 RESEARCH AND PRACTICE 
The Effect of Working for Pay on Adolescent Tobacco Use
| Rajeev Ramchand, PhD, Nicholas S. Ialongo, PhD, and Howard D. Chilcoat, ScD
Although progress has been made in reducing
cigarette smoking among adolescents since
the call to do so in Healthy People 2010, ef-
forts are needed to achieve further reduc-
tions.1,2 Generally, the Centers for Disease
Control and Prevention (CDC) recommends 6
strategies to reduce smoking among adoles-
cents, none of which are geared specifically
toward adolescents who are employed This is
unfortunate because the amount of time
young people spend working for pay has been
positively associated with smoking behaviors
or scales of substance use of which tobacco
use is often a component.4–10
Compared with other social domains com-
monly associated with adolescence (families,
neighborhoods, schools, and peer groups),
working has not been thoroughly investigated
as a risk factor for adolescent tobacco use.11
However, there are significant reasons why the
public health community should focus more at-
tention on this domain. Work is typically per-
ceived as a province of adulthood; however,
studies have indicated that nearly all high
school students work for pay at some point
during the school year.5,12,13 Most research on
work among adolescents has focused on sam-
ples of middle-class adolescents for whom em-
ployment opportunities are plentiful and em-
ployment is common. It is possible that
positive associations between work and to-
bacco use across samples located in diverse so-
cioeconomic communities might indicate that
tobacco use among adolescents may respond
to macrolevel policy changes that restrict the
hours that adolescents can work or the jobs
they can hold legally.14,15 The workplace could
also be considered as a location for antismok-
ing prevention programs or policies.
We investigated whether working for pay
was associated with past, current, and inci-
dent tobacco use among adolescents. Al-
though the relationship among adolescents
between working and substance use has
been seen across many studies, there is a
gap in our understanding about whether
working contributes to the development of
2056 | Research and Practice | Peer Reviewed | Ramchand et al.
Objectives. We investigated the links between working for pay and adolescent
tobacco use to determine whether working for pay increases smoking risk.
Methods. We performed retrospective and prospective analyses using data
from a cohort of 799 predominantly African American students in Baltimore, Md,
who had been followed since the first grade.
Results. At the 10th year of follow-up, when the adolescents were aged 14 to
18 years, there was a positive relationship between the time they spent working
for pay and current tobacco use. This relationship was attenuated somewhat
after adjustment for potential selection effects. Adolescents who spent more than
10 hours per week working for pay also tended to initiate tobacco use earlier
than did their peers. Among adolescents who had not yet used tobacco, those who
started to work 1 year after assessment and those who worked over 2 consecu-
tive assessments had an elevated risk of initiating use relative to adolescents
who did not start working.
Conclusions. There is a strong link between working for pay and adolescent to-
bacco use. Policymakers should monitor the conditions under which young peo-
ple work to help minimize young workers’ tobacco use and potential for initiat-
ing use. (Am J Public Health. 2007;97:2056–2062. doi:10.2105/AJPH.2006.094045)
smoking behaviors and, thus, can be modi- METHODS
fied by workplace policies and prevention
programs. Sample
Using data from an urban-based commu- We obtained data from the second gener-
nity cohort of adolescents followed annually ation of the Baltimore Prevention and Inter-
since the first grade, we investigated whether vention Research Center (PIRC) studies. In
working for pay at or around 10th grade is 1993, 678 students who were entering first
positively associated with current use of to- grade classrooms in participating schools in
bacco. Some researchers suspect that the west Baltimore and their families were in-
positive relationship between work and vited to participate in a randomized con-
smoking is the result of a selection effect, trolled trial aimed at improving young peo-
whereby the same adolescents more likely to ple’s shy and aggressive behaviors. These
use tobacco are those adolescents more students were assigned to 1 of 2 classroom-
likely to work (or devote a significant based prevention programs or to a control
amount of time to work) during high group. The results of the intervention have
school.7,10,16–18 We tested for such an effect been described elsewhere.19–22 Students and
by examining whether these associations re- their families were and continue to be inter-
mained even after adjustment for earlier viewed annually since the first grade, al-
childhood factors and concurrent potential though no assessments were conducted in
confounds, as well as whether adolescent the fourth and fifth years of follow-up. Dur-
workers tended to report an earlier age of ing the first year, 121 additional students
first use of tobacco than their nonworking joined the classrooms after the initial re-
peers. Finally, we tested whether adolescents cruitment, although as of year 11, attempts
who started working or continued to work were made to contact only members of the
between 2 consecutive years had an ele- originally recruited first grade sample be-
vated risk of initiating tobacco use compared cause of the primary goal of the follow-up
with adolescents who did not work during (i.e., to assess distal effects of the first grade
this interval. intervention).
