TRUANCY AND PERCEIVED SCHOOL PERFORMANCE: AN ALCOHOL AND DRUG STUDY OF UK TEENAGERS

Patrick Miller* and Martin Plant

Alcohol and Health Research Centre, City Hospital, Greenbank Drive, Edinburgh EH10 5SB, UK

Received 3 August 1998; in revised form 11 March 1999; accepted 7 May 1999


    ABSTRACT
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
This study is based on a subsample of 15- and 16-year-old school students from the UK, part of the European School Project on Alcohol and Other Drugs (ESPAD). Information was available on truancy rates, perceived school performance, family structure, lifestyle, and usage of alcohol, cigarettes and illicit drugs in 6409 teenagers. Living in a single-parent family, lack of constructive hobbies, presence of psychiatric symptoms, and an aggressive outgoing delinquent lifestyle bore the strongest associations to truancy and to perceived school performance. There were also strong relationships between both these last two variables and use of alcohol, cigarettes, and illicit drugs. However, the effects of alcohol, cigarettes, and illicit drugs were largely accounted for by other variables. Having at least one parent who both supported the respondent and who exercised some control was predictive of better perceived school performance.


    INTRODUCTION
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
In March 1995, a large study was carried out in the UK on school students born in 1979, as part of the European School Project on Alcohol and Other Drugs (ESPAD). This exercise involved the UK and 22 other countries (Hibell et al., 1997). Two of the variables studied were truancy from school in the last 30 days and the students' perceptions of their schoolwork, compared to others of their own age. Both of these, but particularly the former, are currently matters of considerable concern in the UK.

In the literature, truancy and poor academic performance have frequently been considered part of a general deviance or ‘problem behaviour’ syndrome (Jessor and Jessor, 1977Go). Even so, there has been some debate about how general this syndrome actually is (Grube and Morgan, 1990Go; Miller et al., 1995Go). One of the aims of this paper is to examine the possibility that there may be other factors, not normally considered part of general deviance, which might predict the two variables chosen for study. Data were available from the study to try to throw light on these matters. Four of the predictor variables concerned the family. These were: (1) family structure (intact family vs single-parent family); (2) level of parental education; (3) parental caring; (4) parental control. Another set of predictors consisted of the lifetime use of various substances (alcohol, tobacco, solvents, cannabis, and illicit drugs other than cannabis). Finally there were three derived factors measuring psychiatric symptomatology, presence of a sociable/delinquent lifestyle and the extent to which the student had ‘constructive hobbies’.


    METHODS
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
Study population
The original study, more fully described by Miller and Plant (1996), used data from the ESPAD survey of schools in the UK carried out in March 1995. The population sampled consisted of individuals aged 15–16 years who had been born in 1979. Seventy schools were chosen to try to meet four requirements. These were to represent Great Britain and Northern Ireland as a whole, to reflect different regions within the UK, to represent both state and independent schools and to allow analysis of urban vs rural areas. These aims were addressed by randomly choosing 60 state schools from 13 different areas of the UK and by adding a randomly chosen sample of 10 independent schools. All eligible students within each school were approached by a local organizer from the school staff and were asked to complete a standardized questionnaire under ‘exam’ conditions. Where possible, all eligible respondents within a school were tested at the same time.

Thirty-seven of the original 70 schools chosen refused co-operation, mostly on the grounds that they had been over-researched previously. Each of these was replaced by the next school on the list. One school dropped out too late for replacement, due to the illness of the local organizer. Some schools unfortunately were unable to allow enough time for all students to finish the questionnaires. Within the 69 co-operating schools the numbers of pupils were as follows:


View this table:
[in this window]
[in a new window]
 
 

This paper is based on the 6409 (70.2%) who completed the whole questionnaire. There were several indications that the reliability and validity of the responses were acceptable. For instance, inconsistent responses to two versions of the same question were never higher than 2.5% and a fictitious drug, ‘relevin’, was reportedly used by only 0.3% of respondents.

The questionnaire
This was made up of a core section, used by all participating countries and an additional section used in the UK. Demographic details within the core section consisted of the respondent's gender, month of birth, household composition, and level of parental education. There were also questions on school attendance, leisure activities, and a brief self-esteem scale. However, the three main core parts concerned cigarette smoking, alcohol consumption, and illicit drug use. Among the variables explored, were frequency of use of cigarettes and alcohol within various time periods, lifetime frequency of intoxication, frequencies of use of cannabis, inhalants, and a variety of illicit drugs.

