Children and Youth Services Review 32 (2010) 1108–1120
Contents lists available at ScienceDirect
Children and Youth Services Review
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c h i l d yo u t h
Childhood predictors of adult substance abuse
Irma Arteaga a,⁎, Chin-Chih Chen b, Arthur J. Reynolds b,⁎
a
b
Department of Applied Economics, University of Minnesota- Twin Cities
Institute of Child Development, University of Minnesota- Twin Cities
a r t i c l e
i n f o
Available online 2 May 2010
Keywords:
Substance abuse
Drug use
Longitudinal design
Prediction
Child development
a b s t r a c t
Identification of the early determinants of substance abuse is a major focus of life course research. In this
study, we investigated the child, family, and school-related antecedents of the onset and prevalence of
substance abuse by age 26 for a cohort of 1208 low-income minority children in the Chicago Longitudinal
Study. Data on well-being have been collected prospectively since birth from administrative records, parents,
teachers, and children. Results indicated that the prevalence of substance abuse by age 26 was 26% (selfreports or criminal justice system records) with a median age of first use of 17. Probit regression analysis
indicated that substance abuse prevalence was primarily determined by gender (males had a higher rate),
trouble-making behavior by age 12, school mobility, and previous substance use. Family and peer predictors
included involvement in the child welfare system by age 9, parent expectations for school success at age 9,
parent substance abuse by child’s age 15, and deviant peer affiliation by age 16. Age of first substance use
was predicted by gender and race/ethnicity (males and Blacks had earlier incidence), involvement in the
child welfare system, and family risk status at age 8. As with prevalence, the pattern of predictors for males
was similar to the overall sample but the magnitude of effects was stronger. The predictors of the timing of
substance use dependency were gender, family conflict by age 5, involvement in the child welfare system,
social maturity at age 9, adolescent school mobility, and school dropout by age 16. Findings indicate that the
promotion of family involvement and positive school and social behaviors can reduce the risk of substance
abuse.
© 2010 Published by Elsevier Ltd.
1. Introduction
Substance abuse exacts high personal and social costs that have
been well-documented. Besides contributing to school underachievement, antisocial behavior, and mental health problems into adulthood
(Belcher & Shinitzky, 1998; Gilvarry & McArdle, 2007; Hawkins,
Catalano, & Miller, 1992; Merline, O’Malley, Schulenberg, Bachman,
& Johnston, 2004), substance use and abuse are linked to increased
expenditures for treatment in social service, criminal justice, and health
service systems (Monge et al., 1999). Given these serious consequences
and growing costs of treatment, identifying malleable risk factors is
critical for understanding as well as reducing and ultimately preventing
substance abuse and its negative consequences for well-being.
These consequences are magnified by the relatively high prevalence
of substance use among young people. While 10% to 15% of teenagers
have used cannabis by age 15 (Fergusson, Lynskey, & Horwood, 1993),
and 8.9% of 12–17 year olds have been diagnosed with substance use
disorders (Substance Abuse and Mental Health Services Administration,
2004), substance use problems become increasingly prevalent during
⁎ Corresponding authors. Institute of Child Development, University of Minnesota, 51 East
River Road, Minneapolis, MN 55455, United States.
E-mail addresses: [email protected] (I. Arteaga), [email protected] (A.J. Reynolds).
0190-7409/$ – see front matter © 2010 Published by Elsevier Ltd.
doi:10.1016/j.childyouth.2010.04.025
middle and late adolescence. Approximately 25% and 56% of teenagers
by grade 8 and grade 12, respectively, have experienced illicit drug
use (Johnston, O’Malley, & Bachman, 1999). Similar increasing prevalence of drug use was found in the National Clearinghouse on Alcohol
and Drug Information, which reported that 20% of 16–17 year olds have
used marijuana at least once per week compared to 12% of 12–13 year
olds. On the other hand, a declining pattern of use has been found after
young adulthood (Merline et al., 2004). Based on these high prevalence
rates and associated problem behaviors, it is necessary to explore the
connection between substance use and substance abuse, which
influence the long-term well-being.
1.1. Predictors of substance use problems
Many risk and protective factors influencing drug use and abuse in
adolescence and early adulthood have been identified (Hawkins et al.,
1992; Young et al., 2002). Risk factors indicate that the increase in a
specific variable is associated with increased likelihood of substance
use/abuse. Protective factors denote that the increase in a specific
attribute or behavior is associated with decreased likelihood of substance use/abuse. Individual, family, school-based, and peer contexts
are most associated with the development of substance abuse.
Hawkins et al. (1992) and Gilvarry and McArdle (2007) found that
early antisocial behavior is associated with increased risk of substance
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
use. For example, boys exhibiting persistent aggressive behavior are
more likely to use drugs. Girls having higher levels of anxiety are at
increased risk of drug use. Conduct problems and hyperactivity in
childhood and adolescence are more likely to be associated with drug
use. Additionally, Merline et al. (2004) found that individual substance use history during the senior year of high school predicts
substance use in early adulthood.
Early family adversity is predictive of an increased risk of substance
use and abuse in late adolescence and young adulthood (Siebenbruner,
Englund, Egeland, & Hudson, 2006). Specifically, adverse family experiences in childhood including emotional and physical abuse/neglect,
sexual abuse, family dysfunction such as parental separation or divorce,
single-mother status, parental substance abuse, and larger family size
increased the risk of substance abuse (Bennett & Kemper, 1994; Dube
et al., 2003; Hawkins et al., 1992; Osler, Nordentoft, & Andersen, 2006).
Hawkins et al. (1992) found that a family history of alcoholism, parental
use of illegal drugs, poor family management practices, family conflict,
and low bonding to family were associated with drug use in adolescence. Reinhertz, Giaconia, Hauf, Wasserman, and Paradis (2000)
indicated that sibling substance use disorders, larger family size, lower socioeconomic status, and parental substance abuse, and younger
parents were associated with an increased drug abuse in young adulthood. The family impact on the development of substance abuse varied
by gender. Family environment was more influential on drug use problems for girls than boys (Block, Block, & Keyes, 1988).
Besides individual and family factors, school factors play an important role in substance use in adolescence and young adulthood.
Osler et al. (2006) found that participants who disliked school had a
greater risk of substance abuse. Evidence also shows the relationships
between academic risk factors including enrollment status and last
semester’s grades and past year marijuana use. School dropout is
associated with a higher risk of substance use. Students with poor
academic achievement had an increased risk of substance use problems (Hawkins et al., 1992). Negative relationships were also found
between the lack of school attachment and commitment and the
development of substance abuse (Young et al., 2002).
Peer behaviors and attitudes also influence the emergence of
tobacco and drug use (Kandel, Simcha-Fagan, & Davies, 1986). Brook
et al. (1998) found that peer factors have a direct and significant effect
on adolescents’ use of drugs. Peer pressure has impact on adolescent
drug use (Robin & Johnson, 1996). Hawkins et al. (1992) found that
early peer rejection and social forces to use drugs explained the drug
and alcohol problems in adolescence. What probably occurs is that
individuals affiliated with peers who use drugs, consume alcohol, and
get in trouble are more likely to use drugs and become drug-abusers
in young adulthood. For example, Fergusson, Swain-Campbell, and
Horwood (2002) found evidence of consistent associations between
deviant peer affiliations and crime/substance abuse in adolescents
and young adults with the statistical control of confounding factors.
The effect of peer factors on drug use problems, however, decreases
with age.
Other studies have shown that neighborhood adversity is one of the
critical contextual factors related to substance abuse. Residence in highpoverty neighborhoods increases the risk of drug use (Galea, Ahern, &
Vlahov, 2003). Similarly, Cooper, Friedman, Tempalski, and Friedman
(2007) found that residential segregation (isolation and concentration)
is positively related to substance abuse in African-Americans.
Finally, previous research reveals that the most effective interventions for preventing substance abuse as well as delinquency, and
violence target early risk factors before antisocial behavior occurs
(Webster-Stratton & Taylor, 2001). School-age programs that strengthen social skills and positive classroom behavior, parent–child interactions, and a broader set of resistance skills (Belcher & Shinitzky, 1998)
demonstrate effectiveness in reducing substance abuse and associated
problem behaviors. Preschool programs that improve school readiness
and achievement as well as family and school support also show
1109
evidence of reducing later behavior problems that lead to substance
abuse (Campbell, Ramey, Pungello, Sparling & Miller-Johnson, 2002;
Reynolds, Temple, Robertson, & Mann, 2001; Reynolds et al., 2007;
Schweinhart, Barnes & Weikart, 1993; Schweinhart et al., 2005).
However, few if any studies have investigated the long-term effects of
early interventions into adulthood. While preschool programs have
examined adult well-being, substance abuse has rarely been assessed.
Despite the progress in understanding the development of substance abuse and related problems, several limitations in knowledge
remain. Most previous studies were analyzed using cross-sectional
data. Few investigated the antecedents of substance abuse prospectively and with longitudinal designs. In addition, the determinants of
the onset of substance use and progression to substance abuse or
dependency were not well-documented. Moreover, few studies have
examined comprehensive models of the predictors of substance abuse
that include child, family, school, neighborhood, and peer factors
and experiences measured over the entire period of childhood and
adolescence. Such an approach is consistent with a developmental
and life course perspective.
