Testing the Stability of Self-Control

Abstract

Gottfredson and Hirschi suggest that individuals’ levels of self-control remain stable over the life course; however, the empirical status of this proposition remains equivocal. Most tests of the stability hypothesis have employed aggregate assessment methods (e.g., mean-level and correlational analyses) that overlook unique developmental patterns, although some have identified unique developmental patterns in self-control. The current study assesses the stability of self-control across 4 years using both traditional analytic methods and methods that account for the existence of unique developmental patterns (i.e., semiparametric group-based trajectory modeling) and exploring risk factors that differentiate these patterns. The results suggest six unique developmental patterns of self-control: two with high stable trajectories and four that evinced lower, less stable trajectories of self-control. The findings indicate that lower, less stable patterns of development are associated with more delinquent peer association, higher rates of parental criminality, fewer school bonds, and weaker maternal attachment.

Keywords

self-control stability rank-order semiparametric group-based modeling risk factors

Gottfredson and Hirschi (1990) state that individuals who engage in antisocial behavior are unconcerned about the long-term consequences of their actions and possess a trait they termed low self-control that underpins this tendency. Their theory has received a considerable amount of empirical attention. The majority of extant research has examined the association between low self-control and antisocial behavior (see, e.g., Pratt & Cullen, 2000). Research has also tested the claim that self-control is a unidimensional construct comprising six components: impulsiveness, a preference for simple tasks, physical orientation (as opposed to mental), risk-seeking, quick-tempered, and insensitive (see, e.g., DeLisi, Hochstetler, & Murphy, 2003; Longshore, Turner, & Stein, 1996; Piquero & Rosay, 1998). Other research supports the claim that parenting is responsible for the development of self-control (see, e.g., Burt, Simons, & Simons, 2006; Gibbs, Giever, & Higgins, 2003; Hay, 2001; Hope & Chapple, 2004; Perrone, Sullivan, Pratt, & Margaryan, 2004; Unnever, Cullen, & Agnew, 2006; Unnever, Cullen, & Pratt, 2003), although other sources such as school (Beaver, Wright, & Maume, 2008; Burt et al., 2006; Meldrum, 2008; Turner, Piquero, & Pratt, 2005), peer association (Burt et al., 2006; Meldrum, 2008), and biological/genetic factors (Beaver, DeLisi, Vaughn, & Wright, 2010; Beaver, Ratchford, & Ferguson, 2009; Beaver, Schutt et al., 2009; Wright & Beaver, 2005; Wright, Beaver, DeLisi, & Vaughn, 2008) contribute to the development of self-control as well.

Another important tenet of the theory, one that has received considerably less attention, is the stability hypothesis. The theory contends that from the end of childhood, self-control is stable across the life course. This is important as it implies that the same factor (low self-control) is responsible for offending regardless of age. It also implies that once set, one’s level of self-control is not malleable. Thus, the theory implies that interventions aimed at increasing self-control must occur early in life (i.e., before ages 8 to 10). A handful of studies have examined the stability of self-control (Arneklev, Cochran, & Gainey, 1998; Burt et al., 2006; Mitchell & Mackenzie, 2006; Turner & Piquero, 2002; Winfree, Taylor, He, & Esbensen, 2006). These studies have provided valuable insights into the issue of stability, but based on their analytic approach, all are premised on the notion that trajectories of self-control are the same for all individuals.

More recently, studies evaluating the stability hypothesis demonstrated that not everyone follows the same developmental patterns of self-control (Hay & Forrest, 2006; Higgins, Jennings, Tewksbury, & Gibson, 2009). These efforts at the identification of nonstable latent classes are important as they suggest that, for some, self-control is malleable and that treatment efforts aimed at increasing self-control may be effective even later in the life course. These findings also suggest that certain factors are associated with distinct developmental patterns, which may provide guidance for interventions aimed at promoting self-control (see Piquero, Jennings, & Farrington, 2010). The current study supplements prior research by identifying unique developmental patterns of self-control and providing a more rigorous test of the variance both within latent classes and across the sample as a whole. Finally, the current study extends prior research by examining risk factors that differentiate unique developmental patterns of self-control.

The Stability Hypothesis

Gottfredson and Hirschi (1990) unequivocally describe self-control as a stable, enduring trait beyond the ages of 8 to 10. The nature of this stability, however, is less clear. Many times a distinction is made between relative and absolute stability (see, e.g., Caspi, Robert, & Shiner, 2005). Absolute stability suggests that levels of some traits remain constant and that no within-individual change occurs over time. Relative stability, on the other hand, refers to the notion that between-individual differences on some trait within a single cohort remain constant over time. Although normative changes may be occurring, they occur at the same rate and in the same direction for all individuals, maintaining the rank order within that cohort over time. Thus, absolute stability also assumes relative stability, but relative stability can occur without absolute stability.

A closer examination of the theory justifies the common interpretation that it is one of relative stability (e.g., Mitchell & Mackenzie, 2006; Turner & Piquero, 2002). For example, Gottfredson and Hirschi (1990) argue: “As people with low self-control age, they tend less and less to commit crimes; this decline is probably not entirely due to increasing self-control, but to age as well” (p. 111). This statement seems to suggest that changes in self-control may occur and that these changes may contribute, at least somewhat, to normative declines in offending. Elsewhere, they disregard the possibility of decreases in self-control and attribute increases in self-control to socialization that continues to occur into adulthood:

Combining little or no movement from high self-control to low self-control with the fact that socialization continues to occur throughout life produces the conclusion that the proportion of the population in the potential offender pool should tend to decline as the cohort ages. (Gottfredson & Hirschi, 1990, p. 107, emphasis added)

Therefore, it appears that self-control should be considered a relatively stable trait and that normative increases in self-control occur for the entire population. It also would appear that such increases occur at the same rate for all individuals. Specifically, those who originally had the lowest levels of self-control will always be lower in self-control compared to others of the same age and, therefore, will always be more prone to crime.

Studies testing the stability hypothesis have used a variety of methods and analytic procedures. Several studies support the notion of absolute stability, finding that mean levels of self-control either remain the same (Arneklev et al., 1998; Raffaelli, Crockett, & Shen, 2005; Yun & Walsh, 2010) or marginally increase over time (Turner & Piquero, 2002; Vazsonyi & Huang, 2010; Winfree et al., 2006). However, mean-level analyses assess absolute stability at the aggregate level, potentially masking within-individual change that might be occurring. Two studies have attempted to overcome this limitation by employing hierarchical linear modeling (HLM; Arneklev et al., 1998; Hay & Forrest, 2006). Arneklev et al. (1998) found no evidence of within-individual change. However, self-control was assessed over a very short time period and was only assessed at two time points. Using HLM in this way does not allow for the possibility that self-control may change in a nonlinear manner. Hay and Forrest (2006), on the other hand, examined changes in self-control over the course of 9 years measured at five time points. The results from their HLM analysis revealed evidence of within-individual change in levels of self-control, providing support for absolute change in levels of self-control.

