and cessation.21 In two studies, increasing contacts with tobacco retailers, as individuals moved across a city, were related to increased cigarette cravings22 and relapse among would-be quitters.23 Whether these associations hold for young adults remains largely unexplored.

An important shortcoming of existing studies lies in their focus on the residential or school neighbourhood to examine the relationship between tobacco retailers and smoking. This overlooks the fact that individuals are mobile and experience a diversity of settings in the course of their daily activities, all of which may provide exposure to tobacco retailers and subsequent opportunities to purchase cigarettes, or to crave smoking.22–24 Most importantly, young adulthood is traditionally characterised by five transitions: leaving school, leaving the parental home, entering full-time work, entering conjugal relationships and having children.25 These may coincide with decreased social and physical bonds to the residential neighbourhood due to increased mobility, independence and the development of relationships in new settings.26 ,27 Moving beyond single, residential or school neighbourhoods, to examine contextual influences on smoking is thus warranted. This can be carried out by studying individuals’ activity space, defined as the subset of activity locations that they experience as a result of their daily activities.28–30 As such, activity spaces provide a more comprehensive and accurate representation of the contextual features and resources that may be encountered on a regular basis.30

In this paper, we assessed the association between young adults’ smoking status, and the number and proximity of tobacco retailers in their: (1) residential neighbourhood and (2) activity space. Tobacco retailer counts and proximity are two complementary measures of retailer availability.31 While counts represent the overall number of tobacco retailers encountered in a given area, proximity approximates the ease of physical access to them. It is assumed that, all else being equal, the closer one is to a tobacco retailer, the easier it is to buy cigarettes or the more tempted one might be to smoke.7 ,31 We hypothesised that young adults who lived or conducted activities in areas with high numbers of (or in close proximity to) tobacco retailers, would be more likely to smoke than those living or conducting activities in areas with fewer tobacco retailers (or farther from such outlets).


Study design and population

Baseline data from the Interdisciplinary Study on Inequalities in Smoking (ISIS), collected between 23 November 2011 and 4 September 2012, were analysed. 6020 young adults were randomly selected by the Régie de l'Assurance Maladie du Québec, a publicly funded health insurance programme in Quebec, Canada, from all eligible individuals living in each of the 35 Centre Locaux de Services Communautaires (CLSC) on the island of Montreal, Canada. CLSCs are the main health services catchment areas in Québec and served as sampling units to ensure geographic representativity across the study territory, that is, the island of Montreal. Eligibility criteria included being between 18 and 25 years old, fluent in French or English and having lived for at least 1 year at the current address at time of first contact. Full details on procedures are available elsewhere.32 Ethical approval was obtained from the Research Ethics Committee of the Université de Montréal's Faculty of Medicine.

Data collection

Individual-level data

Participants provided sociodemographic, smoking and activity location data in an online questionnaire (90% of respondents), in a paper questionnaire (4.2%) or over the phone with a research assistant (5.8%) in exchange for a $10 gift card. The final sample was 2093 individuals, making for a 37.6% response proportion.32 Online, written or verbal informed consent was obtained prior to questionnaire completion.

An activity space questionnaire was specifically developed to collect information on respondents’ regular activity locations (studying, working, grocery shopping, physical activity, leisure activity and two other activities) in addition to their residential address. Participants provided location details (name, address, street, intersection/landmark, city) for each activity type they conducted. The questionnaire's 2-week test–retest reliability was high, as was its convergent validity when compared to a 7-day continuous Global Positioning System track and a prompted-recall survey completed by an independent sample.33

Area-level data

Geographical coordinates of stores that can legally sell tobacco in Québec, Canada (convenience stores, supermarkets, tobacconist shops and gas stations) were extracted from the 2011 DMTI Enhanced Points of Interest (EPOI) database.34 Field validation of the 2010 version of this database was conducted for convenience stores and supermarkets, which were, respectively, found to have sensitivity values of 0.59 and 0.75, and positive predictive values of 0.73 and 1.00.35 Area-level material deprivation was computed from 2006 Canadian Census data.


