Content uploaded by Harris Hyun-Soo Kim
Author content
All content in this area was uploaded by Harris Hyun-Soo Kim on Aug 16, 2021
Content may be subject to copyright.
© The Author 2015. Published by Oxford University Press on behalf of the Society for Research on Nicotine and Tobacco. All rights reserved.
For permissions, please e-mail: journals.permissions@oup.com.
934
Nicotine & Tobacco Research, 2016, 934–942
doi:10.1093/ntr/ntv172
Original investigation
Advance Access publication August 13, 2015
Original investigation
Examining the Effects of Parental Influence on
Adolescent Smoking Behaviors: AMultilevel
Analysis of the Global School-Based Student
Health Survey (2003–2011)
Harris Hyun-sooKim PhD1, JongSerlChun PhD2
1Department of Sociology, Ewha Womans University, Seoul, Republic of Korea; 2Department of Social Welfare, Ewha
Womans University, Seoul, Republic of Korea
Corresponding Author: Harris Hyun-soo Kim, PhD, Department of Sociology, Ewha Womans University, 52 Ewhayeodae-gil,
Seodaemun-gu, Seoul 120-750, Republic of Korea. Telephone: 822-3277-4622; Fax: 822-364-8019; E-mail: harrishkim@ewha.ac.kr
Abstract
Introduction: Based on a large cross-national dataset, we investigated the role of parental smoking
(a risk factor) and parental supervision (a protective factor) on the frequency of smoking by youths
in resource-poor countries. In addition, we tested for cross-level interactions between these two
predictors and national wealth on the outcome variable.
Methods: Pooled cross-sectional data from the Global School-based Student Health Survey (2003–
2011) were analyzed, which consists of 58 956 students in 31 countries. Hierarchical linear models
were estimated to examine the associations between the two parental influence variables and
adolescent smoking.
Results: Among the control variables, age, gender (male), the experience of being bullied, fre-
quency of getting into physical fights, truancy, and anxiety were significantly related to higher
frequency of smoking. With respect to the main predictors, both at the individual level, parental
supervision was negatively associated with adolescent smoking, while parental smoking was posi-
tively related to it. Two cross-level interaction terms were also observed. National wealth (GDP per
capita) significantly moderated, that is, increased, these effects of parental influence on how often
the adolescents smoked.
Conclusions: We provided new evidence on the factors related to adolescent smoking in low-
income countries, a topic that has received very little attention. We showed that the associations
between parental influences and adolescent smoking behaviors are not constant but vary accord-
ing to the level of economic development. Future research should incorporate this comparative
dimension in elaborating and specifying the conditions under which parental influences and other
predictors differentially affect adolescent smoking.
Implications: Prior research on adolescent smoking focused on developed countries. Based on
the secondary analysis of the Global School-based Student Health Survey (2003–2011), this study
examines the associations between parental influence (parental smoking and parental supervi-
sion) on the frequency of youth smoking behaviors in resource-poor countries. We show that
parental smoking is positively related to adolescent smoking, while parental supervision is nega-
tively related to it. We also find that these two associations vary according to national wealth: both
effects are stronger in a country with higher per capita GDP.
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
935Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
Introduction
Tobacco use is one of the major preventable causes of death through-
out the world.1 Approximately 90% of smokers begin to use tobacco
at the age of 18 or younger, and most smokers become daily smokers
during their adolescence.2 Adolescent smoking increases the likeli-
hood of acute health problems, including respiratory illness, asthma,
decreased tness, reduced pulmonary function, and delayed lung
development.2
It is estimated that tobacco-related deaths will increase from 5.4
million to 8 million by 2030, 80% of which will occur in low- and
middle-income countries.3 Approximately 80% of 1 billion smok-
ers live in developing countries.4 Furthermore, 250 million children
and adolescents are expected to die from tobacco-related causes,
mostly in developing countries.5 Although adult smoking prevalence
has declined in developed countries during the past half century,6
tobacco use has become a serious public health concern in low- and
middle-income countries.7 Despite these facts and trends, relevant
data and research on smoking behaviors have been largely limited to
and focused on developed countries.8,9
Understanding the determinants of adolescent smoking is one
of the major goals in global public health10 for developing effective
prevention and cessation strategies.11 Adolescent smoking has sev-
eral common global determinants. One of the most important and
universal factors is the role of parental inuence. Specically, paren-
tal smoking has been shown to be a signicant predictor of ado-
lescent smoking initiation and current smoking in both developed
and developing countries.12,13 Prior research indicates that parental
inuences on smoking decrease during the span of adolescence,
whereas peer inuences increase.14,15 In comparison, a study by
Bricker etal.16 reports that the inuence of peer smoking on smok-
ing transitions was stable during adolescence, while the inuence of
parental smoking on the transition to daily smoking signicantly
increased. In addition, parental control or supervision is shown to
be a powerful indicator of adolescent smoking behavior in many
parts of the world.17,18 In high-income countries, attachment to par-
ents19 and parental style20 signicantly affect the probability and fre-
quency of adolescent smoking. In a study that surveyed 33 mostly
afuent countries including Canada, Israel, and European nations,
lower family afuence was found to be a risk predictor of adolescent
smoking.21 By contrast, higher household income was associated
with greater smoking prevalence in a study conducted of adolescents
in low-income countries.22
There are other globally common determinants of adolescent
smoking. Gender difference, for example, is frequently witnessed
cross-nationally.9 Age is another risk predictor for adolescent smok-
ing.23,24 In addition, loneliness,25 depression,11,26 and anxiety11,24
have all been shown to raise the odds of adolescent smoking across
countries at varying levels of national wealth. Empirical evidence
also suggests that school truancy is a signicant factor of adolescent
smoking, but only in high-income countries.23,27 Breslau28 reports that
deviant behaviors such as truancy, starting ghting, being expelled or
suspended from school, and running away from home were similarly
related to early smoking initiation and nicotine dependence among
adolescents. Bullying victims, in particular, are also more likely to
start smoking to obtain peer acceptance or to cope with stress and
depression.29 In a related vein, Dijk etal.30 found that greater sup-
port from peers signicantly reduced adolescent tobacco use in six
European countries. Thorlindsson and Vilhjalmsson,31 on the other
hand, show that peer support was positively associated with smok-
ing, which may be caused by the smoking status of friends.32 Having
fewer friends was further associated with more smoking among
adolescents.33
According to the ecological model,34 the process of human
development takes place through the reciprocal interaction between
individuals and their immediate as well as remote environment.
