Hirschi’s Redefined Self-Control

Abstract

The merger of Hirschi’s social bonding and Gottfredson and Hirschi’s self-control theories has resulted in a recent redefinition of self-control as the “tendency to consider the full range of potential costs of a particular act.” The present study clarifies the implications of Hirschi’s redefinition, advances a new measure of redefined self-control, and provides an empirical test of key hypotheses using data from a Midwestern sample of adolescents. Results indicate that the alternative measure of redefined self-control has predictive validity. Although redefined self-control and social bonds are not the same thing, they are moderately correlated. Net of controls, redefined self-control has a significant direct effect on marijuana use and partially mediates the effect of social bonds.

Keywords

redefined self-control Hirschi marijuana use social bonds

Perhaps pressed by scholars seeking to understand the relationship between social- and self-control theories (e.g., see Akers, 1991; Smith, 1995; Taylor, 2001), Hirschi (2004) made an intriguing contribution to the control literature by merging his two control theories. This union has forced difficult choices between competing ideas. Recanting his earlier argument that social bonds are fundamentally fluid in nature (Hirschi, 1969), Hirschi (2004) now contends that they are relatively stable over the life-course. Specifically, he claimed that “differences in social control are stable, that social control and self-control are the same thing” (Hirschi, 2004, p. 543). Furthermore, he indicates that all potential costs, not just long-term costs, are important to self-control. His new conceptualization of self-control is best summarized as follows:

Redefined, self-control becomes the tendency to consider the full range of potential costs of a particular act. This moves the focus from the long-term implications of the act to its broader and often contemporaneous implications. With this new definition, we need not impute knowledge of distant outcomes to persons in no position to possess such information . . . Put another way, self-control is the set of inhibitions one carries with one wherever one happens to go. Their character may be initially described by going to the elements of the bond identified by social control theory. (Hirschi, 2004, p. 543, italics in original)

Although Hirschi (2004) explicitly made the assertion that “self-control and social control are the same thing” (p. 543), he stated elsewhere that social bonds are now thought to be the source of redefined self-control. This latter statement implies that having strong bonds should cause an individual to refrain from engaging in criminal behavior precisely because these bonds lead to the tendency for an individual to consider the possible contemporaneous and long-term social and nonsocial costs of an act (Bouffard & Rice, 2011).

Devoting scholarly attention to understanding and testing the revised self-control is important because redefined self-control may explain why individuals refrain from crime while simultaneously solving several of the key criticisms of the general theory of crime (Akers, 2008, 2009; Akers & Sellers, 2009; Piquero, 2008; Piquero & Bouffard, 2007). Disentangling the exact relationship between social control and redefined self-control is of critical importance to examining the validity of Hirschi’s new statements (Bouffard & Rice, 2011), as this has direct implications for whether the theoretical merger is even appropriate in the first place. Prior to subjecting Hirschi’s revised self-control theory to empirical tests, careful consideration of the meaning of redefined self-control and how best to measure it should be a primary—not secondary—consideration. In explicating how his revised construct should be measured, Hirschi (2004) argued that the factors that inhibit one from engaging in delinquency vary along two dimensions: number and salience.¹ Furthermore, he suggests that these factors “cannot be latent, hidden, or unknown to the actor; nor can they be prior criminal or delinquent acts” (Hirschi, 2004, p. 545).

The present study begins by briefly discussing how the redefinition eliminates key problems of the original concept of self-control and subsequently reviews the limited empirical research on the topic. Following this, the study provides an in-depth discussion of the problems and limitations of measurement strategies and sorts out the meaning of the number and salience dimensions of redefined self-control that ultimately leads to the advancement of an alternative measure of redefined self-control. Finally, a test of the core ideas of the theory using data from a Midwestern sample of adolescents is conducted.

Implications of Hirschi’s Redefinition

Although the jury is still out on the best way to measure Gottfredson and Hirschi’s original concept so that the measure is valid, reliable, and independent of criminal behavior (Marcus, 2004; Piquero, 2008; Ward, Gibson, Boman, & Leite, 2010), a key theoretical advantage to Hirschi’s (2004) redefinition of self-control is that it is no longer privy to the problem of empirical tautology (Akers & Sellers, 2009). Unlike the old definition that identifies a behavioral tendency, a central element of the revised notion is the identification of a cognitive tendency to consider potential costs of a particular act, which is conceptually distinct from the behavioral outcome itself.² That is, individuals with low self-control fail to carry around with them a set of inhibitions meaning they do not consider the short- and long-term consequences of an act; as a consequence, these individuals who are low in self-regulation will engage in delinquent behavior. Interestingly, Hirschi (2004) insists that redefining self-control via social-control elements “should not be construed as an effort to deal with the tautology issue” (p. 550). Regardless, his redefinition results in a construct of self-control that can now be measured without reference to one’s actual behaviors.

Beyond providing an unintended resolution to the tautology problem inherent in the original self-control theory, the inclusion of short-term costs is an additional logical step toward improving the theory (Piquero, 2008; Piquero & Bouffard, 2007). The consideration of the costs of a particular act—whether they are short term and/or long term—seems to be important to the idea of control (see Hirschi, 1969). Although long-term legal costs such as imprisonment may be among the more serious consequences and, therefore, be largely influential in one’s decision making, short-term costs such as parent’s or friend’s reactions also play a significant role in influencing one’s decision to engage in delinquent behavior. That is, although arguably less serious, short-term costs of delinquency are also common considerations among rational (or semirational) actors.

Hirschi’s (2004) redefinition leaves us with a number of old and new hypotheses ripe for empirical research. Perhaps the most controversial of the new claims is that “social control and self-control are the same thing” (Hirschi, 2004, p. 543). The importance of this claim should not be overlooked. If social control is weakly related to redefined self-control, then this would call into question the theoretical integration of the two theories and instead suggest that the two theories would be better left as stand-alones. However, if support is found for Hirschi’s claims, then the confusion concerning the disconnection between the two control theories would be resolved.

To empirically test this hypothesis, one needs to measure redefined self-control by tapping into the number of costs for each particular act and the importance, or salience, of such costs (see Hirschi, 2004). In addition, one needs to also measure the inhibiting factors identified by Hirschi or, in other words, the traditional social bonds, including attachments, commitments, involvements, and beliefs. The merger of the two theories suggests that social bonds and redefined self-control are essentially one in the same, which can be assessed with a simple correlation. Net of controls, redefined self-control should be inversely related to crime and delinquency as originally conceived (Hirschi, 2004). Furthermore, the theory anticipates that any effects of social control on delinquency should be indirect, entirely mediated by redefined self-control (Bouffard & Rice, 2011).

