xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2011 flu ob 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0004962
Ray, James Vance.
Developmental trajectories of self-control :
h [electronic resource] /
b assessing the stability hypothesis
by James Vance Ray.
[Tampa, Fla] :
University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 144 pages.
(Ph.D.)--University of South Florida, 2011.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
ABSTRACT: A key proposition of Gottfredson and Hirschi's (1990) self-control theory is the stability hypothesis which suggests that an individual's level of self-control, once established between the ages of 8-10, is stable over the life course. Empirical results from examinations of the stability hypothesis have been mixed. Prior tests of the stability hypothesis have employed aggregate assessment methods (e.g., mean-level and correlational analyses). Such approaches fail to take into account the possibility that individual developmental pathways may differ. This study employs individual longitudinal data over a four year period for 3,249 7th to 10th grade subjects to assess the stability hypothesis using both traditional stability estimation techniques (e.g., ANOVAs and zero-order correlations), as well as heterogeneity assessment methods semiparametric group-based trajectory modeling (SPGM). Multinomial logistic regression (MLOGIT) of theoretically and empirically relevant risk factors (i.e., parenting, parental criminality, deviant peers, bonds to school) was employed to distinguish between developmental trajectories. SPGM results suggest that self-control is stable for a majority of the sample; however, a sizeable portion of the sample evinced trajectories for which self-control was marked by considerable change. Specifically, 6 unique trajectories in the development of self-control were identified two groups were identified with high stable trajectories of self-control and four groups were identified that had lower, less stable trajectories of self-control. Additionally, several risk factors differentiated these groups. The results indicate that those with lower, less stable trajectories have more deviant peer association, higher rates of parental criminality, less intense bonds to school, and lower levels of parenting. These results indicate that self control is not stable nor is it consistent across groups, leading to a rejection of Hirschi and Gottfredson's explanation.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Jones Lynch, Shayne Michael J.
Jones Lynch, Shayne Michael J.
t USF Electronic Theses and Dissertations.
Developmental Trajectories of Self-Control: Assessing the Stability Hy pothesis By James V. Ray A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Criminology College of Behavioral and Community Sciences University of South Florida Co Major Professor: Shayne Jones, Ph.D. Co Major: Michael J. Lynch, Ph.D. Ojmarrh Mitchell, Ph.D. Tom Loughran, Ph.D. Wesley Jennings, Ph.D. Date of Approval: August 8, 2011 Keywords: antisocial behavior, trajectory analysis, risk factors, absolut e stability, relative stability Copyright 2011, James V. Ray
DEDICATION I dedicate this dissertation to my wife, Phoebe. I could not have done this without her continued financial and moral support. Thank you for being an understanding wife and friend. I also dedicate this to my son, Grant, for whom I am grateful for. I a m deeply indebted to both of you and I love you very much. I would also like to thank the rest of my family for their moral support and encouragement throughout my graduate tra ining.
ACKNOWLEDGMENTS I would like to thank my mentor and major professor, Dr. Shayne Jones for providing me with the guidance and skills necessary to fulfill the dissertation requirement. I would also like to thank the rest of my dissertation committee: Dr Michael Lynch, Dr. Wesley Jennings, Dr. Tom Loughran, and Dr. Ojmarrh Mitchel l for their hard work and time on this project. Also, a special thanks to Dr. Roger Boothroyd for chairing the dissertation defense and providing his comments. Finally, I would li ke to acknowledge the National Institute of Drug Abuse (NIDA) for the funding they pr ovided (Grant #DA-11317) for the project which resulted in the dataset used in the disserta tion and Dr. PamWilcox and Dr. Graham Ousey for providing the data.
TABLE OF CONTENTS LIST OF TABLES ............................................................................................................. iii LIST OF FIGURES ........................................................................................................... iv ABSTRACT ......................................................................................................................v CHAPTER 1: INTRODUCTION ........................................................................................1 CHAPTER 2: BACKGROUND ON SELF-CONTROL THEORY ...................................9 Theoretical propositions of self-control theory........................................................9 Literature review: The empirical status of self-control theory ..............................17 The dimensionality of the self-control construct .......................................17 Research examining the outcomes of low self-control ..............................20 Where does self-control come from? .........................................................27 Empirical status of the stability hypothesis ...............................................35 A summary of the empirical status of self-control theory .........................47 Factors distinguishing developmental patterns of self-control ..................49 CHAPTER 3: METHODS .................................................................................................53 Data collection procedures .....................................................................................53 Participants .............................................................................................................54 Measures: Dependent Variable ..............................................................................58 Self-control ................................................................................................58 Measures: Risk Factors ..........................................................................................58 Parenting ....................................................................................................59 Delinquent peer association .......................................................................61 School bonding ..........................................................................................61 Parental crime ............................................................................................62 Control variables ........................................................................................62 CHAPTER 4: ANALYTIC PLAN ....................................................................................63 Assessing Absolute Stability..................................................................................64 Assessing Relative Stability ...................................................................................64 Identification of Developmental Trajectories ........................................................65 Background on group-based trajectory modeling ......................................65 Statistical formulas.....................................................................................68 Developmental Trajectories of Self-Control ..........................................................69