Other studies have focused on the relative stability of self-control, typically by examining stability coefficients. Stability coefficients are simply correlation coefficients between self-control scores measured at two time points. Several studies taking this approach demonstrate that self-control is relatively stable, at least over the short term. For example, Arneklev et al. (1998) and Beaver et al. (2008) found scores on self-control assessed at two separate time points to be highly correlated (i.e., rs = .82 and .64, respectively). Furthermore, Burt et al. (2006) found self-control assessed when youth were ages 10 to 12 to be moderately correlated with self-control at ages 12 to 14 (r = .48). Using a two-wave panel design, Mitchell and Mackenzie (2006) found scores on self-control to be correlated at .48 over a period of 6 months. In addition, based on data collected from samples during similar developmental periods (i.e., ages 8 to 12), Polakowski (1994), Raffaelli et al. (2005), and Vazsonyi and Huang (2010) all found self-control measured at the beginning of this period to be strongly correlated with self-control measured at the end of this period (rs = .59, .50, and .70, respectively).

Other studies have found that as time increases between assessment periods, the relative stability of self-control decreases. For instance, Yun and Walsh (2010) reported correlations for self-control measured yearly at five different waves. Their analysis revealed strong correlations between Waves 1 and 2 (r = .52) and moderate correlations between Wave 1 and Waves 3, 4, and 5 (rs = .47, .42, and .42, respectively). Winfree et al. (2006) noted similar findings for self-control measured yearly over five waves (rs = .58, .48, .44, and .44, respectively). Turner and Piquero (2002) found correlations ranging from .33 to .68 over seven waves of data, with those measurement periods further apart having correlations with lower magnitudes, suggesting moderate to strong relative stability. Taken together, these findings underscore the important influence of time on stability coefficients and highlight the necessity of having multiple waves of data that offer sufficiently lengthy periods of follow-up.

Despite the advantages of stability coefficients as a way of assessing relative stability, they do possess inherent limitations. First, they assess stability at the aggregate level and have the potential to conceal individual differences in stability (Lamiell, 1981; Mroczek, 2007). Second, they only examine stability between two time points and fail to account for nonlinear developmental patterns (Mroczek, 2007). Thus, the use of stability coefficients cannot ascertain whether or not there are different rates of change that might mask trajectories of smaller yet salient groups that do not follow aggregate trends in the development of self-control. It is possible that not all individuals follow the same developmental patterns in self-control. This is evident in research that has identified distinct developmental patterns in offending (Blokland & Nieuwbeerta, 2005; Chung, Hill, Hawkins, Gilchrist, & Nagin, 2002; D’Unger, Land, McCall, & Nagin, 1998; Fergusson, Horwood, & Nagin, 2000; Laub, Nagin, & Sampson, 1998; Nagin, Farrington, & Moffitt, 1995; Nagin & Land, 1993; Piquero et al., 2001; Sampson & Laub, 2003). If self-control underlies offending, it is possible that developmental patterns in self-control parallel those found for offending behavior.

Research on personality development also provides evidence of distinct developmental patterns of self-control (Blonigen, 2010). Roberts, Caspi, and Moffitt (2001) found groups with discernible patterns in the development of Negative Emotionality (NEM; e.g., anger, antagonism, and anxiety). Additionally, they found that a sizeable portion (12.7%) of their sample evinced significant increases in self-control, as measured by the Control scale of the Multidimensional Personality Questionnaire (MPQ; Tellegen, 2006), between ages 18 and 26, well beyond the age at which self-control should be set. Furthermore, Robins, Fraley, Roberts, and Trzesniewski (2001) found that although the majority of individuals followed normative developmental patterns of global personality traits (e.g., agreeableness, conscientiousness, and neuroticism) from late adolescence into young adulthood, a small group evinced decreases in conscientiousness. Johnson, Hicks, McGue, and Iacono (2007) identified unique patterns of development for some facets of the MPQ among girls measured at four time points from ages 14 to 24. Specifically, and most relevant, they identified four trajectory groups for the Aggression facet (high–decreasing, very high–stable, moderate–steady, and low–decreasing) and three trajectory groups for the Control facet (Low–Steady, Moderate–Increasing, and High–Steady). Cote, Tremblay, Nagin, Zoccolillo, and Vitaro (2002) identified four distinct trajectories of development of impulsivity for a sample of young boys over a span of 7 years. Finally, Schaeffer, Petrus, Ialongo, Poduska, and Kellam (2003) identified unique developmental patterns of aggression. Collectively, these studies indicate that personality traits conceptually related to self-control do not follow uniform developmental trajectories; among some individuals, these traits remain relatively stable, whereas among others, there is evidence of increases and decreases in the level of the trait. This suggests that a more nuanced and appropriate approach requires the disaggregation of developmental patterns.

Trajectories of Self-Control

Only a few studies have attempted to directly identify distinct developmental patterns of self-control. Using group-based trajectory modeling, Higgins et al. (2009) assessed the stability of self-control among a sample of youth between the ages of 12 and 16. Specifically, self-control was assessed using self-report items that represented the risk-taking and impulsivity aspects of self-control. They identified five distinct groups with different developmental trajectories of self-control. For the most part, their findings support the stability hypothesis in that most of the sample maintained their rank-order position, staying within their respective levels of self-control across all five waves. However, they did identify a group that began with the lowest levels of self-control and gradually increased over the five waves.

Using the same analytic method, Hay and Forrest (2006) identified eight trajectories of self-control based on data from the National Longitudinal Study of Youth (NLSY79). Their measure of self-control was based on mother-reports and tapped into theoretically relevant components of self-control as outlined by Gottfredson and Hirschi (i.e., impulsivity, self-centeredness, inability to get along with others, temper, and behavioral problems). Four groups were characterized by considerable stability: a very high-stable (HS) group, a HS group, a medium-stable group, and a low-stable group. They also identified four groups marked by considerable change: a low-increasing (LI) group, a high-decreasing group, a medium-decreasing group, and a low-curvilinear group. It is important to note here that although the number of self-control trajectories was different when comparing Higgins et al.’s (2009) results (e.g., five trajectories) to Hay and Forrest’s (2006) results (e.g., eight trajectories), there was a considerable amount of similarity in the overall percentage of those individuals who exhibited a stable trajectory (80% to 85%). Nevertheless, Hay and Forrest’s findings led them to the conclusion that self-control is not stable in the absolute sense for some individuals. Additionally, they identified two groups that were characterized by absolute losses in self-control—a finding specifically at odds with Gottfredson and Hirschi’s predictions. Also, they found that some of the trajectories intersected with or crossed many of the other trajectories. They interpreted this as showing that some trajectories change enough to alter their rank-order position, questioning the relative stability of self-control. One limitation, however, is that the NLSY79 includes youth from different birth cohorts (1979 to 1990); thus, their analyses may have been affected by potential cohort effects.

An important issue not specifically addressed by either of the aforementioned studies was that neither examined if risk factors could differentiate the developmental patterns of self-control. Although Hay and Forrest (2006) did find that increases in positive parenting resulted in subsequent increases in self-control, this was a separate analysis not directly linked to the developmental patterns identified in their sample. Furthermore, although Higgins et al. (2009) were able to show that self-control trajectories were linked with victimization trajectories, their analysis did not focus on the risk factors that may distinguish self-control trajectories.