Beyond this, little has effectively changed with the general predictions of the original self-control theory. As Hirschi (2004) put it, “Change in conception of the sources of self-control and the cognitive processes it involves should have little effect on the empirical predictions derived from the theory” (p. 548). Although a focused examination is outside the scope of the present study, the stability thesis carries over into the revised theory as well as the predictions concerning between-individual differences in redefined self-control (e.g., race and gender) and between-society differences in the construct (see Gottfredson & Hirschi, 1990).³

Despite the fact there is much known about the empirical state of the general theory of crime (see Pratt & Cullen, 2000), the redefinition of self-control via the merger of the two control traditions has essentially wiped the slate clean. Although some previous research has incorporated an attitudinal measure of trait self-control (e.g., the Grasmick, Tittle, Bursik, & Arneklev, 1993, scale) to compare the predictive validity of the general theory to the revised self-control theory (e.g., see Higgins, Wolfe, & Marcum, 2008; Piquero & Bouffard, 2007), our study takes a different approach and focuses exclusively on the revised theory, thereby permitting a more complete unpacking of the details of the revised theory and, as a result, leading to a more in-depth empirical investigation into Hirschi’s (2004) core ideas. We briefly review the very limited empirical research that has set out to examine the revised self-control theory before discussing lingering concerns regarding the conceptualization, operationalization, and measurement of redefined self-control, and we ultimately progress toward the advancement of an alternative measure believed to be more in line with Hirschi’s (2004) redefinition.

Initial Tests of the Revised Theory

In an initial attempt to assess the merits of his revised theory, Hirschi (2004) measured self-control via a reconfiguration of social bonding items. Using data from the Richmond Youth Study, he dichotomized nine social bonding items and specified a response category (or two) as the mark of high self-control (e.g., “Do you like or dislike school?” [like it], “Does your mother know where you are when you are away from home?” [Usually]). In a bivariate analysis, he found that 73% of individuals reported having committed two or more delinquent acts when they had zero inhibiting factors, whereas only 2% reported engaging in these same behaviors when they had all nine inhibiting factors. There was a systematic decline in the percentage of individuals engaging in two or more delinquent acts committed as the number of self-reported inhibiting factors increased, and this general pattern of findings also held with an alternative data set. Unfortunately, Hirschi’s data did not have measures of the costs of delinquency, let alone a full range of costs, providing little more than a test of the principal ideas underlying social bonding theory. In other words, the self-control measure developed by Hirschi is not consistent with his redefinition (Piquero & Bouffard, 2007). Although a reasonable starting point for testing his revised theory, Hirschi (2004) assumed, as opposed to demonstrated, that those who are strongly bonded refrain from criminal behavior due to a greater tendency to consider the full range of costs of an act. This assumption requires empirical verification.

In addition to Hirschi’s approach, there have been four other tests of the revised theory (see Bouffard & Rice, 2011; Higgins, Wolfe, et al., 2008; Morris, Gerber, & Menard, 2011; Piquero & Bouffard, 2007). In the first independent empirical assessment, Piquero and Bouffard (2007) examined whether redefined self-control was more predictive of crime than a measure of social bonds and the Grasmick et al.’s (1993) trait self-control measure. They captured the number of costs by allowing survey respondents to self-identify up to seven costs after reading a vignette (Piquero & Bouffard, 2007). To measure the salience component, they asked individuals to rank, on a scale from 0 to 100, how important each cost was to his or her decision making. All of these individual salience scores associated with their respective costs were then averaged across the total number of costs identified by the respondent. The theory implies that when the total number of costs considered and the salience attributed to them are both high, self-control is high (Hirschi, 2004). Thus, to create the final self-control measure, Piquero and Bouffard multiplied the total number of costs identified by the respondent by the average salience attributed to all costs. Using vignettes and a college sample of approximately 200 students, Piquero and Bouffard predicted drunk driving and sexual coercion intentions, measured by self-reports of one’s perceived percentage of certainty that he or she would commit the acts in question. Net of controls, the researchers concluded that the redefined self-control had the strongest influence on both drunk driving and sexual coercion intentions among the different measures. In fact, Grasmick et al.’s measure of self-control was no longer statistically significant for either outcome when the redefined self-control measure was in the model.

Higgins, Wolfe, et al. (2008) assessed the extent to which the Piquero and Bouffard self-control measure predicted music piracy. They also incorporated a seven-item scale of social-control items that measured commitment to school and attachment to parents that was very similar to the measure used by Hirschi. Furthermore, they included Grasmick et al.’s measure of trait self-control. They found that all three “self-control” scales predicted music piracy with relatively similar standardized effects, while controlling for the influences of the alternative self-control measures and several control variables.⁴

Using data from the National Youth Survey Family Study (NYSFS), Morris et al. (2011) used a measure of redefined self-control inspired more so by Hirschi’s (2004) original index than by subsequent measurement efforts (e.g., Bouffard & Rice, 2011; Higgins, Wolfe, et al., 2008; Piquero & Bouffard, 2007). In constructing their measure of redefined self-control, they first identified whether each of six bonds (e.g., spouse/parent, children, friends, career, community activities, and religiosity) was present and calculated the corresponding salience of those bonds that were present. For each bond, they computed the product of the bond and salience. They subtracted the average salience from each of these products, and then they summed these quantities across the six bonds. Controlling for basic demographics and trait self-control, redefined self-control exhibited a statistically significant effect on adult offending in both cross-sectional and longitudinal analyses as modeled by both logistic and negative binomial count models. However, in longitudinal models that added prior offending as a control variable, the effect of redefined self-control was rendered marginally significant for offending prevalence and nonsignificant for offense count.

In the only empirical test to date that has explicitly addressed the causal relationship between social bonds, redefined self-control, and delinquency, Bouffard and Rice (2011) used a path analysis and found that social bonds did not have a direct relationship with drunk driving expectations. Their measure of redefined self-control was computed as the average salience of the costs that were identified; they did not incorporate information on the number of costs for reasons that will be subsequently discussed. Their path model provides support for an indirect link (though not a mediated effect) between social bonds and drunk driving expectations through their measure of redefined self-control.

In sum, one study suggests that social bonds have an indirect effect on one form of delinquency through redefined self-control, but further research is needed (Bouffard & Rice, 2011). Collectively, the limited empirical literature provides initial evidence that redefined self-control and social control (i.e., social bonds) are not similar concepts. The constructs, as measured in existing research, appear to be starkly different as the correlations between them are quite low (r = .05, ns, Higgins, Wolfe, & Marcum, 2008; r = .14, p < .05 for drunk driving and r = .01, ns for sexual assault, Piquero & Bouffard, 2007; r = .203, p < .05; Bouffard & Rice, 2011). These weak or nonsignificant relationships might be explained by the operationalizations of redefined self-control, which fail to include one of the two dimensions of redefined self-control and/or measure the salience of a cost to the individual’s decision making generally (but not the salience of “others” opinions to decision making specifically). Concerns with these approaches are now elucidated, which call into question the little that is known about the empirical validity of redefined self-control theory.

