Introduction

There does not seem to be a country in the world that does not experience the problem of students leaving school before attaining that country’s equivalent of a high school diploma. In developing countries a comparison of the gross enrolment ratios (GER’s) Footnote 1 in primary and secondary education suggests that just over 40% of children enrolled in primary education, drop out before proceeding to secondary schooling (EFA Global Monitoring Report 2003/4). Although the GER’s in developed countries are more favourable, high school dropout is still evident. Lahey (2003) cites non-completion rates of 23% in Australia, 9% in Germany, 12% in the United States and Canada, and 6% in Korea. The dropout rate in Norway is 3% among junior high school students, and 10.8% among those who enter the voluntary high school (Wichstrøm 1998). About 55% of high school students in Cape Town, South Africa dropped out before completing their schooling (Flisher et al. 2004).

It is widely recognised that a number of adverse consequences exist for those students who fail to complete their schooling. Governments and the international community view education as fundamental to reducing world poverty and promoting a more equitable, peaceful and sustainable future for all (EFA Global Monitoring Report 2002, 2003/4). In South Africa, where poverty is endemic, and social and economic inequities persist, educational achievement would perhaps go a long way in addressing these disparities.

Given the long-term negative costs, whether to remain in school or whether to drop out, although often beyond the control of the individual adolescent, is one of the most important actions that youth may take (Brooks-Gunn et al. 1993). For many, another perhaps equally important action will involve whether to initiate, and indeed continue the use of substances such as tobacco, alcohol and/or illicit drugs.

A recent World Health Organisation fact sheet (2002) reports that among teens aged 13 to 15 years internationally, approximately one in five smokes cigarettes. Key findings from The National Household Survey on Drug Abuse released in 2000 and conducted in the United States, reveal that roughly 9.7 million 12- to 20-year-olds reported drinking alcohol in the month prior to the survey; and of these, 6.6 million reported binge drinking and 2.1 million reported heavy drinking (Alcohol Policies Project 2003). Prevalence rates from many European countries reveal a similar picture (Gabhai and François 2000; Marieke et al. 2001). Various school surveys among 15- to 16-year-old students conducted in Western European countries by the European Monitoring Centre for Drugs and Drug Addiction (2000) suggest that between 39.8% (United Kingdom) and 4.7% (Portugal) of these youth have used any illegal drug in their lifetimes. Lifetime cannabis use among these adolescents ranges from between 37.5% (United Kingdom) and 3.8% (Portugal); lifetime ecstasy use varies between 9.0% (Ireland) and 1.0% (Sweden); and lifetime cocaine use ranges from 4.3% (Spain) to 0.6% (Sweden). In South Africa high prevalences of the use of tobacco, alcohol and illicit drugs among adolescents has been noted. In Cape Town, for example, Flisher et al. (2004) reported that the proportions of students in Grade 11 who had used tobacco, alcohol and cannabis in the previous month were 27%, 31% and 7% respectively.

The negative health consequences of substance use and abuse are extensively documented in scientific and lay literature internationally and will not be repeated here. The preceding discussion has provided an overview of the prevalence rates of two seemingly dissimilar behaviours among adolescents around the world: dropping out of school and using/abusing substances. An understanding of the association between dropping out and substance use would have valuable implications for prevention of both risky behaviours, and concurrently the promotion of health and well-being among the world’s young persons.

The aims of the review are, therefore, to provide a synthesis of findings from dropout and substance use research; to provide an overview of dominant theoretical paradigms guiding the research; to identify shortcomings in current research and areas for future research; and to provide a valuable resource for those intending to conduct research in these fields.

Review Methodology

The review comprises English language articles that were published in peer-reviewed journals between 1990 and 2006. All articles investigated either a direct or indirect association between high school dropout and substance use/abuse. A database search for relevant articles was conducted on PsycINFO, PsycARTICLES, Medline, Pubmed, Eric, ISAP (The Index of South African Periodicals), Social Science Index and Academic Search Premier. Details of published works were obtained using permutations of the following key words: school dropout, high school dropout, school disengagement, education, education attainment, tobacco, cigarettes, alcohol, drugs, illicit drugs, marijuana, cannabis, heroin, cocaine, crack, LSD, PCP, and hard drugs. The reference lists of retrieved articles were scanned for further relevant publications. A hand search was done of the table of contents of those journals from which articles were accessed for the review.

