Introduction

Rural Left-behind Adolescents refer to minors in rural areas who do not live with one or both parents due to their migration for work1. Adolescence is a critical developmental stage characterized by heightened emotional sensitivity, an increased need for peer acceptance, and significant changes in cognitive and social abilities2,3. These traits make adolescents more susceptible to external influences, particularly in online interactions.

Cyberbullying is defined as intentional and repeated aggressive behavior carried out by individuals or groups through the internet against vulnerable users4. Given the frequent use of the internet and social media platforms among adolescents5, cyberbullying is especially prevalent in this group. Due to their need for social validation and identity exploration, adolescents are more likely to become both perpetrators and victims of cyberbullying. Research indicates that cyberbullying not only negatively impacts adolescents’ academic engagement6 and performance7 but is also closely linked to anxiety, depression, and suicidal tendencies8.

For rural left-behind adolescents, the long-term absence of parental supervision and insufficient emotional support further heighten their vulnerability to cyberbullying, making it a significant behavioral issue in this population. Therefore, early prevention and intervention targeting cyberbullying among rural left-behind adolescents are of great practical importance.

There is a growing interest among researchers in understanding the mechanisms underlying adolescent cyberbullying. Some studies have found that low perceived social support may be linked to adolescent cyberbullying9, and lower individual self-esteem may also contribute to an increase in cyberbullying behavior10. Furthermore, research indicates that adolescent cyberbullying can decrease perceived social support11 and lead to a decrease in self-esteem12. However, there is still debate on whether the relationships between perceived social support, self-esteem, and cyberbullying in adolescents are bidirectional or unidirectional, warranting further in-depth exploration. Additionally, previous studies have primarily used cross-sectional designs and have not fully revealed the longitudinal relationships among these three factors, with limited focus on rural left-behind adolescents. Therefore, this study aims to employ a cross-lagged analysis to examine the longitudinal relationships between perceived social support, self-esteem, and cyberbullying among rural left-behind adolescents. The goal is to provide empirical support and theoretical guidance for understanding the factors associated with cyberbullying among rural left-behind children and for early prevention efforts.

The relationship between perceived social support and cyberbullying

Perceived social support refers to an individual’s cognitive appraisal of external support13. According to social control theory, social bonds—particularly emotional connections—can deter deviant behavior by fostering a sense of responsibility and behavioral regulation14. This theoretical framework suggests that perceived social support, as an indicator of emotional connection, helps individuals cope with challenging situations such as cyberbullying by enhancing resilience and mitigating negative emotions15.

Empirical research supports this perspective. For instance, Fanti et al.16 demonstrated that family support serves as a crucial protective factor against cyberbullying. Similarly, Arao et al.17 found that strong peer support can reduce the risk of adolescents engaging in cyberbullying, and other scholars emphasized the role of social relationships in regulating online behavior18,19,20,21. These findings suggest the need for longitudinal studies to explore the predictive relationship between perceived social support and cyberbullying. Thus, Hypothesis 1a is proposed:

Hypothesis 1a:

Perceived social support significantly and negatively predicts cyberbullying over time.

Furthermore, self-esteem is conceptually linked to perceived social support. The sociometer theory of self-esteem posits that self-esteem, as an affective gauge, reflects an individual’s perceived level of social acceptance and belonging29. Although prior studies have reported correlations between self-esteem and perceived social support, the causal relationship remains unclear30. For example, individuals with higher self-esteem may perceive greater social support due to their positive self-concept, but this relationship may also depend on contextual factors such as social networks31. Therefore, this study revises Hypothesis 1b is suggested:

Hypothesis 1b:

Self-esteem may indirectly influence perceived social support through the mediation of external social factors.

On the other hand, research has shown that cyberbullying can impact adolescents’ perceived social support. According to the Social Buffering Model, when individuals perceive support from others in dealing with negative life events (e.g. being a victim of cyberbullying), they can break free from adversity22. Studies have also indicated that cyberbullying significantly predicts adolescents’ perceived social support23. For example, Shen22 conducted a cross-sectional survey of 622 students from three ordinary middle schools in Jiangsu Province and found a significant correlation between being cyberbullied and perceived social support among middle school students. Hellfeldt et al.11 examined a cross-sectional sample of 1707 young adolescents and observed that cyberbully-victims are less likely to have higher levels of family support. While there is still a lack of further research results on the impact of cyberbullying on perceived social support. Based on existing literature, the Hypothesis 2b is proposed:

Hypothesis 2b:

Cyberbullying can negatively predict perceived social support over time.