Toward a Valid Measure of Redefined Self-Control

Recall, the Piquero and Bouffard’s (2007) measurement strategy of redefined self-control involves obtaining a list of up to seven self-generated costs (number) along with the corresponding importance attributed to each of these costs (salience), averaging the salience scores, and multiplying this average salience score by the total number of costs that were identified. While the Piquero and Bouffard’s measure seems to have good predictive validity, this approach has been the subject of criticism—from one of the measure’s original authors no less. Drawing on the idea that reporting costs might be partially influenced by trait self-control (see Piquero, MacIntosh, & Hickman, 2000), researchers suggest self-generating costs on a survey could reflect either one’s tendency to consider a fewer number of costs or possibly one’s failure to write down a cost that he or she actually did consider (and believed not to be a real cost), the latter of which is thought to be a “response bias” (Bouffard & Rice, 2011, p. 144). In short, the Piquero and Bouffard’s measure confounds trait self-control and redefined self-control.⁵

This observation led Bouffard and Rice (2011) to drop information about the number of costs identified and instead use an average salience measure, which “retain[s] a [italics added] central aspect of Hirschi’s decisional self-control, while eliminating a potential confound in the measure previously used” (p. 146). The elimination of this confounding issue, which arguably improves the scale in this one sense, unfortunately introduces a construct validity problem as one of the two dimensions of redefined self-control identified by Hirschi is omitted. Put differently, this alternative measure may address a potential problem with the measure proposed by Piquero and Bouffard but comes at the expense of failing to capture both dimensions of redefined self-control.

Recently, Morris et al. (2011) explicitly incorporated both the number and salience dimensions of Hirschi’s redefinition. Guided by Hirschi’s (2004) measure of redefined self-control as the presence of strong social bonds, but also incorporating salience, Morris and colleagues (2011) argued their formula results in an improved measure that permits “individual variation in the importance of a particular bond, relative to the respondent’s mean bond salience level” (p. 590). The allowance for individual variation in salience is a noteworthy contribution. However, this self-control measure is subject to the same criticism as Hirschi’s measure; specifically, an essential piece of redefined self-control (i.e., costs) is not included.

Previous research leaves us with four slightly different measures of redefined self-control, none of which is ideal. Some of these measures have construct validity problems, and others have basic concerns with face validity. What is clear is that the details of the salience and number components of revised self-control theory have not been fully scrutinized, though there is important need to do so (Bouffard & Rice, 2011). In moving toward a measure of redefined self-control that incorporates both dimensions of redefined self-control identified by Hirschi, several lingering issues first warrant further consideration.

Likely brought on by Hirschi’s limited guidance on how to best measure it, the conceptualization of the salience dimension in prior research has been problematic. The difficulty involves the attribution of a general salience score to a specific cost. For instance, asking individuals “how important each one of these [self-generated costs] would be when making your decision whether or not to drive home under the circumstances in the story” (see Piquero & Bouffard, 2007; see also Bouffard & Rice, 2011) measures the general salience of a cost to an individual. With this approach, one has to make a conceptual leap that the attributed salience to the cost has to do with one’s concern for others. To this point, Hirschi (2004) clearly argued that the “principal source of control is concern for others. ‘Others’ come in a variety of shapes and sizes: parents, teachers, friends, and policemen” (p. 545). For example, under Piquero and Bouffard’s drunk driving scenario, it is plausible that some persons might identify injury or death as a cost of drunk driving and attribute a high salience to it. However, when asking about the general importance of this cost to one’s decision, one cannot be sure the extent to which others were involved in one’s decision. If the salience dimension happens to be entirely unrelated to a concern for others, the self-control measure would lack construct validity. As a result, a statistically significant relationship between the redefined self-control measure and delinquent behavior in this operationalization might not constitute fair support for the theory.

Concern for others’ opinions when making decisions is conceptually distinct from, though related to, one’s level of social bonds to persons and society. The theory suggests that strong social bonds in the form of attachment, commitment, involvement, and belief ultimately result in other persons being salient to the individual when a decision has to be made regarding whether to engage in delinquent behavior. For instance, in discussing his concept of attachment, Hirschi (1969) claimed,

The important consideration is whether the parent is psychologically present when temptation to commit a crime appears. If, in the situation of temptation, no thought is given to parental reaction, the child is to this extent free to commit the act. (p. 88)

Rather than infer the salience of others from questions pertaining to the strength of the social bond as Hirschi did in his 1969 version of control theory (and again in his 2004 version), the merger of the two control theories necessarily suggests we actually measure whether others are relevant in one’s decision making when temptation to engage in delinquency strikes. Thus, an alternative measurement approach, which appeals directly to Hirschi’s (2004) emphasis on the social aspects of cost consideration, is to capture how important parents, family, friends, and the law (i.e., others) are in helping one make a decision whether to engage in a certain behavior (as opposed to how salient or important the cost is to the individual as that can have many sources distinct from the consideration of others). With a measurement strategy for the salience dimension of redefined self-control, attention is now turned toward measuring and incorporating the number dimension into the construct.

In arguing against the inclusion of the number of costs identified in their measure, Bouffard and Rice (2011) drew a key distinction between facing costs (i.e., the number of costs of committing an act that an individual faces) and considering costs (i.e., the likelihood that an individual will think about the costs that he or she actually faces). They use this distinction to help explain certain difficulties that arise when using the self-generated cost methodology. Specifically, they argue that self-generating a lower number of costs could be the result of truly considering fewer costs (the expected effect), failing to report costs that one actually considers possibly due to low-trait self-control (a response bias), or failing to report costs because one simply does not face them (an effect they argue is not necessarily consistent with the theory).⁶ One interpretation of Hirschi is that “one’s level of self-control (measured within the decision) was reflected in the fact that these individuals ‘failed to consider the full range of consequences’, not that individuals with low self-control necessarily faced fewer bonds (costs)” (Bouffard & Rice, 2011, p. 139). Although we understand why this conclusion may be drawn given Hirschi (2004) does state that “self-control becomes the tendency to consider the full range of potential costs of a particular act” (p. 543), additional information from the theory unambiguously advocates that the number of costs actually faced is an integral component of redefined self-control. Measuring only the salience dimension and omitting information on the number of costs that one actually faces runs counter to Hirschi’s beliefs.