Results

The search resulted in the retrieval of 46 articles that met the inclusion criteria (Table 1). The majority of the studies originated in the United States of America (N = 41), while one was from South Africa (Flisher and Chalton 1995), three from New Zealand (Fergusson and Horwood 1997; Fergusson et al. 2003; Fergusson et al. 1996), one from Norway (Wichstrøm 1998) and one from Australia (Lynskey et al. 2003). The ensuing section of the review is structured in terms of those findings that are theoretically informed and those that are largely a-theoretical in nature. The former group is introduced first followed by the latter grouping. The review concludes with recommendations for future research.

Table 1 Sample description, direction of substance use/dropout relationship and main findings
Full size table

Findings Related to Theoretical Framework/s

For purposes of coherence, an overview of each of the theories is combined with the relevant, pertinent findings.

Social control theory proposes that when the moral bond that ties people to each other and to social norms is broken, mechanisms of social control in the form of restraints on antisocial behaviour become ineffectual (in Aloise-Young and Chavez 2002). When social controls are weak, individuals are more likely to deviate from societal norms for appropriate behaviour, for example, by using illicit drugs and/or drunkenness (Fagan and Pabon 1990). They are also more likely to show poor commitment to conventional society and conventional social groups, school being one example of such a group.

Support for this theory’s propositions is mixed. On the one hand, Aloise-Young and Chavez (2002) found that those adolescents in their sample who left school, because of low school bonding (reflecting a poor commitment to school), did not evidence the highest level of substance abuse, as would be expected. On the other hand, in their predictive models of early school dropout, Battin-Pearson et al. (2000) found that the relationship between school bonding and dropout was mediated by poor academic performance, while Fagan and Pabon (1990), investigating the relative contribution of a variety of factors in distinguishing dropouts from students, found that dropout is in fact a function of a number of social bonds that include school integration, rather than of delinquency and/or substance use behaviours alone.

Problem-prone behaviour and general deviancy theory describe adolescent dropping out and substance use as being two of a constellation of problem behaviours, co-varying with other deviant behaviours to which certain adolescents are prone (in Battin-Pearson et al. 2000; Beauvais et al. 1996; Fagan and Pabon 1990; Fergusson and Horwood 1997; Fergusson et al. 1996; Kaplan and Liu 1994; Newcomb et al. 2002; Obot and Anthony 1999). Specifically, different forms of deviant behaviour co-vary because adolescents who hold non-conforming attitudes and values, and who attempt to establish their identities as independent and adult persons, are more likely to be engaged in a variety of non-conformist behaviours such as smoking, drug and alcohol use, and abandoning the student role (including dropping out of school) amongst others.

This theoretical orientation enjoys the most consistent support among the review’s studies. Fergusson and colleagues found that early cannabis users were a group at risk for a variety of subsequent problems such as abuse of other substances, truancy, school dropout and mental health problems (Fergusson et al. 1996; Fergusson and Horwood 1997). Besides obvious academic problems, Beauvais et al. (1996) found that dropouts in their sample also used all forms of drugs, evidenced current moderate to heavy drug use and were more likely to perpetrate and be victims of violent behaviour. Two studies developed latent constructs of “general deviance” which included delinquency, drug and other substance use, and early sexual involvement (Battin-Pearson et al. 2000; Newcomb et al. 2002). Both studies were able to confirm the general deviance hypothesis. In one, general deviance was found to be a stronger predictor of high school failure (including both truancy and dropping out) than was any single act of deviance (Newcomb et al. 2002). In the other, in spite of a proposed mediational effect of poor academic achievement on high school dropout, general deviance had a direct effect on school dropout beyond the influence of poor academic achievement (Battin-Pearson et al. 2000).

Primary socialisation theory proposes that when ties to school and/or family are weak, and ties to peers are strong (particularly substance-using peers), the adolescent is most at risk (in Aloise-Young and Chavez 2002). These authors found support for this proposition in that substance use was highest for those dropouts who cited “leaving school to be with friends” as a reason for dropping out.

