School-related covariates of adolescent gambling: findings from the Stockholm school survey

ABSTRACT

The present study investigated the associations that student gambling and risk gambling share with a) students’ own experiences of their situation in school, b) school performance, and c) truancy. Data from the 2020 Stockholm School Survey were used, with information collected among 10,901 students in grades 9 (15–16 years) and 11 (17–18 years) from 145 schools in Stockholm Municipality. The studied school-related factors were school satisfaction, perceived teacher caring, perceived school order, school performance, and truancy. Binary logistic regression analyses were performed, with robust errors clustering at the school level. Students’ perceived teacher caring and perceived school order were inversely associated with both gambling and risk gambling, while truancy was positively associated with both gambling and risk gambling, even when mutually adjusting for all school-related factors simultaneously. School satisfaction was inversely associated with gambling, and school performance was inversely associated with risk gambling, when mutually adjusting for all school-related factors simultaneously. In sum, more positive experiences of the situation in school, higher school performance, and not playing truant were linked with a lower likelihood of gambling and risk gambling among students. The findings suggest that students’ situation in school can help to identify those at risk for gambling problems.

KEYWORDS:

• Gambling
• youth
• adolescents
• school
• school performance
• truancy

Introduction

Although it is illegal to offer gambling opportunities to anyone under the age of 18 in Sweden (The Swedish Gambling Act SFS 2018:1138), gambling is still a relatively common activity among adolescents, with some recent Swedish studies reporting that around 15% of students aged 15–18 years have engaged in any type of gambling the past year (Låftman et al., 2020; Spångberg & Svensson, 2022). Swedish adolescents have reported various ways in which they surpass the age-limit on gambling such as using other people’s online accounts, gambling in unregulated environments, or simply not having to disclose their age (The Public Health Agency of Sweden, 2021). A non-negligible proportion of adolescent gamblers can also be categorized as problem or risk gamblers (Calado et al., 2017). Since early initiation into gambling is associated with an increased risk of gambling problems later in life (Carbonneau et al., 2015), and co-occur with other risk behaviors and health problems such as substance abuse, delinquent behavior, and mood disorders (Kryszajtys et al., 2018; Lee et al., 2014; Molinaro et al., 2018; Nigro et al., 2017; Shead et al., 2010), gambling during adolescence can be regarded as a public health concern (Messerlian et al., 2005). Thus, knowledge about risk and protective factors of adolescent gambling is relevant.

Prior research on adolescent gambling has largely concentrated on risk factors at the individual and the family level. There is a clear and consistent gender difference in that both gambling and risk gambling are more common among boys than among girls (Calado et al., 2017; Dickson et al., 2008; Downling et al., 2017; Kraus & Nociar, 2016; Lee et al., 2014; Svensson & Sundqvist, 2019; Turchi & Derevensky, 2006). Other risk factors at the individual level include personality traits such as impulsivity and sensation-seeking, but also anxiety, depression, ADHD, substance use, and delinquency (Lee et al., 2014; Shead et al., 2010). With regards to conditions in the family, earlier research has shown that parental gambling (McComb & Sabiston, 2010; Shead et al., 2010) and low socioeconomic status (Elgar et al., 2018) are examples of risk factors for adolescent gambling. The type of protective factors that have been studied are also usually at the individual level or family level (Dickson et al., 2008; Elgar et al., 2018; McComb & Sabiston, 2010; Shead et al., 2010). In the context of the family, aspects such as family cohesion, family support, parental monitoring, and child disclosure have been shown to be inversely associated with problematic gambling among adolescents (Dickson et al., 2008; Elgar et al., 2018; McComb & Sabiston, 2010; Shead et al., 2010).

