Explaining Persistent Offending Across the Life Course

Although most young offenders “age out” of crime as they enter adulthood, some individuals maintain high levels of offending throughout much of the life course. These chronic, high-rate offenders typically exhibit highly aggressive behavior as young children, engage in high levels of delinquency during adolescence, and persist in serious offending as they grow older (Moffitt, 1993). It is important to explain this pattern of offending, as such offenders commit a disproportionate share of serious crime.

To explain persistent offending, some criminologists highlight the role of stable personality traits, such as low intelligence, impulsivity, or hyperactivity (Moffitt, 1993). Such traits are said to interfere with the development of strong attachments to conventional others and other stakes in conformity. These traits are linked to failure in school, unstable work histories, association with criminal and delinquent peers, and ultimately a pattern of persistent offending. GST offers a similar explanation of life-course-persistent offending but focuses special attention on the trait of “aggressiveness” (Agnew, 1997). Aggressive individuals can be described as having a difficult temperament—they are irritable and have a low tolerance for frustration. The trait of aggressiveness may result, in part, from chronic strains experienced in childhood, such as harsh or erratic parental discipline.

According the GST, the trait of aggressiveness helps to fuel persistent antisocial behavior for at least three reasons. First, aggressive individuals have a propensity to interpret any given situation as frustrating and to blame others for their frustration. As a result, compared to non-aggressive individuals, aggressive individuals are more likely to respond to various situations with anger and delinquent coping. Second, given their difficult temperament, aggressive individuals often provoke negative reactions from others. For example, aggressive children often frustrate their parents and are at risk of emotional and physical abuse, especially when raised by unskilled parents. Aggressive children may also frustrate their peers and teachers, leading to social rejection.

Third, aggressive individuals tend to sort themselves into environments characterized by high levels of strain. As a result of their difficult temperament, aggressive individuals have difficulty maintaining stable relationships and employment. They may end up in poor quality marriages and jobs that prove to be sources of chronic strain, and which further promote persistent high-rate offending (Agnew, 1997).

Explaining Gender Differences in Crime

Males are much more likely than females to engage in crime and delinquency, with the gender difference in offending being greatest for serious offenses. A variety of explanations have been offered to account for this gender gap in offending. Some criminologists argue that, relative to males, females have less freedom or opportunity to offend. For example, females are subject to higher levels of parental supervision, have higher levels of commitment to family and school, and are less likely to associate with delinquent peers. These factors are said to constrain females, limiting their ability to engage in crime. Other theorists emphasize the role of socialization, arguing that males are more likely to engage in crime because they have internalized “masculine” values that are conducive to crime and violence, such as competition and aggressiveness (Agnew & Brezina, 2015).

General strain theorists recognize these factors, but they offer two additional reasons for the gender gap in crime and delinquency (Broidy & Agnew, 1997). First, they argue that the gender gap in crime is related, in part, to the different types of strain that are experienced by males and females. The evidence in this area suggests that, on average, males and females experience the same overall level of strain. But males are more likely to experience those particular types of strain that are strongly related to crime and delinquency, such as harsh parental discipline, negative school experiences, criminal victimization, and homelessness.

Second, certain data suggest that the response to strain is gendered, with males being more likely to cope with strain in a criminal or delinquent manner. Although the reason for this gendered response is not yet clear, several possibilities exist. It is possible, for example, that males are more likely to engage in delinquent coping because they have a greater tendency to associate with delinquent peers, or because they have lower levels of conventional social support. Another possibility is that males are more likely to react to strain with emotions that are conducive to offending, such as moral outrage. It has been suggested that, in response to strain, females are more likely to blame themselves or worry about possible harm to interpersonal relationships. Females still get angry but may also have a high propensity to experience the emotions of depression and anxiety simultaneously. So females may react to strain with a complex combination of emotions that, together, are less conducive to offending.

Explaining Community Differences in Crime

GST is primarily a social psychological theory, focusing on the relationship between the individual and his or her immediate social environment. Yet Agnew (1999) argues that processes related to social psychological strain can be used to explain patterns of crime appearing at the level of schools, neighborhoods, and larger communities.