The primary socialisation premises are mirrored in social learning theory; where it is proposed that learning processes that lead to deviant behaviours, including substance use/abuse, occurs in association with peers (in Fagan and Pabon 1990). These learning processes are thought to be most influential in the face of weak social bonds, where delinquent socialisation becomes the strongest learning influence. In these circumstances, it may be asked, how exactly do peer relationships operate to increase the risk of both substance use and/or dropping out of school? Three closely related theoretical positions—peer cluster theory (in Beauvais et al. 1996), deviant affiliation theory (in Battin-Pearson et al. 2000) and the theory of differential association (in Wichstrøm 1998) provide some answers to this question.

Peer cluster theory proposes that school problems are a major factor in creating deviant peer clusters. Those experiencing problems at school are thought to have a way of seeking each other out and together they form peer groups. These peer clusters encourage, support and normalise a range of deviant behaviours including attitudes sympathetic to dropping out and substance use. From this perspective, and that of deviant affiliation theory, attachment to, and affiliation with deviant peers is likely to lead to deviant attitudes and behaviours through the processes of social learning and attitude formation. Moreover, bonding with anti-social peers will influence adolescents’ own anti-social proclivities.

Swaim et al. (1998) provide the strongest support for a socialisation model of adolescent polydrug use, based on peer cluster theory. Among these researchers’ sample of dropouts, associations with drug-using peers accounted for the largest proportion of variance directly related to adolescent substance abuse for both male and female, Hispanic and non-Hispanic white participants. Furthermore, associations with drug using peers also mediated the effects of school adjustment, religious identification and family sanctions on drug use.

Although not referring directly to deviant affiliation theory as a mechanism to explain their significant results, both Brook and Balka (1999) and Fergusson and Horwood (1997) found that early adolescent drug use was associated with higher affiliations with delinquent and substance-using peers. These findings clearly support the deviant affiliation perspective. In another study, correlations between drug use, bonding to anti-social peers and dropping out of school were all positive (Battin-Pearson et al. 2000). Moreover, their latent construct, “bonding to antisocial peers”, was found to have both a direct and an indirect (mediated by poor academic achievement) effect on dropout.

The theory of differential association views the above pathways marginally differently. Here the proposition is that substance use is the catalyst that brings substance-using adolescents together. The newly found peers act as role models and operate to reinforce behaviour/s that may increase the likelihood of dropping out of school (Wichstrøm 1998). Relevant to this theory, Wichstrøm found that dropouts compared to school completers and in-school students in his sample, were intoxicated significantly more often per year, used narcotics more often, and had more friends who got drunk regularly, used cannabis and had brushes with the police.

Strain theory proposes that school failure causes frustration, which results in students becoming alienated from school (in Aloise-Young and Chavez 2002). In order to acquire status, students experiencing this frustration seek out alternative self-defining behaviours that are often deviant in nature, such as the use of various substances. From this perspective, school failure, poor school performance and dissatisfaction with school, as well as high rates of substance use, should all be strongly related to dropping out of school. Contrary to these expectations, Aloise-Young and Chavez (2002), found that their sample of students who left school because of bad grades, did not evidence the highest rates of substance use as would be expected.

Main Findings from Cross-sectional and Longitudinal Studies

Tobacco Use

Across all the cross-sectional studies reporting on their sample’s use of tobacco (cigarette smoking), dropouts were more likely to report current cigarette smoking than were in-school students (Flisher and Chalton 1995; Wang et al. 1998; Zimmerman and Maton 1992) (regardless of ethnicity (Aloise-Young et al. 2002)), or high school graduates (Gfroerer et al. 1997). They were also more likely to do so at an earlier age and to be heavy smokers (Aloise-Young et al. 2002). In-school students identified as being at risk for dropping out of school were also more likely to smoke cigarettes than were low-risk students (Eggert and Herting 1993) and more likely to be heavy cigarette smokers than normally performing in-school students (Aloise-Young et al. 2002).

A number of longitudinal studies were able to demonstrate the unique effect of cigarette use on high school dropout. After controlling for hypothesised covariates (gender, race, age, family and academic background, problem behaviour and peer influence, and school context), Elickson et al. (1998) found that only cigarette use remained a significant predictor of dropout across all race groups. Independent of prior substance use and problem behaviour, Kogan et al. (2005) found that dropouts were 1.6 times more likely to report cigarette use. Newcomb et al. (2002) were also able to provide evidence of the unique effect of tobacco use on high school failure. Examining cigarette smoking trends over time, Escobedo and Peddicord (1995) found that men who had less than a high school education were more likely to continue cigarette use at similar rates over time, while among women with less than a high school education, cigarette use increased over time. Significant declines in cigarette use were found among high school graduates.

