The Impact of Family Factors on Adolescent Drug Use

1tj ~ ~ 1f '"

*

m

21W3 •

m

19-50J{ •~~931f. 6~ Journal of Health Education

No.21, pp.19-50. June 2004

-*

.f'-<.

~~~.%k~~~~~.~~A~~'~~~32.~~~~'~H.t~#. ~~~~~rn~~~~~zM~~~o~~M*~oT:

1.)1:.(~~.~t~f\ (family structure) ~

1: '

~:1fT ~IJ

ag

~~~1f.1--~~~~m ~~4"-J!.

:1fM.~~~~rno~:~~k~~~~.~ft(*:~.~A~£

•••

ft~)

'£

.~~~£.~~.~~~~-~'~~:1f~~~rn~(~o:£.~~~~~),~~

~.~o~:1f~~~~k~~~m~~.~~~~~J!.:1f~~~*~,~#~~~h'

~ ~:1f

+

ag

~*

1Ji

(effect size) 0

2.)1:.(~~~

1s.

(family relations) ~ i, , ~Ajt ~ag

k

~i. ~ftt. (democratic family) ~.~~~~:1f.~~m~~~~~,~*~~~-.AaA~i.~ftt..~~.~~. m~4h~~~; ~B1fJ~' *Tl~~~ (authoritarian family) ~~4h~m.1--~zrQ'~J!.'J! • ~.if.~B