The relationship between perceived social support and self-esteem

Self-esteem is defined as the self- evaluation of one’s social role, and it has the potential to influence behavioral development24. Ecological Systems Theory emphasizes the crucial influence of the social environment system on individual growth and development25. Therefore, perceived social support plays a significant role in self-esteem. Research has also shown that social support is a strong predictor of self-esteem26. Warmth from family, companionship from peers, and encouragement from teachers can predict stable and positive self-esteem during adolescence12. Qiao et al.27 conducted a three-year longitudinal study on children aged 6–18 in rural China and found that good social support promotes high levels of individual self-esteem. In a cross-sectional survey of 612 Chinese secondary school students, Liu et al.28 found that perceived social support directly influenced students’ self-esteem. Therefore, this study proposes the following hypothesis:

Hypothesis 2a:

Perceived social support can positively predict self-esteem over time.

Furthermore, self-esteem has been found to affect perceived social support. According to the sociometer theory of self-esteem, humans have an innate need for belonging and a motivation to maintain important interpersonal relationships, with self-esteem serving as an emotional indicator of one’s level of integration into these relationships29. Some studies suggest that adolescents’ self-esteem may have a positive impact on perceived social support30. Individuals with low self-esteem lack belief in their self-worth, are overly sensitive to interpersonal relationships and society, struggle to perceive respect and support from others, and have lower levels of understanding social support31. Conversely, individuals with high self-esteem tend to be more confident and self-assured, and they are inclined to view themselves and others in a positive light, making it easier for them to perceive support and recognition from society32. Besides, Liu proposed revealed a significant positive correlation between collective self-esteem and social support among 1,000 university students in China33. Therefore, this study proposes the following hypothesis:

Hypothesis 2b:

Self-esteem can predict perceived social support over time.

The relationship between self-esteem and cyberbullying

Research has shown a close association between adolescents’ self-esteem and cyberbullying, with low self-esteem being linked to an increased risk of cyberbullying34. Individuals with low self-esteem often hold negative self-perceptions, leading to insecurity and a tendency to seek validation through maladaptive means. Specifically, the anonymity and lack of immediate consequences in online interactions provide low-self-esteem individuals with opportunities to express aggression or retaliate against perceived social exclusion35. Unlike traditional bullying—where face-to-face interactions may deter such behavior due to social norms and the risk of immediate retaliation—cyberbullying emphasizes broad social comparison and online popularity, which are particularly salient for adolescents in the stage of identity exploration and peer validation36.

Adolescents with low self-esteem may perceive cyberbullying as a way to enhance their social status or alleviate feelings of inadequacy. For instance, Ding et al.33 observed that Chinese university students with lower self-esteem were more likely to engage in cyberbullying, further illustrating the influence of online social dynamics on such behaviors. Conversely, individuals with high self-esteem typically exhibit stronger self-regulation, enabling them to resist negative influences in online interactions and respond to peer pressure with greater confidence and positivity37. Burns’ research38,39 similarly found a significant negative correlation between self-esteem and cyberbullying among European middle school students.

Self-esteem influences cyberbullying through unique pathways, particularly during adolescence—a developmental stage marked by heightened sensitivity to peer dynamics and social comparison. Therefore, Hypothesis 3a is proposed: Self-esteem negatively predicts cyberbullying over time, and due to the distinct nature of online interactions, its influence operates through significantly different mechanisms compared to traditional bullying.