Alcohol Use

In some cross-sectional studies a clear association between alcohol use and dropout (Aloise-Young and Chavez 2002; Arellano et al. 1998; Fagan and Pabon 1990; Flisher and Chalton 1995; Wichstrøm 1998; Zimmerman and Maton 1992); being at risk for dropout (Eggert and Herting 1993); experiencing academic problems (Beauvais et al. 1996); and completing high school (Dee and Evans 2003) was found. In one cross-sectional study the relationship between alcohol use and dropout was less clear (Gfroerer et al. 1997). This study found that the association between current alcohol use and dropout was no longer evident after adjusting for age, gender and race.

Among the longitudinal studies, Ellickson et al. (1998) found that alcohol use predicted dropping out. However, after controlling for covariates, such as demographic variables, family and academic background, deviance and school environment, alcohol use reduced the odds of dropping out for Asian and Black adolescents. Yamada et al. (1996) were able to conclude that high school graduation is negatively associated with alcohol consumption. Wichstrøm (1998) found that the relationship between dropping out, and intoxication and quantity of alcohol consumption might be mediated by parental care, truancy and peer problem behaviour.

Rosa Crum and colleagues investigated, whether young adults who dropped out of school were at greater risk for later development of an alcohol disorder. In one study, dropping out of school was significantly associated with an increased risk of later alcohol abuse, even after controlling for the possible effects of early onset alcohol problems (Crum et al. 1998). The other found that, compared to school graduates or those who earned a GED, those who dropped out of school were more likely to develop alcohol abuse or dependence (Crum et al. 1993).

Marijuana/Cannabis Use

A number of cross-sectional studies found that dropouts and in-school students identified as being at risk for dropping out of school reported more current marijuana/cannabis use than in-school students (Aloise-Young and Chavez 2002; Eggert and Herting 1993; Zimmerman and Maton 1992), and high school graduates (Gfroerer et al. 1997). They also reported more lifetime use than did in-school students, with female dropouts reporting more lifetime use than their male counterparts (Flisher and Chalton 1995). A combination of lifetime and current marijuana/cannabis use was reported more often by dropouts than in-school students (Aloise-Young and Chavez 2002), and in-school students experiencing academic problems irrespective of gender or ethnicity (Beauvais et al. 1996).

Evidence of a significant association between the frequency of early cannabis use and later dropping out was found in three longitudinal studies (Fergusson et al. 1996; Fergusson et al. 2003; Fergusson and Horwood 1997). In these studies, although the association changed depending on the entry of covariates and/or mediating variables, it remained statistically significant. Three other studies also provided evidence of marijuana/cannabis use proving to be a predictor of dropout, despite the inclusion of covariates (Brook and Balka 1999) and potential confounding variables (Lynskey et al. 2003) in the analyses. Additionally Roebuck et al. (2004) concluded that both chronic (weekly in the past year) and non-chronic (any use in the past year) marijuana use were significantly related to high school dropout even after controlling for other drug use in one model, and alcohol consumption in another. Ensminger et al. (1996) found that marijuana use was directly related to dropping out for both males and females although more so for the latter group. Marijuana use was associated with an increased risk for dropping out of school for both males and females among African-American youth in Chicago (Green and Ensminger 2006). Only one study found that after controlling for a variety of confounding variables, marijuana use no longer remained a significant predictor of dropping out except in the case of Latino youth (Ellickson et al. 1998).

Three longitudinal studies provided evidence of a “reverse causal pattern” whereby the experience of dropping out of school led to an increase in marijuana use (Ensminger et al. 1996; Green and Ensminger 2006; Kogan et al. 2005). However, after adjusting for background factors and pre-existing levels of cannabis use Fergusson et al. (2003) found no significant association between dropping out of school and subsequent frequency of cannabis use. This contrary finding may be a function of the relatively young age (17– 20 years) at which subsequent cannabis use was assessed compared to the age 32–33 years used by Ensminger et al. (1996) and Green and Ensminger (2006).

