More than disconnecting? School smartphone bans and adolescents' social wellbeing

Smartphones are central to adolescents' social lives, but their pervasive presence in schools has raised concerns about students' social well-being. Many schools have therefore introduced smartphone bans, though evidence for their effectiveness remains limited and mixed. This survey study examined whether the introduction of a total school smartphone ban was associated with changes in adolescents' social well-being. Secondary school students (N = 960; Mage = 14.71, SD = 1.81; 66.6% female) were followed across eight waves over one academic year to track changes in three indicators of social wellbeing: social loneliness at school, friend support at school, and cyberbullying. Data were collected once before the introduction of a total smartphone ban (W1), and seven times afterwards (W2-8). We examined both short-term changes from W1 to W2 and longer-term developments across the post-ban period (W2-W8). No short-term changes emerged in social loneliness, friend support, or cyberbullying victimization, although cyberbullying perpetration declined slightly. Over time, social loneliness increased modestly, friend support remained stable, and cyberbullying fluctuated at low levels. Within the specific conditions of this school, a total smartphone ban was not associated with clear improvements in adolescents' social wellbeing; effective school policies may therefore need to combine such restrictions with broader initiatives that foster social connection and digital literacy.

By the first year of secondary school, nearly all adolescents own a smartphone, which they primarily use for social media and peer communication (Vanwynsberghe et al., 2024). Much of this activity occurs during school hours, prompting concerns among educators, parents, and policymakers about how smartphone use affects students' functioning at school (Global Education Monitoring Report Team, 2023; Radesky et al., 2023). Within this context, smartphone use is often seen as a potential contributor to concentration problems, reduced academic performance and mental health problems (Campbell et al., 2024; Wood et al., 2023), while it has simultaneously raised concerns about students' social wellbeing (Böttger and Zierer, 2024; Pouwels et al., 2024). Smartphones provide adolescents with continuous social connection, which can foster support and belonging, yet they also risk displacing face-to-face interactions which may increase social loneliness, and facilitate cyberbullying (Böttger and Zierer, 2024; Pouwels et al., 2024; Selwyn and Aagaard, 2020). This dual role has fueled international debate about whether restricting smartphone use during the school day can help safeguard students' social wellbeing.

In response to these concerns, schools worldwide have adopted a range of smartphone policies, from total bans to partial restrictions. These measures are typically justified by the expectation that reduced smartphone use will improve academic outcomes, increase in-person interaction and mitigate risks such as cyberbullying (Abrahamsson, 2024; Böttger and Zierer, 2024; Pouwels et al., 2024; Selwyn and Aagaard, 2020). Yet, the empirical evidence remains limited and inconsistent (Ferguson, 2025), and lacks the longitudinal depth needed to distinguish short-term reactions from more sustained changes. Notably, the few longitudinal or quasi-experimental studies that do exist have predominantly examined academic outcomes (e.g., King et al., 2024; Kuznekoff and Titsworth, 2013), leaving the longer term social consequences of smartphone bans underexplored.

This study addresses these gaps by evaluating a school policy that introduced a total smartphone ban and examining how it relates to adolescents' social wellbeing across one academic year. We focus on three complementary indicators of adolescents' social wellbeing: social loneliness at school, friend support at school, and cyberbullying (both victimization and perpetration). These indicators capture distinct but related aspects of adolescents' social wellbeing, ranging from positive and supportive peer relationships to deficits in social connectedness and negative or hostile peer interactions, thereby providing a multidimensional perspective on social wellbeing in the school context (Brown and Larson, 2009; Liu et al., 2025; Mitic et al., 2021). Using a multi-wave longitudinal quasi-experimental design with one pretest and seven monthly follow-ups, this study assesses both short-term changes and longer-term trajectories following the introduction of the ban. In doing so, this policy evaluation contributes much-needed empirical evidence to ongoing debates about the social implications of school smartphone policies.

Social wellbeing in the school context and the dual role of smartphones

Social wellbeing refers to individuals' perceptions of being connected, supported, and valued within their communities (Keyes, 1998). It is particularly salient during adolescence, a period marked by increasing autonomy, heightened sensitivity to peer validation, and shifting social priorities (Pfeifer and Berkman, 2018). These developmental changes can make it challenging for adolescents to establish and maintain supportive peer relationships, although they are essential for social wellbeing (Backes and Bonnie, 2019; Lamblin et al., 2017).

The school context plays a decisive role in shaping these experiences, as adolescents spend much of their daily lives at school, where they form and maintain peer relationships (Kumari and Biswas, 2023; Marin and Brown, 2008). Yet, these peer relationships increasingly unfold not only face-to-face but also through digital technologies, particularly smartphones (Nesi et al., 2018). Because most adolescents receive their first smartphone upon entering secondary school, the school environment becomes a primary setting in which digital social interactions unfold (Vanwynsberghe et al., 2024), and smartphones have thereby become deeply embedded in adolescents' daily lives at school (Pouwels et al., 2024; Radesky et al., 2023). While smartphones can facilitate connection and support, they also introduce challenges such as displacement of in-person interactions and exposure to cyberbullying (e.g., Böttger and Zierer, 2024; Selwyn and Aagaard, 2020; Spiratos and Ratanasiripong, 2023; Van Der Meulen et al., 2021).