Risk Factors for Distinguishing Trajectories

Because Gottfredson and Hirschi (1990) did not hypothesize there to be any heterogeneity in developmental patterns of self-control, there is no theoretical basis to identify what risk factors might be most relevant. Having said this, in order to identify distinguishing risk factors, a prudent point of departure would be to examine research focused on developmental trajectories of offending (see Piquero, 2008) as well as those found to be responsible for the development of self-control (e.g., school, parenting, and peers). For example, Nagin (1999) found that individuals identified as having persistent offending trajectories (i.e., chronic offenders) were more likely to come from low-income households, to have experienced poor parenting, and to have parents with criminal histories. Fergusson et al. (2000) found that individuals with distinct developmental trajectories in offending (i.e., nonoffending, moderate, adolescence-limited, and chronic offending groups) had differential levels of family dysfunction (e.g., parental criminality, alcohol, and drug use), social disadvantage, and delinquent peer associations. Fergusson et al. also found that parental criminality, gender, and parental conflict were associated with being in the chronic offending group compared to other offending trajectories. Weisner and Capaldi (2003) found those who were classified as being chronic offenders were more likely to have poor parenting compared to those in the nonoffending or low-offending groups. They also found those who were identified as having decreasing offending trajectories were less likely to associate with delinquent peers than those in the chronic offending groups. Chung et al. (2002) found those youth with desisting trajectories in offending were more likely to experience better parenting, have less family conflict, have higher levels of school commitment and attachment, and have fewer delinquent peers compared to those with increasing offending trajectories. Most recently, Maldonado-Molina, Piquero, Jennings, Bird, & Canino (2009) identified sensation-seeking and exposure to violence as important risk factors for distinguishing group-based delinquency trajectories of children and adolescents.

Factors previously identified as integral in the development of self-control may also be important in distinguishing trajectories of self-control. For instance, proper parenting (Burt et al., 2006; Gibbs et al., 2003; Hay, 2001; Hope & Chapple, 2004; Perrone et al., 2004; Unnever et al., 2006; Unnever et al., 2003) and positive school experiences (Beaver et al., 2008; Burt et al., 2006; Meldrum, 2008; Turner et al., 2005) have been found to contribute positively to its development. Also, peer association has also been implicated in the development of self-control, suggesting that deviant peer association may impede its development (Burt et al., 2006; Meldrum, 2008). Thus, factors associated with the development of self-control and trajectories of offending (i.e., parental characteristics, parenting style, school socialization, and delinquent peer association) may be important in distinguishing distinct patterns of self-control.

Current Study

Although prior research has provided important insight on the stability of self-control, only a few studies have accounted for heterogeneity in developmental patterns of self-control. Of these studies, however, none have directly examined if risk factors could distinguish the identified patterns. Hay and Forrest (2006) did identify parenting as influential in the development of self-control; however, as previously mentioned, they did not link parenting to group membership identified in their trajectory analysis. As such, the current study not only attempts to replicate previous findings by exploring whether trajectories of self-control in a separate sample can be identified but also extends beyond those studies by providing a more rigorous test of the variance both within latent classes and across the sample as a whole, and examining if certain risk factors can distinguish self-control trajectories. Specifically, using group-based trajectory modeling, the current study attempts to identify distinct developmental patterns of self-control.¹ To the extent trajectories are indentified, several key risk factors from multiple domains (i.e., school socialization, parenting, delinquent peer association, race, gender, and parental criminality) will be examined to determine whether they can significantly discriminate unique developmental patterns of self-control from one another using a multinomial logistic regression (MLOGIT) analysis.

Method

Data Collection

The data used in the current study were taken from the Rural Substance Abuse and Violence Project (RSVP; NIDA Grant DA-11317). The RSVP was a longitudinal study in which data were collected from a large sample of youths from middle and high schools located in Kentucky. The RSVP study was designed to collect individual and contextual data about study participants known to influence offending, victimization, and substance abuse. The RSVP used a multistage sampling technique to select study participants. Based on population-based strata, 30 counties located within the state were randomly selected, with rural counties oversampled to ensure representativeness.

To be considered for inclusion schools had to be public and include seventh graders. Within the 30 counties, 74 middle schools were eligible and asked to participate in the study. Of the 74, 9 declined, leaving 65 middle schools from which seventh-grade students were recruited to participate. At the start of data collection, there were 9,488 seventh graders within these schools who were recruited for participation. Because the study design was longitudinal and students were to be followed over multiple waves of data collection, active consent had to be obtained from the parents. Parental consent was obtained for 4,102 (43%) of the students. The response rates for each year were as follows: 90.0%, 88.7%, 74.4%, and 74.1%.

Participants

In the current study, only those individuals for whom self-control was observed at three time points were included in the analyses, as two or fewer observations would make identifying nonlinearity impossible; 727 (17.7%) did not meet this criterion and were omitted from the analyses. This resulted in a final sample size of 3,249 individuals who had data on self-control for at least three time points. The mean age at Time 1 was 13.39 (see Table 1). The racial distribution was as follows: 159 (4.9%) African Americans, 2,791 (85.9%) Whites, 59 (1.8%) mixed race (African American and White), 16 (0.5%) Native Americans, 16 (0.5%) Asian Americans, 46 (1.4%) reported being “other,” and 162 (5%) were missing information on race. A slight majority of the sample was female (n = 1,655; 51%) compared to males (n = 1,431; 44%). The racial profile of the sample matched the Kentucky Department of Education data, although the sample had a somewhat lower proportion of males relative to the population (Wilcox, Tillyer, & Fisher, 2009).

Table 1:

Descriptive Statistics for Study Variables

	Valid n	M	SD	Minimum	Maximum
Minority	3,087	.10	0.29	0	1
Female	3,086	.54	0.50	0	1
Parental crime	3,027	.21	0.41	0	1
Age
Grade 7	3,249	13.39	0.56	12	19
Grade 8	3,249	14.39	0.56	13	20
Grade 9	3,249	15.39	0.56	14	21
Grade 10	3,249	16.39	0.56	15	22
Self-Control
Grade 7	3,038	31.51	6.76	10	40
Grade 8	3,171	31.50	6.86	10	40
Grade 9	2,943	32.77	6.72	10	40
Grade 10	2,940	32.75	6.85	10	40
Parenting
Mother
Attachment	3,073	18.98	4.41	3	25
Supervision	3,072	17.45	3.06	4	20
Father	3,007	36.71	10.45	1	50
School bonds
Commitment	3,072	16.41	2.97	3	20
Rules	3,089	20.11	4.43	4	28
Delinquent peers	2,476	1.47	2.50	0	12

A comparison was made between individuals who were excluded from subsequent analyses and those who were retained based on having a measure of self-control for at least three waves of data on all variables included in the analyses. Those who were excluded were more likely to be male, χ² ₍₁₎ = 0.06, p < .001, φ = .06, and White, χ² ₍₁₎ = 30.60, p < .001, φ = .09, and less likely to have a parent who had been to jail or prison, χ² ₍₁₎ = 53.97, p < .001, φ = -.12. In addition, those who were excluded had significantly lower mother attachment, t _(3,656) = -6.15, p < .001, d = -.20, mother supervision, t _(3,655) = -5.97, p < .001, d = -.20, father parenting, t _(3,576) = -7.37, p < .001, d = -.25, school commitment, t _(3,688) = -6.07, p < .001, d = -.20, and belief in school rules, t _(3,672) = -5.56, p = .003, d = -.18. Finally, those excluded reported that their peers engaged in significantly more delinquency, t _(2,926) = 2.43, p = .003, d = .09, and reported lower levels of self-control measured during the seventh grade, t _(3,569) = -6.69, p < .001, d = -.22, compared with those who remained in the sample. Although it is important to take these differences into account when interpreting the results, all of the effect sizes were small.