Illicit Drug Use (Other than Marijuana/Cannabis)

A number of cross-sectional studies found that, besides marijuana/cannabis, other current illicit drug use was found to be higher among dropouts and students at risk for dropping out than in-school students (Eggert and Herting 1993; Guagliardo et al. 1998; Zimmerman and Maton 1992), and high school graduates (Gfroerer et al. 1997). A combination of lifetime and current use of illicit drugs was also found to be higher among dropouts than in-school students (Aloise-Young and Chavez 2002), and in-school students experiencing academic problems irrespective of gender or ethnicity (Beauvais et al. 1996). Dropouts more often reported having ever used illicit drugs than in-school students in Flisher and Chalton’s (1995) study. After adjusting for a number of confounders, such as age, gender, and Hispanic background, Obot and colleagues found that, compared to high school graduates and in some instances GED holders, dropouts were significantly more likely to have injected a drug recently (Obot and Anthony 1999; Obot et al. 1999) and more likely to have a lifetime history of injecting drug use (Obot and Anthony 2000).

Several longitudinal studies provided evidence of drug use being a significant predictor of dropout, two of which also demonstrated that dropping out of school was found to increase the likelihood of subsequent drug use. Garnier et al. (1997) found a significant positive correlation between teen drug use and dropping out of school, and that teen drug use was among a number of significant predictors of dropout. Other research, however, demonstrated that the relationship between dropping out and substance use was confounded by factors such as impregnating a partner among Latino males (McCLusky et al. 2002), and mediated by factors, such as motivation to perform well in school, being the object of negative social sanctions, and adoption of age-inappropriate roles, such as marriage, pregnancy and parenthood (Kaplan and Liu 1994). Register et al. (2001) also found that the probability of adolescent drug use was a significant and negative determinant of educational attainment for white adolescents, but not for black and Hispanic students, where factors, such as family constellation, parental education levels, number of siblings, and marriage were significant determinants of educational attainment.

Within the construct “general deviance”, Battin-Pearson et al. (2000) combined drug use with measures of non-violent and violent offences and sexual behaviour. These researchers were unable to provide any evidence of the unique effect of drug use on dropping out. Rather, general deviance (in addition to low SES, and bonding to anti-social peers) was a significant and independent predictor of dropout. Newcomb et al. (2002) employed a similar latent construct, encompassing polydrug use, sexual involvement, delinquency and school problems. Although general deviance contributed significantly to high school failure (both truancy and dropping out) only tobacco use, within the polydrug variable had a unique effect on school failure. Drapela (2006) used an additive index of drug use, which included daily cigarette use, and yearly and monthly alcohol, marijuana and cocaine use. She found that although having dropped out was associated with a 30% increase in drug use after controlling for demographic variables and early drug use, once antecedent school problems were entered into the regression model, dropouts did not differ significantly from students in levels of drug use.

A study by Krohn and colleagues explored the bi-directional relationship between dropping out, and drug and alcohol use (Krohn et al. 1997). These researchers found that after controlling for a number of potential confounding variables, the use of drugs and alcohol in early adolescence increased the risk of dropping out of school among males, but not among females. Regressions predicting alcohol and drug use following dropping out, also illustrate a gendered pattern of results. After accounting for a number of potential confounding variables, dropout is significantly related to both later drug and alcohol use for males and later drug, but not alcohol use for females.

Discussion

It is evident that a great deal of research has been undertaken, almost exclusively in the developed world, in an effort to understand the relationship between dropping out of school and substance use. Findings from the cross-sectional studies provide conclusive evidence of higher rates of cigarette, alcohol, marijuana and other drug use among dropouts and students at-risk for dropping out compared to in-school students or graduates.

The findings from longitudinal studies provide evidence of a significant, unique effect of cigarette use on dropout. Similarly, the majority of longitudinal studies that examined the relationship between marijuana use and dropout came to the same conclusion. Only one out of the 10 studies was able to demonstrate the confounding effects of age, gender, family structure and a variety of school related factors in the relationship between marijuana use and dropout. This may be because these researchers included alcohol and cigarette use in their final multivariate model whereas the other studies examined the use of marijuana only. Regarding the relationship between alcohol use and dropping out of school, longitudinal studies found that a number of factors confounded the relationship. These include family variables, school problems including academic background and deviant/problem behaviour including deviant peer behaviour suggesting that these factors are more salient than alcohol use in decisions to dropout of school.

