Rational Choice Theory Applied to an Explanation of Juvenile Offender Decision Making in the Chinese Setting

Abstract

Rational choice theory (RCT) is a classical theory in criminology, with deep roots in the Enlightenment. It has secured a privileged place as a mainstream criminological theory in the United States. Ironically, RCT has not been applied to research on juvenile delinquency and related decision making in China. This study attempts to test the relative utility of RCT among adjudicated juvenile offenders incarcerated in an institution located in a southwestern province of China based on their responses to two hypothetical scenarios of offending. The results of the Tobit model analyses emerged from the two-wave longitudinal data lent strong support to the view that RCT can indeed serve as an important theory for explaining delinquent behaviors in China. More specifically, juvenile offenders used cost-benefit analysis to decide if they want to be involved either in the stealing scenario or in the fighting scenario. A discussion of findings and public policy implications are highlighted at the end of the paper

Keywords

rational choice theory juvenile offender delinquency longitudinal study China

Introduction

Rational choice theory (RCT) is one of the classical criminological theories that has deep roots in the Enlightenment. For example, utilitarian principles and their accompanying psychological assumptions can be traced back to the writings of Jeremy Bentham ([1789] 1948). Bentham posited that a fundamental aspect of human nature is goal-directed and guided by education, with pleasure seeking the primary good being sought. Human happiness can hence be understood as a composite of efforts to maximize pleasure and minimize pain in one’s life. Some 30 years prior to Bentham, Cesare Beccaria ([1764] 1963) defended the privileged status of individual rights arising from a social contract conception of community, and argued that any punishment should be proportionate to the severity of the crime. RCT holds that criminal acts are generally the result of an individual’s rather instantaneous decision making reflecting a quick calculation of likely costs versus gains. Since the second half of the 20th century, and particularly after the publication of Becker’s (1968) influential foundational article, a major revival of interest in RCT research in the field of criminology has taken place (for a review, see Apel, 2013; Bouffard et al., 2008; Loughran et al., 2016; Matsueda, 2013).

Over the past three decades, scholars working on juvenile delinquency in China have typically borrowed criminological theories developed in the West to explain juvenile delinquency in the Chinese context. The theories borrowed from literature developed in the West include social control theory (e.g., Zhang et al., 2014), general strain theory (Bao et al., 2014), labeling theory/re-integrative shaming theory (Chen, 2002), low self-control theory (Ren et al., 2017), and subculture theory (Ren et al., 2016; Zhang et al., 2017; for a review, see Shen & Hall, 2015; Zhang, 2008). Ironically, to the best of our knowledge RCT has not been applied in research on juvenile delinquency, deviance and related decision making in China. We speculate here that cultural differences in core value orientations may offer an explanation for the lack of application of RCT to juvenile delinquency research in China.

RCT arises from the cornerstone of individualism, and a key building block of individualism is the fundamental principle of understanding as the individual right to unrestricted freedom of choice in making personal decisions. Consequently, individuals exercise their right to make their own choices, and in turn they are fully responsible for the actions they take (Becker, 1968). In stark contrast to the political and cultural heritage of the West, group-oriented harmony is the paramount goal in Chinese society (e.g., Anderson & Gil, 1998; Chen, 2004; Jiao, 2001). More specifically, the essence of Confucian philosophy, which has powerfully shaped both historical and contemporary Chinese culture, entails the achievement of social harmony or “greatest unity.” This is a goal that emphasizes the importance of collective behaviors while downplaying individual accomplishments and claims of personal rights in Chinese society (Anderson & Gil, 1998). For example, obedience to and respect for one’s parents and teachers as well as for elders are often considered an appropriate standard for adolescents (Zhang et al., 2014). Exercising individual rights and being at liberty to decide one’s own course of life are strongly discouraged among adolescents, such notions often being depicted as signs of selfishness and a lack of respect for one’s elders. What is characterized as the “hedonistic pursuit of self-interest” is socially condemned, and when it occurs among Chinese students, it is often viewed as just grounds for casting such youth from society (Jiang & Lambert, 2009). The ever-growing body of literature on RCT published in the United States thus has no equivalent in the Chinese setting.

The purpose of this study is to assess the utility of RCT in understanding juvenile delinquency in contemporary China. More specifically, we seek to examine whether RCT is more universal than we might expect, occurring even in the distinctive Chinese cultural setting. The data were collected from two waves of face-to-face interviews with adjudicated offenders incarcerated in the only juvenile prison in a province with 48 million population. The two waves were separated by 6 months. The rationale for the use of an offender sample was in line with the practice of American scholars who have noted the obvious advantage of surveying offenders rather than college students for understanding deviance (e.g., Bouffard et al., 2008; Decker et al., 1993).

This study attempts to provide answers to the following three research questions regarding the utility of RCT in the Chinese setting. First, it examines the degree to which juvenile offenders utilize cost-benefit analyses in their decisions regarding a property crime event (stealing of battery scenario) and a violent crime event (fighting scenario). The subsequent question concerns the relative contributions of cost and benefit-related variables when juvenile offenders make their decisions in association with deviance and/or delinquency. Do benefit-related variables outweigh cost-related variables, or vice versa? This study seeks to determine whether juvenile offenders in China are more likely to value potential gains than fear potential losses. Finally, as American scholars have proposed, we expand the scope of our inquiry by including additional theoretically important variables such as low self-control and association with delinquent friends, assessing the significance of these theory-derived variables in the analysis (e.g., Matsueda et al., 2006; Nagin & Paternoster, 1993).

Literature Review

The contemporary revival of RCT in criminology began with the publication of Becker’s economic approach to crime and punishment. Mehlkop and Graeff (2010) have noted that Becker’s (1968) article can be viewed as the most prominent and influential work on RCT of crime in criminology. The article emphasizes the language of mathematics to characterize the nature of human decision making in terms of maintaining an ongoing balance between rewards and costs (also see Loughran et al., 2016; McCarthy, 2002). For Becker (1968), the central overriding task for scholars seeking to understand and then treat the problem of juvenile delinquency is to focus on the optimal level of proactive deterrence that can be exercised by government agencies, including the police, prosecutors and courts. According to the economic model of RCT, offenders are not driven by ambient social conditions to any great degree. Instead, offenders’ subjective evaluation of likely costs and potential gains is the primary determinant regarding the commission of crimes (Mehlkop & Graeff, 2010). This presumption clearly bears the imprint of the classical school of RCT. Becker notes in this regard: “Lest the reader be repelled by the apparent novelty of an ‘economic’ framework for illegal behavior, let him recall that two important contributors to criminology during the eighteenth and nineteenth centuries, Beccaria and Bentham, explicitly applied an economic calculus” (1968, p. 209). This economic approach to crime deterrence accords primary attention to national policy on crime at the macro-level, such as the workforce level in law enforcement, the use of technology including closed-circuit television, or the harshness of sentencing. However, individuals’ economic calculations ultimately determine the efficacy of a deterrence-oriented policy. The approach taken to crime follows economists’ customary analysis of choices to be made, assuming that a person is likely to commit an offense if the expected utility to him/her exceeds the utility he/she could gain by using his/her time and resources for other activities (Becker, 1968; Coyne & Eck, 2015).