Ml

0 f.~ ~ ~t~ ~ftt. (permission family) i.1"1i~

If-

~

•

1--

~J,t

m

~

4h

~

#t

4"-

0

~Z'~~:1f~ftt.~~~~~~'~~~~~~~h~k'~~'~~~~~.~

~~~~Whk*~~~.~~~m~~~~~z-o

•

*#i*'¥fHfliJf~pJT.!VJi.!I!~~

jMYt{'F~ :*{~Hl

-1U1.it1f~:4a.

.

~ 21 J/Jj . t\.~ 93

-+

6 J)

introduction

Adolescent drug use peaked in the late 1970s, declined somewhat, and then steadily increased during the 1980s and on into the early 1990s (McCoy, Metsch, & Inciardi, 1996). The rates of drug consumption in adolescents today are lower than they were in the early 1970s, but illicit drug use is still more dominant than it was before the late 1970s.

There is a substantial body of research on the relationship between adolescent drug use and family factors. Among these studies, many focus on the association of parental drug use, sibling drug use, single parenting, parents' education, and family income with adolescent substance use (Beardslee, Son, & Vaillant, 1986; Jurich, Polson, Jurich, & Bates, 1985), and several focus on the correlates of parental or family style on adolescent substance use (Brook, Whiteman, Gordon,& Cohen, 1986; Block, Block,& Keyes, 1988).

Emphasizing the correlates of family factors on drug use neither implies that the study of other factors(e.g., school, peer)are not important nor that the role of the family in adolescent drug use is anything but an enormously complex cluster of influences. In general, the family is broadly defined here to include extended families as well as any other patterns whose members identify themselves as a family. Family environment is broadly defined to include(a)family structure and(b) relations. Family structure indicates the physical, concrete, and non-dynamic descriptions of the family, such as family drug use, single parenthood, and socioeconomic status. Family relations are dynamic and abstract, such as family communication, parental style, and affectivity.

Adolescent Drug Use and Family Structure

Family structure consists of (a) family drug use (b) single parent vs intact family, (c) departing of family members, (d) socioeconomic status. Research exploring the correlates of family<structure on adolescent drug use has a long and substantial history. Both Binion (1979) and Schultz and Wilson (1973) have failed to find any relationship between adolescent 'substance use and the number of family members. However, Johnson, Shontz, and Locke (1984) reported a statistically significant association between parental alcohol use and adolescent marijuana and alcohol use. In addition, Tec (1974) reported a positive association

between parental drug use and drug use by children. In contrast, Needle, McCubbin, Wilson, Reineck, Lazar, and Mederer (1986) and Tolone and Dermott (1975) reported that parental drug use is not related to child drug use. Research on the influence of single parenting on adolescent drug use has shown that intact two-parent families have a significantly lower percentage of child substance use than single-parent or non-intact families (Schaffer, 1988). Furthermore, Jurich P, Polson, Jurich A, and Bates (1985) found that most adolescents who use drugs are from divorced families or from families with a high frequency of parental absence or family break-ups. In contrast, Hoffmann (1991) found that one parent's death does not affect the initial use of marijuana in white adolescents.

In sum, the previous research findings on the relationship between adolescent drug use and family structure mayor may not be significant. These contradictory findings imply that advanced study on this topic is recommended.

Adolescent Drug Use and Family Relations

Relatively few researchers have studied the correlates of family relations (e.g., parenting or family style including authoritarian family, democratic family, and permissive family) on adolescent substance use (Rees& Wilborn, 1983). For example, in a study on the correlates of parenting styles upon adolescents, Stanton (1979) associated more laissez-faire parent-adolescent relations with high marijuana use, while moderate marijuana users reported more parental negative control through punishment and autocratic relations with strict rules, and low marijuana users were associated with the democratic families in which their parents made well-established rules for them. Conversely, other investigators have found that the father's negative control was not associated with the use of tobacco, marijuana, and alcohol in male adolescents (Hundleby, 1987). Jurich et aI., (1985) suggested that there is a communication gap between family members of adolescents who are chemically dependent. This is supported by findings that teenagers who abuse drugs typically describe their communication with parents as closed and unclear (Gantman, 1978; Rees & Wilborn, 1983) and have rigid patterns of communication (Streit, Stanton, & Todd, 1982). Parents have also reported communication problems. For example, a number of parents reported that they are inadequate in communicating trust, acceptance, and understanding (Rees& Wilborn,

21-1983). However, Koplin (1987) found that adolescents' use of alcohol, marijuana, and amphetamine was not related to parent-child communication.

In sum, there is no doubt that the relationship between adolescent drug use and family relations exists, but the statistical significance of its relationship is questionable. Just like the contradictory findings on family structure, the research findings on the relationship between adolescent drug use and family relations are inconsistent.

Limitations of Previous Primary Studies

These empirical studies on the relationship between adolescent drug use and family factors are limited by three related problems: (a) the lack of congruent findings among these studies, (b) the need for examining the overall correlates of family factors on adolescent drug use, and (c) the need to re-identify important variables within the family factors which may influence adolescent drug use present in these studies. Hence, there is a need for a method of literature synthesis to help to make sense of these discrepant findings. In general, there are two methods for reviewing literature. These include the narrative approach and the quantitative approach.

The narrative approach involves counting and comparing the number of previous statistically significant and nonsignificant findings in order to indicate whether the results of the research findings are statistically significant or not. For example, Denton and Kampfe (1994) conducted a narrative literature review using the vote-counting method to integrate the contradictory empirical findings related to the correlates of family variables on adolescent substance use. Unfortunately, this review suffered from three problems. One is that the authors selected only a few primary studies to integrate into their review. A second problem is that they attempted to review and integrate studies that were limited to literary, narrative approaches. These have been criticized as being inconsistent and subjective (Glass, 1977). A third problem is that it was not possible for them to examine the magnitude (strength) of the relationship between adolescent drug use and family variables by using vote-counting method. A quantitative method of synthesizing research can not only help to aggregate and analyze the conflicting evidence reported from the previous primary studies concerning adolescent drug use and family variables, but it can also attempt to address the

