Introduction

Bullying, defined as repeated negative behavior in a relationship characterized by an imbalance of power (Olweus, 1994), is a widespread issue in school environments with substantial effects on student development. According to the Progress in International Reading Literacy Study (PIRLS), the prevalence of bullying among students has been a significant concern. According to the report, in 2011, 53% of students experienced bullying, with 33% reporting monthly incidents and 20% reporting weekly incidents. By 2016, these figures showed a slight decrease but remained significant, with 43% of fourth-grade students reporting bullying experiences, 29% occurring monthly, and 14% weekly (PIRLS, 2011; 2016). Experiencing school bullying has been linked to various adverse outcomes, including physical and mental health problems (Asimopoulos et al., 2014; Boyes et al., 2014; Vivolo-Kantor et al., 2021), diminished academic performance (Raza et al., 2020), and long-term negative effects on future development (Wolke et al., 2013; Wolke and Lereya, 2015).

Recent reports of school bullying in China have brought attention to its prevalence and consequences. Studies indicate that bullying is a significant issue in Chinese schools, with rural children more susceptible than their urban counterparts (Li et al., 2019; Zhang et al., 2019). While there is variation in the reported proportions of bullied students across different studies, the general trend points to a high prevalence of bullying in rural schools. For instance, research involving 10,528 students from three rural counties in Jiangxi province found that 73% of the students experienced bullying on a monthly basis (Wang et al., 2022a). Similarly, a study from Zhejiang province reported that 44% of students had been physically bullied at least occasionally (Hesketh et al., 2010). Furthermore, experiencing bullying at school is especially prevalent for “left-behind children,” who are more likely to lack adequate protection due to their parents’ absence (Zhang et al., 2021; Zhu et al., 2020).

While school bullying is a global phenomenon, rural settings often present distinct challenges. International research has shown that bullying in rural areas is influenced by contextual factors such as geographic isolation, limited support services, and closer-knit peer environments. For instance, studies from the United States and Spain have found that although overall prevalence may not always be higher in rural schools, students in these settings often report more severe psychological consequences (Cabrera et al., 2024; Evans et al., 2014). In Thailand, a national survey revealed that nearly half of primary school students had experienced bullying, and that rural students who were bullied were significantly less happy and engaged in school than their urban peers (Aunampai et al., 2022). Similarly, research from Poland highlights the role of social capital in rural communities, where strong neighborhood ties can either buffer or intensify bullying dynamics depending on the local social climate (Mazur et al., 2017).

These findings suggest that rural contexts may amplify the relationships between bullying experiences and student outcomes, and highlight the importance of examining not only its prevalence but also how bullying is associated with students’ social and developmental outcomes in rural settings. However, most existing international studies focus on Western or Southeast Asian countries, and few have systematically investigated rural China. Given the unique socio-cultural characteristics of Chinese rural communities, including widespread labor migration, the prevalence of left-behind children, and limited access to school-based psychological services (Zhang et al., 2021; Zhu et al., 2020), there is a pressing need to understand how bullying operates in this setting and how it is associated with students’ academic and psychological outcomes.

Several factors, including student and family characteristics, are correlated with the likelihood of experiencing bullying, and these factors also influence the consequences of being bullied. Research indicates that the prevalence of bullying varies significantly by gender and age (Hesketh et al., 2010; Wang et al., 2022a; Zhang et al., 2021). Male students are not only more likely to be victims of bullying but also tend to engage in bullying behavior more frequently than their female counterparts (Hesketh et al., 2010; Wang et al., 2022b; Zhang et al., 2019). Furthermore, studies have found a clear age difference in bullying rates, with younger students and students in lower grades experiencing more instances of bullying (Lister et al., 2015; Wang et al., 2009).

Social support has been drawing attention as a potential mediator of negative experiences such as school bullying. Studies have consistently shown that social support is instrumental in improving mental health outcomes for students, particularly in helping them coping with adverse experiences (Chang et al., 2018; Stewart and Suldo, 2011; Ulmanen et al., 2022). Effective social support, which can include emotional backing from family, teachers, and peers, as well as community-based interventions, has been found to be crucial in creating environments that foster resilience and aid recovery from mental health challenges (Li et al., 2018; Lopez-Zafra et al., 2019). This is particularly relevant in rural China, where students, including vulnerable groups such as left-behind children, may face unique mental health challenges.

