Chinese University of Hong Kong
Freie Universität Berlin, Germany
Address correspondence to :
Prof. Dr. Ralf Schwarzer, Freie Universität Berlin , Institut fur Psychologie , Habelschwerdter Alle 45 ,14195 Berlin ,Germany
Running head: Adolescent substance use in Hong Kong
Drug use prevalence data were obtained from 969 adolescents, high school students and imprisoned offenders who reported use of cough medicine, organic solvents, cannabis, heroin, tranquilizers, and narcotics over the past six months. Incarcerated youths, in particular girls, had higher prevalence rates than students. Drug use frequencies were associated with psychosocial variables such as disinhibition, peer drug use, susceptibility to peer pressure, attitudes, encouragement by peers, and perceived availability of drugs. The psychosocial process of the initiation and maintenance of substance use was specified as a path model that considered personality and social environment as distal precursors and a drug-use predisposition and perceived availability as proximal precursors of three kinds of outcome variables: drug use, the intention to try illicit drugs if they were legal, and adverse outcomes of drug use. It was found that the same structural equation model fitted the data of both samples of offenders and students, however, with very different weights assigned to the pathes.
drug use, adolescents, delinquents, disinhibition, peer pressure, social influence, causal modeling
The use of illicit drugs among adolescents is a world-wide issue and many studies have tried to shed more light on the psychological causes and effects of this problem behavior (Baer, Marlatt, & McMahon, 1993; Maton & Zimmerman, 1992; Smith & Nutbeam, 1992; Vega, Gil, & Zimmerman, 1993; Zane & Sasao, 1992; Zimmerman & Maton, 1992). The present research contributes to this evidence by providing data from Hong Kong and from a particular subsample, namely incarcerated delinquents. Moreover, an attempt is made to explore the interplay of various factors that may jointly predict the frequency of drug use and the intention for further use.
Adolescent drug use is determined by various factors in the person and in the social environment. Research provides evidence that drug use is 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). Less attention has been paid to the interplay of these antecedents. Personality traits are remote from the actual behavior in specific situations. Thus, they represent rather distal predictors whereas other factors such as perceived social norms and perceived availability of drugs are more proximal. In the present research the causal order can only be tentatively explored because the study design does not allow a closer examination (see Dembo, Williams, Wothke, & Schmeidler, 1992a, 1992b; Tayman & Pennell, 1992).
According to the statistics released by the Hong Kong Central Registry of Drug Abuse (1992), among 2,904 newly reported addicts in 1992, 52 percent began abusing drugs before they reached the age of 20. Most of them started to use drugs when they were between 14 and 18 years of age. As compared with 1991, there was a marked increase of 34.6% of newly reported addicts aged under 21. Overall, the mean age of first illicit drug use decreased from 22.6 years in 1991 to 21.6 years in 1992. The proportion of newly reported young drug abusers increased from 30% in 1987 to 37.7% in 1991. The drugs most frequently abused by newly reported young persons in 1991 were Heroin (49.4%), followed by Cannabis (24.7%), Cough Medicine (22.7%), Brotizolam (3.5%), Organic Solvents (2.6%) and Flunitrazepam (2.0%). In another survey on the non-medical use of psychotropic substances among secondary school students in Hong Kong, it was revealed that 2.1% of the secondary students in Chinese-speaking schools had abused psych-substances, compared to 1.1% in 1987 (Central Registry of Drug Abuse, 1992). Another change in HK has been the increasing recognition of alcohol and tobacco use, particularly in the young. Approximately one third of 14 year-old youths surveyed by the Community Drug Advisory Council in 1990 reported taking alcohol more than once per month. Further there had been a considerable increase in the regular use of tobacco by those 16 and 17 years old, between 1988 and 1990 (from 14% for each group to 22% and 19% respectively) (C.D.A.C., 1991). -->
There are some personality differences between drug users and abstinent individuals (Hammersley, Lavelle, & Forsyth, 1992a; 1992b; Newcomb & Bentler, 1988a, 1988b). Those who experiment with illicit substances have been found, for example, to be less inhibited (Watson & Clark, 1993; Zuckerman, 1979). Although some of the individual differences can be tracked down to childhood and may even have some genetic predisposition, it is generally assumed that the 'drug-prone personality' is a result of a socialization process that leads into a deviant career. Thus, certain personality characteristics may stem from the same pattern of influence that also triggers drug-related behaviors more directly. This makes it hard to disentangle personal from situational determinants of problem behaviors. In distinguishing between personality and social situation, however, a merely heuristic template is provided to guide research into more or less distinguishable sources of deviant behaviors.