One important arena in adolescents’ lives is the school, which may affect their likelihood of engaging in risk behaviors in several ways. In earlier studies, we have examined the links that adolescent gambling shares with schools’ organizational factors as rated by teachers, including school ethos (Låftman et al., 2020), school leadership (Olsson et al., 2021a), and school collective efficacy (Olsson et al., 2021b). Our studies showed that higher teacher ratings of these school contextual characteristics were linked with a lower prevalence of gambling among the students. There are also studies that have examined gambling in relation to school-related factors at the individual level. For instance, an inverse association has been shown for school performance and gambling (Shead et al., 2010), whereas skipping classes has been shown to be linked with a higher risk of gambling (Svensson & Sundqvist, 2019). Furthermore, Dickson et al. (2008) reported that a higher degree of school connectedness was associated with a lower risk of gambling among students. Taken together, however, there is a relative lack of studies into school-related factors and gambling. Yet, given the general potential of schools to identify students in need of support or interventions, such knowledge is relevant. It is also of interest to examine possible gender and age differences in the link between school-related factors and gambling. Not only do gambling and risk gambling tend to be more common among boys and, although not as consistently, older students (e.g. Calado et al., 2017; Downling et al., 2017; Kaltenegger et al., 2019; Svensson & Sundqvist, 2019), but school performance, experiences, and behaviors can also vary by gender and age (WHO, 2020). Several prior studies have found that the associations between risk and/or protective factors at the school, family, social, and individual level and adolescent gambling differ depending on gender or age (Buja et al., 2022; Chalmers & Willoughby, 2006; Donati et al., 2013; McComb & Sabiston, 2010; Merkouris et al., 2016; Shead et al., 2010; Svensson & Sundqvist, 2019).

In Sweden, a new Gambling Act came into force on 1 January 2019, entailing a license system for gambling companies and a stricter regulation of gambling advertising (Government Offices of Sweden, 2021). Most Swedish studies into factors associated with adolescent gambling have hitherto been based on data collected prior to the implementation of this Act. To obtain an up-to-date picture of correlates of adolescent gambling in Sweden, more recently collected data are needed.

Using survey data collected in lower and upper secondary schools in Stockholm in 2020, the aim of the current study was to investigate the associations that student gambling and risk gambling share with a) students’ own experiences of their situation in school, b) school performance, and c) truancy.

Methods

Data material

The data were obtained from the 2020 Stockholm School Survey (SSS). Every two years, the SSS is performed among students in grades 9 (15–16 years) and 11 (17–18 years) attending schools located in Stockholm Municipality. All public schools and many private schools participate in the survey (Stockholm Municipality, 2020). In total, 16270 surveys were handed out to 145 schools and 11,877 approved responses were collected (response rate 73%). After further exclusion due to internal non-response, the final sample consisted of 10,901 students (92%). Since participation in the SSS is anonymous, the Regional Ethical Review Board of Stockholm has decided that studies based on the SSS are not under consideration for ethical approval (ref. no. 2010/241-31/5).

Measures

Dependent variables

Gambling was captured by the question ‘Have you at any time during the last 12 months gambled for money? (Example: casino, poker, slot machines, Loot-boxes, sports betting, lottery tickets, or similar’. Response alternatives were (1) ‘No’, (2) ‘Yes – online (computer, phone, tablet)’, and (3) ‘Yes – in another way’. Those who ticked (2) or (3) were categorized as having gambled.

One question was used to assess risk gambling and was posed to students who had answered that they had gambled during the last 12 months: ‘How many times during the last 12 months have you … ’ (a) ‘ … tried to reduce your gambling?’, (b) ‘ … felt restless and irritated if you haven’t been able to gamble?’, (c) ‘ … lied about how much you’ve gambled?’. The response alternatives were (1) ‘Never’, (2) ‘1–2 times’, and (3) ‘3 times or more’. Students who answered (2) or (3) on either of the three subquestions were categorized as risk gamblers. The measure has been validated (Kaltenegger et al., 2019) and used in earlier studies (Kaltenegger et al., 2019; Låftman et al., 2020, 2020; Olsson et al., 2021a).

Independent variables

School satisfaction was captured by four items that asked students to rate how much they agreed with the following statements on a four-point Likert scale: (a) ‘I enjoy going to school’, (b) ‘I look forward to going to my classes’, (c) ‘Schoolwork feels pointless’, and (d) ‘Most of my teachers make learning interesting’. Response alternative ranged from (1) ‘very poorly’ to (4) ‘very well’. Item (c) was reversed so that higher scores reflected higher school satisfaction for all items. The responses were then summed to an index ranging from 4–16 (Cronbach’s alpha 0.72). This index has been used in a prior study (Ramberg et al., 2020).

