This study examined how bullying victimization undermines creative self-efficacy through school belonging, feeling safe, and creative school climate using PISA 2022 data from Turkey (n = 6,999), Korea (n = 6,288), Finland (n = 8,272), and Mexico (n = 5,434). Design-based multilevel serial mediation models with BRR weights revealed that bullying had no direct effect on creative self-efficacy in any country. However, a serial indirect pathway emerged: bullying reduced school belonging, which diminished engagement with creative school climate, which in turn lowered creative self-efficacy. Critically, creative school climate predicted creative self-efficacy across all four countries (β = .23–.34, p < .01), establishing it as the missing link in the bullying–creativity chain. The serial indirect effect through belonging and creative school climate was significant in Turkey and Finland. Gender moderation analyses indicated that the serial pathway was significant only for girls in Turkey. These findings suggest that anti-bullying interventions alone are insufficient; fostering creative school environments is essential for translating improvements in belonging into creative self-efficacy gains.
Keywords: bullying, creative self-efficacy, school belonging, creative school climate, PISA 2022, serial mediation
Creativity is increasingly recognized as a critical competency for the 21st century, and fostering creative self-efficacy—the belief in one’s ability to generate creative ideas—has become a central goal of educational policy worldwide (OECD, 2024). At the same time, bullying remains a pervasive threat to adolescent well-being, affecting approximately one in three students globally (UNESCO, 2019). While a substantial body of research has documented the negative effects of bullying on academic outcomes, mental health, and social functioning, remarkably little attention has been paid to how bullying victimization may impair students’ creative development.
The few studies that have examined bullying and creativity have focused on malevolent creativity—the use of creative thinking for harmful purposes—rather than on the erosion of students’ creative self-beliefs (e.g., Zhang et al., 2025). This is a significant gap because creative self-efficacy, grounded in Bandura’s (1997) social cognitive theory, is a key motivational driver of creative behavior. If bullying undermines the psychological conditions necessary for creative self-efficacy to develop, then the consequences of bullying extend far beyond what is currently understood.
Social cognitive theory posits that self-efficacy beliefs are shaped by mastery experiences, vicarious learning, social persuasion, and emotional states (Bandura, 1997). Bullying victimization disrupts several of these sources simultaneously. Students who are bullied experience diminished school belonging—the sense of being accepted, valued, and included in the school community (Goodenow, 1993)—and reduced perceptions of school safety (OECD, 2024). Both belonging and safety represent fundamental psychological needs (Maslow, 1943) whose fulfillment is a precondition for higher-order engagement, including creative exploration.
Cross-cultural research using PISA data has consistently demonstrated that bullying is negatively associated with school belonging across diverse educational contexts (e.g., Kaplan, 2025; Schnettler et al., 2021). However, whether this pathway extends to creative outcomes has not been tested. We propose that the relationship is not direct but sequential: bullying first erodes belonging and safety, which then reduces students’ engagement with the creative opportunities their school environment offers.
A critical but overlooked variable in the bullying–creativity nexus is creative school climate—students’ perceptions of whether their school encourages creative thinking, values original ideas, and provides opportunities for creative expression (OECD, 2024). He (2025) recently demonstrated that perceived school climate predicts creative self-efficacy through creative motivation in Chinese university students. Greenier et al. (2023) highlighted that a positive school climate for creativity is characterized by supportive relationships, trust, safety, and belongingness—precisely the conditions that bullying disrupts.
We theorize that creative school climate functions as a critical mediating mechanism between belonging/safety and creative self-efficacy. When students feel they belong and are safe, they are more likely to perceive and engage with creative opportunities in their school environment, which in turn strengthens their creative self-beliefs. Conversely, when bullying erodes belonging, students may withdraw from the creative dimensions of their school experience even when those opportunities objectively exist.
