Strain, Social Bonds, Deviant Peers and Aggression Among Chinese Adolescents: An Empirical Test and Extension of General Strain Theory

ABSTRACT

General Strain Theory (GST) has been widely applied to understanding deviant behaviors in Western contexts. Nevertheless, compared with the substantial body of research on the direct effect of strain on deviance, empirical evidence for its mediating and moderating mechanisms remains limited and inconsistent, particularly in non‐Western contexts. This gap constrains both the cross‐cultural generalizability of GST and its capacity to account for the full spectrum of deviant behaviors. Drawing on retrospective single‐wave self‐report data from 4497 participants in a Chinese college student sample, this study examined the interrelationships among strain, negative emotion, social bonds, deviant peers, and aggression. The findings indicate that negative emotion fully mediated the relationship between strain and aggression. Social bonds buffered the effect of strain on negative emotion, but strengthened the effect of negative emotion on aggression, with the belief dimension emerging as the key factor underlying this divergence. Deviant peers also significantly intensified the effect of strain on negative emotion. These findings underscore the importance of reducing negative emotional responses and clarifying the mechanisms through which social bonds and deviant peers shape the process by which strain is translated into aggression.

1. Introduction

Adolescent aggression has long been a focal point of scholarly inquiry within the fields of sociology, criminology, and pedagogy. Far from being a monolithic phenomenon, aggression manifests through diverse modalities, including physical violence, verbal harm, and relational exclusion (Crick and Grotpeter 1995), and is also closely associated with school bullying, delinquent behavior, mental health problems, and later antisocial risk in adulthood (Card et al. 2008; Dodge et al. 2008; Roland and Idsøe 2001). Against a backdrop of rapid social transformation, intensifying academic competition, changes in family structure, and increasingly complex patterns of peer interaction, Chinese adolescents are navigating a landscape defined by a confluence of stressors, ranging from academic strain to evolving societal expectations. Consequently, elucidating the developmental mechanisms underlying adolescent aggression holds both profound theoretical significance and substantial practical implications.

In social psychology, the definition of aggression proposed by Baron and Richardson (1994, 28) in Human Aggression remains the most widely cited. They define aggression as any behavior directed toward another organism with the intent to cause harm, where the target is motivated to avoid that harm. Furthermore, aggression is conceptually distinct from anger or hostility. Anger refers primarily to an emotional state, whereas hostility is better understood as a cognitive‐attitudinal disposition; aggression, by contrast, concerns actual harmful behavior or a readiness to engage in such behavior (Baron and Richardson 1994, 30; Ramírez and Andreu 2006). In the development of aggression measures, Buss and Perry (1992) conceptualized aggression as comprising four dimensions: physical aggression, verbal aggression, anger, and hostility. The first two correspond directly to overt behavioral expressions of aggression, whereas the latter two reflect its affective arousal and cognitive readiness. In other words, the Buss‐Perry Aggression Questionnaire does not merely assess anger or hostility in isolation; it is intended to capture a broader dispositional system of aggression that integrates behavioral, emotional, and cognitive components. The original validation study also reported associations with peer‐nominated aggression, suggesting that the scale is not simply a measure of subjective feeling states but one with clear behavioral relevance. In this sense, aggression is best understood as a general aggressive disposition or a broad propensity toward aggressive responding.

GST offers a compelling framework for situating adolescent deviance and aggression within the social sciences. Unlike traditional strain theory restricted to economic frustration, GST emphasizes a broader range of strains and the emotional processes through which they exert their effects, making it especially well suited to explaining the situational conflicts and aggression commonly observed in adolescence (Agnew 2006a). Although criminology frequently subsumes aggression under broader categories rather than treating it as a standalone legal category, a large body of research identifies it as a core behavioral mechanism in pathways to delinquency and violence. Research by Koops and de Castro (2004) highlights the conceptual and empirical overlap between aggression, violence, conduct disorder, and delinquency. This positioning suggests that aggression functions as a vital point on the deviant behavior spectrum, spanning the gap between minor infractions and serious violent crime. Criminologically speaking, aggression is less an internal trait than it is an externalized form of deviance that often serves as a precursor to more severe behavioral escalation.

GST goes beyond the claim that strain leads to delinquency by emphasizing that this relationship is not deterministic. Individuals exposed to similar strains may respond in markedly different ways, suggesting that the effects of strain are shaped by other mediating and moderating factors (Agnew et al. 2008; Thaxton and Agnew 2018). Within this framework, Social Control Theory‐one of the three dominant paradigms in criminological sociology‐posits that robust social bonds, including parental attachment, academic commitment, belief in conventional norms, and involvement in prosocial activities, serve to mitigate the likelihood of aggressive responses to strain (Agnew 1985). At the same time, research on differential association and peer influence shows that deviant peers can facilitate aggression by providing justifications, behavioral models, and opportunities for enactment, thereby making it more likely that strain will be translated into overt aggression (Akers 2017, 21; Benson and Buehler 2012). In other words, social bonds and deviant peers may function, respectively, as protective and risk‐enhancing factors embedded within the strain‐aggression pathway.