Mehlkop and Graeff’s (2010) study of the likelihood of future tax evasion is one that closely approximates the principles of Becker’s economic model. Using a sample gathering for the German General Population Survey, these authors included some key variables derived from RCT, including estimates of the probability of success, the extent of benefits to be realized, the probability of being discovered, and the costs involved. The results from their logistic regression analysis demonstrated that the variables that tapped into the probability of success and potential benefits attained statistical significance in predicting respondents’ future intentions to commit the crime of tax evasion (Mehlkop & Graeff, 2010). It is important to note that no other social or demographic variables except for conventional honesty norms were included in the analysis. Similar studies have lent further support to the relationship between a rational choice calculation of costs and benefits and one’s self-reported intention to offend (Baker & Piquero, 2010; Bouffard & Exum, 2013; Bouffard et al., 2008; Pogarsky, 2004). Although the use of RCT frameworks has gradually expanded in recent decades, this economic approach to crime has drawn considerable criticism from scholars since the 1980s (e.g., Bouffard et al., 2008; Nagin & Paternoster, 1993). Indeed, even though it may effectively explain the connection between financial market health and crime rates, the approach tends to fall short when attention is centered on understanding human behavior (Loughran et al., 2016).

In this specific regard, over 40 years ago Cornish and Clark (1987) issued an eloquent call for an expanded scope of research using RCT in light of the bounded rationality of human reasoning (Coyne & Eck, 2015; Sanders, Kuhns, & Blevin, 2017; Simon, 1955). Similarly, Nagin and Paternoster (1993) appealed for the control of other criminological theories into the RCT research effort on deviance and crime-related decision-making processes. They argued that such control is essential for a comprehensive understanding of juvenile delinquency and adult criminal offenses. Nagin and Paternoster incorporated low self-control as an important predictor of three separate outcome variables (theft, drinking and driving, and sexual assault), and reported that “[i]ntentions to engage in three very distinctive offenses—drunk driving, theft, and sexual assault—are positively and very significantly related to lack of self-control” (1993, p. 489). Over time, many scholars have accordingly included other theoretically relevant factors to test rational choice economic calculation models (e.g., Bouffard et al., 2008; Matsueda et al., 2006; Mehlkop & Graeff, 2010; Nagin & Paternoster, 1994; Nagin & Pogarsky, 2001). Consequently, it can be fairly concluded that we have witnessed a significant extension of Becker’s version of the RCT explanation of crime since the 1990s.

Low self-control theory is a primary factor that has been highlighted in most subsequent studies on the testing of RCT. For example, Gottfredson and Hirschi (1990) began their analysis using a basic assumption of human nature, namely that people are conscious, pleasure-seeking beings. They additionally posited that human beings possess an embedded weakness in the exercise of their free will: the frequent pursuit of individual-oriented hedonism. The authors thus hypothesized that one’s propensity to engage in any crime, delinquency or ‘analogous’ behavior (i.e., excessive drinking, chain smoking, or chronic substance use) is the result of low self-control in conjunction with the presence of opportunity. A crime incident is seen as involving an individual’s spontaneous response to his/her own assessment of self-interested gains in the presence of low self-control. Low self-control tends to result from two primary sources: (a) individuals’ tendency to base their behavior on the short-term positive consequences of their actions; and (b) their inability to consider the long-term negative consequences of their actions (Seipel & Eifler, 2010). Gottfredson and Hirschi (1990) maintained that low self-control is a relatively invariant personal trait, developed during early childhood due to poor parental rearing and inadequate adult supervision. Six specific elements or dimensions are said to constitute the core components of low self-control: being impulsive, insensitive, physical, risk-taking, short-sighted, and nonverbal are all said to be associated with persistent delinquency and criminal offending. Gottfredson and Hirschi (1990) argued that these six dimensions tend to coalesce into a unitary trait that can be labeled as low self-control.¹ Since the initial publication of the general theory of crime, the trait of low self-control has been a frequently used predictor of criminal behaviors in empirical studies, and its robustness has been evidenced in the meta-analysis of 21 empirical studies conducted by Pratt and Cullen (2000). These authors found that the effect-size estimate for crime-related behaviors is 0.20, one of the strongest known correlates of crime among criminological theories. They therefore concluded that “future research that omits self-control from its empirical analyses risks being misspecified” (Pratt & Cullen, 2000, p. 952).

Most studies that have tested RCT have included low self-control as an important analytical variable. For example, Nagin and Paternoster (1993) found that low self-control is a significant predictor of intention to commit all three types of crimes studied: theft, driving under the influence (DUI), and sexual assault. Piquero and Tibbetts (1996) suggested that individuals who lack self-control tend to perceive lower certainty and severity of costs, and are more likely to offend as a result. Most recently, Rebellon et al. (2010) examined a sample of 439 undergraduate college students, finding that low self-control is not only correlated with criminal intent, but is also associated with several intermediate variables such as assessments of certainty and self-assessed shaming. Other studies have attempted to expand the scope of investigation by examining the effect of low self-control on estimating the certainty of punishment and the likelihood of offending. For example, Pogarsky (2002) used a scale of impulsivity to predict self-reported likelihood of offending. In analyzing both the full sample and a deterrable sample, the measure of impulsivity failed to achieve 0.05 level of significance (also see Bouffard, 2007).

A second theoretical perspective that is often included in the “thick” RCT model² is differential association theory, a perspective that emphasizes the link between peer influence and criminal offending (Akers, 1998; Sutherland, 1939). In their elaboration of differential association, Sutherland and Cressey (1966) postulated that criminal motives and techniques are learned from a reference group, particularly from close friends (also see Becker & Mehlkop, 2006). This is particularly true of adolescents, as their decision making can be significantly influenced by the peers whose acceptance is sought. For example, using the Denver Youth Survey, a longitudinal data set, Matsueda et al. (2006) employed two separate variables tapping into peer reference groups—the use of violence, and engagement in theft—in order to predict expected likelihood of arrest. They found that both variables were significant predictors after controlling for many other potential intervening variables, and reached the conclusion that peer influence should be incorporated into future research on RCT. However, they were unable to determine if peer use of violence was associated with violent offending per se. Along these lines, Brezina and Piquero (2003) had previously reported that juveniles’ use of drugs at time 2 is strongly influenced by peer acceptance of substance use at time 1. Rebellon et al. (2010) later reported that the effect of a differential association scale of seven items measuring exposure to deviant peers is associated with intent to offend in a range of delinquent activities.

An often overlooked issue associated with the testing of RCT concerns the research subjects of the studies reported. A conventional approach is to use college students as subjects. The advantage of this approach is its convenience and the high level of educational attainment among subjects. However, an obvious disadvantage is the low offending rate present among college student populations. Consequently, the question is whether the findings derived from students can be generalized to offenders. Indeed, the results from several studies have cast serious doubt on the matter of generalizability. For example, Decker et al. (1993) collected data from two independent samples of offenders and non-offenders for the purpose of comparison, finding significant differences in subjects’ perceptions of potential costs and benefits of burglaries (also see Bachman & Schutt, 2011). Non-offenders are generally found to be largely unaffected by variations in the estimation of the certainty of costs and potential benefits, while offenders are strongly affected by such calculations.

Moreover, in their comparison of 212 college students and 93 adjudicated delinquents, Bouffard et al. (2008) identified several cost- and benefit-related items that varied between college students and juvenile offenders. For example, only 16% of juvenile offenders claimed to “feel guilty” in a shoplifting scenario, compared to 48.1% of college students. More importantly, juvenile offenders rated the item “chance of being caught by clerk” significantly lower than college students, suggesting that deterrence may have greater influence on college students than on juvenile offenders. Finally, the results derived from a multivariate analysis showed that the explanatory power of the models for college students and adjudicated offenders differed remarkably between the two samples (Bouffard et al., 2008). Recently, Loughran et al. (2016) used a panel data set of 1,354 adolescent offenders convicted of a serious offense and reported that these juvenile offenders acted in close accordance with the anticipated rewards and costs of offending. In addition, they found that “[a]lthough consistent with much previous research, the reward behaviors generally carry somewhat greater weight than the costs, both are important influences on the youths’ conduct” (Loughran et al., 2016, p. 107). These findings suggest that juvenile offenders tend to overvalue rewards while underestimating the costs of offending.