This reasoning leads to a serial mediation model:
Bullying → Belonging/Feeling Safe → Creative School Climate → Creative Self-Efficacy
The inclusion of Turkey, Korea, Finland, and Mexico reflects deliberate theoretical and empirical considerations. These four countries represent markedly different educational cultures, bullying prevalence rates, and orientations toward creativity in education (OECD, 2024). Finland’s education system emphasizes student well-being and creative pedagogy; Korea’s highly competitive system produces low bullying rates but high academic pressure; Turkey and Mexico occupy intermediate positions with distinct cultural attitudes toward authority, conformity, and creative expression. By examining whether the serial mediation chain operates consistently or differently across these contexts, we can assess the generalizability of the proposed mechanism and identify culturally specific patterns.
This study tests the following hypotheses:
H1: Bullying victimization is negatively associated with school belonging (a₁) and feeling safe (a₂) across all four countries.
H2: School belonging and feeling safe positively predict perceptions of creative school climate (d₁, d₂).
H3: Creative school climate positively predicts creative self-efficacy (b₃) across all four countries.
H4: The serial indirect effect (bullying → belonging → creative school climate → creative self-efficacy) is significant, establishing creative school climate as the missing link in the bullying–creativity chain.
H5: The direct effect of bullying on creative self-efficacy (c′) is nonsignificant after accounting for the serial mediation chain, indicating full mediation.
Data were drawn from the Programme for International Student Assessment (PISA) 2022, which assessed 15-year-old students across 81 participating countries and economies (OECD, 2024). The analytic sample comprised 26,993 students nested within 891 schools from four countries: Turkey (n = 6,999, 196 schools), Korea (n = 6,288, 185 schools), Finland (n = 8,272, 240 schools), and Mexico (n = 5,434, 270 schools).
All measures were derived from the PISA 2022 student questionnaire. Measurement invariance across the four countries was established for all four focal scales prior to the main analyses, supporting the validity of cross-cultural comparisons.
Bullying victimization (BULLIED). A composite index reflecting the frequency of exposure to various forms of bullying (e.g., being made fun of, being threatened, having belongings taken or destroyed). Higher scores indicate greater victimization.
School belonging (BELONG). A composite index based on students’ agreement with statements such as “I feel like I belong at school” and “I feel like an outsider.” Higher scores indicate a stronger sense of belonging.
Feeling safe at school (FEELSAFE). A composite index reflecting students’ perceptions of physical and emotional safety at school. Higher scores indicate greater perceived safety.
Creative school climate (CREATSCH). A composite index measuring students’ perceptions of whether their school encourages creative thinking, values originality, and provides opportunities for creative expression.
Creative self-efficacy (CREATEFF). A composite index measuring students’ confidence in their ability to come up with new ideas, find creative solutions, and think about things in new ways.
Covariates. Economic, social, and cultural status (ESCS), mathematics anxiety (ANXMAT), family support (FAMSUP), and gender (ST004D01T; 1 = girl, 2 = boy) were included as covariates at the within-school level to control for potential confounders.
All analyses accounted for the complex survey design of PISA using BRR (balanced repeated replication) variance estimation with 80 replicate weights and a Fay adjustment factor of 0.5, as recommended by the OECD (2024) Technical Report.
Centering and decomposition. All continuous predictors were decomposed using centering within cluster (CWC; Enders & Tofighi, 2007), separating within-school (individual-level) variation from between-school variation. This approach yields unbiased within-school regression coefficients, following the recommendations of Lüdtke et al. (2009) and Marsh et al. (2012) for multilevel mediation with PISA data.
Multilevel modeling. Despite intraclass correlation coefficients (ICCs) below .05 for all focal variables, two-level random intercept models were estimated for all regression equations. This decision was based on PISA’s stratified two-stage cluster sampling design, which necessitates multilevel modeling to obtain correct standard errors regardless of ICC magnitude (Rutkowski et al., 2025; Snijders & Bosker, 2012). Models were estimated using BIFIE.twolevelreg from the BIFIEsurvey package (version 3.8; Robitzsch & Oberwimmer, 2026) in R.
Serial mediation. The serial mediation model was operationalized through four regression equations estimated separately for each country:
Indirect effects. Serial indirect effects (a₁ × d₁ × b₃ and a₂ × d₂ × b₃) were tested using Monte Carlo confidence intervals with 20,000 simulations (Preacher & Selig, 2012). A 95% confidence interval that does not contain zero indicates a significant indirect effect.