While some research has focused on school bullying in rural China, there remains a scarcity of evidence regarding its consequences, particularly in the context of primary and junior high schools. Understanding the relationship between bullying experiences and student outcomes at this crucial stage is vital for both individual and national development. Firstly, since bullying is more prevalent among younger students, who are at a critical stage of physical and mental health development, identifying the correlation between experiencing bullying and student outcomes can inform interventions beneficial to their future well-being. Secondly, considering that 75% of China’s children are educated in rural areas (Rozelle and Hell, 2020), the high incidence of bullying in these settings could have significant implications for their development, making it imperative to understand these correlations for the progress of rural areas and the accumulation of rural human capital. Lastly, because students in rural China are more vulnerable than those in urban areas due to having access to fewer public services and resources, research that can guide government and policymakers in implementing measures to enhance rural human capital is necessary for narrowing the rural-urban gap.

This study aims to fill the research gap by examining the association between school bullying and student outcomes—specifically, academic performance and mental health—among rural primary and junior high school students in poor regions of China. We focus on two key questions: (1) Do students who experience bullying perform worse academically and report poorer mental health outcomes? and (2) Do social support moderate these associations? Drawing on existing literature, we hypothesize that bullying is negatively associated with both academic performance and mental health, and that higher levels of social support may buffer these negative relationships. To provide a comprehensive assessment of student development, we focus on both outcome dimensions. While mental health reflects students’ emotional well-being, academic performance—measured by standardized math test scores—captures their cognitive functioning and school success. Examining both outcomes allows for a more nuanced understanding of how bullying experiences are associated with student outcomes in school environments.

Our study makes two main contributions. First, the study reveals the prevalence of bullying in these settings and establishes a negative correlation between bullying experiences and both student academic performance and mental health. Specifically, bullied students are more likely to have lower math scores and experience symptoms of depression, anxiety, and stress, with these associations becoming more pronounced as the frequency of bullying increases. Second, the study explores the moderating effects of social support on the relationship between school bullying and student mental health. Our findings indicate that while support from friends, family, and significant others can mitigate some of the negative associations observed between bullying and students’ outcomes, this support does not fully negate the correlation with adverse mental health outcomes. This highlights the importance of proactive measures by governments and schools to prevent bullying, rather than relying solely on support after incidents occur.

The rest of the paper is organized as follows: Section 2 introduces the data collection and analysis methods used to measure bullying victimization, academic performance, and mental health. Section 3 reports the main analysis results. Sections 4 and 5 discuss the findings and present the conclusions, respectively.

Methods

Participants and procedure

The data used in this study were collected in October 2020 from 30 rural schools in Gansu province, located in northwest China. The sample selection followed two steps. First, we obtained a list of all schools from the local education bureau and randomly selected 30 schools (20 primary schools and 10 junior high schools). We then chose classes randomly from grades four, five, seven, and eight, excluding sixth and ninth grades due to their impending graduation exams and grades one to three because of the younger students’ limited literacy and numeracy skills for survey comprehension. Financial constraints led us to select no more than two classes per grade in each school; if a grade had only one or two classes, we included all classes, while grades with more than two classes had two classes chosen at random.

Second, we selected sample students. Half of the students in each sample class who were present on the day of the survey were randomly selected to participate in the survey. In total, we surveyed 1609 students in 95 sample classes across 30 sample schools. The survey had three sections: a demographic section to collect individual and household data; a section assessing bullying experiences, mental health, and perceived social support using three internationally recognized scales: the Progress in International Reading Literacy Study (PIRLS), the Depression Anxiety Stress Scales (DASS), and the Multidimensional Scale of Perceived Social Support (MSPSS); and a standardized math test to measure academic performance.

Measures

We gathered comprehensive data in five distinct categories for each participating student. This included evaluations of bullying experiences using the PIRLS, mental health assessments utilizing the Depression Anxiety Stress Scales (DASS), standardized math testing, social support assessments using the MSPSS, and a detailed personal questionnaire covering individual, family, and school characteristics.

Bullying experiences. To assess bullying experiences, we utilized the PIRLS-Bullying Scale. This instrument measures the frequency of bullying incidents over a specified period. For this study, we used the version that the survey that was translated into Mandarin Chinese, following PIRLS’ comprehensive translation guidelines, which are designed to maintain the survey’s accuracy and ensure its cultural relevance. The questionnaire prompts students to reflect on their school year and report how often they have been subjected to various forms of bullying. The survey includes eight specific items related to bullying behaviors. Responses to these items are numerically coded to compute a raw score, with “at least once a week” equating to 0, “once or twice a month” to 1, “once or twice a year” to 2, and “never” to 3. Thus, raw scores can range from 0 to 24, we then transform the raw scores using the equivalence table to obtain the transformed scale scores, where higher values indicate less frequent bullying. To interpret the scale, there are two thresholds: transformed scores below 7.9 points suggest the absence of bullying, transformed scores between 7.9 and 9.5 indicate bullying on a monthly basis, and scores exceeding 9.5 denote weekly bullying occurrences. This system allows us to ascertain the presence and regularity of bullying among the surveyed students.