Outcome variable

Smoking status was assessed with questions borrowed from the Canadian Community Health Survey questionnaire. Participants who had responded affirmatively to smoking at least one entire cigarette in the past were asked: “Currently, do you smoke cigarettes (1) every day, (2) sometimes or (3) never?”. This question has been shown to have high sensitivity when compared to urinary cotinine levels.36 Current smoking was defined as smoking daily or occasionally at the time of survey, while non-smokers combined former smokers and never smokers.37


Age, sex, occupational status (neither studying nor working/working/studying (while also working or not)) and educational attainment were considered individual-level covariates. Educational attainment was operationalised as (1) the highest level completed by participants who were not enrolled in studies, or (2) the highest level attained, imputed based on the level taught at the educational establishment attended at the time of survey, for students.38 Three categories were created based on years of schooling: low (≤11 years; secondary school or less), intermediate (12–13 years; CEGEP/trade school (CEGEPs, or Collèges d'enseignement général et professionel, are post-secondary institutions found only in Québec, Canada39)) and high (14+years; university).

Contextual measures

Respondents reported between 0 and 9 activity locations (mean 3.1, SD: 1.5). Out of the 8422 residential and activity locations for which information was provided, 7.5% were not geocoded, for lack of sufficient details. Geographic coordinates for the remaining 7792 locations were obtained with Batch Geocodeur40 using the street address (97.1% of locations), closest intersection (1.7%), landmark (0.3%), place name (0.5%) or street name (0.3%). A total of 8362 tobacco retailers were identified in the DMTI EPOI database.34 Duplicate entries and those geocoded at the city level (n=597) were discarded for a final count of 7765 tobacco retailers. Geocoded residential, activity and tobacco retailer locations were spatialised in ArcGIS V.10.1. They were used to compute two tobacco retailer measures for each participant's residential and activity locations: counts (ie, the number of retailers in a defined area) and proximity (ie, the distance separating a location from the nearest tobacco retailer). In this paper, the activity space was defined as the combination of unique residential and out-of-home activity locations. Activity space measures were operationalised as the mean number of, or proximity to, tobacco retailers across locations, as carried out in similar studies.41 ,42

Tobacco retailer counts

Tobacco retailer counts were computed within 500 m pedestrian road-network buffers anchored on each location. This distance, which corresponds approximately to a 5 min walk, has previously been used to study the tobacco retailer–smoking relationship.20 ,21 Counts were chosen over alternative measures of density (eg, counts by surface area) because road-network buffers inherently integrate aspects of accessibility by limiting measures to locations that can be reached within a given distance.31 For each participant, we (1) counted tobacco retailers within the residential buffer, and (2) computed the mean number of tobacco retailers across unique residential and activity buffers. For example, if a participant had 3, 2 and 10 tobacco retailers in her home, study and work buffers, respectively, her activity space measure was 5. Residential and activity space counts were categorised into tertiles based on their respective distribution: 1 retailer or fewer (low counts), 2–5 retailers and 6 retailers or more (high) for residential counts, and means of fewer than 4 retailers (low), 4 to 7.9 retailers (medium) and 8 retailers or more (high) for activity space counts.

Tobacco retailer proximity

Tobacco retailer proximity was defined as (1) the shortest pedestrian road-network distance separating the residential location from the nearest tobacco retailer (residential proximity) and (2) the mean of distances separating each participant's residential and activity locations from the closest tobacco retailer (activity space proximity). Since residential and mean activity space proximity measures were similarly distributed, they were categorised into three groups: more than 350 m (low proximity), 150–350 m (medium) and less than 150 m (high).

Area-level deprivation was examined as a potential confounder of the tobacco retailer–smoking association. Deprivation was operationalised as the Pampalon relative material deprivation index combining 2006 Census data on education, employment status and income extracted at the dissemination area (DA) scale, the smallest administrative unit at which Census data is available.43 Scores were aggregated within each buffer, and weighted proportionally to the population and surface area of any DA overlap. Residential and mean activity space deprivation scores were classified into four categories based on quartile cut points for the distribution across the Greater Montreal Metropolitan region.32

Statistical analyses

Given the high smoking prevalence in our sample (22.8%), adjusted prevalence ratios and 95% CIs contrasting smokers and non-smokers were estimated using log-binomial regression.44 Generalised estimating equations with an exchangeable correlation matrix were fitted to account for potential clustering given the nested sampling frame (with 35–71 participants per CLSC territory). Analyses were performed with SPSS V.19.0.