Hirschi (2004) argued that the redefinition of self-control is entirely consistent with Gottfredson and Hirschi’s original idea that the various dimensions of self-control can be thought of as “factors affecting calculation of the consequences of one’s act” (Gottfredson & Hirschi, 1990, p. 95). How specifically is the calculation of the consequences affected by self-control? Gottfredson and Hirschi (1990) answered this question by stating,

The impulsive or short-sighted person fails to consider the negative or painful consequences of his acts; the insensitive person has fewer negative consequences to consider; the less intelligent person also has fewer negative consequences to consider (has less to lose). (p. 95)

Thus, those with low self-control both fail to consider costs and have fewer costs to consider in the first place.

Hirschi (1986) sees rational choice theory and social control as sharing important common ground, albeit the former is thought to be a theory of criminal events and the latter of criminality (see also Akers, 1990). In advancing the general theory of crime, Hirschi claims criminologists should not overlook the importance of basic assumptions of classical criminology (Gottfredson & Hirschi, 1990); unless controlled, people will act on innate desires to engage in self-serving behavior such as crime. Consistent throughout his versions of control theory, he maintains that “people add up in an imprecise way the negative consequences of deviant acts and behave accordingly” (Hirschi, 2004, p. 546). Hence, his emphasis on calculation is clearly carried over into the revised theory. In sum, Hirschi explicitly states that redefined self-control manifests itself in both cost consideration and the number of consequences one actually faces. Omitting information on the number of costs faced is inconsistent with Hirschi as he claims that inhibitions vary in both “number and salience.”

Hirschi (2004) states that “self-control becomes the tendency to consider the full range of potential costs of a particular act” and also that “self-control is the set of inhibitions one carries with one wherever one happens to go” (p. 543). Linking these ideas, self-control is both a “tendency to consider” and a “set of inhibitions,” which neatly corresponds to the salience (i.e., tendency to consider costs that one faces) and number (i.e., number of costs faced) dimensions of redefined self-control as discussed herein, respectively. In short, attending to the core ideas laid out by Hirschi’s (2004) redefinition leads to the construction of a measure that incorporates human rationality (number of costs faced) and the clearly important consideration of others’ opinions (tendency to consider costs).

Attending to the details of his revised theory, it has been argued that both the number and salience dimensions are critical components of redefined self-control. Taking this to be true, it remains unclear whether these two components should operate individually or whether these should be used as constituent pieces of an amalgamated measure (see Piquero & Bouffard, 2007). The general predictions of the new self-control theory (i.e., those with higher redefined self-control will engage in lower levels of delinquency) help to clarify this issue. Presumably, self-control manifests itself when one faces costs (number) and has a tendency to consider the costs that one faces (salience). To make this more concrete, consider three different scenarios. First, if one faces negligible costs of a particular act, it does not matter much whether others play a very important role in his or her decision making. The individual has tended to consider the costs, has considered them, but nevertheless faces few (if any) real costs. The behavior is unlikely to be prevented. Second, if others do not play an important role in one’s decision making, then they may be less likely to stop and think about the real-life costs that they do face, even though they might identify a high number of them if they did actually stop to consider costs. In this case, the behavior is also unlikely to be prevented. Finally, if an individual has a tendency to consider the costs of an act (i.e., parents, family, friends, and the law are important to one’s decision) and the individual faces a large number of costs, then delinquency should be controlled in this case. That said, these two dimensions should covary considerably. Thus, individuals should tend to face more costs and tend to consider others’ opinions when they have stronger social bonds. We now discuss the data and measures used to test key hypotheses drawn from the theory.

Data and Method

The present study uses data from the Boys Town Study, run by the Center for the Study of Youth Development in Boys Town, Nebraska (see Akers, Krohn, Lanza-Kaduce, & Radosevich, 1979). The original sample is comprised of 3,065 7th- to 12th-grade adolescents, and the data were collected in seven communities across three Midwestern states. Listwise deletion was used to obtain the final analysis sample (N = 2,243). Descriptive statistics of the sample and measures are reported in Table 1.

Table 1.

Descriptive Statistics (N = 2,243).

	M	SD	Minimum	Maximum
Age	15.383	1.684	10	19
Sex	0.585	0.493	0	1
Race	0.034	0.183	0	1
Religiosity	1.666	0.940	0	3
Delinquent peers	0.873	1.000	0	3
Social bonds	4.897	2.718	0	12
Marijuana use	1.133	1.638	0	5
Redefined self-control	11.178	8.370	0	36
Costs	5.921	2.646	0	12
Salience	1.706	0.818	0	3

Measures

Dependent Variable

Marijuana use

Self-reported marijuana use frequency serves as the outcome measure for the present study. The respondent was asked to identify how often he or she had used marijuana, measured on a Likert-type scale with six response options—“never” (0), “once or twice,” “less than once a month,” “once or twice a month,” “once or twice a week,” and “nearly every day” (5).

Independent Variables

Social bonds (Hirschi’s self-control measure)

Following Hirschi’s construction of a composite index of bonding factors, this measure includes 12 total items with 6 items capturing attachment, 3 measuring belief, and 3 others tapping into commitment. Although we do not measure involvement directly, previous research suggests that the 2 items of commitment and involvement are similar empirically because involvement is seen as the longitudinal dimension of commitment (see Krohn & Massey, 1980). Attachment items asked respondents (a) whether they were close to their mother, (b) whether they were close to their father, (c) whether their mother understands them, (d) whether their father understands them, (e) whether their mother knows their whereabouts, and (f) whether their father knows their whereabouts. Commitment items asked respondents (g) how important schoolwork was to them, (h) how important getting a good job is, and (i) how often they skipped school without an excuse. Belief items asked respondents their level of agreement regarding (j) whether they felt that everyone had a moral duty to abide by the law, (k) whether their parents’ morals were okay for them, and (l) whether they thought that using drugs was wrong because using was illegal. The measure is quite similar to Hirschi’s (2004) proposed 9-item measure. Hirschi’s emphasis on attachment items is echoed in the current measure, and the items were similarly dichotomized here so that a score of 1 on an item was the mark of high social bonds.

Redefined self-control

Drawing from the ideas of rational choice, self-control, and social bonding, a crucial aspect to Hirschi’s redefinition of self-control is that individuals who face a full range of costs (number) and have the tendency to consider them (salience) will have high self-control. The number of perceived cost items in this study is specific to the act of marijuana use and measures both short- and long-term costs. Instead of asking respondents to write in self-generated costs as in prior research (e.g., Piquero & Bouffard, 2007), our cost categories are predetermined so that the respondent can choose whether he or she perceives facing each specific cost. We use 11 cost items that vary considerably and form a reasonable “full range” of costs. The cost items include guilt, effects on grades, health, potential for habit-forming behavior, drug effects (a bad high), negative parental reaction, negative friend reaction, likelihood that parents would catch you, likelihood that police would catch you, celerity of punishment, and whether the individual would face a legal sanction. The number of yes/no costs faced by each respondent was summed to create the costs dimension of the redefined self-control measure.