Dependent Variable

Self-control

A 10-item measure of self-control was used in this study. Select items were drawn from the Dysregulation Inventory (Mezzich, Tarter, Giancola, & Kirisci, 2001), which assesses affective, cognitive, and behavioral dysregulation. The selected items tap into one’s self-reported ability to control their behavior and focus or pay attention, especially when such behavior would be most appropriate. Response sets for items were on a 4-point Likert-type scale (1 = never true, 4 = always true). The items used in the current study are consistent with measures of self-control used in previous studies (e.g., Hay & Forrest, 2006; Meldrum, 2008; Turner & Piquero, 2002; see appendix). These 10 items were subject to a principal components analysis at each wave of data collection, with a one-factor solution identified at each time point (consistent with the full scale; see Mezzich et al., 2001). A visual inspection of the distribution of scores on self-control resembled a normal distribution, with only modest clustering at low and high values. This is important for specifying the correct model in the trajectory analysis (see below). Items were reverse coded and summed, with higher scores representing greater self-control. The full scale demonstrated strong psychometric properties (Mezzich et al., 2001). The abbreviated scale used in the current analysis also demonstrated high internal reliability, with Cronbach’s alpha values across waves ranging from .89 to .92.

Risk Factors

Parenting

Mother and father parenting were measured at Wave 1 in the current study. Both sources of parenting were measured based on 10 items in which individuals responded to questions about their mother’s/father’s parenting style (e.g., My mother/father seem to understand me; My mother/father know where I am/who I am with when I am away from home) using a 5-point Likert-type scale (1 = never and 5 = always). Items reflecting mother parenting and father parenting were each subject to principal components analysis with promax rotation. Observation of the factor loadings and scree plot for the items from the mother scale revealed a two-factor structure with one factor representing supervision (four items; α = .70) and the other representing attachment (six items; α = .84), whereas the father scale revealed a single-factor structure for all 10 items (α = .93). Higher scores on each of the three scales are indicative of better parenting.

Delinquent peer association

Respondents were asked to indicate how many of their closest friends had engaged in 12 specific types of delinquency. Items included behaviors ranging from having smoked marijuana to having physically attacked someone. Each item was binary coded so that responses indicating that none of their friends had engaged in a specific type of behavior were coded as 0 and those indicating that one or more of their friends engaged in a specific type of behavior were coded as 1. The items were then summed to obtain a peer delinquency scale where higher scores are indicative of more variety in peer delinquency (α = .87).

School bonding

School bonding was measured with 12 items in which respondents answered questions about their attitudes toward school. The items were subject to principal components analysis using promax rotation, yielding two distinct components. One component represented one’s “belief in school rules” and comprised seven items (e.g., Everyone knows what the school rules are; The school rules are strictly enforced). The other component comprised five items that captured “school commitment” (e.g., I care a lot what my teachers think of me; Getting an education is important to me). The two school bonding scales evinced good reliability (αs = .77 and .69, respectively).

Parental crime

To account for the intergenerational transmission of criminality, the current study also includes a measure of parental crime. Respondents were asked if either of their parents had ever been in jail or prison (1 = yes, 0 = no).

Control variables

As noted above, the sample was overwhelmingly White (90%), with relatively small percentages representing other racial groups. As such, race was dichotomized (1 = White, 0 = minority). Likewise, gender was binary coded (1 = male, 0 = female).

Analytic Strategy

Rank-order stability was assessed using Spearman’s rank-order correlation coefficients (i.e., correlations between self-control measured at each time point; SPSS Version 19). In order to account for unique developmental patterns in self-control, semiparametric group-based trajectory modeling (SPGM; Nagin, 2005) was applied using the Proc Traj (Jones, Nagin, & Roeder, 2001) procedure designed for SAS. Model selection was done in accordance with Nagin’s (2005) guidelines (i.e., for each number of groups, each trajectory was initially fully parameterized as a quadratic specification). The highest order term of those that were not statistically significant was dropped, and the model was then re-estimated until the highest order parameter estimate was statistically significant for each group. At this point, cross-model comparisons were made based on the Bayesian Information Criteria (BIC) and the mean posterior probabilities of group classification. Once unique trajectories were identified, two types of analyses were conducted. First, we explored the variance both within and between latent classes in an effort to test the stability of self-control (SPSS Version 19). Second, using the “classify-analyze” approach, several variables were used to test whether there were unique risk factors that differentiated developmental patterns of self-control. Specifically, unique associations between risk factors and developmental patterns were assessed using multinomial logistic regression (MLOGIT; STATA 11.0). Separate models were specified in which different developmental patterns were set as the comparison group in order to examine all possible group comparisons on the relevant risk factors. Due to the multi-stage sampling approach, clustering of data was accounted for by using the robust clustered standard errors available in STATA. School identification codes were used as the clustering variable. The school identification code is a unique number associated with each school from which kids were recruited and was included as part of the data collection.

Results

Table 2 reports the results for the Spearman’s rank-order correlations between self-control measured at each time point. Overall, the results show moderate to strong correlations between measures of self-control over time, suggesting a considerable amount of stability. In addition, the observed correlations revealed that the magnitude of the relationships decrease across longer time intervals. Additionally, the magnitude of the correlations between two adjacent time points becomes stronger as individuals get older, suggesting more stability later in adolescence.

Table 2:

Spearman’s Rank-Order Correlations Between Self-Control Measured at Grade

Self-Control	Grade 7	Grade 8	Grade 9
Grade 7	—
Grade 8	.52	—
Grade 9	.48	.53	—
Grade 10	.43	.48	.61

Note. All correlations are significant at the p < .001 level.

The next step was to test the possibility that some groups of individuals follow distinct developmental patterns of self-control that may be masked using aggregate-level analyses such as stability coefficients. Using SPGM and following the model section procedure described above, the six-group trajectory model (see Figure 1) was selected as the best fitting model. Although the BIC was maximized in the seven-group model (-36,301.77), it was only slightly better than that of the six-group model (-36,306.96). In addition, the mean posterior probabilities for the six-group model all exceeded the suggested .70 (Nagin, 2005; range of mean posterior probabilities = .71 to .83). More importantly, the seven-group model was no more substantively meaningful than the six-group model. Thus, the six-group model was selected because it was a more parsimonious approximation of trajectories of self-control.