Why, for example, do some communities have especially high rates of crime and violence? Variation in crime across macro-level social units is typically explained in terms of deviant subcultures or breakdowns in social control. According to subcultural accounts, the characteristics of high crime communities (especially economic disadvantage) foster the development of subcultural orientations, including attitudes and values that are conducive to crime (e.g., Anderson, 1999). According to social disorganization theories, these communities have the lost the ability to control their members due, in part, to the inadequate supervision of young people (e.g., Sampson, Raudenbush, & Earls, 1997).

In addition to subcultural orientations and breakdowns in social control, GST asserts that high-crime communities tend to suffer from a relatively high proportion and angry and frustrated residents (Agnew, 1999). This anger and frustration is thought to be an important source of deviant motivation and is said to be a function of severe and persistent strains. For example, young people in poor inner-city communities experience high levels of family disruption, abuse and neglect, exposure to community violence, school problems, persistent poverty, unemployment, under-employment, and struggle to achieve goals related to money and status (Brezina & Agnew, 2013). They also face much class and racial/ethnic discrimination, including negative experiences with the police and other representatives of the larger society (see Bernard, 1990).

Young people in inner-city communities may also have a tendency to cope with strain in a delinquent manner, given both a lack of coping resources and limited options for legal coping. In particular, they often lack coping resources that are available to those in wealthier communities, such as money, power, and conventional social support. They generally have less control over their lives, having difficulty removing themselves from adverse environments, and have fewer opportunities for legal coping, given poor schools and limited job opportunities. At the same time, inner-city youth often encounter numerous opportunities for criminal coping, such as gang membership and drug selling (Brezina & Agnew, 2013).

Furthermore, the high density of strained individuals in such communities is said to generate much interpersonal friction. It increases the likelihood that residents will interact with others who angry, upset, and potentially hostile. It also increases the likelihood that angry and frustrated individuals will encounter each other, contributing to elevated rates of crime and violence (Agnew, 1999).

Evidence Linking Strain to Offending

Initial tests of GST produced promising results, showing a relationship between various strains and delinquent behavior. Following the initial statement of GST (Agnew, 1992), Agnew and White (1992) examined the effects of various strains on delinquency, based on a large sample of adolescents. In longitudinal analyses, a summary measure of strain predicted future delinquency, even after controlling for measures of social control, delinquent peer associations, and prior delinquent behavior. The summary measure of strain indexed such factors as stressful life events, life hassles, and fights with parents. In the longitudinal analyses, however, strain did not predict drug use (although the effect was significant in cross-sectional analyses).

In another early test of GST, Paternoster and Mazerolle (1994) used data from the National Youth Survey to examine the effect of strain on a measure of general delinquency. The results are generally supportive of GST. In longitudinal analyses that controlled for levels of social control, delinquent peers, and prior behavior, they find that delinquency is predicted by negative life events, negative relations with adults, school/peer hassles, and neighborhood problems. A measure of traditional strain, which indexed the respondent’s perceived chances of going to college and getting a good job, failed to exert a significant effect on future delinquency.

Numerous additional tests of GST have produced similar results, indicating a relationship between various strains and offending behavior (for an overview, see Agnew 2006). Further, certain data indicate that adolescents may resort to delinquency because it allows them to alleviate the negative emotional consequences of strain, at least in the short run (Brezina, 1996, 2000; Novacek, Raskin, & Hogan, 1991). Over the long run, however, delinquent responses to strain are likely to exacerbate problems with parents, teachers, and conventional peers.

Most empirical tests of GST have been conducted in the United States and are based on data from adolescent surveys. Several tests, however, indicate that the central propositions of GST apply to youth in other parts of the world (e.g., Bao, Haas, & Pi, 2007; Moon, Morash, McCluskey, & Hwang, 2009; Sigfusdottir, Kristjansson, & Agnew, 2012; but see Botchkovar, Tittle, & Antonaccio, 2009) and to adult populations (e.g., Jang & Johnson, 2003; Morris, Carriaga, Diamond, Piquero, & Piquero, 2012; Ostrowsky & Messner, 2005; Swatt, Gibson, & Piquero, 2007). Also, while early tests of GST typically examined the relationship between strain and general delinquency, the theory has since been applied to a variety of specific deviant behaviors. Evidence has accrued, for example, linking the experience of strain to aggressive behaviors in school, workplace violence, prison inmate misconduct, substance abuse, suicidal ideation, self-harm, and eating disorders (Brezina, Piquero, & Mazerolle, 2001; Hay & Meldrum, 2010; Hinduja, 2007; Morris et al., 2012; Piquero, Fox, Piquero, Capowich, & Mazerolle, 2010; Sharp, Terling-Watt, Atkins, Gilliam, & Sanders, 2001; Swatt et al., 2007).