The salience dimension is captured by asking respondents, “In deciding to use or not to use marijuana, how important is each of the following in helping you decide?” The others comprising the salience dimension included parents, others in your family, friends, others about your age (not family), others older (not family), and the law. Response options included “not important at all,” “somewhat important,” “important,” and “very important.” Responses to the individual salience items were then averaged. Finally, with the number and salience dimensions independently created, the redefined self-control measure was calculated by multiplying the total number of costs that were faced by the average salience that is attributed to others.

Control Variables

Demographics

Age, gender, and race are controlled for in analyses. Age is measured in years. Gender is coded so that zero indicates male and one indicates female, and race is coded so that zero indicates White and one indicates non-White.

Delinquent peers

Because peer delinquency consistently exerts an effect on behavior (Warr, 2002), we include a measure of delinquent peers to protect against spuriousness. The delinquent peers construct is measured using a three-item scale tapping associations with peers who use marijuana. Individuals were asked how many of their best friends, the friends with whom they associate most often, and the friends with whom they have associated for the longest time use marijuana. Response options included “none, or almost none,” “less than half,” “more than half,” and “almost all.” The items were summed and divided by the total number of items; higher scores indicate greater association with delinquent peers.

Religiosity

Religiosity is measured with a single item asking individuals, “How religious of a person are you?” Responses were coded one through five with the corresponding response options, including “not religious at all,” “a little religious,” “more than a little religious,” “quite religious,” and “very religious.”

Analytic Strategy

Using a measure that is consistent with Hirschi’s redefinition is a prerequisite for examining the theoretical statements of the revised self-control theory. Therefore, several analyses first examine whether the alternative measure of redefined self-control that we contend has face validity appears to also have predictive validity. Both the number and salience dimensions of redefined self-control are standardized and categorized into low, medium, and high groups, creating nine combinations of number of costs and salience (e.g., low cost–low salience, low cost–medium salience). In each of these combinations, marijuana smoking prevalence and mean social bond levels are examined. With the predictive validity examined, the measure of redefined self-control is then used to assess Hirschi’s hypotheses that social control and redefined self-control are the same thing, that social bonds are indeed the source of redefined self-control, and that the effect of social bonds operates through redefined self-control. Accordingly, a simple ordinary least squares (OLS) regression model with robust standard errors is used to estimate the claim of sameness, and a multiple OLS regression model with robust standard errors is used to determine whether one’s social bond level is the source of redefined self-control. Finally, a series of ordered logistic regression models are then estimated to examine the mediation hypothesis that the effect of social bonds on marijuana use operates exclusively through redefined self-control. Empirical findings from supplemental path decomposition models are also presented.

Results

Table 2 examines the predictive validity of the redefined self-control measure on marijuana use. The number of costs faced and average salience scores were standardized and then categorized into approximately equal groups representing low, medium, and high categories. As expected, the proportion of individuals who experience high costs and high salience (i.e., high self-control) are the least likely to have used marijuana (8.7%). At the other end of the continuum, those with low costs and low salience have the highest percentage of respondents having used marijuana (80.9%). Consistent with expectations, when the number of costs faced is low, marijuana use is still fairly likely regardless of whether individuals tend to take others opinions into account. Hence, even when salience is high, 52.6% of respondents who report low costs report using marijuana. Of course, as salience is higher, there is still a general downward trend in the percentage of individuals using marijuana. Somewhat unexpected, however, is the fact that when the salience is low, marijuana use nevertheless becomes somewhat unlikely (26.2%) when individuals report facing a high number of short- and long-term costs of the behavior. Stated differently, when the number of costs of a behavior is especially high, it has a strong influence on behavior even when individuals report they tend not to consider parents, friends, and the law at the time a decision is made. Still, those who face a high number of costs will receive additional inhibition as the salience dimension increases. Overall, the more costs faced and the greater the emphasis placed on others when making decisions (i.e., the higher one’s self-control), the less likely one is to smoke marijuana.

Table 2.

Marijuana Use Means Across High/Medium/Low Salience and Costs Combinations.

	Costs
Salience	Low	Medium	High	Row marginal
Low	0.809 (472)	0.556 (279)	0.262 (84)	0.669 (835)
Medium	0.665 (188)	0.355 (290)	0.145 (179)	0.387 (657)
High	0.526 (76)	0.222 (306)	0.087 (369)	0.186 (751)
Column marginal	0.743 (736)	0.372 (875)	0.126 (632)	0.425 (2,243)

Note: Proportion (number).

Table 3 illustrates that those with the lowest self-control (i.e., low salience and costs) have the lowest average social bonds (3.182), whereas those with the highest self-control (i.e., high salience and costs) have the highest average social bonds (6.277). Consistent with the theory, higher social bonding average scores are found for those individuals who face more costs and tend to consider others’ opinions when making decisions. Taken together, Tables 2 and 3 provide some support for the importance of measuring both the number and salience dimensions of redefined self-control. The correlation between the number and salience dimensions of redefined self-control is moderate (r = .496, p < .001).

Table 3.

Social Bonding Means Across High/Medium/Low Salience and Costs Combinations.

	Costs
Salience	Low	Medium	High	Row marginal
Low	3.182 (472)	3.677 (279)	4.976 (84)	3.528 (835)
Medium	4.234 (188)	5.090 (290)	5.989 (179)	5.090 (657)
High	5.736 (76)	5.824 (306)	6.713 (369)	6.251 (751)
Column marginal	3.715 (736)	4.896 (875)	6.277 (632)	4.897 (2,243)

To assess the claim of sameness between self-control and social-control theories, we estimated a simple OLS regression (see Table 4). To account for clustering of individuals within schools, we report robust standard errors derived from the Huber–White cluster sandwich estimator for all models (Huber, 1967; White, 1980). Social bonds and redefined self-control are significantly associated with one another. Specifically, for a one unit increase in social bonds, redefined self-control increases by 1.445 units (p < .001). The standardized coefficient, which is the correlation in this case, is equal to .469.⁷ This correlation between redefined self-control and Hirschi’s (2004) measure is considerably greater than has been found in prior research. Still, this would hardly suggest that two constructs are one in the same; indeed, the correlation is significantly and substantially different from 1. Furthermore, Model 1 shows that social bonds fall well short of explaining all the variation in redefined self-control. Specifically, social bonds explain approximately 22% of the variance.

Table 4.

OLS Regressions Predicting Redefined Self-control.