Immediate environment refers to the likes of family and school,
whereas more remote environment includes institutional patterns
of culture such as economic activities, customs, and life-styles. In
line with this overarching theoretical framework, an increasing
number of studies show that tobacco prevention or cessation needs
to consider interactions among various determinants at individual,
family, peer, community, and national economic levels.35,36 Flay and
Clayton,36 for example, highlight the importance of simultaneously
considering the interactions between individual and contextual fac-
tors for explaining behavioral patterns. Despite the growing recogni-
tion of their importance, however, only a limited amount of research
has actually investigated the interactive effects among determinants
across multiple domains.37–39 Moreover, little attention has been paid
to the cross-level interactions between the predictors of adolescent
smoking in particular.40
Several exceptions exist in the literature. Thrul et al.40 found
that the prevalence of adult smoking in the community moderated
the relationship between friends’ smoking behavior and adolescent
smoking. Specically, adolescent smoking prevalence had a stronger
association in cities, indicating higher adult smoking prevalence.
Kelly etal.41 examined the inuence of parents, siblings, and friends
on adolescent smoking after accounting for school- and community-
level variances. But no signicant cross-level interactions between
adult community norms/laws about substance use and family or
peer inuences were reported. It has also been shown that among
European and North American countries, national wealth signi-
cantly moderates the effect of family afuence on youth smoking.42
Another study shows that an increase in the country’s economic sta-
tus (GDP) widens educational inequalities in smoking.8 Virtually no
study, however, has examined the interactions between individual-
level and contextual-level variables underlying adolescent smoking
in resource-poor countries, especially with respect to the role of
national wealth.40,42
Due to reasons related to data limitations, previous research has
been largely conned to economically advanced nations. Moreover,
a large proportion of past studies focus on adult populations. Amain
objective of this study, therefore, is to shed light on parental inu-
ences and other key factors linked with youth smoking behaviors
in low-income countries. With limited exceptions, the vast majority
of the extant scholarship also consists of case studies dealing with
single nations. This study takes a more comprehensive approach
by analyzing a pooled cross-sectional dataset consisting of multi-
ple countries. Specically, we concentrate on the impact of parental
smoking behavior (ie, a risk factor) and that of the parent supervi-
sion (ie, a protective factor) on the frequency of adolescent smoking.
We explicitly set out to test whether these relationships hold in the
context of developing countries. In addition, we provide evidence on
the extent to which the above two parental measures interact with a
key national-level economic characteristic (per capita GDP) in inu-
encing smoking behavior of youths.
To our knowledge, this is one of the rst and very few cross-
national investigations of these issues in the context of resource-poor
countries. The current research is guided by the following questions:
(1) in the developing world, to what extent does parental supervi-
sion reduce the frequency of adolescent smoking; (2) to what extent
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
936 Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
does parental smoking (measured at individual and national levels)
increase the frequency of adolescent smoking; (3) and, lastly, how
does the level of national wealth as measured by per capita GDP
moderate the strengths of the above two hypothesized associations?
Methods
Sample
To examine these questions, we analyzed secondary data from the
Global School-based Student Health Survey (GSHS),43 a collaborative
surveillance project implemented mainly in resource-poor countries
since 2003 (www.cdc.gov/gshs). The research was designed to assess
and gauge behavioral risk and protective factors related to adoles-
cent health and mortality. The ultimate goal of the GSHS is to assist
national governments in underdeveloped parts of the world to work
with international agencies in formulating and providing better youth
healthcare programs and policies. It was developed by the World Health
Organization in collaboration with UNICEF and UNESCO, with tech-
nical assistance from the US Centers for Disease Control and Prevention.
Core-questionnaire modules, core-expanded questions, and
country-specic questions are combined to create a self-administered
questionnaire, which is distributed to students and completed in one
regular class period under the supervision of teachers. Atwo-stage
cluster sample design was used to produce data representative of all
students in each country. At the rst stage, schools were chosen with
probability proportional to enrollment size. At the next stage, classes
were randomly selected, with all students in those classes eligible to
participate in the survey. The school response rates ranged from 83
to 100 percent; student response rates varied between 71 to 99 per-
cent (more details are available at www.who.int/chp/gshs/factsheets/
en). Ethical approval was obtained according to the guideline in each
country prior to data collection. Individual country datasets collected
between 2003 and 2011, excluding those that do not contain infor-
mation on smoking behavior, were pooled to create a single data le
consisting of 58 956 students (28 299 boys and 30 657 girls) in 31
low-income countries (Antigua, Argentina, Benin, Botswana, Chile,
China, Costa Rica, Ecuador, Grenada, Guyana, Indonesia, Jamaica,
Jordan, Kenya, Kiribati, Macedonia Maldives, Mauritania, Morocco,
Pakistan, Peru, Philippines, Saint Lucia, Saint Vincent, Samoa,
Solomon Islands, Suriname, Thailand, Trinidad, Tunisia, and Uganda).
Measures
Outcome Variable
The dependent variable of interest is the frequency of adolescent
smoking. In the GSHS, the students were asked to state how often
they had smoked during the past month. The responses were origi-
nally coded on a 7-point scale (eg, 1=0days, 2=1 or 2days, 3=3
to 5days,… 6=20 to 29days, 7=All 30days). The distribution of
the data is positively skewed, with the vast majority of the student
respondents being nonsmokers (89%). The original answers were
log (base 10)transformed to make the distribution less skewed and
to help meet the assumptions of inferential statistics.
