Tobacco Control 1998;7:353–359 353 Tob Control: first published as 10.1136/tc.7.4.353 on 1 December 1998. Downloaded from Gender and ethnic diVerences in young adolescents’ sources of

Leslie A Robinson, Robert C Klesges, Susan M Zbikowski

Abstract function.1 Further, many adolescent smokers Objective—To identify the sources used by will become long-term consumers of young adolescents to obtain cigarettes. products. In fact, research has suggested that Design—In early 1994 a survey assessing between 33% and 50% of those who usual sources of cigarettes and character- experiment with cigarettes escalate to regular istics of the respondents was administered use.3–5 Once tobacco consumption is in homeroom classes. established, these young smokers will be Setting—A large urban, predominantly addicted for an estimated 16 to 20 years.6 African American school system. In an eVort to curb among adoles- Subjects—A population-based sample of cents, all states and the District of Columbia 6967 seventh graders averaging 13 years of have enacted laws banning the sale of tobacco age. to minors. However, active enforcement of Main outcome measure—Reports of usual these tobacco distribution laws has been rare.7 sources of cigarettes. Not surprisingly, then, most children have Results—At this age level, young smokers found cigarettes to be readily accessible. For were more likely to get cigarettes from example, in studies of high school students, friends (31.2%) than buy them in stores between 55% and 82% of those surveyed (14.3%). However, the odds of purchasing thought it would be easy to get cigarettes.8–10 varied for diVerent groups of children. Even among students well below the legal age Regular smokers were much more likely of purchase, access has not been viewed as (48.3%) to have purchased cigarettes than problematic. For instance, over half of a large experimental smokers (9.6%), p<0.001. sample of seventh graders (aged 13 years) Girls were less likely to have bought their recently reported that it would be easy to get cigarettes than boys (p<0.001), and black cigarettes.11 Further, the perception of http://tobaccocontrol.bmj.com/ smokers were less likely to have purchased cigarettes as accessible was a significant risk cigarettes than white children (p<0.001). factor for actual tobacco use: children who Results suggested that family members thought cigarettes were easy to get were more who smoke may constitute a more impor- than twice as likely to experiment with them as tant source of tobacco products than pre- children who viewed cigarettes as less readily viously recognised, particularly for young available.12 Similar results have also been girls. obtained by other researchers examining the In this middle-school sam- Conclusions— relationship between tobacco availability and ple, peers provided the major point of patterns of use among children.13 distribution. However, even at Concern about the continuing availability of this age, direct purchase was not tobacco products and the consequences of uncommon. Sources of cigarettes varied their accessibility has sparked renewed interest significantly with gender, ethnicity, and on September 24, 2021 by guest. Protected copyright. smoking rate. in tobacco regulation. In response, a ( 1998;7:353–359) pioneering Food and Drug Administration (FDA) rule was enacted in the United States in Keywords: gender, ethnicity, adolescents, cigarette 1996 with the aim of reducing tobacco sources consumption in young people by 50% over seven years. In February 1997, the first of its Introduction provisions went into eVect, one which Despite prevention eVorts, cigarette smoking mandated age verification for all over-the- among the young continues to be a major pub- counter sales. (The FDA’s jurisdiction over The University of lic health concern. Although the proportion of tobacco products was called into question by a Memphis Prevention ruling on 14 August 1998, by the Fourth Center, Department of children who smoke declined significantly dur- Psychology, Memphis, ing the 1970s, it stabilised over the next Circuit Court of Appeals. This decision is cur- Tennessee, USA decade1 and in recent years has increased rently stayed pending appeal.) L A Robinson 2 It seems likely that the FDA provision will R C Klesges markedly. For example, in 1991 27.5% of high S M Zbikowski school students smoked, but by 1995 that have its most immediate impact on young number had risen to 34.8%.2 For these young smokers who obtain their cigarettes from Correspondence to: stores. However, how large a proportion of Dr LA Robinson, The smokers the consequences of tobacco use are University of Memphis serious, even in adolescence: Children who adolescent smokers actually buy their Prevention Center, 5350 Poplar Avenue, Suite 430, smoke are more likely to have a variety of cigarettes over the counter remains unclear. To Memphis, Tennessee 38119, health problems, including upper respiratory date, there have been surprisingly few studies USA; l.robinson@ tract infections, reduced lung growth, and of the sources through which children obtain mail.psyc.memphis.edu retardation in the level of maximum lung cigarettes. Further, the studies that have been 354 Robinson, Klesges, Zbikowski Tob Control: first published as 10.1136/tc.7.4.353 on 1 December 1998. Downloaded from

