Characteristics of non-drinking adolescents: a longitudinal Swedish study

ABSTRACT

This study investigated whether non-drinking adolescents (n = 195), with no previous experience with alcohol, differed over time from drinking adolescents (n = 406). Potential differences in psychological health (mental well-being, psychosomatic symptoms, conduct problems), social interaction frequencies (new friends, time spent with friends and social interaction through a screen device), psychoactive substance use, and positive attitudes towards alcohol was investigated. Additionally, the study examined whether these attributes in 9th grade could predict total abstinence two years later. Non-drinkers were consistently characterized by fewer social interactions, less positive attitudes towards alcohol, a lower probability of using other drugs, and fewer conduct problems from 14 to 17 years, which may imply a more introverted personality function. This study contributes insights into predicting factors linked to alcohol abstinence in adolescence, particularly the characteristics of non-drinking adolescents in a society where a declining trend of alcohol consumption is emerging.

KEYWORDS:

• Alcohol use
• psychoactive substance use
• non-drinking adolescents
• psychological health
• social interactions
• positive attitudes towards alcohol

Alcohol consumption among Swedish adolescents has been decreasing since the new millennium began (Pennay et al., 2018), and studies have found a decline in alcohol intake regardless of sociodemographic and cultural contexts (Brunborg et al., 2014; Norström & Raninen, 2018). Additionally, non-drinking adolescents are becoming increasingly common in Swedish society (Norström & Svensson, 2014; Raninen et al., 2014). While 90% of 9th graders consumed alcohol 50 years ago, only 38% of 9th graders consumed alcohol in 2022 (Zetterqvist, 2022). Outside Sweden, the decline in adolescent alcohol consumption has been noted in research from different parts of the world (de Looze et al., 2015; Kraus et al., 2018; World Health Organization, 2018). However, research has so far not distinguished underlying factors influencing this trend (Carlson, 2019; Pennay et al., 2018). To find explanations, several hypotheses have been empirically tested (Bhattacharya, 2016). For example, Gripe et al. (2018) wondered whether alcohol had been replaced with cannabis but could not see any such association with declining alcohol consumption in the Swedish adolescent population. Globally, the most frequently investigated potential influences in relation to the phenomenon were, according to Pape et al. (2018), technology, immigration, parenting style changes, changes in the social/cultural climate, political changes, and economic changes.

The hypothesis regarding the impact of technology/social media and computer gaming on adolescent drinking patterns assumes that adolescents are less exposed to real-life interactions (involving alcohol consumption) due to increased socializing/gaming on internet-based channels, and that this significantly affects alcohol consumption (Pape et al., 2018). One may argue that this is logical and merits exploration, as alcohol consumption among adolescents decreased at the same time as social media networking and video/computer gaming increased (Bjereld et al., 2017; de Looze et al., 2015). Studies focusing mainly on computer/video gamers have found mixed support for this hypothesis (Coëffec et al., 2015; Epstein, 2011; Halkjelsvik et al., 2021; Van Rooij et al., 2014). For example, Larm et al. (2019) found a positive cross-sectional correlation between weekend gaming and non-drinking among Swedish boys in 9th grade, but the association was not evident over a four-year period. Also, Van Rooij et al. (2014) suggested that boys reporting high ‘problematic’ video gaming were twice as likely to use alcohol, nicotine, or cannabis than those who did not play at this problematic level.

Other studies have suggested that social networking sites contribute to adolescent alcohol consumption, meaning that they promote underage drinking (Brunborg et al., 2017; Gommans et al., 2015; Kraus, 2016).

Rogne et al. (2019) studied immigration in relation to adolescent drinking patterns. They found that immigration explained one-fifth of the decline in heavy episodic drinking among adolescents in Norway, but the natives’ drinking pattern mattered more than the immigrants’. Similarly, Svensson and Andersson (2016) could not find an association between immigration and the increase in non-drinking adolescents in the Swedish adolescent population. Beyond this, few studies have empirically investigated the immigration hypothesis, and publicized results do not indicate reliable support for it, meaning that immigration does not seem to have had a significant contribution to the declining trend of adolescent drinking (Pape et al., 2018).

Stricter parenting styles regarding underage drinking have been associated with less inebriation among adolescents in studies by Sharmin et al. (2017) and Yap et al. (2017). Other research has also shown that adolescents who spent more time with their parents drank less (Kim et al., 2019), although Raninen and Livingston (2018) did not find support for this. Overall, this parenting hypothesis does not solely explain the general downward trend in adolescent alcohol consumption, according to Bhattacharya (2016) and Pape et al. (2018). Additional hypotheses from around the world have considered that political, economic, and policy-driven changes, such as age limits and stricter laws (Andersen et al., 2014; Pennay et al., 2015; Trolldal et al., 2020), explain the downward trend in adolescent alcohol consumption. In Sweden, the legal drinking age is 18 years. The government-owned company with statutory monopoly in Sweden (Systembolaget) sells alcoholic beverages with an alcohol content exceeding 3.5 by volume. Systembolaget’s (2021) social mission includes disseminating information about the risks of alcohol and its damaging impacts on adolescent brains. Systembolaget also cooperates with the police in preventing illegal trade in alcohol and reaches out to parents with tools and arguments to persuade them not to provide their children with alcohol.