(2) Specific components of family factors will be associated with adolescent drug use. (2a) An ineffective family structure will favor adolescent drug use.

(2b) Families with dysfunctional relations will lead to increased adolescent drug use. (3) The relationship between family factors and adolescent drug use will be moderated by background characteristics, such as race and gender, and characteristics of the primary study, such as the type of data collection method, the type of publication data, the type of drug(s), and the source of data.

Operational Definitions

The following outline illustrates the operational definitions of family variables and their relatives.

I.Characteristics of Primary Study 1. Types of data collection method

survey only (paper and pencil, telephone) observation or interview included 2. Types of drugs

not specified (not reported) illicit and licit drugs licit drugs

illicit drugs 3. Source of data

journal

thesis or dissertation II. Background Characteristics

1. Gender 2. Race

III. Explanatory Variables Family Structure l.Family drug use

parents (mother, father) siblings (younger, elder) 2. Single parent

divorce remarriage

3. Departing of family members mother or father dead

mother or father not living home with adolescent 4. Socioeconomic status

parents' education parents'/family income Family Relations 1.Parenting styles

authoritarian family: parents control through guilt, punishment, harsh (strict) discipline/rules or authority (dictation)

democratic family: parents reason, well-established rules and discipline, encouraging to express and discuss openly

permissive family: parents not making well-established rules, not making sure that their children follow these rules, less convention

2. Affectivity

adolescent feelings of closeness to parents or family in positive or negative way

positive: the feeling of love, warmth, affection, or support negative: the feeling detached, unsatisfied, disappointed, or angry

METHOD

The purpose of this study is to use the meta-analytic approach to examine the correlates of family factors with adolescent drug use/abuse. The method used in this study is combining and comparing significance levels and effect sizes as described by Rosenthal and Rubin (1986). This methodology consists of (a) trial identification, (b) trial selection, (c) data abstraction, and (d) statistical analysis (Petitti, 1994).

Trial Identification

As many studies as possible are retrieved for this meta-analysis. Since there is no way of deciding whether a group of located studies is representative of the entire population of existing studies on the topic, the best way is to locate as many studies as possible (Jackson, 1980). However, there is no absolute standard for how many of the primary studies are sufficient and appropriate for a meta-analysis (Dulak, 1995; White, 1994).

A comprehensive and thorough search for published studies examining the relationship

25-between family factors and adolescent drug use/abuse is conducted. Computerized searches include Educational Resources Information (ERIC), Psychological Abstracts and PsychINFO (PsychLit), Dissertation Abstracts International (DAI), Sociological Abstracts (SociFile), and LEARN (IBIS). In addition, the bibliographies of review articles, chapters, books, and journal articles located through computer searches are examined for studies which may meet

inclusion criteria.

Trial Selection

The sample of primary studies is collected based on the selection criteria:

I.AII primary studies must have been conducted in the United States and published in the English language or a thesis or dissertation published in the English language. 2.AII studies must include adolescents between the ages of II to 17(or high school

students).

3.The studies may include either licit or illicit drugs.

4.Studies must quantitatively assess outcomes on the correlates of family factors with adolescent drug use or must include quantitative data necessary for calculating effect sizes and significance levels.

5.The search for the studies includes the years from 1960 to 1997. .

Data Abstraction

There are many study characteristics which may influence the results of primary empirical studies. In addition to the five selection criteria previously listed, one fundamental task of collecting data from primary studies is to ensure that the data of the primary studies are coded correctly for computing significance levels and effect sizes. In order to make full use of statistical methods in describing and communicating study findings and accounting for their variance, the characteristics of the primary study and the explanatory variables of family structures and family relations would be coded such that studies with similar characteristics would be consistently coded. All primary studies would be carefully reviewed to develop an appropriate coding sheet.

File Drawer Problem

To deal with this problem, Rosenthal (1979) demonstrated a statistical method to directly assess the threat posed by sampling bias in the literature search. This fail-safe number estimates the number of new, filed, or unretrieved primary

studies averaging null results that would be required to bring the overall combined probability level to the "just" significant convention ofp

=

0.05.

Moderator Effects

The third research hypothesis tests whether selected characteristics of the primary studies are moderators of the overall effect. These assessments would be carried out to test if certain characteristics of the primary studies explain the variation of family factors effects presented in the primary studies. The difference between average effect sizes across primary studies is calculated.

This type of meta-analytic procedure is identified as a focused comparison of effect sizes (Mullen, 1989).

The effects of the background characteristics, such as race and gender, and the characteristics of the primary studies, such as data collection methods, publication date, types of drugs, sample size, and source of data, are estimated to see whether these effects moderate the relationship between adolescent drug use and family factors.

RESULT

The initial search for the studies included the years from 1960 to 1997. Thirty-two studies were conducted and identified on the basis of the study selection criteria from 1970 to 1994. These 32 valid studies were regarded as the study sample and include 30 journal articles, one doctoral dissertation, and one master's thesis, and produced 181 hypothesis tests.

Intercoder Reliability

{tj-1.

ott

lj~:fl . ~ 21 ~ . ~ ~93

of

6 }] Table! (continued) Condition of Family R2 95o/oCI %2 k Z Zfisher Dcohen Nfs Environment _{Qb Qw} Race 69.47· white 15 11.716· 0.2051 0.0409 0.4132 -0.321/+1.125 1022 104.97· black 13 4.949· 0.1443 0.0205 0.2895 -0.492/+ 1.053 184 42.80· mixed 153 -8.816· 0.0118 0.0001 0.0237 -0.200/+0.248 14318 1807.25· Gender 10.93· male 9 7.772· 0.2539 0.0618 0.5133 -0.449/+1.427 191 13.15·

mixed (male&female) 172 -8.224· 0.0151 0.0002 0.0302 -0.181/+0.241 23055 2000.41·

