Illicit drug use in English adolescent students–result of cumulative mediation analyses

ABSTRACT

Background

In England, drug use in young people increased significantly between 2014 and 2017. This upward trend continues despite implementation of drug use policies to reduce supply, possession and manufacture of illicit drugs. Taking the view that drug use is a learnt behaviour, the purpose of this paper is to evaluate whether social learning (SL) factors explain drug use in English adolescents across: a) nine regions b) by age (11 to 15 years) and c) by gender using the Social Structure Social Learning (SSSL) theory as a framework. This study addresses a gap in the literature on English adolescent students by identifying the strongest SL pathway to drug use (imitation, parental reinforcement, attitudes, peer association).

Methods

Cumulative mediation analyses were carried out on data from the Smoking Drinking Drug Use Survey 2016 (N = 12,051) on adolescents aged 11–15 years across England.

Results

The results show that imitation, peer association, attitudes and parental reinforcement mediate drug use at ages 12–14 and for some regions but not for gender.

Conclusion

Drug use is a socially learnt behaviour in adolescents students living in England.

KEYWORDS:

• Illicit drug
• adolescents
• imitation
• peers
• parental reinforcement

Introduction

Illicit drug use is a significant public health concern in the United Kingdom that affects society on a macro and micro level (Stead, Mackintosh, Eadie, & Hastings, 2001). Trade in illicit drugs is a multi-billion-pound industry and it remains very attractive to drug gangs and suppliers, who are constantly shifting and adapting to the market conditions. Despite strict laws and regulations controlling possession and supply of illicit drugs (McCambridge & Strang, 2005; Morgan et al., 2010), the United Kingdom is considered to have the largest and most accessible market for legal highs such as spice, bath salts, Kitkat, Kryptonite, in the whole of Europe (UNODC, 2013, 2020).

Why focus on adolescents?

Although very few children initiate illicit drug use before 8 years of age, the risk of experimentation is steep during adolescence for each year between the age of 10 and 18 (Bennett, 2014; L. Johnston, 2010; Johnston, O’Malley, Bachman, & Schulenberg, 2010; Sloboda et al., 2012). 5.9% of the total UK population, equating to approximately 3.4 million, are adolescents aged between 11 and 15 years (Office for National Statistics, 2021). 15% of 11 to 15 year olds in England alone have used an illicit drug at least once (Health and Social Care Information Centre, H, 2016). This statistic is supported by European research on drug use which shows that young people in the UK not only start taking drugs at an earlier age but are more likely to have consumed illegal drugs than their peers in the rest of Europe (EMCDDA, 2015). In England, drug use in young people increased significantly from 15% to 24% between 2014 and 2017 (NHS Digital, 2017) following a period of steady decline. Similarly, hospital admission data show that illicit drug-related mental and behavioral disorders admissions increased 4% from 840 in 2015/16 to 871 in 2016/17 for adolescents under 15 years (Health and Social Care Information Centre, H, 2016); and that the numbers of young people aged 15 years and under presenting for treatment after consumption of illicit drugs increased by 3% from 2015/16 compared to 2016/17 (Public Health England, 2017). Equivocally, there has been an increase in death rates of those under 20 years in the same time period due to accidental poisoning from illicit drugs from 46 (2016) to 51 (2017) (Office for National Statistics, 2018). Data from the National Drug Treatment Monitoring System (NDTMS) which collects information on young people in treatment for substance use below 18 years (Public Health England, 2017) also show that the number of children under the age of 14 years presenting for treatment has increased by 10%. Almost half of the young people attending the treatment clinics were below the age of 16 years and females in treatment were of a lower median age (15 years) than males (16 years).