Figure 1:

Trajectories of Self-Control

Figure 1 presents the trajectories for each of the six different trajectories along with the percentage of the total sample comprising each trajectory. The first trajectory (3.5% of the sample) was labeled as LI, as its trajectory began with low levels of self-control and gradually increased over time. The next trajectory made up 2.3% of the sample and was labeled as “moderate-decreasing” (MD) as individuals following this trajectory began with moderate levels of self-control and declined over time. A third trajectory composed 8.3% of the sample and was labeled as “moderate-increasing” (MI) as individuals following this trajectory began with a moderate level of self-control and increased over time. A fourth trajectory made up 38.5% of the sample and was labeled as “moderate-high stable” (MHS), as individuals in this group maintained moderate-high levels of self-control relative to the other groups over the four waves. The fifth trajectory represented 12% of the sample and was labeled as “moderate-high decreasing” (MHD), as individuals in this group began with moderate-high levels of self-control and decreased over time. Finally, a sixth trajectory contained 35.4% of the sample and was identified as HS with individuals in this group maintaining high levels of self-control over the study period. Most of the trajectories (LI, MD, MI, and MHS) were defined by a quadratic term, with the MHD and HS trajectories defined by a linear term. Overall, these findings suggest that the majority of individuals in the sample follow stable patterns of self-control; however, there is a nontrivial portion that evinces meaningful change in self-control.

Although the trajectory analysis provides evidence that there are distinct developmental patterns, it does not unequivocally address the stability hypothesis. Specifically, it remains ambiguous whether the identified trajectories are truly stable and/or changing (i.e., increasing or decreasing) and whether some trajectories are evincing greater stability compared to others. In an effort to examine this, we compared the mean levels of variability within each of the six groups to the variability between them. If there is less variability in scores within the stable groups compared to the unstable ones, this would provide evidence of differential stability in scores. In fact, this was found to be the case. Specifically, there were significant mean-level differences in variability, F _{(5, 2,339)} = 131.49, p < .001. The HS group evinced less variability than all other groups (see Table 3). The MHS group was less variable than both decreasing groups and both increasing groups but more variable than the HS group. The MI group had more variability than the LI and MD. Among those groups that failed to demonstrate significant differences in variability across waves, the MD group was equal to LI, MI, and MHD groups, whereas the MI was equivalent to the LI. The results of these analyses provide more confidence in the results of the trajectory model that the patterns that emerged are actually those that the groups followed.

Table 3:

Mean Levels of Variability Across Groups

Groups	Reference Group	Group Comparisons
HS	8.45	< MHS, LI, MI, MD, MHD
MHS	21.61	< LI, MI, MD, MHD
		> HS
LI	35.72	> HS, MHS
		< MI
MI	54.23	> HS, MHS, LI, MHD
MD	56.26	> HS, MHS
MHD	37.04	> HS, MHS
		< MI

Note. HS = high stable; MHS = moderate-high stable; LI = low increasing; MI = moderate increasing; MD = moderate decreasing; MHD = moderate-high decreasing.

Upon examination of the results from the previous analyses, more general patterns of development can be discerned. That is, the trajectory model suggests that approximately 74% of the sample seems to follow a stable self-control trajectory (i.e., the 35.4% of individuals classified as HS plus 38.5% of those classified as MHS).² Next, about 14.3% of the sample appears to display decreasing self-control (12.0% MHD plus 2.3% MD). Finally, 11.8% of the sample exhibits increasing self-control over the age range (8.3% MI plus 3.5% LI). Additionally, the patterns that emerged from the variability analyses suggest that the individual groups comprising each of the respective patterns (i.e., stable, increasing, and decreasing) are more similar to one another in variability compared with other groups. In turn, as was done for the six groups, a similar analysis was done in which we compared the average variance in self-control scores among these broader patterns. The variance for the stable individuals (i.e., MHS and HS) (M = 15.33) was significantly lower, t _(615.49) = -15.29, p < .001; d = .87) compared with those unstable individuals (including both decreasing and increasing patterns; M = 43.81). When comparing the stable, increasing, and decreasing individuals, there were significant differences in average variability, F _{(2, 2,342)} = 234.48, p < .001. The mean variability of the stable individuals (M = 15.53) was significantly lower (p < .001) than the means of the increasing (M = 49.01; d = 1.03) and decreasing (M = 39.94; d = .75) individuals. Moreover, those increasing were significantly more variable than those decreasing (p = .037; d = .22).

As a point of extension, the focus of the current study is to better understand these more general patterns of development by examining the risk factors that distinguish between those that follow increasing, decreasing, and stable patterns of development and not necessarily the “groups” identified in the SPGM. In turn, individuals have been sorted into their respective aggregate pattern of development in the risk factor analyses. Not only is this the focus of the current study, but it provides for more succinct and interpretable results.

To explore if factors exist that can distinguish trajectory patterns (i.e., stable, decreasing, and increasing), we examined the relative importance of risk factors in a multivariate framework using MLOGITs. To examine comparisons between all of the groups, two MLOGITs were conducted: one with the stable developmental pattern as the comparison, and the other with the increasing developmental pattern as the reference category.3

As shown in Model 1 of Table 4, the stable developmental pattern tends to have higher levels of mother attachment and school commitment and lower levels of delinquent peers compared with both of the other developmental patterns and lower levels of parental criminality compared with the increasing developmental pattern (controlling for all other risk factors). However, other sources/types of parenting and school rules did not differentiate the stable developmental pattern from the decreasing and increasing developmental patterns. In addition, the stable developmental pattern was less likely to be of minority status than the decreasing and increasing developmental patterns.

Table 4:

Multinomial Logistic Regression: Risk Factors by Developmental Patterns

	Model 1: Stable Group			Model 2: Increasing Group
Variable	B	SE	exp(B)	B	SE	exp(B)
Gender
Decreasing	.10	.12	1.11	.01	.15	1.01
Increasing	.09	.16	1.09	—	—	—
Stable	—	—	—	−.09	.16	.92
Minority status
Decreasing	−.40*	.18	.67	−.00	.17	1.00
Increasing	−.39**	.13	.67	—	—	—
Stable	—	—	—	.39**	.13	1.48
Delinquent peers
Decreasing	.08**	.03	1.08	−.08**	.03	.93
Increasing	.16***	.02	1.17	—	—	—
Stable	—	—	—	−.16***	.02	.85
Mother attachment
Decreasing	−.04*	.02	.96	.00	.02	1.00
Increasing	−.05*	.02	.96	—	—	—
Stable	—	—	—	.05	.02	1.05
Mother supervision
Decreasing	−.03	.03	.97	−.00	.03	1.00
Increasing	−.03	.03	.97	—	—	—
Stable	—	—	—	.03	.03	1.03
Father parenting
Decreasing	−.01	.01	.99	.02	.01	1.02
Increasing	−.02	.01	.98	—	—	—
Stable	—	—	—	.02**	.01	1.02
Parental criminality
Decreasing	.11	.07	1.12	−.07	.06	.93
Increasing	.19**	.06	1.20	—	—	—
Stable	—	—	—	−.19**	.07	.83
School commitment
Decreasing	−.13***	.03	.88	.02	.03	1.02
Increasing	−.16***	.03	.86	—	—	—
Stable	—	—	—	.16***	.03	1.17
School rules
Decreasing	.03	.01	1.03	.03	.02	1.03
Increasing	−.00	.02	1.00	—	—	—
Stable	—	—	—	−.00	.02	1.00
Wald χ² (R²)	266.63*** (.08)

p < .05. **p < .01. ***p < .001.