Our review of the extant literature suggests that RCT remains a major theory for explaining criminal behavior, and that this theory is supported by the empirical evidence reported in several well-designed studies. For example, Matsueda et al. (2006) concluded from both their own work and that of other researchers that criminal acts of violence and theft conform to a rational choice model. A decade later, Loughran et al. (2016) echoed these sentiments by noting that RCT plays an equivalent role as social learning, social control and strain theories. However, it is important to consider that in a well-designed study on college students’ decision making, Exum (2002) urged caution regarding RCT as an explanation for all types of delinquency. Nevertheless, at this point we feel quite confident that RCT is a theory of wide application and broad use, whose validity should be tested in the Chinese setting.

Methods

Data Collection and Sample

The site of this study is the large autonomous region X in China.³ While the vast majority of Chinese people are of the Han ethnicity, accounting for over 90% of China’s population (National Bureau of Statistics of China, 2011), there are 55 officially recognized non-Han ethnic minority groups scattered across the country’s vast territory. Several major minority groups live in the five areas designated as autonomous regions (equivalent to provinces) where the percentage of a particular ethnic minority group is high. Statistics from the most recent national census show that by the end of 2011, the population of autonomous region X was over 47 million, approximately 41% of whom were ethnic minorities (National Bureau of Statistics of China, 2011).

The subjects of this study were adjudicated juvenile offenders serving time in the juvenile correctional institution in autonomous region X. A longitudinal sample of adjudicated juvenile offenders was required to capture the effect of extent of association with delinquent friends. Out of concern for the relatively low level of literacy present among the targeted population, the method of face-to-face interviews was preferred over the use of self-administered paper-and-pencil surveys. The face-to-face interviews were conducted at two points in time separated by 6 months, and each took an average of 45 min. All interviews were conducted in a secure, designated private room within the correctional institution, where only the researchers and the juvenile offenders were permitted to be present during interview. The members of the research team received an intensive 6-hr training class at a local university prior to starting the project. The interviewer training covered the content of the instrument, the interview techniques to be employed, and the security procedures to be followed within the correctional facility. Participation in the on-site interviews was completely voluntary. The researchers and university-based graduate research assistants informed the juvenile offenders of their right to decline participation and provided assurances of confidentiality and anonymity prior to the start of the interviews. Over the course of the study, many juvenile offenders expressed appreciation for the opportunity to talk with the researchers and to share their thoughts about being in a juvenile prison facility.

The first wave of data collection ran from January 2 to January 14, 2014. The juvenile prison population on January 2, 2014 was 1,540. The interview subjects were juvenile offenders who had been sent to the prison in 2013 (n = 846). Among them, 138 were released in the same year. In addition, 13 of the youth did not participate due to illness or being in the segregation unit (n = 5), while the remainder declined to participate (n = 8). The final eligible sample was 695 at Wave 1. Wave 2 of the survey took place from July 10 to July 25, 2014, with 162 juvenile offenders not participating because they had served their sentence and had been released. Consequently, 533 Wave 1 juvenile offenders remained in prison at the point of the Wave 2 data collection, of whom 17 did not participate due to illness or to being in the segregation unit (n = 7), or alternatively declined to participate (n = 10). The final sample of inmates with participation in two waves of the survey comprised 516 juvenile offenders.

Dependent Variables

The dependent variables used in this study were derived from two hypothetical offending scenarios, namely shoplifting and being involved in a physical fight. Two questions were asked after the researchers had read the hypothetical scenarios to the offenders at Wave 2. The sharing of scenarios involved the oral presentation of a brief fictional depiction of potential offending situations followed by a request for intentions to offend in such instances. The decision to offend was assumed to reflect calculations regarding the perception of potential costs and benefits associated with the scenarios portrayed by the researchers (Bouffard et al., 2008; Decker et al., 1993; Mehlkop & Graeff, 2010; Nagin & Paternoster, 1993). Over the course of the past 25 years, RCT has been used to examine several types of intended criminal behaviors. The most commonly used scenarios include shoplifting (Bouffard et al., 2008; Nagin & Paternoster, 1993), physical fight (Bouffard, 2007; Exum, 2002), DUI (Bouffard, 2007; Nagin & Paternoster, 1993; Pogarsky, 2002, 2004), and a combination of these offenses (Mehlkop & Graeff, 2010).⁴ Nagin and Paternoster (1993) have noted that several advantages of the scenario method outweigh others in the testing of RCT, including the clear link between costs/benefits calculation and intention to offend, the ability to collect respondents’ depictions of their calculations of potential costs and benefits, and the observation of the sequence of specific steps followed to determine situational gains and losses in the ultimate formation of intentions to offend or not.

Two offense scenarios describing the shoplifting of a battery and a physical fight were included in the face-to-face interviews (scenarios presented in Appendix A). The two offending scenarios were adapted from similar studies conducted with college students and offenders (Bouffard et al., 2008). After listening to the shoplifting scenario read by the researcher, the offenders were asked to indicate the probability of “taking the batteries without paying.” The second scenario depicted a hypothetical physical fight situation arising in a bar because another juvenile is flirting with the offender’s girlfriend. Respondents were asked to indicate their likelihood to “pick up the fight” in this situation. The range of response varied from 0% (no chance) to 100% likelihood (definitely would).

Explanatory Variables

Apel (2013) has described how people differ in their perceptions of the consequences of various lines of action, and of how these perceptions shape their choices about participating in the behaviors in question. At Wave 2, data were collected on four scales measuring both the costs and the benefits of offending, alongside the associated certainty and severity of punishment related to the two scenarios. Each of these scales was composed of seven specific items (the scores of the items were later added to constitute a scale in the statistical analyses; please see Table 1). The Cronbach’s alpha statistics for each of the scales with a total of 56 items were very strong, ranging from 0.83 to 0.93. After each scenario was read out and the likelihood of offending questions were answered, respondents were asked to identify seven “good things” (benefits) and seven “bad things” (costs) that might occur if they engaged in the offending behavior described in the two scenarios.⁵ Having provided this information, the juvenile offenders were asked to report their perception of the seven negative outcome items associated with the costs and benefits for each scenario. The answers ranged from 0% (not all important) to 100% (very important) in their decision-making process (Bouffard et al., 2008; Matsueda et al., 2006; Nagin & Paternoster, 1993; Rebellon et al., 2010). The scenario design used in this study represented a hybrid methodology, combining the use of hypothetical scenarios and traditional survey questions. Nagin and Paternoster (1993) have convincingly argued that such a hybrid design is superior to past data collection methods employed in perception-based research on deterrence.

Table 1.

Specific Cost and Benefit Types by Scenario (n = 514).