Mental health. In addition, we employed the DASS, which is a well-established self-report instrument developed at the University of New South Wales to assess depression, anxiety, and stress. Participants rate each item based on a four-point Likert scale ranging from 0 (did not apply to me at all) to 3 (applied to me very much, or most of the time), resulting in a total score for each scale between 0 to 21 and a cumulative score ranging from 0 to 63. The DASS is recognized for its ability to differentiate between the three emotional states, despite their symptomatic overlap. Detailed cutoffs for each category are provided in Appendix Table A1. For our study, we utilized the Chinese version of the DASS-21, which has been validated for the Chinese population by Gong et al. (2010) and is a reliable indicator of mental health status.

Social support. Social support was assessed using the MSPSS, a 12-item scale that categorizes items into three factors corresponding to support from family, friends, and significant others, the latter being a “special person” who can vary from a partner to a teacher or counselor. Responses are measured on a seven-point Likert scale, from 1 (very strongly disagree) to 7 (very strongly agree). Subscale scores ranging from 5.01 to 7.00 are indicative of high social support, and the total score can vary from 12 to 84, with higher scores denoting a greater level of perceived support. The MSPSS has demonstrated excellent reliability, with a Cronbach’s alpha between 0.85 to 0.91. The Chinese version of MSPSS utilized in this study has been validated for its consistency and validity.

Academic performance. Student academic performance was assessed with a 30-minute standardized math test. We selected math as the subject for examination due to its stronger association with school-based learning, as opposed to other subjects that might be more influenced by home learning environments, such as reading or language skills. Test items were tailored to align with the national curriculum for each grade and developed with input from local education bureau educators. The math test used has been employed in previous surveys assessing academic performance in rural China and underwent multiple rounds of pre-testing to ensure its relevance and to establish suitable time limits for the exam (Abbey et al., 2024; Wang et al., 2021). The administration of the test in sample schools was meticulously timed and supervised by trained enumerators. Subsequent to testing, all scores were normalized based on the score distribution within each grade.

Student and household characteristics. Finally, a comprehensive questionnaire collected information on student and household characteristics. This included demographic details such as student age, gender, boarding status, hukou status (rural or urban registration), and whether the student was an only child. Parental education levels were also recorded, with a binary variable created to indicate whether a parent had completed high school or higher education. Additionally, we asked whether a student had a parent who had migrated for work for an extended period (over six months in the past year), labeling these parents as migrants. Household economic status was approximated through ownership of various assets consistent with those listed in the National Household Income and Expenditure Survey, in order to create a family asset index. These variables provide a comprehensive set of controls to isolate the correlations of bullying on mental health and academic performance among rural Chinese students.

Statistical analysis

Our analysis is structured into three distinct parts to address the objectives of the study. The first part focuses on identifying the factors that may increase the likelihood of students being bullied. To achieve this, we employ a linear probability model (LPM), which allows us to estimate the probability of a student experiencing bullying. The linear probability model we used is as follows:

$${bullie}{d}_{i}=\alpha +\beta {X}_{i}+{\varepsilon }_{i}$$
(1)

Where bulliedi is a binary variable indicator where value of 1 representing students who report being bullied at least monthly. and 0 indicates those with less frequent or no reports of bullying. The independent variables Xi encompass a range of individual and household characteristics from our sample. These include the student’s age, gender (male or female), boarding status at school (yes or no), possession of a rural hukou (yes or no), and whether the student is an only child (yes or no). Additionally, we consider the highest education level of the student’s parents, categorized into discrete levels ranging from illiterate to college and above. “Migrant father” and “migrant mother” variables indicate whether a student’s father or mother, respectively, has migrated for work for more than six months within the past year. Parental age and marital status are also controlled for in the model. Economic background is represented by the family asset index, constructed from ownership data on seven household items featured in the National Household Income and Expenditure Survey.