Model 2 (Table 4) presents multivariate results when using the increasing developmental pattern as the comparison group in order to examine differences between the decreasing and increasing developmental patterns. The only risk factor that showed significant differences between the increasing and decreasing developmental patterns was delinquent peer association. Specifically, the increasing developmental pattern showed significantly higher levels of delinquent peer association compared with the decreasing developmental pattern.⁴ In sum, attachment to mother and commitment to school seem to be important protective factors regarding the development of self-control, whereas delinquent peer associations and parental criminality act as risk factors.

Discussion

The current study tested an important yet somewhat overlooked component of Gottfredson and Hirschi’s self-control theory: the stability hypothesis. Although some studies have assessed the stability hypothesis (Arneklev et al., 1998; Burt et al., 2006; Mitchell & Mackenzie, 2006; Turner & Piquero, 2002; Winfree et al., 2006), most have applied methods that neglect the possibility that there may be unique developmental patterns of self-control. Only two studies have applied methods to account for unique developmental patterns (Hay & Forrest, 2006; Higgins et al., 2009); however, neither assessed variance within and/or between groups, nor did they directly examine the association between risk factors and the identified trajectories. The current results support previous efforts in that different groups were identified that evidenced different patterns of self-control. More importantly, the current study advanced the literature by (1) providing additional support for the stability of some trajectories, while noting the instability of others, and (2) noting risk factors that distinguish developmental patterns. More specifically, stability in self-control characterized a large majority of the sample; however, sizeable groups were identified that followed trajectories lacking stability. It was also found that the trajectory patterns marked by change had significantly higher variance than those that were characterized by stability. Also, certain risk factors were associated with general developmental patterns.

The current study identified a considerable portion of the sample as following stable trajectories of self-control (n = 2,465, 73.9%). This finding is consistent with those of Higgins et al. (2009) and Hay and Forrest (2006), who found that the majority of their samples were also marked by stability (85% and 80%, respectively). Although the sample used in the current study is similar to that of the prior two studies with regard to age and level of risk, they were drawn from unique populations and thus provide compelling evidence of unique developmental patterns of self-control. Taken together, these studies suggest stability in self-control is the norm. That is, for most individuals self-control is relatively stable throughout adolescence. Also consistent with those two previous studies, a small but nontrivial portion of the sample were assigned to groups marked by change (n = 704, 26.1%). Given the consistency of these findings across multiple studies, samples, and methodologies, there now exists a reasonable basis to conclude that self-control is not stable for all individuals.

In all, six distinct developmental trajectories of self-control were identified: two that were stable, two that increased, and two that decreased. Although one of the stable groups had a higher mean level of self-control than the other, both of the stable groups evinced higher self-control than the other four groups. Both of the increasing groups started out having the lowest and second lowest mean levels of self-control and increased at similar rates. The decreasing groups both started out having moderate levels, one slightly lower than the other, of self-control, yet one decreased at a slightly higher rate than the other. Beyond simply suggesting that there are different developmental patterns characterized by stable, increasing, and decreasing self-control, these findings underscore the importance of moving beyond traditional methods of assessing stability that appear to be inadequate, as they tend to overlook unique developmental patterns (Blonigen, 2010).

In the current study, there was also an analysis of the variability of scores within and between groups. Although the trajectory analysis suggested the presence of groups marked by stable, increasing, and decreasing self-control, such findings can be supplemented and bolstered by an examination of the stability (or lack thereof) among the groups. The current investigation indicated that the stable groups evinced significantly lower levels of variability than the increasing and decreasing groups. These analyses suggest that the stability of self-control in the groups labeled as stable were, in fact, more stable. That is, they evinced less variability across the 4 years than those that were characterized by increasing and decreasing self-control. To our knowledge, this is the first attempt to directly compare the stability of different groups across time and offers some of the strongest evidence that some individuals vary in self-control much more so than others.

Furthermore, not only does the identification of decreasing and increasing groups suggest changes in absolute stability, but their divergent and sometimes intersecting patterns suggest changes in the rank-order of self-control—a finding consistent with both Hay and Forrest (2006) and Higgins et al. (2009). This small yet relevant body of research suggests that those initially lowest on self-control may not maintain that position, whereas those with moderate levels of self-control may not maintain “functional” levels of self-control as they age. In fact, the identification of decreasing patterns of self-control, despite being consistent with previous research (Hay & Forrest, 2006), is most surprising and in need of explanation, as this is inconsistent with the theory advanced by Gottfredson and Hirschi (1990).

In an effort to extend previous research and potentially provide insights into stability and change, another important contribution provided by this study was the identification of risk factors that were unique across developmental patterns. In this regard, we examined the association between risk factors and broader trajectories of self-control (i.e., the collapsed groups) in a multivariate context. Being minority, having more delinquent peers, lacking commitment to school, poor mother attachment, and parental criminality were more highly associated with lower, less stable trajectories of self-control compared to higher, more stable trajectories. However, there was little evidence that the risk factors examined in this study differentiated the increasing from decreasing groups. It remains unclear how these findings should be interpreted. Perhaps those with more criminogenic risk factors live in environments that are more chaotic and less healthy, and this bombardment of negative influences results in a less stable sense of self. Conversely, being immersed in a more prosocial context might lead to a more stable sense of identity and the ability to develop greater self-control.

Taken together, these findings are consistent with a growing body of literature that has identified factors beyond parenting as being important in the development of self-control (e.g., Burt et al., 2006; Fergusson et al., 2000; Turner et al., 2005). Although more research is needed on this topic before firmer conclusions can be drawn, a clearer understanding of the role of risk factors in the development of self-control could bolster the effectiveness of treatment programs designed to promote the development of self-control (Piquero et al., 2010). For example, children marked by such risk factors may be those who would most benefit from treatment aimed at promoting the development of self-control. Additionally, such treatment programs may be more effective by taking a more well-rounded approach and incorporating these life domains (i.e., peer associations, school, and family) into a multisystemic treatment program designed to aid the development of self-control. However, more research is needed to identify how each of these factors individually and/or collectively work together to impact the development of self-control.

The findings from the current study must be considered in light of some limitations. First, there was some attrition over time, and if this attrition is systematic due to the variable of interest (i.e., self-control), it may bias the results of the current study. That is, if the individuals who have missing values tend to be those who are also low in self-control, then the sample would lack representativeness. However, none of the analyses performed indicated that there were notable differences (in terms of effect sizes) between those with and without missing values on self-control. Second, the participants were students in Kentucky schools, which compose a fairly homogenous group primarily in terms of race and socioeconomic status. Additionally, the stability of self-control was assessed among a fairly low-risk, community sample. Therefore, the findings from the current study may not apply to more high-risk populations with more policy-relevant implications.