Identification of Risk Factors .................................................................................71 Hypotheses .............................................................................................................73 Study Strengths and Potential Limitations .............................................................74 CHAPTER 5: RESULTS ...................................................................................................77 Mean-Level Analysis (absolute stability) ..............................................................77 Stability Coefficients (relative stability) ................................................................79 Group-Based Trajectory Modeling ........................................................................80 Model selection ..........................................................................................80 Group Differences in Self-Control.........................................................................85 Risk Factors ...........................................................................................................87 Mean differences in risk factors .................................................................87 Multinomial Logistic Regression ...........................................................................91 Relative differences in risk-factors across the six identified trajectory groups ............................................................................91 Relative differences in risk-factors for stable, increasing, and decreasing groups...........................................................................96 CHAPTER 6: DISCUSSION .............................................................................................99 Summary of Findings ...............................................................................100 Implications for Theory and Research .....................................................104 Implications for Policy and Practice ........................................................110 Limitations ...............................................................................................113 CHAPTER 7: CONCLUSION ........................................................................................119 CHAPTER 8: LIST OF REFERENCES..........................................................................122 APPENDIX A: LIST OF MEASURES ...........................................................................134
LIST OF TABLES TABLE 1: Descriptive statistics for study variables..........................................................56 TABLE 2: Repeated measure ANOVAs for self-control ..................................................78 TABLE 3: Correlations between self-control measured at each grade ..............................79 TABLE 4: Adjusted and Unadjusted Bayesian Information Criteria (BIC) b y number of groups specified ............................................................................81 TABLE 5: Mean and median posterior probabilities for six-group model ........................85 TABLE 6: One-way ANOVAs for self-control across groups ..........................................86 TABLE 7: One way ANOVAs examining gender, minority status, peer association, parenting, parental criminality, school bonds, and self-control by group .............................................................................................88 TABLE 8: Multinomial logistic regression: group membership by risk factor s ...............92 TABLE 9: Multinomial logistic regression examining differences in risk fa ctors across increasing, decreasing, stable groups ..................................................97
LIST OF FIGURES FIGURE 1: Mean levels of self-control by grade ..............................................................77 FIGURE 2: Trajectories of self-control .............................................................................83
ABSTRACT A key proposition of Gottfredson and HirschiÂ’s (1990) self-control theory is the stability hypothesis which suggests that an individualÂ’s level of self-control, onc e established between the ages of 8-10, is stable over the life course. Empirical r esults from examinations of the stability hypothesis have been mixed. Prior tests of the stability hypothesis have employed aggregate assessment methods (e.g., mean-level a nd correlational analyses). Such approaches fail to take into account the possibility that individual developmental pathways may differ. This study employs individual longitudinal data over a four year period for 3,249 7 th to 10 th grade subjects to assess the stability hypothesis using both traditional stability estimation techniques ( e.g., ANOVAs and zero-order correlations), as well as heterogeneity assessment me thods Â– semiparametric group-based trajectory modeling (SPGM). Multinomial l ogistic regression (MLOGIT) of theoretically and empirically relevant ri sk factors (i.e., parenting, parental criminality, deviant peers, bonds to school) was employed to distinguish between developmental trajectories. SPGM results suggest that s elf-control is stable for a majority of the sample; however, a sizeable portion of the sample e vinced trajectories for which self-control was marked by considerable change. Spe cifically, 6 unique trajectories in the development of self-control were identified Â– two groups were identified with high stable trajectories of self-control and four groups we re identified that had lower, less stable trajectories of self-control. Additionally, several r isk factors
differentiated these groups. The results indicate that those with lower, less stable trajectories have more deviant peer association, higher rates of parental c riminality, less intense bonds to school, and lower levels of parenting. These results indicate that se lf control is not stable nor is it consistent across groups, leading to a rejection of Hir schi and GottfredsonÂ’s explanation.