The stimulation and displacement hypotheses (Valkenburg and Peter, 2007) provide a useful framework for understanding these dynamics. The stimulation hypothesis posits that communication via smartphones, particularly with close peers, can enrich adolescents' social lives at school by helping them maintain support networks and feel connected even when physically apart (Campbell et al., 2024; Pouwels et al., 2024; Valkenburg and Peter, 2007). For adolescents experiencing social loneliness within their classroom or peer group, access to trusted friends via smartphones may foster a sense of belonging (Bae, 2018; Huang et al., 2022; Pouwels et al., 2024). Conversely, the displacement hypothesis warns that digital interactions may replace face-to-face exchanges (Valkenburg and Peter, 2007). When smartphone use becomes excessive, compulsive, or disconnected from offline relationships, its capacity to support social wellbeing diminishes (Lapierre and Zhao, 2021; Pereira et al., 2020; Spiratos and Ratanasiripong, 2023). This risk is particularly salient in schools, where smartphone engagement can crowd out in-person interactions (Pouwels et al., 2024). Over time, such displacement may lead to increased social loneliness, as superficial or asynchronous digital communication substitutes for more meaningful relationships (Mitic et al., 2021; Pouwels et al., 2024; Van Der Meulen et al., 2021).

Smartphones may also shift boundaries that typically constrain aggressive behavior (Smith et al., 2008). Continuous, largely unsupervised connectivity creates conditions in which antagonistic exchanges can emerge, contributing to the prevalence of cyberbullying (Smith et al., 2008; Wang and Jiang, 2022). The online disinhibition effect further explains these patterns: reduced face-to-face interaction and diminished physical presence can lower social inhibitions and facilitate hostile communication (Suler, 2004). Moreover, the constant production and circulation of digital content during school hours, such as photos or messages, may supply material for cyberbullying, even though the harassment itself typically unfolds afterward (Agatston et al., 2007; Singh et al., 2017; Smith et al., 2008). Its consequences nonetheless become woven into students' everyday school experiences (Pennell et al., 2020; Smith et al., 2008; Wang et al., 2010), often in ways that remain invisible to teachers (Englander et al., 2017; Selwyn and Aagaard, 2020; Smale et al., 2021).

Against this backdrop, smartphones may positively and negatively influence adolescents' social wellbeing at school through several interconnected aspects of their peer relationships (Böttger and Zierer, 2024; Mitic et al., 2021; Vanluydt et al., 2026). To capture this dual influence, we focus on three indicators that are particularly relevant to adolescents' social experiences in school: social loneliness, cyberbullying, and friend support. Although conceptually distinct, each indicator individually is well-documented as relevant to adolescents' social experiences at school (Jefferson et al., 2023a; Smith et al., 2008; Tomás et al., 2020), and they capture complementary aspects of adolescents' social wellbeing: friend support reflects the presence of positive and supportive peer relationships, social loneliness reflects perceived deficits in social connectedness, and cyberbullying reflects negative or harmful peer interactions (Brown and Larson, 2009; Liu et al., 2025; Mitic et al., 2021). Considering these indicators together therefore provides a multidimensional perspective on adolescents' social wellbeing that encompasses both positive and adverse aspects of their social experiences within a digitally mediated school context. This approach is supported by Meier and Reinecke (2020) two-continua model of mental health, which posits that wellbeing is defined not only by the presence of positive experiences, but also by the absence of negative ones. Within this framework, social loneliness at school and cyberbullying can be understood as risk indicators, whereas friend support functions as a protective factor (Meier and Reinecke, 2020). These three indicators are discussed in more detail below.

First, social loneliness at school arises when students lack meaningful connections within their immediate school environment, despite being surrounded by peers (De Jong Gierveld and Havens, 2004; Krause-Parello, 2008). Such loneliness may involve feeling excluded by classmates, wishing for greater inclusion, or lacking a sense of belonging (Pouwels et al., 2024). These experiences can undermine mental health (Arslan, 2021; Jefferson et al., 2023b; Santini et al., 2021), whereas strong school belonging and social cohesion are consistently associated with positive psychological and social outcomes (Allen et al., 2016; Craggs and Kelly, 2018; Goodenow, 1993; Packer and Ungson, 2024; Pittman and Richmond, 2007; Veerman and Denessen, 2021).

Second, defined as repeated online harassment or exclusion, cyberbullying can involve posting hurtful content, spreading rumors, or issuing threats (Patchin and Hinduja, 2015). Although cyberbullying often occurs outside school hours, its social and emotional consequences spill over into the classroom (Pennell et al., 2020; Smith et al., 2008; Wang et al., 2010). Victims frequently report diminished belonging, difficulty forming friendships, and heightened social anxiety (Pennell et al., 2020; Smith et al., 2008; Von Marées and Petermann, 2012). Perpetrators, too, may experience strained peer relationships and fewer opportunities to build positive social connections (Camacho et al., 2023; Kircaburun et al., 2018).

Third, friend support encompasses instrumental support (having friends to rely on for help) and emotional support (being able to confide in friends) (Lubben et al., 2006; McLaughlin and Clarke, 2010). Such support enhances adolescents' social wellbeing at school and buffers against risks such as bullying and social exclusion (Phan et al., 2025; Van Der Meulen et al., 2021).

Taken together, these three indicators reflect key dimensions through which smartphones may shape adolescents' social experiences at school, both by facilitating connection and by introducing risks such as displacement and cyberbullying (Liu et al., 2025; Meier and Reinecke, 2020; Mitic et al., 2021; Vanluydt et al., 2026). Because smartphones are embedded in adolescents' peer interactions in ways that can be both beneficial and detrimental, questions have emerged regarding how schools should manage smartphone use during the school day. In response, an increasing number of schools have implemented smartphone bans or restrictions (Böttger and Zierer, 2024; Pouwels et al., 2024; Vanluydt et al., 2026). However, whether such policies improve adolescents' social wellbeing remains unclear.

Smartphone bans: the solution?

Institutions such as UNESCO (Global Education Monitoring Report Team, 2023) and various researchers (e.g., Das and Ahmed, 2023; Lepp et al., 2013; Selwyn and Aagaard, 2020) recommend limiting smartphone use in schools to address concerns about declining learning outcomes, mental health problems, and deteriorating peer relations. Yet, despite these recommendations, definitions and implementations of smartphone bans vary considerably, complicating cross-study comparisons (Campbell et al., 2024). In practice, some schools enforce total bans, requiring students to leave their smartphones at home or store them in lockers, hotels or pouches (Pouwels et al., 2024), whereas others adopt partial bans that prohibit use during lessons but allow access during breaks (Gath et al., 2024). Although such policies are increasingly common, current research offers little consistent evidence that they significantly improve student wellbeing or academic performance (Böttger and Zierer, 2024; Ferguson, 2025).