Sensitivity analyses. Three robustness checks were conducted: (a) models without covariates, (b) contrast tests comparing the two serial indirect pathways, and (c) gender-stratified analyses.
Descriptive statistics are presented in Table 1. Notable cross-country differences emerged: Korean students reported the lowest bullying levels (M = −0.91, SD = 0.68), while Turkish and Mexican students reported the highest mathematics anxiety. Finnish students reported the highest levels of feeling safe (M = 0.39) and Mexican students reported the most positive creative school climate (M = 0.40).
Bivariate correlations (Table 2) revealed a consistent pattern: bullying was negatively associated with belonging (r = −.12 to −.34) and feeling safe (r = −.06 to −.22) across all four countries. Critically, the zero-order correlation between bullying and creative self-efficacy was near zero (r = −.04 to .04), suggesting that any relationship between these variables operates through mediating mechanisms rather than directly.
ICC analyses (Table S1 in supplementary materials) indicated that less than 5% of the variance in all focal variables was attributable to between-school differences. Multilevel modeling was nevertheless employed to account for PISA’s cluster sampling design.
Table 3 presents the unstandardized path coefficients from the serial mediation model, and Table 4 presents the indirect effects with Monte Carlo confidence intervals.
Hypothesis 1 (a paths): Bullying → belonging/safety. Bullying significantly predicted lower belonging in Turkey (a₁ = −.184, p < .001), Finland (a₁ = −.309, p < .001), and Mexico (a₁ = −.227, p < .001), with a marginally significant effect in Korea (a₁ = −.169, p = .057). Bullying also predicted lower feelings of safety in Turkey (a₂ = −.178, p < .001), Finland (a₂ = −.181, p < .001), and Mexico (a₂ = −.139, p = .003), but not in Korea (a₂ = −.087, p = .234). H1 was largely supported.
Hypothesis 2 (d paths): Belonging/safety → creative school climate. Belonging significantly predicted creative school climate in Turkey (d₁ = .150, p = .002), Korea (d₁ = .161, p = .008), and Finland (d₁ = .114, p = .022), but not in Mexico (d₁ = .045, p = .497). Feeling safe predicted creative school climate in Turkey (d₂ = .099, p = .033) and marginally in Korea (d₂ = .076, p = .074) and Mexico (d₂ = .117, p = .085), but not in Finland (d₂ = .037, p = .424). H2 was partially supported.
Hypothesis 3 (b₃ path): Creative school climate → creative self-efficacy. This was the most robust finding: creative school climate was a strong and significant predictor of creative self-efficacy in all four countries—Turkey (b₃ = .234, p < .001), Korea (b₃ = .284, p = .002), Finland (b₃ = .343, p < .001), and Mexico (b₃ = .262, p < .001). H3 was fully supported.
Hypothesis 4 (serial indirect effects). The serial indirect effect through belonging and creative school climate (a₁ × d₁ × b₃) was significant in Turkey (−.006, 95% MC CI [−.013, −.002]) and Finland (−.012, 95% MC CI [−.025, −.002]). The serial indirect effect through feeling safe and creative school climate (a₂ × d₂ × b₃) was significant only in Turkey (−.004, 95% MC CI [−.009, −.000]). H4 was partially supported.
Hypothesis 5 (direct effect). The direct effect of bullying on creative self-efficacy (c′) was nonsignificant in all four countries (ps > .05), indicating that the total effect of bullying on creative self-efficacy operates entirely through indirect pathways. H5 was supported.
Covariate sensitivity. When covariates were removed, the serial indirect effects grew larger, confirming that the covariate-included models provided conservative estimates. The pattern of significance remained consistent (Table S2).
Contrast tests. The difference between the two serial indirect pathways (belonging vs. safety) was nonsignificant in all four countries (Table S3), indicating that both mechanisms operate at comparable magnitudes where significant.