The relationship between other drug use and high school dropout is not as clear. While some studies were able to demonstrate a unique relationship between the two behaviours, in others the relationship appears to differ by race and gender, and to be confounded and/or mediated by factors such as the adoption of age-inappropriate roles (pregnancy, impregnation, and marriage), motivation to perform well at school, and being the object of negative social sanctions. Furthermore, findings from those studies that included other drug use with other theoretically-informed variables such as deviant behaviour, sexual behaviour and school problems where unable to provide evidence of a unique effect of drug use on dropping out of school. This calls into question the usefulness of combining measures of drug use with other problem behaviours when examining the effect of drug use on school dropout.

Those students who had dropped out of school were also more likely to continue to use cigarettes, alcohol and marijuana at significantly elevated levels compared to high-school graduates suggesting the continuing influence of substances in these people’s later functioning.

In summary, there is little doubt that dropouts use substances at elevated levels compared to their in-school peers and/or school graduates. Evidence from the longitudinal studies largely support the unique effect of cigarette and marijuana use on dropping out of school. However, the relationship between alcohol and other drug use and high school dropout is a complex one that may benefit from further research.

Despite findings related to the theoretical frameworks as described above being able to distinguish dropouts from in-school students or graduates, they provide only marginal explanatory power to the substance use/dropout relationship. This may be because, as Drapela (2006) notes, there is no explicit theory of dropout and substance use. Why this is the case, and suggestions as to how future research may address this inadequacy form the final part of this review.

The Complexity of the Dropout Phenomenon

The relationship between substance use and dropping out school is arguably a function of the complexity of the dropout phenomenon and the “bewildering number and diversity” of factors that have been found to be associated with early school leaving (Rosenthal 1998: p. 422).

Recently, researchers in the Netherlands and in Australia have concluded that this array, and complexity of information is unhelpful (Smyth and Hattam 2001) and unsatisfactory (Dekkers and Driessen 1997) as it adds little to our understanding of why so many young people dropout of school. Smyth and Hattam (2001) summarise the situation well:

“The difficulty is that we know what the problem is—and we have for some time—but we only really know the problem from the outside; i.e., in terms of the statistical and calculative extent of non-completion. With few rare exceptions, our knowledge is sparse in terms of the complexity of what is really happening, and why” (p. 401).

There is little doubt that the issue of substance use is interwoven in this complexity and probably explains why it is that the theories were unable to provide any definitive explanation of the relationship between substance use and dropout.

Uncontrolled Confounding Factors

Although a number of studies included a variety of theoretically informed and/or empirically determined confounders in the relationship between substance use and high school dropout, the possibility exists that not all potential confounders were included in their analyses. It is entirely possible that as yet undetected confounders may exist. Lynskey and Hall (2000) make this point when they draw attention to the fact that obvious biological hypotheses exist that explain the relationship between cannabis use and high school dropout, for example reduced motivation and cognitive impairment. However, the same cannot be said for the relationship between cigarette use and dropout, where it is more likely that some or other uncontrolled confounding factor/s are associated with an increased risk of tobacco use and dropout.

Addressing Negative Stereotypes and Advocating for Qualitative Research

The literature and theories on dropouts tend to cast dropouts as an homogenous (often deviant) group of young people who resist conforming to the demands of education systems whose imperative it is to ensure that they reach a prescribed benchmark of educational “success”.

After a careful examination of studies that tended to map the issue of high school dropout statistically or through surveys (Smyth and Hattam 2001), researchers advocating a ‘voiced’ or qualitative approach revealed a focus on the “...individual, personal, familial and cultural deficits...” (Smyth 2005: 3), that seemingly reside within individual students. In this way dropouts are blamed for their circumstances. Framed as “...depressed, helpless, and even without options... as losers” (Fine 1991: 4–5), the stereotypes that develop lead us to believe that dropouts are an homogenous group of “...unworthy, and immoral, or pitiable, victimised and damaged” young people (Fine 1991: 74).

