An Index to Screen Adolescents

An Index to Screen Adolescents at Risk for Substance Abuse

Catherine S. K. Tang

Chinese University of Hong Kong

Ralf Schwarzer

Freie Universität Berlin, Germany

Connie S. Y. Wong

Chinese University of Hong Kong

Address correspondence to :

Catherine Tang, Ph.D.
Chinese University of Hong Kong
Department of Psychology
Shatin, NT, Hong Kong
Fax: +852/2/603 5019
E-Mail: b077764@hp720a.csc.cuhk.hk

Abstract

Previous research suggests that deviant behaviors during adolescence result mainly from social influence, such as peer modeling or pressure. The more that teenagers are susceptible to social influence, the more they are libel to be at risk for deviant behaviors like drug abuse or criminal offense. A brief psychometric index was sought that allows to discriminate between delinquent youths and high school students and predict the probability for being detained in correctional institutions of Hong Kong. Incarcerated youth were compared to high school students in terms of susceptibility to peer pressure. Logistic regression analyses showed that a 3-item index was able to predict group membership for about 90% of cases. Separate equations were established that can be used within a psychological screening program to identify adolescents at risk for substance use. Moreover, this approach was compared to the one that uses a psychometric scale with cut-off scores for cases and noncases. Whatever method is practiced, 3 items on peer pressure appeared to make valid predictions about being deviant or not.

Keywords: drug use, adolescents, delinquents, prevention, peer pressure, social influence, screening

An Index to Screen Adolescents at Risk for Substance Abuse

The local situation of drug abuse among adolescents in Hong Kong shows an undesirable trend, requiring further efforts to screen teenagers at risk as a starting point for a prevention program. The role of social influence in the drug use initiation process is described, with susceptibility to peer pressure being a major factor that predisposes teenagers for subsequent deviant problem behaviors, including substance use. Based on questionnaire data, a screening index is developed that may allow to predict the probability of Hong Kong adolescents becoming a case for correctional institutions.

Adolescent Substance Abuse in Hong Kong

According to statistics released by the Central Registry of Drug Abuse (1993) in Hong Kong, there was a marked increase of 71.5% of newly reported drug addicts under 21 years of age in 1993 compared with the first half of 1992. Overall, the mean age of first illicit drug use decreased from 22.6 years in 1991 to 21 years in 1993. The drugs most frequently used by those under 21 years in the first half of 1993 was heroin (70%), followed by cannabis (20%), cough medicine (15%), and Flunitrazepam (4%). According to statistics released by the Society for the Aid and Rehabilitation of Drug Abusers, there was an increase of 42% in men and 58% in women attending their centers from 1993 to 1994. New drug users under 25 years attending their centers increased by 57.6% in men and 60% in women. The number of young people who abuse drugs may be two or three times higher than the government estimate of 3,000. According to government statistics, the number of new drug abusers under 21 years increased 38% from the first half of 1993 to the first half of 1994, while marijuana abusers increased 30% during the time period from 1985 to 1993. Teachers, social workers, and legislators expressed their concern about the psychosocial and legal problems associated with the recent alarming increase in adolescent substance abuse (Ming Pao Newspaper, December 10, 1994). They argued that adolescent drug abuse was related to families breaking up, accessibility of drugs, peer influences, and influences from Western countries. Adolescent substance abuse was also linked to increasing school dropout and delinquent behavior: Men may be involved in illegal trafficking of drugs, and women may be engaged in prostitution. There is a recent controversy about the decriminalization of marijuana use. Some people argue that marijuana is nonaddictive, has no long-term side-effects, and needs no treatment; thus, it should not be criminalized. Others argue that with the negative psychosocial implications, especially among adolescents, severe penalties should be imposed to discourage its use.

Most of the scant research on substance use among adolescents in Hong Kong was conducted by social workers. It is usually descriptive in nature, aiming at the relationships among sociodemographic characteristics, adolescents' knowledge and attitudes toward taking drugs, or patterns of drug use. The results are, in general, similar to those found in Western literature.

Social Influence on the Initiation of Substance Use

Deviant behavior among adolescents is driven by various factors in both the person and the environment. Research provides evidence that drug use can be associated with personality traits such as self-esteem or sensation seeking, and with social influence variables such as family drug use, peer drug use, and peer pressure to try substances (Bentler, 1992; Botvin et al., 1992; Farrell, Danish, & Howard, 1992; Goddard, 1992; McNeill, 1992; Newcomb & Felix-Ortiz, 1992; Zuckerman, 1979). Temperamental, social, and stress factors increase teenagers’ vulnerability towards substance use, and further maladaptive coping attempts might influence the actual initiation of the problem behavior (Wills & Hirky, 1996). Social influence is regarded as a major factor in the development of deviance, in particular of drug abuse. Adolescents resort to drugs when experiencing social conflicts, thus using substances as a maladaptive coping strategy (Wills, 1986). They also face a number of developmental tasks to prove themselves in their reference groups (Jessor, 1987, 1993; Yu & Williford, 1992). If the dominant reference group consists mainly of peers who value drug use as a mature "adult behavior" or as an indicator of independence from parents, then the individual attempts to meet these expectations. There may be role conflict, for example when family norms contradict peer group norms, and it may take an extended period of time until the youngster eventually complies with one or the other.