The additional section, used only in the UK, consisted mainly of the Achenbach Youth Self-report Schedule (YSR; Achenbach, 1991), tapping emotional and behavioural problems in adolescents. There were also items on parental support, obedience to rules, and social support. The full questionnaire is available from the corresponding author upon request.

Analyses
Most of the analyses were carried out using the Pc Carp package from the University of Iowa (Fuller et al., 1989Go). This allows for clustering of subjects within schools, applies weights to the data according to the inverse probability of school selection and provides many of the features of standard statistical packages for ungrouped random samples. The logistic option of Pc Carp was used to perform the group discriminations described below. Principal components analysis of the Achenbach responses and other variables was carried out using SPSS in the ordinary way, but after the correlation matrix had been checked using Pc Carp and found to be virtually unaffected by clustering within schools.

Dependent variables
The dependent variables were:

  1. Truancy from school. This variable was based on the answers to the question ‘During the last 30 days how many whole days of school have you missed?’


    View this table:
    [in this window]
    [in a new window]
     
     

    The truancy from school variable was based on the answers to part (b) reclassified into no days missed, one day missed, two or more days missed, and no answer given.

  2. Academic performance. This variable concerned the pupils' perceptions of their academic performance and was based on the answers to the question ‘How good do you think you are at schoolwork, compared to people your age?’ Responses from an original seven-point scale were recoded into below average, average, and above average.

Independent variables
The independent variables were derived as follows:

  1. Family structure. Each subject was classified according to whether (s)he lived with both parents, with mother (and possibly other adults), with father or with neither parent. The Pc Carp weighted counts on this variable were: both parents 5533 subjects (76.1%); mother 1574 subjects (20.4%); father 283 subjects (3.7%); neither parent 340 subjects (4.4%).
  2. Tobacco, alcohol and illicit drug outcome variables. There were five of these, namely alcohol intoxication, cigarette smoking, solvents, cannabis, and illicit drugs other than cannabis. On all these, the subjects were classified into three categories; these being, never in a lifetime, at least once, and 40 or more times.
  3. Demographic variables. The subject's sex and the level of education of the parents were included. The latter was measured in two categories — neither parent with any higher education and at least one parent with some higher education.
  4. Personality/lifestyle. Five measures were included. Three of these were derived by principal components analysis of syndrome scores on the YSR (Achenbach, 1991Go) together with self-esteem, social support, and leisure time activity questionnaire items. This analysis has been more fully described by Miller (1997). Five factors emerged with eigenvalues greater than one and were subjected to varimax rotation. Subsequently, the reliability of the factor structure was tested by running separate analyses for the English students and for all other students. Correlations between the factor loadings for these two separate analyses were greater than 0.9 for all five factors.

In the original analysis, the first factor, labelled ‘symptoms' was loaded (in descending order) by the YSR measures anxiety/depression, attention problems, social withdrawal, aggressiveness, social problems, somatic complaints, thought problems, and by low self-esteem. The highest loadings on the second factor, labelled ‘sociable/delinquent', were going out with friends and delinquency. The third one, ‘hobbies', consisted of having constructive hobbies, reading books, and sports activity.

The students' scores on these three factors were divided at the 66th and 33rd percentiles into high, medium, and low categories. Two other factors were originally derived. The first, consisting mainly of computer games and sport is not further reported, as it proved unrelated to the outcome variables. The second, consisting mainly of social support variables was not used, because it was desired to study two of its main constituents separately. These latter two variables (both on five-point Likert scales), reported below, were ‘parental caring’, the answer to the question: ‘I can easily get warmth and caring from my mother and/or father’ and ‘parental control’, the answer to: ‘my parents set definite rules about what I can do outside the home’. Further details of the factors have been described by Miller (1997).


    RESULTS
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
Tables 1–4GoGoGoGo set out the effects of the predictor variables (family structure etc. and alcohol use etc., respectively) separately on the dependent variables [absence from school (Tables 1 and 2GoGo) and perception of academic performance (Tables 3 and 4GoGo)].