CHAPTER 1: INTRODUCTION Self-control theory is one of the most influential individual-level explanations of crime and delinquency (Pratt & Cullen, 2000). Even those who have been most critical of the theory recognize its contributions (Akers, 1991). Less than nine years afte r its publication, Gottfredson and HisrchiÂ’s (1990) A General Theory of Crime the book which introduced self-control theory, was one of the most highly cited criminologi cal works (Cohn & Farrington, 1999). Additionally, Hirschi and Gottfredson have been ranked the first and third most cited criminologists, respectively, since 1991 (C ohn & Farrington, 1998), much of which can be attributed to the attention that self-control theory has received. The major tenet of Gottfredson and HirschiÂ’s (1990) theory states that individua l differences in offending are due to an inability (or unwillingness) to refra in from criminal and analogous behaviors. Specifically, individuals who offend are less concerned about the long-term consequences of their actions. In turn, such individuals take advantag e of criminal opportunity because it is quick and easy and does not require long-term dedication. Gottfredson and Hirschi refer to this tendency as low self-control and t hey suggest that it is the sole explanation of criminal behavior. This component of the theory has received extensive attention. In their meta-analysis, Pratt and Culle n (2000) reported significant weighted mean effect sizes for studies examining the relati onship between
self-control and antisocial outcomes ranging from .22 .29; however, they also found t he effects of social learning variables to explain significant variance i n outcomes while taking into account the effects of self-control. Thus, their findings question the vali dity of Gottfredson and HirschiÂ’s claims regarding self-control as the sole ex planation of crime as well as its explanatory power. Although less abundant, research has recently begun to focus on other tenets of the theory, such as the measurement, origins, and development of self-control. The overall findings from these studies have been inconsistent regarding support for thes e different propositions, questioning its validity. Specifically, Gottfredson a nd Hirschi state that low self-control consists of several behavioral characterist ics and that these components tend to manifest within the same individuals, suggesting that self-control is a unidimensional construct. Research examining the unidimensionality of the self -control construct, however, has been inconsistent (e.g., Arneklev, Grasmick, & Bursik, 1999; Arneklev, Grasmick, Tittle, & Bursik, 1993; Grasmick, Tittle, Bursik, & Arneklev, 1993; Tittle, Ward, & Grasmick, 2003). They also clearly state that individuals are not bor n with self-control and that it must be acquired. They specify parental socializa tion as the primary source of self-control. Again, research regarding this proposition ha s been inconsistent. While empirical evidence exists for parenting as an importa nt factor in the development of self-control (see e.g., Perrone, Sullivan, Pratt, & Margaryan, 2004) other institutions have also been found to be important in its development (see e.g., Turner, Piquero, & Pratt, 2005). Another important proposition of self-control theory and the focus of the current study is the stability hypothesis. Gottfredson and Hirschi imply that selfcontrol is a
stable trait once it becomes set between the ages of 8 Â– 10. Although they are not clear regarding the nature of the stability of self-control, most researchers have interpreted their claims as being one of relative rather than absolute stability (Mi tchell & Mckenzie, 2006; Yun & Walsh, 2010). This means that while there may be some change in self-control, individuals change at the same rate and in the same direction. Thus, once established, individuals maintain their rank-order positioning on the self-control continuum over time within a given cohort. Testing the stability hypothesis has important implications for both policy and theory such as the continuity in offending debate (Cohen & Vila, 1996; Nagin & Paternoster, 2000). Gottfredson and Hirschi clearly state that offending is not a necessary outcome of low self-control, and that behaviors analogous to offending (i.e., drinking, gambling, and risky sexual behavior) may also be manifestations of low self-c ontrol. finding self-control to be stable would suggest that continuity in offending may be explained by low self-control. Therefore, finding self-control to be stable would lend credence to a population heterogeneity perspective. In contrast, finding that self-control is not stable would suggest that a state dependence explanation of continuity in offendi ng may be more appropriate. More importantly, theoretical propositions about the stabi lity of self-control and associated claims (e.g., age at which self-control i s set and socialization agents responsible for its development) would have to be reconsidere d if it were found that changes in self-control occurred beyond the age the theory contends it should remain stable. The stability of self-control also has important policy implications. Finding that self-control is stable would emphasize the need for interventions that target early
developmental periods. Because Gottfredson and Hirschi claim that self-control is set before adolescence, interventions designed to help foster the development of self-c ontrol would need to be implemented prior to that period. In one of the few studies addressing the policy implications of self-control theory, Piquero, Jennings, and Farrington (2010) conducted a meta-analysis assessing the effectiveness of interventions designed to increase self-control among youth. Their findings suggest that interventions can increase self-control and, in turn, reduce offending. The stability hypothesis has been subject to several recent empirical test s. The conclusions of these studies have been mixed. Some studies have found that self-control is stable over time (Arneklev, Cochran, & Gainey, 1998; Turner & Piquero, 2002), while others have concluded it to be only modestly stable (Burt, Simons, & Simons, 2006; Mitchell & MacKenzie, 2006; Winfree, Taylor, He, & Esbensen, 2006). One limitati on of these studies, however, is that they employed non-differentiated stability assessment methods (e.g., mean-level and correlational analyses). Such methods assume that everyone follows the same developmental patterns in self-control and fail to account for the possibility of heterogeneity in developmental patterns within a population. There are also important policy implications associated with finding uni que developmental patterns of self-control. If, contrary to a self-control expla nation, differential developmental patterns exist, (i.e., not everyone follows the sam e developmental trajectory regardless of their level of self-control or the relative stability of social control), then this opens the search for the causes and correlates of deve lopmental patterns. Identifying factors associated with developmental patters i n self-control may
aid in developing policies and programs designed to promote self-control by identi fying those life domains that may be the most effective targets for such intervent ions. Despite its importance, few studies have directly examined if individuals fol low unique self control developmental trajectories; however, indirect evidence provi des support for their existence. For example, research on personality development has found that not everyone follows normative patterns in the development of antisocial personali ty traits (Blonigen, 2010). Studies that have identified unique trajectories in offendi ng provide indirect support for similar trajectories in the development of self-control (e.g., Blokland & Nieuwbeerta, 2005; Chung, Hawkins, Gilchrist, Hill, & Nagin, 2002; DÂ’Unger, Land, McCall, & Nagin, 1998; Fergusson, Horwood, & Nagin, 2000; Laub, Nagin, & Sampson, 1998; Nagin, Farrington, & Moffitt, 1995; Nagin & Land, 1993)Only two studies directly examine if individuals follow different patterns in the development of self-control (Hay & Forrest, 2006; Higgins, Jennings, Tewksbury, & Gibson, 2009). Both studies indicate the existence of distinct developmental trajectories i n self-control. However, neither study examined whether potential risk factors distinguish betw een self control trajectories. The current study attempts to replicate the findings of the prior two studies that have examined trajectories of self-control (Hay & Forrest, 2006; Higgins et al., 2009) and also attempts to expand upon these studies by examining if certain factors dis tinguish development trajectory groups identified. Given that only two studies have examined whether there are different trajectories of self-control, it is important replicate those findings. If similar results are obtained in the current study, it would suggest the general theory requires modification to account for them. The assessment of risk factor s, if