On the other hand, some scholars argue that adolescent cyberbullying can lead to a decrease in self-esteem35. Social comparison theory suggests, individuals typically evaluate their own worth and status by comparing themselves to others36. According to the theory, when experiencing online bullying, individuals may perceive themselves as occupying a lower social position or facing social exclusion, leading to a negative impact on their self-esteem9. Previous research has shown a relationship between cyberbullying victimization and self-esteem, suggesting that individuals who cannot cope with cyberbullying may experience low self-esteem37. Patchin and Hinduja38 surveyed a random sample of 1963 middle school students in the United States and found that students who experienced cyberbullying, both as victims and offenders, had significantly lower self-esteem than those with little or no experience with cyberbullying. Cénat et al.39 conducted a cross-sectional survey of 8,194 teenagers in Quebec and found that cyberbullying victimization contributes to the prediction of low self-esteem. Therefore, this study proposes Hypothesis 3b: Cyberbullying can predict self-esteem over time.

Furthermore, considering the relationships between perceived social support and self-esteem, as well as self-esteem and cyberbullying, it can be inferred that self-esteem may play a crucial connecting role in the relationship between perceived social support and cyberbullying. Due to the relatively stable nature of this role, it may have a longitudinal impact. Therefore, this study integrates the above hypotheses and proposes Hypothesis 4: Self-esteem serves as a longitudinal mediator between perceived social support and cyberbullying.

Research methodology

Participants

This study employed a cluster sampling method to conduct two waves of longitudinal surveys with a three-month interval among left-behind adolescents in grades 7–9 from two rural junior high schools in Jiangsu and Sichuan provinces. The selected schools are located in regions with high rates of parental labor migration, ensuring geographical and demographic representativeness that reflects the general characteristics of rural left-behind children in China.

The study’s inclusion criteria required participants to meet three key conditions: (1) having one or both parents working away from home for at least three months, (2) residing in rural areas while attending the selected schools, and (3) possessing the capability to independently complete the questionnaire survey. Conversely, exclusion criteria eliminated students who either failed to meet the operational definition of left-behind children or were unavailable for follow-up surveys due to school transfers or extended absences. This dual screening approach ensured the sample’s adherence to the study’s specific demographic focus while maintaining methodological rigor in participant selection.

To ensure data completeness and accuracy, the questionnaire surveys were administered face-to-face by trained graduate and undergraduate research assistants with backgrounds in educational psychology. Upon completion, administrators immediately reviewed each questionnaire to verify that all items were answered. If omissions or unclear responses were detected, administrators clarified or supplemented the answers on-site through direct communication with participants. Prior to formal data analysis, incomplete or logically inconsistent questionnaires (e.g., contradictory responses to similar items) were excluded.

The sample size (N = 981) was determined using G*Power 3.1, based on a small-to-medium effect size (f² = 0.02) for cross-lagged modeling, a statistical power of 0.80, and a significance level of 0.05, which indicated a minimum requirement of 88 participants per group. To account for potential attrition and subgroup analysis needs, a larger sample was recruited, with only a 5.49% attrition rate from T1 to T2. Chi-square tests and independent samples t-tests confirmed no significant differences between retained and attrited participants in terms of gender, grade level, perceived social support, self-esteem, or cyberbullying, ensuring the representativeness and reliability of the data.

The first survey was conducted in December 2023 (T1), with 1038 valid participants (545 males, 493 females). The second survey was carried out in March 2024 (T2), yielding 981 valid participants. A total of 57 participants did not take part in the second survey due to reasons such as transferring schools or being on leave, resulting in a participant attrition rate of 5.49%. Chi-square and independent sample t-tests indicated no significant differences between the valid and lost-to-follow-up samples in terms of gender (χ2 = 0.734, P > 0.05), grade (χ2 = 0.966, P > 0.05), perceived social support (t = 0.276, P > 0.05), self-esteem (t = 0.438, P > 0.05), and cyberbullying (t = 0.532, P > 0.05), suggesting there was no structured attrition. Therefore, a final sample of 981 participants (527 males, 454 females; 318 in Grade 7, 380 in Grade 8, 283 in Grade 9), who remained left-behind during both T1 and T2, was selected for data analysis, with an average age of 14.04 ± 0.788 years. This study was approved by the Academic Ethics Committee of Leshan Normal University prior to implementation, with approval number LSN20231105. Inclusion criteria were as follows: participants met the definition of “left-behind children,” meaning that one or both parents were working away from home and unable to live with them for three months or longer; they attended one of the selected schools and were able to complete the questionnaires. Exclusion criteria included: those who did not meet the definition of left-behind children or were unable to participate in follow-up surveys due to reasons such as transferring schools or taking extended leave.