With respect to social wellbeing specifically, findings are particularly mixed (Böttger and Zierer, 2024; Campbell et al., 2024). Qualitative analyses (Gath et al., 2024; Pouwels et al., 2024) highlight this ambivalence: some students report more face-to-face interaction and reduced bullying under bans, whereas others describe feeling excluded when denied access to digital spaces that provide comfort and connection. Quantitative evidence reflects similar inconsistencies. Some studies associate bans with lower levels of cyberbullying or modest improvements in peer relations, while others report no significant or even adverse associations (Campbell et al., 2024; Gath et al., 2024; Kopecký et al., 2021; Porter et al., 2015).

Rapid and scoping reviews reach comparable conclusions. Böttger and Zierer (2024) find that bans may reduce social problems such as bullying, but they emphasize the limited scope and methodological heterogeneity of existing studies. Campbell et al. (2024) note that the field is dominated by descriptive and correlational research that cannot establish causality. Much of the available work focuses on short-term outcomes or lacks experimental and longitudinal designs, restricting insight into potential longer-term effects (Campbell et al., 2024; Wood et al., 2023).

Only a few studies have used longitudinal or quasi-experimental designs. Abrahamsson (2024), for instance, conducted a large-scale event study in Norwegian schools, combining national student survey data on smartphone policies with administrative records on mental health, academic performance, and bullying between 2007 and 2019. The staggered implementation of bans across schools enabled quasi-experimental comparisons over time, and results suggested reductions in psychological complaints and bullying rates. However, several methodological limitations temper these conclusions. Adherence to the policy and consistency of implementation across schools remain unclear; bullying was measured via school-level averages, limiting insight into within-school variation; and the design included only annual data points (2010–2018) without measurements immediately preceding the ban, preventing assessment of short-term effects. Similarly, King et al. (2024) employed a natural experiment. Two groups of schools were compared: the ‘ban' schools, which had already implemented the ban at baseline, and the ‘no-ban' schools, which had not yet implemented the ban at baseline, but introduced it shortly afterward. Students in these schools completed surveys at two time points: initially at baseline, and then again approximately one to two months later, after the no-ban schools had implemented the policy. This design allowed them to capture initial reactions rather than longer-term changes. Results indicated no significant differences between ban and no-ban schools in school belonging, academic engagement or problematic phone use, but bullying declined significantly in both groups over time.

The lack of temporal depth is thus a recurring limitation in the field. Theories of behavioral change and psychological reactance (Steindl et al., 2015; Tatum et al., 2018) suggest that restrictive measures may first elicit frustration or resistance before new social norms emerge and adjustment occurs. Furthermore, effects of policy interventions are unlikely to emerge instantaneously. In particular, indicators of social wellbeing are embedded in longer-term peer relationships and interaction patterns that tend to evolve gradually over time (Rubin et al., 2011). Short-term or single follow-up designs may therefore be insufficient to capture meaningful changes in these outcomes. Evidence from preventive intervention research suggests that early null or mixed findings often reflect follow-up periods that are too limited to detect change, rather than a genuine absence of effects (Hill et al., 2016). Without fine-grained, multi-wave longitudinal data, null or mixed findings may reflect a mismatch between the timing of measurement and the temporal dynamics of the underlying social processes, rather than a true absence of effects. Capturing these dynamics requires repeated measurements within the same school year to assess whether changes emerge gradually, stabilize, intensify, or dissipate over time.

The current study

This study evaluates a school policy by examining how the introduction of a total smartphone ban in secondary education relates to social loneliness at school, friend support at school, and cyberbullying. Despite growing interest in the effects of school smartphone bans, existing research remains limited in both scope and methodology. To date, few studies have employed a multi-wave, long-term longitudinal design capable of capturing both short- and longer-term changes in students' social wellbeing indicators following policy implementation.

Examining developments across an entire academic year offers an opportunity to trace how changes in social wellbeing emerge, develop, and stabilize within the school context. The present study employs a quasi-experimental design with eight waves: one pretest conducted two weeks before the implementation of a total smartphone ban, followed by seven monthly post-ban waves. This design allows us to model short-term changes relatively quickly after the policy change as well as longer-term trajectories across the remainder of the school year. Distinguishing between these time scales is crucial, as school policies are often evaluated shortly after implementation, while their social consequences may only become visible over time (King et al., 2024; Rubin et al., 2011).

By adopting this approach, the study contributes much-needed empirical evidence to academic and policy debates about the social implications of smartphone bans in schools. Specifically, it addresses the following research questions:

What is the short-term impact of the introduction of a smartphone ban on students' social loneliness at school (RQ1), cyberbullying experiences, both as perpetrators and victims (RQ2), and students' sense of friend support at school (RQ3)?

How does students' social loneliness at school (RQ4), cyberbullying experiences, both as perpetrators and victims (RQ5), and students' sense of friend support (RQ6) develop across the subsequent school year following the implementation of the smartphone ban?