To examine the association between bullying experiences and student outcomes, we apply an ordinary least squares (OLS) regression model. This model assesses the correlation between bullying experiences and both mental health and academic performance. The general form of the OLS regression model is as follows:

$${Y}_{{ij}}=\alpha +\beta {{bullied}}_{{ij}}+\gamma {X}_{{ij}}+{c}_{j}+{\varepsilon }_{{ij}}$$
(2)

Where Yij represents either the mental health status or math performance of student i in class j. bulliedij is a dummy variable, we include three types of dummy variables to represent different frequencies of bullying: “being bullied” (which encompasses both monthly and weekly frequencies), “being bullied monthly”, and “being bullied weekly”. Each of these variables takes a value of 1 if the student has experienced bullying in the specified frequency, and 0 otherwise. The coefficient β is of primary interest, as it quantifies the difference in outcomes between students who have been bullied and those who have not. To control for potential confounders, Xij includes various student and household characteristics. Additionally, class fixed effects \({c}_{j}\) are incorporated to control for unobserved heterogeneity within classes, allowing us to compare students in the same class who have experienced different levels of bullying. The error term is denoted by εij. Standard errors in our regression models are clustered at the school level to account for within-school correlation. To justify this clustering approach, we estimated intraclass correlation coefficients (ICCs) for four representative outcome variables using unconditional multilevel models.Footnote 1

In the final part of our analysis, we investigate the moderating effects of social support on the relationship between bullying experiences and student outcomes. The model includes interaction terms between being bullied and social support, where social support is measured from three dimensions: family, friends, and significant others. These interaction terms help us ascertain the effectiveness of social support as a moderator. The extent to which social support moderates the negative association between bullying and student outcomes will be discussed in Section 3. The model for estimating the moderating effects is as follows:

$${Y}_{{ij}}=\alpha +{\beta }_{1}{{bullied}}_{{ij}}+{\beta }_{2}{S}_{{ij}}+{\beta }_{3}{bullie}{d}_{{ij}}\times {S}_{{ij}}+\gamma {X}_{{ij}}+{c}_{j}+{\varepsilon }_{{ij}}$$
(3)

Equation (3) builds on Eq. (2) by introducing the variable Sij, which measures perceived social support using the MSPSS scores, and an interaction term between Sij and bulliedij. To assess the moderating role of social support, we include this interaction term, with the coefficient \({\beta }_{3}\) capturing the extent to which social support buffers the negative relationship between bullying experiences and student outcomes.

To further interpret the interaction effect, we conducted simple slope analyses by estimating the marginal effect of bullying on student outcomes at three levels of perceived social support: one standard deviation below the mean (low), the mean (average), and one standard deviation above the mean (high). This approach provides a clearer understanding of how the relationships between bullying and student outcomes vary at different levels of perceived social support. As a robustness check, we additionally estimated interaction models using dichotomized support levels (high vs. low).

Results

Demographic characteristics

Table 1 presents the demographic and household characteristics of the students in our sample. The mean age of these students was approximately 11.5 years, with a gender distribution of 44.6% females and 55.4% males. A significant majority (91.5%) possessed rural Hukou and a substantial proportion of students (85.1%) did not reside at school. About 86.4% of the students had siblings. In terms of parental background, a considerable percentage of students’ fathers (57.6%) and mothers (25.4%) were migrants, resulting in 19.76% of students having at least one migrant parent. Educational attainment among parents was relatively low, with 75.9% of fathers and 85.5% of mothers not having completed high school. The average age of students’ fathers and mothers was 41 and 38 years, respectively. Additionally, 8.5% of the students had divorced parents. Regarding academic performance, the average standardized math score was 21.3 on a scale that ranged from 4 to 30 points.

Table 1 Summary Statistics of Student Individual and Family Characteristics.
Full size table

Prevalence of being bullied and mental health concerns

Table 2 illustrates students’ incidence of being bullied at school and their mental health status. To begin with, the data indicate a high vulnerability to bullying among the rural students, with 42.64% of them reporting experiences of bullying, and 12.74% of them experiencing it on a weekly basis. Moreover, mental health issues such as depression, anxiety, and stress are prevalent among the students. According to the data, 45.18% of students experienced at least one type of mental health problem. Specifically, 23.18% of the students exhibited symptoms of depression, 42.32% experienced anxiety, and 17.53% suffered from stress. Breaking these figures down further, 14.30% of students experienced moderate depression, and 8.89% had severe depression. Anxiety symptoms were moderate in 18.58% of students and severe in 23.74%. Stress symptoms were moderate in 11.00% of students and severe in 6.53%.