Social influence is commonly regarded as the most conspicuous factor in the development of drug taking. Adolescents face a number of developmental tasks to prove themselves in their reference groups (Jessor, 1987, 1993; Yu & Williford, 1992). If the dominant reference group mainly consists of peers who value drug use as an 'adult behavior' or as indicators of maturity and independence from parents, then the individual may attempt to meet these expectations. There may be role conflict, for example, when the family norms contradict the peer group norms, and it may require considerable coping efforts until the youngster eventually complies with one or the other (Flay et al., 1994; Webster, Hunter, & Keats, 1994; Wills & Hirky, in press). In the following section, several psychosocial antecedents of drug use are discussed in more detail.
Disinhibition. Some individuals are more likely to violate the moral or legal rules of the society in order to pursue a pleasure-oriented lifestyle. They attend to their feelings here and now without contemplating future risks for themselves or others. These individuals display a 'behavioral disinhibition' as opposed to those who are more 'constrained' or self-controlled (Watson & Clark, 1993). The disinhibition construct can be conceived of as a stable and continuous personality trait with the two extremes of behavioral disinhibition and constraint. The former reflects a spontaneous and impulsive approach to life, neglecting long-term consequences of risky actions. The latter describes persons who feel bound to the rules and expectancies of their culture and who rather tend to delay gratifications to attain desirable long-term goals or social rewards. They are more dependable and self-disciplined by committing themselves to generally accepted standards and by avoiding dangerous situations. This personality dichotomy is similar to the notion of being overcontrolled versus undercontrolled.
The disinhibition construct has been mentioned in the literature on psychopathology (Hare, 1970) and has been conceived as being part of the broader personality disposition of sensation seeking. This sensation seeking trait has been defined by Zuckerman (1971, 1979, 1991) as a need for varied, novel, and complex sensations and experiences and the willingness to take physical and social risks for the sake of such experiences. Zuckerman's (1979) Sensation Seeking Scale includes the four subscales of Experience Seeking, Thrill/Adventure Seeking, Disinhibition, and Boredom Susceptibility. It was typically the most powerful predictor of substance use and abuse out of five self-report measures in predicting drug use (Jaffe & Archer, 1987). All four Sensation Seeking subscales were positively correlated with licit and illicit drug use for female subjects, and all the subscales except Experience Seeking correlated with licit drug use for male subjects, whereas Disinhibition was associated with illicit drug use for male subjects (Newcomb & McGee, 1991). Barnea, Teichman and Rahav (1992, 1993) found that sensation seeking was the only personality trait that predicted substance use. It had a strong direct as well as indirect effect on drug use, on behavioral intentions, and on attitudes toward drugs. Stacy, Newcomb and Bentler (1993) found that sensation seeking predicted alcohol use over a nine-year period.
In a study to validate a new disinhibition scale, the following correlations with other variables emerged: impulsivity (.68), irresponsibility (.63), risk taking (.59), low persistence (.56), playfulness (.54), norm rejection (.49), danger seeking (.47), and experience seeking (.47) (Watson & Clark, 1993, p. 512). In a study with 901 university students, the following correlations between disinhibition and drug use were obtained: alcohol (.44), marijuana (.33), cigarettes (.29), psychedelics (.26), narcotics (.19), and tranquillizers (.12) (Watson & Clark, 1993, p. 517). There was also a correlation of .37 with number of sex partners in the past year and, moreover, negative associations with academic achievement emerged. Boys scored generally higher on disinhibition than girls.
Peer drug use. Zucker (1979) considered peer influence to be a particularly potent extra-individual determinant of the transition to drinker status, because drug using peers are both models for drug-taking behavior and sources of availability of the substance. Kandel (1980) concluded that peer related factors, e.g., extent of perceived drug use in the peer group, self-reported drug use by peers, and perceived tolerance for use, are consistently the strongest predictors of subsequent alcohol and marijuana use, even when other factors are controlled. Barnea et al. (1992) found that peer influence on adolescent use is even more significant than that of parents. Friends' smoking has been found to exert a stronger effect on adolescents' smoking than parents'smoking, in particular on initiation (Flay et al., 1994). Peers' drug use behavior appears to have a direct effect both on adolescents' drug use and on their behavioral intentions (Wills et al., in press). 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).
Perceived social norms toward drug use. 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. White, Labouvie and Bates (1985) found friends' norms to be more highly correlated with the adolescents' substance use than friends' substance use. These peer norms, generally measured in terms of peer approval of drug use, are positively correlated with peer drug use. In most studies which assessed both peer norms and peer drug use, the latter was a more salient predictor than the former. The effect of peer behavior on the students' alcohol use, however, was stronger than the effect of peer norms. Rooney (1982) also found 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 factor out of 12 risk factors that were associated 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 more frequently exposed to smoking models, or they may be more likely to get into situations where peer pressure for substance use is prevailing. In addition, it has been suggested that more socially oriented students may be more sensitive to social evaluations and hence more susceptible to conformity or social image pressures (Botvin & McAlister, 1981).