Despite these limitations, the current study findings are consistent with those of prior research (Hay & Forrest, 2006; Higgins et al., 2009) using similar yet different samples. Given this, consistency in findings across studies provides more confidence in the findings based on the current sample despite some issues of generalizability and representativeness. Another potential limitation of the current study is that self-control was observed over a somewhat truncated developmental period: adolescence. This is important in that there might not be considerable change during this period of time. Consistent with this notion, some have found that self-control vacillates during the adolescent years (Burt et al., 2006; Turner & Piquero, 2002). However, there is more evidence of change in early adulthood than during adolescence (despite perceptions of the turbulent teen years in which identity is in flux; see Caspi et al., 2005). This might, in part, explain why the results of the current study (and extant research; see Hay & Forrest, 2006; Higgins et al., 2009) demonstrate considerable stability across this time period. Nonetheless, examining self-control over longer time periods that traverse different developmental stages may provide a better picture of the development of self-control and identify more change and unique factors responsible for that change as individuals enter into different developmental periods. For example, Forrest and Hay (2011) recently linked observed changes in self-control from adolescence to early adulthood to marital status. In sum, it would be informative to take a life course perspective and assess the stability of self-control using longitudinal data that extend from childhood to adulthood and link changes to important life transitions.

In sum, the current study provides important insights into the development of self-control, suggesting that some individuals acquire greater levels of self-control later than others and some develop self-control at different rates. Based on the current analyses and extant literature, self-control does not appear to be stable for everyone, both in an absolute and relative sense. Although it is not possible to determine the process in which the factors identified in the current study lead to different developmental patterns of self-control, it is clear that these risk factors do vary across the different trajectories—factors other than parenting. The current findings, coupled with the few other studies on this topic, indicate that unique developmental patterns of self-control exist. The challenge posed now is to place these empirical findings into a larger, more theoretically diverse framework that considers stability and change, as well as various risk and protective factors.

Footnotes

Appendix

Notes

James V. Ray is a postdoctoral fellow at the University of New Orleans. He received his PhD from the University of South Florida in 2011. His research interests include personality development and antisocial outcomes.

Shayne Jones is an associate professor of criminology at the University of South Florida. He received his PhD from the University of Kentucky in 2003. His research interests include personality and person–environment interactions.

Thomas A. Loughran is an assistant professor of criminology and criminal justice at the University of Maryland. His research interests include offender decision-making, deterrence, and methods for inferring treatment effects from nonexperimental data.

Wesley G. Jennings is an assistant professor in the College of Behavioral and Community Sciences in the Department of Criminology and has a Courtesy Assistant Professor Appointment in the Department of Mental Health Law and Policy at the University of South Florida. He received his doctorate degree in criminology from the University of Florida in 2007. He has published over 80 peer-reviewed articles, and his major research interests include longitudinal data analysis, semiparametric group-based modeling, sex offending, gender, race/ethnicity, and the victim-offender overlap.

References

Arneklev

B. J.

Cochran

J. K.

Gainey

R. R.

(1998). Testing Gottfredson and Hirschi’s “low self-control” stability hypothesis: An exploratory study. American Journal of Criminal Justice, 23, 107-127.

Beaver

K. M.

DeLisi

Vaughn

M. G.

Wright

J. P.

(2010). The intersection of genes and neuropsychological deficits in the prediction of adolescent delinquency and low self-control. International Journal of Offender Therapy and Comparative Criminology, 54, 22-42.

Beaver

K. M.

Ratchford

Ferguson

C. J.

(2009). Evidence of genetic and environmental effects on the development of low self-control. Criminal Justice and Behavior, 36, 1158-1172.

Beaver

K. M.

Schutt

J. E.

Boutwell

B. B.

Ratchford

Roberts

Barnes

J. C.

(2009). Genetic and environmental influences on levels of self-control and delinquency peer affiliation: Results from a longitudinal sample of adolescent twins. Criminal Justice and Behavior, 36, 41-60.

Beaver

K. M.

Wright

J. P.

Maume

M. O.

(2008). The effect of school classroom characteristics on low self-control: A multilevel analysis. Journal of Criminal Justice, 36, 174-181.

Blokland

A. A. J.

Nieuwbeerta

(2005). The effects of life-circumstances on longitudinal trajectories of offending. Criminology, 43, 1203-1240.

Blonigen

D. M.

(2010). Explaining the relationship between age and crime: Contributions from the developmental literature on personality. Clinical Psychology Review, 30, 89-100.

Burt

C. H.

Simons

R. L.

Simons

L. G.

(2006). A longitudinal test of the effects of parenting and the stability of self-control: Negative evidence for the general theory of crime. Criminology, 44, 353-396.

Caspi

Robert

B. W.

Shiner

R. L.

(2005). Personality development: Stability and change. Annual Review of Psychology, 56, 453-484.

10.

Chung

Hill

K. G.

Hawkins

Gilchrist

L. D.

Nagin

D. S.

(2002). Childhood predictors of offense trajectories. Journal of Research in Crime and Delinquency, 39, 60-90.

11.

Cote

Tremblay

R. E.

Nagin

Zoccolillo

Vitaro

(2002). The development of impulsivity, fearfulness, and helpfulness during childhood: Patterns of consistency and change in the trajectories of boys and girls. Journal of Child Psychology and Psychiatry, 43, 609-618.

12.

DeLisi

Hochstetler

Murphy

D. S.

(2003). Self-control behind bars: A validation study of the Grasmick et al. scale. Justice Quarterly, 20, 241-264.

13.

D’Unger

A. V.

Land

K. C.

McCall

P. L.

Nagin

D. S.

(1998). How many latent classes of delinquent/criminal careers? Results from mixed Poisson regression analyses. The American Journal of Sociology, 103, 1593-1630.

14.

Fergusson

D. M.

Horwood

L. J.

Nagin

D. S.

(2000). Offending trajectories in a New Zealand cohort. Criminology, 38, 525-552.

15.

Forrest

Hay

(2011). Life-course transitions, self-control and desistance from crime. Criminology & Criminal Justice, 11, 487-513.

16.

Gibbs

J. J.

Giever

Higgins

G. E.

(2003). A test of Gottfredson and Hirschi’s general theory using structural equation modeling. Criminal Justice and Behavior, 30, 441-458.

17.

Gottfredson

Hirschi

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

18.

Hay

(2001). Parenting, self-control, and delinquency: A test of self-control theory. Criminology, 39, 707-736.

19.

Hay

Forrest

(2006). The development of self-control: Examining self-control theory’s stability hypothesis. Criminology, 44, 739-774.

20.

Higgins

G. E.

Jennings

W. G.

Tewksbury

Gibson

C. L.

(2009). Exploring the link between low self-control and violent victimization trajectories in adolescents. Criminal Justice and Behavior, 36, 1070-1084.

21.

Hope

T. L.

Chapple

C. L.

(2004). Maternal characteristics, parenting, and adolescent sexual behavior: The role of self-control. Deviant Behavior, 26, 25-45.

22.

Johnson

Hicks

B. M.

McGue

Iacono

W. G.

(2007). Most of the girls are alright, but some aren’t: Personality trajectory groups from ages 14 to 24 and some associations with outcomes. Journal of Personality and Social Psychology, 93, 266-284.

23.

Jones

B. L.

Nagin

D. S.

Roeder

(2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods & Research, 29, 374-393.

24.