Table 2 The Prevalence of Bullying Experiences and Mental Health Issues.
Full size table

Linear probability model results

To understand the determinants of student victimization by bullying, we estimated a linear probability model, the results of which are presented in Table 3. The dependent variables—being bullied with any frequency, being bullied monthly, and being bullied weekly—are reported in columns (1) through (3), respectively. Our regressions control for a variety of individual and household characteristics. The findings suggest that younger students, male students, students boarding at school, and those with fathers who attained a higher level of education faced a greater risk of being bullied. Specifically, an additional year of age was associated with a 1.8 to 2.5 percentage points decrease in the likelihood of being bullied, holding other factors constant. Female students were found to have a reduced risk of being bullied by 8.3 to 14 percentage points relative to their male counterparts. The risk of being bullied increased by 7.9 percentage points for students who board at school, and this figure is 7.7 percentage points for monthly bullying incidents compared with non-boarding students. A higher educational level of the father was linked to an increased likelihood of being bullied by 5.6 and 5.7 percentage points.

Table 3 Results of Linear Probability Model.
Full size table

Association between bullying experience and student outcomes

Table 4 presents the OLS regression model results, examining the association between bullying experiences and student academic and mental health outcomes. We considered three dummy variables as independent variables indicating the occurrence of bullying: whether a student was being bullied, and if so, if they were bullied monthly or bullied weekly. These students were then compared against a base group of students who had not been bullied. Dependent variables, spanning columns (1) to (7), included the standardized math score and the presence of moderate or severe levels of depression, anxiety, and stress. Panel A of the results indicates that students who experienced bullying had lower standardized math scores and a higher risk of having moderate depression, anxiety, and stress compared to non-bullied students. Specifically, students who were bullied experienced a 0.22 SD decrease in their standardized math score, significant at the 1% level. Students who were bullied also exhibited a 29.7 percentage points increase in their risk of depression, 33.3 in their risk of anxiety, and 20.6 in their risk of stress compared to non-bullied counterparts. When examining mental health within the severe DASS classification, we observed a similar pattern, with the risk for mental health challenges increasing by 8.9 to 23.8 percentage points.

Table 4 OLS Regression of Bullying Experience and Student Outcomes.
Full size table

The data also suggests a variation in the identified associations according to the frequency of bullying. For students who experienced monthly instances of bullying, their decrease in math score was somewhat smaller at 0.18 SD, significant at the 1% level; yet their risk for having moderate mental health challenges was 14.4 to 26.6 percentage points higher compared to non-bullied students. This pattern holds for risk of having severe mental health issues, though at a smaller magnitude when bullying is experienced monthly. Students subjected to weekly bullying displayed a more significant decline in their math scores by 0.27 SD and faced elevated risks of experiencing moderate mental health issues, with an increase ranging from 35.3 to 48.7 percentage points. Additionally, the risk of having severe mental health issues surged by 20.7 to 36.8 percentage points.

Heterogeneity analysis

To examine the relationship between bullying experiences and student outcomes across subgroups, we introduced interaction terms. These terms combined bullying experiences status, and student characteristics such as age and gender. Table 5 presents the associations between bullying experiences and students’ mental health and academic performance across different subgroups. According to the results, the negative relationship between bullying experiences and mental health is more pronounced among female students, implying that female students experience a greater detriment to their mental health from bullying. Bullied female students exhibited a higher increase in the risk of depression by 11.8 percentage points and anxiety by 9.9 percentage points compared to their male counterparts. Furthermore, the relationship between bullying experiences and anxiety symptoms differed between primary and junior high school students: junior high school students showed a reduced risk of experiencing anxiety by 12.3 percentage points when compared with primary school students. This suggests that younger and female students may be more mentally vulnerable to the bullying experiences.

Table 5 The Association between Experiencing Bullying and Student Outcomes across Student Characteristics.
Full size table

However, no significant differences in the relationship between experiencing bullying and academic performance were found across other individual characteristics of the students or within other subgroups examined.

Figure 1 displays the coefficients and confidence intervals from the heterogeneity analysis, segmented by different groups. The distinctions between the male and female groups are depicted by Fig. 1a, and those between primary and junior high school students by Fig. 1b. The x-axis indicates coefficient values. Labels are added within the figure to denote the various dependent variables assessed, such as standardized math scores, depression, anxiety, and stress. The figure reveals notable differences in coefficient magnitudes between males and females, and between primary and junior high school students. Specifically, the magnitude of the associations for male students was smaller than for female students, and the magnitude for junior high students was smaller than for primary students. Because these figures intuitively display the differences in the regression findings from Table 5, they suggest that female students and primary school students may be more vulnerable and potentially face a higher risk of having adverse mental health outcomes and lower math scores when subjected to bullying.