1.2. Present study
In this study, we investigate the predictors of early adult substance
abuse using data from the Chicago Longitudinal Study, an on-going
prospective investigation of the life course of a cohort of 1539 lowincome minority children growing up in the inner city. Three major
questions are addressed:
A. What factors predict the incidence of substance abuse by age 26?
B. Which factors predict the age at first substance use?
C. Which factors predict substance dependency?
In addition to the focus on children with demographic attributes
that place them at risk of later problematic behaviors, the study has a
number of strengths that can advance knowledge about the development of substance use and abuse. First, extensive data have been
collected from birth to early adulthood on many individual, family,
school, and community influences. A significant number of measures
ranging from participation in intervention to parenting and school
practices are alterable and can be a focus of preventive efforts. Second,
the study has a high rate of sample recovery. Over 80% of the original
sample has provided data on a regular basis from early childhood to
early adulthood. Finally, multiple sources of information are available
on substance use and abuse, including self-reports and administrative
records. This feature strengthens the construct and measurement
validity of indicators.
2. Method
2.1. Sample and data
The sample participates in the Chicago Longitudinal Study (CLS,
2005), a prospective investigation of the life course development of a
cohort of 1539 low-income minority children (93% African American,
7% Hispanic) growing up in the inner city. Although the major focus is
assessing the impacts of the Child–Parent Center program on health
and well-being, the CLS also investigates the influences of child,
family, and school factors on life course development. The original
study sample included a cohort of 989 children enrolled in the CPC
program in 20 sites in preschool and kindergarten during 1985–1986.
It also included a matched cohort of 550 children of the same age
enrolled in alternative preschool programs in 5 different Chicago
public schools, which were randomly selected from 27 sites
participating in the Chicago Effective School Project in similar,
impoverished neighborhoods (Reynolds et al., 2001).
1110
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
The sample for the current study was 1208 participants (78% of
the original sample), who completed the adult survey at ages 22–24
or who had official drug criminal records (i.e., drug conviction) at
county, state, and/or federal levels by age 26. Table 1 shows that
the characteristics of the children and families in the study sample
matched those of the original sample. For example, the samples were
similar on gender composition, parent education and single-parent
status, overall family risk status, CPC participation, and children’s
kindergarten achievement test scores.
Data in the CLS on demographic characteristics and individual,
family and school experiences were collected prospectively. The CLS
includes (among others) the following data sources: teacher surveys
(yearly, kindergarten to grade 7), parent surveys (student grades 2, 4–
6, and 11), student surveys (grades 3–6 and 10), young adult survey
(ages 22–24), the Illinois Department of Public Health, Cook County
Juvenile and Circuit Court records, and the Illinois Public Assistance
Research Database (ILPARD) maintained by Chapin Hall Center for
Children at the University of Chicago. School data and peer variables
were collected from the Chicago Public Schools. Family data that
include public assistance, education, employment, and family structure were collected on application to public assistance from February
1, 1989 to August 31, 2008.
2.2. Explanatory variables
Table 2 shows the explanatory variables of the study by demographic, early and middle childhood, and adolescent attributes including family, school, and peer influences. Many previous studies
have emphasized early childhood predictors (Dube et al., 2003; Osler
et al., 2006; Siebenbruner et al., 2006) while others have focused on
school and peer influences in early adolescence (Hawkins et al., 1992;
Brook et al., 1998; Fergusson et al., 2002). In this study, we focus on a
comprehensive set of predictors spanning preschool to adolescence
and inclusive of child, family, and school-related influences. These are
defined below.
Table 2
Descriptive statistics of explanatory and outcome variables (n = 1208).
Variable
Demographic characteristics
Gender (female)
Race/ethnicity (African-American)
Early and middle childhood factors
CPC preschool participation (ages 3–4)
Child protection services (ages 4–9)
Family conflict (ages 5–10)
Parent substance abuse experience
(ages 5–10)
School mobility (ages 6–9)
CPC school-age participation (ages 6–9)
Social maturity (ages 7–9)
Family risk (age 8)
Parent expectations (ages 8–10)
Intrinsic motivation (ages 9–10)
No trouble-making behavior (ages 9–12)
Reading achievement (age 10)
Adolescence factors
School mobility (ages 10–14)
School quality (ages 10–14)
Parent substance abuse experience
(ages 10–15)
Personal substance use experience
(ages 10–15)
School mobility (ages 14–18)
Deviant peer affiliation (age 16)
School dropout (by age 16)
Outcomes
Age of first substance use
Substance abuse
Substance dependency
(among those with use) a
Length of time from use to dependency
(in years)b
Mean
Std. dev.
Min
Max
.52
.94
.50
.24
0
0
1
1
.65
.10
.06
.05
.48
.30
.24
.21
0
0
0
0
1
1
1
1
.74
.57
19.31
4.23
3.39
−.01
6.06
102.83
.85
.50
4.71
1.79
.87
.79
1.76
14.66
4
1
30
8
5
1.7
12
150
.97
.13
.09
.99
.34
.28
0
0
7
0
1
− 3.2
1
53
0
0
0
0
.03
.17
0
1
.21
.39
.13
.48
.49
.33
0
0
0
3
1
1
17.21
.26
.38
3.20
.44
.49
9
0
0
26
1
1
5.49
2.97
3
16
4
1
1
CPC = Child–Parent Center program.
a
Dichotomous variable that takes a value of 1 if the individual was classified as drug
dependent and 0 otherwise (sample size is 440).
b
Information was available for one-year, two-year or four-year periods. When a
participant satisfied the definition for dependency, he/she was classified as substance
dependent for the whole period (e.g. 4 years). The sample size for length of time from
use to dependency includes only participants who were classified as substance
dependent (n = 168).
Table 1
Background characteristics for the original and study samples.
Variable
Child is female
Child is Black
Mother was younger than age 18
at child’s birtha
School neighborhood with
N60% low incomea
CPC preschool participation (ages 3–4)
Child protection services (ages 4–9)
ITBS word analysis scores (age 6)
ITBS math scores (age 6)
CPC school-age participation (ages 6–9)
Mother is a single-parent (age 8)a
Four or more children in family (age 8)a
Mother did not complete high
school (age 8)a
Mother is not employed (age 8)a
Child eligible for fully subsidized
lunches (age 8)a
Mother received AFDC (age 8)a
Family risk (age 8)
Original sample
(n = 1531)
Study sample
(n = 1208)
Mean
Mean
Std. dev.
2.2.1. Gender
Males were coded 0 and females were coded 1. Data were from
school records upon entry to preschool or kindergarten.
Std. dev.
0.502
0.930
0.167
0.500
0.256
0.373
0.516
0.938
0.172
0.500
0.241
0.377
0.760
0.427

0.760
0.427
0.643
0.098
63.770
56.666
0.552
0.608
0.323
0.440
0.479
0.298
13.300
14.825
0.497
0.488
0.468
0.497
0.646
0.097
64.343
57.228
0.565
0.594
0.329
0.436
0.479
0.296
13.316
14.752
0.496
0.491
0.470
0.496
0.524
0.833
0.500
0.373
0.519
0.836
0.500
0.370
0.600
4.250
0.490
1.792
0.587
4.230
0.493
1.791
Note. Of the original sample of 1539, 8 participants had insufficient information and were
excluded from these statistics. ITBS = Iowa Tests of Basic Skills (end of kindergarten).
CPC = Child–Parent Center program.
a
Included in the family risk index (family risk), which is a sum of 8 dichotomous
indicators.
2.2.2. Race/ethnicity
This was also measured from school records. Hispanic children
were coded 0 and African-American children were coded 1.
2.2.3. CPC preschool participation
This dichotomous variable indicated whether the child attended a
CPC preschool program at age 3 or 4 using administrative data from
the Chicago Public Schools. Children who enrolled at age 3 had two
years of this part-day program and those who enrolled at age 4 had
one year. The comparison group was coded 0 and they participated in
the usual early educational intervention for children at risk, which
was a full-day kindergarten program in the Chicago Public Schools.
Although none participated in CPC preschool, 15% of the comparison
group attended the Head Start preschool program.
2.2.4. CPC school-age participation
This dichotomous variable indicated whether the child attended
the school-age component of the CPC program for one or more years
(ages 6–9) from the first grade to the third grade (ages 6–9). The schoolage programs were affiliated with the same elementary schools as the
preschool program. Data were from school administrative records of
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
1111
the Chicago Public Schools. Those coded 0 did not attend the CPC schoolage program.
ment) showed construct independence from child ratings of school
adjustment and performance.
2.2.5. Involvement in child protective services
The dichotomous variable indicated whether the child or family
received child protection services at ages 4 to 9 based on the reports of
the Child Protective Services Division of the Illinois Department of
Child and Family Services and/or petitions to the Cook County Juvenile
Court.
2.2.11. Intrinsic motivation, ages 9–10
This 10-item scale of achievement motivation was self-reported in
the spring of third and fourth grades as part of the student survey.
Rated on a 3-point scale from not much (1) to (3) often, the items
were as follows: (a) I like to learn things, (b) I like to write stories, (c) I
get bored in school [reverse coded], (d) I like to read, (e) I like science,
(f) I like to have books read to me, (g) like to do math, (h) I have
fun at school, (i) learning is fun, and (j) I like to work on worksheets.
Z-scores were analyzed and those available at both ages were averaged. The internal-consistency reliability coefficient was .66.
2.2.6. Frequent family conflict
This dichotomous variable indicated the presence of frequent
family conflict when the participant was 5 to 10 years of age. In a
retrospective life event checklist in the adult survey, sample members
were asked the following: “We are interested in major events that
have occurred in your life. Please indicate if any of these events have
occurred in your life.” “If yes, how old were you when this happened?”
One life event was “frequent family conflict.” The validity of these
retrospective reports was corroborated by findings that family conflict
was positively correlated with risks measured by other sources (e.g.,
child maltreatment, family risk status) and negatively correlated with
child outcomes such as school achievement.