Note: .p < .05; k

=

number of hypothesis test; Zfisher

=

mean of weighted effect size (weighted by sample size); R2

= the square of correlation

coefficient between adolescent drug use and family environment; Dcohen = mean Cohen's d (effect size); 95%CI = 95% confidence interval for Dcohen; Nfs = fail-safe number; X2 = chi-square = homogeneity statistic test (Qw

=

within-group, Qb

=

between-group).

The fail-safe number of effect sizes for the overall family environment that would be needed to be not statistically significant before one would conclude that the overall result of significance was due to sampling bias in the selected studies is Nfs

=

27,562. The significance levels of the hypothesis tests are significantly heterogeneous and may be thought of as having been sampled from different populations. According to Shadish and Haddock's (1994) caution, the results of the heterogeneity of the effect estimates indicates that the mean of the weighted effect size (Zfisher) should not be explained as an estimate of a single population effect parameter, but rather as a description of the mean of the 181 observed effect sizes.

When the characteristics of the primary study, such as data collection methods, publication date, types of drugs, race, gender, source of data, and sample size, are taken into account, their significance levels are all statistically significant. However, according to Cohen's (1981) estimate of small, medium, and large effect size, each of their effect sizes is small except publication date from 1993 to 1994

(Z

fisher

=

-0.2768,

D

cohen

=-0.5608) and

male gender

(Z

fisher

=

0.2539,

D

cohen

=

0.5133), which have medium effect sizes.

Research Hypothesis Two

In Table 2 of family structure, with k

=

120 hypothesis tests, the combination of significance levels of Z for the family structure was statistically significant(Z

=

-10.839,P

< 0.01). The mean ofthe weighted effect size(Zfisher)was 0.0193 corresponding to an

R

2

=

0.0003 and

D

cohen

=

0.0386, indicating less than a small effect size. Research Question (2a) was confirmed by the expected direction of the effect size (0.0193) indicating that the ineffective family structure has a positive relationship to adolescent drug use. All of the variables of family structure were predictive of adolescent drug use, such as socioeconomic status(2=6.963,e <0.05), departing of family members(2=-13.608,e <0.05), family drug use (2

=

20.860,

e

< 0.05), and single-parent household (2

=

2.034,

e

< 0.05). However, an unexpected direction of effect sizes was found in socioeconomic status

(2fisher= -0.0393) and its subvariables of family income(Zfisher = -0.0401) and parents' education(2fisher

=

-0.0391). That is, adolescents with parents with a high socioeconomic status including parents' high education or high family income have a statistically significantly higher risk to become drug users than adolescents who have parents with a low socioeconomic status. However, departing of family members, family drug use including parental drug use and sibling drug use, and single parent all had effect sizes in the expected directions (0.0630 for departing of family members; 0.1042 for parental drug use; 0.2583 for sibling drug use; 0.0594 for single parent). That is, a parent's death or one or both parents not living at home with the adolescent, parental drug use, sibling drug use, and single-parent household were all statistically significantly related to the adolescent drug use. Only the effect size of sibling drug use ( 2 fisher_- = 0.2583,

_-

D

cohen = 0.5224) was of medium magnitude; the rest of parents' education ( 2 fisher

=

-0.0391,

D

cohen

=

-0.0783), family

income (2fisher

=

-0.0401,Dcohen

=

-0.0802), departing of family members (2fisher

=

0.0630,

D

cohen = 0.1262), parental drug use (2fisher = 0.1042,

D

cohen = 0.2088), and single parent( 2 fisher

=

0.0594,

D

cohen

=

0.1189) were small effects.

Table2 The Effect Size of Family Variables in Family Structure

Family Variables REF DE k Z Zfisher

R

2 Dcohen 95%CI Nfs

Qb Qw Family Structure Socioeconomic status -0.063to fiunily income 0.484 parent(s)' -0.129to + 120 -10.839* 0.0193 0.0003 0.0386 -0.215/+0.292 7927 - 42 6.963* -0.0393 0.0015 -0.0785 -0.506/+0.350 15 6 -4.782* -0.0401 0.0016 -0.0802 -1.206/+1.058 14 - 36 -7.491* -0.0391 0.0015 -0.0783 -0.540/+0.385 94

-

3 1

-15.38* 1553.52* 29.35* 304.66*

education 0.459

Departing of 0.017to family members 0.134

Family drug use parent(s) drug 0.000to use 0.499 sibling drug 0.161to use 0.412 Single parent 0.006to 0.189 + 11 -13.608* 0.0630 0.0039 0.1262 -0.715/+0.958 280 + 60 20.860* 0.1128 0.0126 0.2261 -0.134/+0.584 14235 + 52 19.545* 0.1042 0.0107 0.2088 -0.178/+0.593 9281 + 8 12.059* 0.2583 0.0638 0.5224 -0.5011+1.489 518 + 7 2.034* 0.0594 0.0035 0.1189 -0.937/+1.160 3 19.30* 252.73* 335.42* 10.11 43.94*

Note: •

p

< .05; REF = range of the effect size (Zfisher); DE = direction of effect; "+" indicates expected-direction of effect; "-" indicates unexpected direction of effect; unexpected direction of effect indicates that a directional pattern of the relationship between adolescent drug use and family variables in a hypothesis test disconfrrms the directional pattern of such a relationship in a research hypothesis; expected direction of effect indicates that the directional pattern in a hypothesis test confirms the directional pattern in a research hypothesis; k=number of hypothesis test; Zfisher=mean of weighted effect size (weighted by sample size);

R

2= the square of correlation coefficient between adolescent drug use and family environment; Dcohen

=

mean Cohen's d (effect size); 95%CI

=

95% confident interval for Dcohen; Nfs = fail-safe number;X2 = chi-square = homogeneity statistic test (Qw = within-group, Qb=between-group).

According to the standard of 5k+ 10 suggested by Rosenthal (1984; Rosenthal & Hall, 1981), the fail-safe numbers of family income, parents' education, and single parent all did not exceed this benchmark and can therefore be attributed to sampling bias.

For family relations in Table 3, with k = 61 hypothesis tests, the combination of significance levels of Z were statistically significant(Z= 14.749,p < 0.01). The mean of weighted effect size (Zfisher) was -0.0081, which corresponds to

R

2 = 0.00006 and

D

cohen = -0.0162, indicating a small effect. Only the relationship between authoritarian family and adolescent drug use was found to be statistically nonsignificant. That is, parents' control through guilt, punishment, and strict/harsh rules was not statistically significantly related to adolescent drug use. Democratic family, permissive family, positive affectivity, and negative affectivity were all found to be statistically significantly related to adolescent drug use(Z

=

4.121,Z= 4.428,Z= 15.729, andZ= 9.930, respectively). However, democratic family had a small effect in the unexpected direction (Z fisher = -0.1617). That is, parents' well-established rules, encouraging children to express and discuss family issues openly, and

}:

negative -0.11 0to affectivity 0.872

Note: *p<.05; REF=range ofthe effect size (Zfisher); DE=direction ofthe effect size (Zfisher);"+"