American Journal of Public Health | November 2007, Vol 97, No. 11

Beginning at the sixth year of follow-up,
all the adolescents were asked about their
experiences with tobacco, and beginning at
the 10th year of follow-up, they were asked
about the average time they spent working
for pay during the school year. Of the total
799 adolescents with some information at
grade 1, 570 were interviewed at year 10
(501 of the 678 original recruits; 69 of the
additional 121), 515 were interviewed at
year 11, and 488 were interviewed in both
years 10 and 11 (by definition, all were
from the original 678). Parents received
consent information, and adolescents com-
pleted an assent procedure and were as-
sured of their confidentiality before each in-
terview.
Analysis
In the first stage of the analysis, we exam-
ined the link between working for pay and
substance use at the 10th year of follow-up,
when the adolescents were aged between
14 and 18 years (the mean age was 16
years). Current tobacco use was a dichoto-
mous variable that indicated use in the past
30 days. We collected work intensity (the
average number of hours adolescents
worked per week during the school year) in
5-hour intervals; we subsequently catego-
rized these responses into 1 of 3 categories
of work: not working during the past year,
working an average of 1 to 10 hours per
week (moderate work intensity), or working
an average of more than 10 hours per week
(high work intensity). In separate analyses,
we used other thresholds (15 hours and 20
hours), which yielded qualitatively similar,
although less precise, results (because of the
smaller sizes of the high-intensity group).
We used logistic regression to assess the re-
lationship between working for pay and cur-
rently using tobacco.
To examine the possibility that adoles-
cents more likely to use tobacco were also
more likely to work at year 10 (i.e., a poten-
tial selection effect), early childhood charac-
teristics previously associated with adoles-
cent tobacco use were incorporated into the
logistic model. We included teachers’ re-
ports of adolescents’ early aggressive be-
haviors as exhibited in the first grade in the
model.23 From year 6, when adolescents
November 2007, Vol 97, No. 11 | American Journal of Public Health
 RESEARCH AND PRACTICE 
were aged between 10 and 14 years, scale
values of affiliation with drug-using peers
and affiliation with deviant peers as well as
dichotomous indicators of whether the ado-
lescents were in a grade level lower than
grade 6 (as an indicator of grade retention)
or had any friends aged older than 15 years
were all included in the model.24–26 We
also included a scale of parent monitoring,
in which higher scores indicate less moni-
toring, in the model.27 From year 9, di-
chotomous indicators of whether the ado-
lescents were in a grade level lower than
grade 9 or had any friends aged older than
17 years were included, as was a measure
of parents’ highest level of educational
attainment, a proxy for socioeconomic sta-
tus.28 The difference between year 9 and
year 6 scale values for drug-using peer affil-
iation, deviant peer affiliation, and parent
monitoring were included in the final
model. We also included gender and race
as potential confounds that were collected
at baseline.
To assess the possibility of additional selec-
tion effects, we conducted survival analysis to
examine the age at first use of tobacco strati-
fied by the year-10 work categories specified
above. Because the proportionality assump-
tion for this model was valid when adoles-
cents who worked more than 10 hours were
compared with a combined category of ado-
lescents who did not work and who worked
less than 10 hours, a Cox–proportional haz-
ards regression model was used to empiri-
cally quantify the difference in the hazard
function of smoking based on year-10 work
status.
The final analytic step was to assess the
incidence of tobacco use between years 10
and 11. Adolescents at risk for incident to-
bacco use were defined as never reporting
tobacco use since the year-6 assessment
through year 10. Incidence was then esti-
mated across the entire sample at risk and
then separately across 4 work transition cat-
egories: did not work in both years 10 and
11, did not work in year 10 but worked in
year 11, worked in year 10 but did not work
in year 11, and worked in both years. Relative
risk estimates were calculated with adoles-
cents not working in both years as the refer-
ence group.
Ramchand et al. | Peer Reviewed | Research and Practice | 2057
RESULTS
Sample
Table 1 presents descriptive characteristics
of the baseline and restricted samples (those
interviewed in both years 10 and 11). Across
all samples, close to 55% were male and
85% were African American. Of the total
sample, a little more than half received free
school lunch at the first grade (a proxy of
socioeconomic status for adolescents).29 Com-
plete information on baseline socioeconomic
and intervention status were available on
adolescents at year 11 and in the restricted
sample, which reflects the follow-up strategy.
The prevalence of work increased from year
10 to year 11. At year 10, 26% of the adoles-
cents surveyed reported working for pay; by
year 11, close to 40% worked for pay. There
were no apparent differences between the dis-
tribution of work status among the separate
year-10 or year-11 samples and the restricted
sample of adolescents who were assessed in
both years. Among those who worked in year
10, 30% were employed in fast-food restaurants
and the remainder were employed babysitting
(11%), working in restaurants (9%), working as
a store clerk (12%), or in other occupations, in-
cluding cosmetology and other retail (20%). By
year 11, babysitting reduced to only 4%, al-
though 30% of adolescents who worked still
worked in fast food and the remainder worked
at restaurants (12%), stores (15%), or in other
occupations, such as construction, cosmetology,
retail, service (i.e. catering, teaching dance, etc),
and maintenance and cleaning jobs (24%; data
on type of job not shown).