2.2.7. Parent substance abuse experience
Two dichotomous variables were used to measure the history of
parental problems with substance abuse when the participants were
ages (a) 5 to 10 and (b) 10 to 15. Data came from retrospective reports
on the adult survey at ages 22–24. In the life events checklist, participants were asked the following: “We are interested in major life
events that have occurred in your life. Please indicate if any of these
events have occurred in your life?” “If yes, how old were you when
this happened?” One of these life events was “problem of substance
abuse of parent.” In support of validity, these retrospective reports
were positively correlated with other risk factors and experiences
(e.g., child maltreatment, family risk status) and negatively associated
with indicators of children’s school progress and achievement.
2.2.8. Social maturity, ages 7–9
Teacher ratings of children’s social adjustment were assessed by
a 6-item scale measured from first to third grade. The items were as
follows: concentrates on work, follows directions, is self-confident,
participates in group discussions, gets along well with others, and
takes responsibility for actions. Responses were coded from poor/
not at all (1) to excellent/very much (5). We used the average of the
available scores over the three-year period. Reliability (alphas N .90)
and predictive validity of the scale are high (Ou & Reynolds, 2008).
2.2.9. Family risk index
This continuous variable was the sum of eight dichotomouslycoded family risk factors measured at age 8 from family surveys, school
records, and social service records. They included (a) single-parent
family status, (b) mother did not completed high school, (c) mother
was under age 18 at child’s birth, (d) family participation in the public
assistance programs, (e) mother unemployment status, (f) free lunch
eligibility, (g) 4 or more children in the household, and (h) residence in
low-income neighborhoods.
2.2.10. Parent expectations of child’s progress, ages 8–10
This variable was the average of teacher ratings over grades 2 to
4 on the item “Parent’s expectations of child”. Responses ranged
from “poor/not at all” (1) to “excellent/much” (5). The item was part
of a survey of children’s classroom adjustment and school experiences
(i.e., “Please rate the child on the following characteristics.” Scores
for two or more occasions were averaged. Teacher ratings of parent
expectations (and related parenting measures such as school involve-
2.2.12. No trouble-making behavior, ages 9 to 12
This continuous variable of problem behavior was measured by a
child-rated 4-item scale over grades 3 to 6. The items were (a) I get
in trouble at school, (b) I fight at school, (c) I get in trouble at home,
and (d) I follow class rules [reverse coded]. Item responses ranged
from (1) often to not much (3). The scale was coded to index the
extent of no problem behavior. Higher scores indicated more positive
(less trouble-making) behavior; see also Topitzes et al. (2009). Scores
available on 2 or more occasions were averaged. Given the relatively
few items, the internal-consistency reliability of the scale was .55.
2.2.13. Reading achievement at age 10
This measure was reading comprehension as assessed by the
reading comprehension subtest of the Iowa Test of Basic Skills (ITBS;
Form J, Level 9 or 10; Hieronymus & Hoover, 1990). The subtest
included 49 items on understanding text passages (α = .93). Developmental standard scores were analyzed based on 1988 norms.
2.2.14. Personal substance use experience
This dichotomous variable indicated whether the participants had
ever used illicit drugs up to age 15. Data came from multiple sources
including youth survey at ages 15–16 (“I use alcohol or drugs for
nonmedical reasons.”), retrospective reports from the life event
checklist in the adult survey (history of “problem of personal substance abuse”; see description of frequent family conflict above), and
arrest records from the juvenile justice system.
2.2.15. School moves
The number of school moves was measured at three different age
periods: 6 to 9, 10 to14, and 14 to 18, corresponding to kindergarten to
grade 3, grade 4 to 8, and grade 8 to 12. Data were derived from annual
records of school enrollment in the Chicago Public Schools. Withinyear and normative (expected due to grade promotion) moves were
not included. Many previous studies show that non-normative school
moves are associated with lower levels of school achievement and
attainment (Mehana & Reynolds, 2004; Ou & Reynolds, 2008).
2.2.16. School quality
This was a dichotomous variable indicating whether study participants over ages 10–14 (grades 4–8) attended either (a) selective citywide magnet schools or (b) those in which 40% or more of the student
body was at/above national norms in ITBS reading and math. Data came
from the State of Illinois Report Card from the fifth grade year (1990–
1991). Scores were highly consistent across the measured grades. Magnet schools have selective enrollment policies that require good school
performance and high expectations for success.
2.2.17. School dropout
This dichotomous variable indicated if the child had ever dropped
out of school by age 16. Data came from school system records supplemented with participant surveys. We used this early measure of
1112
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
dropout to avoid confounding the direction of influence between
dropout and substance use.
2.2.18. Deviant peer affiliation
The dichotomous variable indicated whether most of the
participants’ friends exhibited any of these behaviors (“Most of my
friends:”): (a) skip school a lot, (b) drink alcohol, (c) or have
experimented with drugs. Otherwise, the code was 0. Data were
based on self-reports at the age 15–16 survey.
2.3. Outcome measures
At ages 15–16 and 22–24, sample members were surveyed or
interviewed about their well-being including school progress and
performance, attitudes toward education, socio-emotional development, health behavior, school support, peer relations and family experiences. As part of these surveys, participants reported their use of
tobacco, cannabis and other illicit drugs. The ages 22–24 survey/
interview provided the most comprehensive assessment. This selfreport information was combined with arrest records from the
juvenile and criminal justice system on drug possession, manufacturing/delivery, and conspiracy to construct the following measures: age
of first substance use, substance abuse, and substance dependency.
Table 2 shows the summary statistics of the measures. Table 3 shows
the distribution of substance use by age of first use.
2.3.1. Age of first substance use
This was a continuous variable that indicates the age of first use of
illicit drugs including marijuana and harder drugs from self-reports
(excluding alcohol) over ages 15 to 24 from the two major survey/
interview assessments. These data were supplemented with official
juvenile and adult justice system records of arrests for drug possession or manufacturing by age 26. We used the earliest identified
age for any of these sources.
Each self-report assessment included items about current or past
illicit drug use. In the ages 22–24 survey/interview, participants
answered the following questions: “Have you ever: “smoked marijuana” or “used drugs harder than marijuana.” “If yes, how often do you
currently use it?” The frequency of use was classified as “almost every
day”, “a few times a week”, “a few times a month”, “less than once a
month”, “a few times a year”, and “never”. Individuals also were asked
Table 3
Age of first substance use.
Age of first substance use
Frequency
Percent
Cumulative percent
0
Less than 10
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
768
3
3
3
5
6
55
74
49
54
46
41
26
28
20
15
5
5
2
63.6%
0.2%
0.2%
0.2%
0.4%
0.5%
4.6%
6.1%
4.1%
4.5%
3.8%
3.4%
2.2%
2.3%
1.7%
1.2%
0.4%
0.4%
0.2%
63.6%
63.8%
64.1%
64.3%
64.7%
65.2%
69.8%
75.9%
80.0%
84.4%
88.2%
91.6%
93.8%
96.1%
97.8%
99.0%
99.4%
99.8%
100.0%
Note. Participants with no substance use are coded 0.
about their history of “problems of personal substance abuse” over
three age periods (5 to 10, 10 to 15, and 16 to 22–24). This question
was part of a list of 17 life events. For the analysis, participants were
classified as users if they reported any use (regardless of frequency) on
at least one occasion. In some cases, justice system data on drug arrests
and convictions were used to determine the age of first substance
use as were records of DUI (driving under the influence of drugs or
alcohol) from the Illinois Department of Motor Vehicles.
2.3.2. Substance abuse
This was a dichotomous variable of the overall prevalence from
ages 16 to 26 based on self-reports and administrative records. Sample
members were assigned a value of 1 if any of the following behaviors
applied (otherwise 0): (a) self-report at ages 22–24 of “any personal
substance abuse problem from age 16 and above”, (b) self-report
of presently “smoking marijuana almost everyday” or “using drugs
harder than marijuana a few times per week or more”, (c) justice
system records from ages 16 to 26 of being found guilty of drug
possession, manufacturing/delivery, or conspiracy as well as substance-related disorderly conduct, (d) use of substance abuse services
based on the ages 22–24 survey, and (e) Department of Motor Vehicle
record of a DUI conviction since age 18. We also classified individuals
for substance abuse if they used drugs at least a few times a week and
had multiple drug-related arrests, were high school dropouts, or were
not employed after age 18.
Our definition was consistent with the Diagnostic and Statistical
Manual of Mental Disorders (DSM-IV) of the American Psychiatric
Association (2000), which defines substance abuse as “a maladaptive
pattern of substance use leading to clinically significant impairment or
distress, as manifested by one or more of the following, occurring
within a 12-month period…(1) recurrent substance use resulting in
a failure to fulfill major role obligations at work, school, or home…
(2) recurrent substance use in situations in which it is physically
hazardous…(3) recurrent substance-related legal problems…(4) continued substance use despite having persistent or recurrent social or
interpersonal problems…”(p. 199)
2.3.3. Substance use dependency
The primary measure was a continuous variable that indicates
the number of years from the age of first substance use to substance
dependency by age 26. Based on the same self-reports and administrative records listed above, the sample size was the 440 individuals
with any lifetime substance use. From this sample, we identified 168
as having substance dependency. The presence of substance dependency was based on DSM-IV and DSM-IV-TR criteria of the American
Psychiatric Association (2000). These included tolerance, frequency
of substance use, time spent by the individual in activities necessary
to obtain drugs, and important activities that were reduced or given
up due to substance use (e.g. school retention, school dropout).
If three of these criteria were satisfied at any time in the same
12-month period, then the individual was categorized as substance
dependent. In order to categorize dependency for long periods of
time, these criteria had to be satisfied for uninterrupted periods. In
supplemental analysis, we also tested the predictors of the dichotomous measure of substance use dependency among those who had
any history of use since age 16 (n = 440) and for the total study
sample (n = 1208).