When examining this, it is important to also consider gender and age. They were included as control variables as previous research indicates differences in social loneliness, friend support and cyberbullying. Girls generally report greater friend support, more loneliness and are more often victims of cyberbullying, whereas boys tend to report higher social loneliness specifically and are more often perpetrators of cyberbullying (Bokhorst et al., 2009; Craig et al., 2020; Cosma et al., 2022; Vanwynsberghe et al., 2024; Von Soest et al., 2020). Loneliness tends to decrease with age, whereas friend support becomes increasingly important and cyberbullying is somewhat more common among older adolescents (Von Soest et al., 2020; Bokhorst et al., 2009; Vanwynsberghe et al., 2026). Smartphone use may also differ by gender and age. Girls tend to use smartphones more for social purposes, whereas boys use them more for entertainment and gaming. Younger adolescents tend to use smartphones more for entertainment, while older adolescents use them more for communication (Vanwynsberghe et al., 2026). Although evidence on these differences in smartphone use specifically during school hours is limited, similar patterns may apply in the school context. In regard to smartphone bans, girls report more negative social consequences after a school smartphone ban (Vanwynsberghe et al., 2026). Full smartphone bans are associated with reduced school belonging for girls specifically (Vanluydt et al., 2026), but no significant gender difference emerges for decreases in bullying (Abrahamsson, 2024; Vanluydt et al., 2026). Being older also often meant being less compliant with bans (King et al., 2024).

Sample and procedure

This study was part of a larger project entitled “Smartphone Shutdown,” which examined the implications of different types of smartphone bans across multiple outcomes. Several secondary schools in Flanders participated, with each sub study focusing on distinct outcomes and school samples while using broadly similar methodological designs. Ethical approval was obtained G-2024-8306-R9(MIN). During the summer period, schools were invited to participate. The pupils' parents provided passive informed consent, and students gave active assent by signing an informed consent form at the start of each data collection. To encourage engagement, students were informed about the study's relevance for understanding youth wellbeing and informing school policy. Confidentiality and anonymity were ensured throughout. Participants were informed about these safeguards at the beginning of the survey, and identifying information was stored separately from survey responses through the use of a unique code.

For the present study, focusing on social wellbeing, data were collected from the full student population of a relatively large public secondary school, located in a medium-sized town in Flanders, offering general, professional-technical and vocational education tracks. A quasi-experimental longitudinal design was implemented across one school year, consisting of one pretest (W1 = Oct. 2024) and seven posttests (W2 = Nov. 2024, W3 = Dec. 2024, W4 = Jan. 2025, W5 = Feb. 2025, W6 = March 2025, W7 = April 2025, W8 = May 2025). Between W1 and W2, the school introduced a total smartphone ban, prohibiting smartphone use during both lessons and breaks. Prior to the total ban, students were allowed to use smartphones during breaks, while use during lessons was restricted. In practice, however, this restriction was only weakly enforced. According to conversations with the school, teachers varied in whether and how they addressed smartphone use in class. The intervention therefore represented a transition from a partially restricted and weakly enforced smartphone policy to a strongly enforced total smartphone ban, under which smartphone use was prohibited throughout the school day, including during breaks. To provide a rough check on adherence to the new policy, students were asked to what extent they followed the smartphone rules. In W8, 63.3% of students reported adhering to the ban, providing an indication that the intervention was implemented, although this also suggests that a substantial share of students did not fully comply.

A total of 1,045 students participated. Those who completed only one survey were excluded (N = 85). This resulted in 4,792 valid surveys, with students completing on average 4.99 surveys (SD = 1.74). Among the remaining 960 participants, full participation would have yielded 7,680 surveys, corresponding to an overall compliance rate of 62.4%. Most attrition stemmed from practical constraints, such as teachers not always being able to organize survey administration. In each wave, the designated school contact received a link to a LimeSurvey questionnaire, which teachers distributed to students during class, where surveys were completed under supervision. Surveys took approximately 15 min and were pilot tested beforehand.

Demographic information was collected at W1. The mean age was 14.71 (SD = 1.81, range 11–23). Although only 9 students were over 18, they were retained because the transition to adulthood is not defined by age alone and all participants were still enrolled in secondary education. In terms of gender, 66.64% identified as female, 32.37% as male, and.91% preferred not to disclose their gender. Regarding migration background, 81.90% reported that their grandparents originated from Belgium. In line with common approaches in European research contexts, country of origin was used as an indicator of migration background, as racial categories are less commonly applied (Payne-Sturges, 2025; Vietze et al., 2022).

Measures

Social loneliness at school

Social loneliness at school was measured with three items adapted from Pouwels et al. (2024). Students were asked to reflect on the past month and indicate the extent to which the following applied to them: (1) “I felt excluded by other students at school”, (2) “I wished other students would include me more often at school”, and (3) “I felt like I belonged to a group at school”. Responses ranged from 1 (“Did not apply to me at all”) to 5 (“Completely applied to me”). The third item was reverse-coded. A principal axis factoring (PAF) analysis revealed a one-factor solution (Initial Eigenvalue = 1.56; Explained Variance = 52%). Items 1 and 2 loaded strongly (λ = 0.88 for both), whereas the reverse-coded item showed a very low loading (λ = 0.11). Internal consistency was low when including all three items (α = 0.51), but improved substantially when item 3 was removed (α = 0.87). Thus, only items 1 and 2 were retained, and their mean score was used (M = 1.73, SD = 0.87), with higher scores reflecting higher levels of social loneliness.

Friend support at school

Friend support at school was assessed with two items adapted from the Lubben Social Network Scale (Lubben et al., 2006): “How many best friends could you turn to during school hours in the past month if you needed help?” and “How many best friends did you have at school in the past month with whom you could talk about everything, including personal matters?”. Response categories ranged from 0 (“none”) to 5 (“nine or more friends”). Items were summed to create a composite friend support score (M = 6.98, SD = 2.38), with higher values indicating more supportive friendships at school.

Cyberbullying victimization and perpetration

Cyberbullying was assessed using two items based on Patchin and Hinduja (2015). Participants were first presented with a short definition of cyberbullying and then asked to indicate (yes = 1, no = 0): “Were you yourself a victim of any of such online behaviors in the past month?” (92.9% ‘no', 7.1% ‘yes') and “Did you engage in any of such online behaviors yourself in the past month?” (95.5% ‘no', 4.5% ‘yes').

