The role of self-control and cognitive functioning in educational inequalities in adolescent smoking and binge drinking

The current study used the data as described in Junger and Van Kampen [29]. The study was based on a convenience sample consisting of 201 adolescents between 15 and 20 years of age. Ten individuals were excluded due to missing values, which resulted in a study population of 191. Surveys and cognitive tests were performed in 2008, during school hours. Participants were enrolled in eight different schools, located in cities in the centre and south of the Netherlands (Eindhoven, Sleeuwijk, Culemborg, Arnhem, Utrecht, Huizen, and Oosterhout). Schools provided between 8 and 52 respondents (25 on average).

The cognitive tests were performed in either a separate room or a quiet area with three to six students at a time, and were supervised by at least one researcher. Participants first performed the reaction times test, followed by the Raven’s Standard Progressive Matrices, the Corsi block-tapping task, the self-control questionnaire, and finally the general survey (see below). More details on the procedure are published elsewhere [29].

Participation by respondents was entirely voluntary and with minimal risks of any kind, and does therefore comply with the Declaration of Helsinki. The questionnaires and simple tests did not entail observation methods that may be intrusive to subjects. This study does therefore not fall under the Dutch ‘law for medical research’ and ethical approval and active parental consent were therefore not required. To guarantee an ethical procedure, we followed four steps. First, permission for the implementation of the study was implied by the ethical committee of the Faculty of Social Science (FSS) of Utrecht University (which is a former employer of the third author). Second, permission for data collection was requested at each school. For this, all schools discussed the study with their School Board, a board that also included parent representatives. Third, schools informed parents about the study and offered parents the opportunity to opt-out if they disagreed with the participation of their child (passive consent). Fourth, before participating, students were asked to sign a consent form for the use of the data, based on an introductory letter and a brief oral description of the purpose of the study.

Binge drinking (six or more drinks on one occasion) and smoking frequency were defined in the general survey as ‘binge drinking/smoking once a month or less’ or ‘more than once a month’, with the latter coded as one. Sensitivity analyses were performed with the measures ‘ever binge drinking/smoking' (vs. never) and ‘at least weekly smoking’ (vs. less than weekly). The prevalence of weekly binge drinking was too low to analyse it as an outcome.

Educational track measures the level of difficulty of the school curriculum and the lengths of the secondary school career. Schools in the Netherlands may combine different tracks in one building, or offer one specific educational track. In our sample, four schools combined all three tracks, three schools provided only the low track, and one school was a high track school. Respondents were divided into three educational tracks: low, midlevel, and high educational track. Low educational track represents (pre-preparatory) vocational secondary education (provides access ISCED level 3–4), midlevel educational track represents intermediate general education (provides access to ISCED level 5–6) and high educational track represents preparatory university education (provides access to ISCED level 6–8) [30]. High educational track was coded as the reference value.

Self-control was measured with the Dutch Version of the Self-control Scale [21]. This questionnaire consisted of 36 statements (e.g. “I do a lot of things without thinking”) rated on a 5-point Likert scale, ranging from ‘not at all like me’ to ‘very much like me’. Because the relationship between self-control and substance use was examined, we excluded 6 health-related items (“I do certain things that are bad for me, if they are fun”, “I refuse things that are bad for me”, “I can resist temptations well”, “I participate in healthy activities”, “I eat healthy”, “Sometimes I abuse drugs or alcohol”) in order to prevent that self-control directly measured health behaviour. Negatively formulated items were reverse-scored. Cronbach’s alpha for the 30 remaining items was 0.84 indicating a high internal reliability. A sum score was calculated with a maximum of 150. A higher score indicated a stronger self-control.

Cognitive functioning was measured using three variables; cognitive ability, reaction time and memory span [22].

Cognitive ability was measured by the Raven Standard Progressive Matrices (RSPM) [31]. This test measures general cognitive ability (also known as Spearman’s g), the ability to solve problems and understand the relationship of various concepts without using language. The RSPM consisted of 60 items in which participants had to select the missing part (6–8 options) of a sequence of figures. Every correct item accounted for one point resulting in a maximum score of 60. Higher scores indicated higher cognitive ability. The RSPM is a valid and reliable measure for cognitive ability [31].

Reaction time was measured with the Simple Reaction Time (SRT) test. The test measured the time between a circle turning green and the subject pushing a button in reaction to the change in colour. The SRT test tests attention and motor speed [32]. There were two blocks, both consisting of 24 trials. The reaction time for all 48 trials was summed and the median of this sum score was used as the reaction time in the analysis. A lower score corresponds to a better cognitive functioning.

Memory span was measured using the Corsi block tapping test. The test consists of multiple trials in which participants were asked to recall a sequence in forwards and backwards order. The number of items in the sequence was increased with each trial. The longest sequence length that was correctly recalled at least twice was multiplied by the number of correct trials, resulting in two memory span scores (forwards and backwards). The maximum score was 162. A higher score means better memory span. The Corsi test is a valid measure for short term memory and executive control [33].

We used logistic regression models to study the associations of standardised self-control and cognitive functioning variables with smoking and binge drinking, respectively. We applied crude and adjusted models; the adjusted models controlled for age, gender, social disadvantage and educational track. Self-control and cognitive functioning variables were standardized, to enable comparison between these variables. Correlations between self-control and cognitive functioning variables were overall weak, ranging from −0.28 to 0.27.