groups are found, will be an important addition to the existing literature. Specific ally, if risk factors are found that can distinguish between trajectories, then this too w ould suggest the general theory is in need of substantial modification. In order to achi eve these goals, the current study employs a semi-parametric group-based model ing approach specifically designed to account for the identification of groups that follow di stinct developmental patterns in behaviors or traits (Land, McCall, & Nagin, 1996; Nagin, 1999). The results from this analysis will be compared to methods assessing stabi lity using more traditional methods, including stability coefficients and mean-l evel analyses. The current study will also examine whether certain risk and protective fa ctors (i.e., parenting style, parental criminality, delinquent peer associations and sc hool socialization) distinguish developmental trajectories in self-control if the se difference are discovered. The methodology used in the current study to model developmental trajectories is ideal for such an analysis, allowing for groups with disti nct development patterns to be identified and then examining if risk and protective factors differ across these groups. In the following chapters the current study is described in more detail. Befor e doing so, the theoretical background and empirical literature are presented. I n chapter two, the major propositions of Gottfredson and HirschiÂ’s theory are presented. Firs t, the concept of self-control as defined in the theory is described mainly with regards t o its composition. Second, the theoretical claims regarding the relationship between low s elfcontrol and antisocial behavior are presented. Third, the theoretical propositi on regarding the acquisition of self-control (i.e., the development of self-control) is examined Finally, Gottfredson & HirschiÂ’s claim that self-control is a stable trait is discussed.
n In chapter three, the empirical status of each of these major propositions is reviewed. First, the literature examining the construct validity of selfcontrol is reviewed focusing on its structure and dimensionality. Second, a review of the extensive body of research examining the association between self-control and crime is prese nted. Third, research examining Gottfredson and HirschiÂ’s claims regarding the developm ent of selfcontrol is reviewed with a specific focus on studies that assess parenting as the sole contributor and also those that assess the contribution of other social institutions. Final ly, the literature testing the stability of self-control is presented by fi rst discussing those studies that use more traditional methods, followed by a discussion of those that empl oy methods that account for heterogeneity in developmental patterns. Also, resea rch on personality development and offending trajectories is presented as indirect evidence for the identification of distinct development trajectories in self-control. Chapter four presents the current study and its ability to address gaps and limitations of previous research in this area. In order to do this, a detailed descr iption of the methodology used in the current study is presented. Specifically, this section describes the data collection procedures, the characteristics of the sample o btained in the current study, the measures used to operationalize constructs included in the analyse s, and the analytic procedures employed in the current study to identify groups w ith unique developmental patterns of self-control and the effect of associated risk/prot ective factors. Chapter five presents the results from these analyses. This includes a r eview of the major findings and how they fit with existing research, as well as the l imitations of the current study. Also, the implications relevant to both theory and policy will be
discussed, along with suggestions for future research examining the stabilit y of selfcontrol.
CHAPTER 2: BACKGROUND ON SELF-CONTROL THEORY Theoretical Propositions of Self-Control Theory The main component of Gottfredson and HirschiÂ’s theory is the concept of self-control and its relationship to crime. According to the theory, low self-control is a n underlying tendency behind antisocial behavior (also see, Thaler & Shefrin, 1981 for an economic model of self-control). Broadly speaking, low self-control refers t o an inability to refrain from engaging in acts that provide immediate pleasure with lit tle concern for long-term consequences. In contrast, individuals with self-control are more l ikely to refrain from certain precarious opportunities and value the pursuit of long-term go als through conventional means. Based on this definition, Gottfredson and Hirschi distinguish the concept of criminality, which assumes that some individuals have a natural proclivity toward crime, from their concept of propensity or low self-cont rol, which only suggests a propensity to disregard the long-term consequences of certa in behaviors. Gottfredson and Hirschi begin by identifying characteristics common to al l forms of crime. They discuss six elements of crime that are consistent with their concept of low self-control. Specifically, they describe crime as being opportunistic, e asy, physical, exciting/thrilling, immediately gratifying with little long-ter m benefits, and injurious. In turn, they suggest that those who are attracted to crime would be characterized by t raits
r consistent with or complimentary to those elements of crime. In doing so they sugges t that those who engage in crime are likely to be: (1) impulsive; (2) have a prefe rence for easy tasks; (3) to be more physically oriented as opposed to mental activities ; (4) riskseeking; (5) quick-tempered, and (6) insensitive. They suggest that these six elements tend to coalesce within the same individuals. Thus, it is implied that the simultaneous existence of all these traits are required for the formation of low self-c ontrol and that this intersection of traits will more strongly predict antisocial behavior than any one of its lone components. While self-control is considered by Gottfredson and Hirschi as the most importa nt variable in explaining criminal behavior, they also discuss the role of criminal opportunity. Gottfredson and Hirschi see crime as occurring when there is an int ersection between low self-control and opportunity, with those low on self-control more likely to take advantage of criminal opportunities compared to those with high self-control. However, Gottfredson and Hirschi favor a social selection perspective and sugg est that opportunity will only increase rates of offending for individuals with low selfcontrol because they also have a tendency to engage in activities, reside in areas, and opt to hang-out where such opportunities are more prevalent compared to individuals with high self-control. Alternatively, however, a social causation model would suggest t hese are not selection effects, but rather genuine influences on behavior. Thus, while Gottfredson and Hirschi acknowledge that opportunity is a necessary condition of offending, they als o indicate that such opportunities are ubiquitous. Therefore, opportunity should exert littl e, if any, influence relative to self-control.