We also now know that the social environment comprises a dynamic set of interactions between the environment and the individual factors such as age and gender (Bandura & Walters, 1977; Griffin & Botvin, 2010; Sutherland & Shepherd, 2001a, 2001b; Unlu et al., 2014; Verrill, 2005; Vogel et al., 2015; Whitesell et al., 2014; Winfree & Bernat, 1998). These interactions form social networks of parents, children, peers, etc., which according to Baler and Volkow (2011) are modulators of gene expression, cognition, emotion and brain function and development. Under this narrative, drug use in adolescents is very much interlinked with youth culture including, friends, peers and parents and should not be examined in isolation. This research applies the Social Structure Social Learning (SSSL) theory (Figure 1) which is an integrated conceptual theory that embodies these very interlinkages/interactions to examine drug use pathways for each age, gender and region. It builds upon an established classical theory – the social learning theory – that focuses on processes through which drug use is learnt which is empirically supported by a large body of evidence (Akers et al., 1979; Bandura & Walters, 1977; Krohn, 1999; Krohn et al., 2016; Matsueda, 1982; Pratt et al., 2010).

Figure 1. Social Structure Social Learning Theory Framework.

Aims and objectives

The following research questions were addressed in this study as follows: 1) Is there an association between social learning (parental reinforcement, imitation, peer association, attitudes) and drug use? 2) Is social structure (age, gender and region) associated with drug use in the last year? 3) Do social learning constructs (imitation, peer association, family reinforcement and attitudes to drugs) mediate the association between social structure (age, gender and region) on illicit drug use? For the purpose of this study only the first two of the Social Structure constructs, that is differential social location (age, gender) and differential social organization (region) were studied in relation to all four of the social learning variables.

Materials and methods

This research is a deductive theory testing approach delineated by core assumptions on ontology (realism), epistemology (positivism), human nature (determinism) and methodology (nomothetic). That is, the reality of harm from illicit drug use exists external to social actors that it can be prevented or reduced and can be increasingly known by accumulating more complete information (Guba, 1990). The study is based on a quantitative methodology involving secondary data from the most recent Smoking Drinking and Drug Use among Young People Survey (SDDS) dataset carried out in 2016 (SN: 8320). The SDSS 2016 was a cross-sectional national study based on a self-completion survey of 12,051 pupils aged between 11 and 15 years (Grade 7–11) from 177 secondary schools across England in the autumn term of 2016. Details of sampling and weighting can be obtained from UK data research.

Variables of interest

The variables used in the study have been summarized in the Table 1:

Table 1. Variables of interest.

Variable	Question	Original Categories	Code	Recoded Categories	Code
Drug Use	Used any drugs last year. Derived Variable from the dataset	Used drugs Not used	1 2 –8 -9	Used drugs in the last year Not used Don’t Know Missing	0 1
Gender	Are you a boy or girl?	Boy Girl Don’t know Not answered	1 2 –8 -9	Boy Girl Missing	0 1 -8,-9
Age	How old are you now? Derived Variable from dataset	11 12 13 14 15 Don’t know Not answered	11 12 13 14 15 –8 -9	11 12 13 14 15 Missing	11 12 13 14 15 -8,-9
Region	Derived from dataset	North East North West East Midlands Yorkshire & Humber West Midlands East of England London South East South West	1 2 3 4 5 6 7 8 9	North East North West East Midlands Yorkshire & Humber West Midlands East of England London South East South West	1 2 3 4 5 6 7 8 9
Peer Differential Association	How many people your own age do you think take drugs?	All of them Most, but not all About half Some of them None of them Not applicable Don’t know Not answered	1 >0 2 > 0 3 > 1 4 > 1 5 > 2 –1 > 2 –8 -9	Most or all Half or less None Missing	0 1 2 –9/-8
Attitude Cannabis	Do you think it is OK for someone your age to do the following? Try taking cannabis	It’s OK It’s not OK Don’t know Not answered	1 2 3 –9	It’s OK It’s not OK Don’t know Missing	0 2 1 –9
Attitude cocaine	Do you think it is OK for someone your age to do the following? Try taking cocaine	It’s OK It’s not OK Don’t know Not answered	1 2 3 –9	It’s OK It’s not OK Don’t know Missing	0 2 1 –9
Attitude glue	Do you think it is OK for someone your age to do the following? Try sniffing glue?	It’s OK It’s not OK Don’t know Not answered	1 2 3 –9	It’s OK It’s not OK Don’t know Missing	0 2 1 –9
Parental Differential reinforcement	How do you think your family would feel if you started taking drugs? Filter question.	Try to stop me Persuade me not to Do nothing Encourage me Don’t know Not answered Not applicable	1 > 0 2 > 1 3 > 2 4>-9 5 > 2 –9>-9 -1 > 2	Strongly disapprove Disapprove Neither approve or disapprove Missing	0 1 2 –9
Imitation	Why did you try drugs for the first time? (Derived from dataset)	Because friends were doing it No Not answered Not applicable	1 > 1 0 > 0 –9>-9 -1 > 0	Because friends were doing it Other reasons Not answered	0 1 –9