conducted have obtained somewhat inconsist- participation was voluntary. Because the larger ent results. The reasons for these varying study was longitudinal, strict anonymity could results are not entirely clear, but may reflect not be ensured, as some form of identification diVerences in the ages of the children sampled was required to link a given student’s responses or in the exact wording of questionnaires on across years. However, a number of procedures sources. For example, the proportion of young were used to reassure students that their people who directly purchase their cigarettes responses would be confidential. For instance, from any source has been found to vary from students wrote an identification number 41% in a study of high-school students2 to 71% instead of their names on the surveys. Further, in a study of 10–17 year olds.14 Further, a teachers were told not to assist the students recent report2 using the 1995 Youth Risk once the survey began, and after completing Behavior Survey found that only 2.2% of a the questionnaire, the children were instructed large sample of high-school students acknowl- to seal it in an unmarked envelope. These edged using vending machines as a usual envelopes were then sealed in a box source of tobacco, whereas two earlier studies prominently marked “confidential”. Although using ninth9 and tenth grade15 samples the possibility cannot be ruled out that obtained rates of approximately 25% to 50%. smoking was underreported because of student One limitation of most of these studies is that concerns about privacy, it should be noted that their samples consisted primarily of older ado- the rates of smoking reported by our sample lescents, so that little information is available are comparable to those obtained in other regarding the sources through which young large-scale studies.1 teenagers obtain their first cigarettes. Under these conditions students were quite Moreover, with a couple of exceptions,14 16 cooperative: only 3% declined to participate in these studies generally restricted their samples the study. Another 2% of the children were lost to current smokers, such as those smoking at to the study because of parental refusal or our least weekly. Thus, data are needed on the inability to notify the parents of the research. sources used by children engaged in the earliest An additional 16% of the population of 8828 phases of irregular experimental smoking. children did not complete the study because of Such data are critical to develop eVective absenteeism or problems with survey adminis- methods for halting the progression from tration. Notably, in this school system at least casual irregular smoking towards regular 9% of the students are absent on any given day tobacco use. Finally, previous researchers have (K Pruett, personal communication, 20 rarely addressed gender and ethnic diVerences October 1995). Altogether, 79% of the seventh in sources of tobacco. Such diVerences may be graders completed the survey. The characteris- important: the onset of smoking is typically tics of our final sample of 6967 respondents 17 later for girls and young black people, closely match those of the target population in http://tobaccocontrol.bmj.com/ suggesting that they may be less eVective at this school system. The participants averaged accessing the usual routes to tobacco products 13 years of age, with 80.8% black, 16.5% employed by young white males. white, and 2.7% listing other ethnic The purpose of the current study is to assess backgrounds. Males and females were equally the popularity of a variety of sources through represented (49.5% and 50.5%, respectively). which young adolescents obtain cigarettes. To determine whether sources vary with smoking MEASURES frequency, data are presented separately for For the larger study, a questionnaire was experimental smokers and those who consume constructed to measure a number of variables cigarettes regularly. In addition, we explore thought to be related to smoking onset. Among gender and ethnic diVerences in sources of these variables were two dichotomous items

cigarettes. These data are derived from the designed to assess the perceived availability of on September 24, 2021 by guest. Protected copyright. Memphis Health Project, a six-year prospec- cigarettes. On one, students reported whether tive study of the causes of adolescent smoking. or not they believed cigarettes would be easy to obtain. The second item asked children Methods whether or not they thought cigarettes were too OVERVIEW expensive. The entire seventh-grade class in a large, mid- Students were also asked a dichotomous south urban school system comprised the item inquiring about whether or not most of target population. Administrative support for their family members smoked. To allow for the study was excellent, with all 39 eligible non-nuclear family structures—for example, schools participating. Data were collected in children who live with grandparents or other the spring of 1994 in homeroom classes, with extended family members—we intentionally classroom teachers distributing the question- focused on family models in general rather naires to the students. Before collecting the than on parental smoking. Thus, smoking data, passive parental consent was obtained; among any family member, including siblings, that is, parents were notified of the study and is reflected in this item. instructed to call our oYces if they wished to Because we suspected that some sources of remove their child from the study. In addition, cigarettes might be more popular among more children participating in the research signed a deviant youth, a measure of risk-taking and consent form outlining their rights and the rebellious behaviour was developed for use in purpose and procedures of the study. For this study. The scale consisted of five example, the students were told that the survey four-point items scored from 0 to 3, with was about smoking and health, and that higher scores indicating greater rebelliousness Young adolescents’ sources of cigarettes 355 Tob Control: first published as 10.1136/tc.7.4.353 on 1 December 1998. Downloaded from