Zucker (1979) considered peer influence to be a particularly potent extra-individual determinant of the transition to alcoholism because peer drug use is both a model for drug-taking behavior and a source of availability of the substance. Kandel (1980) concluded that peer-related factors, for example extent of perceived drug use in the peer group, self-reported use by peers, and perceived tolerance, are consistently the strongest predictors of subsequent alcohol and marijuana use, even when other factors are controlled. Barnea, Teichman, and Rahav (1992) found that peer influence on adolescent drug use is even more significant than that of parents. Peers' drug use behavior has a direct effect both on adolescents' drug use and on their behavioral intentions. Perceived peer attitudes have a strong effect on the adolescents' own attitudes and on their actual drug use (see also Hansell & Mechanic, 1990; Iannotti & Bush, 1992; Stacy, Newcomb, & Bentler, 1992, 1993; Stanton & Silva, 1992).

Peer norms regarding substance use, considered alone, are also positively related to the adolescents' use of substances (e.g.,. Donovan & Jessor, 1978, 1985; Ellickson & Hays, 1992; Newcomb, Maddahian, Skager, & Bentler, 1987). Biddle, Bank, and Marlin (1980) found in a path analysis that peer norms regarding alcohol use showed a significant direct effect on the students' own norms, which in turn had a significant direct effect on the students' frequency of alcohol use. This was recently replicated in another path analysis by Webster, Hunter, and Keats (1994). White, Labouvie, and Bates (1985) found friends' norms to be more highly correlated with the adolescents' substance use than friends' use. These peer norms, generally measured in terms of peer approval, are positively correlated with peer drug use. In most studies that assessed both peer norms and peer use, the peer use variable was a more salient predictor than peer norms. The effect of peer behavior on the students' alcohol use, however, was stronger than the effect of peer norms. Rooney (1982) also found in a multiple regression analysis that friends' alcohol use is a more salient predictor than the friends' norms. Newcomb et al. (1987) found peer norms to be the most highly correlated out of 12 risk factors with a composite substance use score. Wills, Baker, and Botvin (1989) suggested that adolescents may smoke either because they are exposed to smoking models or because they experience explicit peer pressure. Socially active students may be exposed to smoking models more frequently, or they may be more likely to experience situations where there is peer pressure for substance use. In addition, it has been proposed that socially oriented students are more sensitive to social evaluations and hence more susceptible to conformity or social image pressures. Dielman, Butchart, Shope, and Miller (1990) confirmed and extended the results of the earlier studies, indicating that peer alcohol use and peer norms regarding alcohol use, in combination with susceptibility to peer pressure, accounted for most of the variance in adolescent alcohol consumption. Susceptibility to peer pressure emerged as the most important predictor of both adolescent alcohol use and misuse. Dielman, Campanelli, Shope, and Butchart (1987) found that susceptibility to peer pressure was more highly correlated with all substance use, misuse and intention items than were self-esteem and health locus of control. A conceptual model proposed by Dielman et al. (1990) claims that early childhood exposure to deviant parental norms (e.g., tolerance of adolescent substance use) and behavior with respect to substance use (e.g., immoderate alcohol use or use of illicit drugs) increases the child's tolerance of deviant behavior, which in turn leads to increased susceptibility to peer pressure as well as to attraction to peer groups with deviant norms and behaviors. Increased exposure to deviant peer norms and behavior in combination with increased susceptibility to peer pressure is then hypothesized to result in a greater likelihood of the adolescent's adoption of deviant norms and drug-use behaviors (see also Alberts, Hecht, Miller-Rassulo, & Krizek, 1992; Rose, Bearden, & Teel, 1992).

Research Questions

The question was whether an index, based on susceptibility to peer pressure, could be developed that allows for early screening of adolescents at risk for drug abuse or, possibly, overall deviance. It is assumed that youths who use drugs are deviant in a broader sense because they tend to resort to illegal purchase of drugs. The present study compares high school students with delinquents who were incarcerated for substance abuse among various other reasons. Thus, it is expected that a deviant subculture with high illicit drug use reports a different level of susceptibility to peer pressure than a high school population with low substance use.

The aim of this study was to develop a parsimonious screening instrument based on a minimum of self-report statements about social influence. It should discriminate between incarcerated offenders and school students and, thus, predict the probability of being detained in correctional institutions.