Despite these positive results, most tests of GST have employed rather simple measures of strain. Few studies have assessed the qualities of particular strains as they are experienced by the individual. Although certain strains may be universally stressful or frustrating (e.g., hunger, homelessness, physical pain), most strains have a subjective component. Whether or not a given event or circumstance is subjectively defined as adverse is dependent on the meaning the individual attaches to it (Polizzi, 2011). A particular adverse event may cause intense distress for some individuals but not others, depending on their beliefs, values, life situations, and the techniques at their disposal for minimizing the emotional or cognitive significance of the event (see Leban, Cardwell, Copes, & Brezina, 2016).

As stated earlier, the strains that are said to be most relevant to crime and delinquency tend to be those seen as unjust and high in magnitude (severe, frequent, of a chronic nature, and of central importance to the individual). Certain data indicate that the subjective experience of strain is an important consideration in understanding the relationship between strain and offending, but studies in this area have produced mixed results. In comparison to objective measures of strain, Froggio and Agnew (2007) find that subjective measures of strain are more strongly related to offending. Highlighting the subjective evaluation of fairness, Rebellon and colleagues (2012) observe that the perceived injustice of social relations is a potent predictor of delinquency (see also Scheuerman, 2013). Jang and Song (2015) observe that subjective strain fully mediates the impact of objective strain on the delinquency of middle-school students. Other research, however, does not find the distinction between objective and subjective measures of strain to be consequential (Lin & Mieczkowski, 2011). Clearly, additional research on this issue is needed.

Evidence on Intervening Processes

As predicted by GST, a number of studies indicate that the relationship between strain and offending is partly mediated by anger, and this is especially true of studies that focus on violent behavior (e.g., Aseltine, Gore, & Gordon, 2000; Agnew, 1985; Brezina, 1998; Broidy, 2001; Hay & Evans, 2006; Jang & Johnson, 2003; Mazerolle & Piquero, 1997, 1998; Mazerolle, Piquero, & Capowich, 2003; Moon et al., 2009). As described earlier, Agnew (2006) identifies other possible links between strain and offending. In addition to the generation of anger, strain is likely to have other consequences of a criminogenic nature. Strain may lead to other negative emotions, foster beliefs favorable to crime, reduce social control, and increase attraction to delinquent peers. These alternative links, however, have received less attention.

There is some evidence that negative emotions other than anger may help to account for the relationship between strain and offending, at least for certain deviant outcomes (e.g., Bao, Haas, & Pi, 2007; Ganem, 2010; Hay & Meldrum, 2010; Jang & Johnson, 2003; Kaufman, 2009; Piquero et al., 2010). Based on a national sample of African American adults, Jang and Johnson (2003) find that strain-induced anger best predicts aggression, while strain-induced depression is more strongly associated with substance use. A study by Ganem (2010) highlights the potential complexity of the linkage between strain and offending, indicating that different types of strain may produce different types of negative emotions. Further, emotions such as anger appear to promote criminal tendencies, while emotions such as anxiety and fear appear to inhibit these tendencies (see also Aseltine et al., 2000; Piquero & Sealock, 2004).

An additional level of complexity has been introduced by studies that distinguish between situation-based emotions and trait-based emotions. Theoretically, strain should generate negative emotions that arise in direct response to adverse events or situations. At the same time, certain strains—especially chronic strains experienced in early childhood—may promote the trait of aggressiveness, leading individuals to possess an angry and irritable temperament that transcends particular situations (Agnew, 1997). Limited evidence suggests that situation-based and trait-based emotions may operate differently, with situation-based emotions playing a larger role in mediating the relationship between recent strains and offending (Mazerolle, Piquero, & Capowich, 2003; Moon et al., 2009). In contrast, trait-based emotions—especially “angry disposition” or “negative emotionality”—may play a stronger role in moderating the relationship between strain and offending; that is, individuals who possess these traits seem more likely than others to respond to strain with antisocial behavior (see Agnew et al., 2002; Eitle, 2010). Consequently, tests of GST that fail to distinguish between situation- and trait-based emotions could be problematic.