(Cronbach’s á = 0.78). For example, students Results were asked how often they like to take risks, PREVALENCE OF CIGARETTE USE with responses ranging from “very often” to Results concerning the prevalence of smoking “hardly ever.” Using these same response in our sample have been previously reported.11 options, another item asked students how Briefly, we found that tobacco use was quite frequently it is worth getting into trouble to common, even at this young age. Of those have fun. Scores on the five items were reporting their history of cigarette use (n = averaged, and each student’s average score was 5852), 29.5% were already experimenting with standardised. As expected, analysis revealed cigarettes, and another 4.1% were regular that students scoring high on the rebelliousness smokers. scale felt less supported and successful (Of the total sample of 6967 students, 16% (r[6352] = −0.29, p<0.001), had more friends did not report their smoking status. In who smoked (r[6530] = 0.33, p<0.001), rated comparison with those who did complete this the instrumental value of cigarettes more item, the non-responders were significantly highly (r[6279] = 0.41, p<0.001), and were more rebellious (F[1,6588] = 44.38, p<0.001) more likely to have smoked.12 Thus, the rebel- and more likely to have close friends who liousness measure showed convergent validity smoked (F[1,6878] = 25.90, p<0.001). In with related constructs. addition, non-responders felt less supported Students who smoked were also asked to and successful (F[1,6577] = 69.60, p<0.001), indicate where they usually obtained their and they were more likely to view cigarettes as cigarettes. The following five response options aVordable (OR = 1.22, p<0.05) and as an were provided: (1) “I buy them from a store”; eVective route to enhancing their social image (2) “I buy them from a vending machine”; (3) (F[1,6542] = 9.12, p<0.01). All of these char- “My friends give them to me”; (4) “I take ciga- acteristics have been previously associated with rettes without any adults knowing”; and (5) a greater likelihood of cigarette smoking.12 “other”. Students were encouraged to identify Thus, it should be noted that the base rate of all of the sources that applied. cigarette smoking may be somewhat underesti- Finally, students’ use of cigarettes was meas- mated in this sample.) ured by self-report. These data were then used Further analysis revealed that the probability to group students into three categories. “Non- of experimenting with cigarettes did not vary smokers” included those students who reliably by ethnicity or gender. However, regu- reported never smoking, not even a few puVs. lar smoking was significantly more prevalent “Experimental smokers” consisted of those among boys than girls. Whereas 5.8% of the who admitted to irregular smoking behaviours boys reported smoking at least once a week, ranging from having had a single cigarette “just only 3% of the girls admitted using tobacco at

to try” up to smoking three cigarettes a month. this level. The ethnic diVerence was even more http://tobaccocontrol.bmj.com/ “Regular smokers” were those who reported pronounced: 13.1% of the white children were consuming cigarettes at least weekly. regular smokers, vs only 2.3% of the black youngsters. ANALYSIS In this set of data, three basic questions were NUMBER OF SOURCES OF TOBACCO addressed. First, we calculated the proportions Of the 1969 children who were experimenting of young smokers who used each source, both with cigarettes or smoked regularly, 65.9% for smokers who relied on only a single source acknowledged at least one source of cigarettes. and for the sample as a whole (allowing for (The 34.1% of the children who reported hav- multiple responses). Second, ÷2 analyses were ing smoked without identifying a source were conducted to examine the relationship between much more likely to be irregular smokers than

the use of a given source and level of smoking. were children who did list at least one source on September 24, 2021 by guest. Protected copyright. For example, the proportion of regular (OR = 29.7, p<0.001). These infrequent smokers who bought cigarettes in stores was smokers may not have responded to the compared with that of experimental smokers. sources item because they did not view it as Third, the characteristics of those using diVer- relevant to themselves. As a result, the sources ent sources were explored. For analyses involv- used by infrequent smokers are underrepre- ing continuous data (for example, our measure sented in our data on the total sample; of rebelliousness) analysis of variance was con- however, we should note that, nonetheless, the ducted to identify group diVerences. Dichoto- sample of experimental smokers who did mous data were analysed using ÷2 and logistic respond to the item is quite large (n = 1726).) regression. (In these analyses, we used students The number of sources used by a single rather than schools as the unit of analysis. adolescent ranged from 1 to 5, with most Although this practice is common in studies of young smokers (52.6%) indicating that they adolescent smoking, it should be noted that to had only one usual source of tobacco. the extent that outcomes are correlated within Experimental smokers were significantly less schools, variance may be underestimated and likely to report using multiple sources of tests of significance inflated. However, cigarettes than regular smokers (t[253.80] = intraclass correlations for the various sources 9.01, p<0.001). Black children (who were less of tobacco were modest (ranging from 0.003 to likely to be regular smokers) reported using 0.027), and the group diVerences reported fewer sources than white youngsters here were robust. As such, we opted to report (t[330.99] = 8.05, p<0.001). There was no results from the simpler student-level significant diVerence in the number of sources analyses.) used by boys vs girls. 356 Robinson, Klesges, Zbikowski Tob Control: first published as 10.1136/tc.7.4.353 on 1 December 1998. Downloaded from