Attitudes toward drug use. Attitudes have been hypothesized to influence experimentation with drugs and continued substance use (e.g, Kandel et al., 1978; Krosnick & Judd, 1982). Attitudes, however, are often ill-defined because researchers do not agree upon the components such as affective, cognitive, and conative ones. Fishbein and Ajzen (1975) conceived attitudes as one's affective evaluation of a specific behavior, and redefine the cognitive component as 'behavioral beliefs' that precede the attitude. The prediction of health behaviors by attitudes has yielded mixed results. It has been questioned, for example, whether attitudes were a result of behaviors or vice versa. In a recent longitudinal study of six health behaviors, a cross-lagged panel design was employed to examine the causal predominance ot the two variables over time (Stacy, Bentler, & Flay, 1994). It was found that attitudes exerted a weak influence on alcohol consumption and marijuana use but contradictory or non-significant results emerged for the other four health behaviors. The authors conclude that the attitude construct may be of only limited value compared to alternative predictors within an expectancy framework (see also Schwarzer, 1992, 1994). It was further suggested that attitudes may be of indirect influence, for example, by predicting behavioral intentions that, in turn, would partly determine actual behaviors (Barnea et al., 1992; Ajzen, 1988). Since attitudes are developed during one's socialization process within reference groups they may be closely related to perceived social norms and may not be empirically distinct from them.
Susceptibility to Peer Pressure. The same may be true for susceptibility to peer pressure. Although theoretically a unique construct it can be empirically confounded with the previous two constructs because it also develops during the same socialization process. Dielman, Butchart, Shope and Miller (1990) confirmed earlier studies, indicating that peer alcohol use and peer norms, 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 adolescent alcohol use. Dielman, Campanelli, Shope and Butchart (1987) found that susceptibility to peer pressure was more highly correlated with substance use 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 (e.g., immoderate alcohol or drug use) 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 initiation of drug-use (see also Alberts, Hecht, Miller-Rassulo, & Krizek, 1992; Rose, Bearden, & Teel, 1992).
It was found that initiation into drug use of any substance is preceded by values favorable to its use (e.g, Kandel et al., 1978; Krosnick & Judd, 1982). Barnea et al (1992)'s research indicated that although attitudes of adolescents toward drugs have no direct influence on their drug use, attitudes affect use indirectly, through behavioral intentions. Favorable attitudes increase behavioral intentions which in turn affect the probability of use.
Intentions to use drugs. More proximal cognitive variables such as behavioral intentions are assumed to mediate between rather distal psychosocial antecedents and subsequent behaviors, as stated in the Theory of Reasoned Action (Fishbein & Ajzen, 1975). This theory has been applied with some success in studies on alcohol consumption of adolescents (e.g., Schlegel, Crawford, & Sanborn, 1977). Thus, an intention can be seen as an independent variable that predicts drug-related behavior, but it is also a dependent variable of a set of earlier antecedents. In adopting a process view of substance use, the development of an intention concludes a decision process that is mainly influenced by social and personal dispositions. Whether this intention later translates into actual behavior depends on further circumstances such as social influence or self-regulatory attempts (cf. Schwarzer, 1992).
The exploration of the interplay of drug use antecedents is the main research question. In addition, it is also of interest to examine the prevalence rate of drug use of girls and boys in different educational settings or correctional institutions and to find out whether differences in drug use are reflected by similar differences in antecedent variables. The present sample includes high school students and delinquents who are incarcerated for various reasons. It is assumed that the latter are deviant in a broad sense and more inclined to take drugs or to have taken drugs before imprisonment. Thus, it is expected that there is a deviant subculture with high illicit drug use as opposed to a normal school population with low substance use. It is examined whether these groups differ in personality and social influence variables. Moreover, it could be possible that within these reference groups different patterns of associations exist that point to different pathways in the development of substance use. Adolescents who are socialized in deviant subcultures might become deviant in terms of problem behaviors as well as in terms of attitudes and social norms. Thus, different patterns of influence within the two groups of teenagers are hypothesized.