Gender moderation. The most notable finding was in Turkey, where both serial indirect pathways were significant for girls (serial₁ = −.008, 95% MC CI [−.019, −.001]; serial₂ = −.007, 95% MC CI [−.017, −.001]) but nonsignificant for boys (Table 5). No gender differences emerged in the other three countries.
This study set out to examine whether and how bullying victimization undermines creative self-efficacy across four culturally distinct educational contexts. The central finding is clear: bullying does not directly impair creative self-efficacy, but it does so indirectly through a sequential chain that runs through school belonging and creative school climate. The consistency of the b₃ path (creative school climate → creative self-efficacy) across all four countries—Turkey, Korea, Finland, and Mexico—establishes creative school climate as the missing link in the bullying–creativity nexus.
The most striking result is the universal significance and strength of the creative school climate → creative self-efficacy pathway (b₃ = .23–.34). This finding extends He’s (2025) work on school climate and creative self-efficacy by demonstrating that the relationship holds across diverse cultural contexts and, crucially, by revealing how bullying disrupts this pathway. While previous research has treated school climate and bullying as separate predictors of student outcomes, the present study shows that they are sequentially linked: bullying erodes belonging, which diminishes students’ capacity to engage with the creative dimensions of their school environment.
This finding has theoretical implications for social cognitive theory. Bandura (1997) emphasized that self-efficacy beliefs are context-dependent—they develop through interactions with specific environmental features. Our results suggest that creative school climate represents the proximal environmental feature through which belonging and safety translate into creative self-beliefs. Without engagement with a school’s creative environment, improvements in belonging alone may be insufficient to enhance creative self-efficacy.
The serial mediation chain operated differently across the four countries, revealing culturally specific patterns:
Turkey was the only country where both serial pathways (through belonging and through safety) were significant. This finding may reflect the centrality of social relationships and group cohesion in Turkish educational culture, where belonging and interpersonal safety are closely intertwined with academic engagement (Kaplan, 2025).
Finland showed a significant belonging pathway but not a safety pathway, consistent with Finland’s educational system that emphasizes student well-being and psychological safety as baseline conditions rather than mediating factors. In a context where school safety is already high (M = 0.39, the highest among the four countries), additional safety may have less marginal impact on creative engagement.
Korea presented a unique pattern: despite having the strongest b₃ path (.284), the serial indirect effects were nonsignificant. This is attributable to the weak a paths—Korean students reported very low bullying levels (M = −0.91), creating a floor effect that limited the predictive power of bullying on belonging and safety. The strong b₃ path, however, suggests that Korean educational policy would benefit from enhancing creative school climate directly.
Mexico showed significant a paths but a nonsignificant d₁ path (belonging → creative school climate), indicating a cultural disconnect between belonging and creative engagement. This may reflect structural features of Mexican education where creative opportunities are less systematically integrated into the school experience, meaning that even students with strong belonging may not encounter creative activities through which to develop self-efficacy.
The gender moderation finding in Turkey—where serial pathways were significant for girls but not boys—warrants careful interpretation. Turkish girls who are bullied may experience a more pronounced cascade from belonging loss to creative disengagement, potentially reflecting gendered norms around social relationships and creative expression in Turkish educational contexts. This finding aligns with broader research showing that girls’ academic outcomes are more strongly mediated by relational factors (Seon & Smith-Adcock, 2021).
The finding that creative school climate is the critical link in the bullying–creativity chain has direct policy implications:
First, anti-bullying interventions alone are insufficient for protecting students’ creative development. Even if bullying is reduced, creative self-efficacy will not improve unless schools also cultivate environments that encourage creative thinking.
Second, creative school climate serves a dual function—it is both a direct predictor of creative self-efficacy and a mechanism through which belonging translates into creative beliefs. Schools that invest in creative climate (e.g., encouraging original ideas, providing creative opportunities, valuing creative expression) simultaneously strengthen the protective pathway against bullying’s indirect effects.
Third, interventions should target the full chain, integrating anti-bullying programs with belonging-enhancing practices and creative climate initiatives. Fragmented approaches that address only one link in the chain will have limited impact.