Table 1 Usual sources of cigarettes for seventh graders (13 year olds) Thus, students who used direct purchase as their sole route to cigarettes at this early age Prevalence of smoking showed a number of characteristics that have All smokers Experimental smokers Regular smokers been associated with significant and persistent n = 1969 n = 1726 n = 242 tobacco use. Source of cigarettes % (SE) % (SE) % (SE)

“My friends give them to me.” 31.9 (0.01) 28.4 (0.01) 57.0 (0.03) USE OF ALL SOURCES “I buy them from a store.” 14.3 (0.01) 9.6 (0.01) 48.3 (0.03) Although most students reported using only “I buy them from a vending machine.” 7.0 (0.01) 4.2 (0.01) 26.9 (0.03) “I take cigarettes without any adults one source, many youngsters did endorse mul- knowing.” 16.6 (0.01) 14.8 (0.01) 29.3 (0.03) tiple methods for obtaining tobacco. To allow Other source 17.7 (0.01) 16.3 (0.01) 27.7 (0.03) for those who did use more than one source, Subjects were allowed to report multiple sources for getting cigarettes, and some smokers did we recalculated the rates of use for each source, not report their sources. this time including all of those endorsed by the SE = standard error for proportion. students. Table 1 presents these data for the entire smoking sample, as well as for Table 2 Logistic regression models comparing students that use a given source with those experimental and regular smokers separately. that do not As might be expected, all sources of tobacco Source of cigarettes Predictor Odds ratio 95% CI are more frequently used by regular smokers than experimental smokers. For both groups, “My friends give them to me.”* Ethnicity 1.72 1.78–3.43 friends again constituted the most common Gender 1.54 1.19–1.87 Ethnicity × gender 0.61 0.39–0.96 source of tobacco: almost a third of the “I buy them from a store”† Ethnicity 1.72 1.30–2.28 children reported getting at least some of their Gender 1.54 1.19–2.00 “I buy them from a vending machine.”‡ Ethnicity 4.31 2.99–6.20 cigarettes from peers. In addition, a substantial Gender 2.42 1.63–3.58 16.6% of the sample admitted to taking “I take cigarettes without any adults Ethnicity 1.04 0.78–1.40 cigarettes without an adult’s knowledge. knowing.”§ Gender 0.61 0.48−0.79 Results further indicated that rates of direct n = 1916. The dependent variables are coded so that use of a source = 1 and its non-use = 0. purchase varied considerably, depending on Ethnicity is coded with white students = 1 and black students = 0. Gender is coded with males how regularly a child smoked. For example, = 1 and females = 0. *Model ÷2[3] = 44.82, p<0.001. among the experimental smokers only 9.6% †Model ÷2[2] = 25.80, p<0.001. reported that they usually bought their ‡Model ÷2[2] = 84.04, p<0.001. cigarettes in stores, and vending machine usage §Model ÷2[2] = 15.00, p<0.001. CI = confidence intervals. was even less common (4.2%). In contrast, almost half of the regular smokers bought their USE OF SINGLE SOURCES cigarettes in stores, and over 25% of them When children get cigarettes from a single reported using vending machines regularly.