Perceived teacher caring was captured by seven items: (a) ‘If you don’t understand something, you get help from the teacher straight away’, (b) ‘Teachers praise students who do something good at school’, (c) ‘My teacher doesn’t give me any praise when I work hard’, (d) ‘Adults step in if anyone is harassed or bullied’, (e) ‘The teachers let us know what we can and can’t do’, (f) ‘Students take part in the planning of what we will do in class’, and (g) ‘The school lets my parents know if I’ve done something good’. Response alternatives were the same as for school satisfaction. Item (c) was reversed so that higher scores reflected higher perceived teacher caring for all items. The responses were then summed to an index ranging from 7–28 (Cronbach’s alpha 0.73). This index has been used in an earlier study (Ramberg et al., 2020).

Perceived school order was captured by three items: (a) ‘There’s a lot of noise and rowdiness in class’, (b) ‘When a class starts it takes at least five minutes before we can get started’, and (c) ‘I’m worried about being subjected to crime at school’. Response alternatives were the same as for school satisfaction and perceived teacher caring. All items were reversed so that higher scores reflected higher perceived school order. The responses were then summed to an index ranging from 3–12 (Cronbach’s alpha 0.47).

School performance was captured by students’ self-reported marks in Swedish, English, and mathematics, which were summed to an index. Each mark was given a corresponding number (A = 5, B = 4, C = 3, D = 2, E = 1, F = 0, ‘No mark’ (streck) = 0), resulting in an index ranging from 0–15.

For students with missing values on at most one third of the items included in each index described above, missing values were replaced by the individual mean of the remaining items of the respective index.

Truancy was captured by the question ‘Have you played truant from school for a whole day this school year?’ with response alternatives ranging from (1) ‘No’ to (6) ‘Yes, more than 20 times’. Students reporting truancy more than once this school year were categorized as having played truant.

Control variables

Control variables included gender, grade, family structure, migration background, parental university education, and parental unemployment.

Statistical methods

Associations between the school-related variables and (risk) gambling were assessed though binary logistic regressions. Firstly, bivariate associations between the dependent variable(s) and each of the independent or control variables were assessed in a crude model. Model 1 included school satisfaction, perceived teacher caring, and perceived school order, whilst simultaneously adjusting for all control variables. In Model 2, school performance and all control variables were included, and Model 3 included truancy together with all control variables. Finally, Model 4, included all independent and control variables simultaneously. To take the clustered nature of the data into account, robust standard errors clustering at the school level were estimated. To examine if associations differed by students’ gender or grade we included interaction terms between all school-related factors and gender or grade, one at a time, in Model 4 for both gambling and risk gambling. Furthermore, fully adjusted models stratified by gender and by grade were performed (presented in the Appendix (Tables A1 and A2)).

Results

Descriptive statistics for the study sample (n = 10,901) are presented in Table 1. Having gambled in any way during the last 12 months was reported by 12.1% of the students, and 3.9% were categorized as risk gamblers. The mean values for school satisfaction, perceived teacher caring, perceived school order, and school performance were 11.0 (range:4–16), 18.8 (range:7–28), 8.2 (range:3–12), and 9.5 (range:0–15), respectively. As for truancy, 14.2% of the students had played truant more than once during the current school year. Regarding gender, there were slightly more girls than boys, and additionally, a third gender category was included, comprising of students answering ‘other gender identity’ (n = 105) and those with missing answers (n = 193), making up 2.7% of the study sample. There were also slightly more students in grade 9 than in grade 11, almost two thirds lived with two parents in the same household, 8.4% had lived in Sweden for less than 10 years, 71.1% had at least one parent with a university education, and 4.4% had at least one unemployed parent.

Table 1. Descriptive statistics, n = 10,091.

Variables	Mean	s.d.	Min.	Max.
School satisfaction	11.0	2.57	4	16
Perceived teacher caring	18.8	3.63	7	28
Perceived school order	8.2	1.80	3	12
School performance	9.5	3.35	0	15
	n	%
Truancy
No	9,351	85.8
More than once past school year	1,550	14.2
Gambling last 12 months
No	9,580	87.9
Yes	1,321	12.1
Risk gambling
No	9,669	96.1
Yes	422	3.9
Gender
Girls	5,533	50.8
Boys	5,070	46.5
Other/Missing	298	2.7
Grade
9	5,913	54.2
11	4,988	45.8
Family structure
Lives with two parents in one household	7,074	64.9
Other	3,827	35.1
Migration background
Swedish-born or lived in Sweden 10+ years	9,982	91.6
Lived in Sweden less than 10 years	919	8.4
Parental university education
No parent	3,155	28.9
At least one parent	7,746	71.1
Parental unemployment
No parent	10,422	95.6
At least one parent	479	4.4