Scenario	Cost type	Mean (certainty of cost)	Mean (severity of cost)	Benefit type	Mean (certainty of benefit)	Mean (importance of benefit)
Shoplifting	Caught by the police	58.88	70.54	Possess batteries	44.57	26.82
	Caught by clerk	58.47	67.27	Save money	42.10	27.16
	Guilt	61.96	64.89	Get away with it	38.91	29.02
	Social (e.g., reputation)	74.98	75.02	Convenience	29.06	23.67
	Embarrassment	71.87	71.39	Social (e.g., peer respect)	14.08	16.04
	Morality	73.69	71.66	Caught, learn lesson	40.95	35.14
	Disgrace family	81.41	78.55	Emotions (e.g., confidence)	18.22	17.81
Physical fight	Injure self	61.88	60.37	Deter specific person	47.24	43.11
	Injure other person	63.98	64.02	Emotional benefit (for self)	53.31	49.06
	Bad for the relationship	67.93	64.32	Good for relationship	37.15	44.37
	Legal	67.26	66.89	Win fight	42.17	38.32
	Embarrassment	56.71	55.60	Deter others (in general)	39.25	38.48
	Social (e.g., reputation)	58.04	58.80	Social (e.g., peer respect)	33.76	36.93
	Retribution from others	56.93	55.27	No one gets hurt	36.70	40.84

The distributions of means on the eight scales produced several noteworthy findings. For example, bringing disgrace on the family (M = 81.41), reported in the potential cost of shoplifting, was the highest-rated item among juvenile offenders, surpassing the two deterrence-oriented items (e.g., caught by the police or the clerk). Similarly, “bad for the relationship” (M = 67.93) and “emotional benefit” (M = 53.31), reported in the physical fight scenario, were the two items that received the highest ratings. Overall, the means of items measuring the severity of costs and the importance of benefits were close to the means of costs and benefits identified, aside from the heightened importance of benefit shown in the shoplifting scenario.

The fifth explanatory variable tapped into the concept of low self-control. Akin to other studies testing RCT, we included the commonly used 24-item scale developed by Grasmick et al. (1993). Given the established fact that low self-control is rather invariant across time, Nagin and Paternoster have argued that the link between low self-control and self-reported intentions to offend is “instantaneous” in nature (1993, p. 474). Following this suggestion, we used Grasmick et al.’s 24-item scale of low self-control at Wave 2 in the analysis. The Cronbach’s alpha of this scale was a robust 0.83.

It is important to note that measures of low self-control do vary somewhat across studies. For example, Pogarsky (2002) used six items to measure the concept of impulsivity, a key component of low self-control derived from the NEO (Neuroticism-Extraversion-Openness Inventory) Personality Inventory. The fifth and sixth variables measured the association with delinquent friends by specific types of delinquent activities. In this study we followed Matsueda et al.’s (2006) study by differentiating delinquent friends engaged in property crimes from delinquent friends doing violent crimes. A theoretical reason distinguishing peer influence by the types of crime in which they ‘specialize’ was assumed (Matsueda et al., 2006). For example, in their in-depth study on perceived sanction threats among offenders, Jacobs and Cherbonneau (2017) found that offenders tend to specialize in the particular type of crime with which they feel comfortable. More specifically, they noted that auto thieves are reluctant to embrace the violence of carjacking largely due to their concerns about sanction threat severity (Jacobs & Cherbonneau, 2017). We believe that the data collected in our study on both shoplifting and physical altercations provided a good opportunity to examine the efficacy of differential association theory.

Seven items were used to gauge the level of association with delinquent friends engaging in property crimes (data collected at Wave 1). Juvenile offenders were asked whether they have friends who steal from their homes, engage in shoplifting, steal others’ backpacks or purses, burglarize others’ homes, steal bicycles, steal electric scooters, steal motor vehicles, or steal property inside motor vehicles (alpha = 0.89). The association with friends who engage in violence was measured by six items regarding friends who carry weapons (knife, sword, or iron chain), are involved in gang fights, beat people up, threaten others with a knife or stick, and stab someone (alpha = 0.90). The responses were based on a Likert-type scale ranging from 1 = no such friends through 2 = very few such friends to 5 = almost all my friends. We were particularly interested in the link between the type of delinquent peers identified at Wave 1 and the intention to offend at Wave 2.

We included three variables pertaining to a juvenile offender’s criminal history (data collected at Wave 1). The first variable concerned prior history of arrest. Matsueda et al. (2006) suggested that prior arrests by the police can enhance juveniles’ perceptions of perceived risk, hence this was considered as a baseline effect. Such prior arrest was measured as a dichotomous variable, with 0 = no prior arrest and 1 = prior arrest at Wave 1. The second variable was a measure of the most recent offense for which a juvenile offender had been convicted prior to incarceration at Wave 1. Offense type of crime prior to conviction is often used to capture a glimpse of an offender’s pre-institutional behavior (Cao et al., 1997; Duwe & Johnson, 2016). This variable was coded as 0 = property crime and 1=violent crime. The third variable was the length of time served in prison. The time served in prison and offense type were included because the relevant literature suggests that prison time served is an important variable. For example, Boessen and Cauffman (2016) found that prison time was a significant predictor of institutional offending among juvenile offenders in their longitudinal study (also see Dhami et al., 2007). Similarly, Duwe and Johnson (2016) noted that length of stay in prison was significantly correlated with prison adjustment. Length in prison was measured by the number of days prior to the second face-to-face interview.

Finally, five demographic variables were included in the analysis for the purpose of controls. Ethnicity was coded as 0 = Han and 1 = ethnic minority, while age was measured as the natural age in years of a respondent at the time of interview. Educational attainment was coded from 1 = illiterate through 2 = did not graduate from grade school to 7 = graduated from high school or vocational school. Given that this was a juvenile offender sample, none of the respondents had any college experience. Location of residence was measured by 0 being urban area and 1 being rural area. Economic status was a self-assessed measure of economic condition at home, coded as 1 = poor, 2 = low income, 3 = average, and 4 = affluent. The only child measure was a dichotomous variable assessed by asking a respondent if he or she was the only child in the family, with 0=no and 1=yes.

Statistical Considerations

The Tobit model or a censored regression model is a statistical analytical approach that has been proposed by James Tobin to describe the relationship between a non-negative dependent variable and independent variables. We decided to use the model in the analysis because previous similar studies have consistently shown that it is an appropriate statistical model due to floor and ceiling effects. More specifically, the two dependent variables are measured from 0 to 100 in probabilities. There is a real possibility that there are floor and ceiling effects, as generated by problems in the measurement scale in the estimates as in comparison to the effects generated by a true probability scale that allows negative values or values exceeding unity (Matsueda et al., 2006). The analysis of the data revealed that the percentage of respondents who reported that there was “no chance” that they would commit the specific theft (e.g.,) was 67.7 percent, while 36.2 percent of the study participants selected the same answer in the physical fight scenario. This unique feature of the scenario design was related to the considerable number of zero values reported in the dependent variable, because the variation of dependent variables ranges from 0 to 10 or 0 to 100 (Bouffard et al., 2008). Matsueda et al. noted that “[o]ur Tobit models address violation of three key assumptions of the usual linear regression model. First, the dependent variable, perceived risk, is measured in probabilities, which bound the scales at zero and 100. Consequently, linear models will yield downwardly biased coefficient estimates due to floor and ceiling effects when the data clump at zero or 100” (2006, p. 106). Similarly, Nagin and Paternoster (1993) employed the Tobit modeling approach when a large percentage of their respondents answered zero to three scenarios (theft = 63%, drinking and driving = 33%, and sexual assault = 85%). Other scholars testing the limits of RCT have also used Tobit models in their analyses to appropriately address the potential issue of zero-value bias (Pogarsky, 2002; Rebellon et al., 2010).

Findings

Descriptive statistics for all variables are displayed in Table 2. All of their means are reported for the initial assessment of the appropriateness of the model for the data collected from juvenile offenders to be analyzed. For two dependent variables or scales, the mean probability of shoplifting was 14.73 out of 100, while the probability of fighting was 40.26, given that more respondents reported zero chance to steal (67.7%) than to fight (36.2%).

Table 2.

Descriptive Statistics (n = 514).