source, which one are they most likely to use? Thus, the regular smokers were significantly http://tobaccocontrol.bmj.com/ To address this issue we examined data for the more likely to purchase cigarettes directly from 1035 young smokers who reported that they a store (÷2[1] = 260.31, p<0.001) or a vending used only one source for cigarettes (students machine (÷2[1] = 166.82, p<0.001) than were endorsing multiple sources were excluded less frequent smokers. from this analysis). The results indicated that peers were the most popular source, with GENDER AND ETHNIC DIFFERENCES IN SOURCES 38.4% of the students reporting that they got OF CIGARETTES their cigarettes only from friends. Another To assess gender and ethnic diVerences in 16.3% of the children who endorsed only a sources of cigarettes, a series of logistic regres- single source reported that they took them sions were conducted, using ethnicity, gender, without an adult’s knowledge. Surprisingly, a and their interaction to predict use of each sizeable number of these children reported source. The results of these analyses for each on September 24, 2021 by guest. Protected copyright. buying their cigarettes exclusively, either in a source of cigarettes are presented in table 2. store (15.0%) or vending machines (4.3%). An Turning first to the model predicting students additional 26% of the youngsters reported who used friends as a source, one can see that using some other single source. the main eVects for ethnic group and gender Some might wonder whether particular were both significant, along with their interac- sources were accessed by diVerent types of tion. Follow up of the interaction revealed that children. To explore this possibility, we black boys were 49% more likely to get compared the characteristics of students who cigarettes from their friends than were black used diVerent routes to tobacco products. The girls (OR = 1.49, p<0.001). In contrast, there results indicated that youngsters who bought was no gender diVerence among white their cigarettes (either in stores or vending students (OR = 0.91, p>0.05). machines) were more likely to be regular For the other sources reported in table 2, the smokers than those who used other sources interaction of ethnicity and gender was not sig- (÷2[1] = 72.64, p<0.001). In addition, young nificant; thus, in each equation the interaction purchasers were more rebellious than those term was dropped in favour of the more parsi- who got cigarettes through other means monious main eVects model. As can be seen, (F[1,1123] = 29.59, p<0.001), and they felt both stores and vending machines were much less supported and successful (F[1,1115] = more likely to be used by white children than 18.28, p<0.001). Further, students who by young black smokers. Similarly, boys were bought cigarettes were more likely to view significantly more likely than girls to buy ciga- them as aVordable (÷2[1] = 12.64, p<0.001) rettes at both of these outlets. The odds of and easily accessible (÷2[1] = 9.75, p<0.01). stealing cigarettes, on the other hand, did not Young adolescents’ sources of cigarettes 357 Tob Control: first published as 10.1136/tc.7.4.353 on 1 December 1998. Downloaded from

vary by ethnicity. However, there was a main stealing cigarettes (19.6%) than those with eVect of gender, such that girls were non-smoking families (14.6%) (÷2[1] = 8.54, significantly more likely to report that they p<0.01). However, this eVect was limited to have taken cigarettes without an adult’s knowl- girls: adolescent girls whose families smoked edge. This last result held true even when gen- were significantly more likely to have taken der diVerences in smoking rates were taken cigarettes (23.5%) than those with non- into account. For example, among experimen- smoking families (17.1%) (÷2[1] = 6.31, tal smokers, 18.1% of the girls reported p<0.05). In contrast, the eVect of family smok- stealing cigarettes compared with only 11.3% ing on stealing was not significant for boys of the boys (÷2[1]= 15.85, p<0.001). Similarly, (÷2[1] = 1.22, p>0.05). among regular smokers more girls (37%) than boys (25%) reported taking cigarettes without Discussion permission (÷2[1] = 3.89, p<0.05). To date few studies have provided data on the sources through which young adolescents UNDERSTANDING THE GENDER DIFFERENCE IN obtain tobacco products. The present STEALING investigation was designed to address this Why might girls be more likely to steal issue. Results indicated when all sources were cigarettes? One possibility involves gender considered, a substantial 14% of our sample of diVerences in the personality characteristics of seventh graders routinely purchased cigarettes young smokers. For example, the girls that do in stores, and 7% used vending machines regu- smoke at this young age might be more larly. These findings are consistent with the rebellious than the boy smokers, for whom the large body of research indicating that even very behaviour is more normative in early young children have little diYculty buying adolescence. However, analysis revealed that cigarettes.10 16 18–22 Notably, these data were col- girl smokers were generally less rebellious lected before the FDA regulation requiring (mean = 0.35) than boy smokers (mean = identification during purchase went into eVect. 0.71; F[(1, 1895] = 34.63, p<0.001). Further, As we track these children over the next few there was no significant interaction of gender years, we hope to see these numbers decline in and the tendency to take cigarettes. response to tighter tobacco control regulations. Another reason for girls’ greater propensity Our results further indicated that, at this age for taking cigarettes might involve their inabil- level, by far the majority of young smokers do ity to aVord them. If so, then perceptions of not obtain cigarettes directly from stores. cost should be related to the probability of Instead, peers provide the major point of stealing. However, analysis revealed that distribution at this age, with almost one in perception of cigarettes as too expensive was three young smokers reporting their friends as

unrelated to the probability of stealing both in a usual source. Furthermore, we found the http://tobaccocontrol.bmj.com/ the sample as a whole (÷2[1] = 0.10, p>0.05) prevalence of direct purchase was tied to and for boys and girls separately (÷2[1] = 0.43, smoking frequency; specifically, regular p>0.05 and ÷2[1] = 0.03, p>0.05, respec- smokers were much more likely to buy their tively). cigarettes than experimental smokers. Similar We then considered whether family smoking findings have been obtained by other research- might provide the key to understanding girls’ ers using older samples.28916For example in a propensity for taking cigarettes. Rather than study of high-school students, 35.2% of the stealing them from stores, these children might regular smokers used stores, compared with be taking cigarettes from family members only 15.9% of the smokers.2 Taken without permission. If so, then children whose together, these findings suggest that as children family members smoked would probably be progress toward heavier addiction they are