In the Swedish context, in addition to the trend towards more non-drinkers, adolescents are starting to drink at a later age than before (Guttormsson & Zetterqvist, 2019). In 2021, 71% of high school girls and 64% of high school boys were classified as alcohol consumers; in 2004, the proportion was around 90% for both sexes (Zetterqvist, 2022). In line with other countries (Keyes et al., 2012; Livingston, 2014), this indicates a potential explanation that the age of drinking onset has increased over the past 20 years among 9th graders and other high school students in Sweden (Zetterqvist, 2022). In addition, Gripe (2013) suggested that delayed onset, in parallel with changes in the normative climate, contributed to the decreasing trend in adolescent alcohol consumption seen in Sweden. Törrönen et al. (2019) interviewed Swedish adolescents about alcohol consumption and the age of drinking onset; they suggested that alcohol consumption was no longer as strong a symbol of growing up, and that adolescents were more reflective and responsible, in comparison with earlier generations.

Previous research on non-drinking Swedish adolescents has suggested several main social attributes characteristically associated with non-drinkers. Generally, in comparison with drinkers, they were said to have more problems in social relationships (Kivimäki et al., 2014), feel less close to their friends (Lund & Scheffels, 2019), have fewer friends (Kivimäki et al., 2014), and struggle more to make new friends (Hoel et al., 2004; Larm et al., 2018). Demant and Järvinen (2011) also proposed that non-drinkers differed from drinkers in various social aspects, in that they did not take part in the current normative social climate. Previous research has also indicated that non-drinkers perform better in school and may experience certain health benefits, such as high well-being and high life satisfaction (Larm et al., 2018). However, recent research (Raninen & Livingston, 2018) has questioned these earlier theories and assumptions regarding the characteristics of non-drinking adolescents. A recent cross-sectional study by Raninen et al. (2021) proposes that non-drinking Swedish adolescent 9th graders do not differ in social abilities from drinkers and instead manifest more prosocial behaviour than do drinkers, which does not confirm previous studies (Hoel et al., 2004; Larm et al., 2018).

Prosocial behaviour is described as moral behaviour characterized by sharing, helping, and cooperation, and is fundamental for engaging social relations (Ding et al., 2018). Raninen et al. (2021) therefore argued that alcohol consumption as a social marker among 9th graders in Sweden has shifted, meaning that drinking was no longer the norm (as it was in previous generations) and that non-drinkers’ sociability was not affected by their non-drinking status (as seemed to be the case in previous generations). Raninen et al. (2021) also argued that the scientific research field is outdated regarding knowledge of modern non-drinkers. The present study argues that longitudinal research on various psychological and social attributes of the same non-drinking adolescents over time is required in order to comment further on the social significance of alcohol within the Swedish adolescent population.

A longitudinal perspective would hence contribute to existing data from Raninen et al. (2021) by testing if suggested attributes of a non-drinker could provide a prediction for consistent non-drinking in the same population and age group of adolescents over time. In addition, a measure of adolescents’ attitudes towards alcohol should build on previous research (Törrönen et al., 2019) to investigate how adolescents relate to alcohol in relation to their own perception of it. The rationale for this study is hence to offer an outlook on non-drinking adolescents in a time when the norm climate regarding alcohol consumption is thought to be changing. Psychosocial characteristics of non-drinkers in relation to drinkers within this time of change have not been established yet. At the present time when there is a downward trend in alcohol drinking among young people, we hence need an increased understanding of the potential outcome of the declining trend within adolescent psychosocial characteristics. Nevertheless, to understand the non-drinking individuals within this time in comparison to the drinkers.

This study will examine the same generation of adolescents as studied by Raninen et al. (2021), identifying them at a younger age and following them longitudinally up to the age of 17 years. By following another sample from the same population (i.e. children born in 2001), the objective is to analyse if the outcome would be similar in terms of psychological health and social interactions. Since Raninen et al. (2021) determined that it is now more norm-breaking to drink at the age of 14–15 years (when in 9th grade), it would be interesting also to study an older age group, most of whom are expected to be drinkers (i.e. two years older). A motivation for this interest would be the postponed age of alcohol onset (as estimated occurring later) together with the normative climate’s supposed shift (going from non-normative to drink to normative to drink). Hence, we are studying those adolescents who continue to be non-drinkers as they enter a time when drinking becomes the norm.

The rationale for the present study is also to contribute to the existing literature by studying the potential outcome of the declining trend in adolescent drinking habits and considering non-drinkers in an older age group than previous research has done. This would potentially contribute to knowledge beyond non-drinking 9th graders to also apply to an older adolescent age group. This is important because, even though the non-drinking group has grown (Pennay et al., 2015), there is a gap in the literature examining non-drinking adolescents longitudinally.