While extensive scholarship has utilized General Strain Theory, Social Control Theory, and Social Learning Theory to analyze adolescent deviance, several critical gaps remain. Primarily, most existing studies either test the direct effects of social bonds and deviant peers on delinquency (Bao et al. 2016; Jin et al. 2023) or treat them as mediators (Bao et al. 2014; Gao et al. 2016; Gao and Wong 2018), thereby addressing the question of why strain leads to aggression. Methodologically, however, moderators and mediators serve different functions. A moderator indicates that the strength, direction, or significance of the relationship between an antecedent and an outcome varies as a function of some condition, whereas a mediator explains the internal mechanism through which an antecedent affects an outcome (Fairchild and MacKinnon 2009; Hayes and Rockwood 2020). Viewed through this lens, social bonds function as a restrictive resource and a protective condition. Their role is to shape whether risk factors such as strain and anger are further externalized into aggression, rather than to account for the immediate psychological mechanism through which strain affects aggression (Agnew 1985; Hirschi 1969). This point becomes even clearer when social bonds are situated within the framework of GST. GST argues that strain gives rise to aggression or delinquency primarily because it elicits negative emotions, especially anger, which in turn increase the likelihood of aggressive or deviant coping. Negative emotion, therefore, is more appropriately conceptualized as a mediator (Agnew 1992). By contrast, constructs such as social support, social control, and deviant peers are typically viewed as conditional factors that weaken or amplify the pathway from strain to emotion to deviance. Capowich et al. (2001), for example, explicitly characterized social support as a “conditioning influence” on the strain‐delinquency relationship. Likewise, Jang and Rhodes (2012) argued that variables linked to GST, social bonds, and social learning may operate either as mediators or as moderators, but when the question concerns the conditions under which the effect of strain becomes stronger, such variables are more appropriately treated as moderators. The same logic applies to deviant peers. Rooted in Social Learning Theory, deviant peers affiliation primarily functions by providing definitions, modeling, reinforcement, and opportunity structures for aggression (Sutherland et al. 1992, 89–90). As Hoeben et al. (2016) noted, deviant peers represent a critical dimension of social influence, impacting behavior through the transmission of techniques, rationalizations, and reinforcement mechanisms. Deviant peers thus act as situational intensifiers: when individuals experience strain within a peer network that tolerates or encourages violence, that strain is more readily externalized as aggression. In the absence of such an environment, strain may not manifest as overt deviance. Deviant peers thus act as situational intensifiers: when individuals experience strain within a peer network that tolerates or encourages violence and deviance, that strain is more readily externalized as aggression. In the absence of such an environment, strain may not manifest as overt deviance. In other words, deviant peers alter both the magnitude of the effect of strain and the probability that it will be behaviorally expressed, which is precisely the logic of moderation. Taken together, social bonds, as a resource of social control, and deviant peers, as an element of the social learning environment, function primarily as either buffers against risk or amplifiers of risk. Because they shape the likelihood and intensity with which strain is translated into aggression, the present study treats them as moderators.

Furthermore, while GST has been rigorously validated within Western contexts (Agnew 2010; Patchin and Hinduja 2011; Rocheleau et al. 2025), empirical applications concerning Chinese adolescent populations often rely on traditional Western path models (Agnew 2015). There is still considerable room to deepen this line of inquiry, particularly by examining more closely the processes of emotional transformation and the mechanisms of conditional moderation within China's sociocultural context.

Accordingly, the present study focuses on Chinese adolescents to empirically examine and theoretically extend the triggers, process mechanisms, and conditional factors of adolescent aggression within the framework of GST. By incorporating two pivotal variables‐social bonds and deviant peer associations‐this research elucidates the complex dynamics underlying aggression. These findings provide empirical evidence relevant to family education, school governance, and adolescent intervention policies.

2. Literature Review

2.1. Theoretical Overview

Drawing upon interdisciplinary scholarship, particularly research on stress in medical sociology and psychology, fairness and justice in social psychology, and aggression in behavioral psychology, Agnew expanded the conceptualization of strain and transcended the limitations of traditional strain theories. In his seminal article, Agnew (1992) formally introduced GST. Unlike earlier macro‐level formulations, GST emphasizes micro‐level negative relationships, namely adverse life circumstances and detrimental social interactions. He posits that the probability of delinquency is contingent upon a multifaceted array of factors, including both the objective and subjective attributes of strain, as well as distinct individual characteristics (Agnew 2002; Agnew et al. 2002). Indeed, numerous stressors may elevate the risk of criminal or deviant behavior. Agnew delineated several defining features of criminogenic strains: they are perceived as unjust, high in magnitude, associated with weak social control, and likely to foster criminal coping. Strain may exert a direct effect on aggression because exposure to severe, chronic, or subjectively unjust strain can foster hostile cognitive appraisals, heighten irritability, diminish inhibitory control, and increase the tendency to construe social cues as threatening or provocative. Under these conditions, aggression may function as an immediate and oppositional coping response to adverse experiences. In this sense, strain should not be understood solely as a distal background risk factor for aggression; it may also operate as a proximal precipitant of aggression (Antonaccio and Botchkovar 2015).

Beyond its direct impact on crime and deviance, strain may also exert indirect effects through the mediation of negative emotions (Agnew 1992, 2006b). Each category of strain increases the likelihood that individuals will experience one or more negative emotional states, including but not limited to anger, frustration, and depression. Agnew (1992) identified anger as the most pivotal emotional response, arguing that it emerges when individuals attribute their adverse experiences to others. This attribution amplifies perceived victimization, intensifies retaliatory desires, undermines self‐control, and serves to rationalize aggression (Averill 1982, 30). The experience of negative emotions typically elicits a motivation to take corrective action. Deviant behavior may thus serve as a coping mechanism to mitigate strain by facilitating the attainment of positively valued goals, protecting or restoring positive stimuli, or terminating and escaping from negative stimuli. Furthermore, deviant behavior may be undertaken as expressions of retaliation. Research indicates that retaliatory behaviors often occur even when eliminating the underlying source of adversity is impossible (Berkowitz 1982, 283–284). Consequently, GST provides a comprehensive explanatory framework for a broad spectrum of delinquency and deviant behaviors, including theft, aggression, and drug use.