Given the traits characteristic of self-control, it is logical to expect t hat individuals with low self-control also engage in a wide variety of behaviors analogous to cr ime, and are more likely to be victims of crime, suffer from health problems, and experienc e unemployment. Therefore, crime is not the only, nor is it a necessary outcome of low self-control but may also include gambling, risky driving, and precarious sexual behaviors among other hazardous behaviors. Because low self-control is seen as the common factor present among all types of crime and analogous behaviors, self-contr ol theory is the quintessential attempt to develop a general explanation of crime and deviance. Gottfredson and Hirschi clearly state the sources of self-control. Aki n to control theories, self-control theory implies that all individuals are born with hedonisti c tendencies and engage in self-serving behaviors (alternatively see Kilpi nen, 1999; Mansbridge, 1990; Miller, 1999; Sen, 1977; Veblen, 1898). As a result, criminal propensity (i.e. low self-control) is not something that is gained or acquired, but rat her is inhibited. The theory, therefore, implies that all individuals are born lacking in sel fcontrol and in order for them to become respectful, law abiding citizens they must develop self-control through effective socialization. More specifically, pare nting is implicated as the sole socializing agent responsible for the development of selfcontrol. Gottfredson and Hirschi outline the components of parenting essential for the development of self-control. After a review of the empirical literature on par ental behaviors found to be consistently important in preventing deviant and criminal behaviors, they identify three specific components of parenting that are import ant in the development of self-control. First, they state that parents must monitor or supervis e their
children in order to bear witness when antisocial behavior occurs. Second, parents mus t be capable of recognizing certain behaviors as deviant or antisocial. And final ly, parents must be willing and able to apply consistent and fair punishment when that behavior occurs. All three of these aspects of parenting must occur on a consistent basis for the proper development of self-control. Self-control will not be instilled if one or more of these aspects of parenting are missing or inconsistently delivered. Gottfredson and Hirschi address the idea that parental attachment is an importa nt component of parenting necessary for the development of self-control. Specifical ly, they suggest that attachment is necessary for the implementation of the other three components of parenting (i.e., monitoring, recognition, discipline). Parents who are attached to their children will monitor, supervise, and discipline their children be cause they are invested in their children. Parents who are not attached to their child a re unconcerned about the whereabouts, associations, and behaviors of their children and, in turn, will not effectively monitor or discipline them. Thus, in the eyes of Gottfre dson and Hirschi, the parent-child attachment does not need to be reflected in measures a ttempting to capture parenting since it can be assumed when effective monitoring, recogni tion, and discipline are present. For the most part, Gottfredson and Hirschi dismiss alternative sources of socialization as being important to the development of self-control. Originally they suggested that schooling could effectively socialize children and facilit ate the development of self-control; however, they quickly dismissed this notion on the grounds that poor parenting would undermine the socialization that children experience at school. Essentially, they suggest that schools could only act to supplement parenting i n the
development of self-control and is not solely responsible for its development. More recently, however, they have suggested that the school acts more independently in the development of self-control and imply that the school may actually compensate for poor parenting (Gottfredson & Hirschi, 2003). Thus, even though they state that parental behavior is the primary determinant of self-control, they acknowledge other possibl e sources might exist. Gottfredson and Hirschi also dismiss biological or genetic factors as possi ble determinants of oneÂ’s level of self-control. Because they suggest that everyo ne is born lacking self-control and it can only be developed through proper parenting practice s, they also imply that oneÂ’s level of self-control is not determined before birth or that de fects existing at birth contribute to oneÂ’s development of self-control. From Gottfredson a nd HirschiÂ’s standpoint, the common finding that criminal parents produce crime prone children is not due to the intergenerational transmission of criminal tendencies (e .g., genetic or learned). Instead, they attribute this to the idea that parents who a re criminal are also low in self-control themselves. Parents with low self-control will a lso practice poor parenting and ultimately produce children who are low in self-control. Once self-control is set near the end of childhood, socializing agents, even parenting, are ineffective in altering oneÂ’s level of self-control. Even though e mpirical evidence exists marking the effectiveness of policies and treatment progr ams targeted at reducing antisocial behavior (see e.g., Lipsey, 1992, 1999), such interventions are not effective means of actually altering criminal propensity (i.e., low se lf-control). Accordingly, Gottfredson and Hirschi dismiss the idea that self-control can be lost or that criminal propensity can be acquired as individuals age. For example, criminal prope nsity