Control variables

Gender was assessed by asking students to self-identify as a girl, boy, or prefer not to say. Because few participants selected the third option, gender was recoded into a binary variable for use as a control. Participants also reported their year of birth; this was converted into age, which served as an additional control variable.

Analytical strategy

Descriptive analyses in R included means and standard deviations for continuous outcomes (friend support and social loneliness at school) and frequencies for binary outcomes (cyberbullying victimization and perpetration) at each wave. Correlations among all study variables were computed at both the between- and within-person levels using the utils package in R (Bengtsson, 2005). Mixed-effects models were first estimated to determine whether age and gender should be retained as control variables.

Analyses proceeded in two phases: (1) short-term changes following the smartphone ban (RQ1–RQ3), and (2) long-term developments afterward (RQ4–RQ6). Linear mixed-effects models were used for continuous outcomes, and logistic mixed-effects models for binary outcomes. For analyses requiring complete data, a subset of 3,099 observations from 805 participants was used. This complete-case sample was used for all multilevel models examining longer-term trajectories (waves 2–8), including empty models, control models, and linear and non-linear growth models.

Short-term changes following the smartphone ban (RQ1-RQ3)

Short-term changes from W1 (pre-ban) to W2 (two weeks post-ban) were assessed by coding time categorically (W1 = 0, W2 = 1), with repeated measures nested within participants. With only two waves, models included random intercepts but not random slopes.

For social loneliness (RQ1), a linear mixed-effects model initially specified, failed to converge due to near-zero within-person variance; mean loneliness scores were nearly identical across waves (W1: M = 1.70, SD = 0.83; W2: M = 1.69, SD = 0.85). The fixed effect of time was also zero [ t (df = 614.70) = 0, p = 1], indicating minimal change over time. Therefore, a paired-samples t -test was used to test for mean-level differences.

Cyberbullying victimization and perpetration (RQ2) were analyzed using logistic mixed-effects models with random intercepts. Odds ratios, pairwise comparisons, and predicted probabilities were derived using the emmeans package (Lenth, 2025).

Friend support at school (RQ3) was analyzed using a linear mixed-effects model with random intercepts using the lme4 package (Bates et al., 2015), with estimated marginal means and predicted trajectories calculated via emmeans (Lenth, 2025).

Long-term developments after the implementation of the smartphone ban (RQ4-RQ6)

To examine developments from W2 to W8 multilevel growth models were estimated using lme4 (Bates et al., 2015). This approach accounts for the hierarchical data structure (repeated measures nested within participants) and allows modeling both interindividual differences and intraindividual change. Intercept-only models were first estimated to compute intra-class correlation coefficients (ICCs) and determine the proportion of variance at the between- and within-person levels. Time was coded continuously (W2 = 0, W8 = 6) and linear, quadratic, and cubic terms were tested to capture potential non-linear patterns.

Models began with random intercepts and fixed slopes; random slopes for time were added to account for individual variability in change rates. If models failed to converge, specifications were simplified to random-intercept-only. Model fit was compared using likelihood ratio tests, AIC, and BIC. Linear mixed-effects models were used for social loneliness (RQ4) and friend support (RQ6), and logistic mixed-effects models for cyberbullying victimization and perpetration (RQ5). ICCs for binary outcomes were computed on the latent (logit) scale using random intercept variance and the theoretical level-1 variance (π 2 /3) (Nakagawa et al., 2017). All materials, including survey, data, code can be found on the Open Science Framework (https://osf.io/zmywd?view_only=7a406f91e0e84d3c9eab5ded4d1ed2b6).

Descriptive statistics

We first computed descriptive statistics, including means and standard deviations, for all outcome variables across waves (Table 1). Between- and within-person correlations showed that the social wellbeing indicators were moderately to strongly related at both levels (Table 2). Social loneliness at school correlated negatively with friend support and negatively with both cyberbullying victimization and perpetration at the within- and between-person levels. Cyberbullying victimization and perpetration were positively correlated with each other at both levels. Within-person correlations were generally weaker than between-person correlations.

The ICC for social loneliness at school was 0.52, indicating that 52% of the variance reflected between-person differences, with the remainder attributable to within-person variation and error variance. The ICC for friend support at school was 0.61. For cyberbullying victimization and perpetration, ICCs were estimated on the latent (logit) scale (Nakagawa et al., 2017), yielding ICCs of 0.65 and 0.48, respectively, consistent with substantial between-person stability in these binary outcomes.

Model testing

We first examined whether age and gender should be included as controls. For friend support and social loneliness, linear mixed-effects models including age and gender as fixed effects were compared to the corresponding empty models using likelihood ratio tests. Adding these controls significantly improved model fit for friend support [χ 2 (2) = 31.05, p < 0.001] and social loneliness [χ 2 (2) = 11.83, p = 0.003]. Age negatively predicted friend support (b = −0.21, SE =0.04, p < 0.001). For social loneliness, gender was a significant predictor (b = −0.18, SE =.05, p = 0.001), with boys reporting lower loneliness than girls.

For cyberbullying victimization and cyberbullying perpetration, adding age and gender did not improve model fit (p = 0.915 and p = 0.116, respectively). Accordingly, age was included as a covariate in the friend support models, gender in the social loneliness models, and no control variables were retained for the cyberbullying models.

Short-term changes of the smartphone ban (W1-W2)

To assess short-term effects of the smartphone ban, social loneliness and friend support were modeled from W1 (pre-ban) to W2 (two weeks post-ban) using mixed-effects models with time as a categorical predictor. Short-term changes in cyberbullying were examined using two generalized linear mixed models (binomial, logit) for victimization and perpetration.

Model with social loneliness (RQ1)

Due to convergence issues and near-zero within-person variance across the first two waves, a paired samples t -test was conducted. Results confirmed no significant change from W1 to W2 [ t (588) = 0.34, p = 0.737], with a negligible mean difference of 0.009.