Variable	Mean	S.D.	Min/Max	α
Dependent variable
Probability of shoplifting	14.73	26.44	0–100
Probability of physical fight	40.26	38.71	0–100
Independent variable
Certainty of benefit of shoplifting	32.56	25.58	0–100	0.84
Importance of benefit of shoplifting	25.36	26.38	0–100	0.86
Certainty of cost of shoplifting	68.75	27.24	0–100	0.90
Severity of cost of shoplifting	49.93	20.85	0–100	0.93
Benefit of physical fight	41.37	25.95	0–100	0.83
Importance of benefit of physical fight	41.59	29.26	0–100	0.87
Cost of physical fight	61.82	27.28	0–100	0.88
Severity of cost of physical fight	60.75	29.50	0–100	0.91
Low self-control	1.96	0.64	1–4	0.83
Delinquent violent friends	10.37	5.06	5–25	0.90
Delinquent property friends	9.15	3.51	7–33	0.89
Length of incarceration	183	102.76	17–378
Ever arrested before	0.34	0.47	0–1
Offense type	0.94	0.24	0–1
Age	16.81	0.96	14–19
Education attainment	3.78	0.96	1–7
Economic status	2.47	0.73	1–4
Only child	0.20	0.40	0–1
Location of residence	0.69	0.46	0–1
Ethnicity	0.41	0.49	0–1

The means of the next four explanatory variables reveal that the respondents believed that the overall cost of shoplifting (M = 68.75) and its overall severity (M = 49.93) are greater than the two benefit variables associated with shoplifting (M = 32.56 and 25.36). Accordingly, the means of these variables representing the physical fight scenario suggest that the costs of fighting (M = 61.82) and their collective severity (M = 60.75) are higher than the overall benefits of fighting (M = 41.37) and their collective importance (M = 41.59). The perceived benefits of fighting are higher than the perceived benefits of shoplifting, whereas the overall perceived costs of fighting and stealing were fairly close.

The means of low self-control (M = 1.96, alpha = 0.83), association with friends who participate in delinquent violent offenses (M = 10.37, alpha = 0.90), and association with friends who participate in delinquent property offenses (M = 9.15, alpha = 0.89) are also reported. This survey-based finding is consistent with the nature of the sample universe, as 94%of the incarcerated juvenile offenders in this facility had committed violent offenses. The average length of incarceration was 183 days at the time of the survey, and 34% of the study participants reported having a prior arrest history.

The control variables indicate that the average age of the juvenile offenders participating in the survey was 16.81 (ranging from 14 to 19). Ethnic minority juvenile offenders accounted for 40.66% of the surveyed respondents. The mean of educational attainment was 3.78, a scale value falling between grade school completion and not graduating from middle school. Family economic status was measured by their own perception on a continuum ranging from poor to well off, with a mean of 2.47 out of 4. Finally, 20.4% of the juvenile offenders reported being the only child in the family, and 69% grew up in a rural area.

The results from the Tobit analysis of probability of shoplifting are reported in Table 3. Model 1 (the second column) presents the results of the demographic variables alone. The results in Model 2 include the demographic controls as well as the variables measuring low self-control, differential association, and offender-related variables. Finally, the results for all variables—including those measuring costs and benefits—are shown in Model 3. The findings suggest that none of the demographic variables are statistically significant in Model 1, and the variable of association with friends who participate in delinquent property offenses is a significant predictor of probability of shoplifting. The coefficient for association with friends who participate in delinquent property crimes indicates that a unit change leads to a 3.16 corresponding change in the probability of stealing the battery in Model 2. The addition of benefit- and cost-related variables in Model 3 reveals that two of the benefit-related variables are significant predictors, but not the two cost-related variables. Of particular relevance is that the association between perceived benefits of shoplifting and the dependent variable for probability of shoplifting is statistically significant, indicating that a unit change in benefits of stealing (0–100) can lead to a 0.75 change in the dependent variable on the same 0 to 100 scale. In comparison, the magnitude of the 0.001 level of statistical significance for the importance of the benefits of shoplifting is about the half the strength. Neither of the two cost-related scales yield any statistically significant effects on the dependent variable. Overall, the probability > chi-square shows that the model is highly significant at the 0.00001 level.⁶

Table 3.

Probability of Shoplifting.

	Model 1			Model 2			Model 3
Variables	Coefficient	S.E.	t-value	Coefficient	S.E.	t-value	Coefficient	S.E.	t-value
Constant	−41.39	65.44	−0.63	−112.91	74.52	−1.52	−109.36	67.44	−1.62
Age	0.50	3.54	0.14	3.11	3.79	0.82	4.19	3.42	1.23
Education attainment	4.96	3.58	1.38	3.52	3.51	1.01	4.51	3.23	1.40
Economic status	−4.30	4.70	−0.91	−5.34	4.67	−1.14	−4.92	4.21	−1.17
Only child	−5.99	8.31	−0.72	−5.09	8.22	−0.62	−7.58	7.44	−1.02
Location of residence	−4.14	7.32	−0.56	−3.41	7.27	−0.47	−1.74	6.59	−0.26
Ethnicity	10.89	6.86	1.59	12.01	6.76	1.78+	11.64	6.17	1.89+
Low self-control				9.92	5.29	1.87+	2.49	4.91	0.51
Delinquent violent friends				0.63	0.78	0.81	0.97	0.71	1.37
Delinquent property friends				3.16	1.04	3.04**	2.23	0.94	2.38*
Ever arrested before				−4.41	7.01	−0.63	−6.59	6.37	−1.03
Offense type				−9.47	14.05	−0.67	−12.66	12.96	−0.98
Length of incarceration				−0.04	0.03	−1.17	−0.05	0.03	−1.47
Certainty of benefit of shoplifting							0.75	0.16	4.75***
Importance of benefit of shoplifting							0.34	0.15	2.31*
Certainty of cost of shoplifting							−0.23	0.18	−1.29
Severity of cost of shoplifting							−0.26	0.23	−1.1
Log likelihood	−1104.8961			−1074.9736			−1040.3396
Prob > Chi2	0.44901			0.0024			0.0000

Note. +p < .10. *p < .05. **p < .01. ***p < .001.

Table 4 reports the results derived from a Tobit model analysis of the probability of engaging in a physical fight in a bar scenario. The pattern of demographic variable effects is very similar to that for the shoplifting case. Moreover, there are only two significant variables present in the Model 2 results: the variable for delinquent friends who resort to violence, and the variable for low self-control. The result for self-control indicates that a unit increase in low self-control (1–4) can result in a corresponding 15.16 increase in the dependent variable, suggesting that the effect of this association is substantial. The results reported in Model 3 imply that delinquent friends who resort to the use of violence remains statistically significant, and the perceived benefit of fighting and the importance of the benefits of fighting are significant predictors of the likelihood of engaging in fighting. Particularly noteworthy here is the perceived benefit of physical fighting, attaining a t-value of 5.86. The corresponding unit change between benefits of fighting and probability of fighting is substantial as well. Again, the two cost-related scales are not found to be significant influences. The model fit value shows that the model is statistically significant at the 0.00001 level.

Table 4.

Probability of Physical Fight.