Main Predictors
Two main predictors of adolescent smoking were considered:
whether one of the parents/guardians or both smoke and the quality
of the relationship between parents and children. To measure the
former, the following question was used: “Which of your parents or
guardians use any form of tobacco?” (1=Neither, 2=My father or
male guardian, 3=My mother or female guardian, 4=Both). The
responses were dichotomized where the choice “Neither” was coded
as 0 and 1 otherwise (ie, if at least one of the parents or guardians
was a smoker). This variable (Parental smoking) is also conceptual-
ized at the country level, since the proportion of adult smoking, as a
proxy measure for general smoking norm, has been known to shape
adolescent smoking behavior at the individual level.39 This aggre-
gate variable (Adult smokers) is created by adding up the student-
level responses and then calculating the national average. For the
latter measure concerning the adult monitoring of children (Parental
supervision), three separate survey items were combined: During the
past 30days, how often did your parents or guardians check to see
if your homework was done?”; “During the past 30days,… under-
stand your problems or worries?”; and “During the past 30days,…
really know what you were doing with your free time?” The answers,
which were originally coded on a 5-point scale (eg, 1 = Never,
3 = Sometimes, 5 = Always), were merged to create a scale vari-
able (Cronbach’s alpha=.67). To examine any possible cross-level
interactions, we also considered national wealth in terms of logged
per capita GDP based on the purchasing power parity converted to
US dollars from the World Bank’s Development Indicators database,
which is the average value of all goods and services produced in a
country valued at prices prevailing in the United States (http://data.
worldbank.org).44 Table1 reports country-specic description of the
outcome and main predictor variables, as well as samplesize.
Sociodemographic and Other Background Variables
In keeping with earlier research9,11,23–25,30 several background factors
were considered in examining the independent and cross-level asso-
ciations between adolescent smoking and the main predictors. Two
demographic factors included were age and gender. Whether the stu-
dent respondent had been a victim of bullying and, as a proxy for
family economic background, how often she/he had gone hungry in
the past month were also controlled for. In addition, variables meas-
uring the frequency of truancy and physical ghting were considered,
as were two psychological variables gauging the subjective levels of
loneliness and anxiety. As protective factors, two social capital vari-
ables (peer support and friendship size) were also included. Finally,
to control for the possible confounding effect of the time of data col-
lection, which varies from 2003 to 2011, the survey year was taken
into account. The weighted descriptive statistics for all the variables
used are presented in Table2. The question wording and the coding
schemes for variable construction are summarized in Table3.
Statistical Analyses
The structure of data in GSHS (2003–2011) are such that individual
students are nested across multiple countries, creating a methodo-
logical problem of nonconstant variance across contextual units
(ie, countries) and, as a result, underestimation of standard errors.45
To correct for this, we ran multilevel models using the latest ver-
sion of HLM,46 which adjusts for clustering of individual observa-
tions within countries and also accounts for country-level effects.
Following a standard procedure to avoid collinearity problems47,48
all nondichotomous student-level variables were group-mean cen-
tered; all country-level variables were grand-mean centered. As a rst
step, a one-way analysis of variance was performed to see whether
the variability in the outcome variable (adolescent smoking) across
countries is signicantly different from zero. The chi-square test (P <
.001) from the null model (not shown) without any of the covariates
showed a signicant variation in adolescent smoking across coun-
tries, thereby justifying the use of hierarchical linear modeling. Three
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
937Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
nested models were estimated to empirically test the aforementioned
research questions. Model 1 includes only the level-1 covariates;
Model 2 introduces the two variables measured at the country level,
GDP per capita and the proportion of parents who smoke. Finally,
Models 3 and 4 incorporate each of the two cross-level interaction
terms. Consistent with recent cross-national studies on adolescent
health,49–51 we estimated linear multilevel regression models, which
are formally expressedas:
Individual-levelmodel:
Yij=β0j + β1 (Age) + β2 (Male) + β3 (Bullied) + β4 (Fight) + β5 (Lonely)
+ β6 (Anxiety) + β7 (Friendship size) + β8 (Physical hunger) + β9
(Truancy) + β10 (Peer Support) + β11 (Parental supervision) + β12
(Parental smoking) + rij,
where Yij is the frequency of adolescent smoking for the respondent
i in the country j.
Country-level direct effectsmodel:
β0j=γ00 + γ01 (GDP) + γ02 (Adult smokers) + u0j
Cross-level moderating effectsmodel:
Β1j=γ10 + γ11 (GDP) + u1j
Β2j=γ20 + γ21 (GDP) + u2j
Results
Model 1 in Table 4 contains results from estimating the effects of
main individual-level predictors on adolescent smoking, while
Table1. Country-Level Breakdown of the Outcome and Main Predictor Variables
Country
Year
of survey
Sample
size
Response
rate (%)
Number
of boys/
girls
Number
of schools
Number
of classes
Per capita
GDP
Smoke
(ln)
Parental
smoking
Parental
supervision
Antigua 2009 1235 71 577/658 16 56 13 829 .08 .26 8.88
Argentina 2007 1881 82 924/957 23 47 7666 .33 .42 9.77
Benin 2009 2676 90 1742/934 15 30 713 .03 .21 9.90
Botswana 2005 2073 95 931/1142 12 25 5294 .09 .64 9.07
Chile 2004 1967 85 915/1052 12 25 6224 .40 .59 9.69
China 2003 2306 99 1114/1192 12 25 1274 .12 .64 9.51
Costa Rica 2009 2667 72 1285/1382 15 30 6386 .13 .18 9.30
Ecuador 2007 2411 87 1145/1266 13 27 3575 .12 .31 9.27
Grenada 2008 1438 82 646/792 19 53 7947 .07 .34 8.38
Guyana 2004 1164 80 490/674 12 25 1038 .07 .36 9.59
Indonesia 2007 2991 95 1427/1564 12 25 1871 .10 .72 9.14
Jamaica 2010 1576 72 769/807 12 24 4917 .21 .43 8.81
Jordan 2004 2289 95 991/1298 12 13 2156 .17 .56 9.23
Kenya 2003 3479 87 1694/1785 23 19 440 .18 .54 9.78
Kiribati 2011 1510 89 658/852 12 37 1539 .27 .66 7.37
Macedonia 2007 2026 93 1003/1023 15 30 3892 .24 .50 10.72
Maldives 2009 3110 80 1409/1701 25 73 6209 .15 .54 8.79
Mauritania 2010 1933 73 916/1017 12 24 861 .22 .31 8.76
Morocco 2006 2537 84 1274/1263 12 25 2128 .06 .31 8.17
Pakistan 2009 5147 87 3871/1276 22 44 1018 .07 .34 10.20
Peru 2010 2855 85 1399/1456 25 50 5075 .17 .24 9.06
Philippines 2003 6948 85 2953/3995 73 50 1016 .15 .49 8.21
Saint Lucia 2007 1232 82 524/708 19 47 6072 .08 .30 8.75
Saint Vincent 2007 1262 84 597/665 20 51 5611 .08 .29 8.85
Samoa 2011 2136 96 882/1254 13 29 3080 .46 .75 8.71
Solomon Islands 2011 1353 88 702/651 14 29 1295 .34 .72 8.94
Suriname 2009 1681 89 859/822 12 25 7450 .13 .42 9.09
Thailand 2008 2665 93 1317/1348 15 30 4118 .11 .38 9.24
Trinidad 2007 2832 78 1374/1458 19 50 16 664 .01 .32 8.63
Tunisia 2008 2717 83 1323/1394 12 25 4343 .11 .36 9.85
Uganda 2003 3079 76 1583/1496 22 23 236 .05 .30 9.28
Data source: Global School-based Student Health Survey (2003–2011).