Model with cyberbullying victimization (RQ2)

A generalized logistic mixed model with time as predictor indicated no significant change in victimization between W1 and W2 (β = −0.04, SE = 0.42, p = 0.922). The odds ratio for W2 relative to W1 was OR = 1.04, reflecting stability at very low prevalence levels.

Model with cyberbullying perpetration (RQ2)

A similar model for perpetration showed significant decrease at W2 (β = −1.75, SE = 0.68, p = 0.010). The odds at W2 were OR = 5.78 lower than at W1, indicating a drop in perpetration. However, given the extremely low prevalence, this change should be interpreted cautiously. Figure 1 shows the predicted probabilities, which remain near zero despite the statistical significance.

Model with friend support (RQ3)

A linear mixed-effects model controlling for age indicated a small, non-significant decline in friend support at W2 (β = −0.14, SE = 0.079, p = 0.076). Estimated marginal means were M = 7.01 (SE = 0.08) at W1 and M = 6.87 (SE = 0.09) at W2. Age remained a significant predictor (β = −0.14, SE = 0.04, p = 0.001), with older students reporting slightly less friend support. Overall, no significant short-term change in friend support following the ban was observed (p = 0.076).

Long-term developments after the implementation of the smartphone ban (W2-W8)

Model with social loneliness (RQ4)

Longer-term changes were modeled using linear mixed models with random intercepts and slopes. A random-slopes model fitted better than a random-intercepts-only model [χ 2 (df = 2) = 34.57, p < 0.001]. Adding quadratic or cubic time terms did not improve fit (p = 0.412 and p = 0.114). The final linear model indicated a modest but significant increase in social loneliness over time (β = 0.02, SE = 0.01, p = 0.002; Figure 2). Gender remained significant, with boys reporting lower levels than girls (β = −0.18, SE = 0.05, p < 0.001). A negative intercept-slope correlation (r = −0.29) indicated that students who began with lower loneliness tended to increase more over time.

Model with cyberbullying victimization (RQ5)

Logistic growth models indicated that adding a quadratic time term improved model fit [χ 2 (df = 1) = 18.55, p < 0.001], and adding a cubic term improved it further [χ 2 (df = 2) = 23.27, p < 0.001]. In the final model, only the cubic time effect was statistically significant (β = −0.07, SE = 0.03, p = 0.033), while both the linear (β = −0.12, p = 0.861) and quadratic (β = 0.34, p = 0.223) terms were not. Although this suggests a slight undulating pattern, absolute changes in predicted probabilities were minimal due to the low prevalence of victimization (7.10%), and should not be interpreted as meaningful change.

Model with cyberbullying perpetration (RQ5)

Growth models for perpetration also favored a quadratic over a linear model [χ 2 (df = 1) = 15.31, p < 0.001]. However, prevalence was extremely low (4.45%), and predicted probabilities for later waves had very wide confidence intervals (extending up to 0–99%), indicating substantial numerical instability. Although both linear (β = 1.14, SE = 0.01, p < 0.001) and quadratic (β = −0.42, SE = 0.01, p < 0.001) terms were significant, the resulting estimates cannot be meaningfully interpreted.

Model with friend support (RQ6)

Friend support at school was examined across waves using a linear mixed model. A random-slopes model fitted better than an intercept-only model [χ 2 (2) = 31.62, p < 0.001], while the inclusion of quadratic or cubic time terms did not improve model fit. Overall, friend support remained stable over time (β = −0.003, SE = 0.01, p = 0.864). Age was a significant negative predictor, indicating slightly lower levels of friend support among older adolescents (β = −0.21, SE = 0.04, p < 0.001). A negative intercept-slope correlation (r = −0.33) indicated that students who initially reported higher friend support at school tended to show a minor decline over time.

Smartphone bans are increasingly implemented in schools, based on the belief that they can improve students' social wellbeing, mental health or academic performance. However, empirical support for this assumption remains limited and inconsistent. Although some evidence points to potential benefits of smartphone restrictions for academic outcomes or classroom distraction (Campbell et al., 2024; King et al., 2024; Kuznekoff and Titsworth, 2013; Wood et al., 2023), evidence regarding their effects on students' social wellbeing specifically remains scarce and inconsistent. The present study therefore evaluated a school policy by examining the effects of a transition from a partial, weakly enforced smartphone policy to a strongly enforced total smartphone ban in a secondary school on several indicators of students' social wellbeing. Specifically, we assessed changes in social loneliness at school, friend support at school, and cyberbullying (both victimization and perpetration) from pre-ban to two weeks post-ban, and we explored longer-term developments in these outcomes across the school year following the ban's implementation. Overall, we found no short-term changes apart from a small decline in cyberbullying perpetration, and a modest longer-term increase in social loneliness at school. These findings underline the complexity of using smartphone bans as a strategy to promote social wellbeing, but, as elaborated below, should be interpreted with appropriate caution given the design of the present study. The main results and their implications for school smartphone policies are discussed.

Key findings

First, there were no significant short-term changes in social loneliness two weeks after the ban compared to pre-ban levels. This is consistent with the notion that social loneliness is rooted in relatively stable relational contexts, such as perceived belonging, peer acceptance, and friendship stability, which tend to change gradually and require sustained opportunities for interaction (Goossens, 2018; Jefferson et al., 2023b; Laursen and Hartl, 2013). A two-week window following policy implementation is therefore likely too brief for detectable shifts in loneliness to emerge.

Over the longer term, however, a small but statistically significant increase in social loneliness was observed. Though modest, the gradual rise suggests that restrictive smartphone policies may intersect with adolescents' broader social adjustment over the school year. Such changes in social wellbeing may unfold slowly and therefore become apparent only across the longer-term, rather than shortly after policy implementation (Rubin et al., 2011; Hill et al., 2016).