Relatively few studies have examined intervening processes that involve factors other than emotions. Based on a national sample of adolescents, Paternoster and Mazerolle (1994) find that the effect of strain on delinquency is partly mediated by social control and association with delinquent peers. In particular, strain appears to reduce social control and increase involvement with delinquent peers, thereby resulting in elevated rates of offending (see also Brezina, 1998). Jang and Rhodes (2012), however, find that the effects of strain are partly mediated by social bonds and self-control, but not delinquent peers.

A study by Brezina (2010) highlights the fact that strain may have both emotional and cognitive consequences of a criminogenic nature (see also Konty, 2005). In a national sample of male adolescents, it was observed that angry arousal exerts both direct and indirect effects on violent behavior. Chronic anger tends to foster attitudes that favor aggression, which in turn increase the likelihood of violent offending. These findings support previous theoretical arguments that linked angry arousal to cognitive processes that promote aggression. According to Bernard (1990), angry/frustrated individuals often have difficulty trusting others, attribute hostile motives to strangers, and view aggression as appropriate or justifiable in many different circumstances (see also Agnew, 2006).

Evidence on Conditioning Factors

People differ in their response to strain and only some strained individuals—perhaps a small percentage of strained individuals—respond with offending behavior. According to the initial statement of GST (Agnew, 1992), the likelihood of a deviant response to strain is shaped or conditioned by the individual’s coping skills and resources, availability of social support, association with criminal/delinquent peers, social control, beliefs about crime, and possession of certain traits such as self-control. However, research on the conditioning effects of these factors has produced mixed results.

According to some studies, delinquent peer associations, deviant beliefs, low self-control, and other factors increase the likelihood that strain will lead to deviant outcomes (e.g., Agnew & White, 1992; Keith, 2014; Mazerolle & Maahs, 2000). Other studies, however, fail to observe the predicted conditioning effects (e.g., Hoffmann & Miller, 1998; Mazerolle & Piquero, 1997; for an overview, see, Agnew, 2006). In hindsight, these mixed results may not be surprising. As Agnew (2013) observes, the coping process is very complex:

[The] examination of the coping process is difficult. There are hundreds of coping strategies; individuals often employ several strategies, contemporaneously and over time; the strategies they employ often vary, depending on the stressors they experience and other factors.

(p. 660)

To address the mixed body of results produced by empirical studies, Agnew (2013) has further specified the conditions under which a deviant response to strain is more or less likely. Given the complexity of the coping process, he now argues that a single factor examined in isolation—such as social support or association with delinquent peers—is not likely, by itself, to shape the likelihood of deviant coping. Rather, deviant responses to strain are most likely when multiple factors converge: “The choice of a coping strategy such as crime is likely influenced by the convergence of several factors, including the characteristics of the individual, the characteristics of the stressor, the appraisal of the stressor, and the circumstances surrounding the stressor” (Agnew, 2013, p. 660).

In particular, criminal coping is said to be most likely when highly criminogenic strains are experienced by individuals who have a strong overall propensity to offend and who are in circumstances or situations in which the opportunities for legal coping are limited (Agnew, 2013). This population could include, for example, individuals who are low in social and self-control, belong to gangs, are strongly committed to street culture or live on the street. To advance research in this area, Agnew (2013) now recommends that quantitative studies be based on samples that contain a sizable number of individuals who possess a strong propensity to offend. In contrast, most studies that have examined conditioning factors are based on general population samples. These samples are likely to comprise mostly individuals who have a low overall propensity to offend—individuals who would be unlikely to respond to strain with crime regardless of their standing on particular variables (but see Baron, 2004; Leban et al., 2016; Morris et al., 2012).