2.4. Missing data
We used multiple imputation with the expectation-maximization
(EM; Schafer, 1997) method for the explanatory variables. The assumption is that values are missing at random (MAR). We imputed
the variables depicted in Appendix A. Each variable was imputed on a
set of explanatory variables by an appropriate model based on the
imputed variable. The regression model is an OLS if the imputed
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
variable is a continuous variable or a logit model if it is a binary
variable. The set of explanatory variables used for our imputation are
background characteristics: participant’s gender, school site, years
attending CPC preschool, kindergarten dosage (half-day or full-day
kindergarten) and risk index at child’s birth (marital status of parents,
mother completed high school, mother’s age, free lunch eligibility,
mother’s participation in the public aid program AFDC (Aid to Families
with Dependent Children), school neighborhood poverty, mothers
employment status, and number of children in the household).
2.5. Data analysis
The first research question examined the factors that predict the
incidence of substance abuse by age 26. Consistent with prior research
(Fergusson et al., 1993; Merline et al., 2004), we used probit regression analysis for model estimation. The probit model assumes
that the probability distribution function is normal. We used the
maximum likelihood estimation method using STATA in which the
log-likelihood function is maximized as follows:

n
o
ln L = ∑ ln Φ xj β + ∑ ln 1−Φ xj β
j∈S
j∉S
ð1Þ
1113
males. Consequently, we tested the predictors of abuse for males as
well as the overall sample.
As shown in Table 3, the age of first use ranged from 9 to 26 with a
mean of 17.2 years of age (median of age 17). Most of the cases had an
age of first use from 14 to 19. Of the 440 participants with identified
use, 38% (168 participants) met the definition of dependency by age
26. The average number of years from first use to dependency (among
168 with dependency) was 5.5 with a range of 3 to 16.
Appendix B shows the correlations among the explanatory and
outcome variables. In general, the correlations among the explanatory
variables were low to moderate. The highest correlations included
teacher ratings of social maturity and parent expectations for child’s
school progress (r = .61) and reading comprehension and social
maturity (r = .56). The association between personal substance use
and parent substance use at child ages 10 to 15 was modest (r = .23).
The explanatory variables most correlated with substance abuse
prevalence were gender (r = − 0.47 [females had lower prevalence]),
personal substance use (r = 0.25), social maturity (r =−0.24), and
parent expectations (r = − 0.22).
3.2. Factors predicting substance abuse in early adulthood
3. Results
Findings of the explanatory model for substance abuse are shown
in Table 4. Marginal effects (b) are reported and were converted from
the probit coefficients. Marginal effects denote the change in the
outcome in percentage points for each 1-unit change in the predictor
after controlling for the influence of other model variables.
For the overall sample, three predictors were associated with a
decreased likelihood of substance abuse by age 26: gender (b = −0.35
[in favor of females]; p b .001), no trouble-making behavior (b = −0.03;
p b .001), and parent expectations of child’s school progress (b = −0.05,
p b .05). For example, an improvement of 2 points on the no trouble scale
is associated with a 6 percentage point reduction in substance abuse.
A similar change in parent expectations corresponds to a 10 percentage point reduction in substance abuse.
Five other predictors were linked to an increased likelihood of
substance abuse: involvement in child protection services (b = 0.17,
p b .001), parent substance abuse at child’s ages 10 to 15 (b = 0.10,
p b .05), school mobility at ages 14 to 18 (b = 0.04, p b .05), previous
personal substance use at ages 10 to 15 (b = 0.46, p b .001), and
deviant peer affiliation (b = 0.08, p b .01). These findings indicate
that family and school contexts contribute substantially to substance abuse even after previous substance use is taken into
account.
As further shown in Table 4, the pattern of findings for males
was similar to that of the overall sample (because of sample size
limitations the model was not estimated for females). School mobility
at ages 10–14 was the only additional predictor for males (b = 0.06,
p b .05). Parent expectations, however, did not predict substance
abuse. With the exception of substance use experience at ages 10 to
15, the size of the coefficients was larger for males than for the overall
sample. For the school mobility measures, the magnitude of influence
was double. The influence of no trouble-making (b = −0.04, p b .001),
child protective services (b = 0.22, p b .01), and deviant peer affiliation
(b = 0.13, p b .01) also were stronger.
3.1. Descriptive findings
3.3. Factors predicting the age of first substance use
By age 26, the study sample had a prevalence rate of substance
abuse of 26% (see Table 2). The rate of substance use was 36% (a less
restrictive definition). The rate of substance abuse included not only
self-reports of frequent use or treatment but justice system records of
convictions for drug possession. Given the similarity between the
study sample and the original sample on many child and family
characteristics (see Table 1), the prevalence rate would likely hold for
the original sample. Nearly 90% of those with substance abuse were
We used survival analysis to examine the predictors of the age of
onset of substance use. Findings are reported in Table 5 and are based
on the Cox proportional hazard model. The hazard ratio (HR) indicates
the relative risk of substance use per 1-unit change in the explanatory
variable controlling for other model variables. Four variables were
significant predictors of substance use onset. Involvement in child
protective services at ages 4 to 9 was associated with an increased
hazard (HR = 1.39; p b .01) or relative risk of earlier substance use.
In the model, Φ represents the cumulative normal and (xβ) the
probabilities of substance abuse. The primary metric for interpretation
is the marginal effect, the change in the probability of substance abuse
associated with a one unit change in x.
For the second (predictors of age of first use) and third (time from
first use to substance dependency) research questions, we used
survival analysis. This approach has been rarely used to investigate a
comprehensive set of predictors into adolescence. Based on the Cox
(1972) proportional hazard model (see also Gutierrez, 2002), the
hazard is the probably (relative risk) that participants will be a
substance user as they age. The model assessed the predictors of “how
long does it take?” to become a user. The duration of time is up to age
26. To analyze the use to substance dependency question, survival
analysis was used to test the predictors of being classified as substance
dependent given that an individual was already a substance user. The
hazard (relative risk) is the probability that an individual will become
dependent as the length of time (years is unit time interval) between
use and dependency increases.
To illustrate the Cox proportional model for research question 3,
let λ denote the hazard rate, λ0 be the baseline hazard (individual
heterogeneity), xi represents a constant term and a set of variables that
are assumed not to change from time T = 0 until the “failure time”,
T = ti, where T denotes the total length of duration and t denotes time.
The approach is a semi-parametric method that analyzes the effect of
covariates (βs), without requiring the estimation of λ0, on the hazard
rate.
0
λðti Þ = exp xi β λ0 ðti Þ
ð2Þ
1114
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
Table 4
Factors predicting the incidence of substance abuse by age 26.
Total sample (n = 1208)
Marginal effect
Gender (female)
Race/ethnicity (African-American)
CPC preschool participation (ages 3–4)
Child protection services (ages 4–9)
Family conflict (ages 5–10)
Parent sub abuse experience (ages 5–10)
School mobility (ages 6–9)
CPC school-age participation (ages 6–9)
Social maturity (ages 7–9)
Family risk (age 8)
Parent expectations (ages 8–10)
Intrinsic motivation (ages 9–10)
No trouble-making behavior (ages 9–12)
Reading achievement (age 10)
School mobility (ages 10–14)
School quality (ages 10–14)
Parent substance abuse experience (ages 10–15)
Personal substance use experience (ages 10–15)
School mobility (ages 14–18)
Deviant peer affiliation (age 16)
School dropout (by age 16)
Wald Chi-squared
Prob. N Chi-squared
Pseudo R-squared
Male group (n = 585)
Robust std. error
p-value
0.025
0.033
0.023
0.057
0.041
0.059
0.017
0.032
0.004
0.006
0.021
0.018
0.006
0.001
0.016
0.044
0.055
0.115
0.020
0.025
0.032
0.000***
0.151
0.509
0.000***
0.943
0.687
0.098
0.219
0.896
0.752
0.012*
0.181
0.000***
0.345
0.060
0.237
0.037
0.000***
0.018*
0.002**
0.218
− 0.350
0.060
− 0.010
0.170
− 0.010
− 0.030
− 0.030
0.040
0.000
0.000
− 0.050
− 0.020
− 0.030
0.000
0.030
− 0.060
0.100
0.460
0.040
0.080
0.030
2,968.73
0.00
0.33
Marginal effect
Robust std. error
p-value
0.071
0.035
0.067
0.076
0.137
0.028
0.042
0.006
0.014
0.035
0.030
0.011
0.002
0.025
0.081
0.094
0.081
0.045
0.048
0.050
0.322
0.303
0.002**
0.873
0.689
0.211
0.283
0.573
0.874
0.075
0.270
0.000***
0.390
0.016*
0.285
0.150
0.000***
0.005**
0.006**
0.575
0.080
− 0.030
0.220
− 0.020
− 0.060
− 0.040
0.050
0.000
0.000
− 0.060
− 0.040
− 0.040
0.000
0.060
− 0.080
0.140
0.400
0.110
0.130
0.020
2,207.66
0.00
0.16
Note: ***p b 0.001, **p b 0.01, *p b 0.05. Probit coefficients are transformed to marginal effects (dy/dx), which are the change in substance abuse (in percentage points) per 1-unit
change in the explanatory variables.
Relative to those not involved in child protective services, children
in protective services had a 39% increased risk. Family risk status
(HR = 1.06; p b .05) also was associated with an increased risk of earlier
onset as was school mobility at ages 10–14 (HR = 1.32, p b .001). Each
additional school move increased the risk of earlier substance use by
32%. Conversely, Black participants had a significantly lower risk of
earlier onset (HR = 0.74, p b .05). This is a 26% lower risk than Hispanic
participants.