Because the study lacked pre-ban developmental trajectories and a control group, it remains unclear whether this increase represents a deviation from typical developmental patterns or a continuation of pre-existing trends. Prior work has shown that loneliness can fluctuate naturally during adolescence and across the school year, as peer groups reorganize and academic or social pressures vary (Goossens, 2018; Jefferson et al., 2023a; Maes et al., 2019; McElhaney et al., 2008). Nonetheless, the timing of the increase, emerging after the introduction of a total smartphone ban, raises the possibility that restricting smartphone use may have unintentionally amplified students' feelings of loneliness at school.

Drawing on the stimulation and displacement hypotheses (Valkenburg and Peter, 2007), smartphones can both bolster and hinder social wellbeing. For well-integrated adolescents, removing smartphones may have little impact, as they can rely on established offline networks (Böttger and Zierer, 2024; Ihm, 2018). For others, however, smartphones may function as a crucial social scaffold, providing access to support, contact with friends outside the classroom, or a sense of inclusion during breaks; especially for students who struggle to feel included offline (Bae, 2018; Pouwels et al., 2024). When this scaffold is removed, these students may become more acutely aware of their social isolation, leading to heightened feelings of social loneliness at school (Pouwels et al., 2024). The increase found in this study may therefore reflect the presence of students who rely on smartphones to compensate for limited offline belonging and who experience the ban as socially restrictive. Although the increase emerged after the introduction of the smartphone ban, the findings should be interpreted with caution due to a lack of a control group and pre-ban developmental trajectories. They indicate a rise in loneliness following the ban, but do not provide definitive evidence that the ban itself caused this increase.

Second, friend support at school showed no significant short-term changes following the smartphone ban, nor any systematic long-term trajectory. This stability suggests that the ban neither enhanced nor undermined adolescents' friend support at school. Friendships typically develop and consolidate through repeated positive interactions and shared experiences (Algoe, 2019; Mitic et al., 2021), and it may therefore be that the structural opportunities for building and maintaining friendships within this school context remained largely unchanged (Jefferson et al., 2023b). The stable pattern over time is consistent with the idea that friend support reflects relatively enduring relational ties that are less susceptible to disruption or reinforcement through a single policy measure such a smartphone ban.

Third, we observed no significant change in cyberbullying victimization in the short term, but a small reduction in perpetration two weeks after the ban. The stability of victimization aligns with evidence that much cyberbullying occurs outside school contexts or shifts to other devices (laptops, tablets) and thus continues despite in-school smartphone restrictions (Campbell et al., 2024; Selwyn and Aagaard, 2020; Smith et al., 2008; Pennell et al., 2020). The modest decline in perpetration may indicate that limiting smartphone access temporarily reduces opportunities to engage in such behaviors, for example by constraining students' ability to record, share, or post harmful content during school hours (Böttger and Zierer, 2024; Gath et al., 2024).

Across the school year, more complex, non-linear trajectories appeared to emerge for both victimization and perpetration. However, because both forms of cyberbullying occurred at very low base rates and several models were numerically unstable, these results should be interpreted with caution and not treated as robust evidence of meaningful changes.

Taken together, and contrary to widespread expectations that smartphone bans will directly strengthen peer relationships and reduce online risks (Campbell et al., 2024; Global Education Monitoring Report Team, 2023), our findings suggest that, within the specific conditions of this school, adolescents' social wellbeing at school may be less directly shaped by smartphone access than often assumed. It is possible that social wellbeing reflects a broader set of contextual and relational conditions within the school environment (Allen et al., 2016; Crosnoe, 2021; Meier and Reinecke, 2020). Allen et al. (2016), for instance, emphasize supportive peer and teacher relationships as key building blocks of belonging, while Crosnoe (2021) points to how school climates and social structures influence how students navigate their social worlds. Social wellbeing thus emerges from multiple, interrelated layers (Keyes, 1998; Ryff and Singer, 2006), which together shape how students feel connected and supported at school.

From this perspective, it may be that a single policy intervention such as a smartphone ban exerts only limited effects, at least within the context of this specific study, on such complex, multifaceted outcomes. While bans may limit specific risks (e.g., opportunities for in-school cyberbullying), they are unlikely to transform the deeper relational systems in which social wellbeing is embedded (Rubin et al., 2011). Providing meaningful offline alternatives for connection and participation alongside smartphone restrictions may therefore be helpful in avoiding unintended consequences, such as increased loneliness (Böttger and Zierer, 2024; Pouwels et al., 2024; Selwyn and Aagaard, 2020).

At the same time, schools may pursue such policies for multiple reasons, including reducing distraction or supporting mental health (Campbell et al., 2024; Wood et al., 2023), which were not evaluated in the present study. Our findings therefore speak specifically to the social wellbeing dimension of implementing a smartphone policy rather than to the overall effectiveness or desirability of school smartphone bans.

Practical implications

Although the present findings are limited to a single school and cannot establish the broader effectiveness of smartphone bans, they nonetheless point toward some tentative practical considerations.

Previous studies show that students experience smartphone restrictions in highly diverse ways; for some, bans can reduce distractions and ease social pressures, whereas for others they may increase feelings of exclusion or frustration (Campbell et al., 2024; Das and Ahmed, 2023; Gath et al., 2024; Pouwels et al., 2024). This diversity highlights that the social meaning of smartphones varies across students and contexts. School policies could therefore be more effective when designed with sensitivity to students' heterogeneous social needs and to the specific characteristics of each school.

Schools may benefit from developing context-sensitive strategies that explicitly acknowledge how smartphones function in the everyday social lives of their students. Providing structured offline alternatives for connection, such as organized group activities during breaks or accessible communal spaces, could help fill the gap left by restricted smartphone use and reduce the risk of heightened loneliness (Pouwels et al., 2024; Selwyn and Aagaard, 2020).