Method

Participants

A sample of 1,001 adolescents in Hong Kong were contacted. Thirty-two of the questionnaires (3.1%) had to be discarded for obvious response sets. Finally, 969 valid questionnaires were used for data analysis. Among these 969 participants, 59.8% (n = 579) were secondary school students studying from Form 2 to Form 6 in two different schools. One of the schools was classified by the Education Department as around the "Band Four" to "Band Five" school, that is, students' academic performance was among the lowest in Hong Kong. Another was a "Band Two" school, where students' academic performance was about average to high average.

In addition to the school sample, incarcerated delinquents were also included in the study. Four institutions of the Correctional Services Department (CSD) were approached. Almost all of the young offenders in these institutions, who were able to read and had about Primary 6 education level, participated in the research. Three correctional homes of the Social Welfare Department (SWD) were also approached. The response rate was 30 to 40%. The data collected in the CSD and SWD were collapsed and are referred to as the "delinquent sample" throughout this paper.

Table 1

Composition of Sample: Number of Participants Broken Down by Institution and Gender

	Boys	Girls	Subtotal
School sample	227 (39.2%)^a	352 (60.8%)^a	579 (59.8%)^b
Delinquent sample	302 (77.4%)^a	88 (22.6%)^a	390 (40.2%)^b
Subtotal	529 (54.6%)^a	440 (45.4%)^a	Total = 969 (100%)
c² (1) = 135.86, p < .001

Note. ^a Row percentages. ^b Column percentages.

As shown in Table 1, 59.8% (n = 579) of the participants were high school students and 40.2% (n = 390) were incarcerated delinquents, many of them convicted of drug-related crime. Among the students, 39.2% (n = 277) were boys and 60.8% were girls (n = 352). Of the delinquent sample, 77.4% (n = 302) were boys and 22.6% (n = 88) were girls. Thus, more girls were in the school sample, whereas more boys were in the delinquent sample.

The mean age of the school sample was 15.87 years (SD = 1.75), 15.91 years for boys and 15.84 years for girls. In the delinquent sample, the overall mean age was 17.32 (SD = 1.75). Among them, boys in average were 17.28 years old and the girls 17.43.

Measures

The data set included a variety of measures to assess constructs such as self-esteem, attitudes, sensation seeking, and susceptibility to peer pressure. From previous analyses (Wong, Tang, & Schwarzer, 1997), it was evident that the most promising set of items for the present purpose was susceptibility to peer pressure. Seven items on perceived susceptibility to peer pressure by Dielman et al. (1987) were used (e.g., "If a friend offered you a drink of alcohol, would you drink it?"). Participants were asked how far they agreed on a four-point scale from strongly disagree (1) to strongly agree (4). Other measures used in an exploratory stage did not emerge as fruitful for the present analysis.

Procedure

The data collection took place from November 1993 to February 1994. Self-report questionnaires in Chinese were administered in group format by reading standardized instructions, including a brief description of the study. The subjects were reminded of the anonymous nature of the questionnaire and were assured that their responses would be confidential and would not be revealed to parents, school or institution personnel. In addition, it was emphasized that the drug use items pertained to nonmedical use of substances only (e.g., cough medicine). On average, the questionnaire took about 30 minutes to complete.

Analyses

The aim of the present analysis was to identify a parsimonious set of items that "predicted" group membership. For this purpose, logistic regression analysis as available in the SPSS statistics package (Norusis, 1992) was chosen. This procedure requires fewer assumptions than discriminant analysis. It estimates the probability for an event to occur or not to occur. Thus, a dichotomous criterion is predicted by a number of independent variables (or "covariates"). In the present study, the criterion was an adolescent being in a correctional institution (CI = 1) or not (CI = 0). The relationship between the covariates and the probability is nonlinear and can be written as:

Prob (event) = 1 / (1 + e ^-z)

with z = (b₀ + b₁X + b_pX_p...) (where e is the base of the natural logarithms, approximately 2.718). The parameters of the model are estimated in an iterative manner using the maximum-likelihood method (Hosmer & Lemeshow, 1989). By this method, regression coefficients (logits) are estimated that make the observed group membership most likely. Separately for both genders, being incarcerated or not was chosen as the dependent variable with the seven susceptibility items as the independent variables or covariates. The stepwise forward method was chosen to search for the most parsimonious model, thus not including those items that did not make an additional significant contribution.

Results

Search for Most Predictive Items in Male Subsample

The stepwise logistic regression procedure selected three out of seven susceptibility items with the following results: -2 log likelihood c 2 = 293.5, model c 2 = 422.9 (3 df) (p < .001), and goodness of fit c 2 = 592.4 (521 df) (p = .016). Table 2 displays the classification results.