Questionnaires were given to 1,001 Chinese adolescents. Due to obvious response sets or mainly blank forms , thirty two (3.1%) were discarded. Among the remaining 969 participants, more than half were normal secondary school students attending Form 2 to Form 6 in two schools (Table 1). One of the schools was classified by the Education Department as around the "Band Four" to "Band Five" school, i.e., students' academic performance was among the poorest in Hong Kong. Another was a "Band Two" school, where students' academic performance was about average to high average. The second sample consisted of incarcerated delinquents. Four facilities of the Correctional Services Department were approached. Almost all of the young offenders in these facilities, who were eligible to read and had about Primary 6 education level, participated in the data collection. Three Correctional Homes of the Social Welfare Department were also approached.Written consent from both participants and their parents was obtained. The response rate was 36%
Insert Table 1 about here
As shown in Table 1, 59.8% (N=579) of the participants were school students and 40.2% (N=390) were incarcerated delinquents. Among the school students, 39.2% (N=277) were boys and 60.8% were girls (N=352). In the delinquent sample, 77.4% (N=302) were boys and 22.6% (N=88) were girls. Chi-square analysis confirmed significant differences. 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 whereas in the delinquent sample, the overall mean age was 17.32 years (SD=1.75). On average, delinquent boys were 17.28 years, and girls were 17.43 years of age. There was no sex by institution interaction (F[1,961]=.73, p>.05), or main effect of sex (F([,961].000, p>.05) on age. However, age differences were significant between the school and delinquent teenagers (F[1,961]=136.54, p<.001).
The following nine measures were part of a questionnaire given in Chinese to the research participants.
Drug Use Behavior. Frequency of drug use measures were adapted from Segal (1990). They were obtained for the following illicit substances: cough medicine, organic solvents, marijuana, heroin, tranquillizers, and narcotics. Frequency of drug use over the preceding six months was rated on 7-point scales anchored with Never (1), Less than once a month (2), Two or three times a month (3), Once a week (4), 2-5 times a week (5), Once a day (6), More than once a day (7).
Peer Drug Use. Following Stein, Newcomb and Bentler's (1987) format in assessing peer drug use, participants were asked three items on how many of their good friends used illicit substances at least once a month on a 5-point scale that ranged from None (1), One to Two of them (2), Some of them (3), Most of them (4) to All (5) :(1) How many of your good friends used cigarettes or alcohol at least once a month? 2) How many of your good friends used heroin at least once a month? (3) How many of your good friends used other illicit substances at least once a month? The peer drug use scale was a composite of the three items (Cronbach's a = .85).
Disinhibition. As a measure of the personality trait of disinhibition, a four-item subscale was adapted from Zuckerman's (1971, 1979) Sensation Seeking Scale (see Huba, Newcomb, & Bentler, 1981): (1) I like wild "uninhibited" parties. 2)I enjoy the company of real "swingers". (3) I could conceive of myself seeking pleasures around the world with the "jet set". (4) I enjoy watching many of the "sexy" scenes in movies. Participants answered on a four-point scale ranging from Strongly Disagree (1) to Strongly Agree (4). A high score on the composite scale indicated a high level of disinhibition (Cronbach's a = .51).
Perceived availability of drugs. Participants' perceived availability of drugs was assessed by three items on how easy they could obtain illicit substances, based on a four-point scale ranging from Very Easy (1) to Very Difficult (4). (1) How far do you find it easy to obtain cigarettes or alcohol? (2) How far do you find it easy to obtain heroin? (3) How far do you find it easy to obtain other illicit substances? A high score referred to a higher perceived availability or easiness in obtaining these substances (Cronbach's a = .84).
Intention to use drugs if they were licit. Intention to use drugs was assessed by two items on how likely participants would use heroin and other illicit substances if they became licit, based on a five-point scale ranging from Definitely Will (1) to Definitely Will Not (5). The Intention to Use Drugs scale was computed after reverse-scoring the sum of these two items. High scores indicated a higher likelihood to experiment with drugs if they were licit.
Adverse Consequences of Drug Use. Participants were asked four items on how often they had been negatively affected by drug use over the preceding six months (1) How far did your using alcohol or other illicit substances affect your working/ academic performance? (2)How far did your using alcohol or other illicit substances affect your family life? (3) How far did your using alcohol or other illicit substances affect your social or recreational activities? (4) How far did your using alcohol or other illicit substances affect situations in which use is physically hazardous, e.g., driving while being intoxicated? The total score on these four items was the Adverse Consequences scale, with a high score indicaing the experience of more negative consequences of drug use (Cronbach's a = .91).
Attitude Toward Drug Taking. The Attitude Toward Drug Taking scale constructed by Ong (1988) was adapted. Only those items that were more relevant to the Hong Kong situation were selected, resulting in eight items (e.g., 'Sometimes I feel like taking some illegal drugs to stimulate myself'). Three additional items on the drug's effect on productivity, one's future, and one's determination to resist dependency were constructed (e.g., 'Illegal drugs can improve my working performance'). All items were measured on a four-point scale ranging from Strongly Disagree (1) to Strongly Agree (4). This 11-item scale aimed at measuring a person's attitude toward illicit drug use, their legalization, and their effects. High scores indicated a favorable attitude toward using unlawful drugs (Cronbach's a = .92).