Moreover, rather than viewing smartphone restrictions as a standalone solution, schools may consider how these policies fit within broader frameworks of digital literacy and socio-emotional learning (Ferguson, 2025; Selwyn and Aagaard, 2020; Global Education Monitoring Report Team, 2023). Rather than imposing a complete ban, some policies allow limited or context-dependent smartphone use, for instance when it serves a clear educational purpose (Fitz et al., 2019; Gath et al., 2024), reflecting the possibility of tailoring smartphone restrictions to different contexts and purposes. Actively involving students, teachers, and parents in policy design can further enhance legitimacy, compliance, and awareness of potential unintended outcomes (Randhawa et al., 2024).

Finally, only 63.3% of students in the present study reported adhering to the ban, which may partly explain the largely null findings observed here. Recent research suggests that compliance depends less on the strictness of enforcement than on the extent to which students internalize and endorse the rationale underlying the rules (Van de Casteele et al., 2026). Need-supportive communication, combining clear and consistent guidance with attentiveness to students' perspectives, may foster greater ownership of and adherence to the policy, more so than chaotic or controlling approaches (Van de Casteele et al., 2026). More attention to how a total ban is communicated, rather than to its formal strictness alone, may therefore be critical.

Limitations and future research

Several limitations should be acknowledged. First, the quasi-experimental design relied on a single school pre-post comparison and lacked a parallel control group of students who were not exposed to a smartphone ban during the same time period, which constrains causal inference. Contextual and temporal factors (e.g., changes in classroom composition or exam periods and school holidays), may also have contributed to the observed trajectories (Allen et al., 2016; Campbell et al., 2024; Crosnoe, 2021). Future studies should therefore compare schools implementing smartphone bans with comparable schools without bans, or with different types and timings of restrictions, to better disentangle policy effects from natural developmental and contextual changes. Relatedly, because this school moved from a partial, weakly enforced, restriction to a total smartphone ban, the findings may not be generalizable to schools implementing a smartphone restrictions for the first time. This may yield different outcomes, and future studies should explicitly compare schools with differing baseline policies.

Second, there is a limited availability of theories that specify how smartphone bans are expected to influence adolescents' social wellbeing. As a policy evaluation, this study examined an intervention that was implemented in practice rather than derived from a predefined theoretical model. The present findings can therefore be viewed as providing empirical input for further theory development on the social consequences of smartphone bans in school contexts. Future work may benefit from frameworks that more explicitly account for differences in time scale, student heterogeneity, and school context helping to align policy assumptions with observed outcomes.

Third, the study was conducted in a single school, limiting generalizability. School-level factors such as climate, size, and enforcement practices likely shape how such policies play out in practice (Campbell et al., 2024). However, no such measures were included in this study. Future studies should therefore incorporate baseline measures of such factors prior to policy implementation to better disentangle the role of such contextual factors in shaping the effects of smartphone bans.

Fourth, students' awareness of both the policy implementation and the accompanying research may have affected self-reported behaviors and attitudes, potentially interacting with recent national legislation regulating smartphone use in Flemish schools.

Fifth, we focused exclusively on cyberbullying, capturing online manifestations of peer aggression. Removing smartphones during breaks, however, may also affect in-person bullying dynamics. Future research should therefore include both online and offline forms of bullying to obtain a more comprehensive picture of how smartphone bans shape peer interactions. In addition, multi-school comparative designs and investigations of different types of bans, from total bans to more flexible, partial restrictions, are needed to identify under which contextual conditions such policies yield meaningful effects.

Sixth, several constructs were assessed using brief measures. Specifically, cyberbullying victimization and perpetration were each measured with a single item, while social loneliness was assessed using two items after one item was removed because of poor internal consistency. Although this was a pragmatic choice to minimize participant burden in an extensive, repeated classroom-based survey, and one that is not uncommon in research on cyberbullying (Jia et al., 2026; Menesini and Nocentini, 2009), brief measures may reduce reliability and sensitivity to detect both short-term and gradual changes over time. Future research should consider using more comprehensive measures where feasible while balancing the practical constraints of repeated school-based data collection.

Seventh, gender was assessed using a rather brief measure, with ‘boy', ‘girl', and ‘prefer not to say' response options. Given the repeated administration of the survey during school hours and evidence that some adolescents may provide mocking or non-serious (joke) responses to gender identity questions (Teufl et al., 2025), we opted for a brief measure to reduce participant burden and potential mocking. However, this approach does not adequately capture the diversity of adolescents' gender identities. Future studies should employ more inclusive measures of gender identity and may benefit from explicitly reiterating confidentiality and privacy safeguards, as well as using careful framing, when introducing sensitive questions such as gender identity, particularly when data can be collected in settings that offer greater privacy.

Finally, this study did not include measures related to smartphone use itself, such as time spent on smartphones or types of use. Prior research suggests that restrictive school smartphone policies can reduce smartphone use during school hours but may not substantially decrease adolescents' overall use outside the school context (Goodyear et al., 2025). Including such measures both during and after school could therefore help determine whether reduced access at school is accompanied by compensatory use outside school hours. Future studies could incorporate such measures to examine whether changes in the amount and type of smartphone use following a ban are associated with changes in students' social wellbeing.

Overall, the present study evaluated a school policy and revealed minimal short-term and only modest long-term changes following a total smartphone ban. Within the context of this single-school, quasi-experimental design, these findings suggest that smartphone bans alone are not sufficient to substantially reshape students' social dynamics at school. While in-school smartphone restrictions may temporarily limit opportunities for cyberbullying, they do not appear to enhance broader indicators of social wellbeing and may even coincide with rising social loneliness at school.

These findings highlight the complexity of adolescents' social experiences and the limitations of relying on single, technology-focused policy measures. Effective strategies will likely require integrating smartphone regulations into broader frameworks that promote digital literacy, social connection, and emotional support, thereby targeting the multiple relational layers that underpin students' social wellbeing at school.