The prevalence of current tobacco use in-
creased from approximately 13% at year 10
(which is similar to state-level estimates of
public school–attending 10th graders in Mary-
land) to 17% at year 11.30 Among the 488
adolescents assessed in both years, 50% did
not work in both years, 25% did not work in
year 10 but worked in year 11, and the re-
mainder was roughly evenly distributed be-
tween working in year 10 but not in year 11
and working in both years (data not shown).
Cross-Sectional Analysis
Of the 570 adolescents assessed at year
10, complete information was available
for 400 on the first-grade, sixth-year, and
  RESEARCH AND PRACTICE 

TABLE 1—Descriptive Characteristics of Samples, by Year of Follow-Up: Second Generation Constructs that have been previously asso-
of the Prevention and Intervention Research Center Study, Baltimore, Md, 1993–2004 ciated with substance-using behaviors were
entered into the logistic model to assess for
Restricted a potential selection effect. Estimates from
Baseline, No. Year 10, No. Year 11, No. Sample,a No.
the multivariable model (also presented in
(Category %) (Category %) (Category %) (Category %)
Table 2) revealed that after including these
Total 799 (100.00) 570 (100.00) 515 (100.00) 488 (100.00) variables, the relationship between high work
Baseline characteristics intensity and current tobacco use was some-
Gender what attenuated and the corresponding CI
Male 435 (54.44) 315 (55.26) 283 (54.95) 263 (53.89) now included the null value (adjusted OR
Female 364 (45.56) 255 (44.74) 232 (45.05) 225 (46.11) [AOR] = 1.98; 95% CI = 0.84, 4.67). Further
Raceb analyses revealed that 1 construct was largely
African American 678 (84.86) 492 (86.32) 452 (87.77) 429 (87.91) responsible for attenuating the relationship
White 120 (15.02) 78 (13.68) 63 (12.23) 59 (12.09) between high work intensity and tobacco use:
Proxy measures of socioeconomic status changes in drug-using peers between years
Not available 6 (0.75) 4 (0.70) 5 (0.97) 4 (0.82) 6 and 9, which itself remained linked with
Received free school lunch 411 (51.44) 305 (53.51) 317 (61.55) 297 (60.86) current use of tobacco (AOR = 1.18; 95%
Received reduced-price school lunch 52 (6.51) 41 (7.19) 40 (7.77) 39 (7.99) CI = 1.07, 1.31). The only other variable in
Paid by child 209 (26.16) 151 (26.49) 153 (29.71) 148 (30.33) the multivariable model that remained linked
Missing 121 (15.14) 69 (12.11) 0 (0.00) 0 (0.00) with current use of tobacco was affiliation
Intervention status with drug-using peers in year 6 (AOR = 1.23;
Control group 261 (32.67) 181 (31.75) 169 (32.82) 157 (32.17) 95% CI = 1.05, 1.44).
Classroom-centered group 258 (32.29) 182 (31.93) 167 (32.43) 163 (33.40)
Family–school partnership group 260 (32.54) 192 (33.68) 179 (34.76) 168 (34.43) Age of First Tobacco Use
No design 20 (2.50) 15 (2.63) 0 (0.00) 0 (0.00) Kaplan–Meier (product–limit) estimates for
Year 10 follow-up characteristics age of first use of tobacco by 10th-year work
Work status intensity are presented in Figure 1. Aside
No work 419 (73.51) 361 (73.98) from a few sparse respondents who initiated
1–10 h/wk 61 (10.70) 54 (11.07) tobacco use earlier than age 7 years, the ear-
> 10 h/wk 90 (15.79) 73 (14.96) liest tobacco users initiated use at age 7
Current tobacco use 74 (12.98) 58 (11.89) years. The median failure time (age of first
Year 11 follow-up characteristics use) among adolescents who worked more
Work status than 10 hours at year 10 was at age 13 years.
No work 320 (62.14) 300 (61.48) Nonworkers at year 10 had a median failure
1–10 h/wk 54 (10.49) 52 (10.66) time at age 14 years, whereas moderate
> 10 h/wk 141 (27.38) 136 (27.87) workers had a median failure time at age 15
Current tobacco use 89 (17.28) 75 (15.37) years. Estimates from a Cox–proportional
a hazards model indicated an increased hazard
The restricted sample included only those adolescents surveyed in both years 10 and 11.
b
One student was of Hispanic ethnicity at baseline and was lost to follow-up at years 10 and 11. ratio for high-intensity workers compared
c
These students joined the classrooms during first grade. with a collapsed category of moderate-
intensity workers and nonworkers (hazards
ratio = 1.34; 95% CI = 1.01, 1.78).