more likely to view cigarettes as easy to get more likely to resort to buying cigarettes. Thus, on September 24, 2021 by guest. Protected copyright. than those with non-smoking families. Indeed, the most immediate (but not the only) eVect of analysis of our sample supported just such a age verification enforcement might be in relationship: whereas 54.7% of the children preventing the escalation of casual experimen- whose family members smoked believed it tation to regular smoking. However, as of yet would be easy to get cigarettes, only 45.5% of few studies have examined the actual eVect of those with non-smoking families thought tobacco regulation on adolescent smoking access would not be problematic (÷2[1] = habits, and the limited findings to date have 50.72, p<0.001). Furthermore, the relation- proven contradictory. For example, one study ship between family smoking and perceived found that legislation restricting cigarette sales ease of access appeared to be much stronger did significantly decrease adolescent tobacco for girls than boys. For example, only 41.9% of use and that, as might be expected, the girls with non-smoking families thought decrease was greater for regular smoking than cigarettes would be easy to get, compared with experimental use.23 However, another recent 55.3% of girls with smoking families (÷2[1] = investigation found that well-enforced tobacco 55.34, p<0.001). In contrast, 49.1% of boys sales laws did not alter adolescents’ perceived with non-smoking families vs 54.1% of those access to tobacco or their smoking habits.24 with smoking families thought that access Thus, further evaluation research on the would be easy (÷2[1] = 7.18, p<0.01). empirical eVects of such legislation is clearly Are children whose families smoke more needed. likely, then, to report taking cigarettes? Our In addition to getting cigarettes from friends results confirm that youngsters with smoking and merchants, a surprisingly large proportion families were indeed more likely to report of our sample (almost 17%) reported that they 358 Robinson, Klesges, Zbikowski Tob Control: first published as 10.1136/tc.7.4.353 on 1 December 1998. Downloaded from

had taken cigarettes without an adult’s knowl- ple we found that white girls whose families edge. In contrast, only 4.2% of students smoked were over three times more likely to try responding to the recent Youth Risk Behavior cigarettes than girls with non-smoking families. Survey (CDC) admitted stealing tobacco In contrast, family smoking was unrelated to products,2 and in another investigation, only experimental tobacco use for white boys. 9.3% reported shoplifting cigarettes.25 Al- Further, for both genders, family smoking had though the reasons for our diVering results are little impact on the progression to regular not clear, it should be noted that these studies smoking. Thus, the primary impact of family diVered from ours in three important respects. smoking on girls appeared to lie in prompting First, in both of the previous studies, youths their initial experimentation with cigarettes.12 were allowed to report only one source for Obviously, both boys and girls have cigarettes. It seems reasonable to expect that opportunities to take cigarettes from parents relatively fewer adolescents would use theft as and siblings who smoke. Why, then, might girls their primary means of access than as an ancil- be particularly responsive to these opportuni- lary method. Second, the adolescents surveyed ties? Part of the answer may lie in social in the previous studies were in high school attitudes towards tobacco use among girls. rather than middle school, and older children Although smoking in women has become may be less likely to steal cigarettes because increasingly acceptable over the past few direct purchase is easier for them. Third, the decades, there is still evidence that young girls previous studies225 inquired about stealing per feel more pressure to remain smoke-free than se, whereas our own question was much less boys.31 For example, in this sample of children, direct (we asked students whether they had girls viewed cigarettes as having less social “taken” cigarettes without adults’ knowledge). reinforcement value than did boys,11 and other This diVerence in wording may be important researchers have found that girls are more to students, in that they may be less likely to likely to believe their parents would be upset if acknowledge directly an illegal act. they smoked that boys.32–34 Not surprisingly, Our analyses also revealed clear ethnic and then, girls have been found to be more likely to gender diVerences in sources of tobacco that keep their smoking a secret,35 and they paralleled smoking rates. For example, white generally view cigarettes as being harder to get youngsters and boys (groups more likely to than do boys.10 11 Because they are less smoke regularly) were more likely to get comfortable accessing the public sources of cigarettes from friends and to buy their tobacco that boys favour, young girls may be cigarettes directly than were black youngsters particularly responsive to opportunities to and girls. The gender diVerence in buying rate secretly “lift” cigarettes from family members. was somewhat surprising, in that a number of Conversely, boys may be less likely to engage in