GST also seeks to identify potential moderating variables, although empirical examinations of these factors remain relatively scarce (Agnew 2013; Antonaccio and Botchkovar 2015). Importantly, GST does not assert that strain inevitably leads to crime or deviance. Instead, it posits several possible coping strategies, which Agnew (2013) classified into cognitive, behavioral, and emotional strategies. Cognitively, individuals may reinterpret strain‐inducing situations in ways that minimize their subjective adversity, such as by downplaying the importance of the adversity, emphasizing positive outcomes, or engaging in self‐attribution. Behavioral coping strategies typically fall into two types: those aimed at minimizing or eliminating the source of strain, and those directed toward satisfying retaliatory impulses. When both cognitive and behavioral strategies prove ineffective, individuals may resort to emotional coping strategies that directly target the negative emotions associated with strain. Collectively, these coping strategies indicate that strain can be managed through a range of adaptive and maladaptive responses, only some of which involve criminal or deviant behavior. Empirical evidence from stress research indicates that individuals differ in their coping preferences under specific conditions (Moon and Morash 2017; Thaxton and Agnew 2018). These findings help to explain why GST shows weaker explanatory power for certain types of criminal or deviant behavior. In most instances, individuals tend to respond to strain and negative emotions through non‐deviant means. However, under particular conditions, such as a lack of social support (Agnew and Brezina 2018, 125) or weakened social bonds (Watts 2017), individuals may be more inclined to engage in overtly criminal coping behaviors. While the direct effects of strain on deviance have been extensively supported empirically, consensus regarding mediating and moderating mechanisms remains limited.

2.2. Strain and Adolescent Aggression in the Chinese Context

Adolescent aggression represents a salient issue within the socialization process of youth. Within China's unique cultural context, the relationship between strain and aggression exhibits distinct pathways and characteristics that differ from those observed in Western individuals.

First, under the combined influence of Confucian traditional culture and intense modern social competition, academic achievement is endowed with familial significance that transcends individual value (Fu et al. 2022). Children's success embodies the value of bringing honor to the family and constitutes a primary means of fulfilling filial piety. This familial expectation elevates academic performance to a core metric of adolescents' self‐worth and familial status. Engaged in long‐term, intense competition centered on the Gaokao (National College Entrance Examination), adolescents endure pressure not only from academic performance but also from the heavy psychological burden of carrying family honor and parental expectations. These expectations form a persistent and high‐magnitude source of strain.

Second, Confucian ethics emphasize a hierarchical order encapsulated in the principle of “ruler guides subject, father guides son,” which manifests within the family as unidirectional obedience of children toward parents (Q. Wang 2016). Parents are traditionally regarded as inherent authority figures, while children's autonomy and individuality have historically been afforded limited respect (Costa et al. 2016). However, as adolescents develop psychologically and socially, they increasingly seek autonomy in personal matters, friendships, and future planning. This tension between the internal need for autonomy and the externally imposed demand for obedience generates profound psychological strain and relational conflict. Suppressed autonomy and accumulated negative emotions can readily be externalized as aggression.

Finally, within a broader context of fierce employment competition, and driven by long‐term considerations for their children's future, Chinese parents tend to adopt an authoritarian parenting style characterized by high standards, strict demands, and low responsiveness (Zhu and Chang 2019). This approach emphasizes absolute obedience and frequently employs psychological control (e.g., inducing guilt) and behavioral control to dominate children's lives (Q. Wang et al. 2007). Such parenting practices severely undermine the children's sense of self‐worth and security, leading to heightened feelings of shame, guilt, and emotional distress (Yu et al. 2021). Moreover, limited educational welfare in China means that beyond the 9‐year compulsory education system, families bear the costs of extracurricular tutoring, high school, and university education. Parents often work long hours to meet these financial demands, resulting in reduced parent‐child interaction and emotional closeness (Zhao et al. 2015).

These cumulative risk factors exert a compounding effect on adolescents' psychological adjustment. Once strain surpasses a critical threshold, individuals may resort to deviant or aggressive coping behaviors as a means of restoring internal equilibrium and psychological balance (Antonaccio and Botchkovar 2015).

Prior studies have successfully applied General Strain Theory (GST) to the study of Chinese adolescent delinquency, providing empirical support for its cross‐cultural validity. A growing body of work has shown significant associations between different forms of strain and delinquent or deviant behavior among Chinese youth. Notably, Bao et al. (2007) demonstrated how strain interacts with coping strategies to drive deviance in a sample of 615 students, while Gao and Wong (2018) found similar correlations among migrant populations. Chen and Cheung (2020) further confirmed the role of school‐level strains. By integrating a macro‐level perspective, Zhang et al. (2018) analyzed the effects of cumulative strains, significantly broadening the analytical hierarchy of GST in Chinese research.

However, the emotion‐based mediating mechanism emphasized by GST has received mixed empirical support in Chinese samples. While certain studies (e.g., W. Wang et al. 2025) find that situational negative emotions mediate the relationship between multiple strains and delinquency, others, such as Gao et al. (2016), report non‐significant mediating effects for state‐level negative affect. These discrepancies indicate that while GST is useful for understanding Chinese adolescent deviance, its underlying mechanisms still require further validation across different samples, types of strain, and measurement frameworks.

2.3. Social Bonds and Deviant Peers as Moderators

Since the latter half of the 20th century, driven by developments in systems theory and structural functionalism, the social sciences have witnessed a methodological shift from reliance on singular theoretical frameworks toward integrative multi‐theoretical approaches. Cross‐theoretical integration has emerged as a significant trend in understanding the mechanisms underlying complex social behaviors (Cullen et al. 2014, 158). Within the field of criminology, Social Control Theory (SCT), Social Learning Theory (SLT), and GST constitute the three dominant empirically supported perspectives, each demonstrating substantial explanatory power for deviant and criminal behavior. This raises an important question: does the relationship between strain and aggression vary as a function of social bonds and deviant peers?