Table 2

Classification Results for Boys

	Predicted Group Membership
	School	Incarcerated	% correct
Observed school	197	27	88%
Group membership incarcerated	30	271	90%
Overall			89%

Overall, 89% of the boys were classified correctly. Of 224 students, 197 were classified as such (88%), whereas of 301 incarcerated boys 271 were classified correctly (90%). Based on three items, this procedure resulted in the misclassification of only 30 incarcerated adolescents and 27 school students. The susceptibility to peer pressure items selected were the following:

If a friend wants you to smoke a cigarette and your parents don't want you to, would you smoke it?
If your best friend is skipping school, would you skip too?
If your friends offer you pills, heroin or cough medicine, would you try them?

The logistic regression equation for the boys was:

Probability (CI) = 1 / (1 + e ^-z)

with z = -8.97 + 1.60 * item 1 + 1.59 * item 2 + 1.08 * item 3.

The corresponding odds ratios were 4.95, 4.91, and 2.94, respectively. When applied to the subsample of male school students, this formula resulted in an average probability of being in the correctional institutions (CI) of .20 (SD = .24) compared to an average probability of .85 (SD = .23) for the incarcerated boys. The higher the probability, the more an individual can be regarded as being at risk for deviance.

Cross-Validation in the Female Subsample

Usually, findings based on one sample are hard to cross-validate in a different one. But most surprisingly, the stepwise logistic regression for girls yielded exactly the same three items in the same order with the following results:
-2 log likelihood c 2 = 246.5 (433 df) (p = 1.0), model c 2 = 187.0 (3 df) (p < .001), and goodness of fit c 2 = 387.2 (433 df) (p = .94). Table 3 displays the classification results.

Table 3

Classification Results for Girls

	Predicted Group Membership
	School	Incarcerated	% correct
Observed school	329	22	94%
Group membership incarcerated	32	54	63%
Overall			88%

Overall, 88% of the girls were classified correctly. Of 351 school students, 329 were classified as such (94%), whereas of 86 incarcerated girls 54 were correctly classified (63%). Based on three items, this procedure resulted in the misclassification of 32 incarcerated girls, but only 22 school students. The logistic regression equation for the girls was:

Probability (CI) = 1 / (1 + e ^-z)

with z = -7.28 + 1.17 * item 1 + 0.75 * item 2 + 0.77 * item 3.

The corresponding odds ratios were 3.23, 2.11, and 2.15, respectively. When applied to the subsample of female school students, this formula yielded an average probability of .11 (SD = .18) compared to an average probability of .56 (SD = .28) for the incarcerated girls. Although this is a considerable difference between the two groups of girls, it is not as impressive as the one for the boys. The overall classification being satisfactory, the selected index was, however, less successful in discriminating incarcerated girls from school girls.

Further Examination of the Resulting 3-Item Index

In the following section, a more traditional approach is chosen to illustrate how the three items operate jointly and separately in distinguishing between the two groups of teenagers. Also, psychometric characteristics of the index are determined.

Having found that three items are optimal in predicting group membership, the four redundant items were discarded for further analysis. A two-group multivariate analysis of variance was computed that employed the index as three dependent variables. The multivariate results were Wilks’ lambda = .46, F[2,963] = 323.5, p < .001, which translates into 54% of variance accounted for by the index. The univariate results are reported in Table 4, along with the means and standard deviations for the two groups.

Table 4

Univariate Results of the MANOVA

		Normals	Delinquents
		(n = 577)	(n = 389)	F	p <	eta ²
Item 1	M	1.67	2.91	658	.001	.41
Item 1	SD	(0.78)	(0.66)
Item 2	M	1.70	2.84	526	.001	.35
Item 2	SD	(0.74)	(.78)
Item 3	M	1.45	2.71	759	.001	.44
Item 3	SD	(0.63)	(0.79)

Instead of using three single predictors within a logistic regression equation, one could resort to the more straightforward and familiar approach of developing a brief psychometric scale and use its total score for predictions. Its internal consistency was very high considering its shortness (Cronbach’s alpha = .83), and the corrected item-total correlations were .70, .67, and .70, respectively. The scale mean was 6.28 (SD = 2.45, n = 966). Girls and boys differed significantly in their level of susceptibility to peer pressure (girls: M = 5.6, SD = 2.41; boys: M = 6.8, SD = 2.32; t[964] = 8.1, p < .001).

Using the latter approach, the most difficult question is where to set the cut-off score for assigning persons to a risk group for further diagnostics. This is the issue of sensitivity and specificity. Sensitivity is the proportion of true cases correctly identified, whereas specificity is the proportion of true normals correctly identified. The theoretical and empirical range of the psychometric scale is 3 to 12 (since each item had a four-point response format). Setting the optimal cut-off score for the boys between 6 and 7 resulted in a specificity of 87% and a sensitivity of 93%. Setting the optimal cut-off score for girls between 8 and 9 yielded a specificity of 89%, but a sensitivity of only 76%. This confirms the gender difference in predictability already found above in the logistic regression analysis. The scale is very sensitive in identifying deviant boys, but much less so in identifying deviant girls.