is unaffected by the association with delinquent peers during adolescence Â– li ke-minded youth associate with one another and those low on self-control associate with other s who are low in self-control (see e.g., Akers, 1991). Because self-control is determ ined prior to adolescence and the emergence of peer groups, self-control determines peer group formation and not the other way around. Therefore, Gottfredson and Hirschi see socia l factors that are said to be associated with crime, such as poor school attendance, de viant peer association, and drug use, as spurious since they are caused by low self-c ontrol. Even parenting is said to have only an indirect affect on offending through its effect on the development of self-control and, even then, losing its importance once self-control becomes set. Thus, beyond childhood, self-control becomes an immutable trait that is unaffected by the influences of positive or negative socialization according to t he theory. Since self-control becomes set near the end of childhood and is unaffected by socializing agents beyond this point, it is seen as a stable and enduring trait. T hose who are low on self-control will remain low on self-control over the life-course. Ac cordingly, the same individuals who are low on self-control will continue to be those most likely to engage in crime. Therefore, Gottfredson and Hirschi dismiss any causal relat ionship between prior and subsequent criminal involvement. The reason why the same individuals continue to engage in crime is because they pervasively lack self-contr ol throughout their life, not because crime increases the chances of subsequent cri me. Although Gottfredson and Hirschi state that oneÂ’s level of self-control rema ins stable, they do acknowledge normative declines found in offending that occur into adulthood and suggest that they are not necessarily reflecting increases in le vels of selfcontrol. They suggest that as individuals move into adulthood opportunities for
Â“expressingÂ” a lack of self-control differ from those in previous developmental pe riods. For example, truancy becomes failure to maintain steady employment; poor pe rformance in school becomes poor work history; and risky driving becomes gambling. While some of these acts may still be considered illegal, they are less likely to be detected. Because crime is not a necessary outcome of low self-control, explicit illegal act s committed in adolescence are replaced by other, non-criminal, or at least more covert f orms of illegal behaviors (e.g., drinking, gambling, and risky sex) in adulthood. Therefore, according t o self-control theory, it is not actually offending that is stable, but the propensi ty to do so. Moreover, according to Gottfredson and Hirschi, declines in offending do not reflec t increases in self-control. While Gottfredson and Hirschi clearly suggest that self-control is a stable trait, they are less clear regarding the nature of this stability. When referri ng to the stability of traits, many times a distinction is made between relative and absolute stabi lity (see e.g., Caspi, Roberts, & Shiner, 2005). On the one hand, absolute stability suggests that levels of some trait 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. Whi le 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 stabi lity can occur without absolute stability. It is unclear whether Gottfredson and HirschiÂ’s stability hypothesis is one of absolute or relative stability. However, a close examination of their theory justifies the
common interpretation that it is one of relative stability (e.g., Mitchell & Mackenzie, 2006; Turner & Piquero, 2002). For example, in their theory they state: Â“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.Â” Inconsistent wi th other claims about the likelihood of changes in self-control and its relationship to offending (i.e., crime is replaced with other types of behaviors), it seems they do suggest that changes in self-control may occur and that these changes may contribute, at least some what, to normative declines in offending. Elsewhere, they disregard the possibility of decr eases 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, pg 107, emphasis added). Based on these statements, it appears that Gottfredson and Hirschi hypothesize that selfcontrol is a Â“relativelyÂ” stable trait and that normative increases in self -control occur for the entire population, yet individuals experience these increases at the same rate. Those who originally had the lowest levels self-control will always be lowe r in self-control compared to others of the same age and, therefore, will always be more prone to cr ime. In sum, there are four major propositions that stem from Gottfredson and HirschiÂ’s theory. First, they see self-control as being a unidimensional con struct
n comprised of six components (i.e., impulsiveness, preference for simple tasks over complex, more physical as opposed to mental, risk-seeking, quick-tempered, and insensitive). Second, self-control is able to explain all types of criminal an d deviant behavior above and beyond other criminological variables. Third, three specific aspec ts of parenting (i.e., monitoring, supervision, and punishment) are the sole source of self-control up to ages 8 to 10. Finally, once set, oneÂ’s level of self-control will remain st able across the life-course. The strong claims made by Gottfredson and Hirschi re garding their theory and their competing approach with other theoretical perspectives have led to each these four propositions being subjected to extensive empirical scrutiny. Literature Review: The Empirical Status of Self-Control Theory The dimensionality of the self-control construct. Since its conception, there has been a considerable amount of research regarding the structure (i.e., dimensi onality) of self-control (Hirschi & Gottfredson, 1993); however, research in this area ha s sparked considerable debate and little consensus has emerged regarding the nature of se lf-control (Longshore, Turner, & Stein, 1996; Piquero & Rosay, 1998). Contributing to this lack of consensus is that interpretations of Gottfredson and HirschiÂ’s (1990) definition of se lfcontrol have been less than consistent (see e.g., Marcus, 2004). Some researchers interpret their definition as suggesting that self-control is unidimensional (e.g ., Longshore al., 1996; Piquero & Rosay, 1998), while others suggest that they describe self-control is a multidimensional construct made up of several distinct traits (e.g., Arnekl ev, Grasmick, & Bursik, 1999; Vazsonyi, Pickering, Junger, & Hessing, 2001). Thus, it is not surprising that research examining the dimensionality of self-control is som ewhat inconclusive.