Some research has shifted focus from examining what characterizes those who drink to examining those who do not drink during adolescence (Lund & Scheffels, 2019; Scheffels et al., 2020); however, the scientific literature still lacks longitudinal studies of psychological health, social interactions, and alcohol attitudes among Swedish adolescent non-drinkers, taking this updated information into consideration (i.e. delayed onset age and change in alcohol’s cultural position).

The first aim of this study was to examine the characteristics of Swedish adolescent non-drinkers aged 12–17 years. The second aim was to analyse whether psychological health (mental well-being, psychosomatic symptoms, conduct problems), social interaction frequency (new friends, time spent with friends, and social interaction through a screen device), the number of psychoactive substances used in the last 12 months, and positive attitudes towards alcohol at age 14–15 could predict non-drinking two years later (i.e. continuously being categorized as alcohol-naïve). We controlled for potential confounding variables impact on drinking status, including age, sex, parental monitoring, and religious affiliation.

The two research questions were:

1. Do non-drinking adolescents differ from drinking adolescents in psychological health, psychoactive substance use, social interaction frequency, and positive attitudes towards alcohol?

2. Can psychological health, psychoactive substance use, social interaction frequency, and positive attitudes towards alcohol in 9th grade predict non-drinking two years later?

Methods

This study was performed using data originally collected within the framework of the five-wave programme Longitudinal Research on Development in Adolescence (LoRDIA; Boson, 2016). The primary purpose of the programme was to follow adolescents over time with a particular focus on social networks, school attendance, well-being, mental health, alcohol use, and drug consumption. Data from four of the total of five data collections were used as the foundation for this study.

Participants

This study used longitudinal data from four data collections: surveys of adolescents from time point 1 (T1; n = 1515, age 12.5 years), time point 2 (T2; n = 1467, age 13.3 years), time point 3 (T3; n = 1322, age 14.3 years), and time point 4 (T4; n = 949, age 16.9 years). The analytical sample of this study included those adolescents who had answered the questions about drinking experiences at all measured time points (T1, T2, T3, and T4) in the LoRDIA programme. See a thorough description in the Measures section on Drinking status. Those who had not answered the questions on any measurement occasion were automatically excluded (79 individuals from T1, seven from T2, 13 from T3, and three from T4). The analytical sample for this study was 601 participants (245 boys and 356 girls). The mean age at T4 was 16.0 years (SD = 0.4), with 32% of participants with no alcohol experience consistently over all four time points and therefore identified as ‘non-drinkers.’ The remaining 67% answered that they had been drinking alcohol at T1 and/or T2 and/or T3 and/or T4 and were identified as ‘drinkers.’ For an overview, see Figure 1.

Figure 1. Flowchart of the analytical sample.

Measures

Drinking status

Drinking status was measured using the question ‘Have you ever drunk alcohol, more than just a single sip (don’t count light beer or low-alcohol cider)?’ with the response alternatives being: ‘Yes/No’ at T1, T2, and T3. At T4 the participants were asked ‘How often do you drink alcohol?’ with the response alternatives being: ‘Never/Once a month/2–4 times a month/2–3 times a week/4 or more times per week.’ The response ‘Never’ was coded as No = 1 and all other responses at T4 were coded as Yes = 0. Hence, answers were dichotomized for every measurement occasion. Additionally, to examine these groups independently, a binary variable was created capturing all those who answered (1) ‘No’ on all four test occasions and those who answered (0) ‘Yes’ on at least one test occasion. This means that non-drinkers were categorized as ‘1’ and drinkers as ‘0’. The questions were originally formulated for the Swedish Council for Information on Alcohol and Other Drugs (CAN) annual study of Swedish school students’ drug and alcohol habits (Gripe, 2013), which is based on the Swedish government’s strategy on alcohol, narcotics, and tobacco use (Ministry of Social Affairs, 2013).

Psychosomatic symptoms

Psychosomatic symptoms were measured at T1, T3 and T4, using the Swedish version of Psychosomatic Problem Scale (eight items; Hagquist, 2008). Example of items are: Have you been bothered by headaches? Have you felt that you had difficulty sleeping? The items are rated as Never (1), Seldom (2), Sometimes (3), Often (4) and Always (5). The total psychosomatic symptoms scale was calculated as the mean score of the eight items, with a higher score indicating higher frequency of the symptoms; Cronbach’s alphas at T1 (.81), T3 (.91) and T4 were .88.