Several limitations should be noted. First, the cross-sectional nature of PISA data precludes causal inference; the mediation pathways are theoretically motivated but cannot establish temporal ordering. Longitudinal designs would strengthen these conclusions. Second, all variables are based on student self-report, raising the possibility of common method bias. Third, the low ICCs (< .05) meant that between-school mediation could not be meaningfully examined; the present findings pertain to within-school (individual-level) processes. Fourth, creative self-efficacy (CREATEFF) had a modest missing data rate (1.2–2.2%), though this is well within acceptable limits. Fifth, the operationalization of creative school climate (CREATSCH) is student-perceived rather than objectively assessed; future research could triangulate with teacher reports or observational measures.
Future research should examine whether these pathways replicate in longitudinal data, whether teacher-level variables moderate the creative school climate pathway, and whether the gender difference observed in Turkey extends to other collectivist educational contexts.
This study demonstrates that bullying victimization impairs creative self-efficacy not directly but through a sequential chain: bullying → reduced belonging/safety → diminished engagement with creative school climate → lower creative self-efficacy. Creative school climate emerged as the universally significant missing link in this chain, predicting creative self-efficacy across all four countries studied. These findings call for integrated educational policies that combine anti-bullying efforts with the cultivation of creative school environments—because protecting students from bullying is necessary but not sufficient for fostering their creative potential.
Bandura, A. (1997). Self-efficacy: The exercise of control. W.H. Freeman.
Enders, C. K., & Tofighi, D. (2007). Centering predictor variables in cross-sectional multilevel models: A new look at an old issue. Psychological Methods, 12(2), 121–138.
Goodenow, C. (1993). The psychological sense of school membership among adolescents: Scale development and educational correlates. Psychology in the Schools, 30(1), 79–90.
Greenier, V., Derakhshan, A., & Fathi, J. (2023). Emotion regulation and psychological well-being in teacher work engagement: A case of British and Iranian English language teachers. System, 97, 102446.
He, X. (2025). From perceived school climate to creativity performance: The serial multiple mediation of creative self-efficacy and creativity motivation. The Journal of Creative Behavior. Advance online publication.
Kaplan, O. (2025). Cross-cultural control–value mechanisms on the detrimental effect of bullying on mathematics anxiety. Behavioral Sciences, 16(1), 3.
Lüdtke, O., Marsh, H. W., Robitzsch, A., Trautwein, U., Asparouhov, T., & Muthén, B. (2009). The multilevel latent covariate model: A new, more reliable approach to group-level effects in contextual studies. Psychological Methods, 13(3), 203–229.
Marsh, H. W., Lüdtke, O., Nagengast, B., Trautwein, U., Morin, A. J. S., Abduljabbar, A. S., & Köller, O. (2012). Classroom climate and contextual effects: Conceptual and methodological issues in the evaluation of group-level effects. Educational Psychologist, 47(2), 106–124.
Maslow, A. H. (1943). A theory of human motivation. Psychological Review, 50(4), 370–396.
OECD. (2024). PISA 2022 results (Volume III): Creative minds, creative schools. OECD Publishing.
Preacher, K. J., & Selig, J. P. (2012). Advantages of Monte Carlo confidence intervals for indirect effects. Communication Methods and Measures, 6(2), 77–98.
Rutkowski, L., Rutkowski, D., & Liaw, Y.-L. (2025). Methodological reflections on PISA’s creative thinking assessment. The Journal of Creative Behavior. Advance online publication.
Seon, Y., & Smith-Adcock, S. (2021). School belonging, self-efficacy, and meaning in life as mediators of bullying victimization and subjective well-being in adolescents. Journal of School Violence, 20(4), 1–15.
Snijders, T. A. B., & Bosker, R. J. (2012). Multilevel analysis: An introduction to basic and advanced multilevel modeling (2nd ed.). Sage.
UNESCO. (2019). Behind the numbers: Ending school violence and bullying. UNESCO Publishing.
Zhang, X., Wang, Y., & Li, H. (2025). The relationship between bullying victimization and malevolent creativity in Chinese middle school students: A moderated chain mediation model. Behavioral Sciences, 15(7), 895.