The association between educational track and smoking and binge drinking, respectively, was analysed using four logistic regression models. Model 1 included educational track, age, gender and social disadvantage. In Model 2 self-control was added and in Model 3 cognitive ability was added. Finally, Model 4 included reaction time and memory span (forwards and backwards). The fit of the models was compared using likelihood ratio tests, in order to determine whether the addition of variables improved the model. In a sensitivity analysis, the explanatory variables were added to Model 1 separately. We found that this alternative approach did not change our conclusions with regards to the extent to which the model fit changed for each of these variables.

The regression analyses were repeated for ever binge drinking/smoking and for weekly smoking. Statistical analyses were performed in Stata version 14.2.

Smoking and binge drinking were more prevalent in students in the low educational track compared to those in the high educational track. The association between educational track and binge drinking was slightly reduced after adjustment for self-control and cognitive ability, but remained unchanged with all the cognitive functioning variables included. The association between educational track and smoking was reduced to a greater extent after adjustment for self-control and cognitive functioning, especially cognitive ability. However, inequalities in smoking remained substantial.

The study used a convenience sample of eight schools. The study population may therefore not be representative for the general Dutch adolescent population. For example, ethnic minority groups were underrepresented. However, we included schools from cities and towns from different parts of the country, that had a mean income that was comparable with the Netherlands as a whole. Also, the three main educational tracks in the Dutch educational system were all included. Studies with larger and more diverse samples are needed to assess the generalizability of our findings for the Netherlands.

The cross-sectional study design does not provide information to assess the causality of the observed associations. As self-control and cognitive ability develop in early childhood and substance use is initiated later in life [34, 35], it is likely that self-control and cognitive ability predict substance use. A longitudinal study from New-Zealand found that childhood self-control predicts substance dependence in adulthood [36]. However, binge drinking may affect the cognitive performance of young people [37] and in a sample of adult smokers, smoking seemed to enhance self-control [38]. Moreover, a twin study in Finnish adults showed that smoking is negatively related to long-term labour market outcomes [39]. Reverse causation or bi-directionality of associations can therefore not be ruled out. Moreover, we cannot rule out that drinking and smoking are only indirectly related to self-control and cognitive functioning, due to common underlying factors related to these concepts, such as parenting practices or personality traits that were not measured in the data.

Smoking and binge drinking were self-reported. Reporting of socially desirable answers may lead to an underestimation of the prevalence of smoking and drinking [40]. However, in our study, students were assured that the survey was anonymous and that neither teachers nor parents would be informed of individual answers. Moreover, the dichotomous classifications of substance use that we applied in this paper are less vulnerable to reporting errors than more detailed classifications. Likewise, self-control was measured via a self-report survey and may therefore be subject to reporting bias. However, a in a meta-analysis of 236 studies, self-report questionnaires for self-control had a higher internal validity than delay of gratification tasks and executive function tasks [41].

We found that self-control was inversely associated with smoking and binge drinking, also after controlling for socio-demographic variables and cognitive functioning. Previous studies also found an inverse relationship between self-control and smoking and binge drinking [23‐25]. For example, a longitudinal study by King et al. found that adolescents who had developed self-control problems over time were more likely to report smoking and binge drinking [25].

Cognitive ability was associated with smoking and educational track, and therefore weakened the association between educational track and smoking. However, this was not found for binge drinking. This might be due to the difference in the health beliefs of alcohol use and smoking in Western society [42]. While tobacco smoking is nowadays viewed by most adolescents as a health hazard and irresponsible behaviour, alcohol use is considered by many as a normal part of social life [42, 43]. Adolescents with higher intelligence or in high educational tracks may therefore be more inclined to protect their health against smoking than against alcohol use [43].

In our data, reaction time and memory span did not play a significant role in educational inequalities in smoking and binge drinking. These measurements for cognitive functioning have rarely been used in alcohol and tobacco research and their association with substance use may be converse to what we expected. Binge drinking during adolescence has a negative influence on memory span tests among females, but a positive influence among males [37]. Moreover, young adults who were heavy drinkers had faster reaction times than non-heavy drinkers [44]. Further studies should elucidate the nature of these unexpected relationships.

Self-control and cognitive functioning did not play a role in differences between educational tracks in binge drinking, and played a minor role in differences in smoking. Various other explanations have been developed for educational differences in substance use. Differences may originate from variations in social context and socialization processes, with those in lower tracks experiencing greater peer pressure to start using alcohol, tobacco, and other substances [45, 46]. High prevalence rates of smoking and binge drinking within the lower educational track may sustain itself because of its direct effect on social norms and peer pressure, resulting in freshmen adopting the view that drinking and smoking are normal behaviours.

School environment may also be an important factor to take into consideration. Stronger school policies may be associated with lower adolescent smoking rates [47] and lower adolescent binge drinking rates, especially among adolescents with a lower school attachment [48]. Moreover, parental substance use and parenting style may also influence their offspring’s substance use [49, 50]. Parenting styles characterized by warmth, care and positive emotional attachment lower the odds of adolescent substance use [50] and may be more common in some educational groups than others [51, 52].

Some authors argue that educational differences in smoking and binge drinking are entrenched in inequalities in society at large. These inequalities already start by differentiating young people into educational tracks at an early age [53, 54]. According to Elstad [15] the unpromising social prospects of those in lower tracks may lead to higher drinking and smoking prevalence due to “more need for stress-alleviating behaviours, less interest in the future … attempts to compensate lack of recognition in school by excelling in alternative social fields, and deliberate opposition to social authorities because of the experience of being rejected by them”.