In sum, whether one uses the simple, familiar method of applying a psychometric scale, or the much more sophisticated logistic regression equation does not make much of a difference in terms of clinical decision-making, after the most relevant and parsimonious subset of three variables had been detected by a multivariate rationale.

Discussion

It is well known that social influence by peers represents the primary determinant for social deviance in adolescents. Susceptibility to peer pressure can be assessed psychometrically by a few self-report items and may serve as an indication for deviance, in particular for substance abuse. The present study investigated whether incarcerated deviant teenagers can be successfully discriminated by a minimum of self-report statements from those who attend high school. The striking result was that it is possible with only three items to achieve a 90% prediction of group membership. Susceptibility to peer pressure seems to be a sensitive variable that allows a valid separation of deviant from nondeviant adolescents. Moreover, it was striking that the results found in the subsample of males were cross-validated in the subsample of females, although with some constraints. The overall prediction was even better, but the most crucial classification of deviant girls into the group of incarcerated girls was less successful, with 37% of them being misclassified. Depending on the purpose of further screenings, however, this Type II error might be tolerable.

It was also examined whether the present results were dependent on the kind of statistical analysis chosen. For the given problem, one can either turn to discriminant analysis or logistic regression. The former focusses more on the optimal separation of groups, whereas the latter aims at the prediction of an event to occur. Both methods have been compared for this data set with exactly the same results. For the presentation, however, logistic regression was the preferred choice since it makes fewer assumptions and performs very well under most circumstances.

In addition to the multivariate approaches, more familiar psychometric methods have been employed, in particular to allow for comparison with other research. It was found that the three-item psychometric scale not only had a very high internal consistency, but also a very high sensitivity coefficient for boys. That is, the instrument can validly identify not only true normals (specificity), but also true cases (sensitivity) in the latter subsample. The new measure compares very well with other screening instruments. For example, a four-item alcohol screening measure (Ewing, 1984) yielded sensitivity coefficients between 84% and 89% in different samples, with specificities between 95% and 100%. The ten-item brief Michigan Alcoholism Screening Test had a sensitivity of 91%, but only a specificity of 83% (Willenbring, Christensen, Spring, & Rasmussen, 1987). The 12-item version of the General Health Questionnaire yielded a median sensitivity of 86% and a median specificity of only 80% (Goldberg & Williams, 1991, p. 50). These comparison data have to be interpreted in light of the preferred cut-off scores. Moving this score up and down results in a trade-off between sensitivity and specificity.

Some limitations have to be taken into account. It has to be kept in mind that this was a cross-sectional study on a unique sample of adolescents in Hong Kong. Therefore, the above results and the equations in particular must be understood as preliminary, requiring validation by further studies. The scope of validity remains somewhat ambiguous because there was no detailed information about the variety of criminal offenses by the teenagers, including history of deviant behavior, multiple behavior problems, previous detention, and so forth. Also, information about problem behaviors among the high school students was lacking. It could be, for example, that some of the students are in fact offenders who have not been identified or reported so far.

For practical considerations, the present results represent a promising first step to quickly identify adolescents at risk for legal problem behaviors. Those who endorse the three items are at risk for a deviant career since they share a similar profile with incarcerated youths. The practical procedure for the application of these results is as follows: (a) Administer the three items with a four-point format to the target groups, pointing out the aim of the screening study and underscoring the confidentiality of the responses (if this is unrealistic at the individual level, the analysis can be performed at the school class level based on anonymous data); (b) insert the three obtained raw scores into the logistic regression equations mentioned above for girls and boys separately; (c) consider all individuals with a probability above .50 to be at risk (other probability values can be chosen, of course, depending on the particular screening policy); (d) provide educational or psychological prevention programs, or apply a two-stage assessment strategy with a more elaborated diagnostic tool at the second stage of testing.

Alternatively, the more simple cut-off score method can be chosen based on the reliable three-item psychometric scale. Thresholds can be moved up and down to increase either sensitivity or specificity, whichever is desired more.

References

Alberts, J. K., Hecht, M. L., Miller-Rassulo, M., & Krizek, R. L. (1992). The communicative process of drug resistance among high school students. Adolescence, 27, 203-226.

Barnea, Z., Teichman, M., & Rahav, G. (1992). Personality, cognitive, and interpersonal factors in adolescent substance use: A longitudinal test of an integrative model. Journal of Youth and Adolescence, 21, 187-201.