In addition, Agnew (2006, 2013) recommends that future studies make an effort to measure the overall standing of individuals on dimensions related to deviant coping, including overall availability of coping resources, total opportunities for legal coping, and general disposition to crime. Lin and Mieczkowski (2011) constructed a composite measure to index the overall standing of young people on various conditioning factors, including moral beliefs, delinquent peer associations, self-control, and self-esteem. The results of study were mixed, as this composite measure conditioned the impact of certain strains on delinquency but not others. Similar findings are reported by Jang and Song (2015) and Ousey, Wilcox, and Schreck (2015). It should be noted, however, that all three studies were based on samples of students in middle school. It is possible that these samples contained few individuals with a strong propensity to offend.

Evidence on Strain and Persistent Offending

Limited evidence indicates that GST has some potential to explain continuity and change in offending behavior. Giordano, Schroeder, and Cernkovich (2007) follow a sample of adolescents into adulthood and observe that changes in trait-based anger are associated with changes in offending even after controlling for social bonds, prior behavior, and other variables. Based on another longitudinal sample of adolescents, Eitle (2010) finds that increases in strain over time are associated with an increase in future offending, while decreases in strain promote desistance from crime. Further, the association between strain and future offending appears to be especially strong for individuals who possess an angry disposition.

Slocum (2010) examines longitudinal data on substance use and finds partial support for the GST explanation of continuity and change. The data indicate that children who obtain high scores on a combined measure of negative emotionality/low constraint tend to report elevated levels of illicit drug use during adolescence and adulthood, controlling for other relevant variables. Further, individuals who possess this trait are more likely than others to respond to strain with depression and substance use. Substance use, in turn, appears to exacerbate problems. It is associated with higher levels of stress, which promotes more substance use in the future. According to GST, negative emotionality/low constraint is partly of function of harsh or erratic parenting. In addition, GST predicts that individuals who possess this trait will tend to provoke negative reactions from others, leading to elevated levels of strain. However, these predictions were not supported.

Evidence on Gender Differences

Numerous studies have examined gender differences in the experience of strain and its emotional and behavioral consequences (e.g., Baron, 2007; Cheung & Cheung, 2010; De Coster & Zito, 2010; Francis, 2014; Hay, 2003; Hoffmann & Su, 1997; Jang, 2007; Jennings, Piquero, Gover, & Pérez, 2009; Kaufman, 2009; Mazerolle, 1998; Morash & Moon, 2007; Piquero et al., 2010; Piquero & Sealock, 2004). Evidence indicates that males are more likely than females to experience certain strains conducive to crime, such as violent victimization, and that this difference partly explains gender differences in offending (e.g., Hay, 2003).

Further, as predicted by Broidy and Agnew (1997), most studies in this area conclude that the reaction to strain is gendered. Males and females appear to react differently to strain or to the emotional consequences of strain. These differences, in turn, are linked to the gender gap in delinquent involvement. However, the exact nature of the observed gender differences varies across studies. This outcome could reflect the different populations that have been sampled across studies, which include adolescents in the general population, justice-involved youth, and adults.

Piquero and Sealock (2004) analyze data from a small sample of justice-involved youth and find that males and females generally experience the same amount of overall strain. Unexpectedly, they observe that the females in their sample exhibit higher levels of anger and depression. Strain predicted anger in both males and females, but it predicted depression in males only. Strain also predicted violence and property crime among males but not among females. Jennings and colleagues (2009) report similar results based on a study of Mexican American adolescents. They examine various individual strains and conclude that some strains affect males and females differently (see also Hay, 2003; Jang, 2007). For example, females are more likely than males to react to academic problems with anger and depression.

In contrast, based on a sample of students in middle school, De Coster and Zito (2010) find that males and females exhibit similar levels of anger, although females exhibit higher levels of depression (see also Kaufman, 2009). They also find that the co-occurrence of anger and depression is more common among females. In addition, they observe that the combined effects of anger and depression have criminogenic consequences but in ways not predicted by GST. Whereas Broidy and Agnew (1997) predicted that depression mitigates the criminogenic effect of anger, thereby suppressing female delinquency, DeCoster and Zito (2010) find no such mitigating effect. Among females, depression does not alter the effect of anger. Among males, however, depression exacerbates the criminogenic effect of anger. According to the authors of this study, the differential impact of emotions across gender likely reflects cultural norms regarding the proper display of emotions. In comparison to their female counterparts, angry and depressed males are more likely to engage in delinquency because the outward expression of anger and depression is consistent with masculinity norms.