Conduct problems

Conduct problems were measured at T1, T3 and T4 using a scale (12 items) from the Swedish Crime-preventing Counsel which repeatedly examines crime statistics in Sweden (Ring, 1999, 2013) Example of items are: ‘Stole (shoplifted) something in a shop or department store?’, ‘Had a knife with you (as a weapon) when you went out or another weapon’, ‘Threatened someone with a beating or a weapon to get money or other valuables?’. The items are rated as Never (1), Seldom (2), Sometimes (3), Often (4) and Always (5). The total conduct problems scale was calculated as the mean score of the 12 items, with a higher score indicating higher level of conduct problems; Cronbach’s alphas at T1 (.80), T3 (.91) and T4 were .83

Mental well-being

A two-item mental well-being measure was used at T1, T3 and T4, concerning satisfaction with life and purpose and meaning in life (Berlin et al., 2012). The two items are: ‘In general, how happy are you with life at the moment’? with the score rated as: Very happy (4), Quite happy (3), Quite unhappy (2), Very unhappy (1), and ‘I think that my life has purpose and meaning’ with the score rated as; Completely agree, (4), Partly agree (3), Partly disagree (2), Completely disagree (1).

Positive attitudes toward alcohol

A Swedish scale on positive attitudes towards alcohol was measured in LoRDIA by six statements at T3 and T4: (1) ‘Alcohol generally has a positive effect on people (makes them feel good, happy)’; (2) ‘Alcohol can both help and harm how well a person works with others (makes them want to have fun together/makes them mean to others)’; (3) ‘Alcohol makes people think better and strengthens their coordination (people understand things better/do things better)’; (4) ‘Alcohol improves sexuality (more enjoyable/feels more romantic and sexy/makes it easier to have sex)’; (5) ‘Alcohol makes people feel stronger and more powerful (easier to fight/speak in front of others/stand up for others)’; and (6) ‘Alcohol helps people relax, feel less tense, and stop thinking about their mistakes at school/work.’ The items were rated as Totally incorrect (1), Incorrect (2), Uncertain (3), Correct (4), and Absolutely correct (5). The total positive attitude towards alcohol was calculated as the mean score of the six items, with a higher score indicating a more positive attitude towards alcohol use; Cronbach’s alphas at T3 and T4 were .70.

Social interaction frequency

Social interaction frequency was measured by two questions originally part of a constructed and validated Swedish scale that explores social activities among adolescents and children (Arvidsson, 2013; Arvidsson et al., 2012; Kerr & Stattin, 2000; Stattin & Kerr, 2000). The first question was ‘How often do you make new friends?’ The response alternatives were ‘Seldom’, ‘Sometimes’, and ‘Often’. This question was measured at T1, T2, T3, and T4 and was treated as categorical. The second question was ‘During a regular week, how many days do you usually spend with friends?’ The response alternatives were ‘Never’, ‘Less than one day per week’, ‘One day per week’. This question was asked at T3 and T4 and was treated as categorical.

Social interaction through screen device was measured at T3 using the question ‘How much do you use your computer or mobile phone to send emails, chat, surf the net or keep in touch (through e.g. Facebook, Skype) on a normal weekday? With the response alternatives being: Never/seldom,’ ‘Less often than every week,’ ‘Approximately once per month,’ and ‘Every week’ and treated as categorical. The question is part of a larger Swedish scale measuring children’s leisure time (Statistics Sweden, 2009).

Psychoactive substance use was measured using the question ‘How many psychoactive substances have you been using the last 12 months?’ Alcohol use was not included. The question was measured at T4. The item is part of the Swedish version of Drug Use Disorders Identification Test (DUDIT; Berman et al., 2005). The response alternatives were ‘Never’, ‘Once a month or less often’, ‘2–4 times a month’, ‘2–3 times a week’, and ‘4 times a week or more often’ and coded from 0 to 4 on a likert scale.

Covariates

To control for potential impact on the outcome of the main analyses and on the dependent variable, non-drinkers and drinkers were compared in terms of confounding variables regarding age, sex (girl/boy), religious affiliation and parental monitoring. Religious affiliation was measured at T3 using the question ‘What’s your religious affiliation?’ with the response alternatives being: ‘I don’t have a religious affiliation,’ ‘Christian,’ ‘Muslim,’ ‘other’. The ‘other’ alternative included Judaism, Buddhism and other religion. The variable was treated as categorical. Parental monitoring was measured at T3 and T4 by the Swedish versions of the scales: Parental knowledge (6 items, e.g. ‘Do your parents know what you are doing during your free time?’), Parental solicitation (6 items, e.g. ‘Do your parents ask you to talk about your friends’), Parental control (6 items, e.g. ‘Do you need permission to stay out late on a weekday evening?’, and Adolescent disclosure scale (5 items, e.g. ‘When you have been out in the evening, do you talk to your parents about what you have done that evening.’ The items are rated as Never (1), Sometimes (2), and Often/always (3). The subscales of parental monitoring was calculated as the mean score of the including items, with a higher score indicating higher occurency of the behaviour; Cronbach’s alphas at T3 were Parental knowledge; .52, Parental solicitation; .57, Parental control; .81, Adolescent disclosure; .06 and at T4: Parental knowledge; .55, Parental solicitation; .59, Parental control; .79, Adolescent disclosure; .11

Procedure

The first data collection (T1) was in 2013, when the participants from the two age cohorts were about 12 and 13 years old, respectively; they were followed up yearly until 2017 and 2018, when the participants were about 17 years old. Before participation, summary information was presented about the purpose of the study and written consent was obtained. Participants who were younger than 15 years old at the time of recruitment were also required to obtain participation approval from their parents. The participants completed the questionnaires during school hours in their classrooms, while project researchers were present to answer questions. The survey, consisting of about 350 questions in total, took approximately one to two hours to complete. Project scientists were responsible for the data collection, with guidance from the principal investigator of the LoRDIA project. For a more in-depth presentation of the data collection procedure, see Boson (2016) and Boson (2018).