studies have demonstrated that underage girls this behaviour because they are more comfort- http://tobaccocontrol.bmj.com/ are actually more successful at buying tobacco able with public purchasing. than boys.10 18 21 However, just because girls can Overall, our findings have important buy cigarettes does not mean that they do. implications for tobacco control. For example, Although the success rate per attempt may be given that regular smokers are more likely to be lower for boys, through sheer persistence they purchasers, they appear to be the group most may actually complete more purchases than do likely to be immediately aVected by the FDA girls. age verification procedures. Experimental ciga- Whereas girls were less likely to buy rette use may also be influenced, but additional cigarettes or to get them from friends, our data steps may be required to significantly reduce indicated that they were more likely than boys lower-level tobacco consumption. For exam- to take cigarettes without an adult’s ple, public health eVorts to discourage

knowledge. This proclivity toward stealing was cigarette access through the family may consti- on September 24, 2021 by guest. Protected copyright. not tied to greater rebelliousness among girls tute a powerful supplement to legislative nor to concerns about the costs of cigarettes. control, especially in light of other research However, family smoking did appear to play an suggesting that tobacco acquisition through important role. Specifically, we found that family members is common.14 25 Toward this young girls whose family members smoked end, further research is needed to verify more were more likely to have stolen cigarettes than specifically the typical sources of tobacco those with non-smoking families. Among boys, within the family. For example, do young girls however, family smoking was unrelated to primarily take cigarettes from parents who reports of taking cigarettes. This pattern of smoke, or are older siblings the main source? findings is similar to that obtained in a recent By clarifying this issue, educational eVorts to survey of 230 tenth graders.10 In that investiga- stop access to tobacco through the family can tion, 41% of girls vs only 18% of boys whose be targeted more precisely. parents smoked got their cigarettes from home. Several limitations of the present study Viewed together, these results suggest that should be acknowledged. In this investigation family members who smoke provide an impor- only a few sources for cigarettes were tant source of tobacco for young girls. examined, and our data suggested that The significance of these findings becomes additional routes to tobacco products are clearer when viewed against the backdrop of important to consider. For example, almost previous research. Studies have repeatedly 18% of our sample indicated that they had shown that family smoking is more likely to obtained cigarettes through a method other prompt tobacco use in girls than boys.12 26–30 than peers, stores, vending machines, and For example, in other analyses using this sam- theft. It may be, for example, that a substantial Young adolescents’ sources of cigarettes 359 Tob Control: first published as 10.1136/tc.7.4.353 on 1 December 1998. Downloaded from