Fig. 1: Coefficients and Confidence Intervals across Different Student Characteristics.
Fig. 1: Coefficients and Confidence Intervals across Different Student Characteristics.
Full size image

a The association between experiencing bullying and student outcomes by gender. b The association between experiencing bullying and student outcomes by school type. This figure displays the regression results by different student characteristics. Sub-figures a and b show the coefficients and confidence intervals corresponding to the coefficients in Table 5. The dependent variables, which include standardized math scores, depression, anxiety, and stress, are represented in panels A, B, C, and D, respectively. Sub-figure a illustrates the association between experiencing bullying and student outcomes for male and female students. Sub-figure b demonstrates the association between bullying and student outcomes for primary and junior high school students, highlighting the differences in the regression results by gender and school type.

Moderating role of social support

Providing support to students who are being bullied is a challenging problem. To determine whether social support can be a moderating factor, we added the interaction between social support and bullying experiences to our model. Students’ social support was measured by the MSPSS, which consisted of three dimensions: support from significant others, friends, and family.

The results, summarized in Table 6, indicate that perceived social support significantly moderates the relationship between bullying and the likelihood of experiencing moderate symptoms of depression and anxiety. The interaction terms between bullying and each dimension of social support are consistently negative and statistically significant across all models (columns 1–8), suggesting that higher levels of social support buffer the negative associations between bullying and mental health. For instance, in column (1), the interaction between bullying and support from significant others is –0.014 (p < 0.01), indicating that as perceived support increases, the marginal effect of bullying on depression decreases. Similarly, significant moderating effects are observed for support from family and friends. The total support index also shows a robust buffering effect, with the coefficient of interaction term being –0.006 for depression and –0.004 for anxiety, both significant at the 1% level. These findings are further illustrated by simple slopes analysis (see Appendix Fig. A1), which estimates the marginal effects of bullying on mental health at low, average, and high levels of social support. The results show that the correlation between bullying and the likelihood of experiencing moderate symptoms of depression or anxiety is stronger with low perceived support, while the association weakens, and in some cases becomes statistically insignificant among those with higher support levels.

Table 6 Analysis of the Moderating Role of Social Support.
Full size table

Overall, the evidence supports the role of social support as a protective factor. While it does not fully eliminate the negative correlations between bullying and student outcomes, it clearly mitigates its association with students’ mental health. Additional interaction results for academic performance and stress are available in Appendix Table A3. However, the interaction terms in those models were not statistically significant.

As a robustness check, the results for the interaction models using dichotomized version of the social support variable (high vs. low) are presented in Appendix Table A4. The results are generally consistent with those based on continuous support measures: bullying is more strongly associated with symptoms of depression and anxiety among students with low levels of support, while higher support levels are linked to attenuated, though not fully eliminated, negative associations. The direction and magnitude of the coefficients are consistent with the main results, reinforcing the moderating role of social support.

Discussion

This study enhances the literature’s understanding of bullying victimization among Chinese primary and junior high school students in less developed areas, as well as the negative associations between bullying and student academic performance and mental health. Firstly, utilizing survey data collected from a less affluent province in northwestern China, we found that bullying is prevalent among our sample, with 42.64% of students having experienced bullying and 12.74% experiencing bullying on a weekly basis. Being bullied is associated with specific student and household characteristics: younger students, male students, those who board at school, and students whose fathers have higher education levels are more likely to be bullied. Second, our findings indicate a negative association between bullying experiences and student outcomes. Students who have had bullying experiences have lower math scores and a higher risk of mental health issues compared to non-bullied students. Furthermore, this negative association appears to intensify with an increased frequency of bullying. The associations between bullying and mental health also vary across different genders and grade levels. Third, social support moderates the association between experiencing bullying and its related outcomes, with higher levels of social support associated with a reduction in the negative associations. However, even higher levels of social support do not completely negate the associations between bullying and adverse outcomes.