Data analysis

In the first part of the analyses, descriptive data analyses were carried out to understand the characteristics of the English adolescents (Table 2). The socio-demographic data included: age, region, gender of the sample and drug use in the last year for the 12,051 students in the dataset. After testing for frequencies, cross tabulation was carried out to analyze the patterns and trends between the subgroups in the independent (age, gender and region) and mediator variables (peer association, peer imitation, parental reinforcement and attitudes) with drug use. Following this, a series of mediation analyses were carried out as follows: Social Learning Model (Model 1), the first step involves using binomial logistic regression to test for association between the mediator variables (differential association, differential reinforcement, imitation, and attitude) and drug use Social Structure Model (Model 2), the second step involves testing for association between the independent variables (age, gender and region) and drug use using binomial logistic regression to obtain an estimate of the association between X (Social Structure) and Y (drug use); and the final step is to test the aggregate SSSL Mediation model (Model 3) using binomial regression. Multinomial regressions were first carried out between Social Structure (SS) and Social Learning (SL) variables to test for association between X (SS) and M (SL) separately.

Table 2. Descriptive Statistics.

Variable φc/(p-value)	Total cases & Missing cases	Category	Not Used Drugs Frequency/(%)	Used Drugs Frequency/ %
Gender*DrugsUsed (φc = .01 p = .26)	9137 (75.8%) 2914 (24.2%)	Boy (1) % Total %within Gender Girl (2) %Total %within Gender	3625 39.7% 83.2% 4017 44.0% 84.1%	733 8.0% 16.8% 762 8.3% 15.9%
Age 11–15*DrugsUsed (φc = .21 p < .001)	9176 (76.1%) 2875 (23.9%)	11 (1) %within Total %within Age 12 (2) %within Total %within Age 13 (3) %within Total %within Age 14 (4) %within Total %within Age 15 (5) %within Total %within Age	1261 13.7% 93.3% 1472 16.0% 91.0% 1801 19.6% 86.7% 1513 16.5% 80.7% 1623 17.7% 72.0%	91 1.0% 6.7% 145 1.6% 9.0% 276 3.0% 13.3% 362 3.9% 19.3% 632 6.9% 28.0%
Region*DrugsUsed (φc = .06 p < .001)	9302 (77.2%) 2749 (22.8%)	North East (1) %within total %within Region North West (2) %within Total %within Region York&Humber (3) %within Total %within Region East Midlands (4) %within Total %within Region West Midlands (5) %within Total %within Region East England (6) %within Total %within Region London (7) %within Total %within Region South East (8) %within Total %within Region South West (9) %within Total %within Region	1332 14.3% 84.4% 1031 11.1% 80.9% 1222 13.1% 85.2% 617 6.6% 83.7% 551 5.9% 84.1% 972 10.4% 83.7% 237 2.5% 77.2% 803 8.6% 81.6% 1008 10.8% 86.2%	247 2.7% 15.6% 244 2.6% 19.1% 212 2.3% 14.8% 120 1.3% 16.3% 104 1.1% 15.9% 189 2.0% 16.3% 70 0.8% 22.8% 181 1.9% 18.4% 162 1.7% 13.8%
Attitude Cannabis*DrugsUsed (φc = .45 p < .001)	9121 (75.7%) 2930 (24.3%)	It’s Ok to try (1) %within Total %within Att. Can It’s not OK (2) %within Total %within Att. Can Don’t Know (3) %within Total %within Att. Can	391 4.3% 38.7% 6699 73.4% 90.8% 555 6.1% 75.7%	620 6.8% 61.3% 678 7.4% 9.2% 178 2.0% 24.3%
Attitude Cocaine*DrugsUsed (φc = .22 p < .001)	9087 (75.4%) 2964 (24.6%)	It’s Ok to try (1) %within Total %within Att. Coc It’s not OK (2) %within Total %within Att. Coc Don’t Know (3) %within Total %within Att. Coc	100 1.1% 39.4% 7087 78.0% 85.8% 426 4.7% 74.6%	154 1.7% 60.6% 1175 12.9% 14.2% 145 1.6% 35.4%
Attitude Glue*DrugsUsed (φc = .26 p < .001)	9149 (75.9%) 2902 (24.1%)	It’s Ok to try (1) %within Total %within Att. Glue It’s not OK (2) %within Total %within Att. Glue Don’t Know (3) %within Total %within Att.Glue	392 4.3% 53.6% 6262 68.4% 87.9% 1011 11.1% 78.4%	340 3.7% 46.4% 866 9.5% 12.1% 278 3.0% 21.6%
Imitation *DrugsUsed (φc = .25 p < .001)	9044 (75.0%) 3007 (25.0%)	Friends were taking (1) %within Total %within Imitation Other reason (2) %within Total %within Imitation	30 0.3% 18.5% 7628 84.3% 85.9%	132 1.5% 81.5% 1254 13.9% 14.1%
Peer Association *DrugsUsed (φc = .29 p < .001)	4699 (39.0%) 7352 (61.0%)	Most or all (1) %within Total %within peer assoc. Half or less (2) %within Total %within peer assoc. None(3) %within Total %within peer assoc.	101 2.1% 45.5% 2418 51.5% 80.7% 1410 30.0% 95.3%	121 2.6% 54.5% 580 12.3% 19.3% 69 1.5% 4.7%
Parental Reinforcement *DrugsUsed (φc = .21 p < .00)	8871 (73.6%) 3180 (26.4%)	Strongly disapprove (1) %within Total %within Parental Reinf. Disapprove (2) %within Total %within Parental Reinf. Neither (3) %within Total %within Parental Reinf.	2928 33.0% 93.4% 321 3.6% 89.9% 4136 46.6% 76.9%	208 2.3% 6.6% 36 0.4% 10.1% 1242 14.0% 23.1%
Total (N)	12,051