	Model 1			Model 2			Model 3
Variables	Coefficient	S.E.	t-value	Coefficient	S.E.	t-value	Coefficient	S.E.	t-value
Constant	16.60	49.77	0.33	−73.85	55.96	−1.32	−87.86	48.55	−1.81+
Age	−0.13	2.70	−0.05	2.56	2.85	0.90	2.34	2.47	0.95
Education attainment	4.06	2.73	1.49	3.42	2.64	1.29	1.87	2.31	0.81
Economic status	−5.98	3.61	−1.66+	−5.88	3.52	−1.67+	−2.71	3.07	−0.88
Only child	7.82	6.53	1.20	8.28	6.33	1.31	10.26	5.48	1.87+
Location of residence	−1.81	5.67	−0.32	0.89	5.54	0.16	6.01	4.86	1.24
Ethnicity	0.75	5.30	0.14	2.53	5.14	0.49	3.71	4.48	0.83
Low self-control				15.16	4.04	3.75***	6.88	3.59	1.92+
Delinquent violent friends				1.99	0.58	3.42**	1.18	0.51	2.32*
Delinquent property friends				−0.43	0.83	−0.52	−0.26	0.72	−0.36
Ever arrested before				6.55	5.39	1.22	5.30	4.68	1.13
Offense type				−10.98	10.79	−1.02	−12.87	9.32	−1.38
Length of incarceration				−0.04	0.03	−1.46	−0.05	0.02	−2.23*

Certainty of benefit of physical fight							0.76	0.13	5.86***
Importance of benefit of physical fight							0.36	0.12	3.15**
Certainty of cost of physical fight							−0.08	0.14	−0.58
Severity of cost of physical fight							−0.17	0.13	−1.25
Log likelihood	−1948.683			−1911.1522			−1842.4028
Prob > Chi2	0.3475			0.0000			0.0000

Note. +p < .10. *p < .05. **p < .01. ***p < .001.

Discussion and Conclusion

The present study has attempted to fill the void by applying RCT to the Chinese setting, based on the premise that the popularity of group-oriented culture is different from RCT, which has deep roots in the Enlightenment. Consequently, the first research question concerned whether or not Chinese juvenile offenders make use of a cost-benefit analysis when asked about their intentions related to two hypothetical scenarios involving theft and fighting. Four scales (variables) were operationalized to assess the influence of RCT in two distinct scenarios. The results from the Tobit model analysis have lent strong support to the view that RCT can indeed serve as an important theory for explaining delinquent behaviors in China.

Adding to our belief in the utility of RCT for understanding deviance in the Chinese setting is the fact that we took a careful look at the decision-making mechanism of cost-benefit analysis exhibited among Chinese juvenile offenders. A unique characteristic of the scenario research design, for example, concerns the number of zeros that participants tend to select (e.g., Nagin & Paternoster, 1993; Rebellon et al., 2010). Becker (1968) argued in his foundational article that cost and benefit analysis is the primary factor determining the outcome of a deviance decision. If this is truly the case, then those who indicated an intent to steal a battery or to fight for one’s reputation were more likely to see the benefits of doing so than those who declined to engage in either action. Accordingly, we conducted additional analyses to examine differences in costs versus benefits between the zero group of juvenile offenders (desistors) and the non-zero group (deviants) among the Chinese juvenile offenders. The results reveal a highly significant difference in mean ratings between the zero group and the non-zero group. Indeed, juvenile offenders in the zero group rated the perceived benefit of stealing a battery (M = 28.15) significantly lower than their counterparts in the non-zero group (M = 41.80). At the same time, they were more likely to rate the costs (M = 71.52) higher than juvenile offenders in the non-zero group (M = 62.96). The same pattern of differences held in the physical fight scenario: juvenile offenders in the zero group rated the benefits of engaging in a physical fight (M = 27.96) lower than did their counterparts in the non-zero group (M = 48.97). This set of findings is consistent with a primary assumption of RCT that juvenile offenders take into consideration the calculation of benefits versus costs in their decision-making process (Becker, 1968; Bouffard et al., 2008). The F-values of the two scenarios suggest that the between-group difference of the benefits of physical fighting reported in the physical fight scenario (F = 91.51) was more than double that documented in the shoplifting scenario (F = 34.04).

Our second research question pertained to the relative contribution of benefit-related scales versus cost-related measures. The results reported in Tables 2 and 3 provided sufficient evidence that Chinese juvenile offenders tended to pay more attention to benefits or “good things” than to costs or “bad things” in forming their intentions. Consistently across the two tables, the two coefficients tapping into the benefits and the importance of benefits were found to be significant predictors of intention to be engaged in a common specific form of theft or to be involved in a physical fight in a bar setting, while none of the cost-related variables reached statistical significance. In addition, the magnitudes of the coefficients of benefit-related variables were uniformly large and consistent across the two scenarios. A unit change in benefit of either shoplifting or physical fight led to 0.76 or 0.75 change in the likelihood of intention to steal or fight. Loughran et al. have noted in this regard that benefit calculation is the core of RCT: “It is the third component of Becker’s (1968) model, the benefits from crime Y, which highlights the most compelling distinction between deterrence studies and the broader concept of rational choice” (2016, p. 90). This is consistent with the relevant literature that offenders tend to exemplify the effects of benefits while underestimating the cost-associated factors (e.g., Loughran et al., 2016).

Our third research question concerned the inclusion of data pertaining to additional theories that represent alternatives to RCT in the production of deviance (Nagin & Paternoster, 1993; Rebellon et al., 2010). In fact, almost all studies that have tested RCT have incorporated variables measuring low self-control, previous delinquent behaviors, and social learning theory (Matsueda et al., 2006). The results from the multivariate analyses undertaken here strongly suggested that the degree of association with delinquent peers was indeed a significant predictor of intent to offend. Moreover, we found that the variable of degree of association with friends who participate in delinquent property offenses was linked to the shoplifting scenario outcomes among Chinese juvenile offenders, and the degree of association with delinquent friends who resort to violence was likewise associated with outcomes from the physical fight scenario. The t-values of the two delinquent friend-related variables were above 2.3, well over the cut-off margin of 1.96 for achieving statistical significance. In addition, we used delinquent peer variables at W1 to predict intention to steal or fight in W2. Taking these factors into consideration, we reached the conclusion that questions regarding different types of delinquent peers must be included in future research. The key point found in this study is that specific types of delinquent peers are linked to specific crime scenarios. Relevant research on fear of crime, for example, shows a positive link between the types of crime incidents that occur in one’s neighborhood and the specific type of fear that citizens experience. In this regard, Wilcox Rountree (1998) found that burglary incidents are associated with fear of residential burglary in Seattle neighborhoods. Similarly, Moore and Shepherd (2007) have suggested that fear of crime should not be viewed as a monolithic construct, but instead be assessed with respect to several speciﬁc crime types, such as personal crime (e.g., robbery) and property crime (e.g., burglary; also see Lai et al., 2011). It is important to note that the results derived from the three research questions revealed that Chinese juvenile offenders exhibited similar decision-making processes to their counterparts reported in the United States.

Low self-control is a measure that has been discussed extensively in the literature testing RCT (for a review, see Rebellon et al., 2010). It has been assumed that low self-control remains largely invariant after childhood (Gottfredson & Hirschi’s, 1990). The literature is replete with studies showing that participants who have low self-control are more likely to commit delinquent behaviors, but findings on the utility of the self-control measure when testing RCT are rather mixed (e.g., Exum, 2002; Nagin & Paternoster, 1993; Pogarsky, 2002). In the present study, low self-control had no effect on intention to steal a battery, but it did have an almost statistically significant effect in the fighting scenario (t-value = 1.92). Relevant literature suggests that burglars are likely to select their targets very carefully prior to undertaking a burglary. For example, burglary incidents tend to cluster in time and space, especially as regards repeated burglary incidents, necessitating pre-planning and the careful selection of suitable targets (Bowers & Johnson, 2005; Townsley et al., 2003; Zhang et al., 2015). In contrast, one of the primary characteristics of physical fighting is a lack of self-control, particularly simultaneous anger stimulated by a specific situation escalating out of control. The results reported in this study were somewhat inconclusive, and we urge caution in drawing any firm conclusions. Future research must investigate the link between low self-control and different types of RCT-related scenarios. Different measures of self-control should be carefully assessed in future research.

