Cybervictimization, self-esteem, and cyberaggression endorsement in Thai university students: evidence from a two-part zero-inflated model

A comprehensive empirical investigation conducted among undergraduate and graduate populations at Chiang Mai University in January 2023 has shed new light on the complex psychological mechanics connecting online harassment to retaliatory behavior. The study, spearheaded by a multi-institutional research team and published in Frontiers in Psychology, examined how cybervictimization influences cyberaggression endorsement, and crucially, how individual self-esteem mediates or moderates this trajectory. Utilizing an advanced statistical architecture known as a two-part zero-inflated Poisson (ZIP) framework, the findings separate the act of initial cyberaggression involvement from the subsequent frequency or extent of aggressive behaviors among at-risk individuals.

With global internet penetration among young adults aged 15 to 24 hovering near 82%, digital communication platforms have become ubiquitous environments for academic, social, and personal interaction. However, this hyper-connectivity carries a heavy social cost. In heavily digitized nations such as Thailand, university students spend substantial portions of their daily routines embedded in peer-based online networks, rendering them exceptionally vulnerable to hostile digital acts, including targeted humiliation, exclusion, privacy violations, and malicious dissemination of rumors. Despite this high exposure, empirical research into adult cyberbullying has traditionally lagged behind adolescent-focused studies, leaving a significant blind spot in campus mental health frameworks and institutional prevention programming.

Methodological Architecture and Sample Characteristics

The cross-sectional study deployed an online survey methodology in January 2023, capturing responses from an analytic sample of 414 students at Chiang Mai University. Rigorous data-screening procedures—including checks for atypically rapid completion times, response straight-lining, and duplicate submissions—were enforced to protect data integrity. The resulting demographic profile revealed that 61.4% of the participants identified as female, 35% as male, and 3.6% as other, with 28.3% identifying within the LGBTQIA+ community. The median age of the cohort was 20 years old. Over half of the respondents (54.8%) reported spending more than six hours online daily.

To operationalize the core behavioral variables, researchers utilized validated Thai adaptations of established psychometric tools. Cybervictimization and cyberaggression perpetration were evaluated using specific item pools from the Cyberaggression–Perpetration and Victimization Scale, while self-esteem was measured via the Thai version of the Rosenberg Self-Esteem Scale, which yielded a median score of 30. Notably, the distribution of cyberaggression endorsement was heavily zero-inflated, with 83.1% of the sample reporting zero perpetration behaviors over the preceding month, while a smaller subset acknowledged multiple acts. This highly skewed distribution justified the implementation of the zero-inflated Poisson model over standard linear regressions, allowing scientists to distinctly analyze two different behavioral hurdles: crossing the threshold into initial involvement versus escalating the frequency of aggressive acts once involved.

Empirical Findings: Distinguishing Involvement from Escalation

The results of the ZIP modeling provided nuanced insights into the psychological pathways of cyberbullying. Model comparisons firmly established the superiority of the ZIP specification over traditional Poisson and negative binomial models, registering an Akaike Information Criterion (AIC) of 484.143 and a Bayesian Information Criterion (BIC) of 516.350.

When evaluating the count component—which measures the extent of cyberaggression endorsement among individuals already at risk—cybervictimization itself did not exert a statistically significant main effect ($b = 0.048, p = 0.386$). However, self-esteem emerged as a significant protective factor within this group ($b = -0.078, p = 0.005$). For every single-unit increase in self-esteem, the expected count of endorsed cyberaggressive behaviors decreased by approximately 7.5% among at-risk individuals. This suggests that robust self-worth functions as a vital internal regulator, dampening the escalation or repetition of aggressive retaliation once a person enters the risk pool.

Conversely, the zero-inflation component—which models the probability of belonging to a structural-zero or non-involvement subgroup—revealed a critical interactive effect. The interaction between cybervictimization and self-esteem was statistically significant ($b = -0.106, p = 0.050$), while the main effect of cybervictimization approached traditional significance thresholds ($b = -1.406, p = 0.054$). Sensitivity analyses incorporating demographic covariates such as age, gender, and internet usage confirmed the stability of these findings, with the interaction maintaining significance ($p = 0.006$).

These statistical interactions indicate that self-esteem actively conditions the association between being victimized online and crossing the threshold into initial cyberaggression involvement. Interestingly, simple-slopes probing demonstrated that victimization increased the likelihood of any cyberaggression endorsement more steeply among individuals reporting higher levels of self-esteem. Researchers propose that for certain students, higher self-esteem may coincide with a heightened sensitivity to ego or status threats, prompting defensive or retaliatory actions—such as "pushing back" or protecting social standing—when publicly humiliated or wronged online.

Implications for Campus Mental Health and Institutional Policy

The bifurcation of these findings carries substantial implications for public health strategies and university-based support systems. Historically, campus anti-bullying initiatives have treated cyberaggression as a monolithic behavior, applying broad, one-size-fits-all intervention models. The empirical evidence generated by this Chiang Mai University study demonstrates that the psychological engines driving initial involvement in cyberaggression differ fundamentally from those driving the escalation or frequency of aggressive acts.

Public health analysts and academic administrators note that prevention frameworks must pivot toward a dual-pronged approach. First, because cybervictimization strongly predicts initial behavioral involvement, institutions must prioritize early-intervention mechanisms immediately following reported victimization. Providing rapid psychological first aid, accessible and confidential reporting channels, and peer support networks can interrupt the transition from victim status to initial perpetrator status.

Second, intervention programming must address the nuanced role of self-esteem and emotional regulation. While higher self-esteem helps restrain the frequency of aggressive acts among those already engaged, it does not uniformly prevent initial retaliatory involvement under conditions of acute digital provocation. Therefore, holistic campus mental health initiatives should integrate cognitive-behavioral strategies focused on adaptive coping, anger rumination reduction, and emotional appraisal. By teaching students alternative mechanisms to process online harassment and social devaluation without resorting to digital hostility, universities can foster safer, more resilient digital communities.

Limitations and Future Research Directions

Despite its methodological rigor, the study’s authors noted several limitations that should temper broad generalizations. The cross-sectional design prevents causal inferences, meaning that while the data aligns with the hypothesis that self-esteem moderates responses to victimization, bidirectional or reciprocal trajectories—such as chronic cyberaggression eroding self-esteem over time—cannot be definitively ruled out. Furthermore, reliance on self-report instruments introduces potential social desirability biases and underreporting of sensitive behaviors.

Future inquiries are recommended to deploy longitudinal or ecological momentary assessment methodologies to establish temporal precedence and capture real-time behavioral shifts across academic semesters. Additionally, expanding research beyond a single institutional setting via multi-university, regionally diverse samples across Thailand will enhance external validity and allow for deeper cross-demographic comparisons regarding gender and sexual orientation. Through continued empirical refinement, educators and public health officials can better safeguard emerging adults navigating increasingly complex digital ecosystems.