On the contrary, as the numbers of dropouts increases, and in fact become the majority of high school-aged young people as is the case in South Africa and many developing nations, the imperative needs to shift to exploring the complex interdependent factors that cause early school leaving as experienced and lived by the leavers themselves. The notion of using student voices as data in dropout research has a very recent history (see Shacklock et al. 1998; Smyth 2005; Smyth and Hattam 2001), and it is interesting to note that only one of the reviewed studies used a qualitative method, combining it with a survey (Aloise-Young and Chavez 2002).

Furthermore qualitative research may also well uncover other salient factors that research to date has not yet contemplated. As Rumberger (1995) suggested, a qualitative enquiry into cultural and other factors would be important in developing a better understanding of differences within and among ethnic groups in the tendency to complete high school.

Resilience in the Face of Risk

A further reason for the inconsistent findings from the theories and some of reviews’ studies may have to do with the fact that within the dropout population, there are large numbers of young people who, despite similar antecedents to dropping out of school, are not “depressed, helpless or losers” as the common stereotype suggests. Large numbers of dropouts also do not use substances (Beauvais et al. 1996 and Garnier et al. 1997). In every one of the review studies whose samples consisted of dropouts only, many more of these young people were not using substances than were. It is evident that protective factors may operate to provide some dropouts with the ability to resist substance use, and indeed other maladaptive behaviours. It is lamentable that only one study investigated the role of protective factors as confounders in the substance use/dropout relationship (Swaim et al. 1998). This study was able to demonstrate that family sanctions serve as a protective factor against polydrug use; dropouts were more likely to perceive their parents as “caring” if their parents held negative attitudes towards drug use.

Having suggested a change of focus from risk to protection and resilience, it is important to note that an exclusive focus on protective factors would be an incomplete strategy for reducing dropout and substance use (Pollard et al. 1999). These authors suggest that studies investigating both risk and protective influences would better inform prevention policies and programmes.

Methodological Issues

A number of methodological issues also need to be mentioned. Firstly, all but one of those studies whose samples consisted of both urban and rural youth (N = 29), did not provide some analysis focusing on a comparison among the groups—particularly with regard to dropping out of school where there is a lack of evidence suggesting urban/rural differences. The one study that did provide some comparative data, did so on the basis of a significantly higher rate of early school leaving among rural compared to urban adolescents in Victoria, Australia (Lynskey et al. 2003). Future studies should take note of potential urban/rural differences and where possible incorporate comparative data.

Secondly, the studies employing longitudinal designs, all suffered varying degrees of attrition in their samples: a difficulty endemic to most longitudinal research. Those studies where attrition rates were relatively high (> 20%), evidence of differential attrition and non-random sample loss is apparent. Evidence suggesting that socially disadvantaged (Fergusson and Horwood 1997; Fergusson et al. 1996; Fergusson et al. 2003) and poor persons (Ensminger et al. 1996), dropouts (Crum et al. 1998; Lynskey et al. 2003; Winstrøm 1998), and drug users (Crum et al. 1998; Kaplan and Liu 1994) are over-represented in some of the loss to follow-up groups is cause for concern. It is these groups of persons who are most at risk for dropout and/or substance abuse. That some studies did not provide an analysis or description of their lost participants, and/or that others only explored certain demographic characteristics of their lost participants, is a cause to evaluate their findings with some caution. It is suggested that more rigorous mechanisms to retain participants in longitudinal studies and lower the attrition rates should be employed.

Thirdly, there appears to be two distinct definitions of school dropout. The first encompasses those who had not attended school in more than 30 days, had not sought a transfer to another school, and had not sought re-admission. Only five of the studies employed this definition, which clearly allows for a possible return to schooling and subsequent graduation. The other definition is more loosely constructed, but encompasses those not enrolled in school, or having received a high school diploma. Some variation of this definition involves those students who had received the General Equivalency Diploma (GED). In five studies these youth were not considered high school dropouts (Brooks-Gunn et al. 1993; Ellickson et al. 1997; Garnier et al. 1997), while one included GED holders with the dropout group (McCluskey et al. 2002). Most often the data from those who had gained a GED was analysed separately and/or comparatively to dropouts and in-school students (Crum et al. 1993; Crum et al. 1998; Obot and Anthony 1999; Obot et al. 1999; Obot and Anthony 2000).