Table2. Descriptive Statistics for the Variables Used
Variable
Mean/
proportion SD Min. Max.
Outcome measure
Smoke 21% __ 0 1
(Level-1 variables; N=58 956)
Age 4.30 1.22 1 6
Male 48% __ 0 1
Bullied 36% __ 0 1
Truancy 28% __ 0 1
Fight 36% __ 0 1
Loneliness 2.24 1.13 1 5
Anxiety 2.14 1.10 1 5
Peer support 3.15 1.29 1 5
Hunger 1.83 1.07 1 5
Friendship size 3.35 0.96 1 4
Parental smoking 41% __ 0 1
Parental supervision 9.27 3.45 3 15
(Level-2 variables; N=31)
GDP (Ln) 7.96 1.02 5.46 9.72
Adult smokers 0.43 0.16 0.18 0.75
Survey year 5.26 2.56 1 9
Data source: Global School-based Student Health Survey (2003–2011). The
gures presented are survey-adjusted and weighted to account for the prob-
ability of selection.
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
938 Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
adjusting for control variables. Several background factors were
related to the outcome variable. Older and male adolescents are
more likely to smoke. Those who experienced bullying or were
more prone to physical ghts are also more susceptible to smok-
ing. Skipping school, a sign of delinquency, is another signicant
predictor, as is the level of psychological anxiety (ie, being chroni-
cally worried). Turning to the rst research question, we found a
signicant and negative association between parental supervision
and adolescent smoking: the greater the parental supervision, the
less likely she/he is to smoke. With respect to the second question,
it was observed that parental smoking contributes signicantly to
adolescent smoking behavior. Specically, a student who lived with
at least one parent or guardian who smoked is more likely to do so
as well, while holding constant a large number of sociodemographic,
psychological, and interpersonal factors. To illustrate the effect size,
we ran additional models by dichotomizing the dependent variable
(0=never smoked; 1=otherwise). The odds are, for example, more
than 60% greater for adolescents to smoke if s/he lived with an adult
smoker, while they are 8% lower for youths who are better super-
vised by their parents or guardians. The ndings are summarized in
Supplementary Appendix A.
To examine whether living in a country with a higher propor-
tion of adult smokers and with greater national wealth is associated
with adolescent smoking, we added two country-level variables
(Adult smokers and GDP) in the analysis. Survey cycle was also
included to control for unobserved heterogeneity associated with the
timing of survey. According to Model 2, the parameter estimate for
national wealth reaches the conventional level of statistical signi-
cance (β=.089; P < .001), while the aggregate proportion of adult
smokers (parents and guardians who smoke, as reported by survey
respondents) does not. That is, only living in a country with higher
GDP per capita increases the frequency of adolescent smoking. The
association between the overall economic level and adolescent smok-
ing can be partly attributed to the fact that access to cigarettes may
be nancially prohibitive in very poor countries, a nding consist-
ent with an earlier study showing that youths are more sensitive to
cigarette prices in lower-income countries.52 That there may be a
positive connection between adolescent smoking and the proportion
of adult smokers, on the other hand, does not receive empirical sup-
port. Also insignicant is the survey cycle variable. While there has
been a growing trend over time cross-nationally in instituting more
antismoking laws and policies,53 this macro variable was not related
to smoking behaviors of adolescents.
The third and last research question asked whether the associa-
tions between the two main predictors (parental smoking and paren-
tal supervision) and the outcome variable are moderated by the level
Table3. Summary of Variable Definitions
Variable name Survey questions and coding
Smoke “During the past 30days, on how many days did you smoke cigarettes?” (1=0days, 2=1 or 2days, 3=3 to
5days,… 6=20 to 29days, 7=All 30days). The original answers are log transformed due to right-tailed
skewed distribution.
Level-1 variables (N=58 956)
Age “How old are you?” (eg, 1=11years old or younger; 2=12years old; 3=13years old; 6=16years old).
Male “What is your sex?” (male=1)
Bullied “During the past 30days, on how many days were you bullied?” (1=0days; 2=1 or 2days; 3=3 to 5days;
4=6 to 9days; 5=10 to 19days; 6=20 to 29days; 7=All 30days) The responses are dichotomized where
“0days”=0 and 1 otherwise.
Truancy “During the past 30days, on how many days did you miss classes or school without permission?” (1=0days,
2=1 to 2days, 3=3 to 5days, 4=6 to 9days, 5=10 or more days). Responses recoded so that “0days”=0
and 1 otherwise (ie, if the student missed at least 1day of school).
Fight “During the last 12months, how many times were you in a physical ght?” (eg, 1=0 times, 2=1 time, 3=2 or
3 times,… 7=10 or 11 times, 8=12 or more times) Recoded so that “0 times”=0; 1 otherwise.
Loneliness “During the past 12months, how often have you felt lonely?” Coded on a 5-point scale (1=Never, 2=Rarely,
3=Sometimes, 4=Most of the time, 5=Always).
Anxiety “During the past 12months, how often have you been so worried about something that you could not sleep at
night?” (Same as above)
Peer support “During the past 30days, how often were most of the students in your school kind and helpful?” Coded on a
5-point scale (Same as above)
Hunger “During the past 30days, how often did you go hungry because there was not enough food in your home?”