Table 2 displays the results from binary logistic regressions with gambling as the dependent variable. In the crude analyses, all school-related factors were associated with gambling (school satisfaction: OR 0.91, p < 0.001; perceived teacher caring: OR 0.95, p < 0.001; perceived school order: OR 0.89, p < 0.001; school performance: OR 0.95, p < 0.001; truancy: OR 1.96, p < 0.001). All of these associations remained robust and statistically significant in Models 1–3. In Model 4, all associations except for school performance remained robust and statistically significant. Turning to the control variables in Model 4, boys and students in the third gender category had an elevated risk of gambling compared with girls, and students living in ‘other’ households had an increased risk of gambling compared with students living with two parents in the same household.

Table 2. Odds ratios (OR) from binary logistic regressions of gambling, n = 10,901.

	Gambling
	Crude	Model 1	Model 2	Model 3	Model 4
	OR	OR	OR	OR	OR
School satisfaction	0.91***	0.92***			0.94***
Perceived teacher caring	0.95***	0.97**			0.96**
Perceived school order	0.89***	0.91***			0.92***
School performance	0.95***		0.97**		1.00
Truancy	1.96***			2.61***	2.28***
Gender
Girls (ref.)	1.00	1.00	1.00	1.00	1.00
Boys	7.26***	7.94***	7.21***	8.06***	8.57***
Other/Missing	6.58***	6.27***	6.47***	6.44***	6.31***
Grade
9 (ref.)	1.00	1.00	1.00	1.00	1.00
11	0.83	0.97	0.86	0.80**	0.90
Family structure
Two parents, same household (ref.)	1.00	1.00	1.00	1.00	1.00
Other	1.21**	1.25***	1.26***	1.21**	1.18*
Migration background
Born in or lived 10+ years in Sweden (ref.)	1.00	1.00	1.00	1.00	1.00
Lived in Sweden less than 10 years	1.06	0.94	0.88	0.89	0.91
Parental university education
No parent	1.13	0.91	0.90	0.94	0.89
At least one parent (ref.)	1.00	1.00	1.00	1.00	1.00
Parental unemployment
No parent (ref.)	1.00	1.00	1.00	1.00	1.00
At least one parent	1.23	1.28	1.26	1.20	1.21

***p < 0.001 **p < 0.01 *p < 0.05.

In Table 3, results from binary logistic regressions with risk gambling as the dependent variable are presented. All school-related factors were associated with risk gambling in the crude analyses (school satisfaction: OR 0.91, p < 0.001; perceived teacher caring: OR 0.95, p < 0.001; perceived school order: OR 0.83, p < 0.001; school performance: OR 0.88, p < 0.001; truancy: OR 2.33, p < 0.001). In Model 1, the associations with both school satisfaction and perceived school order were attenuated but remained statistically significant while the association with perceived teacher caring was no longer statistically significant. The association with school performance was attenuated but remained statistically significant in Model 2, and the association with truancy was amplified and remained statistically significant in Model 3. In Model 4, the associations between risk gambling and perceived teacher caring, perceived school order, school performance, and truancy were statistically significant, whereas the estimate for school satisfaction was non-significant. Regarding the control variables in Model 4, boys and students in the third gender category were more inclined to engage in risk gambling compared with girls, and students in grade 9 were more likely to report risk gambling compared with those in grade 11.

Table 3. Odds ratios (OR) from binary logistic regressions of risk gambling, n = 10,901.

	Risk gambling
		Crude	Model 1	Model 2	Model 3	Model 4
		OR	OR	OR	OR	OR
School satisfaction		0.91***	0.94**			0.98
Perceived teacher caring		0.95***	0.97			0.97*
Perceived school order		0.83***	0.87***			0.88***
School performance		0.88***		0.91***		0.93***
Truancy		2.33***			3.03***	2.53***
Gender
Girls (ref.)		1.00	1.00	1.00	1.00	1.00
Boys		9.42***	9.88***	8.76***	10.30***	10.33***
Other/Missing		9.94***	9.02***	9.30***	9.01***	8.92***
Grade
9 (ref.)		1.00	1.00	1.00	1.00	1.00
11		0.59**	0.68***	0.60***	0.54***	0.61***
Family structure
Two parents, same household (ref.)		1.00	1.00	1.00	1.00	1.00
Other		1.39***	1.36**	1.30**	1.29**	1.20
Migration background
Born in or lived 10+ years in Sweden (ref.)		1.00	1.00	1.00	1.00	1.00
Lived in Sweden less than 10 years		1.70**	1.41*	1.25	1.39*	1.26
Parental university education
No parent		1.46**	1.12	0.94	1.15	0.94
At least one parent (ref.)		1.00	1.00	1.00	1.00	1.00
Parental unemployment
No parent (ref.)		1.00	1.00	1.00	1.00	1.00
At least one parent		1.52*	1.41	1.34	1.30	1.23