Bentler, P. M. (1992). Etiologies and consequences of adolescent drug use: Implications for prevention. Journal of Addictive Diseases, 11, 47-61.

Biddle, B., Bank, B., & Marlin, M. (1980). Social determinants of adolescent drinking. Journal of Studies on Alcohol, 41, 215-241.

Botvin, G. J., Baker, E., Goldberg, C. J., Dusenbury, L., et. al. (1992). Correlates and predictors of smoking among black adolescents. Addictive Behaviors, 17, 97-103.

Central Registry of Drug Abuse (1993). Thirty-second report: January 1984 - June 1993. Government Secretariat (Narcotics Section): Hong Kong.

Dielman, T. E., Butchart, A. T., Shope, J. T., & Miller, M. (1990). Environmental correlates of adolescent substance use and misuse: Implications for prevention programs. The International Journal of the Addictions, 25, 855-880.

Dielman, T. E., Campanelli, P. C., Shope, J. T., & Butchart, A. T. (1987). Susceptibility to peer pressure, self-esteem, and health locus of control as correlates of adolescent substance use. Health Education Quarterly, 14, 207-221.

Donovan, J. E., & Jessor, R. (1978). Adolescent problem drinking: Psychosocial correlates in a national sample study. Journal of Studies on Alcohol, 39, 1506-1524.

Donovan, J. E., & Jessor, R. (1985). Structure of problem behavior in adolescence and young adulthood. Journal of Counseling and Consulting Psychology, 53, 905-912.

Ellickson, P. L., & Hays, R. D. (1992). On becoming involved with drugs: Modeling adolescent drug use over time. Health Psychology, 11, 377-385.

Ewing, J. A. (1984). Detecting alcoholism: The CAGE questionnaire. Journal of the American Medical Association, 252, 1905-1907.

Farrell, A. D., Danish, S. J., & Howard, C. W. (1992). Risk factors for drug use in urban adolescents: Identification and cross-validation. American Journal of Community Psychology, 20, 263-286.

Goddard, E. (1992). Why children start smoking. British Journal of Addiction, 87, 17-18.

Goldberg, D., & Williams, P. (1991). A user’s guide to the General Health Questionnaire (2nd ed.). Windsor, UK: NFER-Nelson.

Hansell, S., & Mechanic, D. (1990). Parent and peer effects on adolescent health behavior. In K. Hurrelmann & F. Lösel (Eds.), Health hazards in adolescence (pp. 43-66). Berlin, Germany: de Gruyter.

Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic regression analysis. New York: Wiley.

Iannotti, R. J., & Bush, P. J. (1992). Perceived vs. actual friends' use of alcohol, cigarettes, marijuana, and cocaine: Which has the most influence? Journal of Youth and Adolescence, 21, 375-389.

Jessor, R. (1987). Problem-behavior theory, psychosocial development, and adolescent problem drinking. British Journal of Addiction, 82, 331-342.

Jessor, R. (1993). Successful adolescent development among youth in high-risk settings. American Psychologist, 48, 117-126.

Kandel, D. B. (1980). Drug and drinking behavior among youth. Annual Review of Sociology, 6, 235-285.

McNeill, A. (1992). Why children start smoking: The need for a comprehensive tobacco control policy. British Journal of Addiction, 87, 24-25.

Ming Pao Newspaper, December 10, 1994.

Newcomb, M. D., & Felix-Ortiz, M. (1992). Multiple protective and risk factors for drug use and abuse: Cross-sectional and prospective findings. Journal of Personality and Social Psychology, 63 (2), 280-296.

Newcomb, M. D., Maddahian, E., Skager, R., & Bentler, P. M. (1987). Substance abuse and psychosocial risk factors among teenagers: Associations with sex, age, ethnicity, and type of school. American Journal of Drug and Alcohol Abuse, 13, 413-433.

Norusis, M. J. (1992). SPSS for Windows. Advanced statistics. Release 5. Chicago, IL: SPSS Inc.

Rooney, J. F. (1982). The influence of informal control sources upon adolescent alcohol use and problems. American Journal of Drug and Alcohol Abuse, 9, 233-245.

Rose, R. L., Bearden, W. O., & Teel, J. E. (1992). An attributional analysis of resistance to group pressure regarding illicit drug and alcohol consumption. Journal of Consumer Research, 19, 1-13.

Stacy, A. W., Newcomb, M. D., & Bentler, P. M. (1992). Interactive and higher-order effects of social influences on drug use. Journal of Health and Social Behavior, 33, 226-241.

Stacy, A. W., Newcomb, M. D., & Bentler, P. M. (1993). Cognitive motivation and sensation seeking as long-term predictors of drinking problems. Journal of Social and Clinical Psychology, 11, 1-21.

Stanton, W. R., & Silva, P. A. (1992). A longitudinal study of the influence of parents and friends on children's initiation of smoking. Journal of Applied Developmental Psychology, 13, 423-434.