(Same as above)
Parental smoking “Which of your parents or guardians use any form of tobacco?” (1=Neither, 2=My father or male guardian,
3=My mother or female guardian, 4=Both). Responses dichotomized where “Neither”=0 and 1 otherwise.
Friendship size “How many close friends do you have?” (1=0 friend, 2=1 friend, 3=2 friends, 4=3 or more friends)
Parental supervision “During the past 30days, how often did your parents or guardians check to see if your homework was done?”
“During the past 30days,… understand your problems or worries?” “During the past 30days,… really
know what you were doing with your free time?” Coded on a 5-point scale (eg, 1=Never, 3=Sometimes,
5=Always). The three survey items combined to create a scale variable (Cronbach’s Alpha=.67).
Level-2 variables (N=31)
GDP (Ln) Per capita Gross Domestic Product from the year prior to the data collection
Adult smokers National average for (percentage of) the parents/guardians who smoke
Survey year Year in which the national survey was conducted, with values ranging from 1 to 9 (eg, 2003=1, 2004=2,
2005=3… 2011=9)
Data source: Global School-based Student Health Survey (2003–2011).
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
939Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
of national wealth. To address this question, we estimated Models
3 and 4 that successively introduced the two cross-level interaction
terms. The results indicated signicant interaction effects. First, the
positive association between parental smoking and adolescent smok-
ing was greater in countries with higher income level (β=.003; P
< .001), as shown in Model 3.Second, according to Model 4, the
negative association between parental supervision and adolescent
smoking was also stronger in countries characterized by greater per
capita GDP (β=−.003; P < .001).
Discussion
Despite the amount of research devoted to understanding the pro-
tective and risk factors associated with adolescent smoking, there is
relatively little evidence concerning the issues underlying smoking
behaviors among youths in low-income countries. By analyzing the
GSHS (2003–2011), we sought to bridge this empirical gap in the
extant literature. We focused on two family inuence factors: one
risk-oriented (whether or not the child lived with an adult smoker)
and another protective (degree of parental supervision). Consistent
with some of the previous results based on developed countries,17,18
multilevel analyses revealed that these two predictors were signi-
cantly associated with adolescent smoking behaviors in resource-
poor countries. The student respondents who lived with an adult
smoker under the same household are more susceptible to smoking.
By contrast, adolescents who were better integrated in terms of their
relationships with parents or guardians, or better supervised and
monitored by them, exhibited lower frequency of smoking.
In the current study, we sought to move beyond previous research
by examining possible interaction effects between individual-level
and contextual-level variables. As has been pointed out, understand-
ing how risk and protective factors interact across different levels
of analysis in inuencing adolescent smoking behaviors is critical.40
This inquiry, however, has not received adequate attention due to the
limited availability of data. One of the main goals of this study, there-
fore, was to address this issue using a large cross-national dataset.
While previous efforts have demonstrated the importance of paren-
tal role in shaping the smoking behaviors of adolescents in economi-
cally advanced nations,12,41 our study is the rst attempt to examine
how the magnitudes of this association vary according to national
wealth, specically in less developed regions of the world. Results
showed that the effects of parental inuence on adolescent smoking
behaviors are not constant but uctuate across a key national-level
characteristic, namely GDP per capita.
Why are the associations between the parental inuence factors
and adolescent smoking stronger in countries with higher levels of
national wealth? One possible explanation concerns the issue of
smoking norms, operating at both household and national levels. In
prior research, the social norm or societal attitude toward smoking
was found to be a strong predictor of tobacco use among youths.54 In
contrast to the situation in developed countries, public perception of
the health risks of smoking is relatively low55 and antismoking poli-
cies are very limited in resource-poor countries.56,57 Consequently,
the adolescents living there are more prone to have a positive percep-
tion of smoking.55 As such, they are also less likely to be signicantly
affected by the smoking behaviors of their parents. The situation
Table4. HLM Models Estimating the Frequency of Adolescent Smoking, GSHS (2003–2011)
Individual
level (N=58 956)
Country
level (N=31)
Model 1 Model 2 Model 3 Model 4
β (SE)β (SE)β (SE)β (SE)
Intercept β0Intercept, γ00 .021 (.029) .014 (.021) .001 (.021) .064 (.024)*
Level-1
Age, β1Intercept, γ10 .039 (.01)*** .039 (.01)*** .039 (.01)*** .039 (.01)***
Male, β2Intercept, γ20 .085 (.03)*** .086 (.03)*** .087 (.03)*** .086 (.03)***
Bullied, β3Intercept, γ30 .021 (01)*** .022 (01)*** .022 (01)*** .022 (.01)***
Fight, β4Intercept, γ40 .083 (.02)*** .083 (.02)*** .081 (.02)*** .080 (.02)***
Lonely, β5Intercept, γ50 .003 (.00) .003 (.00) .003 (.00) .003 (.00)
Anxiety, β6Intercept, γ60 .020 (.00)*** .021 (.00)*** .021 (.00)*** .021 (.00)***
Friendship size, β7Intercept, γ70 −.003 (.00) −.004 (.00) −.004 (.00) −.004 (.00)
Hunger, β8Intercept, γ80 .001 (.00) .001 (.00) .001 (.00) .001 (.00)
Truancy, β9Intercept, γ90 .116 (.02)*** .117 (.02)*** .116 (.02)*** .116 (.02)***
Peer support,β10 Intercept, γ100 .002 (.00) .002 (.00) .000 (.00) .000 (.00)
Parental supervision, β11 Intercept, γ110 −.008 (.00)*** −.008 (.00)*** −.007 (.00)*** −.009 (.00)***
Parental smoking, β12 Intercept, γ120 .051 (.01)*** .055 (.01)*** .066 (.01)*** .055 (.01)***
Level-2
GDP (Ln), γ01 .089 (.00)*** .076 (.02)** .089 (.00)***
Survey year,γ02 −.021 (.01) −.021 (.01) −.021 (.01)
Adult smokers,γ03 −.001 (.16) −.004 (.15) −.004 (.15)
Cross-level interaction
Parental smoking × GDP, γ11 .003 (.01)***
Parental supervision × GDP, γ21 −.003 (.00)***
Variance component (L-1) .138 .138 .138 .138
Variance component (L-2) .009*** .005*** .005*** .005***
ICC (%) 6.8 3.5 3.5 3.5
Deviance 50608.71 50587.32 50507.61 50552.19
GSHS=Global School-based Student Health Survey; SE =standard error. Parameter estimates are adjusted using person-weights and calculated from the unit-
specic models.