The pattern of findings for males was nearly identical to that of the
overall sample. School mobility, family risk status, and child protective
service involvement. Although race (HR = 0.73 [Black participants had
lower risk]; p = .06) and frequent family conflict (HR = 1.4; p = .07)
were marginally associated with age of onset but the magnitude
of effects was relatively large. Overall, findings suggest that family
adversity exerts a sizable impact on the onset of substance use.
3.4. Factors predicting substance dependency
Survival analysis was also used to predict the hazard (relative risk)
from first substance use to dependency. This is a negative indicator of
desistence. The sample size for this analysis was the 440 participants
who had substance use by age 26. As shown in Table 6, there were 6
significant predictors of relative risk of substance dependency. Gender
(HR = 0.29, p b .001 [females had lower risk]1) and social maturity
(HR = 0.97, p b .05) were associated with a decreased risk of substance
dependency by age 26. A 5-point increase in social maturity, for
example, would decrease the hazard of dependency by 15% ([(1 −
0.97) × 5]). Being female and social maturity are protective against
substance use problems and increase desistence.
On the other hand, involvement in child protective services by
age 9 (HR = 1.83, p b .01), frequent family conflict at ages 5 to 10
Table 5
Factors predicting age of first substance use.
All
Hazard ratio
Female
Black
CPC preschool participation (ages 3–4)
Child protection services (ages 4–9)
Family conflict (ages 5–10)
School mobility (ages 6–9)
CPC school-age participation (ages 6–9)
Social maturity (ages 7–9)
Family risk (age 8)
Parent expectations (ages 8–10)
Intrinsic motivation (ages 9–10)
No trouble-making behavior (ages 9–12)
Reading achievement (age 10)
School mobility (ages 10–14)
School quality (ages 10–14)
Parent substance abuse experience (ages 10–15)
Deviant peer affiliation (age 16)
Wald Chi-squared
Prob. N Chi-squared
Note: ***p b 0.001, **p b 0.01, *p b 0.05.
1.132
0.735
0.867
1.394
1.203
1.044
1.027
1.005
1.055
1.034
1.056
1.023
0.999
1.315
0.820
0.981
1.098
41.64
0.00
Males
Std. err.
z
pNz
0.140
0.104
0.086
0.167
0.169
0.056
0.099
0.011
0.029
0.062
0.058
0.029
0.004
0.096
0.125
0.145
0.095
1.010
− 2.170
− 1.450
2.780
1.310
0.800
0.280
0.440
1.960
0.550
1.000
0.820
− 0.240
3.750
− 1.300
− 0.130
1.080
0.314
0.030*
0.148
0.005**
0.189
0.426
0.781
0.662
0.050*
0.579
0.315
0.413
0.814
0.000***
0.195
0.898
0.279
Hazard ratio
Std. err.
z
pNz
0.732
0.891
1.407
1.353
0.998
1.038
1.005
1.067
0.932
1.060
1.027
1.000
1.257
0.771
0.832
1.095
32.71
0.0081
0.121
0.101
0.204
0.226
0.061
0.120
0.013
0.035
0.065
0.066
0.031
0.004
0.109
0.139
0.155
0.109
− 1.880
− 1.020
2.360
1.810
− 0.020
0.320
0.420
2.000
− 1.010
0.940
0.890
− 0.110
2.650
− 1.440
− 0.990
0.910
0.060
0.309
0.018*
0.070
0.980
0.747
0.675
0.046*
0.314
0.349
0.375
0.914
0.008**
0.150
0.324
0.362
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
1115
Table 6
Factors predicting substance dependency.
All
Female
Black
CPC preschool participation (ages 3–4)
Child protection services (ages 4–9)
Family conflict (ages 5–10)
CPC school age participation (ages 6–9)
Social maturity (ages 7–9)
Family risk (age 8)
Parent expectations (ages 8–10)
Intrinsic motivation (ages 9–10)
Reading comprehension (age 10)
No trouble-making behavior (ages 9–13)
School quality (ages 10–14)
School mobility (ages 14–18)
Parent substance abuse experience (ages 10–15)
Deviant peer affiliation (age 16)
School dropout (by age 16)
Wald Chi–squared
Prob. N Chi–squared
Males
Haz. ratio
Std. error
z
pNz
0.287
2.235
0.800
1.833
1.682
0.914
0.973
1.018
1.025
1.087
1.001
0.972
0.493
1.447
1.754
1.213
1.518
0.077
1.438
0.129
0.379
0.433
0.178
0.012
0.060
0.129
0.096
0.005
0.052
0.202
0.226
0.587
0.236
0.317
− 4.68
1.25
− 1.39
2.93
2.02
− 0.46
− 2.16
0.30
0.20
0.95
0.21
− 0.52
− 1.73
2.37
1.68
0.99
2.00
0.000***
0.212
0.166
0.003**
0.044*
0.644
0.031*
0.762
0.843
0.343
0.831
0.600
0.084
0.018*
0.093
0.320
0.046
116.200
0.000
Haz. ratio
1.950
0.823
1.600
2.014
0.909
0.967
1.045
1.037
1.040
0.962
1.002
0.582
1.121
1.417
1.107
1.458
Std. error
z
pNz
1.021
0.159
0.337
0.432
0.172
0.020
0.050
0.120
0.100
0.052
0.007
0.216
0.286
0.229
0.192
0.347
1.28
− 1.01
2.23
3.26
− 0.50
− 1.60
0.91
0.31
0.41
− 0.71
0.33
− 1.46
0.45
2.16
0.59
1.58
0.202
0.312
0.025*
0.001**
0.615
0.110
0.364
0.757
0.683
0.479
0.738
0.145
0.655
0.031*
0.557
0.113
50.340
0.000
Note: ***p b 0.001, **p b 0.01, *p b 0.05.
(HR = 1.68, p b .05), school dropout by age 16 (HR = 1.52, p b .05) and
school mobility at ages 14 to 18 (HR = 1.45, p b .05) were associated
with an increased hazard of substance dependency. Each of these
experiences substantially increased by about 50% or more the likelihood of earlier ages of dependency or faster progression from use to
dependency. While involvement in child protective services nearly
doubled the relative risk of substance dependency, 3 school moves
increased the relative risk (compared to no moves) by more than
double. That the majority of the major predictors of dependency occur
prior to age 10 is suggestive of the need for earlier preventive services.
We report the predictors of the dichotomous measure of
substance dependency in Appendix C. This model does not account for
age of dependency. There were some differences compared to the age
hazard measure in Table 6. Besides gender, the significant predictors in
the overall sample were race (Blacks had higher rates), frequent family
conflict at ages 5–10, no trouble-making behavior, and parent substance
abuse experience at child’s ages 10–15. For example, each additional point
on the no trouble-making scale reduced the rate of substance dependency
by 3.6 percentage points. For males, frequent family conflict and race were
most predictive of substance dependency. Appendix D shows that the
predictors of substance dependency using the entire study sample of 1208
were consistent with those of the substance use sample of 440.
4. Discussion
This study investigated a comprehensive set of early childhood and
adolescent predictors of three measures of substance use problems in
early adulthood for a large urban cohort. We found that substance abuse
is primarily determined by individual (trouble-making behavior and
personal substance use experience), family (child protection services,
parent expectations, and parent substance abuse), and school-related
(mobility, deviant peer affiliations) measured up to the middle of
adolescence. Early school dropout linked to a faster progression to
dependency but not to overall prevalence or age at first use. Similar
predictive patterns of substance abuse were found for males. About 90% of
substance abusers were males. Not surprisingly, and consistent with
extant research, substance use history and parent substance use predicted
substance use problems (Young et al., 2002; Merline et al., 2004).
Consistent with previous studies (Hawkins et al., 1992; Osler et al.,
2006; Reinhertz et al., 2000), early family adversity (i.e., child protection
services and family conflict) had a key role in predicting the onset of
substance use and substance dependency. Alternatively, being female
and having greater social maturity decreased the likelihood of
progression to substance dependency. School factors in adolescence
such as school mobility and school dropout also decreased the likelihood
of progression to substance dependency. These findings are unique as
previous research has not investigated factors that predict progression
from substance use to dependency. The predictors of substance use and
abuse, however, have been extensively investigated. Findings have
indicated that early family adversity and disadvantage, early antisocial
behavior, and deviant peer affiliation are especially salient (Fergusson
et al., 2002; Gilvarry & McArdle, 2007; Siebenbruner et al., 2006) even
after substance use history is taken into account.
In this study, we found limited support that early school-age factors
predicted substance use problems. Early school-age social maturity
predicted substance dependency and teacher ratings of parent expectations for children’s success predicted the prevalence of abuse at age 26.
The comprehensiveness of the model–prior to substance use and a host of
other factors–may be responsible. The relatively large influence of home
environment, for example, is likely to reduce or account for the effects of
other variables on substance abuse and dependency (e.g. peer affiliation,
see Appendix B correlation matrix). Frequent family conflict and
involvement in child protective services due to child maltreatment were
predictors of one or more of the substance use outcomes. This is consistent
with the accumulated knowledge (Hawkins et al., 1992; Dube et al., 2003;
Osler et al., 2006; Siebenbruner et al., 2006; Young et al., 2002).
Although associated with adverse early child and family experiences,
school mobility was found to be a significant predictor of each measure
of adult substance use. The impact of the number of moves was also
largest for males—about double the effect size of the overall sample (see
Tables 4 and 6). These findings are consistent with ecological
(Bronfenbrenner, 1989) and social capital (Coleman, 1988) theories of
life course development, in which school transitions, especially if
frequent, can disrupt peer networks, school learning environments, and
relationships with teachers that culminate in higher risk of substance
use problems in adulthood. Much previous research has found that
school mobility links to lower school achievement and school dropout
(Mehana & Reynolds, 2004; Reynolds, Mathieson, & Topitzes, 2009;
Rumberger, 2003). Because we found that early school dropout by age
16 independently predicted substance dependency, school failure may
mediate the relation between mobility and substance abuse. Nevertheless, our study found that many individual, family, and school-related
factors predicted substance use problems in adulthood and together
they suggest approaches for prevention and cost savings.