Research instruments

Perceived social support scale

Perceived social support was measured using the Multidimensional Scale of Perceived Social Support, developed by Zimet40. The scale comprises 12 items, divided into three dimensions: family support (e.g., “My parents can provide me with practical help.”), friend support (e.g., “I can rely on my friends when facing difficulties.”), and other support (e.g., “I can share joys and sorrows with certain individuals.”). A 7-point Likert scale was utilized, with responses ranging from 1 to 7 indicating “Strongly disagree” to “Strongly agree”, where higher scores indicate greater social support. The scale has demonstrated reliability and validity among Chinese adolescents41. The Cronbach’s α coefficients for perceived social support in this study were 0.930 and 0.950 for the two measurement occasions.

Self-esteem scale

The Self-Esteem Scale, developed by Rosenberg24, was employed to assess adolescents’ self-esteem. This scale consists of 10 items (e.g., “I feel that I have a lot of good qualities.”) rated on a 4-point Likert scale, with scores ranging from 1 to 4, representing “strongly disagree” to “strongly agree”, and higher scores indicating higher self-esteem. The scale showed good reliability and validity in previous studies42. The Cronbach’s α coefficients for self-esteem in this study were 0.932 and 0.940 for the two measurement occasions.

Cyberbullying scale

Adolescents’ cyberbullying was measured using the revised Chinese version of the Cyberbullying Scale by You43 This scale comprises the Cyberbullying Victimization Scale (12 items) and the Cyberbullying Perpetration Scale (8 items). The study utilized the Cyberbullying Victimization subscale for measurement. The subscale includes three dimensions: cyber verbal bullying (e.g. “Someone has said mean things or spread rumors about me online), anonymous identity (e.g. “Someone has anonymously harassed me online.”), and cyber impersonation Bullying (“Someone has sent me infected files or images through email.”). Responses are rated on a 4-point scale from 1 (Never happened) to 4 (Frequently happened), with higher scores indicating a higher frequency of cyberbullying. The reliability and validity of scale has been vilified among adolescents28. The Cronbach’s α coefficients for cyberbullying in this study were 0.955 and 0.961 for the two measurement occasions.

Research procedure

The research procedure consisted of two stages. Firstly, contact with schools before the first assessment to obtain consent from schools, teachers, and participants, with students obtaining written parental consent. Secondly, conduct group testing in each class. Each class was administrated by two education psychology graduate or undergraduate students who had undergone rigorous training as test administrators. During the testing, they provided uniform instructions and explanations of the questionnaire content. All completed questionnaires were collected on-site by the test administrators.

Data analysis

Prior to employing the Full Information Maximum Likelihood (FIML) method21 for missing data handling, we conducted a Missing Completely at Random (MCAR) test for all variables following Little and Rubin’s (2002) guidelines. The MCAR test, performed using SPSS 25.0, yielded χ² = 58.72 with df = 50 (χ²/df = 1.17, p = 0.187). The χ²/df ratio below the conventional threshold of 3, coupled with a non-significant p-value, confirmed that the data met the MCAR assumption. These results justified the application of FIML, which provides unbiased parameter estimates under MCAR conditions. All subsequent analyses, including the cross-lagged panel models examining perceived social support, self-esteem, and cyberbullying, were conducted using AMOS 24.0 with FIML estimation.

Results

Common method Bias

All data in this study were collected through self-report measures, necessitating an examination of common method bias. First, Harman’s single-factor test44 was conducted on all items measuring perceived social support, self-esteem, and cyberbullying at both T1 and T2. The results revealed 10 factors with eigenvalues greater than 1, with the first factor accounting for only 28.745% of the variance - below the critical threshold of 40%.

To further validate the absence of common method bias, a confirmatory factor analysis (CFA) was performed to test a two-factor model. In this model, all items were simultaneously loaded onto two latent factors: one representing the substantive measurement content and the other representing method bias. The results demonstrated good model fit (χ²/df = 2.87, CFI = 0.93, TLI = 0.91, RMSEA = 0.05, SRMR = 0.04). Importantly, the method factor contributed only 21.3% of the variance, which is below the commonly accepted threshold of 30%. These findings collectively suggest that significant common method bias was not present in this study.