*P < .05; **P < .01; ***P < .001 (two-tailed tests).
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
940 Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
is different, however, for adolescents in more developed countries,
characterized by more stringent antismoking attitudes and policies.53
Where smoking has an obvious social stigma, parental smoking, seen
as an act of validation or even approval, thus has a greater impact
on the probability and frequency of adolescent smoking behavior. Its
inuence is also greater in higher-income countries because children
living with adult smokers have easier access to cigarettes that oth-
erwise might not be available because of the legal restrictions that
prohibit minors from purchasingthem.
The second main nding is that the negative relationship between
parental monitoring of children and their smoking behaviors is also
more pronounced in resource-rich countries. In a relatively poor
country, with fewer antismoking policies and greater tolerance for
smoking in general, the impact of parental supervision is weaker.
Ayouth living in such an environment may, for instance, experience
greater peer pressure to smoke or be more favorably predisposed to
smoking due to public perception at large, which could undermine
the protective function of close parental supervision. Areverse sce-
nario is more likely in a higher-income country which typically has
stronger antismoking sentiments and better-instituted antismoking
policies.53,57 There, monitoring of children by parents would not run
into the kinds of countervailing forces present in lower-income coun-
tries where views and attitudes regarding smoking are much more
permissive. Consequently, in a country with higher per capita GDP,
parental care, intervention and supervision would be more effective
in reducing adolescent smoking. In sum, it is when the broader exter-
nal environment sets up more cultural, legal and institutional barri-
ers against smoking that the two measures of parental inuence take
on greater potency, albeit in opposite directions.
Findings from the secondary analysis performed in this study
need to be interpreted in light of some limitations. First, given the
cross-sectional nature of the data, inferring causality should be done
with caution. In particular, the direction of causation between ado-
lescent smoking and the quality of the parent-child relationship can-
not be drawn conclusively. Second, the data collection for national
surveys took place over a 9-year period. Though dummy variables
were included to control for this time variation in the analysis, there
may still be some confounding effects associated with it. Third, the
GSHS did not ask the parents or guardians directly about their own
smoking behaviors. Rather, the information was based on the stu-
dent reports, which may have been biased or inaccurate. Fourth, the
measurement for parental supervision was based only on three sur-
vey items. Some of the control variables (eg, loneliness, anxiety, and
peer support) were also measured using a single item. Finally, since
the GSHS consists of resource-poor countries only, the ndings of
this study cannot be extended to other countries.
Limitations notwithstanding, our research provides a novel look
at how family inuence variables shape the degree to which adoles-
cents in developing countries engage in smoking behavior, as well as
how the magnitude of this association varies according to the level
of national wealth. In light of these results, smoking cessation inter-
ventions for youths must take into account the contingent role of
parental smoking and monitoring. That is, adolescent smoking ces-
sation programs need to recognize the level of national wealth as a
critical moderating factor, meaning attempts to create a single “one-
size-ts-all” strategy that does not consider country-level character-
istics (eg, per capita GDP) may fall short of reaching its objective.
Amajor policy implication is that an effective antismoking measure
requires paying close attention to both micro and macro variables.
Despite the efforts of domestic governments and international health
organizations, adolescent smoking in resource-poor countries has
been on the rise. More research and evidence are needed to provide
a clearer understanding of how and to what extent individual-level
factors associated with adolescent smoking (eg, parental smoking
and supervision) differentially exert inuence by interacting with
broader sociocultural and economic contexts.
Supplementary Material
Supplementary Appendix A can be found online at http://www.ntr.
oxfordjournals.org
Funding
This research was partially supported by a grant from the National
Research Foundation of Korea given to the corresponding author
(2014S1A5A2A03066021).
Declaration of Interests
None declared.
Acknowledgments
The data and coding schemes are available by the authors to anyone who
wishes to replicate this study.
References
1. World Health Organization. WHO Report on the Global Tobacco
Epidemic, 2011: Warning about the Dangers of Tobacco. 2011. http://
whqlibdoc.who.int/publications/2011/9789240687813_eng.pdf?ua=1.
Accessed July 24, 2015.
2. U.S. Department of Health and Human Services. Preventing Tobacco
Use among Youth and Young Adults: A Report of the Surgeon General.
Atlanta, GA: U.S. Department of Health and Human Services, Centers
for Disease Control and Prevention, National Center for Chronic Disease
Prevention and Health Promotion, Ofce on Smoking and Health; 2012.
http://www.surgeongeneral.gov/library/reports/preventing-youth-tobacco-
use. Accessed March 20, 2015.
3. Mathers CD, Loncar D. Projections of global mortality and burden of
disease from 2002 to 2030. PLoS Med. 2006;3(11):e442. doi:10.1371/
journal.pmed.0030442.
4. World Health Organization. Tobacco Fact Sheet N.339. 2015. http://www.
who.int/mediacentre/factsheets/fs339/en/. Accessed July 24, 2015.
5. Murray CJ, Lopez AD, ed. The Global Burden of Disease: AComprehensive
Assessment of Mortality and Disability from Disease, Injuries and Risk
Factors in 1990 and Projected to 2020. Cambridge, MA: Harvard School
of Public Health; 1996.
6. David A, Esson K, Perucic AM, Fitzpatrick C. Tobacco use: equal-
ity, and social determinants. In: Blas E, Kurup AS, ed. Equality, Social
Determinants, and Public Health Programmes. Geneva, Switzerland:
World Health Organization; 2010:199–217.
7. World Health Organization. WHO Report on the Global Tobacco
Epidemic, 2008: The MPOWER Package. 2008. http://whqlibdoc.who.
int/publications/2008/9789241596282_eng.pdf?ua=1. Accessed July 24,
2015.