Susceptibility to peer pressure. An eight-item scale on perceived susceptibility to peer pressure by Dielman et al. (1987) was used (e.g., 'If a friend offers you a drink of alcohol, would you drink it?'). Participants were asked how far they agreed on a four-point scale ranging from Strongly Disagree (1) to Strongly Agree (4). The higher the score, the more susceptible to peer pressure one is supposed to feel (Cronbach's a = .80).
Peer Encouragement. For the peer encouragement on drug use, participants were asked three items indicating to what extent their good friends encouraged them to smoke or drink, take heroin and other illicit substances on a four-point scale ranging from Strongly Encouraged (1) to Strongly Discouraged (4) (e.g., 'How far do your friends encourage you to take heroin?'). The peer encouragement scale is computed after reverse-scoring of the sum of these three items (Cronbach's a = .84).
Peer Attitudes Towards Drugs. With respect to the peer attitude on drug use. participants were asked three items on how far their good friends approve adolescent smoking or drinking, taking heroin and other illicit substances on a four-point scale ranging from Strongly Disapprove (1) to Strongly Approve (4) (e.g., 'How far do your good friends approve your taking heroin?'). (Cronbach's a = .84)
The data collection took place from November to February. Self-report questionnaires written in Chinese were administered in group format by reading a standardized instruction. The first author collected the data of the incarcerated participants in a face-to-face situation. For the normal secondary school sample, the questionnaires were administered to the class during their weekly assembly meetings or free lessons. The instruction included a brief description of the study and reminded subjects of the anonymous nature of the questionnaire. Participants were assured that their responses would be completely confidential and would not be revealed to parents, or school or institution personnel. In addition, it was emphasized that the drug use items pertained to the non-medical use of substances only (e.g., cough medicine). On average, completion of the questionnaire took about 30 minutes.
Frequency of drug use data were broken down by gender and institution. Means and standard deviations for the four subgroups are displayed in Table 2.
Insert Table 2 about here
By two-factorial analyses of variance with gender and institution as factors, main effects for gender were found on all uses of substances except of organic solvents. In general, girls reported higher use of substances than boys. Significant main effects for institution were also found. Delinquents used more substances of any kind than the school students. Significant interaction effects were found in two of the reported drug use measures, namely organic solvents (F[1,950]=27.54, p < .01) and cough medicine (F[1,953]=3.99, p < .05). Girls in the delinquent sample used more organic solvents (F[1,939]=4.59, p<.05) than their male counterparts. On the other hand, boys in the delinquent sample and girls in the school sample used more cough medicine than their opposite sex counterparts.
Data on psychosocial variables were also broken down by gender and institution. Means and standard deviations for the four subgroups are displayed in Table 3.
Insert Table 3 about here
In a two-way ANOVA, main effects of gender and institution emerged. There were significant gender differences on all psychosocial measures. Boys were higher than girls in susceptibility to peer pressure, attitude toward drug taking, disinhibition, perceived availability of drugs, intention to use drugs, peers' drug use and attitudes, as well as consequences of drug taking. In addition to the gender effects, the delinquent and school samples differed significantly on all psychosocial scales. Delinquents had more favorable attitudes toward drug taking, were more susceptible to peer pressure, less inhibited, had more perceived availability of drugs, greater intention to use drugs if they were licit, more friends using drugs, more favorable peers' attitude toward drug taking as well as more adverse consequences of drug taking. An interaction was found for perceived availability of drugs (F[1,923]= 4.82, p<.05). Girls in the delinquent sample had a slight tendency to score higher than boys whereas school boys obtained significantly higher scores than school girls (t  = 3.4, p < .01).
Frequencies of illicit drug use were significantly correlated with all psychosocial variables (Table 4).
Insert Table 4 about here
Use of cough medicine was most closely associated with attitudes (r = .56), intentions to try (.53), and peer drug use (.53). For organic solvents, most correlations were moderate (r ranged from .24 to .38). Cannabis and heroin use were most closely associated with peer drug use, intentions, attitudes, and perceived availability of drugs (r ranged from .56 to .64). Use of tranquillizers, intentions (.57), attitudes (.57), and peer drug use (.58) were closely interrelated. For narcotics, the correlations were a bit lower, with intentions to try as the closest psychosocial variable (r ranged from .22 to .47).
Disinhibition, peer drug use, perceived availability, and intention to try were closely linked to cannabis and heroin use. The latter two can be seen as the typical illicit drugs in today's youth subcultures where peer models seduce adolescents to comply with their norms. On the other hand, the self-reported drug use consequences are the worst for these two substances as is reflected by the high correlations in the last row of Table 4.
The preceding section has focussed on the correlation of reported drug use behaviors with a number of psychosocial variables. In the following section, a structural model is specified that may account for the complex relationships among the factors that influence drug use. The basic assumption is that there are various antecedents and consequences of drug use, each of which can be represented in a more or less "causal" order. According to the model, some factors are considered to be more distal, others to be more proximal to the dependent variable(s).