View this table:
[in this window]
[in a new window]
 
Table 1. Absence from school in the past 30 days in relation to family structure, level of parental education, psychiatric symptoms, lifestyle, parental caring and control
 

View this table:
[in this window]
[in a new window]
 
Table 2. Absence from school in the past 30 days and lifetime use of alcohol, cigarettes, and illicit drugs
 

View this table:
[in this window]
[in a new window]
 
Table 3. Perception of academic performance in relation to family structure, level of parental education, psychiatric symptoms, lifestyle, parental caring and control
 

View this table:
[in this window]
[in a new window]
 
Table 4. Perception of academic performance and lifetime use of alcohol, cigarettes, and illicit drugs
 
These separate effects were analysed in two ways. First logistic discriminations were run using the Pc Carp package, discriminating between all four truancy groups and all three perceived performance groups. The F-values and design effects in the tables are based on this set of analyses. Second, logistic regressions were run with the dependent variables dichotomized. For this second analysis, truancy was dichotomized into no days missed and any days missed, while perceived performance was categorized as below average and the rest. These two sets of analyses produced virtually identical findings. Nearly all the predictor variables were significant beyond the P = 0.001 level in predicting the two dependent variables; e.g. having one or both parents absent is strongly associated with absence from school in the past 30 days. There were three exceptions: gender and parents setting rules of conduct were not significant in predicting absence from school, and lifetime alcohol intoxication was significant at the P = 0.05 level only in predicting perceived academic performance.

Tables 5 and 6GoGo set out a more detailed examination of the family structure variable in relation to truancy and perceived school performance respectively. For both dependent variables, it is clear that the main differences were between the group with both parents present and the other groups. When the former group was discarded from the analyses, no significant differences were found.


View this table:
[in this window]
[in a new window]
 
Table 5. Absence from school in the past 30 days in relation to family structure
 

View this table:
[in this window]
[in a new window]
 
Table 6. Perception of academic performance in relation to family structure
 
Next, all the predictor variables were entered together into logistic discriminations of truancy and perceived school performance. Once again two sets of analyses were run, in one of which the dependent variables were dichotomized. The results were very similar. Results from the analyses on the non-dichotomized variables are reported here. The results of this showed that there was overlap, with some predictor variables being unnecessary after others were entered. Absence from school was predicted by ‘sociable/delinquent lifestyle', ‘symptoms', ‘lack of hobbies' (all P < 0.001), ‘parental education', ‘cannabis use', ‘any drug bar cannabis' (all P < 0.01), ‘family structure’, ‘cigarette smoking’ (both P < 0.05). Gender, parental caring, parental control, alcohol intoxication, and use of solvents were not significant predictors. The average design effect was 1.73. Regarding perception of academic performance, the significant variables were ‘gender’, ‘parental education’, ‘lack of symptoms’, ‘hobbies’, ‘parental caring’, ‘lack of sociable/ delinquent lifestyle’ (all P < 0.001), ‘family structure’, ‘parental control’ and ‘low cigarette smoking’ (all P < 0.01). Use of cannabis, solvents, alcohol, and drugs other than cannabis were all non-predictors. The average design effect was 1.71. Possible problems of multi-collinearity were examined by running parametric versions of these discriminations. The minimum tolerance found in any of these other analyses was 0.45. Multi-collinearity is therefore unlikely to affect the solutions.

Finally analysis was run of perceived school performance entering the significant variables and also truancy (as three dummy variables) as predictors. Truancy was a highly significant predictor of poor perceived performance (P < 0.001) and all the other variables remained significant.


    DISCUSSION
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
Before reflecting on the main results, a word of caution must be entered. It was only possible to perform these analyses on a subsample (6409 out of 7722) of the total group surveyed. This was because the personality/symptom measures came at the end of the questionnaire and not all students managed to complete it in the time available. This may have biased the sample, meaning, in particular that slower working pupils would have been excluded. On some variables on which information was available for the whole 7722 group, it was clear that those failing to complete did differ from the rest. Notably, a disproportionate number from single-parent families failed to finish. However, it is probable that, had these respondents been included, the results would have been more, rather than less, significant.