ninth-year variables used in the multivariable 95% confidence interval [CI] = 1.40, 6.11), al-
logistic regression model. Univariate associa- though there was not strong evidence to link Longitudinal Analysis
tions between constructs included in the ad- moderate work with current use (OR = 1.38; The incidence of tobacco use between years
justed model and current tobacco use for 95% CI = 0.50, 3.83). There were additional 10 and 11 across the entire sample and the 4
these 400 adolescents are presented in positive associations in univariate analyses be- work transition categories is presented in Table
Table 2 (estimates are qualitatively similar tween current tobacco use and high levels of 3. A total of 488 adolescents were surveyed in
in univariate analyses that included all ado- aggression in first grade, reductions in parent both the 10th and 11th year of follow-up. Of
lescents). These results showed that high- monitoring between years 6 and 9, affiliation these, 214 had not yet reported using tobacco
intensity workers at year 10 were 3 times with peers who used drugs in year 6, and in- at year 10. Overall, the 1-year incidence of to-
more likely to report current use of tobacco creases in affiliations with peers who used bacco use in this “at risk” sample was 14%.
than were nonworkers (odds ratio [OR] = 2.93; drugs between years 6 and 9. However, there were significant differences in

2058 | Research and Practice | Peer Reviewed | Ramchand et al. American Journal of Public Health | November 2007, Vol 97, No. 11
  RESEARCH AND PRACTICE 

TABLE 2—Prevalence for Categorical Variables and Logistic Regression Results, Tobacco the incidence of tobacco use across the work-
Use at Year 10: Second Generation of the Prevention and Intervention Research Center transition categories. Incidence of smoking was
Study, Baltimore, Md, 1993–2004 (N = 400) 0.29 among adolescents who did not work in
year 10 but did work in year 11 versus 0.04
Current Tobacco Usea among the adolescents who did not work in
Prevalence, OR AOR both periods (relative risk = 8.0; 95% CI = 2.8,
No. % (95% CI) (95% CI) 22.9). In addition, the incidence of tobacco use
Work status among adolescents who worked in both peri-
No work (Ref) 299 8.5 1.00 1.00 ods was 0.31, which yielded a relative risk of
1–10 h/wk 44 11.4 1.38 (0.50, 3.83) 1.50 (0.46, 4.82) 8.7 (95% CI = 2.9, 15.8) compared with non-
> 10 h/wk 61 21.3 2.93 (1.40, 6.11) 1.98 (0.84, 4.67) workers. Although both relative risk estimates
Gender have corresponding 95% CIs that are wide,
Male (Ref) 215 11.6 1.00 1.00 the lower bounds for both are far from the
Female 185 9.7 0.82 (0.43, 1.55) 1.04 (0.48, 2.22) null, which signals strong evidence of an ele-
Race
vated risk.
African American (Ref) 352 9.7 1.00 1.00
White 48 18.8 2.16 (0.96, 4.83) 1.25 (0.43, 3.63)
DISCUSSION
Academic performance
The results from this study indicated a
≥ Grade 6 or above (year 6; Ref) 366 11.2 1.00 1.00
clear relationship between working for pay
< Grade 6 34 5.9 0.49 (0.11, 2.14) 0.22 (0.04, 1.32)
and adolescent tobacco use. It is noteworthy
≥ Grade 9 or above (year 9; Ref) 331 9.4 1.00 1.00
that this association held among adolescents
< Grade 9 69 17.4 2.04 (0.99, 4.20) 2.07 (0.76, 5.63)
in a primarily urban environment. Previous
First grade behaviorsb
research indicated that adolescents who find
Authority acceptance . . .c . . .c 1.54 (1.08, 2.18) 1.43 (0.94, 2.19)
c c jobs in urban areas are unique because they
Parent monitoring—year 6 ... ... 1.02 (0.94, 1.10) 0.98 (0.88, 1.11)
have found work where jobs are likely
Change in parent monitoring—from year 6 to year 9 . . .c . . .c 1.07 (1.00, 1.15) 1.03 (0.95, 1.12)
scarce.14,15,31 They also may be more likely to
Peer affiliationd . . .c . . .c
work to pay for transportation and supple-
Peers who use drugs—year 6 . . .c . . .c 1.11 (1.02, 1.21) 1.23 (1.05, 1.44)
c
mental school materials.15,32 However, we
Change in peers who use drugs—from year 6 to year 9 ... . . .c 1.14 (1.06, 1.22) 1.18 (1.07, 1.31)
found that the time adolescents report work-
Deviant peers—year 6 . . .c . . .c 1.07 (0.99, 1.16) 1.01 (0.88, 1.16)
ing for pay during the school year was posi-
Change in deviant peers—from year 6 to year 9 . . .c . . .c 1.03 (0.96, 1.10) 0.99 (0.90, 1.10)
tively associated with the likelihood of cur-
No peers aged older than 15 years—year 6 (Ref) 198 9.1 1.00 1.00
rent tobacco use. Thus, it appears that the
Peers aged older than 15 years—year 6 202 12.4 1.41 (0.74, 2.68) 1.32 (0.62, 2.82)
relationship between work and tobacco use
No peers aged older than 17 years—year 9 (Ref) 134 7.5 1.00 NA 1.00 NA
seen previously in both nationally representa-
Peers aged older than 17 years—year 9 266 12.4 1.76 (0.84, 3.68) 1.08 (0.46, 2.53) tive and community surveys of more-affluent
Parents’ educational attainment adolescents exists even in urban areas.