Webster, R. A., Hunter, M., & Keats, J. A. (1994). Peer and parental influences on adolescents' substance use: A path analysis. The International Journal of the Addictions, 29, 647-657.

White, R. H., Labouvie, E. W., & Bates, M. E. (1985). The relationship between sensation seeking and delinquency: A longitudinal analysis. Journal of Research in Crime and Delinquency, 22, 197-211.

Willenbring, M. L., Christensen, K. J., Spring, W. D., & Rasmussen, R. (1987). Alcoholism screening in the elderly. Journal of the American Geriatrics Society, 35, 864-869.

Wills, T. A. (1986). Stress and coping in early adolescence: Relationships to substance use in urban school samples. Health Psychology, 5, 503-529.

Wills, T. A., Baker, E., & Botvin, G. J. (1989). Dimensions of assertiveness: Differential relationships to substance use in early adolescents. Journal of Consulting and Clinical Psychology, 57, 4, 473-478.

Wills, T. A., & Hirky, E. (1996). Coping and substance abuse: A theoretical model and review of the evidence. In M. Zeidner & N. S. Endler (Eds.), Handbook of coping: Theory, research, and application (pp. 279-302). New York: Wiley.

Wong, C. S. Y., Tang, C. S. K., & Schwarzer, R. (1997). Psychosocial correlates of substance use: Comparing high school students with incarcerated offenders in Hong Kong. Journal of Drug Education, 27(2), 147-172.

Yu, J., & Williford, W. R. (1992). The age of alcohol onset and alcohol, cigarette, and marijuana use patterns: An analysis of drug use progression of young adults in New York State. International Journal of the Addictions, 27, 1313-1323.

Zucker, R. A. (1979). Developmental aspects of drinking through the young adult years. In H. T. Blane & M. E. Chafetz (Eds.), Youth, alcohol, and social policy (pp. 91-146). New York: Plenum.

Zuckerman, M. (1979). Sensation seeking: Beyond the optimal level of arousal. Hillsdale, NJ: Erlbaum.