The association between each tobacco retailer measure and smoking was modelled separately for the residential neighbourhood and activity space. Adjusted models included the individual-level covariates age, sex, occupational status and educational attainment. Tests for linear trend across categories of tobacco retailer measures were carried out by modelling tertiles as continuous variables. To compare models’ goodness-of-fit, we present the quasi-likelihood under the independence model criterion (QICu) statistic, with smaller QICu values considered indicative of better model fit.45 Sensitivity analyses using alternative expressions of activity space measures (the sum and maximum counts and minimum proximity across locations), as well as counts aggregated within 800 m buffers, were performed.


Of the 2093 young adults who completed the questionnaire, 37 were excluded because their main work or study location was outside the Greater Montreal Metropolitan Region. This exclusion criterion sought to ensure that the sample would closely represent the spatial behaviour of respondents more likely to experience the study territory on a daily basis. An additional 62 participants were excluded because no residential buffer zone could be created (n=2), due to insufficient details to geocode their activity locations (n=26), or due to missing values for smoking status (n=10) and/or occupational status (n=19) and/or educational attainment (n=10). Compared to the analysis sample (n=1994), excluded participants were more likely to be exclusively in employment and to conduct activities in areas that were characterised, on average, by higher tobacco retailer counts and higher disadvantage (data not shown).

The sample was 56.9% women and mean age was 21.5 years (SD 2.3; table 1). Smokers composed 22.8% of the sample. They were more likely to be male, of lower educational attainment and non-students, compared to non-smokers. A significantly higher proportion of smokers than non-smokers lived and conducted activities in areas characterised by high tobacco retailer counts and proximity. While smokers were more likely to conduct activities in areas that were, on average, more disadvantaged compared to non-smokers, no significant difference was found for residential deprivation level.

Table 1

Descriptive statistics for 1994 young adults from the ISIS study (Montreal, Canada, 2011–2012), full sample and by smoking status

Most activity spaces included a place of study or main employment, but also grocery shopping destinations and sports and leisure locations (table 2).

Table 2

Percentage of participants whose activity space included each type of activity

Prevalence ratios for smoking by categories of residential and mean activity space tobacco retailer counts are shown in table 3 (upper half). High numbers of retailers in the residential area and activity space were associated with an increased prevalence of smoking. Participants encountering six tobacco retailers or more in their residential neighbourhood or a mean of eight retailers or more in their activity space were, respectively, 53% and 46% more likely to be smokers compared to those exposed to fewer retailers. Trends were statistically significant. Goodness-of-fit statistics suggested the residential model was slightly better than the activity space model.

Table 3

Adjusted* prevalence ratios and 95% CIs for the association between tobacco retailer counts and proximity in the residential neighbourhood and activity space,† and smoking status among 1994 participants from the ISIS study (Montreal, Canada, 2011–2012)

Proximity models are also shown in table 3 (bottom half). Residential proximity to a tobacco retailer was not associated with smoking, whereas medium (150–350 m) and high (<150 m) mean proximity in the activity space were, respectively, associated with prevalence ratios of 1.32 (95% CI 1.03 to 1.68) and 1.42 (95% CI 1.09 to 1.86), compared to low mean proximity. The activity space model showed the strongest associations with smoking, a statistically significant trend and had a better fit compared to the residential model. Adjusting for deprivation did not change the associations between tobacco retailer counts or proximity and smoking (data not shown).

In all fully adjusted models, increasing age was associated with a higher likelihood of being a smoker. An educational gradient was also found, with decreasing educational attainment associated with an incrementally higher smoking prevalence (data not shown).