	Model 1			Model 2
	b	RSE	β	b	RSE	β
Constant	4.100	0.464***	—	11.864	1.992***	—
Social bonds (Hirschi’s SC)	1.445	0.081***	.469	0.908	0.059***	.295
Age				−0.252	0.132	−.051
Sex				0.720	0.269*	.042
Race				−0.982	0.653	−.021
Religiosity				0.640	0.182***	.072
Delinquent peers				−3.108	0.129***	−.372
F statistic	315.35***			338.72***
R ²	0.220			0.362

Note: OLS = ordinary least squares; RSE = robust cluster standard errors; SC = self-control.

p ≤ .05. **p ≤ .01. ***p ≤ .001.

Model 2 includes the control variables in the analysis, yielding two important findings. First, social bonds maintain a significant and substantively meaningful influence on redefined self-control (p < .001). Second, however, other factors are related to redefined self-control, including gender (p < .05), religiosity, and delinquent peers (p < .001). The combined model explains more than 36% of the variation in redefined self-control, which is considerably more than Hirschi’s measure explains alone. In sum, social bonds seem to influence redefined self-control, but other factors might also contribute to one facing more costs and considering others when making decisions. Although social bonds and self-control seem to be two distinct constructs, support for the mediation aspect of the theory is garnered if the influence of social bonds on delinquent behavior operates through redefined self-control (Bouffard & Rice, 2011). We now investigate this possibility.

Table 5 contains the results of a series of ordered logistic regression models regressing marijuana use on the constructs of interest and the control variables.⁸ Model 3 demonstrates that those with greater social bonds are less likely to use marijuana, controlling for age, sex, race, religiosity, and delinquent peer associations. This finding is not surprising and is essentially supportive of Hirschi’s (1969) original social-control argument. To assess the mediation hypotheses, Model 4 adds redefined self-control to the equation. An initial assessment is that redefined self-control exhibits a significant effect on marijuana use (p < .001). For a one unit increase in redefined self-control, the log of the odds decreases by −0.095. As this variable does not violate the parallel lines assumption, this coefficient is valid for all five comparisons across the ordered categorical variable (i.e., k > 0 . . . k > 4). The coefficient representing the effect of social bonds on marijuana is weakened by nearly 40% but remains statistically significant. This finding provides evidence consistent with partial mediation.

Table 5.

Ordered Logistic Regressions Predicting Marijuana Use.

	Model 3			Model 4
	b	RSE	e^b(stdX)	b	RSE	e^b(stdX)
Social bonds (Hirschi’s SC)	−0.164	0.030***	0.641	−0.102	0.031***	0.758
Age^a	0.162	0.028***	1.313	0.147	0.030**	1.280
Sex^a	−0.536	0.086***	0.768	−0.473	0.081***	0.792
Race	0.185	0.280	1.034	0.129	0.274	1.024
Religiosity	−0.211	0.075**	0.820	−0.162	0.073*	0.858
Delinquent peers	1.920	0.062***	6.831	1.733	0.066***	5.664
Redefined self-control				−0.095	0.008***	0.451
Thresholds
1	2.956	0.485	—	2.015	0.511	—
2	4.332	0.483	—	3.464	0.513	—
3	5.035	0.467	—	4.194	0.507	—
4	6.019	0.478	—	5.199	0.503	—
5	7.472	0.496	—	6.654	0.506	—
Wald χ²	1,461.16***				1,643.36***
Pseudo R²	.303				.323

RSE = robust cluster standard errors; e^b(stdX) = partially standardized odds ratio, which indicates the change in the odds for a standard deviation increase in X.

These variables violated the parallel lines assumption in Models 1 and 2. The five unconstrained parameter estimates for these variables for each model are available on request.

p ≤ .05. **p ≤ .01. ***p ≤ .001.

Supplemental path analysis estimated in Mplus v6 that uses OLS and probit regressions with a weighted least squares estimator (WLSMV) and robust Huber–White cluster standard errors corroborated these findings. The indirect effects represent approximately 45% of the total effect of social bonds on marijuana use. The indirect effect, direct effect, and total effects of social bonds on marijuana use are statistically significant (p < .01; see appendix). For comparison purposes, the appendix also shows results from three alternative path analysis models that use as mediators only the cost dimension, only the salience dimension, and both dimensions (multiple mediator model). It is important to note that the appendix provides further evidence that both the number and salience dimensions are relevant to redefined self-control. The indirect effects through costs only and salience only represent approximately 29% and 26% of the total effects, which are considerably less than for the models that incorporate both dimensions.

Discussion and Conclusion

The present study has explicated important theoretical and measurement considerations regarding Hirschi’s (2004) redefined self-control and provided a test of key hypotheses with an alternative measure of the construct that is argued to be consistent with Hirschi’s redefinition. Results indicated that a measure of redefined self-control that captures both the salience and number dimensions has predictive validity. In general, those who tend to consider more costs (i.e., they consider others when making decisions) and those who face more costs (i.e., they identify more things as actual costs) are less likely to use marijuana and have higher social bonds. However, one unanticipated finding regarding the functioning of the measure warrants further discussion.

Individuals who faced a high number of costs tended to not use marijuana, even if they were not inclined to consider others when making decisions. Conversely, those with high salience but low costs did not have similarly low levels of marijuana use. Specifically, approximately 26% of persons with high costs and low salience used marijuana, whereas about 53% of individuals with low costs and high salience used it (see Table 2). Interestingly, the former group had somewhat lower social bonds than the latter group (see Table 3). The theory anticipates that delinquency is controlled when one faces numerous costs, and he or she stops to think about them because of concerns for others. The general idea that behavior is constrained when one actually faces costs and tends to consider them is sound. In explaining these findings, it is possible that the tendency to consider component may stem from nonsocial sources, including, for example, religion. To be clear, religion is not purely a nonsocial factor. The institutional nature of religion necessarily results in it having a social component. However, the personal relationship with one’s deity is nonsocial. Hirschi and Stark (1969) found religion to be irrelevant to explaining delinquency, but Burkett and White (1974) noted that religion has a sizable influence on victimless offenses such as marijuana and alcohol use. If one believes in an omniscient entity or entities, by definition, there is a guarantee that the deity/deities will know of any immoral acts (at least in the mind of the person who has faith). From this logic, religion might conceivably cause one to consider the various consequences of any particular act. Interestingly, Akers (2010) suggested that religion could have been incorporated into each of the elements of social bonding theory. Beyond others in the social sense then, there might be additional sources that cause individuals to “stop and think” about the costs that they actually face. In the interest of explaining behavior, additional consideration should be given to the potential factors beyond social relationships that give the costs that are faced their salience. At the same time, Hirschi specifically emphasizes others, including parents, teachers, peers, and police officers. Thus, redefined self-control as measured herein provides an appropriate and fair test of Hirschi’s revised theory.