Model Specification. The model was composed of seven latent variables (factors), each of which was qualified by a number of observed variables (indicators), i.e., there were seven measurement models on top of which a structural model with latent variables was specified. . Figure 1 displays the complete structural model.
Insert Figure 1 about here
The two factors at the far left are the exogenous latent variables disinhibition and peer drug use. It is obvious that personality as well as social context play a role in the motivation to act and in the actual behavior itself, across a wide range of human functioning, by setting the stage for behavioral cognitions and emotions. As a potential personality dimension relevant to drug use, disinhibition was chosen. The four items of the disinhibition scale described above served as indicators. As a social context factor, the reported peer drug use was seen as appropriate. The three items served as indicators. (All indicators in this section are the same as described in detail in the Method section).
On the right hand side of Figure 1, there are five endogenous latent variables. Starting far right, the most 'dependent' factor was specified to be the adverse consequences of drug use. Participants were asked what effects of drug use occurred to them during the last six months, after they had experience with drugs. Four items were used as indicators.
A logical precursor of these adverse effects is the actual frequency of drug use. Respondents were asked how often they had taken certain substances over the preceding six months. For the composition of this factor, reported frequencies of the six illicit substances were used: cough Medicine, organic solvents, marijuana, heroin, tranquillizers, and narcotics.
Behavior is usually guided by intentions. In the present study, behavioral intentions to use drugs have not been directly measured because the drug use behavior was reported retrospectively for the last six-months period, and thus, intentions were no longer appropriate. Instead, it was asked whether participants would be willing to try either (a) heroin or (b) "pills", cannabis, cough medicine or organic solvents, in case they were legal. The intention to try was specified as a latent outcome variable based on these two indicators.
The three factors intention to try, reported frequency of drug use, and perceived consequences of drug use can be seen as the most "dependent" variables whereas a social-cognitive mindset and the perceived behavioral control can be seen as mediators. In other words, the latter represent proximal antecedents of drug-related outcomes while personality (disinhibition) and social context (peer drug use) represent more distal antecedents.
The factor social-cognitive drug-use predisposition has been composed of the sum scores of four psychometric scales, namely Attitude Toward Drug Taking, Susceptibility to Peer Pressure, Peer Encouragement Toward Drug Taking, and Peer Attitude Toward Drug Taking. These scales were very closely related to each other and seem to reflect the representation of social norms and pressures of one's peer group in terms of deviant behaviors and, as a consequence, the readiness to act accordingly.
The factor perceived availability of drugs reflects the phenomenon that some individuals can make drugs easily available to themselves whereas others have difficulty obtaining them. It is composed of three indicators at the item level.
Central to the present model is the mediator hypothesis. It is assumed that the three drug-related outcome factors result from a psychological process that operates through the establishment of a a social-cognitive mindset as well as the perception of control over drugs. These psychological process factors should be dependent on social influence such as peer drug use and personality dispositions such as lack of inhibition. The research question was whether this model would fit the data and to what degree the parameter estimates would differ in the two subsamples of normal and delinquent adolescents.
Model Testing. The model was tested with the LISREL 8 program (Jöreskog & Sörbom, 1993). Input was a correlation matrix of the 26 observed variables. Accordingly, the parameters were estimated by the unweighted least squares method. The model fit was evaluated in terms of chi-square, root mean square residuals, and various goodness of fit indices. Since chi-square is dependent on sample size, it was only considered in hierarchical tests at different stages of the model specification process but not in the final evaluation. The chi-square divided by the degrees of freedom, may be seen as a less biased fit estimate (2/df). This quotient should be small, and values below 3 are usually considered to be satisfactory (Bentler, 1980; Bollen, 1989; Bollen & Long, 1993; Byrne, 1990). The root mean square residual should be very small with values of .05 being excellent. The goodness of fit index (GFI) should be close to 1.00. The same applies to the adjusted GFI (AGFI; adjusted for degrees of freedom). These are rough fit indicators only. A more comprehensive assessment includes further fit indices as well as a careful inspection of parameter estimates, accounted variance, and modification indices.
The structural model was tested separately in the two subsamples of normal and delinquent adolescents. An overall test collapsing both samples was too unlikely to fit well because of substantially different correlations in these groups. Rather, it was of interest to identify differences in parameter estimates to explore the unique dynamics of interrelationships within both samples. First, the delinquent sample was examined. Input was a matrix based on 332 adolescents incarcerated for drug use who had complete data (listwise deletion of missing values). The following fit indices were obtained: 2(288) = 826.52, 2 /df= 2.87, RMR=.084, GFI=.93, and AGFI=.92. This reflects a satisfactory model fit. The parameter estimates of the measurement model (factor loadings) are presented in Table 5.