We investigated the association between the number and proximity of tobacco retailers in the residential neighbourhood and activity space, and smoking status among young adults. To our knowledge, this study is the first to consider individuals’ activity space when examining contextual correlates of smoking. It addresses shortcomings related to the common focus on single, usually residential or school neighbourhoods, in contextual studies of smoking46 and fills an important gap regarding risk factors for smoking among young adults.47

Consistent with other studies,8–10 ,14 we found that living in a neighbourhood with a large number of tobacco retailers was associated with smoking. Unlike one study showing that individuals with the best access (measured as driving time to the closest tobacco retailer) had higher odds of being smokers compared to those with the worst access,19 we did not find a statistically significant association between residential proximity and smoking. Unlike this latter study, we also did not observe a confounding effect of neighbourhood deprivation. This is probably due to the lack of an association between the highest level of residential deprivation and smoking in our sample (Cf table 1), despite the fact that tobacco retailers tended to be more concentrated in more disadvantaged residential neighbourhoods (data not shown).

Our study is innovative in its consideration of individuals’ regular activity settings, including their residential area and responds to calls to examine whether smoking is related to tobacco retailer availability not only in residential neighbourhoods, but also where people spend time.19 We found that the mean number of, and proximity to, tobacco retailers in the activity space were significantly associated with smoking. Clear gradients were observed, with increasing mean counts or proximity being increasingly strongly associated with the likelihood of smoking. These findings were robust across definitions of retailer availability in the activity space. In sensitivity analyses, we found the highest tertiles of the sum and maximum number of tobacco retailers in the activity space to be associated with smoking (prevalence ratio (PR) and 95% CI of 1.33 (1.10 to 1.62) and 1.36 (1.14 to 1.63), respectively) (data not shown). High proximity, measured as the minimum distance separating any residential or activity location from the closest tobacco retailer, was also associated with smoking (PR and 95% CI of 1.47 (1.23 to 1.75)) (data not shown).

Retailer counts and proximity may influence smoking through similar, but also distinct, mechanisms. Generally, in addition to providing opportunities for purchasing cigarettes, tobacco retailers may tempt smokers through provision of visual cues and point-of-purchase advertisement. Two recent studies have suggested that the mere sight of a tobacco retailer could result in an impulse to buy cigarettes or to smoke.22 ,23 In addition, the sight of smokers gathering in proximity to retailers could trigger contagion or normalisation effects.18 High retailer counts may be more specifically indicative of a more price-competitive market, as well as of smoking being more socially acceptable as a practice,7 ,48 while proximity lowers the travel costs of smoking.49 These putative mechanisms linking tobacco retailer availability and smoking have mostly been discussed with regard to residential and school neighbourhoods, but they could also be at play in other settings, as individuals go about their daily activities.19 ,23

Interestingly, the number of tobacco retailers in the residential area seemed more important for smoking than residential proximity, while in the activity space both retailer measures were significantly associated with smoking. This could potentially be explained by the differential depth of knowledge people usually have of their residential neighbourhood compared to their regular activity settings. People tend to have strong knowledge of the diversity of resources found in their residential neighbourhood due to time spent there and to their home being the main anchor from which most trips to other destinations originate.50 They may thus be more aware of the overall availability of tobacco retailers from whom to purchase cigarettes, therefore not necessarily relying on the one closest to home. On the other hand, the areas where regular activities are performed may be more or less thoroughly known to individuals, supporting the idea that both the number of retailers and their proximity, or ‘how close one gets to a tobacco retailer’ in the course of daily activities, may act as a trigger for smoking or a deterrent to quitting.

Overall, our findings suggest that (1) focusing exclusively on the residential area only partially informs us of the link between tobacco retailers and smoking, and (2) other significant places of exposure merit consideration in contextual studies of smoking. Our findings further provide support to recent discussions on zoning policies limiting the sale of tobacco products at large, including in employment or leisure activity nodes.19 ,51

An important strength of our study lies in its focus on young adults involved in a variety of combinations of study, work, shopping and leisure activities, rather than a more narrow focus on ‘students’ and their place of study or ‘workers’ and their workplace. This makes our findings generalisable to a heterogeneous young adult population. Several sensitivity analyses were performed, including aggregating retailer counts within 800 m buffers and limiting our sample to those for whom more than 80% of activity locations were successfully geocoded, and results were found to be robust across model specifications.