Descriptive statistics

A test of homogeneity of variance was conducted for gender and grade across two time points. The results indicate that there were significant differences in perceived social support and self-esteem across different grade levels (P < 0.05). Additionally, cyberbullying did not show significant differences across different grade levels, and these differences did not change over time (P > 0.05). Furthermore, there were no significant differences in perceived social support, self-esteem, and cyberbullying based on gender, and these differences did not change over time (P > 0.05). The specific details are presented in Table 1.

Table 1 Analysis of differences in demographic variables across different periods.
Full size table

Correlation analysis

The correlation analysis of the three variables—perceived social support, self-esteem, and cyberbullying (as shown in Table 2) indicate that, whether viewed from a simultaneous or delayed perspective, all pairs of the three variables show significant correlations. Specifically, both the perceived social support exhibits significant positive correlations with self-esteem (P < 0.01), as well as significant negative correlations with cyberbullying (P < 0.01) in T1 and T2. Furthermore, self-esteem in T1 and T2 also demonstrate significant negative correlations with cyberbullying (P < 0.01), indicating the stability of these variables over time. Therefore cross-lag analysis can be further conducted.

The Pearson correlation coefficients were used to examine the correlation between the variables, as shown in Table 2. Perceived social support and self-esteem from both time points are significantly positively correlated, with correlation coefficients ranging from 0.215 to 0.432. Additionally, the results from both time points show significant negative correlations (P < 0.01) between perceived social support, self-esteem, and cyberbullying, with correlation coefficients ranging from − 0.420 to − 0.197.

Table 2 Correlation coefficient matrix between variables.
Full size table

Cross-lagged analysis of perceived social support, self-esteem, and cyberbullying

Considering the significant correlation between grade and perceived social support, between grade and cyberbullying, grade was included as a control variable in the model. Four models (M1, M2, M3, M4) were constructed using longitudinal data to examine cross-lagged relationships. M1 is the baseline model containing only the autoregressive paths for perceived social support, self-esteem, and cyberbullying; M2 adds the predictive paths from perceived social support and self-esteem to cyberbullying to M1; M3 adds the predictive paths from cyberbullying to perceived social support and self-esteem to M1; and M4 adds the cross-lagged model for perceived social support, self-esteem, and cyberbullying to M1. Comparison of the models (see Table 3) revealed that while the fit of M3 was within an acceptable range, M4 yielded a better fit. Therefore, the cross-lagged model (M4) was used to describe the longitudinal relationships between perceived social support, self-esteem, and cyberbullying.

Table 3 Comparison of model fit Results.
Full size table

Using structural equation modeling, a cross-lagged analysis was conducted to examine the relationships between perceived social support, self-esteem, and cyberbullying among left-behind rural adolescents. The cross-lagged analysis results (as shown in Fig. 1) reveal that the autoregressive paths of perceived social support, self-esteem, and cyberbullying from T1 to T2 are all significant (β = 0.184, P = 0.002; β = 0.208, P < 0.001; β = 0.170, P = 0.003). The cross-paths indicate that perceived social support at T1 significantly positively predicts self-esteem (β = 0.124, P = 0.030) and negatively predicts cyberbullying (β=–0.199, P < 0.001) at T2. Self-esteem at T1 does not significantly predict perceived social support positively (β = 0.100, P = 0.085) and negatively predicts cyberbullying (β=–0.048, P = 0.405) at T2. Furthermore, cyberbullying at T1 does not negatively predicts perceived social support (β=–0.096, P = 0.100), but negatively predicts self-esteem (β=–0.165, P = 0.003) at T2.

Fig. 1
figure 1

The cross-lagged model for the relationships between perceived social support, self-esteem, and cyberbullying. Note: The path coefficients in the figure are standardized path coefficients; * represents P<0.05, ** represents P<0.01, *** represents P<0.001; dashed lines indicate insignificant, solid lines indicate significant; demographic control variables are not included in the figure.