Insert Table 5 about here
The parameter estimates of the structural model is displayed in Figure 2. Of the variation in adverse consequences of drug use, 33% (1 minus the residual) have been accounted for by drug use. Drug use itself could be well explained jointly by the drug-use predisposition and perceived availability (58%). The latter two proximal predictors accounted for 72% of the intention variance. The two distal precursors, disinhibition and peer drug use, accounted for 71% of the predisposition variance and for 32% of the perceived availability variance.
Insert Figure 2 about here
Second, the school sample was examined. Input was a matrix based on 494 adolescents in high school who had complete data (listwise deletion of missing values).The following fit indices were obtained: 2 (288) = 814.28, 2/df= 2.83, RMR=.069, GFI=.95, and AGFI=.94. This reflects a slightly better model fit than the one for the delinquent sample. The parameter estimates of the measurement model (factor loadings) are reported in Table 6.
Insert Table 6 about here
The parameter estimates of the structural model are represented in Figure 3. The variation in adverse consequences of drug use remained unexplained (0%). Drug use could only be explained by the drug-use predisposition (11%). The latter proximal predictor accounted for 32% of the intention variance. The two distal precursors disinhibition and peer drug use accounted for 71% of the predisposition variance and 36% of the perceived availability variance.
Insert Figure 3 about here
Statistical Two-Group Comparison. The proposed structural model fitted in both samples separately which indicates that the structure of relationships holds across different groups. Obviously, the parameter estimates differ from sample to sample. To test the inequality of these estimates, a two-group comparison was specified with the constraint that all relationships were set invariant from one group to the other. The following fit indices were obtained: Chi-Square (638) = 3236, Chi-Square/df=5.07, RMR=.092, GFI=.91. This is clearly less satisfactory than the results for the two individual models and, thus, confirms the distinctiveness of the two groups in terms of parameter estimates. In general, the operating mechanisms seem to be the same in both kinds of samples but some of the involved constructs are of different importance as indicated by the weights given to the pathes.
The present study has surveyed the illicit drug use prevalence of adolescents in Hong Kong schools and correctional institutions and further explored the role that psychosocial antecedents may play in the initiation or maintenance of such risk behaviors. For six kinds of illicit drugs, gender differences and institution differences were found.
It was counterintuitive to detect higher levels of drug use in girls (cf. Pascale & Evans, 1993). This was mainly due to those girls who were incarcerated. They represent an unusual group of multiple offenders with severe behavior problems. Girls are less likely to become imprisoned than boys. Usually they are first put on probation, and only after repeated violation of legal norms they are referred to corrective institutions. These girls are typically characterized by multiple problem behaviors. Anecdotal evidence suggests that many of them have held jobs of prostitutes or waitresses in the red-light districts where they were exposed to various deviant behaviors including polysubstance use. It might also play a role that they earned enough in this kind of business to be able to afford drugs.
There was also a tendency for girls in the student sample to report somewhat more illicit drug use than boys. It could be that girls were more 'mature' in the sense that some of them socialized with older boys or men from deviant reference groups. The present data set does not allow, however, a more detailed analysis of the gender issue. It did not come as a surprise that the delinquents reported higher drug use of all kinds than the students. Delinquents were imprisoned for diverse problem behaviors that may have included dealing with or abusing substances.
More interesting for psychological research is to look beyond prevalence data and to identify psychosocial variables that are associated with risk behaviors. It turned out that all psychosocial variables, no matter whether being more dispositional or social, were moderately to highly associated with drug use. It is tempting to conclude that these personality and social influence factors represent causal precursors, and this is why they have been labeled 'antecedents' in this paper. This conclusion, however, is less based on the present empirical evidence but more so on theoretical assumptions. The opposite causal pathway cannot be ruled out, i.e., that drug use would create attitudes, social norms, dispositions, etc. that are in line with one's deviant behaviors, a hypothesis in accord with Self-Perception Theory (Bem, 1972). However, this seems to be less likely. Disinhibition, for example, is commonly regarded as being a partly genetic temperament factor (Watson & Clark, 1993; Zuckerman, 1971, 1979, 1991), and peer drug use is normally understood as a major social source of behavioral change, e.g., as suggested by the modeling paradigm of Social Cognitive Theory (Bandura, 1992; Bochner, 1994; Webster, Hunter, & Keats, 1994).
In line with these assumptions, a complex structural model has been specified that employs these two constructs as the most distal precursors of three latent outcome variables: intentions, drug use, and adverse consequences of drug use. The two precursors have an unbiased intercorrelation of .33 in the delinquents' sample and .42 in the students' sample. The more disinhibited the adolescents are, the more they socialize with deviant reference groups where substance use is prevalent. It remains a question for further longitudinal research to examine the causality within these relationships.