***p < 0.001 **p < 0.01 *p < 0.05.

Subsequently, we included interaction terms between all school-related factors and gender or grade, one at a time, in Model 4 for both gambling and risk gambling to check for any moderating effects (not shown in table). These revealed that gender moderated the association between risk gambling and truancy (p < 0.05) as well as the association between risk gambling and school performance (p < 0.01). Gender-stratified analyses (presented in Appendix Table A1) showed that the association between risk gambling and truancy was weaker and not statistically significant for girls compared with boys and the third gender category, while the inverse association between risk gambling and school performance was stronger for girls. Additionally, grade moderated the associations between perceived teacher caring and both gambling (p < 0.001) and risk gambling (p < 0.05). Analyses stratified by grade (presented in Appendix, Table A2) showed that perceived teacher caring was associated with gambling and risk gambling only among students in year 9.

Discussion

The current study examined associations between student gambling and risk gambling and five school-related factors: school satisfaction, perceived teacher caring, perceived school order, school performance, and truancy. Using data collected in 2020 among more than 10,000 students in grade 9 and 11 attending schools in Stockholm Municipality, we found that all studied school-related factors were associated with either gambling or risk gambling, or both. Perceived teacher caring, perceived school order and truancy were associated with both gambling and risk gambling, even when mutually adjusting for all school-related factors simultaneously. School satisfaction was associated with gambling but not with risk gambling and school performance was associated with risk gambling but not with gambling, when mutually adjusting for all school-related factors simultaneously. These findings corroborate previous research that have found that gambling and risk gambling are associated with school factors such as school achievement, truancy, and school connectedness, and indicate that a positive school experience and environment may be beneficial in protecting against gambling engagement among adolescents (Dickson et al., 2008; Shead et al., 2010; Svensson & Sundqvist, 2019) (although Svensson and Sundqvist (2019) did, in contrast to the current study, not report any statistically significant association between school satisfaction and gambling). Worth highlighting is that support from adults in school, in the form of perceived teacher caring, can play a role in protecting against student gambling. To the best of our knowledge, this is a novel finding in relation to gambling. Yet, earlier studies have provided evidence for a link between teacher support and adolescent outcomes such as alcohol use (Chen & Feeley, 2018; Dimitrova et al., 2020; Wormington et al., 2013), smoking and marijuana use (Wormington et al., 2013), self-reported mental health and well-being (Wahlström et al., 2021a, 2021b), and academic achievement (Tennant et al., 2015). Our results also fit in a larger framework of studies that have showed that teacher-reported school conditions at the contextual level, such as school ethos, school leadership, and school collective efficacy, are associated with gambling among students (Låftman et al., 2020; Olsson et al., 2021a, 2021b).

In the current study, 12.1% and 3.9% of the students were categorized as gamblers and risk gamblers, respectively. These figures are generally lower compared to other recent studies conducted in Sweden (Claesdotter-Knutsson et al., 2022; Låftman et al., 2020; Spångberg & Svensson, 2022). However, different instruments have been used across studies (Spångberg & Svensson, 2022) and according to The Public Health Agency of Sweden (2019) there is no general definition of youth problem gambling, which complicates comparison between studies. Additionally, the findings showed that both gambling and risk gambling were more common among grade 9 than among grade 11 students. This was unexpected and in contrast to findings from other studies. For instance, a national Swedish school study from 2021 showed that for boys, gambling was more common among 11th graders than among 9th graders, whereas no clear difference by grade was seen for girls (CAN (The Swedish Council for Information on Alcohol and Other Drugs), 2021). Also, in light of the new, more restrictive gambling regulation in Sweden which includes, e.g. a prohibition of marketing gambling activities to minors (Government Offices of Sweden, 2021), our finding is unanticipated and not easily interpreted.