High school graduate (Ref) 160 10 1.00 1.00 Evidence of a selection effect is supported
Eighth grade or less 5 60 . . .e . . .e by these analyses, which indicate that the
Some high school 51 15.7 1.67 (0.67, 4.18) 1.48 (0.51, 4.27) cross-sectional relationship between work
Vocational/college/graduate 184 8.7 0.86 (0.41, 1.77) 0.92 (0.41, 2.07) hours and tobacco use was somewhat attenu-
Note. OR = odds ratio (crude); CI = confidence interval; AOR = adjusted odds ratio. All estimates were derived from respondents ated in models in which we controlled for a
with no missing data and who were assessed at the 10th year of follow-up. number of earlier childhood predictors of
a
Current tobacco use was defined as having used tobacco in the past 30 days. adolescent substance use. However, because
b
Early childhood characteristics previously associated with adolescent tobacco use were incorporated into the logistic model.
Teachers’ reports of adolescents’ early authority acceptance, an indicator of aggressive behaviors, as exhibited in the first work information was not collected until year
grade were included in the model. A scale of parent monitoring, in which higher scores indicate less monitoring, was also 10 and changes in drug-using peers between
included in the model. years 6 and 9 largely drove this attenuation,
c
Continuous variables; OR represents change per unit.
d
From year 6, when adolescents were aged between 10 and 14 years, scale values of affiliation with drug-using peers and these results should be interpreted cautiously.
affiliation with deviant peers as well as dichotomous indicators of whether the adolescents were in a grade level lower than We can not exclude the possibility that a
grade 6 (as an indicator of grade retention) or had any friends aged older than 15 years were all included in the model.24–26 young person started working between years
From year 9, dichotomous indicators of whether the adolescents were in a grade level lower than grade 9 or had any friends
aged older than 17 years were included. 6 and 9 and developed friendships with
e
Small sample sizes impeded ability to calculate estimates. coworkers who used drugs, which would
clearly implicate working as a risk factor

November 2007, Vol 97, No. 11 | American Journal of Public Health Ramchand et al. | Peer Reviewed | Research and Practice | 2059
  RESEARCH AND PRACTICE 

and to investigate the threshold that differen-

tiates “high” work intensity from “moderate”
work intensity.
The most striking finding of this research
concerns the incidence of tobacco use be-
tween the 10th and 11th years of follow-up.
Among adolescents who had not yet used
tobacco, those who worked during 2 consec-
utive years as well as those who started to
work between these years were at least 3
times more likely to report tobacco use initia-
tion compared with nonworkers. These find-
ings suggest that working has some impact
on adolescent tobacco use independent of se-
lection effects. Young workers may resort to
cigarette use in order to take breaks from the
duties and tasks they are required to perform
FIGURE 1—Kaplan–Meier failure estimates of age of first use of tobacco, by year-10 work
at work, because they have access to com-
intensity: Prevention and Intervention Research Center Study, Baltimore, Md, 1993–2004
mercial sources of cigarettes vis-à-vis their
job earnings, or to cope with the stresses of
balancing the responsibilities of being a stu-
dent and a worker. Further research of this
TABLE 3—Incident Tobacco Use Between Years 10 and 11: Second Generation of the sample, and of others that contain more-
Prevention and Intervention Research Center Study, Baltimore, Md, 1993–2004 detailed information on the types and charac-
Ever Use At Risk Event Incidence RR (95% CI) teristics of the jobs adolescents hold, will
help test these and other hypotheses.