The two distal precursors are seen to create a 'general drug-use predisposition', a more proximal social-cognitive mindset that consists of attitudes, susceptibility to peer pressure, perceived peer encouragement, and perceived peer attitudes. These four variables were so highly intercorrelated in the present data set that they represent one common factor. Whether this would be the case with different operationalizions or in different populations remains to be investigated. This drug-use predisposition is similarly strongly associated with peer drug use and disinhibition in both samples. Adolescents with high levels of behavioral disinhibition who socialize with drug using peers are likely to develop a readiness towards using substances themselves. This accounted for 71% of the variance of this predisposition within both samples. In particular, the intention to try further substances if they were licit, is strongly associated with this mindset. This accounted for 72% of the variance in the delinquent group and for 32% in the student group.
A behavioral intention is usually one of the best predictors of an action (Ajzen, 1988). In the present study, the intention does not refer directly to drug use behavior but to the hypothetical scenario that drugs might become licit. Therefore, it is an outcome variable in its own right and not merely a precursor of substance use. Frequency of drug use had been measured retrospectively for a six-months interval which is another reason why it has been specified without relation to future intentions.
Frequency of drug use is very closely related to the social-cognitive drug use predisposition in the delinquent sample, less so in the student sample. Furthermore, perceived availability of drugs is a predictor in the delinquent sample and not at all in the student sample. This is obviously a result of the low prevalence rate of substance use in the student sample.
Perceived availability did not turn out to be an important mediator. It was, however, closely related to peer drug use. Social influence is not merely seen as a psychological process of persuasion or peer pressure. It also includes real opportunity structures and the material availability of substances.
Another reflection of the different prevalence rates was found in the path leading from drug use to its adverse consequences. There was a strong effect in the delinquent sample but no effect at all in the student sample. The incarcerated drug users may have realized that they suffer from their problem behavior whereas the students did not perceive adverse life changes since they only experiment with very small amounts of substances. In the correlation analyses it was found that mainly cannabis and heroin use were followed by distress or performance deficits.
In sum, there is evidence that at least two of the three outcome constructs (intention and drug use) are well predictable by peer drug use and disinhibition as distal precursors jointly with a social-cognitive drug-use predisposition as a proximal precursor (or mediator). This was valid across two very different groups of adolescents. Moreover, there are additional features for the incarcerated youths by including effects from perceived availability and by linking adverse consequences to the frequency of substance use.
This study has made a contribution mainly by examining two samples of adolescents within a unique culture in terms of drug use prevalence and psychosocial antecedents. It hopefully stimulates further research to model the pattern of influence that leads to the initiation and maintenance of deviant problem behaviors. But there are some limitations in this research that have to be pointed out to interpret the results with caution. Firstly, this is a cross-sectional study that does not allow to make any causal inferences. The structural model can be seen as a heuristic for other studies but does not represent a test of a causal order. Longitudinal studies and those including interventions would be desirable. Secondly, the data are self-report data. There is no validation of drug use by physiological methods, observation, or peer ratings (cf. Anglin, Hser, & Chou, 1993; Belkin & Miller, 1992; Skog, 1992; Wagenaar et al., 1993). This is difficult to achieve, in particular with such large samples, without compromising anonymity. Research designs with higher internal validity are needed to complement this kind of large-scale field study. Thirdly, it might be considered to use alternative measures to those employed here to see whether the operationalization affects the results. Fourth, additional constructs should be included that have been proven useful in research on risk behaviors such as self-efficacy (Bandura, 1992; Schwarzer, 1992, 1994).
It is premature to draw practical implications from the present findings but it endorses the commonly held view that adolescents need to be trained in how to resist social influence. This should not be pursued by 'fear appeals' or 'information appeals' but rather by the systematic development of 'resistance self-efficacy' (Marlatt, Baer, & Quigley, 1995). Risk communication and the adoption of more realistic drug expectancies may set the stage for a more stable motivation to refrain from substances in favor of alternative actions that result in peer approval. It may be added, that disinhibition, although having presumably a genetic component, need not be taken for granted. Instead, it can be seen as an 'alterable variable'. Individuals become more constrained with increasing age and maturity (Watson & Clark, 1993; Zuckerman, 1971, 1979). This natural process can be promoted with the appropriate educational messages. It is, by the way, an unresolved issue whether prevention measures should focus on complete abstinence, for example, in promoting 'peer resistance self-efficacy', or whether it should acknowledge the pervasiveness and inevitability of drug use and focus rather on damage control, for example, by promoting 'recovery self-efficacy' after a drug has been tried out (Marlatt et al., 1995). The former approach typically results in a later onset of the problem behavior whereas the latter allows youngsters to have the experience and then to overcome it.
LISREL Parameter Estimates for the Structural Model in the Delinquent Sample
LISREL Parameter Estimates for the Structural Model in the School Sample