In general, research has approached examining the dimensionality of self-cont rol in two ways: examination of the definitional (internal) validity of self-c ontrol and examining the predictive (external/criterion) validity of self-control The definitional validity of self-control has typically been examined through the use of factor anal ysis. Factor analytic approaches can address questions related to the internal structure of selfcontrol as captured by a specific measure. While related, predictive validity a ddresses the extent to which self-control and its separate components are related to those behaviors that it proposes to explain. Grasmick, Tittle, Bursik, and Arneklev (1993) developed one of the most widely used measures of self-control. This measure, often referred to as the Grasm ick scale, has been the subject of most of the research on the dimensionality of self-control. They created 24 items each of which was designed to be a valid indicator of one of the si x elements of self-control as described by Gottfredson and Hirschi. After admini stering their measure to a sample of university students, Grasmick and colleagues conducte d an exploratory factor analysis of all twenty-four items and found that their meas ure represented a one factor solution. Their findings were replicated by Nagin a nd Paternoster (1993), Piquero and Tibbetts (1996), and Wood, Pfefferbaum, & Arneklev (1993), who also administered the Grasmick scale to samples of university students a nd found evidence for a one-factor solution using exploratory factor analysis. Afte r factor analyzing the Grasmick scale, Piquero and Rosay (1998) extended the evidence for s elfcontrol as a unidimensional construct among an offending sample. Alternatively, studies have also found support for a multidimensional structure of self-control also using the Grasmick scale. For example, both Cochran et al ., (1998) and
Arneklev et al. (1999) found self-control to be multidimensional with six lower-or der factors that all load on one higher-order factor (i.e., Â“self-controlÂ”) amon g a sample of undergraduates. Similarly, Longshore et al. (1996), using the same data as Piquero and Rosay (1998), yet taking a slightly different approach (e.g., inclusion of corr elated error terms, modification of response sets from 4 to 5 options), also found that self-control is multidimensional, with subscales that load on a single factor. Vazsonyi et al. ( 2001) also found evidence for six lower-order factors that load on one higher-order factor of s elfcontrol and found this factor structure to be tenable across samples from different n ations. DeLisi, Hochstetler, and Murphy (2003) also found support for the hierarchical st ructure of self-control (i.e., six lower-order factors and one higher-order fac tor) as measured by the Grasmick scale. In addition, Piquero, MacIntosh, and Hickman (2000) found support for a hierarchical factor structure through the use of an item response theory Rasch model. While the findings from these studies suggest that self-control is multidimensional with lower-order subscales, they still find that they conver ge on one higher-order factor. Thus, they support the notion that these characteristics tend t o manifest within the same individuals. With regards to predictive validity, Longshore et al. (1996) found that a composite self-control scale was just as or even less predictive of crime co mpared to some of the identified subscales (e.g., risk-seeking and impulsiveness/self -centered scales). Piquero and Rosay (1998) confirmed these findings, revealing that the r iskseeking and impulsive subscales predicted offending just as well or better than the composite self-control scale. Similarly, Wood et al. (1993) found that the subscales of the Grasmick scale had differential associations with antisocial outcome s. Particularly,
r the temper scale was the best predictor of antisocial behavior compared to othe r subscales. DeLisi et al. (2003) found the temper subscale to be the only significant predictor of delinquency suggesting, again, that self-control is no better an expla natory variable than existing concepts. Alternatively, Vazsonyi et al. (2001) found tha t each of the subscales had differential associations with different forms of antisocia l behavior; however, the total self-control scale proved to be the best predictor of antisocial behavior. With the exception of one study (Vazsonyi et al., 2001), these findings suggest tha t selfcontrol may not have explanatory power beyond that of existing concepts within criminological theory. In sum, research tends to suggest that the six traits identified by Gottfre dson and Hirschi coalesce in the same individuals. However, it is unclear whether the structure of self-control is unidimensional or if it is characterized by a hierarchical structure. More concerning, however, is the fact that studies examining the relative importance of each components of self-control, as well as the global trait of self-control, tend to sugge st that they have differential explanatory power. These findings question the unique predic tive ability of self-control beyond traits already identified as important predict ors of antisocial behavior (e.g., risk-seeking, impulsivity, or anger). Research examining the outcomes of low self-control The empirical status of the relationship between self-control and antisocial behavior is presented in a me taanalysis conducted by Pratt and Cullen (2000). They empirically examined the explanatory power of low self-control based on findings from studies published within the first decade the theory was introduced. Specifically, they examined the results of 21 empirical studies, which contained 126 effect size estimates, representing t he integration
of 49,727 individual cases. The results indicated that the effect sizes of the relati onship between self-control and antisocial outcomes ranged from .22 to .29. They also conducted moderator analyses in an effort to explore what factors might influe nce the relationship between self-control and offending. The moderators in their analys es included the effects of attitudinal versus behavioral measures of self-control ; longitudinal versus cross-sectional research designs; including competing variables ( e.g., social learning and strain); controlling for opportunity; explaining crime versus anal ogous behaviors; and using samples with different characteristics (e.g., gende r, community, juveniles). Contrary to Gottfredson and HirschiÂ’s claims, it was found that the relationship between self-control and crime was not as strong when examined using longitudinal research designs as compared to cross-sectional designs. It wa s also found that social learning variables remained significant predictors of crime in studies including variables from both theories. These set of findings are not consistent with the theory Despite the salience of Pratt and CullenÂ’s (2000) meta-analysis as a mar ker for the empirical status of the theory, another decade of research on self-control has e lapsed. During this time a number of notable studies have examined the association between l ow self-control and various types of antisocial behavior using a variety of methodol ogical approaches. Studies have examined the association between self-control and offi cial reports of offending. DeLisi (2001) found low self-control to be a correlate of a varie ty of index offenses based on arrest records, including violent, property, white collar, and nuisance offenses (e.g., prostitution and disorderly conduct). Junger, West, and Timman (2001) found that individuals who engaged in risky driving behavior, which they used as a behavioral indicator of low self-control, were more than twice as likely to ha ve an