indicates expected direction of effect size; "." indicates unexpected direction of effect size; unexpected direction of effect indicates that a directional pattern of the relationship between adolescent drug use and family variables in a hypothesis test disconfmns the directional pattern of such a relationship in a research hypothesis; expected direction of effect indicates that the directional pattern in a hypothesis test confmns the directional pattern in a research hypothesis; k=number of hypothesis test; Zfisher=mean of weighted effect size (weighted by sample size); R2

= the square of correlation coefficient

between adolescent drug use and family environment; Dcohen=mean Cohen's d (effect size); 95%CI=

95% confident interval for Dcohen; Nfs = fail-safe number; x2 =chi-square =homogeneity statistic test (Qw=within-group, Qb=between-group).

The mean of the weighted effect size for permissive family was 0.1318, in the expected direction, which, corresponding to

R

2

=

0.017 and

D

cohen

=

0.264, was a small effect. That is, the parents who did not make well-established rules and ensure that their children followed these rules were more likely to have children who became drug users.

Both positive and negative affectivity have small effects in the expected direction (positive affectivity,

Z

fisher

=

0.11 0 I; negative affectivity,

Z

fisher

=

0.1747). That is, positive affectivity, adolescents' feeling of love, warmth, affection or support in the family, are associated with increased the resistance of adolescent drug use. In contrast, negative affectivity, adolescents' feeling of detachment, dissatisfaction, disappointment, or anger in the family, are associated with increased risk of adolescent drug use. The Nfs of democratic family (Nfs

=

7 ) and permissive family (Nfs

=

16) are less than 60 (e.g., Sk + 10

=

60, for democratic family) and 30 (e.g., Sk

+

10

=

30, for permissive family), respectively; therefore, these results may be due to sampling bias. That is, it may be relatively easy to find statistically nonsignificant results for democratic family and permissive family in future studies, although both democratic family and permissive family are statistically significantin this study.

Research Hypothesis Three

Table 4 shows'that significance levels and effect sizes are predictable by the .characteristics of the primary study, such as data collection methods, publication date, types of drugs, race, sample size, source of data, and gender. Among these characteristics of the primary study and background characteristics, one moderator, data collection methods, was

1tr1.~1f":4l' ~21JtJJ· ~1fl93.f6}]

FDUS 8 0.9120 0.4182 0.4133

SPI 7 2.181· 0.4980 0.5076 Zfisher=-16.426+(8.330E-03) • PUBYEAR Cohen's D=-37.334+(1.891E-02) • PUBYEAR DFM 11 9.750· 0.4095 -0.6821 Zfisher=-11.755+(5.967E-03) • PUBYEAR

Cohen's D=86.315 +(-4.358EE-02) • PUBYEAR SOEE 36 3.037· -0.0279 -0.0297 Zfisher= 1.193 +(-6. 153E-04) • PUBYEAR

Cohen's D=2.786+(-1.389E-03) • PUBYEAR SOEI 6 1.214 0.4611 0.4507 PSD PSP AFFN AFFP 10 4 10 23 2.671· -0.1742 -0.5577 -0.5059 0.2445 -0.0575 -0.0667 -0.7223 0.2446

Zfisher= 14.663+(-7.489E-03) • PUBYEAR Cohen's D= 19.789+(-9.983E-03) • PUBYEAR Zfisher= 46.829+(-2.352E-02) • PUBYEAR Cohen's D= 17.430+(-8.649E-03) • PUBYEAR Zfisher= 156.843+(-7.876E-02) • PUBYEAR Cohen's D=451.679+(-0.227) • PUBYEAR Zfisher= -23.343+(1.187E-02) • PUBYEAR Cohen's D=-45.902+(0.0234) • PUBYEAR Note: •p <.01; FDUP = parental drug use; FDUS = sibling drug use; SPI = single parent; DFM = departing of family members; SOEE = parents' education; SOEI = family income; PSD = democratic family; PSP=permissive family; AFFN =negative affectivity; AFFP =positive affectivity; Czfisher = correlation coefficient between Zfisher and publication year; Ccohen'sD = correlation coefficient between Cohen's D and publication year. PUBYEAR= publication year.

Summary

The findings of this study are presented as follows:

1. Based on the overall effect estimate for each of the family environment factors, family relations, and family structure, it was found that dysfunctional family environment, ineffective family structure, and dysfunctional family relations were all statistically significantly associated with adolescent drug use.

2. For family structure, a high socioeconomic status including family income and parents' education, one or both parents not living at home with the adolescent or parental death, family drug use including parental and sibling drug use, and living with a single parent significantly increased the opportunity for the adolescent to use drugs. Only the influence of sibling drug use had a medium effect on adolescent drug use; the rest of the subvariables, family income, parents'education, departing of family members, parental drug use, and single parent, had a small effect. However, the Nfs of family income, parents'

-Itr1.ft1f~• . ~ 21 $,}) • ~ ~ 93

-+

6 Jl

Discussion

In this study, adolescents in a family with a higher socioeconomic status were found to be at risk for substance use. This result challenges a view prevalent in social control theory that would predict that lower socioeconomic status would be related to increased problems with adolescent drug use because many lower class families may lack the motivation and skills needed to encourage a child into a conventional line of action and, as a result, fail to provide the "opportunity" needed for successful achievement (Miller, 1958; Kornhauser, 1978). However, this result may show support for strain theory because the adolescent in a family of higher socioeconomic status who feels the most pressure to succeed academically and who fears that he or she may not fulfill family expectations because he or she lacks efficient management technique to successfully cope with frustration may engage in the greatest amount of delinquency (Stinchcombe, 1964).

Consistent with previous research (Kaufman & Kaufmann, 1979), the results of living with a single parent or of the departing of family members showed a statistically significant relationship to adolescent drug use. These results display that a family crisis may develop through a family's experiencing a parent's death, divorce, or the separation of family members and stress each of the family members. Depending on the family members' overall coping capacity, a family crisis can facilitate a family environment for maladaptive behaviors by the children, and drug use is among possible maladaptive behaviors.

Social learning theory asserts that the decision to engage in deviant behavior is the result of differential "reinforcement." According to this theory, adolescents learn through interaction in which drug use is encouraged and rewarded by parental and sibling drug use. That is, the deviant behaviors of parent(s) and siblings provide reinforcement for adolescent delinquency as well as a model for imitation (Akers & Cochran, 1985). The findings of the relationship between adolescent drug use and family drug use show support for the social learning approach in which parental or sibling drug use leads to a significant increase in the level of reported adolescent drug use. Within family drug use, the correlation of sibling drug use and adolescent drug use is the greatest among the family variables.