Total 214 31 0.14 Our study consisted largely of African
No work in both years (Ref) 111 4 0.04 1.00 American adolescents, and much research
No work to work 52 15 0.29 8.00 (2.79, 22.93) has been devoted to smoking behaviors
Work to no work 19 2 0.11 2.92 (0.57, 14.85) among this group. In general, African
Work both periods 32 10 0.31 8.67 (2.91, 15.81) American adolescents are less likely to re-
Note. RR = relative risk; CI = confidence interval. We assessed the incidence of tobacco use between years 10 and 11. port smoking compared with White adoles-
Adolescents at risk for incident tobacco use were defined as never reporting tobacco use since the year-6 assessment cents.33–35 Proposed reasons for this vary,
through year 10. Incidence was then estimated across the entire sample at risk and then separately across 4 work transition ranging from a reduced likelihood of devel-
categories: did not work in both years 10 and 11, did not work in year 10 but worked in year 11, worked in year 10 but did
not work in year 11, and worked in both years. Relative risk estimates were calculated with adolescents not working in both oping patterns of habitual or daily use,
years as the reference group. stronger bonds to religious organizations,
higher levels of parental disapproval of
smoking, or a reduced likelihood of posi-
tive smoking-related attitudes compared
versus a potential selection effect. On the The Kaplan–Meier curves indicated that with White adolescents.36–40 Informal so-
other hand, survival analysis indicates that moderate-intensity workers tended to initiate cial controls, such as bonds with religious
adolescents who worked more than 10 hours tobacco use later than did both nonworkers entities and familial bonds that protected
per week at year 10 did tend to report earlier and high-intensity workers. This is similar to adolescents from substance use in the past
ages of tobacco use initiation. To explain findings from the Monitoring the Future might weaken when adolescents start
similar results, Newcomb and Bentler pro- study, in which a greater share of nonworkers working and thereby increase substance-
posed a theory of “precocious development”— reported using tobacco heavily, marijuana, using behaviors.41,42 In addition, the role
a developmental trajectory in which adoles- and cocaine than did adolescents who of cigarette breaks and increased exposure
cents seek out the rewarding aspects of worked between 1 and 10 hours per week.5 to coworkers who smoke may promote
adulthood ahead of their counterparts by as- This suggests that perhaps a moderate more positive smoking-related attitudes.
suming social roles (e.g., worker) and adult- amount of work is protective against smoking Future research on these topics is crucial
like behaviors (e.g., tobacco use), although and other drug involvement among adoles- for persons interested in using the work-
further longitudinal research is needed to dis- cents. However, more research is needed to place as a venue for adolescent tobacco
cern such a trajectory.16 examine the benefits to adolescents of work prevention programs.

2060 | Research and Practice | Peer Reviewed | Ramchand et al. American Journal of Public Health | November 2007, Vol 97, No. 11

Strengths and Limitations
There are 2 significant limitations of this
study. In the 10th and 11th years of follow-up,
adolescents were asked only limited informa-
tion about their work experiences. Other work-
related constructs, such as work duration or
work schedules, also likely impact young work-
ers’ tobacco use, but were not ascertained in
these surveys. Additional information about
adolescents’ work experiences has been col-
lected in year 12, although these data are not
yet available for analysis. The other significant
limitation is temporal sequencing. Although our
longitudinal study assessed adolescents annu-
ally, if adolescents moved from not working to
working between the year-10 and year-11 as-
sessments and also initiated tobacco use be-
tween these 2 surveys, it is unclear whether
working preceded smoking or vice versa. Fi-
nally, potentially important confounders associ-
ated with adolescents smoking in the past, such
as discretionary income and parental smoking
behaviors, are not included in the survey or
analysis.
Our study carries with it strengths that off-
set these limitations. First, the cohort of ado-
lescents came from a well-defined population
of adolescents who entered first grade class-
rooms in a specific geographic region, thereby
reducing biases inherent in convenience or
other sampling procedures. The strengths of
the longitudinal structure were exploited to
control for childhood constructs assessed
prospectively and, thus, less subject to recall
bias, including measures of first-grade aggres-
sion and age at first use of tobacco. The longi-
tudinal design also allowed us to look at inci-
dent tobacco-using behaviors specific to
general work transition categories, which to
our knowledge, have never been performed
elsewhere. Finally, the sample was an urban-
based community cohort of primarily African
American youth. Research on work in this
population is limited, and our study lends ev-
idence to the notion that the link between
working and tobacco is generalizable to set-
tings of varying economic conditions.
Conclusions
We recommend 3 public health responses
to these findings. First, given that young work-
ers in this study and in most other studies are
concentrated among certain industries (such
November 2007, Vol 97, No. 11 | American Journal of Public Health
 RESEARCH AND PRACTICE 
as fast food, other restaurant work, and retail),
efforts should be made to encourage these in-
dustries be smoke-free. Previous research has
indicated that totally smoke-free workplaces
are associated with reductions in prevalence
of smoking, and we believe that these policies
might also lead to reduced smoking incidence
among young people.43 Second, efforts should
be made to establish workplace-based preven-
tion programs. These programs would ideally
incorporate prevention strategies pertinent to
young workers, such as fiscal management
and ways to cope with stress while at work.
Finally, this research should highlight the need
for future studies to more carefully and sys-
tematically evaluate the impact that working
for pay has on the substance-using behaviors
of adolescents.
About the Authors
At the time of the study, Rajeev Ramchand, Nicholas S.
Ialongo, and Howard D. Chilcoat were with the Department
of Mental Health, Johns Hopkins Bloomberg School of
Public Health, Baltimore, Md.
Requests for reprints should be sent to Dr Rajeev Ramc-
hand, Rand Corporation, 1200 S Hayes St, Arlington, VA
22202-5050 (e-mail: rajeev_ramchand@rand.org).
This article was accepted November 14, 2006.