Although all the studies stated their definition of dropout clearly, the transitory nature of the first-mentioned definition as opposed to the more permanent nature of the second does limit the ability to integrate findings across studies employing different definitions.

Related to the problem of varying definitions of dropout, the fluidity of educational attainment also needs to be considered. In all but one study, have no information about the various studies’ dropouts beyond the point at which they were classified as having dropped out of school. This is regrettable given that Aloise-Young et al. (2002) found that a 3- to 6-year follow-up of their sample found that one third of dropouts had graduated, and one-third of students at risk for dropping out had dropped out of school.

Fourthly, the way in which the use of various substances is measured varies greatly. Among those studies that measured the use of just one substance or a number of substances but provided an analysis for each, most measured the use of their respective substance/s of interest in terms of recent or current use. The majority of the studies’ measurements of substance use entailed constructing varying composite measures of substance use/abuse and/or categories of substance users, for example, an additive index of drug use employed by Drapela (2006). Some studies combined various measures of a number of substances into a single construct, for example, “drug positive” (Guagliardo et al. 1998). Two studies measured substance use as a continuous variable encompassing the cumulative measure of the number of times participants reported using any of a variety of illegal substances (Krohn et al. 1997; McCluskey et al. 2002). Other studies incorporated various measures of a number of substances into a broader category that included other aspects of drug affiliation or drug involvement (Eggert et al. 1993). Some studies combined lifetime and current measures of a variety of substances to create user categories, for example “problem drug user” (Ellickson et al. 1997). The range and inconsistent measurement of the use of substances among study participants makes comparison and an integration of findings difficult. However, if one assumes that any substance use comprises risk for these young people, then all have equal relevance and comparison and integration of findings may be undertaken with some confidence.

Fifthly, the accuracy of self-report of substance use is questionable in all but one of the studies reviewed. This is unfortunate given that there is no way in which self-report responses to items assessing substance use can be deemed entirely accurate. As an example, Brasseux et al. (1997) (in Guagliardo et al. 1998) found that almost half of respondents who tested positive for marijuana use denied having used marijuana in the past 2 days, and 31% denied using the substance in the past 30 days. Guagliardo et al. (1998) subsequently employed a biological determinant of substance use (urinanalysis) rather than a self-report indicator. Furthermore, Gfroerer et al. (1997) suggest that under-reporting may plague household studies where the anonymity of responses cannot be ensured. Fourteen of the reviewed studies were household surveys calling into question the accuracy of their substance use findings. Some measure to ensure anonymity of reporting can be exercised in school and/or college settings, but would exclude dropouts.

Finally, Lynskey and Hall (2000) draw attention to the variable reliability and validity of measures used in studies assessing the effects of cannabis use on educational attainment. They suggest that “...any substantial limitations in the reliability and validity of observed measures is likely to reduce the power of a study to control for the effects of [covariates]” (p. 1626). It is unfortunate that among this review’s studies, only 14 reported this information when available.

Directions for Future Research

Given the foregoing, directions for future research include:

  • Qualitative research may well identify contexts and reasons for dropping out and substance abuse, and the relationship between them, that have to date been overlooked in the empirical literature, and theories on high school dropout. Furthermore, qualitative data would greatly enrich the array of empirical findings that exist. Triangulation, where participant’s own perceptions of, and reasons for dropping out of school is used to illustrate empirical findings would be particularly valuable.

  • Theories and research that examine factors that protect dropouts from substance use is sorely needed. Future research should attempt to combine their investigation of risk factors with those that offer some students protection against risk.

  • Research needs to be undertaken in developing countries. It is these countries that appear to have the least favourable school outcomes and yet there is a dearth of research to inform causes, consequences, and avenues for intervention/prevention. It is debatable whether findings from countries in the developed world would provide any insight into the dropout phenomenon in developing countries, such as the one from South Africa in this review, where it is the majority rather than the minority of high school learners who dropout of school, and where the social context is characterised by pervasive poverty, unemployment, and social inequality.

  • The methodological issues mentioned earlier should be addressed where possible. For example, designing research among rural adolescent dropouts, ensuring retention in longitudinal studies, defining dropout consistently and perhaps following-up on dropouts to assess their continued dropout status or otherwise, using consistent measures of substance use across studies and perhaps adding biological confirmation of substance use if possible, and determining and reporting the reliability and validity of measures used.