The models

1. Social Learning Model – Model 1 (Figure 2)

   Figure 2. (SL Model 1).

   

2. Social Structure Model – Model 2 (Figure 3)

   Figure 3. (SS Model 1).

   

3. Aggregate SSSL Mediation model – Model 3 (Figure 4)

   Figure 4. (SSSL Model 3).

There must be a significant path of association from X to M and then from M to Y (Model 2) for mediation to be considered between X and Y by M (Model 3). Therefore, multinomial (attitude and imitation) and ordinal (peer association and parental reinforcement) regressions were first carried out between the SS and SL variables to test for association between X and M before running model 3.

Results

Reported next are the results of the mediation analyses.

Model 1: social structure model

The logistic regression model was statistically significant, X² (13) = 438.3 p < .001 and the model explained 8% (Nagelkerke R²) of the variation in drug use in the last year and correctly classified 84% of the cases (Table 3). The model fit, Wald X² statistic shows a significant contribution by region (X² =30.2, p < .001), age (X² = 375.7 p < .001) but not by gender (X² = .88, NS (p < .349)). Increasing age was associated with an increased likelihood of drug use in the last year. More specifically, 11 year old adolescents were 81% less likely to have used drugs in the last year than their 15 year old counterparts. Adolescents in the London were 103% were more likely to have used drugs in the last year as were those living in North West (44%), East of England (30%) and South East (45%) than those living in the South West of the country. Gender is not significantly associated with drug use in this model.

Table 3. Social Structure Model.