8. Pampel FC, Denney JT. Cross-national sources of health inequality:
education and tobacco use in the world health survey. Demography.
2011;48(2):653–674. doi:10.1007/s13524-011-0027-2.
9. Global Youth Tobacco Survey Collaborative Group. Tobacco use among
youth: a cross country comparison. Tob Control. 2002;11(3):252–270.
doi:10.1136/tc.11.3.252.
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
941Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
10. Schnohr CW, Kreiner S, Rasmussen M, Due P, Currie C, Diderichsen
F. The role of national policies intended to regulate adolescent smok-
ing in explaining the prevalence of daily smoking: a study of adoles-
cents from 27 European countries. Addiction. 2008;103(5):824–831.
doi:10.1111/j.1360-0443.2008.02161.x.
11. Flay BR, Petraitis J, Hu FB. Psychosocial risk and protective factors
for adolescent tobacco use. Nicotine Tob Res. 1999;1(supl 1):S59–S65.
doi:10.1080/14622299050011821.
12. De Vries H, Engels R, Kremers S, Wetzels J, Mudde A. Parents’ and
friends’ smoking status as predictors of smoking onset: ndings from six
European countries. Health Educ Res. 2003;18(5):627–636. doi:10.1093/
her/cyg032.
13. Leonardi-Bee J, Jere ML, Britton J. Exposure to parental and sibling
smoking and the risk of smoking uptake in childhood and adolescence:
a systematic review and meta-analysis. Thorax. 2011;66(10):847–855.
doi:10.1136/thx.2010.153379.
14. Berndt TJ. Developmental changes in conformity to peers and parents.
Dev Psychology. 1979;15(6):608–616. doi:10.1007/BF02087623.
15. Kandel DB, Lesser G. Youth in Two Worlds. San Francisco, CA: Jossey-
Bass Inc; 1972.
16. Bricker JB, Peterson AV, Sarason IG, Andersen MR, Rajan KB. Changes
in the inuence of parents’ and close friends’ smoking on adolescent
smoking transition. Addict Behav. 2007;32(4):740–757. doi:10.1016/j.
addbeh.2006.06.020.
17. Chun J. Determinants of tobacco use among Korean female adolescents:
longitudinal test of the theory of triadic inuence. Child Youth Serv Rev.
2015;50:83–87. doi:10.1016/j.childyouth.2015.01.009.
18. Blokland E, Hale W, Meeus V, Engels R. Parental support and control
and early adolescent smoking: a longitudinal study. Subst Use Misuse.
2007;42(14):2223–2232. doi:10.1080/10826080701690664.
19. Scragg R, Reeder A, Wong G, Glober M, Nosa V. Attachment to par-
ents, parental tobacco smoking and smoking among year 10 students
in the 2005 New Zealand national survey. Aust N Z J Public Health.
2008;32(4):348–353. doi:10.1111/j.1753-6405.2008.00253.x.
20. Chassin L, Presson C, Rose J, Sherman S, Davis M, Gonzalez J. Parenting
style and smoking-specic parenting practices as predictors of adolescent
smoking onset. J Pediatr Psychology. 2005;30(4):333–344. doi:10.1093/
jpepsy/jsi028.
21. Pförtner T-K, Moor I, Rathmann K, etal. The association between fam-
ily afuence and smoking among 15-year-old adolescents in 33 European
countries, Israel and Canada: the role of national wealth. Addiction.
2015;110(1): 162–173. doi:10.1111/add.12741.
22. World Bank. Curbing the Epidemic: Governments and the Economics of
Tobacco Control. Washington, DC: The World Bank; 1999.
23. Yu M, Hahm HC, Vaughn MG. Intrapersonal and interpersonal determi-
nants of smoking status among Asian American adolescents: ndings from
a national sample. Nicotine Tob Res. 2010;12(8):801–809. doi:10.1093/
ntr/ntq100.
24. Ozer EJ, Fernald LCH. Alcohol and tobacco use among rural
Mexican adolescents: individual, familial, and community level
factors. J Adolesc Health. 2008;43(5):498–505. doi:10.1016/j.
jadohealth.2008.04.014.
25. DeWall CN, Pond RS. Loneliness and smoking: the costs of the desire to
reconnect. Self Identity. 2011;10(3):375–385. doi:10.1080/15298868.201
0.524404.
26. Grenard JL, Guo Q, Jasuja GK, et al. Inuences affecting adolescent
smoking behavior in China. Nicotine Tob Res. 2006;8(2):245–255.
doi:10.1080/14622200600576610.
27. Markham WA, Aveyard P, Bisset SL, Lancashire ER, Bridle C, Deakin S. Value-
added education and smoking uptake in schools: a cohort study. Addiction.
2008;103(1):155–161. doi:10.1111/j.1360-0443.2007.02020.x.
28. Breslau N. Psychiatric comorbidity of smoking and nicotine dependence.
Behav Genet. 1995;25(2):95–101.
29. Weiss JW, Mouttapa M, Cen S, Johnson CA, Unger J. Longitudinal
effects of hostility, depression, and bullying on adolescent
smoking initiation. J Adolesc Health. 2011;48(6):591–596. doi:10.1016/j.
jadohealth.2010.09.012.
30. Dijk F, Reubsaet A, de Nooijer J, de Vries H. Smoking status and peer
support as the main predictor of smoking cessation in adolescents
from six European countries. Nicotine Tob Res. 2007;9(3):S495–S504.
doi:10.1080/14622200701587060.
31. Thorlindsson T, Vilhjalmsson R. Factors related to cigarette smoking and
alcohol use among adolescents. Adolescence. 1991;26(102):399–418.
32. Filippidis FT, Agaku IT, Vardavas, CI. The association between peer, paren-
tal inuence and tobacco product features and earlier age of onset of regu-
lar smoking among adults in 27 European countries [published online
ahead of print March 31, 2015]. Eur J Public Health. 2015. doi:10.1093/
eurpub/ckv068.
33. Moor I., Rathmann K, Lenzi M, etal. Socioeconomic inequalities in ado-
lescent smoking across 35 countries: a multilevel analysis of the role of
family, school and peers. Eur J Public Health. 2015;25(3):1–7. doi:http://
dx.doi.org/10.1093/eurpub/cku244.