Full size image

Longitudinal mediation analysis

Based on the first-order longitudinal mediation analysis proposed by Fang et al.45, a cross-time analysis was conducted to examine the mediating effect of self-esteem on the relationship between perceived social support and cyberbullying among rural left-behind adolescents. While this two-wave design with a three-month interval provides valuable insights into the longitudinal associations, it does not meet the criteria for Granger causality, which requires multiple time points to establish robust temporal precedence. Therefore, the results should be interpreted as indicative of longitudinal relationships rather than definitive causal inferences. The Bootstrap method was utilized with a sample size of 5000, and significance testing was conducted within a 95% confidence interval. According to the results of Model M5, the path coefficient of perceived social support at T1 to cyberbullying at T2, denoted as c’ =–0.199, with a 95% confidence interval of [–0.379, − 0.074]; the path coefficient of perceived social support at T1 to self-esteem at T2, denoted as a = 0.124, with a 95% confidence interval of [0.003, 0.213]; the path coefficient of self-esteem at T1 to cyberbullying at T2, denoted as b=–0.048, with a 95% confidence interval of [–0.256, 0.103]; and the path coefficient of self-esteem at T1 to perceived social support at T2, denoted as b = 0.100, with a 95% confidence interval of [–0.062, 0.328]. The longitudinal mediating effect was calculated as \(a \times b = − 0.006\), with a 95% confidence interval of [–0.025, 0.002]. Since the confidence interval includes zero, the mediating effect is not statistically significant. Consequently, the direct effect (\(c’ = − 0.199\), 95% CI: [–0.379, − 0.074]) remains significant, indicating no partial mediation in this model. Furthermore, as c’ is significant, this mediating effect is considered a partial mediating effect.

Discussion

Relationship between perceived social support and cyberbullying

This study revealed that perceived social support negatively predicts cyberbullying among rural left-behind adolescents, supporting Hypothesis 1a. This finding aligns with Social Control Theory, which posits that stronger social ties can reduce deviant behaviors by fostering emotional connections and discouraging harmful actions15. For left-behind adolescents, the absence of parental companionship weakens their social bonds, leaving them more vulnerable to engaging in cyberbullying as a means of coping with frustration or asserting control.

Compared to non-left-behind adolescents, rural left-behind adolescents experience prolonged parental absence, which may lead to feelings of social disconnection and neglect. This highlights the unique role of peer and teacher support in compensating for the lack of family support. Consistent with Yang et al., peer support has been shown to effectively reduce cyberbullying risks1, suggesting that interventions targeting peer relationships may be particularly effective for this group.

Additionally, cyberbullying at T1 did not negatively predict perceived social support at T2, not supporting hypothesis 1b. This result is not consistent with previous research17. One possible reason for it could be the presence of other sources of social support that offset the negative impact of cyberbullying. Specifically, this could be explained by the theory of social support, which suggests that individuals can receive support from various sources such as family, friends, and community, and the presence of these alternative sources of support may buffer the negative effects of cyberbullying on perceived social support.

Relationship between perceived social support and self-esteem

The study also found that perceived social support positively predicts self-esteem, supporting Hypothesis 2a. This finding is consistent with Ecological Systems Theory, which emphasizes the influence of social environments, including family, peers, and schools, on individual psychological development25. For left-behind adolescents, family support is often limited, making peer and teacher support critical for fostering self-esteem.

Furthermore, the rural context of left-behind children exacerbates their sense of isolation, highlighting the importance of external support systems. As demonstrated by Qiao et al.,27 consistent peer and teacher encouragement can help left-behind adolescents develop positive self-perceptions, even in the absence of parental involvement. These results suggest that schools in rural areas should prioritize strengthening peer support networks and teacher-student relationships to enhance the self-esteem of left-behind adolescents.

Furthermore, this study did not find support for hypothesis 2b, as adolescents’ self-esteem did not significantly predict perceived social support. There could be two reasons for this result. First, perceived social support may have a certain degree of stability, limiting the short-term impact of self-esteem. Perceived social support involves cognitive and experiential perceptions of the social support one receives48,47,48, and previous research has shown that these perceptions exhibit strong stability over time49. As a result, even though perceived social support is still developing, its relative stability may reduce the likelihood of being significantly influenced by self-esteem in the short term.