In fact, social bonds serve as an inhibiting force in the process through which stressful events trigger aggression among adolescents. Social bonds denote individuals' relational boundaries with family, friends, and other conventional institutions. Hirschi identified four critical components of social bonds, including attachment, involvement, commitment, and belief (Hirschi 1969, 4–16). Based on the framework proposed by Hirschi, Agnew (1993, 2001) posits that social bonds can buffer the impacts of strain on aggression through three specific mechanisms. Firstly, as social bonds amplify the social costs of delinquency, adolescents with a higher level of social bonds may display lower tendencies toward violent behaviors. Secondly, social bonds can provide individuals with emotional, relational, and financial support, which may mitigate the impact of strain on negative emotions, further reducing an individual's aggressive inclination. Finally, individuals with higher social bonds tend to internalize more traditional moral and legal beliefs, possessing higher self‐control, thereby reducing the effects of negative emotions on aggression. However, empirical research showed mixed findings regarding the aforementioned theoretical arguments. For instance, Jang and Rhodes (2012) found that social bonds cannot significantly moderate the relationship between strain and violent behavior in adolescents. Kuptsevych‐Timmer et al. (2019) found that among the three types of social bonds (parental support, parental supervision, and school bonds), only parental supervision can significantly moderate the relationship between strain and criminal behavior. Conversely, Robbers (2004) found a significant moderation effect of social bonds on the relationship between strain and criminal behavior, with this moderation differing by gender.

In contrast, association with deviant peers serves as a driving force in the process by which stressful experiences trigger adolescent aggression. As adolescents gradually detach from parental supervision and protection during their developmental years, peers become the most salient and influential social group in their daily lives. Deviant peers refer to peer groups who break conventional rules, ethical principles, or laws (Sutherland et al. 1992, 7; McGloin and Thomas 2019). Furthermore, the hostile attributions, deviant norms, and negative interactional climate fostered by deviant peers may intensify the effects of strain on negative emotions such as anger and frustration, thereby heightening emotional reactivity (Herd and Kim‐Spoon 2021). In addition, once adolescents have already developed negative emotions, deviant peers may, through mechanisms of modeling, identification, and reinforcement, increase the likelihood that those emotions will be externalized in the form of aggression or other delinquent behavior (Guo 2021). Agnew (1992, 2013) therefore argued that deviant peers can moderate the relationships between strain, negative emotion, and aggression.

This suggests that social bonds, as a resource of social control, and deviant peers, as an element of the social learning environment, function primarily as either buffers against risk or amplifiers of risk, shaping both the likelihood and the intensity with which strain is translated into aggression. Yet empirical studies using integrated theoretical models have provided only limited support for the mediating roles of social bonds and deviant peers in the link between strain and aggression (Bao et al. 2014; Gao et al. 2016), while paying insufficient attention to their potential moderating roles, particularly among Chinese adolescents.

3. The Current Study and Research Hypotheses

To date, most studies on GST have been conducted primarily within Western contexts, while studies in Asian societies remain relatively limited, particularly in the Chinese context (Agnew 2015). Given that China's unique cultural environment shapes adolescents' perceptions of strain and expressions of aggression in culturally specific ways, a more in‐depth empirical and theoretical examination of the process mechanisms and conditional factors underlying adolescent aggression within the framework of GST is of considerable theoretical and practical importance.

Accordingly, the present study aims to test and extend the theoretical framework of GST by clarifying the logical relationships among the core variables of GST, SCT and SLT. To this end, a dual‐moderated mediation model was constructed to explore the internal mechanism through which strain influences adolescent aggression via negative emotions, under the joint moderating effects of social bonds and deviant peers (Figure 1). The research hypotheses are as follows:

FIGURE 1.

A dual‐moderated mediation model.

H 1

Strain has a positive effect on aggression among Chinese adolescents.

H 2

Negative emotions mediate the relationship between strain and aggression among Chinese adolescents.

H 3a

Social bonds mitigate (moderate) the relationship between strain and aggression among Chinese adolescents.

H 3b

Social bonds mitigate (moderate) the relationship between strain and negative emotions among Chinese adolescents.

H 3c

Social bonds mitigate (moderate) the relationship between negative emotions and aggression among Chinese adolescents.

H 4a

Deviant peers amplify the relationship between strain and aggression among Chinese adolescents.

H 4b

Deviant peers amplify the relationship between strain and negative emotions among Chinese adolescents.

H 4c

Deviant peers amplify the relationship between negative emotions and aggression among Chinese adolescents.

4. Methods

4.1. Participants

We conducted a cross‐sectional survey in Zhejiang Province in March 2022. Participants were recruited through purposive cluster sampling. Three vocational colleges registered by the Ministry of Education were selected. These institutions offer majors such as artificial intelligence, creative design, digital commerce, landscape technology, environmental engineering technology, electrical automation technology, engineering cost, and architectural design, exhibiting a reasonable coverage of higher vocational students. Prior to data collection, the ethics committee (Student Affairs and Administrative Offices) comprehensively reviewed the ethical and methodological aspects of the research plan based on sociological research ethical standards. To ensure ethical research, all research assistants received structured guidance about questionnaire distribution, research support, and ethical considerations. Participants were informed that their involvement in any phase of the study was voluntary and were aware that they could withdraw at any time. The online questionnaire was distributed via class‐built social media groups using a URL link. At the top of the questionnaire, a confidentiality and ethical statement was provided, illustrating research protocols and ethical considerations. It should be noted that the population examined in this study consists primarily of vocational college students in late adolescence (approximately 18–20 years old), a developmental period situated at the transition from adolescence to emerging adulthood (Arnett 2007; Sawyer et al. 2018).

In addition, the present study relies on retrospective, single‐wave self‐report data that incorporate information on temporal ordering, rather than on strictly prospective longitudinal data. By asking respondents to recall experiences, attitudes, behaviors, and other relevant information from earlier periods, this design attempts to distinguish the temporal sequence of variables at the measurement stage and thereby offers an initial basis for examining their temporal associations (Menard and Elliott 1990). Compared with cross‐sectional designs in which all variables are measured simultaneously, this approach has some advantage in terms of temporal ordering. That advantage, however, lies primarily in describing the sequence in which variables occurred and is not sufficient to support strong causal claims (Hardt and Rutter 2004). Retrospective data in this study were collected through a questionnaire survey. Specifically, the recall period for the independent variable, strain, was defined as the past 6–12 months; for the mediator, negative emotion, as the past 3–6 months; for the moderators, social bonds and deviant peers, as the past 12 months; and for the dependent variable, aggression, as the past 3 months. The purpose of this design was to avoid measuring all variables at the same time point and to provide at least a preliminary temporal basis for subsequent analyses. At the same time, it must be acknowledged that this temporal ordering depends on participants' recollections of past experiences and is therefore inevitably subject to recall bias.