Contributors
R. Ramchand originated and performed all analyses
and led the writing. N. S. Ialongo was the principal in-
vestigator of the larger cohort study and contributed to
the writing and analyses. H. D. Chilcoat also contributed
to the writing and analyses. All authors helped to con-
ceptualize ideas, interpret findings, and review drafts of
the article.
Acknowledgments
This work was supported by grants from the National
Institute on Drug Abuse (DA018013 and DA11796)
and from the National Institute of Mental Health
(MH57005).
Human Participant Protection
This study was approved by the Johns Hopkins
Bloomberg School of Public Health institutional review
board.
References
1. Healthy People 2010: Understanding and Improving
Health. 2nd ed. Washington, DC: US Department of
Health and Human Services; 2000.
2. Grunbaum JA, Kann L, Kinchen S, et al. Youth
risk behavior surveillance—United States, 2003.
MMWR Surveill Summ. 2004;53:1–96.
3. Centers for Disease Control and Prevention. Ciga-
rette use among high school students—United States,
1991–2003. MMWR Morb Mortal Wkly Rep. 2004;
53:499–502.
Ramchand et al. | Peer Reviewed | Research and Practice | 2061
4. Resnick MD, Bearman PS, Blum RW, et al. Pro-
tecting adolescents from harm. Findings from the Na-
tional Longitudinal Study on Adolescent Health. JAMA.
1997;278:823–832.
5. Bachman JG, Schulenberg J. How part-time work
intensity relates to drug use, problem behavior, time
use, and satisfaction among high school seniors: are
these consequences or merely correlates. Dev Psychol.
1993;29:220–235.
6. Valois RF, Dunham AC, Jackson KL, Waller J.
Association between employment and substance abuse
behaviors among public high school adolescents. J Ado-
lesc Health. 1999;25:256–263.
7. Safron DJ, Schulenberg JE, Bachman JG. Part-time
work and hurried adolescence: the links among work
intensity, social activities, health behaviors, and sub-
stance use. J Health Soc Behav. 2001;42:425–449.
8. Wu LT, Schlenger WE, Galvin DM. The relation-
ship between employment and substance use among
students aged 12 to 17. J Adolesc Health. 2003;32:
5–15.
9. Steinberg LD, Fegley S, Dornbusch SM. Negative
impact of part-time work on adolescent adjustment:
evidence from a longitudinal study. Dev Psychol. 1993;
29:171–180.
10. Paternoster R, Bushway S, Brame R, Apel R. The
effect of teenage employment on delinquency and
problem behaviors. Soc Forces. 2003;82:297–335.
11. Turner L, Mermelstein R, Flay B. Individual and
contextual influences on adolescent smoking. Ann N Y
Acad Sci. 2004;1021:175–197.
12. Light A. High School Employment. In: Bureau of
Labor Statistics, ed. National Longitudinal Survey Dis-
cussion Paper, report NLS 95-27. Washington, DC: US
Department of Labor; 1995.
13. Board on Children, Youth and Families. Protecting
Youth at Work: Health, Safety, and Development of
Working Children and Adolescents in the United States.
Washington, DC: National Academy Press; 1998.
14. Mortimer JT. Working and Growing Up in America.
Cambridge, Mass: Harvard University Press; 2003.
15. Newman KS. No Shame in My Game: The Working
Poor in the Inner City. 1st ed. New York, NY: Knopf and
the Russell Sage Foundation; 1999.
16. Newcomb MD, Bentler PM. Consequences of Ado-
lescent Drug Use: Impact on the Lives of Young Adults.
Newbury Park, Calif: Sage Publications; 1988.
17. Staff J, Uggen C. The fruits of good work: early
work experiences and adolescent deviance. J Res Crime
Delinquency. 2003;40:263–290.
18. Bachman JG, Safron DJ, Sy SR, Schulenberg JE.
Wishing to work: new perspectives on how adoles-
cents’ part-time work intensity is linked to educational
disengagement, substance use, and other problem
behaviors. Intl J Behav Dev. 2003;27:301–315.
19. Storr CL, Ialongo NS, Kellam SG, Anthony JC. A
randomized controlled trial of two primary school in-
tervention strategies to prevent early onset tobacco
smoking. Drug Alcohol Depend. 2002;66:51–60.
20. Furr-Holden CD, Ialongo NS, Anthony JC, Petras H,
Kellam SG. Developmentally inspired drug prevention:
middle school outcomes in a school-based randomized
prevention trial. Drug Alcohol Depend. 2004;73:
149–158.
  RESEARCH AND PRACTICE 

21. Ialongo NS, Werthamer L, Kellam SG, Brown CH, 38. Catalano R, Morrison DM, Wells EA, Gillmore MR,
Wang S, Lin Y. Proximal impact of two first-grade pre-
ventive interventions on the early risk behaviors for
later substance abuse, depression, and antisocial behav-
Iritani B, Hawkins JD. Ethnic differences in family
factors related to early drug initiation. J Stud Alcohol.
1992;53:208–217.