	Social Structure Model
	Exp(B)/Odds Ratio	Significance (p value)	B	S.E
11 years	.19	.00	−1.69	.12
12 years	.25	.00	−1.38	.10
13 years	.38	.00	−.978	.08
14 years	.61	.00	−.50	.08
15 years	-	-	-	-
Female	1.06	.35	.06	.06
Male	-	-	-	-
North East (1)	1.18	.16	.16	.11
North West (2)	1.44	.00	.37	.12
York & Humber (3)	1.12	.32	.12	.12
East Midlands (4)	1.30	.06	.26	.14
West Midlands (5)	1.06	.70	.06	.14
East England (6)	1.30	.03	.26	.12
London (7)	2.03	.00	.71	.17
South East (8)	1.45	.00	.37	.12
South West (9)	-	-	-	-
Model Fit
Chi-square	438.28 (df. 13)	.00
Nagelkerke Pseudo R^2	.08

^*(values rounded off to 2 dp)

Model 2: social learning model

The logistic regression model was statistically significant, X² (11) = 1626.60 p < .001 and the model explained 59% (Nagelkerke R²) of the variance in drug use and correctly classified 92% of the cases (Table 4). The model fit Wald X² statistic show that attitudes to cannabis (X² = 189.63 df = 2), p < .001), cocaine (X² = 7.34 df(2), p < .05) and glue use (X² = 19.58 df(2), p < .001), peer association (X² = 81.30 df(2) p≤<.001), family reinforcement (X² = 418.43 df(2), p < .001) and imitation (X² = 57.09 df(1), p < .001), contributed significantly to the model. All social learning measures were significantly associated with drug use.

Table 4. Social Learning Model.

	Social Learning Model
	Exp(B)/ Odds Ratio	Significance (p value)	B	S.E
Attitude Cannabis- OK	9.78	.00	2.28	.17
Attitude Cannabis- Not sure	4.22	.00	1.44	.23
Attitude Cocaine-OK	1.08	.81	.08	.33
Attitude Cocaine- Not sure	.47	.01	−.76	.29
Attitude Glue- OK	2.13	.00	.76	.19
Attitude Glue Not sure	.81	.28	−.21	.20
Fam Reinf. Strongly disapprove	.06	.00	−2.75	.14
Fam Reinf disapprove	.07	.00	−2.74	.27
Peer Assoc. Most or all	14.40	.00	2.67	.30
Peer Assoc. Some or less than half	3.47	.00	1.24	.20
Imitation- Yes	8.46	.00	2.14	.28
Model Fit
Chi-square	1626.61 df (11)	p ≤ .001
Nagelkerke Pseudo R^2	.59

^*(values rounded off to 2 dp)

The peer association variable has the strongest influence on drug use compared to any of the other social learning variables. Adolescents who perceived that most or all of their peers own age were taking drugs were themselves 14 times more likely to have taken drugs in the last year compared to adolescents who perceived that none of their peers took drugs. Consistent with previous research and the correlation matrix results above, drug use is highly correlated with peer association (Akers, 2011), more than any other social learning variable.

Adolescents with a positive attitude to cannabis use were 9.8 times more likely to have used drugs in the last year compared to those with a negative attitude. Furthermore, those who had ambivalent attitudes to cocaine had lower odds of drug use in the last year. Comparatively, those with positive attitudes to glue who had twice the odds of having used drugs in the last year. This data indicates that the acceptability of cannabis is the greatest followed by glue and cocaine.

Those who reported having taken drugs for the first time because of imitating friends, had 8.5 times the odds of drug use in the last year. Adolescents who reported strong parental disapproval (stop me from taking drugs) and disapproval (persuade me not to take drugs) were 93% less likely to have used drugs in the last year (OR = .07).

Model 3: aggregate social structure social learning model

The logistic regression model was statistically significant, X² (24) = 2283.37, p < .001. The social learning model explained 42% (Nagelkerke R²) of the variance in drug use and correctly classified 89% of the cases (Table 5). The results indicate that the while the aggregate model is an improvement on the SS model (Model 1), it is not better than the SL model (Model 2).

Table 5. Cumulative Social Structure Social Learning Model.