In addition, none of the five demographic variables included in this study reached statistical significance at the 0.05 level. The only childhood variable was close to the significance level at 0.10 in the model of probability of engaging in a physical fight (t-value = 1.87). A possible explanation for the lack of significant impact of the demographic variables is the issue of limited variance among the juvenile offenders who participated in the interviews, as they were very similar in background. Of particular interest is that 94% of the study participants had been convicted of violent offenses. Unsurprisingly, most preferred engagement in a physical fight to stealing from a convenience store. Future research should pay close attention to the relationship between demographic background factors and juvenile offenders whose criminal offenses are very similar.

Finally, the results derived from this study were to a certain extent surprising. The possible difference between group-oriented culture and individually oriented culture piqued our interest in the testing of RCT in the Chinese setting. Certainly, recently published studies have shown that traditional Chinese culture has made a significant contribution to the variation in students’ attitudes toward the police. In their empirical study on the utility of Chinese culture, for example, Wang et al. (2020) found that traditional Chinese culture—and especially collective-oriented values—was positively related to juveniles’ favorable perceptions of the police. In contrast, Western popular culture was a negative predictor of attitudes toward the police. Similarly, the utility of a unique Chinese culture has been reported in many articles (Jiao, 2001; Ren et al., 2016; Zhang et al., 2014). This is the first study to identify the practical implication of RCT, a theory with Western origins but used among juvenile offenders in the Chinese setting. We speculate that rapid economic development and globalization have significantly altered the perspectives of young people in China, enabling them to feel less constrained by traditional values (e.g., Cao, 2007). In addition, the influence of traditional values is very limited among juvenile offenders whose educational attainment and mainstream values are particularly low.

Policy Implications and Limitations

The findings of this study suggest three very important policy implications for professionals who work in juvenile justice agencies. The first policy implication is that prison educational programs that attempt to prepare juvenile offenders for reentry into civil society after release should maximize awareness of the consequence of costs incurred in crime commission while minimizing the exaggerated “dream” of benefits. “No luck” is the best assumption to be made to remain free; this is the theme of such programs. Next, prison administrators should identify “habitual” juvenile offenders who become well-known to the authorities. The findings from this study are unequivocal regarding the peer effect at Wave 1 and the subsequent impact on intention to steal or fight at Wave 2 (also see Brezina & Piquero, 2003). It follows that avoidance of contact with former delinquent peers should be a goal in reentry and community supervision after release. The third policy implication is related to the concept of individual responsibility. Juvenile prison education programs need to expand their content to include coverage of the individual responsibility lessons to be learned from the punishment of incarceration. Traditional Chinese culture emphasizes social harmony, while the protection of individual rights and responsibility for one’s own future are not viewed as noteworthy objectives of learning. Within this collectivist cultural society, an important means of highlighting the costs of crime is to promote self-responsibility for social harmony in society. Juvenile offenders need to be more fully aware of the fact that every decision they make regarding deviance and crime generates greater costs for themselves and society than the benefits associated deviance and crime.

This study has several limitations that need to be acknowledged. First, we were unable to incorporate culture-oriented measures in the analysis. The present study has only proven the utility of rational choice theory in the Chinese setting. Little is known regarding the potential contribution of unique Chinese culture in juvenile offenders’ decision-making process. Future research should include culture-oriented variables in the test of RCT in China. Second, costs and benefits are two key but excessively broad concepts in RCT. In this regard, Loughran et al. (2016) have noted that the concept of rewards resulting from crime requires a comprehensive measure. In this study we incorporated the items that were developed and tested in studies conducted by others (Bouffard, 2007; Nagin & Paternoster, 1993). Future work in comparative criminology research needs to develop culturally sensitive measures of costs and rewards to be used to assess theories developed in Western societies. Another limitation of this study concerns the difference of rational decision making between juvenile offenders and non-offenders (e.g., students). In their study, Bouffard et al. (2008) found that compared with adjudicated juvenile offenders, students are more likely to believe the chance of being caught by the police is high and that students tend to have a stronger sense of guilt. This suggests that there are some perceptual differences of risk and benefit between juvenile offenders and students. Future research should include a student sample and answer this question of difference in the Chinese setting. Finally, the sample was collected from an autonomous region containing a population of 47 million. Coastal regions of China can differ quite substantially from inland regions in many ways, including in terms of the dynamics associated with deviance, delinquency, and crime. Future studies should therefore feature samples from both inland and coastal regions of China.

Footnotes

Appendix A

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Xingting Wang

Hongwei Zhang

Ruohui Zhao

Notes

References

Akers

R. L.

(1998). Social learning and social structure: A general theory of crime and deviance. Northeastern University Press.

Anderson

A. F.

Gil

V. E.

(1998). China’s modernization and the decline of communitarianism. Journal of Contemporary Criminal Justice, 14, 248–261.

Apel

(2013). Sanctions, perceptions, and crime: Implications for criminal deterrence. Journal of Quantitative Criminology, 29(1), 67–101.

Bachman

Schutt

P. K.

(2011). The practice of research in criminology and criminal justice (4th ed.). Sage Publications.

Baker

Piquero

A. R.

(2010). Assessing the perceived benefits-criminal offending relationship. Journal of Criminal Justice, 38(5), 981–987.

Bao

Haas

Chen

(2014). Repeated strains, social control, social learning, and delinquency: Testing an integrated model of general strain theory in China. Youth and Society, 46(3), 402-424.

Beccaria

([1764] 1963). On crimes and punishments. Bobbs-Merrill.

Becker

G. S.

(1968). Crime and punishment: An economic approach. Journal of Political Economy, 76, 169–217.

Becker

Mehlkop

(2006). Social class and delinquency: An empirical utilization of rational choice theory with cross-sectional data of the 1990 and 2000 German General Population Surveys (ALLBUS). Rationality and Society, 18(2), 193–235.

10.

Bentham

([1789] 1948). An introduction to the principle of morals and legislations. Blackwell.

11.

Boessen

Cauffman

(2016). Moving from the neighborhood to the cellblock: The impact of youth’s neighborhoods on prison misconduct. Crime and Delinquency, 62(2), 200–228.

12.

Bouffard

J. A.

(2002). Methodological and theoretical implications of using subject generated consequences in tests of rational choice theory. Justice Quarterly, 19(4), 747–771.

13.

Bouffard

J. A.

(2007). Predicting differences in the perceived relevance of crime’s costs and benefit in a test of rational choice theory. International Journal of Offender Therapy and Comparative Criminology, 51(4), 461–485.

14.

Bouffard

J. A.

Bry

Smith

, & Bry, Rhonda. (2008). Beyond the “Science of sophomores”: Does the rational choice explanation of crime generalize from university students to an actual offender sample? International Journal of Offender Therapy and Comparative Criminology, 52(6), 698–721.

15.

Bouffard

J. A.

Exum

M. L.

(2013). Rational choice beyond the classroom: Decision making in offenders versus college students. Journal of Criminal Justice, 41(6), 438–447.

16.

Bowers

K. J.

Johnson

S. D.

(2005). Domestic burglary repeats and space-time clusters: The dimensions of risk. European Journal of Criminology, 2(1), 67–92.

17.

Brezina

Piquero

(2003). Exploring the relationship between social and non-social reinforcement. In Akers

R. L. A.

Jensen

G.F.

(Eds.), Advances in criminological theory, social learning and the explanation of crime (pp. 265–288). Transaction Publishers.

18.

Cao

L. Q.

(2007). Returning to normality: Anomie and crime in China. International Journal of Offender Therapy and Comparative Criminology, 51(1), 40–51.

19.