Statistical analyses

Descriptive comparisons using t-tests and chi-square tests were conducted between non-drinkers and drinkers for the independent variables (i.e. psychological health, social interaction frequency, psychoactive substance use and positive attitudes towards alcohol) and the confounding variables (i.e. age, sex, parental monitoring, and religious affiliation). For independent sample t-tests, information about effect sizes was estimated using Cohen’s d and can be interpreted as follows: .20 = small effect, .50 = medium-sized effect, and .80 = large effect (Cohen, 1988). For chi-square tests, information about effect sizes was estimated using Cramér’s V and can be interpreted as follows: <.20 = small effect, .20–.39 = medium-sized effect, and .40–.59 = large effect (Rea & Parker, 2014). As expected, not all members of the total study sample (N = 601) answered all survey questions; those who did not answer all questions were kept in the dataset and included in all analyses but were reported as ‘internal missing’ for the specific missing items.

To examine within-group differences, between-group differences, and interaction effects over time, for variables with three measurement points [T1, T3, T4]: (psychosomatic symptoms, conduct problems, mental well-being and having new friends), mixed ANOVA was used. The variables were z-transformed for easier interpretation of the data in the presented figures.

To explore predictors of the non-drinking group, two binary logistic regression models were computed, Model 1 without covariates and Model 2 adjusted with covariates. Information about odds ratio (OR) was included. Before the logistic regressions, analyses were conducted to explore multicollinearity, along with inspections for outliers using standardized residuals. Preliminary analysis further suggested that the assumption of multicollinearity was met regarding all independent variables. Inspection of residual values revealed outliers, that were tested to rule out any potential extreme effects. However, analyses did not reveal any effects. Accordingly, all outliers were kept in the dataset and included in the main analyses. Since 15 comparisons were carried out, the significance level was determined at a Bonferroni-adjusted alpha level of p < .003 (.05/15) for the independent sample t-tests/chi-square-tests; for all other analyses conducted here, the significance level was determined at p < .05.

Results

Research Question 1:

Do non-drinking adolescents differ from drinking adolescents in psychological health, psychoactive substance use, social interaction frequency, and positive attitudes towards alcohol?

Analyses using t-tests at T3 and T4 showed a significant difference in conduct problems (p < .001), positive attitudes towards alcohol (p < .001), and number of psychoactive substances used last 12 months (p < .001), indicating that drinkers had significantly more of these experiences and attitudes. Analyses using the chi-square test indicated that non-drinkers were significantly less likely to have new friends (p < .001). They were also significantly less likely to spend as much time with friends as drinkers (p < .001). Non-drinkers were also significantly less likely to have as much social interactions through a screen device, as drinkers (p < .001). Non-drinkers described their parents as having significantly more knowledge (p < .001) and control (p < .001). The non-drinkers were significantly more open to their parents (adolescent disclosure) than were the drinkers (p < .001). There were significantly more drinkers in the group with no religious affiliation (p < .001). Table 1 shows the descriptive statistics in the analytical sample and group mean differences between adolescents categorized as Drinkers or Non-drinkers.

Table 1. Descriptive statistics and comparisons of means between drinkers and non-drinkers for the analytical sample.

	Drinkers		Non-drinkers
Psychosocial factors (Time point)	n	M (SD)/%	n	M (SD)/%	t/Chi-2	p		Cohens d/Cramérs V
Conduct problems (T4)	394	6.36 (.95)	192	6.03 (.24)	6.60	<.001^a		.424
Psychosomatic symptoms (T4)	396	2.47 (.80)	193	2.33 (.83)	1.96	.051		.172
Mental well-being (T4)	402	6.64 (1.24)	195	6.66 (1.34)	−.22	.824		−.019
Positive alcohol attitudes (T3)	405	3.26 (.54)	195	2.93 (.67)	5.95	<.001^a		.560
Number of psychoactive substances last 12 months (T4)	406	.22 (.70)	195	.03 (.16)	5.28	<.001^a		.331
New friends (T4)					17.50	<.001^a		.172
Never	60	59.4	41	40.6
Sometimes	248	65.1	133	34.9
Often	90	´84.1	17	15.9
Time with friends (T4)					31.08	<.001^a		.229
Never	12	35.3	22	64.7
Less than once a week	62	59.0	43	41.0
Once a week	109	63.4	63	36.6
Several times a week	215	76.5	66	23.5
Social interaction through screen device (T3)					30.12	<.001^a		.225
Never	5	35.7	9	64.3
Less than once a week	59	56.7	45	43.3
Once a week	109	59.6	74	40.4
Several times a week	226	77.4	66	22.6
Covariates
Parental knowledge (T4)	402	2.50 (.40)	190	2.61 (.33)	−3.58	<.001^a		−.295
Parental control (T4)	402	1.79 (.53)	190	1.99 (.59)	−4.07	<.001^a		−.374
Parental solicitation (T4)	400	2.29 (.41)	191	2.23 (.48)	1.293	.197		.120
Adolescent disclosure (T4)	402	2.41 (.42)	191	2.53 (.39)	−3.54	<.001^a		−.311
Age (T4)	406	16.94 (.45)	195	16.92 (.42)	0.661	.509		.058
Sex (T1)					.070	.791		.011
Girl	239	67.1	117	32.9
Boy	167	68.2	78	31.8
Religion (T3)					15.43	<.001^a		.162
No religion	161	75.2	53	24.8
Christian	214	64.7	117	35.3
Muslim	15	51.7	14	48.3
Other	6	40.0	9	60.0