	Alberts, J. K., Hecht, M. L., Miller-Rassulo, M., & Krizek, R. L. (1992). The communicative process of drug resistance among high school students. Adolescence, 27, 203-226.
	Barnea, Z., Teichman, M., & Rahav, G. (1992). Personality, cognitive, and interpersonal factors in adolescent substance use: A longitudinal test of an integrative model. Journal of Youth and Adolescence, 21, 187-201.
	Bentler, P. M. (1992). Etiologies and consequences of adolescent drug use: Implications for prevention. Journal of Addictive Diseases, 11, 47-61.
	Biddle, B., Bank, B., & Marlin, M. (1980). Social determinants of adolescent drinking. Journal of Studies on Alcohol, 41, 215-241.
	Botvin, G. J., Baker, E., Goldberg, C. J., Dusenbury, L., et. al. (1992). Correlates and predictors of smoking among black adolescents. Addictive Behaviors, 17, 97-103.
	Central Registry of Drug Abuse (1993). Thirty-second report: January 1984 - June 1993. Government Secretariat (Narcotics Section): Hong Kong.
	Dielman, T. E., Butchart, A. T., Shope, J. T., & Miller, M. (1990). Environmental correlates of adolescent substance use and misuse: Implications for prevention programs. The International Journal of the Addictions, 25, 855-880.
	Dielman, T. E., Campanelli, P. C., Shope, J. T., & Butchart, A. T. (1987). Susceptibility to peer pressure, self-esteem, and health locus of control as correlates of adolescent substance use. Health Education Quarterly, 14, 207-221.
	Donovan, J. E., & Jessor, R. (1978). Adolescent problem drinking: Psychosocial correlates in a national sample study. Journal of Studies on Alcohol, 39, 1506-1524.
	Donovan, J. E., & Jessor, R. (1985). Structure of problem behavior in adolescence and young adulthood. Journal of Counseling and Consulting Psychology, 53, 905-912.
	Ellickson, P. L., & Hays, R. D. (1992). On becoming involved with drugs: Modeling adolescent drug use over time. Health Psychology, 11, 377-385.
	Ewing, J. A. (1984). Detecting alcoholism: The CAGE questionnaire. Journal of the American Medical Association, 252, 1905-1907.
	Farrell, A. D., Danish, S. J., & Howard, C. W. (1992). Risk factors for drug use in urban adolescents: Identification and cross-validation. American Journal of Community Psychology, 20, 263-286.
	Goddard, E. (1992). Why children start smoking. British Journal of Addiction, 87, 17-18.
	Goldberg, D., & Williams, P. (1991). A user’s guide to the General Health Questionnaire (2nd ed.). Windsor, UK: NFER-Nelson.
	Hansell, S., & Mechanic, D. (1990). Parent and peer effects on adolescent health behavior. In K. Hurrelmann & F. Lösel (Eds.), Health hazards in adolescence (pp. 43-66). Berlin, Germany: de Gruyter.
	Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic regression analysis. New York: Wiley.
	Iannotti, R. J., & Bush, P. J. (1992). Perceived vs. actual friends' use of alcohol, cigarettes, marijuana, and cocaine: Which has the most influence? Journal of Youth and Adolescence, 21, 375-389.
	Jessor, R. (1987). Problem-behavior theory, psychosocial development, and adolescent problem drinking. British Journal of Addiction, 82, 331-342.
	Jessor, R. (1993). Successful adolescent development among youth in high-risk settings. American Psychologist, 48, 117-126.
	Kandel, D. B. (1980). Drug and drinking behavior among youth. Annual Review of Sociology, 6, 235-285.
	McNeill, A. (1992). Why children start smoking: The need for a comprehensive tobacco control policy. British Journal of Addiction, 87, 24-25.
	Ming Pao Newspaper, December 10, 1994.
	Newcomb, M. D., & Felix-Ortiz, M. (1992). Multiple protective and risk factors for drug use and abuse: Cross-sectional and prospective findings. Journal of Personality and Social Psychology, 63 (2), 280-296.
	Newcomb, M. D., Maddahian, E., Skager, R., & Bentler, P. M. (1987). Substance abuse and psychosocial risk factors among teenagers: Associations with sex, age, ethnicity, and type of school. American Journal of Drug and Alcohol Abuse, 13, 413-433.
	Norusis, M. J. (1992). SPSS for Windows. Advanced statistics. Release 5. Chicago, IL: SPSS Inc.
	Rooney, J. F. (1982). The influence of informal control sources upon adolescent alcohol use and problems. American Journal of Drug and Alcohol Abuse, 9, 233-245.
	Rose, R. L., Bearden, W. O., & Teel, J. E. (1992). An attributional analysis of resistance to group pressure regarding illicit drug and alcohol consumption. Journal of Consumer Research, 19, 1-13.
	Stacy, A. W., Newcomb, M. D., & Bentler, P. M. (1992). Interactive and higher-order effects of social influences on drug use. Journal of Health and Social Behavior, 33, 226-241.
	Stacy, A. W., Newcomb, M. D., & Bentler, P. M. (1993). Cognitive motivation and sensation seeking as long-term predictors of drinking problems. Journal of Social and Clinical Psychology, 11, 1-21.
	Stanton, W. R., & Silva, P. A. (1992). A longitudinal study of the influence of parents and friends on children's initiation of smoking. Journal of Applied Developmental Psychology, 13, 423-434.
	Webster, R. A., Hunter, M., & Keats, J. A. (1994). Peer and parental influences on adolescents' substance use: A path analysis. The International Journal of the Addictions, 29, 647-657.
	White, R. H., Labouvie, E. W., & Bates, M. E. (1985). The relationship between sensation seeking and delinquency: A longitudinal analysis. Journal of Research in Crime and Delinquency, 22, 197-211.
	Willenbring, M. L., Christensen, K. J., Spring, W. D., & Rasmussen, R. (1987). Alcoholism screening in the elderly. Journal of the American Geriatrics Society, 35, 864-869.
	Wills, T. A. (1986). Stress and coping in early adolescence: Relationships to substance use in urban school samples. Health Psychology, 5, 503-529.
	Wills, T. A., Baker, E., & Botvin, G. J. (1989). Dimensions of assertiveness: Differential relationships to substance use in early adolescents. Journal of Consulting and Clinical Psychology, 57, 4, 473-478.
	Wills, T. A., & Hirky, E. (1996). Coping and substance abuse: A theoretical model and review of the evidence. In M. Zeidner & N. S. Endler (Eds.), Handbook of coping: Theory, research, and application (pp. 279-302). New York: Wiley.
	Wong, C. S. Y., Tang, C. S. K., & Schwarzer, R. (1997). Psychosocial correlates of substance use: Comparing high school students with incarcerated offenders in Hong Kong. Journal of Drug Education, 27(2), 147-172.
	Yu, J., & Williford, W. R. (1992). The age of alcohol onset and alcohol, cigarette, and marijuana use patterns: An analysis of drug use progression of young adults in New York State. International Journal of the Addictions, 27, 1313-1323.
	Zucker, R. A. (1979). Developmental aspects of drinking through the young adult years. In H. T. Blane & M. E. Chafetz (Eds.), Youth, alcohol, and social policy (pp. 91-146). New York: Plenum.
	Zuckerman, M. (1979). Sensation seeking: Beyond the optimal level of arousal. Hillsdale, NJ: Erlbaum.