1116
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
4.1. Contributions to previous research
The prospective investigation of developmental precursors in
early childhood, childhood, and adolescence contributes to the integration of knowledge about the predictors of substance use and
its progression to abuse. Our study contributes to greater understanding of the development of substance use problems for at-risk
students living in impoverished neighborhoods. It also provides relevant knowledge for the development and implementation of interventions to forestall and prevent substance abuse. This is one of the
first studies using prospective longitudinal data to test a comprehensive model of the development of substance abuse, including the
onset of substance use to dependency. Besides early childhood and
family risk factors, individual, family, school, and peer variables were
collected from multiple sources, which generated a full spectrum of
ecological variables into early adulthood. One of the advantages of
the longitudinal design is that the links between early precursors and
later substance abuse problems can be more confidently defined as
predictors rather than as mere correlates. Consequently, the relatively
strong predictive power of involvement in child protective services,
parent expectations, trouble-making behavior, school mobility, and
deviate peer affiliation is unlikely to be affected by other confounding
influences. Indeed, the breadth of the model reduces the risk of model
specification errors. The consistently negative influence on substance
use problems of school mobility beginning at age 10 is a further
example of the strength of prediction from the model.
Besides the investigation of the factors associated with the onset of
substance use and factors predicting later substance abuse problems
in adulthood, this is one of the first studies to examine the factors
predicting the time from substance use to substance dependency. The
findings showed factors related to the persistence of and resistance
from substance use. The major predictors, including family conflict,
social maturity, involvement in child protective services, school
mobility, and early school dropout, can all be altered by program or
policy intervention. School mobility has not been previously identified
as a predictor of substance use problems or dependency (Hawkins
et al., 1992; Young et al., 2002).
More generally, the identification of malleable predictors in
childhood through adolescence in the onset of substance use and
progression to dependency contributes to knowledge about not
only etiology of substance abuse but avenues of prevention and
intervention. At an individual level, engaging children’s pro-social
behavior in classroom settings during the early school-age years
may reduce the likelihood of subsequent substance abuse. At the
family level, reducing economic stresses associated with poverty
and related factors as well as promoting positive parenting
practices that reduce family conflict and the risk of child
maltreatment are likely to promote positive health behaviors. At
the level of the school context, reducing school mobility and its
detrimental effects was found to be particularly salient. This
suggests that unstable learning environments may weaken the
school bond and commitment to education (Hawkins et al., 1992;
Osler et al., 2006) and thus lead to substance use problems.
Moreover, the strong influence of deviant peer affiliations, which
develops as a function of weakened school commitment (Fergusson
et al., 2002; Galea et al., 2003; Hawkins et al., 1992), suggests that
building supportive and positive social networks in adolescence
may be especially beneficial to the prevention of substance use
problems (Monge et al., 1999).
sample was almost exclusively African-American, which may limit
generalizability to other racial and ethnic groups.
The second limitation is that although our study tested a comprehensive set of predictors of substance use problems, it is
possible that other important factors were omitted. For example,
neighborhood characteristics were only indirectly measured
and early home environment experiences were limited to family
demographics, involvement in child protective services and preschool participation. Of course, given the breadth of our model, the
added value of these or other omitted variables would likely be
small.
The third limitation is that some of our explanatory variables were
based on retrospective reports including previous substance use,
parent substance use, and family conflicts early in life. Although such
reports for salient life experiences are often accurate, the threat of
recall bias cannot be ruled out completely. Moreover, other measures,
such as parent expectations for children’s school success, were from
teachers and these reports have lower construct validity than parent
or child reports.
Finally, by emphasizing the predictors of substance use problems,
we did not investigate the mediators and developmental pathways
leading to substance abuse problems and the extent to which these
processes may vary by participant subgroups. Future research should
more fully integrate predictors and mechanisms in the development
of substance abuse.
4.3. Implications for policy and practice
The findings of the study are suggestive of a variety of interventions, policies, and practices at different ages and at multiple levels
that can contribute to the prevention of substance use, progression to
abuse, and related behavioral problems. These include early childhood
interventions (Manning, Homel, & Smith, 2010; Ou & Reynolds, 2006;
Reynolds et al., 2007; Schweinhart et al., 2005) and parenting
programs (Sweet & Appelbaum, 2004; Webster-Stratton & Taylor,
2001), home visiting programs to prevent child maltreatment
(Reynolds, Chen, & Herbers, 2009), social skills training to strengthen
peer relationships and resistence skills (Hawkins, Kosterman, Catalano, Hill & Abbott, 2005; Botvin et al., 2001), school-based programs
to reduce mobility, coordinate services, and provide continuity in
learning environments (Finn-Stevenson & Zigler, 1999; Takanishi &
Kauerz, 2008; Titus, 2007), and youth mentoring (Rhodes & DuBois,
2006). Unfortunately, most programs to prevent substance use and
related problems are implemented too late to be effective. The most
cost-effective programs and those likely to have enduring effects are
usually implemented prior to adolescence and follow established
principles of effectiveness (Nation et al., 2003; Reynolds & Temple,
2008).
More broadly, multi-component programs that are implemented relatively early in the life course and have sufficient duration,
scope, and intensity are most likely to be effective not only for the
prevention of substance abuse but the promotion of well-being in
education, social behavior, health, and mental health. Given the
breadth of predictors of substance abuse found in this study and
others, collaboration across education and human service systems
in the development and implementation of programs and policies
also may provide unique opportunities to improve the health and
well-being of young people in more enduring and sustainable
ways.
4.2. Limitations
Using existing data from the CLS, the sample was representative of
at-risk children in low-income areas in Chicago, who attended early
childhood intervention programs. Findings may not generalize to
youth in non-urban or less disadvantaged contexts. Furthermore, the
Acknowledgment
Funding support for the study was provided by the National
Institute of Child Health and Human Development (R01 HD034294).
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
1117
Appendix A. Descriptive statistics of the model variables before and after imputations
Variable
Before imputation
Child protection services (ages 4–9)
Family conflict (ages 5–10)
Parent substance abuse experience (ages 5–10)
School mobility (ages 6–9)
Social maturity (ages 7–9)
Family risk (age 8)
Parent expectations (ages 8–10)
Intrinsic motivation (ages 9–10)
No trouble- making behavior (ages 9–12)
Reading achievement (age 10)
School mobility (ages 10–13)
School quality (ages 10–14)
Parent substance abuse experience (ages 10–15)
Personal substance use experience (ages 10–15)
School mobility (ages 14–18)
Negative peer affiliation (age 16)
School dropout (by age 16)
Obs
Mean
1140
1142
1142
1152
1142
1142
1072
985
1077
1007
1102
1103
1142
1142
698
1142
1168
0.103
0.058
0.041
0.740
19.324
4.067
3.388
− 0.017
6.062
102.906
0.959
0.130
0.087
0.028
0.385
0.037
0.124
After imputation (n = 1208)
Std. dev.
Mean
0.304
0.233
0.199
0.871
4.829
1.856
0.872
0.870
1.863
15.884
0.976
0.336
0.282
0.165
0.487
0.188
0.330
Std. dev.
0.097
0.061
0.046
0.741
19.311
4.230
3.388
− 0.014
6.064
102.833
0.965
0.129
0.089
0.031
0.393
0.031
0.125
0.296
0.240
0.209
0.851
4.709
1.791
0.872
0.786
1.763
14.664
0.989
0.335
0.284
0.175
0.489
0.175
0.331
Note. Imputations based on the expectation-maximization method. The selection variables included children’s gender, family demographic characteristics, kindergarten school, and
CPC preschool participation. See text for further information.