Second, factors influencing perceived social support (e.g., social support networks) may primarily stem from external environments rather than psychological traits such as self-esteem. However, prior studies suggest that self-esteem may have long-term effects on perceived social support: individuals with high self-esteem are more likely to focus on their strengths and values, enhancing self-affirmation and making them more inclined to recognize and perceive social support in their interactions with others50. The discrepancy between our findings and previous studies may be due to the shorter time span of this study, which might not fully capture the potential long-term effects of self-esteem on perceived social support. Additionally, cultural background or sample characteristics (e.g., the unique context of rural left-behind adolescents) may serve as moderating factors in the relationship between perceived social support and self-esteem.

Relationship between self-esteem and cyberbullying

This study found that the self-esteem of rural left-behind adolescents did not significantly negatively predict their cyberbullying over time, thus hypothesis 3a was not supported. This result is inconsistent with previous research findings39,51. The reasons for this inconsistency may be that rural left-behind adolescents may mitigate the negative impact of cyberbullying on their self-esteem through positive psychological adaptation mechanisms, thereby weakening the influence of self-esteem on cyberbullying. Another contributing factor is that cultural backgrounds shape individuals’ perceptions of self-esteem and coping strategies regarding cyberbullying, thereby influencing the relationship between self-esteem and cyberbullying. For instance, in collectivist cultures (e.g., China), individuals’ self-worth is more strongly tied to interpersonal relationships and social approval, whereas in individualist cultures (e.g., Western countries), self-worth tends to derive from independence and personal achievement. These cultural differences may lead adolescents in collectivist cultures to prioritize group harmony and conformity when facing cyberbullying, adopting more passive or avoidant coping strategies. In contrast, adolescents from individualist cultures may be more inclined to confront bullies directly or seek external support.

The current study focuses on rural left-behind adolescents, who are embedded in a particularly pronounced collectivist cultural context. This may further shape the self-esteem-cyberbullying relationship in two key ways: First, these adolescents may rely more heavily on peer relationships and teacher support to build self-esteem; when such social support is lacking, their ability to cope with cyberbullying may be significantly compromised52,53. Second, collectivist cultural norms may predispose them to endure bullying silently to maintain group harmony54,55.

Another reason lies in the fact that individuals’ perceptions of self-esteem and their coping strategies for cyberbullying differ across cultural contexts, which further influences the relationship between self-esteem and cyberbullying. For instance, in collectivist cultures (e.g., China), self-worth is more closely tied to interpersonal relationships and social recognition56, whereas in individualistic cultures (e.g., Western countries), individuals tend to derive their sense of self-worth from independence and personal achievements57. These cultural differences may lead adolescents in collectivist cultures to prioritize group harmony and social approval when facing cyberbullying, adopting more compliant or avoidant coping strategies. In contrast, adolescents in individualistic cultures may be more inclined to confront the issue directly or seek external support58. This study focuses on rural left-behind adolescents, whose collectivist cultural traits are particularly pronounced, potentially further shaping the relationship between self-esteem and cyberbullying. On one hand, rural left-behind adolescents may rely more heavily on peer relationships and teacher support to build self-esteem, and the absence of such social support could significantly impair their ability to cope with cyberbullying. On the other hand, collectivist norms may predispose them to endure bullying in silence to avoid disrupting group relationships59,60.

The findings of this study indicate a positive correlation between cyberbullying and their self-esteem among left-behind rural adolescents, supporting hypothesis 3b. This result aligns with previous research37,50, which suggests that individuals experiencing more cyberbullying tend to have lower self-esteem. This may stem from the comparison with others and feelings of social exclusion or disadvantage in the online environment, leading to a decrease in self-esteem9. Consequently, cyberbullying could emerge as a potential factor influencing self-esteem, particularly among left-behind rural adolescents.