Shortage?
ior. Am J Community Psychol. 1999;27:599–641.
39. Clark PI, Scarisbrick-Hauser A, Gautam SP,
22. Lambert SF, Brown TL, Phillips CM, Ialongo NS. Wirk SJ. Anti-tobacco socialization in homes of African-
The relationship between perceptions of neighborhood American and white parents, and smoking and non-
characteristics and substance use among urban African smoking parents. J Adolesc Health. 1999;24:329–339.
American adolescents. Am J Community Psychol. 2004; A Warning Shot:
34:205–218. 40. Sussman S, Dent CW, Flay BR, Hansen WB,
Johnson CA; Centers for Disease Control and Preven- Influenza and the
23. Kellam S, Ialongo N, Brown H, et al. Attention tion. Psychosocial predictors of cigarette smoking onset
problems in first grade and shy and aggressive behav- by White, Black, Hispanic, and Asian adolescents in
2004 Flu Vaccine
iors as antecedents to later heavy or inhibited sub-
stance use. NIDA Res Monogr. 1989;95:368–369.
Southern California. MMWR Morbid Mortal Wkly Rep.
1987;36(suppl 4):11S–16S.
Shortage By Tim
24. Kandel D. Adolescent marihuana use: role of par- Brookes
41. Hirschi T. Crime and the family. In: Wilson J, ed.
ents and peers. Science. 1973;181:1067–1070.
Crime and Public Policy. San Francisco, Calif: Institute
25. Lloyd JJ, Anthony JC. Hanging out with the wrong for Contemporary Studies; 1983:53–68.
crowd: how much difference can parents make in an
urban environment? J Urban Health. 2003;80: 42. McMorris BJ, Uggen C. Alcohol and employment
383–399. in the transition to adulthood. J Health Soc Behav.
2000;41:276–294.
26. Bryant AL, Schulenberg J, Bachman JG, O’Malley PM,
Johnston LD. Understanding the links among school 43. Fichtenberg CM, Glantz SA. Effect of smoke-free
misbehavior, academic achievement, and cigarette use: workplaces on smoking behaviour: systematic review.
a national panel study of adolescents. Prev Sci. 2000;1: BMJ. 2002;325:188.
71–87.
27. Chilcoat HD, Dishion TJ, Anthony JC. Parent
monitoring and the incidence of drug sampling in
urban elementary school children. Am J Epidemiol.
1995;141:25–31.
28. Miech RA, Hauser RM. Socioeconomic status and
health at midlife. A comparison of educational attain-
ment with occupation-based indicators. Ann Epidemiol. This important book gives the facts
2001;11:75–84.
and the analysis necessary to under-
29. Ensminger ME, Forrest CB, Riley AW, et al. The
validity of measures of socioeconomic status of adoles-
stand what happened and how to
cents. J Adolesc Res. 2000;15:392–419. avoid similar crises in the future.
30. 2002 Maryland Adolescent Survey. Baltimore: ORDER your copy TODAY!
Maryland State Department of Education; 2003. ISBN 0-87553-049-4 •Softcover • 2005
31. Wilson WJ. When Work Disappears: The World of $13.97 APHA Members • $19.99 Nonmembers
the New Urban Poor. 1st ed. New York, NY: Knopf; 1996. plus shipping and handling
32. Entwisle DR, Alexander KL, Olson LS. Early work
histories of urban youth. Am Sociol Rev. 2000;64: American Public
279–297. Health Association
33. Johnston LD, O’Malley PM, Bachman JG, Publication Sales
Schulenberg JE. Monitoring the Future: National Survey Web: www.aphabookstore.org
Results on Drug Use, 1975–2005, Volume 1: Secondary E-mail: APHA@pbd.com
School Students. Bethesda, Md: National Institute on Tel: 888-320-APHA
Drug Abuse; 2006. NIH publication 06-5883. Fax: 888-361-APHA
34. Substance Abuse and Mental Health Services Ad-
ministration. Results From the 2004 National Survey on Don’t get caught with the Flu!!!
Drug Use and Health: National Findings. Rockville, Md:
Office of Applied Studies; 2004. NSDUH series H-28;
DHHS publication SMA 05-4062.
35. Wallace JM Jr, Bachman JG, O’Malley PM,
Schulenberg JE, Cooper SM, Johnston LD. Gender and
ethnic differences in smoking, drinking and illicit drug
use among American 8th, 10th, and 12th grade stu-
dents, 1976–2000. Addiction. 2003;98:225–234.
36. Ellickson PL, Orlando M, Tucker JS, Klein DJ.
From adolescence to young adulthood: racial/ethnic
disparities in smoking. Am J Public Health. 2004;94:
293–299.
37. Griesler PC, Kandel DB. Ethnic differences in
correlates of adolescent cigarette smoking. J Adolesc
Health. 1998;23:167–180.

2062 | Research and Practice | Peer Reviewed | Ramchand et al. American Journal of Public Health | November 2007, Vol 97, No. 11