Turning to the results, it is no surprise to find that the best predictor both of truancy from school and of perception of below average school performance was a ‘sociable/delinquent lifestyle’. This is similar to Jessor and Jessor's (1977) ‘general deviance’. Slightly more surprisingly, family structure still strongly predicted both variables, even when ‘sociable/delinquent lifestyle’ was entered. In other words, absence of one or both parents predicted (e.g.) absence from school, regardless of whether or not the respondent scored high on the lifestyle factor. Neither lack of parental caring nor lack of parental control predicted school absence directly, probably because of their association with ‘sociable/delinquent lifestyle’. However, caring and control did seem to be of direct importance in the students' perceptions of their school performance. One slightly unexpected result was the positive effect of having hobbies. Those who did have hobbies seemed to be absent from school less and to have a better perception of their school performance. Having symptoms of anxiety/depression was another highly significant predictor of both dependent variables, but the symptoms could, of course, be partly the results of non-attendance at school and poor schoolwork, rather than their causes. Finally, the relative lack of importance in relation to these analyses of psychoactive substance use was striking. Once the other variables were entered, solvents and alcohol intoxication predicted neither dependent variable. The use of cannabis failed to predict academic performance and was only just significant in predicting absence from school. Cigarette smoking, however, remained highly predictive of both variables and it may be because of the very strong association between cigarette smoking and cannabis use that the latter is so unimportant. (Cigarette smoking, it should be noted, is inversely associated with socioeconomic status.) Similarly, the ‘sociable/delinquent lifestyle’ variable may have also helped to account for the modest or non-existent significant findings for the drug-use variables.

To put it another way, these drug-use variables might well be best seen as part of a general syndrome of deviance or risk-taking behaviours (Plant and Plant, 1992Go; Plant, 1998aGo,bGo). However, it was also clear that other variables, in particular family structure, psychiatric symptoms, and pursuit of hobbies, may have played a part in truancy and perception of low academic performance among these UK teenagers.


    ACKNOWLEDGEMENTS
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
The authors are grateful to many colleagues for help and advice throughout the conduct of this survey. Mr Björn Hibell and Ms Barbro Andersson of the Swedish Council for Information on Alcohol and Other Drugs, Stockholm, played a major role in this exercise, which was conducted under the auspices of the Pompidou Group of the Council of Europe. Generous support for the UK part of this study was provided by the Gannochy Trust, the Alcohol Education and Research Council, the John M. Archer Charitable Trust, the Craignish Trust, the Hope Trust, the Miller Group Ltd, the Robertson Trust, and the Wates Foundation. The Alcohol and Health Research Centre, a registered charity, was originally established under the name of ‘Alcohol Research Group’. The work of the research team has been supported by the beverage alcohol industry, charities, government departments, research councils, health boards, the European Union, and the World Health Organization. Particular thanks are due to Allied Domecq plc, Diageo plc, the North British Distillery Company Ltd, and the PF Charitable Trust.


    FOOTNOTES
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
* Author to whom correspondence should be addressed. Back


    REFERENCES
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 ACKNOWLEDGEMENTS
 REFERENCES
 
Achenbach, T. (1991) Integrative Guide for the 1991 CBCL/4–18 YSR, and TRF Profiles. University of Vermont, Department of Psychiatry, Burlington, VT.

Fuller, W. A., Kennedy, W., Schnell, D., Sullivan, G. and Jin-Park, H. (1989) Manual for PcCarp. University of Iowa Ames, IA.

Grube, J. and Morgan, M. (1990) The structure of problem behaviours among Irish adolescents. British Journal of Addiction 85, 667–675.[ISI][Medline]

Hibbell, B., Andersson, B., Bjarnason, T., Kokkevi, A., Morgan, M. and Narusk, A (1997) The 1995 ESPAD Report: Alcohol and Other Drug Use Among Students in 26 European Countries. Swedish Council for Information on Alcohol and Other Drugs, Stockholm.

Jessor, R. and Jessor, S. (1977) Problem Behavior and Psychosocial Development: A Longitudinal Study of Youth. Academic Press, New York.

Miller, P. (1997) Family structure, personality, drinking, smoking and illicit drug use: a study of UK teenagers. Drug and Alcohol Dependence 45, 121–129.[ISI][Medline]

Miller, P. and Plant, M. (1996) Drinking, smoking and illicit drug use among 15 and 16 year olds in the United Kingdom. British Medical Journal 313, 394–397.[Abstract/Free Full Text]

Miller, P., Plant, M. L., Plant, M. A. and Duffy, J. (1995) Alcohol, tobacco, illicit drugs and sex: an analysis of risky behaviors among young adults. International Journal of Addiction 30, 239–258.[ISI][Medline]

Plant, M. A. (1998a) Young people and alcohol. In Young People and Mental Health, Aggleton, P., Hurry, J. and Warwick, P. eds. John Wiley, London.

Plant, M. A. (1998b) Learning by experiment. In Learning to Drink, Grant, M. ed. Taylor and Francis, Washington, DC.

Plant, M. A. and Plant, M. L. (1992) Risktakers: Alcohol, Drugs, Sex and Youth, London: Tavistock/Routledge.