Adolescents who are more involved in, and enjoy, school, i.e. students who do not play truant, are satisfied with school, have high grades, report being cared for by teachers, and feel that their school provides a safe and orderly environment, may be less incentivized to engage in health risk behaviors such as gambling, compared with students who are more disengaged with school, because they have more to lose by doing so and are often surrounded by peers for whom school is important as well. School-focused students might, for example, regulate their own behavior and avoid activities such as gambling that do not fit in their own perception of themselves as well-behaved adolescents or that they believe might upset or disappoint their teachers (Chen & Feeley, 2018; Rudasill et al., 2010). Similar mechanisms could explain the relationship between risk gambling and the studied school variables. Additionally, risk gambling may in itself negatively affect school involvement and enjoyment as gambling activities become more time-consuming, reduce well-being, and lead to engagement in other health risk behaviors, thus resulting in a reciprocal relationship between risk gambling and various school-related factors.

Additional analyses presented in the Appendix revealed that the association between risk gambling and school performance was more pronounced among girls, while the association between risk gambling and truancy was weaker and not statistically significant for girls. Other studies have, for example, found that the association between school performance and subjective health complaints is stronger for girls (Låftman & Modin, 2012) and that fear of failure, which can predict, e.g. stress, anxiety and depression, is also more prevalent among girls (Borgonovi & Han, 2021). However, risk gambling among adolescent girls is rare and more knowledge is needed about potential gender differences in relation to student gambling.

Strengths and limitations

The current study’s main strength is the large up-to-date data material that enables robust analyses of gambling as well as risk gambling. The inclusion of a third gender category, which to the best of our knowledge has not been analyzed in relation to adolescent gambling before, might also not have been possible without such large-scale data. However, several limitations have to considered as well. Firstly, causal interpretations of the reported associations are severely limited by the fact that cross-sectional data were used. While school-related factors might indeed alter the risk of adolescent gambling, engaging in gambling could also negatively affect the experiences of attending school or school performance. Additionally, it is also possible that gambling, risk gambling, and school-related factors share common determinants, such as personality traits or family conditions other than those included in the current study. Secondly, since the data was collected among students attending schools in Stockholm Municipality, the generalizability of this study’s findings to adolescents in other contexts is limited. The data could also be susceptible to self-report biases, such as differences in interpreting the question. For example, the question about engaging in any kind of gambling is followed by examples of various gambling activities. These range from gambling types that may be considered less harmful such as buying lottery tickets or Loot-boxes to more severe ones such as gambling in a casino and on slot machines. It is not possible for us to know what types of gambling the respondents have engaged in and students who are categorized as gamblers may differ significantly in their gambling experiences. Thirdly, the data were collected in the spring of 2020 which coincides with the first major outbreak of Covid-19 in Stockholm and subsequent government responses, including the decision to transition to only distance education for upper secondary schools. Although this clearly affected the data collection due to absent students and closed schools, and might also have impacted students’ experiences of their school situation, the total response rate is similar to the survey from 2018 (Stockholm Municipality, 2020). Lastly, the measure of school order is unvalidated and has not been used in any prior studies. The low Cronbach’s alpha value for the school order index could also be of some concern. However, school order has theoretically been described as a multifaceted concept, encompassing factors ranging from disruptive behavior in the classroom to the presence of gangs in the school (Cornell & Mayer, 2010; Peguero & Bracy, 2014). With such disparate factors a low Cronbach’s alpha value is not in itself a reason to disregard the measure.

Future research

More research on the connection between school-related factors and gambling behaviors are needed to deepen the knowledge on youth gambling and its determinants. Future studies should consider school-related factors together with conditions at the family level and individual level simultaneously to investigate how these determinants interact and relate to gambling independently. To disentangle causal directions between school-related factors and youth gambling, additional studies using longitudinal data would be beneficial (see for example Fröberg et al., 2015).

Conclusion

Drawing from a sample of adolescents attending schools in Stockholm Municipality, the current study showed that more positive experiences of the situation in school and higher school performance were linked with a lower likelihood of gambling and risk gambling among students. Conversely, truancy was clearly and strongly associated with increased risks of both gambling and risk gambling. The findings suggest that students’ situation in school can help to identify those at risk for gambling problems.