While a pennissive family and negative affectivity facilitate the chance of adolescent substance use, positive affectivity diminishes the possibility of adolescent drug use. These

findings are consistent with social control theory in which substance use is more likely when attachment to family/parents is low. However, the findings of parental style show somewhat surprisingly that adolescents in a democratic family are at high risk to become drug users, and no support is found for the expected effect of authoritarian parents on adolescent drug use. These findings are inconsistent with common sense and with the previous findings that authoritarian parenting style increases and democratic parenting style decreases the opportunity for adolescent deviant behavior. Although this surprising finding in the democratic family results from synthesizing four primary articles (Block et aI., (1988), Brook et aI., (1981), Anderson & Henry (1994), and Brook, Whiteman, Gordon, & Cohen (1986», the surprising result of the positive relationship between democratic family and adolescent drug use may be attributed to the profound influence of the Block et aI., (1988) study, which includes four positive correlations between adolescent drug use and parents' open communication with their children andpare~ts' reasoning about their children's misbehavior and two negative correlations between adolescent drug use and parents' well-established rules. Furthermore, the sample size of Block et aI., (1988) was only 105 adolescents including 54 girls and 51 boys. Within this sample, only girls were surveyed about whether their parents had a democratic parenting style. This unique situation may make it difficult to generalize about the results concerning democratic parenting style.

.'\

Additionally, sampling bias could have produced these unexpected results.

However, when we consider the findings of "parental styles" and "affectivity"

;> simultaneously, one idea may occur that may imply that it is not the different conditions of

parental styles that influence adolescent drug use but the feelings of adolescents created by the process of different parental styles that may make an adolescent become a drug user or remain drug free. Presumably, it is in the feeling the adolescent attaches to the events (i. e., different parental styles) that the combination of parental styles and affectivity may be more appropriately explained by social control theory than by strain theory or social learning theory.

Elliott, Huizinga, and Ageton (1985) had tried to synthesize the perspectives of strain, social control, and social learning. They argued that adolescents have different socialization experiences, resulting in different degrees of "attachment" to conventional groups. The limited opportunities for achieving legitimate success induce "strain" on the individual, and

39-as a result the strength of the social "attachment" weakens. Nevertheless, at the same time they acknowledge that a social learning model would include all of the predictors used in the test of the integrated model. However, any theoretical integration runs the risk of doing

what Hirschi (1969) warns against: denying or ignoring crucial differences in each of theoretical key concepts and overemphasizing one or more of the theories in the interest of producing a harmonious model. Moreover, such a conceptual integration does not deal with differences in assumptions underlying the concepts (Akers& Cochran, 1985).

In sum, the basic premise of social control theory is that delinquency occurs when an individual's bond to society is weakened or broken (Hirschi, 1969); social learning theory assumes that delinquency occurs when there is an overflow defiance against the law, and this defiance is usually learned through interaction with siblings, parents, or peers (Matsueda, 1982). Strain theory assumes that delinquency is the result of pressures generated by a society that emphasizes success and in the meantime does not provide equal access to the means of achievement (Agnew, 1991). Each theory may prevail in one condition but may be flawed in other situations. Therefore, they are complementary rather than competing perspectives. In other words, theories selected to interpret the influence of family environment on adolescent drug use may depend on the kind of subjects and their conditions, the research designs, the selection of variables, the emphasis of the current study, and the interpretation of previous studies related to the similar current study.

Finally, in this study the statistical significance or nonsignificance across the relationship between adolescent drug use and family variables was examined by combining the Z of each primary study, but the effect size (Zfisher) indicating the magnitude or strength of the relation was not examined to demonstrate whether the magnitude/strength (Zfisher) of the relationship between adolescent drug use and family variables were statistically significant or not. That is, whether Zfisher in this study is statistically significant or not is unknown. It is possible that the statistical significance of the relationship between adolescent drug use and family variables may be associated with the nonsignificant strength/magnitude of their relations.

drawer problem. Although the fail-safe number shows an exact numerical value, it does not really indicate how many unpublished studies remain in file drawers; in fact, it needs to be compared to the standard of 5k+lO suggested by Rosenthal (1984) to see whether or not the research results will be tolerant to future null results. In a word, the fail-safe number only indirectly indicates the threshold number of null results that are needed to overturn the meta-analytic findings.

Conclusion and Suggestion

First, it is obvious that the conditions of some significant variables in the family structure, such as parents' education, family income, departing of family members, and single-parent family, cannot possibly be changed or controlled immediately or dramatically. However, parents should not become the models of drug use in the family (i.e., parental drug use), and they should show their children their love, wannth, affection, and support in an acceptable way in order to reduce the chance of the adolescents becoming involved in drug use (i.e., positive affectivity).

Second, at first glance at the variables of (a) parental style, democratic family, authoritarian family, and permissive family and at (b) positive affectivity and negative affectivity, a general picture of "family interaction and atmosphere" will appear. "Family interaction and atmosphere" will be interpreted in a common sense way because family interaction represents the parental style and affectivity implies the family atmosphere during the time of family interaction.

However, the results of tests of significance are somewhat different: positive affectivity, negative affectivity, and permissive family are statistically significantly related to adolescent drug use, but authoritarian family is not, and democratic family is statistically significantly related to adolescent drug use, but in an unexpected direction. From this point of view, parents should pay more attention to what their children feel during the time of family interaction. These unusual results mean that the adolescent psychological mechanisms between the conditions of parental style and affectivity are worthy of being investigated in the future.