	Model1 SS–>DU		Model 2 SL–>DU		Model 3 SS–>SL–>DU
Drugs used Last Year	Exp(B)	Sig	Exp(B)	Sig	Exp(B)	Sig	B	SE
11 years	.19	.00			.50	.02	−.70	.30
12 years	.25	.00			.66	.08	−.42	.24
13 years	.38	.00			.99	.96	−.01	.20
14 years	.61	.00			.83	.29	−.19	.18
15 years	-	-			-	-	-	-
Male	1.06	.35			1.13	.39	.12	.14
Female	-	-			-	-	-	-
North East (1)	1.18	.16			1.22	.43	.20	.25
North West (2)	1.44	.00			.94	.83	−.06	.26
York&Humber (3)	1.12	.32			.97	.91	−.03	.26
East Midlands (4)	1.30	.06			.86	.64	−.15	.33
West Midlands (5)	1.06	.70			1.38	.29	.33	.31
East England (6)	1.30	.03			.72	.24	−.33	.28
London (7)	2.03	.00			1.93	.07	.66	.36
South East (8)	1.45	.00			1.00	.99	−.00	.28
South West (9)	-	-			-	-	-	-
Attitude Cannabis- OK			9.78	.00	8.57	.00	2.15	.19
Attitude Cannabis- Don’t know			4.22	.00	3.92	.00	1.37	.24
Attitude Cocaine-OK			1.08	.81	1.18	.64	.17	.35
Attitude Cocaine- Don’t know			.47	.01	.46	.01	−.77	.30
Attitude Glue- OK			2.13	.00	2.00	.00	.71	.20
Attitude Glue Don’t know			.81	.28	.85	.40	−.17	.20
Fam Reinf. Strongly disapprove			.06	.00	.06	.00	−2.78	.14
Fam Reinf. Strongly disapprove			.07	.00	.07	.00	−2.73	.27
Peer Assoc. All or most			14.40	.00	12.04	.00	2.49	.31
Peer Assoc. Half or some			3.47	.00	3.00	.00	1.10	.27
Imitation			8.46	.00	8.40	.00	2.13	.28
Model Fit
Chi-square	1610.64 df(24)	p ≤ .001
Nagelkerke Pseudo R^2	.600

^*(values rounded off to 2 dp)

As expected, the odds of drug use became insignificant for all social structural variables once the social learning variables were introduced except for 11 year old adolescents. 11 year old adolescents are now 50% (compared to previously being only 20%) more likely to have used drugs in the last year after the introduction of social learning variables compared to 15 year old adolescents.

Similarly, the odds of drug use in the North West, East of England, London and South East regions became insignificant after the introduction of social learning variables. Drug use by adolescents in the having the perception that peers were using drugs mediated drug use in the North West (both categories) and London (half or less). Drug use was also mediated in the North West by positive attitudes to glue use. As there were no associations between East England and South East with social learning variables, mediation cannot be claimed.

Gender as with the remaining other regions remained insignificant after the introduction of social learning variables.

Discussion

The conceptual underpinning in this study is the SSSL, which posits that an adolescent’s age, gender or the location in which they live in affects their chances of learning drug use through four social learning processes which are imitation, peer association, parental reinforcement and attitudes to drugs.

Social learning model

Imitation of friends and peer association

Having the perception that most or all peers take drugs was the strongest social learning variables associated with drug use. Perceiving that most or all peers were taking drugs was associated 14 times with the odds of drug use in the last year and perception of having less than half of peers taking drugs was associated with only 3.5 times the risk of taking drugs. Adolescents who reported imitation were 8.5 times likely to have used drugs in the last year compared to those who did not imitate their friends. The findings corroborate the findings of Young and Weerman (2013; Young, Rebellon, Barnes, & Weerman, 2014) that overestimating peer use has the strongest effect of delinquency and this is particularly true of adolescents who value social approval. Researchers employing the SSSL framework have also found perception of peer drug use to be the most robust and consistent predictor of drug use and when pitted against imitation, this construct performs much better (Cooper & Klein, 2018; Duncan et al., 2014; Giannotta et al., 2014; Holland-Davis, 2006; Hwang & Akers, 2006; Solakoglu & Yuksek, 2020). Imitation on the other hand, having previously been one of the weaker social learning pathways in previous studies using the SSSL framework (Hwang & Akers, 2006; Kim, 2010; Lanza-Kaduce & Capece, 2003; Lee, Akers, & Borg, 2004) was found to be a strong social learning pathway in this study. This could be because the studies measured imitation indirectly in that they measured imitation through observation of role models taking drugs and stopped short of asking about the respondent’s own drug use as a result of observing drug use (Cooper & Klein, 2018; Kim, 2010; Kim et al., 2013).