In this study, the percentage of students in our sample experiencing bullying is 42.64%, a figure that is close to those reported in developed countries such as the United States (45%), England (48%), Canada (49%), and Singapore (49%), all measured using similar sample scales. Notably, in these countries, the proportions exceed 45%, as indicated by PIRLS 2016 data. This suggests that the prevalence of bullying is not necessarily dependent on the development level of an area. However, students in more developed areas typically have access to more resources compared to those in rural areas, meaning that students being bullied in less developed regions may suffer more, both physically and mentally. Similar findings have been reported in other studies, highlighting that school bullying is a more significant problem in rural settings than in urban ones (Dulmus et al., 2004; Leadbeater et al., 2013).

While our findings align with global concerns about school bullying, their implications may differ depending on socioeconomic and institutional contexts. For instance, although the prevalence of bullying in our sample is comparable to that reported in developed countries such as the United States or England, its consequences in rural China may be more severe due to limited access to psychological services, a shortage of trained counselors, and a lack of comprehensive anti-bullying policies in schools. International studies, including those conducted in the U.S. (Evans et al., 2014) and Thailand (Aunampai et al., 2022), highlight the importance of institutional support systems in mitigating the impacts of bullying. In contrast, our findings suggest that even with high levels of social support from peers or family, only partially mitigate the negative associations in the rural Chinese setting, suggesting a greater vulnerability among students in under-resourced rural environments. Although our data come from Gansu Province—a relatively underdeveloped region in northwestern China—the findings have broader relevance. China is characterized by substantial regional disparities in economic development, educational resources, and public services. As one of the least developed provinces, Gansu reflects structural challenges common to many rural areas across the country, including high proportions of left-behind children, limited access to school counselors, underdeveloped mental health infrastructure, and inadequate anti-bullying mechanisms (Rozelle and Hell, 2020; Zhang et al., 2021; Zhu et al., 2020). Therefore, while our findings may not be fully generalizable to all rural settings, they provide important insights into the vulnerabilities faced by students in economically disadvantaged areas and underscore the urgent need for targeted policy interventions in these contexts.

Results from the linear probability model indicate that in our sample, male students, younger students, students who board at school, and those with highly educated fathers are more likely to be bullied. These findings are supported by several studies, which report similar factors influencing bullying experience likelihood. Male students tend to be bullied more often than females (Obregon-Cuesta et al., 2022; Smith et al., 2019), and a study in rural China also found that male students are more likely to be bullied (Li et al., 2019). Additionally, bullying victimization rates are generally higher among younger children, particularly between the ages of 6 and 11, compared to older age groups (Ding et al., 2020; Lebrun-Harris et al., 2019). Bullying tends to increase until secondary school age and then decrease as students get older (Olweus, 1994; Rigby, 1999). Boarding students also appear to be at a higher risk of being bullied, indicating a need for increased vigilance from boarding school staff (Pfeiffer and Pinquart, 2014). Interestingly, our study found that students whose fathers have a higher education level are more likely to be bullied, which might be related to the correlation between higher socioeconomic status (SES) and bullying. The relationship between SES and bullying is complex, with some studies suggesting higher rates of bullying among students from lower SES families (Chaux and Castellanos, 2015; Fu et al., 2013), while others report higher bullying rates among children from middle and high SES families (Jankauskiene et al., 2008). Overall, our findings underscore the importance for school administrators and policymakers to implement appropriate measures to reduce the risk of being bullied for vulnerable subgroups such as male students, younger students, and those who board at school.

Our OLS regression analysis indicates a significant negative correlation between bullying experiences and both academic performance and mental health among primary and junior high school students in rural China. This study particularly focuses on these demographics due to the higher prevalence and frequency of bullying incidents in rural areas compared to urban settings, and the increased vulnerability of younger students. Consistent with existing literature, our findings corroborate the notion that bullying acts as a risk factor for students’ academic success and psychological well-being (Hysing et al., 2021; Källmén and Hallgren, 2021; Kowalski and Limber, 2013; Lin et al., 2022; Riffle et al., 2021; Yen et al., 2014). Predictably, the analysis reveals that the adverse associations of bullying vary with its frequency: experiencing monthly bullying had smaller negative correlations compared to weekly occurrences, meaning that more frequent bullying was associated with more severe negative academic and mental health outcomes. Furthermore, our heterogeneity analysis shows that these negative associations are more pronounced among female students and those in primary school. This aligns with findings from prior research (Bannink et al., 2014; Rigby, 1998) and suggests that these groups may be particularly vulnerable when bullied, experiencing more severe symptoms of depression and anxiety. Developmental and psychological literature suggests that younger children and girls may have fewer emotional resources to cope with adverse social experiences. These findings highlight the importance of tailoring anti-bullying interventions in rural China to account for the intersection of gender and age. Targeted support should be provided for these particularly at-risk groups to mitigate the negative associations with bullying (Hébert et al., 2016; Scheithauer et al., 2006).