Ethical approval

The study used data from the Stockholm School Survey (SSS). Since participation in the SSS is anonymous, the Regional Ethical Review Board of Stockholm has decided that studies based on the SSS are not under consideration for ethical approval (ref. no. 2010/241-31/5).

Pre registration statement

The authors declare that there is no pre-registration in relation to this study.

Acknowledgements

We are grateful to the students who participated in the data collection. We also thank Stockholm Municipality for giving us access to the data.

Disclosure statement

The study was financed by Svenska Spel’s Research Council (ref. FO2020-0015). Open access funding provided by Stockholm University.

Data availability statement

The data that support the findings of this study are available from Stockholm Municipality. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors with the permission of Stockholm Municipality.

Additional information

Notes on contributors

Joakim Wahlström

Joakim Wahlström has a bachelor’s degree in Social Work and works as a research assistant at the Department of Public Health Sciences at Stockholm University. His research interests include adolescent health and health behaviors.

Sara Brolin Låftman

Sara Brolin Låftman is an Associate Professor of Sociology and Senior Lecturer at the Department of Public Health Sciences, Stockholm University. Her main research interests revolve around young people’s living conditions and health outcomes

Gabriella Olsson

Gabriella Olsson, PhD in Sociology, has focused her research on youth health behaviors, particularly the influence of conditions in schools and the family for such behaviors.

Appendix

Table A1. Odds ratios (OR) from binary logistic regression analyses of gambling and risk gambling, stratified by gender. Fully adjusted models.

	Gambling			Risk gambling
	Girls^a	Boys^b	Other/missing^c	Girls^a	Boys^b	Other/missing^c
	OR	OR	OR	OR	OR	OR
School satisfaction	0.93*	0.94***	1.04	0.90	0.98	1.07
Perceived teacher caring	0.96	0.97*	0.90*	0.96	0.97	0.87*
Perceived school order	0.92	0.92***	0.86*	0.82	0.89**	0.83
School performance	0.99	1.00	1.00	0.83***	0.93***	1.06
Truancy	1.53*	2.53***	2.86**	0.85	2.92***	3.36*
Grade
9 (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
11	0.90	0.92	0.52	0.87	0.60***	0.21
Family structure
Two parents, same household (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Other	1.17	1.18*	1.20	0.81	1.28*	0.64
Migration background
Born in or lived 10+ years in Sweden (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Lived in Sweden less than 10 years	1.39	0.81	1.44	2.39*	1.10	2.25
Parental university education
No parent	1.37*	0.79**	1.21	1.20	0.88	1.51
At least one parent (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
Parental unemployment
No parent (ref.)	1.00	1.00	1.00	1.00	1.00	1.00
At least one parent	1.67	1.02	1.96	1.97	1.18	1.14

***p < 0.001 **p < 0.01 *p < 0.05.

^an = 5,533 ^bn = 5,070 ^cn = 298.

Table A2. Odds ratios (OR) from binary logistic regression analyses of gambling and risk gambling, stratified by grade. Fully adjusted models.

	Gambling		Risk gambling
	Grade 9^a	Grade 11^b	Grade 9^a	Grade 11^b
	OR	OR	OR	OR
School satisfaction	0.96*	0.92***	0.98	0.95
Perceived teacher caring	0.94***	1.01	0.95**	1.02
Perceived school order	0.92***	0.93*	0.87**	0.91
School performance	1.01	0.97*	0.94**	0.91**
Truancy	2.59***	2.06***	2.86***	2.20***
Gender
Girls (ref.)	1.00	1.00	1.00	1.00
Boys	8.83***	8.38***	11.75***	8.36***
Other/Missing	7.78***	4.40***	12.75***	3.17
Family structure
Two parents, same household (ref.)	1.00	1.00	1.00	1.00
Other	1.24*	1.09	1.26	1.02
Migration background
Born in or lived 10+ years in Sweden (ref.)	1.00	1.00	1.00	1.00
Lived in Sweden less than 10 years	1.07	0.74	1.20	1.33
Parental university education
No parent	0.78**	1.08	0.79	1.32
At least one parent (ref.)	1.00	1.00	1.00	1.00
Parental unemployment
No parent (ref.)	1.00	1.00	1.00	1.00
At least one parent	1.24	1.23	1.18	1.35

***p < 0.001 **p < 0.01 *p < 0.05.

^an = 5,913 ^bn = 4,988.