Attitudes to cannabis, cocaine and glue

This is the first study to have tested the effects of attitudes to three different types of drugs separately in an SSSL framework. Not only is cocaine the most difficult to get hold of comparatively but it is thought of as “hard drug” by users which implies that it is addictive, needs to be injected and is taken for dramatic/profound effects; cannabis is regarded as being in the “soft drug” category and is a suppressant/ depressant which means it has the opposite effect to cocaine (Fothergill et al., 2009; Palamar, 2014; Schaefer et al., 2015; Seddon, 2008). Class A drugs such as cocaine according to the Misuse of Drugs Act 1971 are the most harmful of the three possible drug classes; this class also carries the most severe penalties for possession, supply and production. Cannabis is a class B drug and carries less severe penalties than class A drugs. Glue on the other hand is not a scheduled or illegal substance, is known to give users a high upon inhalation and is the easiest to access of all the three.

In this study, positive attitudes to cannabis had a stronger association to drug use than positive attitudes to glue. The association between positive attitudes to cocaine and drug use was insignificant. The most important point to note here is that having a positive attitude toward and being ambivalent to trying both cannabis and glue were risk factors for drug use, that is, they increased the odds of taking drugs compared to having a negative attitude. 6.3% of the respondents reported that they did not know if it was acceptable to try cocaine, 14.1% did not know if it was acceptable to try glue and 8.1% did not know if it was acceptable to try cannabis.

Parental reinforcement

The results showed that both measures of parental disapproval: “strong parental disapproval” and “disapproval” had significant associations with drug use. 2.3% of adolescents reported strong parental disapproval and drug use in the last year compared to 0.4% who reported less authoritative parental disapproval and drug use. However, 14% of adolescents who reported that their parents would neither disapprove nor approve of their drug use had taken drugs in the last year. Corroborating the first set of findings from the descriptive analyses, is research showing that high levels of authoritarian parenting control and restrictions (Becoña et al., 2013; Calafat, García, Juan, Becoña, & Fernández-Hermida, 2014) tend to exacerbate adolescent substance.

Social structure (SS), SSSL models

Although the descriptive data analysis revealed that more boys than girls had used drugs in the last year (albeit the difference was only 0.3%), the association with drug use was not significant as evident in the crosstabulation, correlation matrix and model 1 regression analysis. That is there was no difference between males and females for drug use.

The results of the cumulative analysis show that drug use was mediated by social learning variables for all ages except for age 11 where drug use appears to be moderated, for some regions but not for gender.

Conclusions

The first part of the paper was able to demonstrate an association between a) social learning (parental reinforcement, imitation, peer association, attitudes) and drug use b) social structure (age, gender and region) and drug use and to an extent c) that imitation, peer association, family reinforcement and attitudes to drugs mediate the association between age (12–14), some regions (but not gender) and illicit drug use.

Limitations

The findings of this study should be viewed in light of several limitations. First, the SDDS 2016 survey is based on self-report measures and on self-incriminating behaviors that can generate methodological response bias and inaccurate reporting (Delaney-Black et al., 2010; Percy et al., 2005; Williams & Nowatzki, 2005). Second, this study employs secondary data analysis which means that the data is situational and not collected for the intents and purposes of this study (M. P. Johnston, 2017); however, the dataset was of considerable breadth and also contained factors known to be associated with drug use in young people (Koziol & Arthur, 2012). Third, the results pertaining to the region London should be interpreted with caution due to the low response rate. Finally, what these analyses do not show are the specific social learning mediators or pathways to drug use for each of the social structural factors (age, gender and region).

Disclosure statement

No potential conflict of interest was reported by the author(s).