4.2. Measures

4.2.1. Strain

The General Strain Scale developed by Patchin and Hinduja (2011) was adopted to measure strain. This scale encompasses various sources of strain that are prevalent or individual in the daily lives of adolescents over the past 6 months, such as “breaking up with a significant other,” “receiving a bad grade,” “getting into a disagreement with a family member,” “moving to a new school”, “being the victim of a crime.” Consisting of nine items, respondents rate each item on a Likert scale ranging from 0 (completely disagree) to 9 (completely agree). A higher score indicates that the individual perceives strain more intensely. The original scale demonstrates a Cronbach's α coefficient of 0.77, while in this study, the Cronbach's α coefficient for the scale was 0.87.

4.2.2. Aggression

Buss‐Perry Aggression Questionnaire (BPAQ), compiled by Buss and Perry (1992), was implemented to measure vocational students' aggression. Because of linguistic and cultural differences, the reliability and validity of the scale needed to be reassessed in the Chinese context. Therefore, Lv et al. (2013) selected university students from nine cities in China as the target population, conducting large‐scale sampling nationwide to localize and revise the BPAQ. They removed items with low factor loadings, such as “I am suspicious of overly friendly strangers,” “Once in a while, I can't control the urge to strike another person,” “When frustrated, I let my irritation show,” and “I am an even‐tempered person.” Eventually, they derived a Chinese version of the BPAQ for university students, comprising 22 items. Respondents use a five‐point Likert scoring method (1 for strongly disagree, 5 for strongly agree), with higher scores indicating higher levels of individual aggression. The original questionnaire demonstrated a Cronbach's α coefficient of 0.89, while in this study, the Cronbach's α coefficient for the questionnaire was 0.96.

4.2.3. Negative Emotions

Negative Emotions Scale, extracted from the General Strain Scale developed by Patchin and Hinduja (2011), was adopted for measuring negative emotions. The scale comprises five items: “I let little things irritate me,” “I feel like yelling at a parent or teacher,” “I feel like other people are always lucky and they get all of the breaks in life,” “I feel like life has given me a raw deal (has been unfair),” and “I feel like a powder keg ready to explode.” The scale employs a five‐point Likert scoring method (0 for strongly disagree, 4 for strongly agree), with higher scores indicating a heightened level of individual negative emotions. In this study, the Cronbach's α coefficient for the scale was 0.87.

4.2.4. Social Bonds

The Social Bonds Questionnaire, crafted by Longshore et al. (2004) and examined in Chinese juveniles' samples (B. Xu 2025), was employed in this study. This questionnaire encompasses four dimensions: “attachment,” “commitment,” “involvement,” and “belief,” featuring a total of 14 items. Respondents utilize a Likert four‐point scoring system (1 for not at all, 4 for completely), with higher scores indicating a stronger positive bond between the individual and society. The Cronbach's α coefficient for the questionnaire was 0.93.

4.2.5. Deviant Peers

Utilizing the measurement developed by Nelson and Boisvert (2011) and later applied in the Chinese context (Y. Zhou et al. 2024), we measured deviant peers. This tool consists of four questions, measuring an individual's perception of the number of peers engaged in deviant behaviors. These behaviors include activities like “truancy, alcohol abuse, physical altercations, and involvement in criminal activities.” Respondents employ a five‐point Likert scale (1 for having no friends engaged in such behaviors, 5 for having nearly all friends engage in these activities). A higher score denotes a heightened degree of association with deviant peers. The Cronbach's α coefficient for the questionnaire was 0.94.

4.2.6. Control Variables

To minimize potential confounding as far as possible, we also included a range of control variables: age, gender, household registration (hukou), monthly household income, academic achievement, grade level, and ethnicity. It should be noted that, although this study controlled for relevant variables as far as possible to reduce potential confounding, the number of control variables remained limited.

These variables were not entered mechanically. Rather, they were selected because each may plausibly influence emotional states, social bonds, and aggression at the same time, and thus serve as a potential source of confounding. First, gender, age, and grade level reflect individuals' developmental stage and social positioning. Prior research has shown that males and females differ in the expression of aggression, emotional responses, and social adjustment (J. Zhou et al. 2017). At the same time, as adolescents grow older and move through school, the developmental tasks they face, the academic pressures they experience, and the structure of their peer relationships all change in systematic ways. These factors may therefore jointly shape the relationship between negative emotion and aggression (Yang et al. 2014; J. Zhou et al. 2017).

Second, monthly household income per capita represents an important component of family socioeconomic status. A large body of research has shown that family socioeconomic status is closely related to adolescents' mental health, academic performance, and behavioral adjustment, and may affect the core variables in this study through differences in access to resources, family stress, and developmental opportunities. It was therefore important to account for it statistically (Agnew et al. 2008; Liu et al. 2020).

Third, academic achievement reflects students' level of adaptation within the school context and is also an important indicator of academic pressure, self‐evaluation, and peer status. Previous studies have found significant links between academic achievement, aggression, psychological adjustment, and peer relationships. Including it as a control variable therefore helps reduce the possibility that differences in school adjustment distort the focal associations under study (Yang et al. 2014; W. Zhou and McLellan 2021).

Finally, in the Chinese context, hukou status and ethnicity are not merely demographic characteristics, they also carry important structural and sociocultural significance. Hukou status, in particular, is closely tied to educational opportunity, resource allocation, social integration, and experiences of discrimination (D. Xu and Wu 2022), while ethnic background may also be associated with the cultural environment individuals inhabit and with their experiences of social identity (Liu et al. 2020). Both variables may therefore shape emotional experience and behavioral adjustment.