Regarding effect sizes Cohens d and Cramérs V; ‘–’ indicates higher scores in the non-drinking group.

^ap < 0.003(Bonferroni adjusted alpha level).

We examined within-group differences, between-group difference, and interaction effects over time, for variables with three measurement points [T1, T3, T4]: Conduct problems, psychosomatic symptoms, mental well-being, and having new friends (see Figure 2).

Figure 2. Mean values over time of the variables conduct problems, psychosomatic problems, mental well-being and having new friends divided into drinkers and non-drinkers.

The adolescents had a significant increase in conduct problems over time, F (2, 563) = 4.635 p = .010. There was also a difference between groups in conduct problems, F (1, 565) = 15.50 p < .001, where drinkers had more of these problems. There was also a significant interaction effect, F (2, 565) = 3.020 p = .049; the drinkers increased in conduct problems over time, whereas the non-drinkers did not have any increase in conduct problems over time. It should be noted that the values were near zero for both groups, meaning that no group had severe conduct problems (see Figure 2).

The adolescents had a significant increase in psychosomatic symptoms over time, regardless of group, F (2, 565) = 470.48 p < .001. Hence, there was no difference between the groups in psychosomatic symptoms, F (1, 565) = 1.65 p = .199, neither was there any interaction effect, F (2, 565) = 1.86 p = .157. The adolescents had neither an increase, nor a decrease in mental well-being over time, F (2, 565) = 0.070 p = .932. There was no difference between the groups in mental well-being, F (1, 565) = 2.53 p = .165, neither was there any interaction effect, F (2, 565) = 1.93 p = .82. The adolescents had neither an increase, nor a decrease in having new friends over time, F (2, 550) = 1.60 p = .203. There was a difference between the groups in having new friends, F (1, 550) = 22.45 p < .001, where drinkers to a greater extent made new friends. There was no interaction effect in having new friends, F (2, 550) = 0.98 p = .374.

Research Question 2:

Can psychological health, psychoactive substance use, social interaction frequency, and positive attitudes towards alcohol in 9th grade [T3] predict non-drinking two years later [T4]?

Binary logistic regressions were performed to assess the impact of the psychosocial factors on the likelihood that participants would report that they had never drunk at T4: Firstly, bivariate logistic regressions were made for the independent variables separately. Model 1 contained only the independent variables; Model 2, was an adjusted model containing both the independent variables and the covariates. The models will be referred to as Model 1 and Model 2, below.

Model 1 contained eight psychosocial factors: Conduct problems, psychosomatic symptoms, mental well-being, positive attitudes towards alcohol, the number of psychoactive substances last 12 months, new friends, days spent with friends per week and social interaction through a screen device. Model 1 containing all psychosocial factors was statistically significant (χ² [13, N = 562] = 150.99, p < .001), indicating that the model could distinguish between participants who did and did not report drinking throughout their adolescent years. The model explained between 23.6% (Cox and Snell R²) and 32.7% (Nagelkerke R²) of the variance in drinking status and correctly classified 71.5% of cases. As shown in Table 2, three of the psychosocial factors (i.e. number of psychoactive substances last 12 months, days spent with friends per week and social interaction through a screen device) made unique and statistically significant contributions to the model.

Table 2. Bivariate associations and logistic regression analyses predicting being a non-drinker at T4 by psychosocial factors at T3. Odds ratios and confidence intervals are presented (n = 601).