6: 323-46.

Amoateng,A.Y.,& Bahr, S. J. (1986). Religion, family, and adolescent drug use.Sociological Perspectives, 29(1), 53-76.

Anderson, A. R., & Henry, C. S. (1994). Family system characteristics and parental behaviors as predictors of adolescent substance use.Adolescence, 29(114), 405-420.

Bachman,1. G., Johnston,L. D.,& O'Malley, P. M. (1981). Smoking, drinking, and drug use among American high

~choolstudents: Correlates and trends, 1975-1979.American Journal ofPublic Health, 71(1),59-69.

*Bauman, K. E., Foshee, V.A.,Linzer, M.A.,& Koch, G. G. (1990). Effect of parental smoking classification on the association between parental and adolescent smoking.Addictive Behaviors, 15, 413-422.

Beardslee, W. R., Son,L.,& Vaillant, G. E. (1986). Exposure to parental alcoholism during childhood and outcome in adulthood: A prospective longitudinal study.British Journal ofPsychiatry, 149,584-591.

Binion, V. J., (1979). A descriptive comparison of the families of origin of women heroin users and nonusers. In: Addicted Women: Family Dynamics. Self Perceptions, and Support Systems. Washington, D. C. : U.S. Government Printing Office (Dhew Publication No. (ADM) 80-762).

Block, J., Block, J. H.,& Keyes, S. (1988). Longitudinally foretelling drug usage in adolescence: Early childhood personality and environmental precursors.Child Development, 59, 336-355.

Brook, J. S., Lukoff,I.F.,&Whiteman, M. (1977). Peer, family, and personality domains as related to adolescent's drug behavior.Psychological Reports, 41, 1095-1102.

Brook, J. S., Lukoff,I.F.,&Whiteman, M. (1980). Initiation into adolescent marijuana use. The Journal ofGenetic Psychology, 137, 133-142.

Brook, J. S., Nomura, C., & Cohen, P. (1989). A network of influences on adolescent drug involvement: Neighborhood, school, peer, and family. Genetic, Social, and General Psychology Monographs, 115,(1),

125-145.

Brook, J. S., Whiteman, M.,& Gordon,A. S. (1981). Maternal and personality determinants of adolescent smoking behavior.The Journal ofGenetic Psychology, 139, 185-193.

Brook, J. S., Whiteman, M., Gordon,A. S.,& Brenden, C. (1983). Older brother's influence on younger sibling's drug use.The Journal ofPsychology. 114, 83-90.

Brook, J. S., Whiteman, M., Gordon,A.S.,& Cohen, P. (1986). Some models and mechanisms for Explaining the impact of maternal and adolescent characteristics on adolescent stage of drug use. Developmental Psychology, 22, 460-467.

Brook, J. S., Whiteman, M., Gordon,A.S., Nomura, c.,& Brook, D. W. (1986). Onset of adolescent drinking: A longitudinal study of intrapersonal and interpersonal antecedents.Advances in Alcohol and Substance Abuse, 5, 91-110.

Brunswick,A. F.,& Messeri, P. (1984). Causal factors in onset of adolescent's cigarette smoking: A prospective study of urban black youth.The Addictive Behaviors, 3(0.5), 35-52.

Chassin, L., Presson, C. C., Sherman, S. 1., Montello, D., & McGrew, J. (1986). Changes in peer and parent influences during adolescence: Longitudinal versus cross-sectional perspectives on smoking initiation.

Developmental Psychology,22(3), 327-334.

Cohen, 1. (1960). A coefficient of agreement for nominal scales.Educational and Psychological Measurement, 20, 37-46.

Cohen, J. (1981).Statistical power analysis forthe behavioral science.New York: Academic Press.

Cook,T.,& Leviton,L.(1980). Reviewing the literature: a comparison of traditional methods with meta-analysis. Journal ofPersonality, 48, 449-472

Denton, R. E.,& Kampfe, C. M. (1994). The relationship between family variables and adolescent substance abuse: A literature review.Adolescence. 29(114).

Dulak, 1. A (1995). Understanding meta-analysis. In: Grimm, L. G. & Yamold, P. R. (Eds.), Reading and understanding multivariate statistics(pp. 319-352). Washington, D.

c.:

American Psychology Association. Elliott, D. S., Huizinga, D.,& Ageton, S. S. (1985).Explaining delinquency and drug use.Beverly Hill, California:

Sage.

Fagan, 1., Weis, 1. G.,& Cheng, V-To (1990). Delinquency and substance use among inner-city students.Journal of Drug Issues,20(3),351-402.

Friedman,A S., Pomerance, E., Sanders, R., Santo, Y., & Utada, A (1980). The structure and problems of the families of adolescent drug abusers.Contemporary Drug Problem, 327-356.

Gantman, C. A (1978). Family interaction patterns among families with normal, disturbed, and drug-abusing adolescents.Journal of Youth and Adolescence,7, 429-440.

Glass, G. V. (1977). Integrating findings: The meta-analysis of research. Review of Research in Education, 5, 351-379.

Hirschi,T.(1969).Causes ofdelinquency.Berkeley: University of California.

Hoffmann (1991).Family Structure, Family Relationships, and Adolescent Drug Use: A Theoretical and Empirical Examination.Doctoral dissertation, State Uinversity of New York at Albany.

Hundleby,1. D. (1987). Family and friends as social environments and their relationship to young adolescents' use of alcohol, tobacco, and marijuana.Journal ofMarriage and the Family, 49, 151-164.

Jackson, G.B.(1980). Methods of integrative reviews.Review ofEducational Research, 50, 438-460.

Jessor,R.,Chase, J.A,&Donovan, 1. E. (1980). Psychosocial correlates of marijuana use and problem drinking in a national sample of adolescents.American Journal ofPublic Health, 70(6), 604-613.

·Johnson, G., Shontz, F., & Locke, T. (1984). Relationships between adolescent drug use and parental drug behavior.Adolescence, /4,295-299.

Jurich,A P., Polson,C.1., Jurich, 1.A,& Bates, R.A (1985). Family factors in the lives of drug users and abusers. Adolescence, 20, 143-159.

Kandel, D.B.,Kessler, R.,& Margulies, R. (1978). Antecedents of adolescent initiation into stages of drug use: A developmental analysis. In: Kandel, D. B. (Ed.), Longitudinal research on drug use: Empirical findings and methodological issues.(pp.73-101). Washington, D.C.: Hemisphere-John Wiley.

Kaufman, E., & Kaufmann, P. (1979). From a psychodynamic orientation to a structural family approach in the treatment of drug dependency. In: Kaufman, E& Kaufmann, P. (Eds),Family Therapy of Drug and Alcohol

45-Abuse. (pp. 43-54). New York: Gardner.

Koplin, M. D. (1987). Family, Religious, and Peer Influence on Adolescent Drug-use. Unpublished doctoral dissertation, Brigham Young University, Provo, UT.

Kornhauser, R. (1978).Social Sources ofDelinquency: An Appraisal ofAnalytical Models. Chicago: University of Chicago Press.

Levitt, E. E., & Edwards, J. A. (1970). A multivariate study of correlative factors in youthful cigarette smoking. Developmental Psychology, 2(1), 5-11.

Levy, S. D. (1980).Adolescent drug users and their parents: a study of their personality and family interaction. Unpublished doctoral dissertation, Northwestern University, Evanston, IL.

Margulies, R. Z., Kessler, R. C., & Kandel, D. B. (1977). A longitudinal study of onset of drinking among high-school students.Journal ofStudies on Alcohol, 38(5), 897-912.