After eliminating the cases with missing values, we finally arrived at 4497 valid samples, with an average age of 19.23 ± 1.07 years, 2254 male students (50.12%), and 2243 female students (49.88%). Urban registered students accounted for 16.61% (747 individuals), while rural registered students accounted for 83.39% (3750 individuals). Among the participants, 2957 were first‐year students (65.75%), 1262 were second‐year students (28.06%), and 278 were third‐year students (6.18%). The ethnic composition included 4347 Han Chinese students (96.66%) and 150 students from ethnic minorities (3.34%). It should be noted that first‐ and second‐year students primarily engage in on‐campus curriculum studies, while third‐year students participate in social practices and employment preparation, resulting in fewer questionnaire responses from third‐year students than first‐ and second‐year students. Table 1 presents the descriptive statistics for all sample variables.

TABLE 1.

Descriptive analysis (N = 4497).

	M	SD or %	Min	Max
Age	19.23	1.07	16	25
Gender (1 = male)	1.50	0.50	1	2
Parental income	4.36	1.64	1	6
Academic performance	2.96	0.95	1	5
Urban residence (1 = urban)	1.17	0.37	1	2
Ethnic group (1 = han)	1.03	0.18	1	2
Strain	13.38	14.19	0	81
Negative emotions	7.89	7.88	0	40
Social bonds	41.62	7.69	14	56
Deviant peers	5.28	2.70	4	20
Aggression	46.38	16.81	22	110
Grade 1	2957	65.8	0	1
Grade 2	1262	28.1	0	1
Grade 3	278	6.2	0	1

4.3. Analytic Strategies

This study employs a stratified analytical strategy for statistical testing and results presentation. First, mediation effects were examined using Structural Equation Modeling in Mplus 8.0, allowing measurement error associated with multi‐item indicators to be controlled within a latent‐variable framework (Cheung et al. 2021). Second, moderation effects were tested using the PROCESS MACRO Package in SPSS 24.0. Within a regression‐based framework, PROCESS provides a relatively standardized and transparent way to estimate and present moderation effects, making it easier to identify the boundary conditions of variable relationships and to improve the interpretability of the results (Hayes 2017; Hayes and Rockwood 2020). The Johnson‐Neyman (J‐N) technique was further used to identify the regions of significance across the continuous range of a single moderator (Johnson and Neyman 1936). Finally, given that the proposed model involves dual moderating effects, the J‐N method is insufficient for visualizing the simultaneous impact of two moderators. Consequently, Polynomial Curved Surface Fitting (PCSF), as proposed by Y. Zhou et al. (2023), was employed to generate three‐dimensional visualizations of these dual interactions. Compared with the simple‐slopes approach, which depends on plotting effects at arbitrarily selected low, medium, and high values, PCSF captures the global trend and confidence intervals of dual moderating effects across the entire parameter space. This renders it a superior tool for elucidating the complex conditional process mechanisms inherent in this study.

5. Results

5.1. Analyzing the Mediation Effect

We employed a Structural Equation Model to analyze the mediation role of negative emotions in the relationship between strain and aggression. The standardized results of the path analysis are depicted in Figure 2. Specifically, the fit indices of this structural equation model are satisfactory: CFI = 0.92, TLI = 0.91, RMSEA = 0.05, and SRMR = 0.06, indicating a reasonable model fit (Hair et al. 2009). The estimation results reveal a significant positive effect of strain on negative emotions among Chinese adolescents (β = 0.84, p < 0.001), while negative emotions exhibit a significant positive effect on aggression (β = 0.49, p < 0.001). However, strain does not significantly predict aggression (p > 0.05), failing to support H1. Therefore, Hypotheses H3a and H4a were not supported. Regarding control variables, adolescents from urban areas tend to exhibit higher negative emotions than their rural counterparts (β = 0.02, p < 0.05). Compared with their male counterparts, female adolescents exhibited a higher level of aggression (β = 0.05, p < 0.05).

FIGURE 2.

Standardized path model of the mediation effects.

Using the Bootstrap analysis method, the path effect values and their confidence intervals in the model were estimated through 2000 resampling iterations. The results are depicted in Table 2. Regarding the total effect of strain on aggression (β = 0.41, p < 0.001), the indirect effect (β = 0.41, p < 0.001) accounted for 100% of the total effect. The Bootstrap analysis results confirmed the full mediation effect within the proposed model. Accordingly, Hypothesis 2 was supported, whereas Hypothesis 1 was not supported.

TABLE 2.

Bootstrap analysis of path effects.

Paths	β	Std. error	Z	95% CI		p
				Lower	Higher
Total effect: Strain → aggression	0.41	0.02	20.11	0.37	0.45	***
Indirect effect: Strain → negative emotions → aggression	0.41	0.04	11.59	0.34	0.48	***
Direct effect: Strain → aggression	0.00	0.04	0.04	−0.08	0.09	0.97

Note: *p < 0.05, **p < 0.01, ***p < 0.001, with 2000 times bootstrap resampling.

5.2. Analyzing the Moderation Effect

The significance of the moderation effects was tested using Model 76 of the PROCESS MACRO in SPSS 24.0. The results showed that social bonds did not significantly moderate the relationship between strain and aggression (p > 0.05), nor did deviant peers significantly moderate the relationship between strain and aggression (p > 0.05). Therefore, Hypotheses H3a and H4a were not supported.

The results reveal that social bonds significantly mitigate (moderate) the relationship between strain and negative emotions (b = −0.03, p < 0.001), with the 95% confidence interval (CI) of [−0.04, −0.01]. Besides, we found deviant peers significantly amplify (moderate) the relationship between strain and negative emotions (b = 0.02, p < 0.001), with the 95% confidence interval of [0.01, 0.03]. Hence, H3b and H4b receive support, suggesting the dual moderation effect on the first mediation path exists.