Cao

L. Q.

Zhao

van Dine

(1997). Prison disciplinary tickets: A test of the deprivation and importation models. Journal of Criminal Justice, 25(2), 103–113.

20.

Chen

(2002). Social control in China: Applications of labeling theory and the reintegrative shaming theory. International Journal of Offender Therapy and Comparative Criminology, 46(1), 45–63.

21.

Chen

(2004). Social and legal control in China: A comprehensive perspective. International Journal of Offender Therapy and Comparative Criminology, 48(5), 523–536.

22.

Cornish

D. B

Clarke

R. V.

(1987). Understanding crime displacement: An application of rational choice theory. Criminology, 25(4), 933–947.

23.

Coyne

M. A.

Eck

J. E.

(2015). Situational choice and crime events. Journal of Contemporary Criminal Justice, 31(3), 12–29.

24.

Decker

Wright

Logie

(1993). Perceptual deterrence among actual residential burglars: A research note. Criminology, 31(1), 135–147.

25.

Dhami

M. K.

Ayton

Loewenstein

(2007). Adaptation to imprisonment. Criminal Justice and Behavior, 34, 1085–1110.

26.

Duwe

Johnson

B. R.

(2016). The effects of prison visits from community volunteers on offender recidivism. The Prison Journal, 96(2), 279–303.

27.

Exum

M. L.

(2002). The application and robustness of the rational choice perspective in the study of intoxicated and angry intentions to aggress. Criminology, 40(4), 933–966.

28.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford University Press.

29.

Grasmick

H. G.

Tittle

C. R.

Bursik

R. J.

Arneklev

(1993). Testing the core implication of Gottfredson and Hirschi’s general theory of crime. Journal of Research in Crime and Delinquency, 30(1), 5–29.

30.

Jacobs

B. A.

Cherbonneau

(2017). Perceived sanction threats and projective risk sensitivity: Auto theft, carjacking, and the channeling effect. Justice Quarterly, 35(2), 191–222.

31.

Jiang

Lambert

E. G.

(2009). Views of formal and informal crime control and their correlates in China. International Criminal Justice Review, 19(1), 5–24.

32.

Jiao

A. Y.

(2001). Police and culture: A comparison between China and the United States. Police Quarterly, 4(2), 156–185.

33.

Lai

Y. L.

Zhao

Longmire

D. R.

(2011). Specific crime-fear linkage: The effect of actual burglary incidents reported to the police on residents’ fear of burglary. Journal of Crime and Justice, 34(1), 1–22.

34.

Loughran

T. A.

Paternoster

Chalfin

Wilson

(2016). Can rational choice be considered a general theory of crime? Evidence from individual-level panel data. Criminology, 54(1), 86–112.

35.

Matsueda

R. L.

(2013). Rational choice research in criminology: A multi-level frame work. In Wittek

Snijders

Nee

(Eds.), Handbook of rational choice social research (pp. 283–321). Stanford University Press.

36.

Matsueda

R. L.

Kreager

D. A.

Huizinga

(2006). Deterring delinquents: A rational choice model of theft and violence. American Sociological Review, 71(1), 95–122.

37.

Mehlkop

Graeff

(2010). Modelling a rational choice theory of criminal action: Subjective expected utilities, norms, and interactions. Rationality and Society, 22, 189–222.

38.

McCarthy

(2002). New economics of sociological criminology. Annual Review of Sociology, 28, 417–442.

39.

Moore

Shepherd

(2007). The elements and prevalence of fear. The British Journal of Criminology, 47(1), 154–162.

40.

Nagin

D. S.

Paternoster

(1993). Enduring individual differences and rational choice theories of crime. Law and Society Review, 27(3), 467–496.

41.

Nagin

D. S.

Paternoster

(1994). Personal capital and social control: The deterrence implications of a theory of individual differences in criminal offending. Criminology, 32(4), 581–606.

42.

Nagin

D. S.

Pogarsky

(2001). Personal capital and social control: The deterrence implications of a theory of individual differences in criminal offending. Criminology, 39(4), 865–891.

43.

National Bureau of Statistics of China. (2011). 2011 China Statistical Yearbook. China Statistics Press.

44.

Piquero

A. R.

Tibbetts

S. G.

(1996). Specifying the direct and indirect effects of low self-control and situational factors in offenders’ decision making: Toward a more complete model of rational offending. Justice Quarterly, 13(3), 481–510.

45.

Pogarsky

(2002). Identifying “deterrable” offenders: Implication for research on deterrence. Justice Quarterly, 19(3), 431–452.

46.

Pogarsky

(2004). Projected offending and contemporaneous rule-violation: Implications for heterotypic continuity. Criminology, 42(1), 111–136.

47.

Pratt

T. C.

Cullen

F. T.

(2000). The empirical status of Gottfredson and Hirschi’s general theory of crime: A meta-analysis. Criminology, 38(3), 931–964.

48.

Rebellon

C. J.

Piquero

N. L.

Piquero

A. R.

Tibbetts

S. G.

(2010). Anticipated shaming and criminal offending. Journal of Criminal Justice, 38(5), 988–997.

49.

Ren

Zhao

Zhang

(2017). Self-control, risky lifestyles, and victimization: A study with a sample of Chinese school youth. Criminal Justice & Behavior, 44(5), 695–716.

50.

Ren

Zhang

Zhao

(2016). Delinquent subculture, traditional models and juvenile offenders’ attitudes toward the police in China. Police Quarterly, 19(1), 87–110.

51.

Sanders

A. N.

Kuhns

J. B.

Blevin

K. R.

(2017). Exploring and understanding differences between deliberate and impulsive male and female burglars. Crime and Delinquency, 63(12), 1547–1571.

52.

Seipel

Eifler

(2010). Opportunities, rational choice, and self-control: On the interaction of person and situation in a General Theory of Crime. Crime and Delinquency, 56(2), 167–197.

53.

Shen

Hall

(2015). The same the whole world over? A review essay on youth offending from the 1980s and youth justice in contemporary China. International Journal of Law, Crime and Justice, 43(2), 273–291.

54.

Simon

H. A.

(1955). A behavioral model of rational choice. Quarterly Journal of Economics, 69(1), 99–118.

55.

Sutherland

E. H.

(1939). Principles of criminology. Lippincott.

56.

Sutherland

E. H.

Cressey

(1966). Principles of criminology. Lippincott.

57.

Townsley

Homel

Chaseling

(2003). Infectious burglaries: A test of the near repeat hypothesis. British Journal of Criminology, 43(3), 615–633.

58.

Wang

Zhao

Zhang

(2020). The impact of two different cultures on juveniles attitudes toward police in China. International Journal of Offender Therapy and Comparative Criminology, 64(1), 124–143.

59.

Wilcox Rountree

. (1998). A reexamination of the crime-fear linkage. Journal of Research in Crime and Delinquency, 35(3), 341–372.

60.

Zhang

Zhao

Ren

Zhao

(2017). Subculture, gang involvement, and delinquency: A study of incarcerated youth in China. Justice Quarterly, 34(6), 952–977.

61.

Zhang

Zhao

Ren

(2014). Social attachment and juvenile attitudes toward the police in China: Bridging Eastern and Western wisdom. Journal of Research in Crime and Delinquency, 51(6), 703–734.

62.

Zhang

(2008). Juvenile delinquency and justice in contemporary China: A critical review of the literature over 15 years. Crime, Law and Social Change: An Interdisciplinary Journal, 50(3), 149–160.

63.

Zhang

Zhao

Ren

Hoover

(2015). Space-time clustering of crime events and neighborhood characteristics in Houston. Criminal Justice Review, 40, 340–360.