Matsueda, R. L. (1982). Testing control theory and differential association: a causal modeling approach.American Sociological Review, 47 (4): 489-504.

McCoy, C. B., Metsch, L. R., & Inciardi, J. A. (1996). Intervening with Drug-Involved Youth. Thousand Oaks, California: Sage.

McDermott, D. (1984). The relationship of parental drug use and parents' attitude concerning adolescent drug use to adolescent drug use.Adolescence, XIX, 73, 89-97.

McGee, Z. T. (1992). Social class differences in parental and peer influence on adolescent drug use. Deviant Behavior, 13, 349-372.

McLaughlin, R.J.,Baer, P. E., Burnside, M.A.,& Pokorny,A.D. (1985). Psychosocial correlates of alcohol use at two age levels during adolescencejournalofStudies on Alcohol, 46(3), 212-218.

Miller, W. B. (1958). Lower-class culture as a generating milieu of gang delinquency.Journal ofSocial Issues, 14, 5-19.

Mills, C.J.,& Noyes, H. L. (1984). Patterns and correlates of initial and subsequent drug use among adolescents. Journal ofConsulting and Clinical Psychology, 52(2), 231-243.

Mullen,B. (1989).Advanced BASIC meta-analysis. Hillsdale, NJ: Lawrence Erlbaum.

Needle, R., McCubbin, H., Wilson, M., Reineck, R.,Lazar,A.,& Mederer, H. (1986). Interpersonal influencesin adolescent drug use - The role of older siblings, parents, and peers. The International Journal of the Addictions, 21, 739-766.

Norem-Hebeisen,A.,Johnson, D. W., Anderson, D., & Johnson, R. (1984). Predictors and concomitants of changes in drug use patterns among teenagers.The Journal ofSocial Psychology, 124,43-50.

Petitti, D. B. (1994). Meta-analysis, decision analysis and cost-effectiveness analysis in medicine: Methods for quantitative synthesis in medicine. Oxford: Oxford University Press.

Rees, C. D.,& Wilborn, B. L. (1983). Correlates of drug abuse in adolescents: A comparison of families of drug abusers with families of nondrug users.Journal of Youth and Adolescence, 12,55-63.

Robinson, T. N., Killen, J. D., Taylor, C. B., Teich, M.J.,Bryson, S. W., Saylor,K.E., Marson, D.J.,Maccoby, N.,

& Farquhar, J. W. (1987). Perspectives on adolescent substance use: A defined population study.Journal of

the American Medical Association,258( 15), 2072-2076.

Rosenthal, R. (1979). The "file drawer problem" and tolerance for null results. Psychological Bullentin, 86, 638-641.

Rosenthal, R. (1984).Meta-analytic procedures for social research. Beverly Hills, CA: Sage.

Rosenthal, R.& Hall, J. (1981). Critical values of Z for combing independent probabilities.Replications in Social Psychology, 1(2),1-6.

Rosenthal, R. &Mullen, B. (1985). BASIC meta-analysis: Procedures and programs. Hillsdale, NJ: Lawrence Erlbaum.

Rosenthal, R.,& Rubin, D. B. (1986). Meta-analytic procedures for combining studies with multiple effect sizes. Psychological Bulletin,99, 400-406.

Schaffer, Y. L. (1988).Adolescent substance use: family adaptability and cohesion, family structure, and gender. Unpublished master's thesis, Northern Illinois Univesity, Dekalb, Illinois.

Schultz, D. A., & Wilson, R. A. (1973). Some traditional family variables and their correlations with drug use among high school students.Journal ofMarriage and the Family,35, 628-631.

Shadish, W. R.,&Haddock,C.K. (1994). Combining estimates of effect size. In H. Cooper&L.V. Hedges (Eds.), The handbook ofresearch synthesis(pp. 261-281). New York: Russell Sage Foundation.

Stanton, M. D. (1979). Drugs and the family.Marriage& Family Review,2, 2-10. Stinchcombe,A. L. (1964).Rebellion in a High School. Chicago: Quadrangle.

Streit, F., Stanton, M. D., & Todd, T. C. (1982). Patterns of tum-taking and alliance formation in family communication.Journal ofCommunication, 32, 148-160.

Tee, N. (1974). Parent-child drug abuse: Generational continuity or adolescent deviancy?Adolescence,4,351-364. Tolone, W. L.,& Dermott, D. (1975). Some correlates of drug use among high school youth in a midwestern rural

community. The International Journal ofthe Addictions,10,761-777.

Tudor, C. G., Petersen, D. M., & Elifson, K. W. (1980). An examination of the relationship between peer and parental influences and adolescent drug use.Adolescence, 783-798.

White,H. D. (1994). Scientific communication and literature retrieval. In: Cooper, H., & Hedges, L. 15(60)V.

(Eds.),The handbook ofresearch synthesis(pp.41-55). New York: Russell Sage Foundation. Wilkinson, K. (1974). The broken home and juvenile delinquency.Social Problem, 21,726-740.

*studies included in the meta-analysis

47-1'"""'''

I

I

i

APPENDIX

Formulas for the Transformation ofInferential Statistics intors

fort, r =

~t2/(t2

+df)

for F(l,dt),

r

=

~

p/(p

+

df)

for chi-square= %2(1),

r

=

~

%2/

N

for Z,

r

=

~

Z2 /

N

Formulas for the Transformation of Inferential Statistics into Zs for t,

Z

=

~df

{In[l

+

(t2 / df)]} .

~l-

(l/2df)

for F(l,dt),

Z

=

~df

{In[l

+

(P /

df)]} .

~l-

(l/2df)

for chi-square=%2(1),

Z

=

R

forr, t

=

r.JN

-2/~

thenz=~df{ln[1+(t2

/df)]} .

~1-(l/2df)

Note: Logarithms are natural logarithms (i.e., logeand not loglO)'

Formula for Convertingr to ZFISHER or ZFISHER tor

ZFISHER=.5(In[(I+r)/(l-r)]}

formula for the fail-safe number for thep

=

.05 level of significance

Formula for the moderate effects

Note:The statistical software of Advanced BASIC Meta-Analysis (Mullen, 1989) and DSTAT (Johnson, 1993) would be used for data analysis.

~fttj 2003.3.18

~lE 2003.11.24

1i~ 2004.2.27

49-The Impact of Family Factors on Adolescent

Drug Use -

A Meta-analytic Approach

Shin-Liang Li

Abstract

Findings regarding the influence of family environment on adolescent drug use were as follows:

I.Regarding the findings of family structure, the opportunity for adolescent drug use was significantly increased by a high socioeconomic status family income and parents' education, one or both parents not living at home with the adolescent or parental death, family drug use including parents' and sibling drug use, and single parent. Only the influence of sibling drug use has a medium effect size on adolescent drug use; the rest of the variables have a small magnitude of influence.

2.Regarding the findings of family relations in an authoritarian family, the parents' control through guilt, punishment, or strict rules was not significantly associated with adolescent drug use. In a pennissive family, their parents may not make well-established rules, but even if they do, they may not make sure that their children follow these rules, resulting in some adolescents becoming likely drug users. In a democratic family where parents reason with their children and may make well-established rules and discipline or encourage their children to express their opinions and discuss family events openly, adolescents are surprisingly likely to use drugs. This finding is somewhat different from the common-sense idea that the democratic parenting style would decrease the opportunity of adolescent drug use. The more adolescents felt their parents' support, affection, love or warmth (positive affectivity), the less likely they were to become drug-users.

In sum, sibling influence is the largest than any other of family variables. Hence, in the -future, it is highly suggestible to decrease the sibling negative influence that means to reduce

the risk of adolescent drug use.