34. Bronfenbrenner U. Ecological models of human development. In: Husen
T, Postlethwaite TN, eds. International Encyclopedia of Education. Vol 3.
2nd ed. Oxford, United Kingdom: Elsevier; 1994.
35. Schnohr CW, Kreiner S, Rasmussen M, Due P, Currie C, Diderichsen
F. The role of national policies intended to regulate adolescent smok-
ing in explaining the prevalence of daily smoking: a study of adoles-
cents from 27 European countries. Addiction. 2008;103(5):824–831.
doi:10.1111/j.1360-0443.2008.02161.x.
36. Flay BR, Clayton R. Context and adolescent tobacco use trajectories.
Addiction. 2003;98(suppl 1):iii–iv. doi:10.1046/j.1360-0443.98.s1.1.x.
37. Ennett, ST, Foshee VA, Bauman KE. A social contextual analysis of youth
cigarette smoking development. Nicotine Tob Res. 2010;12(9):950–962.
doi:10.1093/ntr/ntq122.
38. Leatherdale ST, McDonald PW, Cameron R, Jolin MA, Brown KS. A multi-
level analysis examining how smoking friends, parents, and older students
in the school environment are risk factors for susceptibility to smoking
among non-smoking elementary school youth. Prev Sci. 2006;7(4):397–
402. doi:10.1007/s11121-006-0049-y.
39. Murnaghan DA, Leatherdale ST, Sihvonen M, Kekki P. A multilevel analy-
sis examining the association between school-based smoking policies,
prevention programs and youth smoking behavior: evaluating a provin-
cial tobacco control strategy. Health Educ Res. 2008;23(6):1016–1028.
doi:10.1093/her/cyn034.
40. Thrul J, Lipperman-Kreda S, Grube JW, Friend KB. Community-level adult
daily smoking prevalence moderates the association between adolescents’
cigarette smoking and perceived smoking by friends. J Youth Adolesc.
2014;43(9):1527–1535. doi:10.1007/s10964-013-0058-7.
41. Kelly AB, O’Flaherty M, Connor JP, et al. The inuence of par-
ents, siblings and peers on pre- and early-teen smoking: a
multilevel model. Drug Alcohol Rev. 2011;30(4):381–387.
doi:10.1111/j.1465-3362.2010.00231.x.
42. Pförtner TK, Moor I, Rathmann K, et al. The association between fam-
ily afuence and smoking among 15-year-old adolescents in 33 European
countries, Israel and Canada: the role of national wealth. Addiction.
2015;110(1):162–173. doi:10.1111/add.12741.
43. Centers for Disease Control and Prevention. Global School-based Student
Health Survey. 2015. http://www.cdc.gov/gshs/. Accessed April 6, 2015.
44. World Bank Open Data. Free and open access to data about development
in countries around the globe. 2015. http://data.worldbank.org. Accessed
April 6, 2015.
45. Snijders TAB, Bosker RJ. Multi-level Analysis: An Introduction to Basic
and Advanced Multilevel Modeling. 2nd ed. London, United Kingdom:
Sage; 2011.
46. Raudenbush SW, Bryk AS, Cheong YF, Congdon RT, du Toit M. HLM 7
for Windows. Chicago, IL: Scientic Software International; 2011.
47. Paccagnella O. Centering or not centering in multilevel models? The
role of the group mean and the assessment of group effects. Eval Rev.
2006;30:66–85. doi:10.1177/0193841X05275649.
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018
942 Nicotine & Tobacco Research, 2016, Vol. 18, No. 5
48. Raudenbush, S, Bryk A. Hierarchical Linear Models: Applications and
Data Analysis Methods. 2nd ed. Newbury Park, CA: Sage; 2002.
49. Rathman K, Ottova V, Hurrelmann K, etal. Macro-level determinants of
young people’s subjective health and health inequalities: a multilevel anal-
ysis in 27 welfare states. Maturitas. 2015;80(4):414–420. doi:10.1016/
jmaturitas.2015.01.008.
50. Elgar FJ, Pfortner T, Moor I, De Clercq B, Stevens G, Currie C.
Socioeconomic inequalities in adolescent health 2002–2010: a time-series
analysis of 34 countries participating in the Health Behaviour in School-
aged Children study. Lancet. 2015:385(9982):2088–2095. doi:10.1016/
S0140-6736(14)61460-4.
51. Paek HJ, Hove T, Oh HJ. Multilevel analysis of the impact of school-
level tobacco policies on adolescent smoking: the case of Michigan. J Sch
Health. 2013;83(10):679–689. doi:10.1111/josh.12081.
52. Kostova D, Ross H, Blecher E. Is youth smoking responsive to cigarette
prices? Evidence from low- and middle income countries. Tob Control.
2011;20(6):419–424. doi:10.1136/tc.2010.038786
53. Fong GT, Cummings KM, Borland R, etal. The conceptual framework of
the International Tobacco Control (ITC) Policy Evaluation Project. Tob
Control. 2006;15(suppl 3):iii3–iii11. doi:10.1136/tc.2005.015438.
54. Chen YH, Chen PL, Huang WG, Chiou HY, Hsu CY, Chao KY. Association
between social climate for smoking and youth smoking behaviors in
Taiwan: an ecological study. Int J Nurs Stud. 2010;47(10):1253–1261.
doi:10.1016/j.ijnurstu.2010.02.007.
55. Kendel D, Chen L. Consumer information and tobacco use. In: Jha P,
Chaloupka F, eds. Tobacco Control in Developing Countries. New York,
NY: Oxford University Press; 2000:178–214.
56. Abdullah ASM, Husten CG. Promotion of smoking cessation in develop-
ing countries: a framework for urgent public health interventions. Thorax.
2004;59(7):623–630. doi:10.1136/thx.2003.018820.
57. World Health Organization. WHO Report on the Global Tobacco
Epidemic 2013: Enforcing Bans on Tobacco Advertising, Promotion and
Sponsorship. 2013. http://www.who.int/tobacco/global_report/2015/
report/en/. Accessed July 24, 2015.
Downloaded from https://academic.oup.com/ntr/article-abstract/18/5/934/2510874
by Ewha Womans Univ. Library user
on 15 June 2018