Figure 3 demonstrates the results of the PCSF analysis for the dual moderation effect of social bonds and deviant peers on the relationship between strain and negative emotions. As illustrated in the figure, the bottom flat surface represents the value of 0; the upper and lower curved surfaces represent the upper and lower bounds of the 95% confidence interval; the middle curved surface represents the effect of strain on negative emotions under the moderation of social bonds and deviant peers. The nine black dots on the middle curved surface represent the path effects at the points of low (M−1SD), medium (M), and high (M + 1SD) levels of social bonds and deviant peers, respectively. Figure 3 suggests that within the entire value range of social bonds and deviant peers, the effect of strain on negative emotions is consistently significant, as the 95% confidence intervals do not span zero. Specifically, the positive relationship between strain and negative emotions diminishes with the increase of social bonds and intensifies with the rise of deviant peers.

FIGURE 3.

The PCSF results: The effect of strain on negative emotions with the dual moderation effect of social bonds and deviant peers.

The significance test suggests social bonds significantly amplify (moderate) the relationship between negative emotions and aggression (b = 0.20, p < 0.001), with the 95% confidence interval of [0.14, 0.27]. This finding is contradictory to our H3c. Meanwhile, the moderation effect of deviant peers on the relationship between negative emotions and aggression is not significant (p > 0.05). Thus, H4c is also rejected.

To comprehensively examine the contradictory findings regarding the moderation role of social bonds on the second mediation path, we utilized ordinary least squares (OLS) regression to evaluate the moderation effects of each sub‐dimension of social bonds (attachment, involvement, participation, belief). The findings reveal that, when controlling for each social bond sub‐dimension and their interaction variables with negative emotions, only the belief dimension significantly moderates the relationship between negative emotions and aggression (b = 0.16, p < 0.001), with the 95% confidence interval of [0.11, 0.21]. In particular, the stronger the belief, the greater the predictive effect of negative emotions on aggression. We further implemented the JN approach to visualize this moderation effect, which is shown in Figure 4. It reveals that when the belief exceeds 1.6, the 95% confidence interval does not span zero, indicating a significant moderation effect. Hence, belief emerges as a pivotal factor in exerting the moderation effect within the broader social bonds construct.

FIGURE 4.

The Johnson‐Neyman (JN) results of the moderation effect of belief on the relationship between negative emotions and aggression.

Figure 5 illustrates the PCSF results of the dual moderated mediation effect. As depicted in Figure 5, when social bonds and deviant peers are positioned in the upper‐right quadrant of the “ab” intersection, the 95% confidence intervals do not encompass 0, indicating a significant mediation effect. The significance region is primarily sensitive to social bonds rather than deviant peers, with most of the significance region located in the interval when social bonds range from 2 to 4. Within the significance region, social bonds and deviant peers both amplify (moderate) the mediation effect. Besides, social bonds show a greater moderation effect than deviant peers as the curvature for the former is larger than that for the latter.

FIGURE 5.

The PCSF results of the dual moderated mediation effect.

6. Discussion and Conclusion

Over the past 3 decades, GST has gained increasing prominence in explaining adolescent deviant behavior. Within Western contexts, the direct effects of strain on deviance have been extensively validated through empirical research. However, in non‐Western societies, GST has yet to spark comparable empirical interest. From this perspective, although Western scholars have examined the causal relationships between various types of strain and specific forms of deviant behavior, the absence of evidence from non‐Western samples makes it difficult to assert the cross‐cultural universality of GST. Expanding empirical research in non‐Western contexts is therefore of substantial scholarly value for refining and advancing the theoretical scope of the framework. Furthermore, while prior studies have primarily focused on the direct effects of strain on deviance, there remains limited consensus regarding the mechanisms underlying its mediating and moderating effects (Agnew 2006a, 2006b; Agnew and Brezina 2018; Antonaccio and Botchkovar 2015). Addressing these research gaps, the present study pursues two primary objectives: (1) to evaluate the applicability and explanatory power of GST in accounting for aggression among Asian adolescents; and (2) to explore how the restraining forces proposed in SCT (i.e., social bonds) and the driving forces emphasized in SLT (i.e., deviant peers) interact with strain to influence adolescent aggression. Through this analytical approach, the study seeks to extend the theoretical boundaries and contextual relevance of General Strain Theory within the Chinese cultural milieu. The principal findings of the study are summarized as follows.

Firstly, aligned with prior research findings (Bao et al. 2004; Li et al. 2023; Z. Wang et al. 2022), the current study found that negative emotions play a mediation role in the relationship between strain and aggressive behaviors among Chinese adolescents. This finding supports the perspective of the GST, affirming that strain strengthens adolescent aggression by amplifying negative emotional experiences (Agnew 1992). We further identified a full mediation role of negative emotions, diverging from existing research. This finding underscores the potent explanatory power of the GST, specifically for Chinese adolescents. From a cultural perspective, Chinese society is characterized by collectivism and relationalism (Liu 2016), which socialize adolescents to suppress direct expressions of external conflict or hostility. When experiencing strains such as academic pressure or interpersonal tension, adolescents tend to internalize these stressors, leading to negative emotions. Once these emotions accumulate beyond a tolerable threshold, they are likely to be released through aggressive behavior as a form of emotional catharsis.

Secondly, we found that social bonds did not show a significant moderating effect on the direct pathway between strain and aggression, and their moderating effects differed across the other two pathways, which corresponds with the theoretical propositions of the GST (Agnew 2001). Within the cultural context of China, ruptured family structures, fractured interpersonal relations, and harsh parenting styles often constitute strains for vocational students. Therefore, adequate social bonds from family members (e.g., healthy parent‐child relationships, effective parental supervision, and sufficient social support) play a pivotal role in mitigating the impacts of strain on negative emotions among adolescents (Chan 2010; Juan et al. 2020). Similarly, school bonds can also alleviate adolescents' negative emotions. High levels of teacher‐student attachment, academic engagement, and participation in extracurricular activities foster positive teacher‐student relationships, peer friendships, and prosocial attitudes. When individuals confront external stressors, having more coping resources reduces the likelihood of developing adverse emotions (Agnew 2001; Robbers 2004). Therefore, for adolescents with stronger social bonds, the impact of strain on negative emotions is diminished.