Evidence Linking Strain to Community-Level Differences

Can GST explain why some communities (or other macro-level social units) have high rates of problem behavior? Only a handful of studies have tested the macro-level implications of GST, with mixed results. Consistent with GST, Brezina et al. (2001) find that schools harboring a relatively high percentage of angry students tend to have high rates of aggressive behavior, especially fights between students. They also find support for the “interpersonal friction” argument, noting that, in such schools, students in general (not just angry students) have an elevated risk of becoming involved in fights. In another multilevel study of problem behavior in schools, de Beeck, Pauwels, and Put (2012) find that a school-level measure of strain, based on negative future prospects, predicts violence but not other delinquencies. Other school-context variables (such as mean level of negative affect) exhibit little or no relationship to problem behavior. Likewise, a study by Hoffmann and Ireland (2004) produced mixed results regarding the impact of school-context variables on delinquency. Certain school-level measures (such as perceived fairness) predicted delinquency at the individual level, but others did not (such as school problems). Measures of individual-level strain, however, exerted significant effects on delinquent behavior.

Warner and Fowler (2003) assessed the ability of GST to account for rates of violence across neighborhoods. Several findings from this study are noteworthy. First, they find that neighborhood disadvantage and instability are associated with elevated levels of neighborhood strain. Second, neighborhood-level strain exerted a significant effect on violence in communities characterized by a low level of social support. Although it was anticipated that strain would have a stronger effect on violence in neighborhoods characterized by low social control, the findings did not support such a pattern. Hoffmann (2003) examined delinquent behavior across census tracks in the United States and found limited empirical support for the role of strain. In particular, rates of poverty and male joblessness predicted delinquent behavior in urban communities. Moreover, in urban communities with high rates of male joblessness, the effect of individual-level stress on delinquency was magnified.

Summary of Empirical Findings

Overall, empirical tests of GST are generally supportive of the theory’s core propositions. Measures of strain typically predict crime and delinquency, even after factoring in variables from rival crime theories. Moreover, many studies—especially those focusing on aggression—find that the effect of strain is partly mediated by anger. Research on other aspects of the theory, however, has produced inconsistent results. These mixed findings may reflect methodological challenges and the limitations of individual studies (see Mazerolle & Maahs, 2000). They may also reflect a need for greater theoretical specification.

It is important to recognize that GST is an evolving theoretical framework. In response to inconsistent empirical findings, Agnew (2001, 2013) has further specified the types of strain that are relevant to offending as well as the factors that are said to condition the effects of strain on crime. The full potential of these revisions has yet to evaluated, as few studies have fully incorporated the recommended specifications. Tests of GST, however, point to other areas where further specification may be required. For instance, different types of strain may have distinct emotional consequences, leading to distinct behavior outcomes (Ganem, 2010). Likewise, the gender gap in offending may reflect gender differences in the experience of, and reaction to, strain, but unanticipated findings in this area of research suggest that the gendered pathways may be more complex than originally specified (De Coster & Zito, 2010). And initial research suggests that the application of GST to other areas of scholarly inquiry is likely to be fruitful, including, for example, research on racial/ethnic differences in crime (Brezina & Agnew, 2013; Kaufman, Rebellon, Thaxton, & Agnew, 2008) and terrorism (Agnew, 2010). These and other issues provide opportunities for further theoretical development and are likely to stimulate additional research on GST.

It is important to develop a more complete understanding of the relationship between strain and crime because such research may guide crime-control efforts. To the extent that strain contributes to crime, it may be possible to prevent or reduce crime by alleviating the strains that promote offending behavior, by equipping individuals with the skills they need to avoid such strains, or by reducing the likelihood that individuals will cope with strain in a criminal or delinquent manner. Examples of interventions that may have potential in this area include parenting and anger-management programs (for overviews, see Agnew, 1995b, 2006).