	Bivariate associations	Model 1	Model 2
			Full Model
Psychosocial factors	OR (95% CI)	OR (95% CI)	OR (95% CI)
Conduct problems	.03 (.01–.34)**	1.28 (.78–2.10)	1.44 (.82–2.52)
Psychosomatic symptoms	.92 (.75–1.14)	1.22 (.86–1.72)	1.01 (.68–1.48)
Mental well-being	1.02 (.89–1.16)	.98 (.79–1.22)	1.00 (.79–1.27)
Positive alcohol attitudes	.86 (.79–.93)**	.92 (.83–1.01)	0.93 (.83–1.04)
Number of psychoactive substances last 12 months	.07 (.02–.20)**	.08 (.02–.24)**	.08 (.02–.27)**
New friends (ref: never)
Sometimes	.89 (.54–1.44)	1.19 (.64–2.21)	1.27 (.65–2.49)
Often	.40 (.21–.79)	.93 (.41–2.10)	.90 (.38–2.15)
Time with friends (ref: never)
Less than once a week	.38 (.17–.85)*	.39 (.16–.99)*	.41 (.14–1.14)
Once a week	.32 (.15–.68)**	.38 (.16–.94)*	.34 (.12–.94)*
Several times a week	.17 (.08–.36)**	.24 (.10–.59)**	.23 (.08–.63)**
Social interaction through screen device (ref: never)
Less than once a week	.42 (.13–1.35)	.12 (.01–.1.01)	.10 (.01–.91)*
Once a week	.38 (.12–1.17)	.12 (.01–.95)*	.08 (.01–.76)*
Several times a week	.16 (.05–.50)**	.07 (.01–.55)**	.04 (.01–.39)**
Covariates
Parental knowledge			1.03 (.89–1.19)
Parental control			1.11 (1.02–1.21)*
Parental solicitation			.96 (.87–1.06)
Adolescent disclosure			.94 (.83–1.08)
Age			1.42 (1.02–2.00)
Sex (ref: girl)
Boy			.67 (.41–1.11)
Religion (ref: no religion)
Christian			1.26 (.77–2.07)
Muslim			2.37 (.82–6.82)
Other			5.46 (1.23–24.16)*

Models 1 and 2 are adjusted, but covariates are included only in Model 2.

*p < 0.05.

**p < 0.01.

Model 2, the full model, contained the eight psychosocial factors together with the covariates (i.e. age, sex, parental monitoring, and religious affiliation). The full model containing all predictors was statistically significant (χ² [22, N = 547] = 161.69, p < .001), indicating that the model could distinguish between participants who did and did not report drinking throughout their adolescent years. The model explained between 26.4% (Cox and Snell R²) and 36.8% (Nagelkerke R²) of the variance in drinking status, and correctly classified 73.8% of cases.

As shown in Table 2, the number of psychoactive substances last 12 months, days spent with friends per week and social interaction through a screen device continued to be significant in the adjusted model.

Discussion

The overall rationale for this study was to investigate the characteristics of Swedish non-drinkers during adolescence and psychosocial predictors of total abstinence. The study aimed to determine whether non-drinking adolescents differed from drinking adolescents in psychological health (mental well-being, psychosomatic symptoms, conduct problems), social interaction (new friends, time spent with friends, and social interaction through screen devices), psychoactive substance use, and positive attitudes towards alcohol. Furthermore, it sought to establish whether psychological health, social interaction frequency, psychoactive substance use, and positive attitudes towards alcohol in 9th grade (T3) could predict continued non-drinking two years later (T4). Analyses revealed a significant difference in conduct problems, with drinkers having more conduct problems, and there was an increase in problems over time. It should be noted, however, that conduct problems were low at the group level in both groups, despite the differences, implying that adolescents in general reported low levels of conduct problems. Psychosomatic symptoms were similar between drinkers and non-drinkers, both experiencing a significant increase over time. Mental well-being remained stable and comparable in both groups. Hence, self-reported psychosomatic symptoms and mental well-being appeared independent of alcohol use in our sample. Drinkers had significantly more positive attitudes towards alcohol and more experience using psychoactive substances in the last 12 months. Non-drinkers were significantly less likely to have new friends, spend as much time with friends as drinkers, and have social interactions through screen devices. These results supported previous research on adolescent non-drinkers, indicating fewer social relations, fewer friends, and more difficulty making new friends (Hoel et al., 2004; Kivimäki et al., 2014; Larm et al., 2018). However, these findings contrasted with Raninen et al. (2021) conclusion that non-drinkers are more prosocial than drinkers. Non-drinkers in this study also exhibited fewer conduct problems, consistent with previous studies (Boson, 2016; Kuperman et al., 2013) associating non-drinking with fewer externalizing problems. Additionally, non-drinkers described their parents as having significantly more knowledge and control, and they were more open to their parents (i.e. adolescent disclosure) than drinkers. This result aligns with other studies emphasizing the importance of parenting styles for adolescent alcohol consumption (Sharmin et al., 2017; Yap et al., 2017). Significantly more drinkers belonged to the group with no religious affiliation compared to non-drinkers. Results also suggested that non-drinkers are distinguished by less positive attitudes towards alcohol, consistent with recent studies (Scheffels et al., 2020; Törrönen et al., 2019) arguing that contemporary adolescents are more responsible and reflective, regarding alcohol less as a symbol of growing up than previous generations. Moreover, both non-drinkers and drinkers scored fairly low on positive attitudes towards alcohol, indicating that neither group had particularly positive attitudes towards alcohol consumption in general, but non-drinkers had significantly less positive attitudes. Analyses on the predictive value of psychosocial factors measured in the 9th grade on being categorized as non-drinking two years later suggested associations between several factors and non-drinking. Conduct problems, positive attitudes towards alcohol, use of psychoactive substances, and more time spent with friends per week decreased the likelihood of being categorized as a non-drinker two years later. These psychosocial factors remained important even when controlled for relevant covariates. This may imply that a more introverted personality function is linked to alcohol abstinence during adolescence and could be seen as a factor protecting against alcohol use. Links between a more introverted (low novelty seeking and high harm avoidance) personality function have been found previously (Boson et al., 2019).