Appendix B. Correlation matrix (values of 0.06 and above in absolute value are significant at p b 0.05)
1 Female
2 Black
3 CPC preschool participation
4 Child protection services (ages 4–9)
5 Family conflict (ages 5–10)
6 Parent substance abuse experience (ages 5–10)
7 School mobility (ages 6–9)
8 School-age CPC participation (ages 6–9)
9 Social maturity (ages 6–9)
10 Family risk (age 8)
11 Parent expectations (ages 8–10)
12 Intrinsic motivation (ages 9–10)
13 No trouble-making behavior (ages 9–12)
14 Reading comprehension
15 School mobility (ages 10–13)
16 School quality (ages 10–14)
17 Parent substance abuse experience (ages 10–15)
18 Personal substance use experience (ages 10–15)
19 School mobility (ages 14–18)
20 Deviant peer affiliation (ages 16)
21 School dropout (by age16)
22 Age of first substance use
23 Substance abuse
24 Drug dependency
1
2
3
4
5
6
7
8
9
10
11
12
1.00
0.03
0.08
0.00
− 0.04
− 0.03
− 0.05
0.03
0.27
0.01
0.17
0.08
0.15
0.20
− 0.09
0.07
− 0.02
− 0.15
− 0.05
0.01
− 0.11
− 0.10
− 0.47
− 0.21
1.00
0.01
− 0.01
− 0.02
− 0.03
0.04
0.03
− 0.06
0.05
− 0.07
− 0.02
0.03
− 0.06
− 0.02
0.01
0.00
0.01
0.06
− 0.01
− 0.04
0.04
0.01
0.11
1.00
− 0.02
− 0.01
0.02
− 0.05
0.37
0.16
− 0.05
0.14
0.00
0.10
0.18
− 0.17
0.16
0.07
− 0.04
− 0.02
− 0.01
− 0.07
0.04
− 0.09
− 0.03
1.00
0.09
0.12
0.12
− 0.01
− 0.08
0.15
− 0.17
− 0.01
− 0.04
− 0.05
0.109
− 0.06
0.14
0.10
0.02
− 0.02
0.10
− 0.16
0.13
0.06
1.00
0.34
0.00
0.03
0.01
0.02
− 0.03
− 0.02
− 0.04
0.07
0.071
− 0.01
0.14
0.19
0.02
0.06
0.09
− 0.05
0.10
0.15
1.00
0.03
0.00
0.03
0.03
0.02
0.03
− 0.04
0.05
0.049
− 0.01
0.53
0.19
− 0.03
− 0.01
0.09
− 0.08
0.12
0.10
1.00
− 0.27
− 0.13
0.19
− 0.13
− 0.02
− 0.10
− 0.11
0.176
− 0.13
0.06
0.04
0.04
0.02
0.00
− 0.08
0.03
− 0.05
1.00
0.12
− 0.01
0.11
− 0.01
0.08
0.13
− 0.16
0.13
0.00
− 0.02
− 0.05
0.00
− 0.03
0.02
− 0.01
− 0.07
1.00
− 0.20
0.61
0.07
0.19
0.56
− 0.18
0.09
0.05
− 0.06
− 0.11
− 0.02
− 0.12
0.00
− 0.24
− 0.15
1.00
− 0.28
0.03
− 0.13
− 0.28
0.17
− 0.18
0.00
0.06
0.08
0.00
0.07
− 0.14
0.07
0.04
1.00
0.07
0.08
0.43
− 0.21
0.14
0.02
− 0.03
− 0.12
− 0.03
− 0.14
0.03
− 0.22
− 0.11
1.00
0.12
0.00
− 0.03
0.02
0.03
− 0.04
− 0.01
− 0.02
0.00
− 0.04
− 0.09
− 0.02
1118
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
Appendix B (continued)
1 Female
2 Black
3 CPC preschool participation
4 Child protection services (ages 4–9)
5 Family conflict (ages 5–10)
6 Parent substance abuse experience (ages 5–10)
7 School mobility (ages 6–9)
8 School-age CPC participation (ages 6–9)
9 Social maturity (ages 6–9)
10 Family risk (age 8)
11 Parent expectations (ages 8–10)
12 Intrinsic motivation (ages 9–10)
13 No trouble-making behavior (ages 9–12)
14 Reading comprehension
15 School mobility (ages 10–13)
16 School quality (ages 10–14)
17 Parent substance abuse experience (ages 10–15)
18 Personal substance use experience (ages 10–15)
19 School mobility (ages 14–18)
20 Deviant peer affiliation (ages 16)
21 School dropout (by age16)
22 Age of first substance use
23 Substance abuse
24 Drug dependency
13
14
15
16
17
18
19
20
21
22
23
24
1.00
0.16
− 0.13
0.10
0.00
− 0.05
− 0.05
− 0.03
− 0.04
0.00
− 0.17
− 0.14
1.00
− 0.19
0.24
0.09
− 0.05
− 0.14
0.03
− 0.07
0.02
− 0.19
− 0.09
1.00
− 0.17
0.02
0.06
0.09
0.03
0.15
0.03
0.22
0.11
1.00
− 0.02
− 0.06
− 0.08
0.00
− 0.05
0.09
−0.10
− 0.12
1.00
0.23
0.00
− 0.01
0.06
− 0.04
0.10
0.10
1.00
0.03
0.03
0.05
− 0.16
0.25
0.34
1.00
0.04
− 0.09
− 0.10
0.10
0.08
1.00
0.04
− 0.05
0.08
− 0.03
1.00
− 0.11
0.13
0.03
1.00
0.21
− 0.04
1.00
0.39
1.00
Appendix C. Factors predicting substance dependency (dichotomous specification, n = 440)
All
Female
Black
CPC preschool participation (ages 3–4)
Child protection services (ages 4–9)
Family conflict (ages 5–10)
CPC school-age participation (ages 6–9)
Social maturity (ages 7–9)
Family risk (age 8)
Parent expectations (ages 8–10)
Intrinsic motivation (ages 9–10)
Reading achievement (age 10)
No trouble-making behavior (ages 9–12)
School quality (ages 10–14)
School mobility (ages 14–18)
Parent substance abuse experience (ages 10–15)
Deviant peer affiliation (age 16)
School dropout (by age 16)
Wald Chi-squared
Prob. N Chi-squared
Pseudo R-squared
Male
Marginal effect
Std. Err.
z
PNz
− 0.256
0.267
0.003
0.047
0.263
− 0.063
− 0.009
− 0.007
0.006
0.019
0.001
− 0.036
− 0.164
0.050
0.165
− 0.016
0.008
319.97
0.00
0.12
0.045
0.075
0.055
0.070
0.083
0.063
0.006
0.014
0.035
0.033
0.001
0.016
0.094
0.047
0.072
0.049
0.065
− 5.130
2.540
0.050
0.670
3.090
− 1.010
− 1.540
− 0.490
0.170
0.570
0.800
− 2.190
− 1.520
1.070
2.330
− 0.330
0.120
0.000***
0.011*
0.963
0.502
0.002**
0.312
0.124
0.621
0.868
0.571
0.425
0.029*
0.129
0.282
0.020*
0.739
0.903
Marginal effect
Std. Err.
z
PNz
0.292
− 0.013
0.059
0.249
− 0.081
− 0.013
− 0.004
0.032
0.011
0.002
− 0.038
− 0.152
0.055
0.123
− 0.012
0.002
250.06
0.00
0.08
0.098
0.060
0.105
0.108
0.075
0.008
0.016
0.048
0.044
0.002
0.020
0.117
0.054
0.067
0.066
0.064
2.290
− 0.220
0.560
2.190
− 1.080
− 1.550
− 0.270
0.660
0.240
1.130
− 1.930
− 1.210
1.020
1.840
− 0.180
0.020
0.022*
0.829
0.573
0.028*
0.279
0.120
0.790
0.509
0.810
0.257
0.054
0.227
0.310
0.065
0.854
0.980
Note: ***p b 0.001, **p b 0.01, *p b 0.05. Probit coefficients are transformed to marginal effects (dy/dx), which are the change in substance dependency (in percentage points) per 1unit change in the explanatory variables.
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
1119
Appendix D. Factors predicting substance dependency (dichotomous specification, n = 1208)
All
Marginal effect
Female
Black
CPC preschool participation (ages 3–4)
Child protection services (ages 4–9)
Family conflict (ages 5–10)
CPC school-age participation (ages 6–9)
Social maturity (ages 7–9)
Family risk (age 8)
Parent expectations (ages 8–10)
Intrinsic motivation (ages 9–10)
Reading achievement (age 10)
No trouble-making behavior (ages 9–12)
School quality (ages 10–14)
School mobility (ages 14–18)
Parent substance abuse experience (ages 10–15)
Deviant peer affiliation (age 16)
School dropout (by age 16)
Wald Chi-squared
Prob. N Chi-squared
Pseudo R-squared
− 0.176
0.068
0.005
0.061
0.127
− 0.013
− 0.003
0.000
− 0.012
− 0.003
0.000
− 0.017
− 0.056
0.027
0.075
0.023
0.017
1836.46
0.00
0.24
Male
Std. err.
z
PNz
0.023
0.015
0.014
0.030
0.046
0.016
0.002
0.004
0.010
0.008
0.000
0.005
0.020
0.014
0.029
0.015
0.024
− 10.82
2.62
0.36
2.41
3.44
− 0.81
− 1.44
− 0.11
− 1.1
− 0.39
− 0.22
− 3.49
− 1.89
1.95
3.22
1.61
0.74
0.000***
0.009**
0.722
0.016*
0.001**
0.416
0.149
0.915
0.270
0.696
0.828
0.000***
0.058
0.051
0.001**
0.108
0.459
Marginal effect
0.169
0.006
0.109
0.213
− 0.042
− 0.008
0.001
− 0.016
− 0.019
0.000
− 0.034
− 0.119
0.074
0.093
0.047
0.018
655.15
0.00
0.11
Std. err.
z
PNz
0.051
0.035
0.070
0.086
0.042
0.006
0.010
0.031
0.022
0.001
0.012
0.063
0.035
0.050
0.041
0.045
2.32
0.16
1.63
2.64
− 1.00
− 1.44
0.10
− 0.51
− 0.86
0.39
− 2.99
− 1.54
2.06
2.07
1.15
0.39
0.020*
0.869
0.102
0.008**
0.318
0.151
0.920
0.613
0.387
0.695
0.003**
0.123
0.040*
0.039*
0.248
0.693
Note: ***p b 0.001, **p b 0.01, *p b 0.05. Probit coefficients are transformed to quantify marginal effects of the selected independent variables on the probability of outcomes.
References
American Psychiatric Association (2000). Diagnostic and statistical manual of mental
disorders: DSM-IV-TR. Washington, DC: Author.
Belcher, H. M. E., & Shinitzky, H. E. (1998). Substance abuse in children. Archives of
Pediatrics & Adolescent Medicine, 152, 952−960.
Bennett, E. M., & Kemper, K. J. (1994). Is abuse during childhood a risk factor for
developing substance abuse problems as an adult? Developmental and Behavioral
Pediatrics, 15, 426−429.
Block, J., Block, J., & Keyes, S. (1988). Longitudinal drug use in adolescence: Early childhood
personality and environmental precursors. Child Development, 59, 336−355.
Botvin, G. J., Griffin, K. W., Diaz, T., & Ifill-Williams, M. (2001). Drug abuse prevention
among minority adolescents: Posttest and one-year follow-up of school-based
preventive intervention. Prevention Science, 2, 1−13.
Bronfenbrenner, U. (1989). Ecological systems theory. Annals of Child Development, 6,
187−249.