Finally, our study finds significant moderating effects of social support on the relationship between bullying experiences and mental health outcomes, specifically in depression and anxiety. We found that the level of social support can alter the magnitude of this relationship, a conclusion that aligns with findings from other research (Guo et al., 2020; K. Holt and L. Espelage, 2007). Our results also indicated that the negative association between bullying and mental health is stronger when perceived social support is low, but becomes weaker and in some cases statistically insignificant when social support is high. Similar results have also been indicated in other studies, which suggests that the moderating effect of social support varies across different levels of support (Lin et al., 2022; Rothon et al., 2011; Šmigelskas et al., 2018). One possible reason is that high levels of social support from families raise student resilience (Bowers et al., 1992). In addition, bullying risk has been found to decrease with high levels of social support from friends, thereby limiting the extent of bullying (Lynn Hawkins et al., 2001). This suggests that providing more robust social support can reduce, though not entirely eliminate, the severity of depression and anxiety symptoms in students who have been bullied. The mitigating role of high levels of social support on mental health underscores the importance of fostering supportive environments in schools and communities. However, it is crucial to recognize that while increased social support can help alleviate some of the negative associations of bullying experiences, it does not address the root problem. Therefore, preventive measures aimed at stopping bullying from occurring in the first place are essential. Governments and policymakers should focus on implementing comprehensive anti-bullying programs that not only provide support to victims but also work towards creating a culture of respect and empathy to prevent bullying incidents.

The limited moderating role of social support in our findings may be partly explained by the unique cultural and social dynamics in rural China. First, mental health issues are widespread in rural areas but often underrecognized, which reduces the likelihood that students will seek help from adults or mental health professionals. Second, the high proportion of left-behind children—due to parental labor migration—may weaken the quality and availability of emotional support within the family. Additionally, teachers in rural schools may lack the training or resources needed to identify and respond effectively to bullying. These contextual factors may intensify the negative associations with bullying and highlight the urgent need for context-specific interventions in rural China.

Strengths and limitations

Our study enriches the literature in three fundamental ways. First, our analysis of data from rural primary and junior high school students, an understudied but vulnerable group, confirms the prevalence of bullying in these communities. Bullying experiences strongly correlate with reduced academic performance and increased risk of mental health issues, with these correlations intensifying as the frequency of bullying increases. Second, our moderation analysis sheds light on the protective roles of high social support from family, friends, and other significant individuals, which can significantly alleviate the mental health challenges faced by bullied students. This finding highlights the crucial role of educators and school administrators in providing focused attention and support to vulnerable groups, such as female and younger students. Third, although support from friends, family, and significant others can mitigate some of the negative associations observed between bullying and students’ outcomes, it does not fully negate the correlation with adverse mental health outcomes. This underscores the importance of proactive measures by governments and schools to prevent bullying, rather than relying solely on support after incidents occur. Prevention strategies should be prioritized to stop the occurrence of bullying in the first place.

Despite these contributions, our study is subject to limitations stemming from its cross-sectional design, which constrains our ability to draw causal inferences about the identified risk and protective factors. Although we find robust associations between bullying victimization and students’ mental health and academic outcomes, we cannot rule out potential reverse causality—for instance, students with existing mental health issues may be more likely to become targets of bullying. Additionally, unobserved confounding factors such as teacher behaviors, peer dynamics, or school climate, may influence both bullying and outcomes. Another possibility is that higher levels of social support may not only buffer the psychological impacts of bullying but also reduce the likelihood of being bullied in the first place, thereby affecting both the exposure and the outcome. Future research using longitudinal or experimental designs is needed to disentangle these complex relationships. Nevertheless, we attempt to mitigate bias by including extensive controls for student demographics and household characteristics, as well as class fixed effects.

Conclusion

In this study, we utilized data from 1609 primary and junior high school students in rural China to conduct regression analyses exploring the relationship between bullying and student well-being. Our findings demonstrate significant negative correlations between bullying and both mental health and academic performance, with a moderating role played by social support. Moreover, our results emphasize the high prevalence of bullying among young students in China’s rural areas. There is an urgent need for targeted anti-bullying interventions and enhanced support mechanisms in rural schools to mitigate adverse mental health and academic outcomes.