Relationship between perceived social support, self-esteem and cyberbullying

The study explored the longitudinal mediation effects among perceived social support, self-esteem, and cyberbullying. The results indicate that self-esteem does not significantly mediate the relationship between perceived social support and cyberbullying among rural adolescents, as the confidence interval of the indirect effect (\(a \times b\)) includes zero. However, perceived social support directly influences cyberbullying (\(c’ = − 0.199\), 95% CI: [–0.379, − 0.074]). This highlights the importance of focusing on perceived social support as a direct intervention target to reduce cyberbullying behaviors. Future studies should include additional time points to better capture the mechanisms underlying these relationships. According to social control theory, individuals are inherently restrained from deviant behavior by social bonds15. Applied to cyberbullying, the perceived social support an individual receives directly influences their engagement in cyberbullying behavior. Moreover, social cognitive theory explains how individuals develop, maintain, and modify their behavior within their environment. With increased perceived social support, adolescents are more likely to develop higher self-esteem, thus negatively impacting their online behavior. Therefore, perceived social support among rural adolescents not only directly influences cyberbullying but also acts through self-esteem.

Limitations and future research directions

There are some limitations in the study. Firstly, all questionnaires in this study were conducted in a self-assessment format. In the future, different evaluation subjects, such as parents and teachers, could be included to enhance the objectivity and effectiveness of the measurements. Secondly, the interval between the two assessments in this study was three months, and only two time points were included. This design, while sufficient for preliminary longitudinal mediation analysis, may not fully capture the dynamic and reciprocal relationships among perceived social support, self-esteem, and cyberbullying. Additionally, Granger causality requires multiple time points to establish robust temporal precedence and predictive power, which is beyond the scope of this study. Future research should consider incorporating additional waves of data collection with longer intervals to better explore the temporal dynamics and causal mechanisms underlying these relationships. Lastly, this study focused on rural adolescent groups, and it did not reveal the relationships among perceived social support, self-esteem, and online bullying in normal adolescent populations. Future research could supplement this aspect.

Implications

Implications arising from this study are significant for both theoretical advancements and practical interventions in addressing cyberbullying among rural adolescents. From a theoretical perspective. This study offers a novel perspective on understanding the longitudinal relationships among perceived social support, self-esteem, and cyberbullying among left-behind rural adolescents. It enriches our comprehension of the intricate mechanisms underlying these psychological factors in rural adolescent online behavior and lays a theoretical foundation for exploring effective strategies to prevent cyberbullying.

From a practical perspective, the findings hold practical significance for intervention and prevention efforts. Addressing cyberbullying among left-behind rural adolescents should prioritize enhancing their perceived social support and self-esteem to mitigate cyberbullying incidents. Specifically, in terms of perceived social support, parents, despite their busy work schedules, should make efforts to regularly communicate with their children, listen to their needs and feelings, and provide support and understanding. Providing psychological support, teachers should pay attention to the mental well-being of left-behind rural adolescents, promptly identify and assist them in dealing with issues, and offer necessary psychological support and guidance. Peers should support and encourage each other in daily life, fostering a sense of mutual support. Regarding self-esteem, parents and teachers should timely provide students with positive feedback and affirmation, acknowledging their efforts and achievements in both academic and non-academic pursuits. Positive encouragement can boost students’ confidence and self-esteem. Parents and teachers should guide students in cultivating positive attitudes and thinking habits, encouraging them to remain optimistic and resilient in the face of challenges and difficulties. A positive attitude contributes to strengthening students’ confidence and self-esteem.

Conclusion

This study employed a two-wave longitudinal design and cross-lagged model to analyze the relationships among perceived social support, self-esteem, and cyberbullying in rural left-behind adolescents. The findings revealed that: (1) Perceived social support demonstrated significant temporal stability, exerting a positive influence on self-esteem and a significant negative predictive effect on cyberbullying behavior; (2) Self-esteem served as a partial mediator between perceived social support and cyberbullying, indicating that perceived social support not only directly reduced cyberbullying but also indirectly inhibited it by enhancing self-esteem; (3) While cyberbullying did not show a significant negative predictive effect on perceived social support, it significantly negatively influenced self-esteem, suggesting that cyberbullying may have a lasting adverse impact on individuals’ psychological well-being.

These findings provide valuable longitudinal evidence for understanding the mechanisms and psychological dynamics underlying cyberbullying behaviors among rural left-behind adolescents.