Overall, mixed support was garnered for the hypotheses examined in this study. Social- and self-control are significantly and substantially related, albeit they are a far cry from being the same thing. This suggests that Hirschi’s (2004) analysis linking social bonds to delinquency directly without specifically measuring self-control as he redefines it results in an inappropriate test of the theory. Thus, the assumption that redefined self-control and social bonds are the same thing is untenable. As a result, future research should probably not use Hirschi’s bonding measure or any variant when testing the direct effects of redefined self-control on delinquency.

Results indicated that social bonds influence marijuana use through our measure of redefined self-control, albeit not entirely. Approximately 45% of the total effect of social bonds on marijuana use is indirect through redefined self-control. These findings suggest that social bonds may ultimately maintain a direct influence or may also operate through other nonhypothesized mechanisms to thwart marijuana use. The fact that social control and redefined self-control are not the same thing, coupled with the finding that the latter only partially mediates the former, seems to call into question the validity of the theoretical merger of the two theories. That is, although Hirschi (2004) has made strong statements regarding the nature of the associations between social bonds, redefined self-control, and delinquency, the data fail to support such bold assertions and rather suggest there is bounded efficacy of these relationships. Still, the fact there is moderate support for the theory may provide important guidance for public policies aimed at curbing delinquent behavior. Increasing the number of costs that one faces and, to a lesser extent, increasing the salience of others in the decision-making process are likely to lower marijuana use among adolescents. One way in which the costs of marijuana smoking and the tendency to consider them can be augmented is through the development of stronger attachments and commitments to prosocial persons and organizations along with stronger beliefs in conventional society. If future research confirms that social bonds are only partially mediated by redefined self-control, additional studies will be needed to examine the various ways in which social bonds provide added benefit by controlling behavior above and beyond their effects through redefined self-control.

Although the present study has sought to clarify key theoretical and measurement issues of redefined self-control and provide a test of the main hypotheses of the theory, there are limitations that warrant consideration. First, the data used in the present study come from a Midwestern sample of adolescents that is quite homogeneous with respect to race and focuses on substance use. Concerning the demographics, approximately 1 of every 20 individuals in these data is a racial or ethnic minority. Thus, additional empirical tests are needed using more diverse samples. Furthermore, given our data are confined to the Midwest, future studies should seek to use data that are nationally representative. In addition, studies should seek to explore the effects of redefined self-control on other forms of delinquency. It is yet to be determined whether the effects of redefined self-control on different types of delinquency such as violent and property offenses are similar. Of course, given the limitations of Likert-type scales for measuring crime and delinquency (e.g., see Elliott & Ageton, 1980), it is recommended that count and frequency measures be used when available. Given that a person can conceivably face a different number of costs for different acts, future research might explore whether those who face a high number of costs for one delinquent act tend to be the same individuals who face a high number of costs across different offenses. The theory anticipates that this is the case, but it should be subjected to empirical testing. It is also an empirical possibility that others might be salient to one’s decision making for some types of offenses but not others. In short, additional research is needed to examine the generality of redefined self-control theory across distinct persons and divergent crime types.

Second, the cross-sectional nature of our data precludes an investigation into the stability thesis of redefined self-control. Furthermore, although we examined the influence of social bonds on marijuana use through redefined self-control, longitudinal data are required to determine causal order. Nagin and Paternoster (1993) argued that “the strength of social bonds are themselves determined by self-control and have no independent influence on offending behavior. Weak social bonds are thus not a cause of crime but just another manifestation of low self-control” (p. 472, italics in original). Of course, Nagin and Paternoster were at the time speaking of trait self-control, but it is, at least, an empirical possibility that redefined self-control might influence one’s bonds. But, given self-control is now cost centered rather than trait centered, it seems logical that redefined self-control stems from one’s social bonds rather than the reverse. Nevertheless, future research should bring to bear longitudinal data to test whether redefined self-control indeed mediates the influences of social bonds on substance use and delinquency more generally.

Finally, there appears to be considerable disagreement concerning how best to measure redefined self-control, and the measure advanced here is not without its own limitations. Asking individuals to report whether they face a number of costs, even if they cover a wide range of possible consequences, does not identify all possible costs that one could face. However, this wide range of negative outcomes of marijuana smoking appears to be a reasonable approximation of the full range of costs, and the measurement strategy is able to capture the elements of rational choice and cost consideration that are implicit and explicit with the merger of self-control and social-control theories.

It is worth reiterating that the measure of redefined self-control used in this study is more strongly correlated to social bonds (i.e., the measure that Hirschi used to assess self-control) than the three alternative measures of redefined self-control used in previous studies (cf. Bouffard & Rice, 2011; Higgins, Wolfe, et al., 2008; Piquero & Bouffard, 2007). This suggests that there are important differences in findings when using the self-generated cost methodology and/or using only one dimension of self-control as compared with the measurement strategy used herein. We do not believe the measure of redefined self-control used in this study to be the final word on the modern-day self-control measurement problem and strongly encourage future independent exploration into the measurement of redefined self-control. Importantly, it would be wise to avoid getting married to a measure prematurely, have it fall under severe criticism years later, and ultimately, have an entire body of empirical evidence called into question—as was the case with the general theory of crime (see Marcus, 2004). Although we have argued that previous measurement strategies have key limitations, certain measures have some unique properties. For instance, Morris et al. (2011) allowed individual variation in the salience dimension. Perhaps incorporating that aspect of their measure with the benefits of one presented in this study may be fruitful. Nevertheless, the limited empirical evidence that exists on the revised self-control theory, along with the results of this study, shows redefined self-control to have some utility in explaining substance use and, at the least, to be worthy of additional theoretical and empirical attention.

Footnotes

Appendix

Path Decomposition of Primary Model and Three Alternative Models (N = 2,243)

Measure	Effect	Coefficient	p value
1. Redefined self-control (costs × salience)	Total effect	−0.156	<.001
	Total indirect effect^a	−0.070	<.001
	Direct effect	−0.086	<.01
2. Costs	Total effect	−0.156	<.001
	Total indirect effect^a	−0.046	<.001
	Direct effect	−0.110	<.001
3. Salience	Total effect	−0.156	<.001
	Total indirect effect^a	−0.040	<.001
	Direct effect	−0.116	<.001
4. Multiple mediator model (costs and salience)	Total effect	−0.156	<.001
	Total indirect effect	−0.066	<.001
	Specific indirect effect (B → C → M)	−0.041	<.001
	Specific indirect effect (B → S → M)	−0.025	<.001
	Direct effect	−0.090	<.01

Note: B = bonds; C = costs; M = marijuana use; S = salience. Standardized effects reported. All models control for age, race, gender, delinquent peers, and religiosity. Models are fully saturated and, therefore, provide perfect fit to the data. Full-model details are available on request.

Also the specific indirect effect.