As reported previously, both non-drinkers and drinkers indicated similar levels of mental well-being, and both groups had somewhat high scores, suggesting that good mental well-being appears unrelated to drinking status in our findings. While previous studies have distinguished non-drinkers by their greater mental well-being (Hoel et al., 2004; Larm et al., 2018), this study found no significant difference in well-being between non-drinkers and drinkers. However, it is noteworthy that indications of high well-being scores among adolescents in general (at the same time point as the data on alcohol use were collected) have been observed in previous research (Boson et al., 2016; WHO, 2016).

Furthermore, the present results are not aligned with the description of adolescents from a previous cross-sectional study of non-drinking 9th-grade Swedish adolescents (Raninen et al., 2021), which found no difference in social position between non-drinkers and drinkers. Additionally, the study reported that non-drinkers displayed more prosocial behaviour. While Raninen et al. (2021) further suggested a shift in the social significance of alcohol consumption among Swedish adolescents, this study is sceptical of that proposition. Instead, the present results indicate that non-drinking adolescents significantly differ from drinking adolescents in their social and psychological attributes, as evidenced by fewer social interactions in 9th grade predicting non-drinking two years later. This study notes that Raninen et al. (2021) determined that it is now more norm-breaking to drink at age 15 when in 9th grade. Therefore, it would be interesting to study older non-drinkers at an age when the expectation is that most adolescents are drinkers (e.g. two years later). In view of the present results, this study suggests that the characteristics of non-drinkers (i.e. fewer social interactions, fewer conduct problems, and less positive attitudes towards alcohol) remained consistent over the studied period from 9th grade to two years later, given that the adolescents were strictly non-drinkers throughout their adolescent years. This means that, while the normative climate changed between 9th grade and two years later, the characteristics of the non-drinkers did not. The results could also extend previous suggestions (Gripe, 2013) that changes in social climate, in interaction with delayed age of drinking onset, could have affected the trend of decreasing adolescent alcohol consumption observed in Swedish society (Norström & Raninen, 2018).

Limitations

First, the analytical sample contained more girls than boys (59% girls) which might have influenced the results. However, we found no differences in distribution between the sexes regarding the outcome variable (i.e. being a drinker or a non-drinker). Second, there are general considerations about the participants, for example, that the participants might have initially differed from the non-participants in some ways, simply by being more prone to participate longitudinally in the study. For example, exclusion analyses from one of the original studies within the LoRDIA project, presented by Boson et al. (2016), indicate that especially boys with externalizing/conduct problems dropped out over time. The same study revealed gender differences in adolescents’ health problems, meaning that boys had more externalizing/conduct problems and girls more internalizing problems (Boson et al., 2016). This suggests that boys with externalizing/conduct problems might not be sufficiently represented in this study, and the influence this could have on the results merits further discussion.

Conclusions

This study aimed to fill the gap in knowledge on the characteristics of non-drinkers during adolescence. Non-drinkers were consistently characterized by fewer social interactions, less positive attitudes towards alcohol, less probability of using other drugs, and fewer conduct problems from 14 to 17 years. Additionally, non-drinkers experience higher control and self-disclosure in relation to their parents. In summary, given that a trend of declining adolescent alcohol consumption has been observed in Sweden (Norström & Raninen, 2018) and other parts of the world (Pennay et al., 2018), it is crucial to distinguish meaningful social and psychological factors underlying the behavioural characteristics of adolescent non-drinkers. The results of this study may suggest that total alcohol abstinence appears to be linked to more introverted social functioning, less overall problem behaviour, and more control and dialogue with caregivers. Therefore, the importance of supporting parents throughout adolescence, possibly through preventive interventions, can be valuable, as can providing youth with access to alcohol- and drug-free social arenas. Distinguishing these characteristics and profiling adolescent non-drinkers might additionally benefit other scientific research areas associated with adolescent developmental factors. By examining non-drinkers longitudinally, this study advances our knowledge of the characteristics of non-drinking Swedish adolescents aged 12–17 years.

Disclosure statement

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

Additional information

Funding

This work was supported by the [Alcohol Research Council of the Swedish Alcohol Retailing Monopoly] under Grant [No. 2019-0029]. The LoRDIA programme provided the bulk of funding through a combined grant from four Swedish research foundations: the Swedish Research Council (VR); the Swedish Research Council for Health, Working Life and Welfare (FORTE), Sweden’s Innovation Agency (VINNOVA) and the Swedish Research Council Formas [No. 259-2012-25].