Lamiell

J. T.

(1981). Toward an idiothetic psychology of personality. American Psychologist, 36, 276-289.

25.

Laub

J. H.

Nagin

D. S.

Sampson

R. J.

(1998). Trajectories of change in criminal offending: Good marriages and the desistance process. American Sociological Review, 63, 225-238.

26.

Longshore

Turner

Stein

J. A.

(1996). Self-control in a criminal sample: An examination of construct validity. Criminology, 34, 209-228.

27.

Maldonado-Molina

M. M.

Piquero

A. R.

Jennings

W. G.

Bird

Canino

(2009). Trajectories of delinquent behaviors among Puerto Rican children and adolescents at two sites. Journal of Research in Crime and Delinquency, 46, 144-181.

28.

Meldrum

R. C.

(2008). Beyond parenting: An examination of the etiology of self-control. Journal of Criminal Justice, 36, 244-251.

29.

Mezzich

A. C.

Tarter

R. E.

Giancola

P. R.

Kirisci

(2001). The dysregualtion inventory: A new scale to assess the risk for substance use disorder. Journal of Child and Adolescent Substance Use, 10, 35-43.

30.

Mitchell

Mackenzie

D. L.

(2006). The stability and resiliency of self-control in a sample of incarcerated offenders. Crime & Delinquency, 52, 432-449.

31.

Mroczek

D. K.

(2007). The analysis of longitudinal data in personality research. In Robins

R. W.

Fraley

R. C.

Krueger

R. F.

(Eds.), Handbook of research methods in personality psychology (pp. 543-556). New York: Guilford Press.

32.

Nagin

(1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4, 139-157.

33.

Nagin

D. S.

(2005). Group-based modeling of development. Cambridge, MA: Harvard University Press.

34.

Nagin

D. S.

Farrington

D. P.

Moffitt

T. E.

(1995). Life-course trajectories of different types of offenders. Criminology, 33, 11-40.

35.

Nagin

D. S.

Land

K. C.

(1993). Age, criminal careers, and population heterogeneity: Specification and estimation of a nonparametric, mixed Poisson model. Criminology, 31, 327-362.

36.

Nagin

D. S.

Tremblay

R. E.

(2005). Developmental trajectory groups: Fact or useful statistical fiction? Criminology, 43, 873-904.

37.

Perrone

Sullivan

C. J.

Pratt

T. C.

Margaryan

(2004). Parental efficacy, self-control, and deviance: A test of a general theory of crime on a nationally representative sample of youth. International Journal of Offender Therapy and Comparative Criminology, 48, 298-312.

38.

Piquero

A. R.

(2008). Taking stock of developmental trajectories of criminal activity over the life course. In Liberman

A. M.

(Ed.), The long view of crime: A synthesis of longitudinal research (pp. 23-78). New York: Springer Science + Business Media.

39.

Piquero

A. R.

Blumstein

Brame

Haapanen

Mulvey

E. P.

Nagin

D. S.

(2001). Assessing the impact of exposure time and incapacitation on longitudinal trajectories of criminal offending. Journal of Adolescent Research, 16, 54-74.

40.

Piquero

A. R.

Jennings

W. G.

Farrington

D. P.

(2010). On the malleability of self-control: Theoretical and policy implications regarding a general theory of crime. Justice Quarterly, 27, 803-883.

41.

Piquero

A. R.

Rosay

A. B.

(1998). The reliability and validity of Grasmick et al.’s self-control scale: A comment on Longshore et al. Criminology, 36, 157-174.

42.

Polakowski

(1994). Linking self- and social control with deviance: Illuminating the structure underlying a general theory of crime and its relation to deviant activity. Journal of Quantitative Criminology, 10, 41-78.

43.

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.

44.

Raffaelli

Crockett

L. J.

Shen

(2005). Developmental stability and change in self-regulation from childhood to adolescence. The Journal of Genetic Psychology, 166, 54-75.

45.

Raudenbush

S. W.

(2005). How do we study “what happens next”? Annals of the American Academy of Political and Social Science, 602, 131-144.

46.

Roberts

B. W.

Caspi

Moffitt

T. E.

(2001). The kids are alright: Growth and stability in personality development from adolescence to adulthood. Journal of Personality and Social Psychology, 81, 670-683.

47.

Robins

R. W.

Fraley

R. C.

Roberts

B. W.

Trzesniewski

K. H.

(2001). A longitudinal study of personality change in young adulthood. Journal of Personality, 69, 618-640.

48.

Sampson

R. J.

Laub

J. H.

(2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 301-340.

49.

Sampson

R. J.

Laub

J. H.

(2005). Seductions of method: Rejoinder to Nagin and Tremblay’s “Developmental trajectory groups: Fact or fiction.” Criminology, 43, 905-913.

50.

Schaeffer

C. M.

Petras

Ialongo

Poduska

Kellam

(2003). Modeling growth in boys’ aggressive behavior across elementary school: Links to later criminal involvement, conduct disorder, and antisocial personality disorder. Developmental Psychology, 39, 1020-1035.

51.

Tellegen

(2006). Manual for the Multidimensional Personality Questionnaire. Minneapolis: University of Minnesota Press.

52.

Turner

M. G.

Piquero

A. R.

(2002). The stability of self-control. Journal of Criminal Justice, 30, 457-471.

53.

Turner

M. G.

Piquero

A. R.

Pratt

T. C.

(2005). The school context as a source of self-control. Journal of Criminal Justice, 33, 327-339.

54.

Unnever

J. D.

Cullen

F. T.

Agnew

(2006). Why is “bad” parenting criminogenic? Implications from rival theories. Youth Violence and Juvenile Justice, 4, 3-33.

55.

Unnever

J. D.

Cullen

F. T.

Pratt

T. C.

(2003). Parental management, ADHD, and delinquent involvement: Reassessing Gottfredson and Hirschi’s general theory. Justice Quarterly, 20(3), 471-500.

56.

Vazsonyi

Huang

(2010). Where self-control comes from: On the development of self-control its relationship to deviance over time. Developmental Psychology, 46(1), 245-257.

57.

Weisner

Capaldi

D. M.

(2003). Relations of childhood and adolescent factors to offending trajectories of young men. Journal of Research in Crime and Delinquency, 40, 231-262.

58.

Wilcox

Tillyer

M. S.

Fisher

B. S.

(2009). Gendered opportunity? School-based adolescent victimization. Journal of Research in Crime and Delinquency, 49, 245-269.

59.

Winfree

L. T.

Taylor

T. J.

Esbensen

(2006). Self-control and variability over time: Multivariate results using a 5-year, multisite panel of youths. Crime & Delinquency, 52, 253-286.

60.

Wright

Beaver

(2005). Do parents matter in creating self-control in their children? A genetically informed test of Gottfredson and Hirschi’s theory of low self-control. Criminology, 43, 1169-1202.

61.

Wright

Beaver

DeLisi

Vaughn

(2008). Evidence of negligible parenting influences on self-control, delinquent peers, and delinquency in a sample of twins. Justice Quarterly, 25, 544-569.

62.

Yun

Walsh

(2010). The stability of self-control among South Korean adolescents. International Journal of Offender Therapy of Comparative Criminology, 55, 1-15.