Brook, J. S., Brook, D. W., de la Rosa, M., Duque, L. F., Rodriguez, E., Montoya, I. D., et al. (1998).
Pathways to marijuana use among adolescents: Cultural/ecological, family, peer, and
personality influences. Journal of the American Academy of Child and Adolescent Psychiatry,
37, 759−766.
Campbell, F. A., Ramey, C. T., Pungello, E., Sparling, J., & Miller-Johnson, S. (2002). Early
childhood education: Young adult outcomes from the Abecedarian Project. Applied
Developmental Science, 6, 42−57.
Chicago Longitudinal Study (2005). Chicago longitudinal study: User’s guide (Version 7).
Madison, WI: University of Wisconsin.
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of
Sociology, 94(Supplement), S95−S120.
Cooper, H. L. F., Friedman, S. R., Tempalski, B. T., & Friedman, R. (2007). Residential
segregation and injection drug use prevalence among Black adults in U.S. Metropolitan
areas. American Journal of Public Health, 97, 344−352.
Cox, D. R. (1972). Regression models and life-tables (with discussion). Journal of the
Royal Statistical Society, Series B34, 187−220.
Dube, S. R., Felitti, V. J., Dong, M., Chapman, D. P., Giles, W. H., & Anda, R. F. (2003). Childhood
abuse, neglect and household dysfunction and the risk of illicit drug use: The Adverse
Childhood Experience Study. Pediatrics, 111, 564−572.
Fergusson, D. M., Lynskey, M. T., & Horwood, L. (1993). Conduct problems and attention
deficit in middle childhood and cannabis use by age 15. The Australian and New Zealand
Journal of Psychiatry, 27, 673−682.
Fergusson, D., Swain-Campbell, N. R., & Horwood, L. J. (2002). Deviant peer affiliations,
crime and substance use: A fixed effects regression analysis. Journal of Abnormal
Child Psychology, 30, 419−430.
Finn-Stevenson, M., & Zigler, E. (1999). Schools of the 21st century: Linking child care and
education. Boulder, CO: Westview Press.
Galea, S., Ahern, J., & Vlahov, D. (2003). Contextual determinants of drug use risk behavior:
A theoretic framework. Journal of Urban Health, 80, 50−58.
Gilvarry, E., & McArdle, P. (2007). Determinants of substance misuse. Developmental
Medicine and Child Neurology, 49, 636−640.
Gutierrez, R. G. (2002). Parametric frailty and shared frailty survival models. Stata Journal,
2, 22−44.
Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective factors for alcohol and
other drug problems in adolescence and early adulthood: Implications for substance
abuse prevention. Psychological Bulletin, 112, 64−105.
Hawkins, J. D., Kosterman, R., Catalano, R. F., Hill, K. G., & Abbott, R. D. (2005). Promoting
positive adult functioning through social development intervention in childhood:
Long-term effects from the Seattle Social Development Program. Archives of Pediatric &
Adolescent Medicine, 159, 25−31.
Hieronymus, A. N., & Hoover, H. D. (1990). Iowa tests of basic skills: A manual for school
administrators. University of Iowa. Chicago: Riverside Publishing Company.
Johnston, L. D., O’Malley, P. M., & Bachman, J. D. (1999). National survey results on drug
use from the monitoring of the future study 1975–1998. The University of Michigan
Institute of Social Research: US Department of Health and Human Services, National
Institutes of Health.
Kandel, D. B., Simcha-Fagan, O., & Davies, M. (1986). Risk factors for delinquency and illicit
drug use from adolescence to young adulthood. Journal of Drug Issues, 16, 67−90.
Manning, M., Homel, R., & Smith, C. (2010). A meta-analysis of the effects of early
developmental prevention programs in at-risk populations on non-health outcomes
in adolescence. Children and Youth Services Review, 32, 506−519.
Mehana, M., & Reynolds, A. J. (2004). School mobility and achievement: A metaanalysis. Children and Youth Services Review, 26, 93−119.
Merline, A. C., O’Malley, P. M., Schulenberg, J. E., Bachman, J. G., & Johnston, L. D. (2004).
Substance use among adults 35 years of age: prevalence, adulthood predictors, and
impact of adolescent substance use. American Journal of Public Health, 94, 96−102.
Monge, P., Fulk, J., Parnassa, C., Flanagan, A., Rumsey, S., & Kalman, M. (1999). Cooperative
Interagency approaches to the illegal drug problem. International Journal of Police
Science and Management, 2, 229−241.
Nation, M., et al. (2003). What works in prevention: Principles of effective prevention
programs. The American Psychologist, 58, 449−456.
Osler, M., Nordentoft, M., & Andersen, A. M. (2006). Childhood social environment and
risk of drug and alcohol abuse in a cohort of Danish men born in 1953. American
Journal of Epidemiology, 163, 654−661.
Ou, S., & Reynolds, A. J. (2006). Early childhood intervention and educational attainment:
Age 22 findings from the Chicago Longitudinal Study. Journal of Education for Students
Placed at Risk, 11, 175−198.
Ou, S., & Reynolds, A. J. (2008). Predictors of educational attainment in the Chicago
Longitudinal Study. School Psychology Quarterly, 23, 199−229.
Reinhertz, H. Z., Giaconia, R. M., Hauf, A. M. C., Wasserman, M. S., & Paradis, A. D. (2000).
General and specific childhood risk factors for depression and drug disorders by
early adulthood. Journal of American Academy of Child and Adolescent, 39, 223−231.
Reynolds, A. J., Chen, C. -C., & Herbers, J. E. (2009, June). School mobility and educational
success: A research synthesis and evidence on prevention. Paper presented at the
workshop on the impact of mobility and change on the lives of young children, schools,
and neighborhoods. Washington, D.C.: National Research Council.
Reynolds, A. J., Mathieson, L. C., & Topitzes, J. W. (2009). Can early childhood intervention
prevent child maltreatment? A review of research. Child Maltreatment, 14, 182−206.
Reynolds, A. J., & Temple, J. A. (2008). Cost-effective early childhood development programs
from preschool to third grade. Annual Review of Clinical Psychology, 4, 109−139.
Reynolds, A. J., Temple, J. A., Ou, S., Robertson, D. L., Mersky, J. P., Topitzes, J. W., et al.
(2007). Effects of a school-based, early childhood intervention on adult health and
well being: A 19-year follow up of low-income families. Archives of Pediatric &
Adolescent Medicine, 161(8), 730−739.
Reynolds, A. J., Temple, J. A., Robertson, D. L., & Mann, E. A. (2001). Long-term effects of
an early childhood intervention on educational achievement and juvenile arrest — A
15-year follow-up of low-income children in public schools. Journal of the American
Medical Association, 285, 2339−2346.
1120
I. Arteaga et al. / Children and Youth Services Review 32 (2010) 1108–1120
Rhodes, J., & DuBois, D. (2006). Understanding and facilitating the youth mentoring
movement. Social policy report, XX, No. III. Ann Arbor, MI: Society for Research in
Child Development Available from http://www.srcd.org/spr.html.
Robin, S. S., & Johnson, E. O. (1996). Attitude and peer cross pressure: Adolescent drug
and alcohol use. Journal of Drug Education, 26, 69−99.
Rumberger, R. W. (2003). The causes and consequences of student mobility. Journal of
Negro Education, 72, 6−21.
Schafer, J. L. (1997). Analysis of incomplete multivariate data. Monographs on statistics
and applied probability, 72,London: Chapman & Hall.
Schweinhart, L. J., Barnes, H. V., & Weikart, D. P. (1993). Significant benefits: The high/scope
Perry preschool study through age 27. Monographs for the high/scope educational
research foundation, No. 10,Ypsilanti, MI: High/Scope Educational Research Foundation.
Schweinhart, L. J., Montie, J., Xiang, Z., Barnett, W. S., Belfield, C. R., & Nores, M. (2005).
Lifetime effects: The high/scope Perry preschool study through age 40. Monographs
for the high/scope educational research foundation, No. 14,Ypsilanti, MI: High/Scope
Educational Research Foundation.
Siebenbruner, J., Englund, M. M., Egeland, B., & Hudson, K. (2006). Developmental
antecedents of late adolescence substance use patterns. Development and Psychopathology, 18, 551−571.
Substance Abuse and Mental Health Services Administration (2004). Results from the
2003 national survey on drug use and health: National findings (NHSDA Series H25, DHHS Publication No. SMA 04-3964). Rockville, MD: Author.
Sweet, M. A., & Appelbaum, M. I. (2004). Is home visiting an effective strategy? A metaanalytic review of home visiting programs for families with young children. Child
Development, 75, 1435−1456.
Takanishi, R., & Kauerz, K. (2008, March). PK inclusion: Getting serious about a P-16
education system. Phi Delta Kappan, 89(7), 480−487.
Titus, D. (2007). Strategies and resources for enhancing the achievement of mobile
students. NASSP Bulletin, 91(1), 81−97.
Topitzes, J. W., Godes, O., Mersky, J. P., Tiwari, G., Ceglarek, S., & Reynolds, A. J. (2009).
Educational success and adult health: Findings from the Chicago Longitudinal Study.
Prevention Science, 10(2), 175−195.
Webster-Stratton, C., & Taylor, T. (2001). Nipping early risk factors in the bud: Preventing
substance abuse, delinquency, and violence in adolescence through interventions
targeted at young children (0 to 8 years). Prevention Science, 2, 165−192.
Young, S. E., Corley, R. P., Stallings, M. C., Rhee, S. H., Crowley, T. J., & Hewitt, K. J. (2002).
Substance use, abuse, and dependence in adolescence: Prevalence, symptom profiles,
and correlates. Drug and Alcohol Dependence, 68, 309−322.

Purchase answer to see full
attachment