number of children receive cigarettes as gifts 11 Robinson LA, Klesges RC. Ethnic and gender diVerences in risk factors for smoking onset. Health Psychol 1997; from adults. Thus, future investigations should 16:499–505. explore other possible sources of tobacco 12 Robinson LA, Klesges RC, Zbikowski SM, et al. Predictors of risk for diVerent stages of adolescent smoking in a bi- products. Second, sources should be specified racial sample. J Consult Clin Psychol 1997;65:653–62. more precisely. Although our findings 13 Bauman KE, Fisher LA, Bryan ES, et al. Antecedents, sub- regarding the relationship between parental jective expected utility, and behavior: a panel study of ado- lescent cigarette smoking. Addict Behav 1984;9:121–36. smoking and students’ reports of “taking” 14 DiFranza JR, Eddy JJ, Brown LF, et al. Tobacco acquisition cigarettes are interesting, further research and cigarette brand selection among youth. Tobacco Control 1994;3:334–8. directly inquiring about theft from parents vs 15 Forster JL, Klepp K.-I, JeVery RW. Sources of cigarettes for stores is needed. Third, defining frequency of tenth graders in two Minnesota cities. Health Educ Res use more clearly would be helpful. In this Theory Pract 1989;4:45–50. 16 Forster JL, Wolfson M, Murray DM, et al. Perceived and study, we asked students to endorse all sources measured availability of tobacco to youths in 14 Minne- sota communities: the TPOP Study. Am J Prev Med 1997; that they usually used; however, students may 13:167–74. vary in what they consider to be a “usual” 17 Klesges RC, Robinson LA. Predictors of smoking onset in source. A more precise method would be to adolescent African American boys and girls. J Health Educ 1995;26:1–7. inquire separately about all sources that have 18 US Centers for Disease Control. Minors’ access to ever been used and then about the single most tobacco—Missouri and Texas. MMWR 1993;42:125–8. 19 US Centers for Disease Control. Minors’ access to cigarette frequently used source. Finally, data are vending machines—Texas. MMWR 1994;43:625–7. needed to determine how children’s sources of 20 DiFranza JR, Norwood BD, Garner DW, et al. Legislative tobacco products change as they age. By eVorts to protect children from tobacco. JAMA 1987; 257:3387–9. understanding the access routes at each age 21 Erickson AD, WoodruV SI, Wildey MB, et al. Baseline level, more eVective legislative and educational assessment of cigarette sales to minors in San Diego, Cali- fornia. J Commun Health 1993;18:213–24. programmes can be designed to prevent the 22 Wildey MB, WoodruV SI, Agro A, et al. Sustained eVects of onset of cigarette smoking throughout educating retailers to reduce cigarette sales to minors. adolescence. Public Health Rep 1995;110:625–9. 23 Jason LA, Ji PY, Anes MD, et al. Active enforcement of ciga- rette control laws in the prevention of cigarette sales to This study was supported by a grant (HL50723) form the minors. JAMA 1991;266:3159–61. National Heart, Lung, and Blood Institute (NHLBI). Support 24 Rigotti NA, DiFranza JR, Chang Y, et al. The eVect of was also received from the Centers of Excellence awarded by enforcing tobacco-sales laws on adolescents’ access to the State of Tennessee to the Department of Psychology at the tobacco and smoking behavior. N Engl J Med 1997; University of Memphis. 337:1044–51. 25 Wildey MB, WoodruV SI, Pampalone SZ, et al. Self-service sale of tobacco: how it contributes to youth access. 1 US Department of Health and Human Services. Tobacco Preventing Control 1995;4:355–61. tobacco use among young people. A report of the Surgeon Gen- eral, 1994. Atlanta, Georgia: Public Health Service, Cent- 26 Chassin L, Presson CC, Sherman SJ, et al. Changes in peer ers for Disease Control and Prevention, OYce on Smoking and parent influence during adolescence: longitudinal ver- and Health, 1994. (US Government Printing OYce Publi- sus cross-sectional perspectives on smoking initiation. Dev cation No S/N 017-001-00491-0.) Psychol 1986;22:327–34. 2 US Centers for Disease Control and Prevention. Tobacco 27 Malkin SA, Allen DL. DiVerential characteristics of adoles- cent smokers and non-smokers. 1980;10:437– use and usual source of cigarettes among high school J Fam Pract http://tobaccocontrol.bmj.com/ students—United States, 1995. MMWR 1996;45:413–8. 40. 3 Hirschman RS, Leventhal H, Glynn K. The development of 28 Stanton WR, Silva PA. Children’s exposure to smoking. Int smoking behavior: conceptualization and supportive cross- J Epidemiol 1991;20:933–7. sectional survey data. J Appl Soc Psychol 1984;14:184–206. 29 Swan AV, Creeser R, Murray M. When and why children 4 McNeill AD. The development of dependence on smoking first start to smoke. Int J Epidemiol 1990;19:323–30. in children. Br J Addict 1991;86:589–92. 30 Williams AF. Personality and other characteristics associ- 5 Henningfield JE, Cohen C, Slade JD. Is nicotine more ated with cigarette smoking among young teenagers. J addictive than cocaine? Br J Addict 1991;86:565–9. Health Soc Behav 1973;14:374–80. 6 Pierce, JP, Gilpin, E. How long will today’s new adolescent 31 Murray M, Cracknell A. Adolescents’ views on smoking. J smoker be addicted to cigarettes? Am J Public Health Psychosom Res 1980;24:243–51. 1996;86:253–6. 7 US Department of Health and Human Services, OYce of 32 Kelson SR, Pullella JL, Otterland A. The growing epidemic: a survey of smoking habits and attitudes toward smoking Inspector General. Youth access to tobacco. Bethesda, Mary- land: Department of Health and Human Services, Public among students in grades 7 through 12 in Toledo and Health Service, National Institutes of Health, 1992. (NIH Lucas County (Ohio) public schools. Am J Public Health Publication No. OEI-02-91-00880.) 1975;65:923–38. 8 US Centers for Disease Control. Accessibility of cigarettes 33 Palmer AB. Some variables contributing to the onset of to youths aged 12–17 years—United States, 1989. cigarette smoking among junior high school students. Soc MMWR 1992;41:485–8. Sci Med 1970;4:359–66. on September 24, 2021 by guest. Protected copyright. 9 Cummings KM, Sciandra E, Pechacek TF, et al. Where 34 Urberg K, Robbins RL. Adolescents’ perceptions of the teenagers get their cigarettes: a survey of the purchasing costs and benefits associated with cigarette smoking: sex habits of 13–16 year olds in 12 US communities. Tobacco diVerences and peer influence. J Youth Adolesc 1981; Control 1992;1:264–7. 10:353–61. 10 Forster JL, Hourigan M, McGovern P. Availability of 35 Goddard E. Why children start smoking: an enquiry carried out cigarettes to underage youth in three communities. Prev by Social Service Division of OPCS on behalf of the Med 1992;21:320–8. Department of Health. London: HMSO, 1990.