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.

Notes

Author Biographies

Jeffrey T. Ward is an assistant professor in the Department of Criminal Justice at the University of Texas at San Antonio. His research interests include sanction effects, gangs, developmental and life-course criminology, and quantitative methodology. His work has recently appeared in Criminology, Criminal Justice and Behavior, and Journal of Criminal Justice.

John H. Boman, IV is a doctoral student at the University of Florida in the Department of Sociology and Criminology & Law. His research interests include peers and peer influence, gender, measurement, quantitative methods, and construct validation. He has recently published in Journal of Research in Crime & Delinquency, Substance Use and Misuse, and Criminal Justice and Behavior.

Shayne Jones is an associate professor in the Department of Criminology at the University of South Florida. His research interests include personality and legal decision making. His work has recently appeared in Crime & Delinquency, Journal of Criminal Justice, and Journal of Abnormal Psychology.

References

Akers

R. L.

(1990). Rational choice, deterrence, and social learning theory in criminology: The path not taken. Journal of Criminal Law and Criminology, 81, 653-676.

Akers

R. L.

(1991). Self-control as a general theory of crime. Journal of Quantitative Criminology, 7, 201-211.

Akers

R. L.

(2008). Self-control and social learning theory. In Goode

(Ed.), Out of control: Assessing the general theory of crime (pp. 77-89). Stanford, CA: Stanford University Press.

Akers

R. L.

(2009). Social learning and social structure: A general theory of crime and deviance. New Brunswick, NJ: Transaction Publishing.

Akers

R. L.

(2010). Religion and crime. Criminologist, 35, 2-6.

Akers

R. L.

Krohn

M. D.

Lanza-Kaduce

Radosevich

(1979). Social learning and deviant behavior: A specific test of a general theory. American Sociological Review, 44, 636-655.

Akers

R. L.

Sellers

C. S.

(2009). Criminological theories: Introduction, evaluation, and application. Los Angeles, CA: Roxbury Publishing.

Burkett

S. R.

White

(1974). Hellfire and delinquency: Another look. Journal for the Scientific Study of Religion, 13, 455-462.

Bouffard

J. A.

Rice

S. K.

(2011). The influence of the social bond on self-control at the moment of decision: Testing Hirschi’s redefinition of self-control. American Journal of Criminal Justice, 36, 138-157.

10.

Elliott

D. S.

Ageton

S. S.

(1980). Reconciling race and class differences in self-reported and official estimates of delinquency. American Sociological Review, 45, 95-110.

11.

Gottfredson

M. R.

Hirschi

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

12.

Grasmick

H. G.

Tittle

C. R.

Bursik

R. J.

Jr Arneklev

B. J.

(1993). Testing the core empirical implications of Gottfredson and Hirschi’s general theory of crime. Journal of Research in Crime & Delinquency, 30, 5-29.

13.

Higgins

G. E.

(2007). Examining the original Grasmick scale: A Rasch model approach. Criminal Justice and Behavior, 34, 157-178.

14.

Higgins

G. E.

Ricketts

M. L.

Vegh

D. T.

(2008). The role of self-control in college student’s perceived risk and fear of online victimization. American Journal of Criminal Justice, 33, 223-233.

15.

Higgins

G. E.

Wolfe

S. E.

Marcum

C. D.

(2008). Digital piracy: An examination of three measurements of self-control. Deviant Behavior, 29, 440-460.

16.

Hirschi

(1969). Causes of delinquency. Berkeley: University of California Press.

17.

Hirschi

(1986). On the compatibility of rational choice and social control theories of crime. In Cornish

Clarke

(Eds.), The reasoning criminal: Rational choice perspectives on offending (pp. 105-118). New York, NY: Springer-Verlag.

18.

Hirschi

(2004). Self-control and crime. In Baumeister

R. F.

Vohs

K. D.

(Eds.), Handbook of self-regulation: Research, theory, and applications (pp. 537-552). New York, NY: Guilford.

19.

Hirschi

Gottfredson

M. R.

(1994). The generality of deviance. New Brunswick, NJ: Transaction Publishing.

20.

Hirschi

Stark

(1969). Hellfire and delinquency. Social Problems, 17, 202-213.

21.

Huber

P. J.

(1967). The behavior of maximum likelihood estimates under non-standard conditions. In Proceedings of the fifth Berkeley symposium on mathematical statistics and probability (pp. 221-233). Berkeley: University of California Press.

22.

Krohn

M. D.

Massey

J. L.

(1980). Social control and delinquent behavior: An examination of the elements of the social bond. Sociological Quarterly, 21, 529-544.

23.

Marcus

B. N.

(2004). Self-control in the general theory of crime: Theoretical implications of a measurement problem. Theoretical Criminology, 8, 33-55.

24.

Morris

R. G.

Gerber

Menard

(2011). Social bonds, self-control, and adult criminality: A nationally representative assessment of Hirschi’s revised self-control theory. Criminal Justice and Behavior, 38, 584-599.

25.

Nagin

D. S.

Paternoster

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

26.

Piquero

A. R.

(2008). Measuring self-control. In Goode

(Ed.), Out of control: Assessing the general theory of crime (pp. 26-37). Stanford, CA: Stanford University Press.

27.

Piquero

A. R.

Bouffard

J. A.

(2007). Something old, Something new: A preliminary investigation of Hirschi’s redefined self-control. Justice Quarterly, 24, 1-27.

28.

Piquero

A. R.

MacIntosh

Hickman

(2000). Does self-control affect survey response? Applying exploratory, confirmatory, and item response theory analysis to Grasmick et al.’s self-control scale. Criminology, 38, 897-929.

29.

Pratt

T. C.

Cullen

F. T.

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

30.

Sampson

R. J.

Laub

J. H.

(1993). Crime in the making: Pathways and turning points through life. Cambridge, MA: Harvard University Press.

31.

Sampson

R. J.

Laub

J. H.

(2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press.

32.

Smith

(1995). Criminology for social work. Hampshire, UK: Macmillan.

33.

Taylor

(2001). The relationship between social and self-control: Tracing Hirschi’s criminological career. Theoretical Criminology, 5, 369-388.

34.

Ward

J. T.

Gibson

C. L.

Boman

Leite

W. L.

(2010). Assessing the validity of the retrospective behavioral self-control scale: Is the general theory of crime stronger than the evidence suggests? Criminal Justice and Behavior, 37, 336-357.

35.

Warr

(2002). Companions in crime: The social aspects of criminal conduct. New York, NY: Cambridge University Press.

36.

Watkins

A. M.

Melde

(2007). The effect of self-control on unit and item nonresponse in an adolescent sample. Journal of Research in Crime & Delinquency, 44, 267-294.

37.

White

(1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 50, 1-25.