Our study is limited in its cross-sectional design, preventing us from making causal claims regarding the links between tobacco retailers and smoking. Given the relatively low response proportion (37.6%), we also cannot discount the possibility that selection bias affected our results, since non-respondents were more likely to be men and to live in the highest quartile of deprivation than respondents. Selective daily mobility bias, whereby residual confounding by unmeasured characteristics of young adults that are related to both smoking and exposure to tobacco retailers in the activity space would explain the associations we observed, could also be at play.52 This is, however, unlikely, since the activity space questionnaire assessed routine activities that had no direct link with smoking. It did not include secondary activities such as visits to the convenience store, where 83.3% of smokers in our sample purchased their cigarettes. Although grocery stores sell cigarettes, most activity locations were unlikely to be tobacco retailers. In this paper, we defined activity spaces as the combination of home as well as out-of-home activity locations.28 This resulted in 4.3% of participants with an activity space equivalent to their residential neighbourhood, which could have led to part of the activity space-smoking associations to be attributable to residential exposure. However, in analyses excluding these participants, results were not found to differ, supporting the importance of the non-residential context for smoking. The sole reliance on a secondary database to locate tobacco retailers is a limitation, in that it excluded other potentially important cigarette purchasing sources such as contraband. However, given that tobacco retailers may not only influence smoking through purchase-related mechanisms, not considering contraband is unlikely to have influenced our findings. We also could not verify that all outlets actually sold cigarettes. Relying on a provincial register of tobacco retailers might have been more reliable. Finally, we operationalised the activity space as a non-contiguous space without considering the routes connecting respondents’ activity locations and along which they may encounter tobacco retailers. This was carried out because regular activity spaces are more likely to be composed of a combination of daily life nodes rather than being continuous, and since there may be more regularity in the places where people conduct activities than in the paths used to travel between them.53


This study extends knowledge regarding the association between tobacco retailer availability and smoking, and adds much needed evidence to the limited literature regarding smoking among young adults from the general population, a subgroup that has suffered from a lack of attention in tobacco research and policy.47 Moving beyond the more traditional work on residential neighbourhoods, our findings suggest that the residential and activity space contexts, as well as measures of counts and proximity, should all be considered when investigating contextual correlates of smoking. Although further research is required to disentangle causal pathways, selection processes and the activity setting(s) most relevant for smoking, we recommend that individuals’ activity space be accounted for in future research on the influence of context on smoking, as it may be particularly insightful when targeting action on settings.

What this paper adds

  • Young adults between 18 and 25 years consistently register the highest smoking prevalence among all age groups.

  • The presence of tobacco retailers in the home and school neighbourhoods has been associated with smoking in youth and adults.

  • Past studies have failed to consider the fact that individuals are mobile and may encounter tobacco retailers in the diversity of settings they experience in the course of their daily activities, that is, in their activity space. This is especially relevant in studies of young adults, since they are a particularly mobile population.

  • Living or conducting activities in areas characterised by high numbers of tobacco retailers is associated with smoking among 18–25-year-olds.

  • While residential proximity to a tobacco retailer is not associated with smoking, conducting activities in places that are, on average, within 150 m, or between 150 and 350 m of such an outlet, is associated with a higher likelihood of smoking.

  • Considering individuals’ regular activity locations in contextual studies of smoking may be particularly insightful for tobacco control interventions targeted at settings, such as zoning policies.


The authors acknowledge Yuddy Ramos and Benoît Thierry for extracting area-level variables; Tarik Benmarhnia for providing comments on an earlier version of this paper; and the ISIS research team as well as participants. MS is the recipient of a Doctoral Research Award from the Canadian Institutes of Health Research (CIHR) in partnership with the Public Health Agency of Canada. KF holds a CIHR New Investigator Award. YK is the recipient of a Young Investigator Award from Fonds de la Recherche du Québec en Santé. Thus work was supported by a CIHR operating grant #DCO150GP.



  • Contributors MS conceptualised the study, analysed the data, interpreted the results and wrote the manuscript. YK, JV, GD and KLF contributed to conceptualising the study and interpreting the results, and read several versions of the manuscript.

  • Funding This work was supported by a CIHR operating grant #DCO150GP.

  • Competing interests None declared.

  • Ethics approval Research Ethics Committee of the Université de Montréal's Faculty of Medicine.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Data are available to bona fide researchers from the investigators on request.