Contrary to the initial hypotheses and findings of prior research (e.g., Bao et al. 2007), this study reveals that social bonds do not alleviate, but rather exacerbate, the impact of negative emotion on adolescent aggression. Specifically, the belief dimension, reflecting an individual's commitment to social rules, normative order, and moral standards, emerged as the pivotal factor driving this positive moderation. This unexpected finding may be explained by the characteristics of our sample. It may be understood in light of the specific characteristics of the sample. While strong normative beliefs typically function as a deterrent to deviance, they may heighten sensitivity to perceived injustice, normative violations, and identity devaluation when an individual is already experiencing high levels of negative emotion (Miller 2001). In other words, individuals with higher levels of belief are more likely to interpret adverse situations through a lens of “what ought to be” or “others' failure to adhere to proper conduct.” This interpretive framework tends to activate blame, indignation, and punitive orientations, thereby catalyzing the transition from negative emotion to overt aggression. This moderating process appears to be particularly salient in the present sample. Students in higher vocational colleges are at a stage of heightened social comparison and identity formation, and are therefore especially likely to experience pressure from social evaluation, identity devaluation, or a lack of recognition in everyday life. For individuals with stronger beliefs, expectations of fairness, respect, and deserved treatment are often more pronounced. Accordingly, once they are already in a negative emotional state, the tension between these expectations and the stigmatizing experiences they encounter in reality is more readily interpreted as an instance of injustice. Prior research suggests that individuals who are more sensitive to justice and social norms are more likely to respond to perceived unfairness or norm violations with anger, hostile attributions, and punitive reactions, all of which may further increase the likelihood of aggression (Jiang et al. 2022; Swit and Harty 2025).

Thirdly, the research findings indicate that deviant peers amplify the effect of strain on negative emotions. Nevertheless, its moderation effect remained non‐significant regarding both the direct pathway from strain to aggression and the link between negative emotions and aggression. In other words, the impact of strain on negative emotions is enhanced with the increase in individuals' engagement with deviant peers. Yet, the effect of negative emotions on aggression remains unaffected by deviant peers. This observation partially validates the GST's theoretical arguments regarding the moderation role of deviant peers (Agnew 2001). More importantly, this finding fills the current research gap by highlighting the detrimental impacts of deviant peers among Chinese adolescents. Deviant peer association is not uncommon among adolescents. Affected by interpersonal similarity, adolescents tend to form friendships with peers with similar life experiences and behavioral tendencies (Byrne and Nelson 1965). Peer relationships, in turn, shape individual behavior patterns through moral neutralization, behavioral imitation, and differential reinforcement. However, adolescents could be more susceptible to deviant peers because this vulnerable population has faced multidimensional life‐stress events like childhood abuse, academic failure, teacher neglect, and societal discrimination, which can lead them to deviate from mainstream prosocial social circles (Yoon 2020).

In sum, this study provides partial support for the applicability of GST to aggression among Chinese adolescents, although the moderating mechanisms proposed by the theory have yet to receive consistent and robust empirical support (Agnew 2006a, 2006b, 2013; Agnew and Brezina 2018; Antonaccio and Botchkovar 2015). The results demonstrate that negative emotion fully mediates the relationship between strain and aggression. Furthermore, social bonds exhibited divergent moderating effects across different pathways: while they buffered the direct impact of strain on aggression, they unexpectedly exacerbated the influence of negative emotion on aggression. The belief dimension emerged as the pivotal factor driving this discrepancy, representing one of the most illuminating findings of this research. Additionally, deviant peers significantly intensified the effect of strain on negative emotion. Taken together, these findings underscore not only the importance of alleviating negative emotional responses, but also the ways in which social bonds and deviant peers, as protective and risk factors respectively, can buffer or intensify the process through which sources of strain are translated into aggression. In this respect, the study offers new empirical support for the applicability of GST to the explanation of aggression among Chinese adolescents.

Nevertheless, the present study is still subject to the following limitations. First, the study utilizes retrospective single‐wave self‐report data, which relies heavily on the participants' recollections of past experiences. Over time, memory may become blurred or inaccurate, leading to potential biases in the data. Particularly, negative emotions are highly state‐dependent, and recalling experiences from “3–6 months ago” is difficult, making this one of the study's main limitations. Future research could use longitudinal tracking designs to better identify temporal sequences of variables and provide more robust causal inferences for the model. Furthermore, the Cronbach's α for the aggression scale is 0.96, which reflects measurement stability but may also suggest some redundancy or overlap in the content of certain items. Future studies could further examine and optimize the scale's structure to improve the precision and economy of the measurement tool. Finally, although the study used an overall sampling method, the sample was limited to a single educational institution. Therefore, caution should be exercised when generalizing the results to a broader group of Chinese adolescents. Future research could expand the sample scope to include middle school students, high school students, undergraduates, and a wider range of adolescents to test the applicability and robustness of the conclusions across different educational settings.

Author Contributions

Di Chen: conceptualization, methodology, data analysis, formal analysis, writing – original draft. Boyang Xu: supervision, conceptualization, methodology, validation, writing – review and editing, project administration, funding acquisition. You Zhou: supervision, methodology, visualization, writing – review and editing. Xiuyun Lin: supervision, writing – review and editing.

Funding

This work was supported by “National Social Science Fund of China” (24CFX098).

Ethics Statement

The procedures carried out in the studies involving human participants were approved by Academic Programs for School of Criminal Justice at China University of Political Science and Law research review board. These procedures were in full compliance with the 1964 Helsinki Declaration and its subsequent revisions, or equivalent ethical principles that govern the ethical treatment of human subjects in research.

Consent

All participants provided informed assent and consent, respectively.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Interested researchers may contact the author via email at 407653751@qq.com to obtain access to the data.