officially recorded arrest within the last five years compared to those w ho have not engaged in risky driving behavior. While these studies found support for the relationship between low self-control and official offending, they were both based on crosssectional data. A larger number of studies have examined the relationship between self-contr ol and various types of self-reported offending. Tittle, Ward, and Grasmick (2003) found that both a behavioral and attitudinal measure of low self-control to be associat ed with various types of self-reported illegal (e.g., theft and assault) and analogous (e.g., gambling) behaviors. Holtfreter, Reisig, Piquero, and Piquero (2010) also found that l ow self-control was significantly related to self-reported offending, anal ogous behaviors (e.g., academic dishonesty), and victimization. Kerley, Hochstetler, and Cope s (2009) found an association between low self-control and both self-reported prison infracti ons and self-reported victimization in prison among a sample of incarcerated offender s. While some of these studies have examined the association between low self-control and analogous behaviors (e.g., academic dishonesty and gambling), others ha ve found an association between low self-control and specific types of behaviors. For example, Gibson and Wright (2001) examined the association between low self-control and occupational delinquency. Occupational delinquency was defined as antisocial behavior occurring as part of work (e.g., stealing from employers, giving aw ay employersÂ’ goods). They found that low self-control interacted with cowor ker delinquency to have a significant effect on occupational delinquency. However, the direct effect of low self-control on occupational delinquency was not significant once other theoretically relevant variables were included in the model. More importa ntly, the
direct effect of coworker delinquency remained significant and its effect was stronger than the interaction between co-worker delinquency and self-control. Unnever and Cornell (2003) found a relationship between low self-control and bullying; however, they did not find low self-control to be associated with being a victim of bullying. Higgins and colleagues have conducted several studies examining the association betwe en low self-control and media piracy (e.g., software and movies; Higgins, 2005; Higgins, Fe ll, & Wilson, 2007; Higgins & Makin, 2004). In all three studies low self-control was found to have a significant effect on piracy; however, all three studies also showed tha t the association between low self-control and piracy was higher for those who associat e with peers who pirate media. Finally, Cochran, Aleska, and Chamlin (2006) found low self-control to be related to self-reported academic dishonesty. Other studies have attempted to examine the relationship between self-c ontrol and antisocial behavior using longitudinal data. Some studies have looked at this longitudinal relationship using officially recorded offending. Langton (2006) found that low se lfcontrol assessed at baseline predicted parole violations during a two-year f ollow-up period among a sample of youth released from secure juvenile detention faciliti es. Similarly, Piquero, MacDonald, Dobrin, Daigle, and Cullen (2005) examined the ability of self-control to predict violent offending and homicide victimization based on offic ial records of youth released from secure detention facilities. They found that low selfcontrol measured prior to institutionalization was a significant predictor of both violent offending and homicide victimization during a five-year follow-up period. Piquero, Moffitt, and Wright (2007) examined whether self-control measured at several time points across childhood and through adolescence (i.e., ages 3 Â– 18) was predictive of
delinquent and criminal convictions assessed in early adolescence (i.e., ages 13 Â– 17) a nd adulthood (i.e., ages 18 Â– 26) using official court data. Based on these data, they derived four outcomes of offending including participation (i.e., having a conviction or not), frequency (i.e., number of convictions), persistence (i.e., 2 or more convictions by age 26), and desistence (i.e., having one juvenile but no adult convictions) and found low self-control to be a predictor of all outcomes. Consistent with the work of Pratt a nd Cullen (2000), and inconsistent with Gottfredson and HirschiÂ’s (1990) suggestion, the findings indicate that the relationship between self-control and offending using longitudinal designs are not as strong as those found in cross-sectional resear ch. Other studies have examined the relationship between self-control and self-reported antisocial behavior. Cretacci (2008) found low self-control to be a predic tor of self-reported property and drug offending over two waves of data collection whi le accounting for the effects of other theoretically relevant constructs (e.g ., rational choice, bonding, and learning); however, low self-control was unrelated to violent criminal outcomes. Smith (2004) found that low self-control assessed at the beginning of the semester among a group of university students had a significant effect on acade mic dishonesty later in the semester. De Kemp et al. (2009) found that prior levels of self-control exerted a significant effect on both self-reported aggression and delinquenc y. They also found that the relationship between self-control and delinquency was reciprocal for boys, suggesting that engaging in delinquency contributes to a decr ease in self-control and subsequently leads to more offending. Other studies using longitudinal designs have examined both the direct and indirect effects of self-control on self-reported behaviors. Longshore, Chan g, Hsieh, and
Messina (2004), for example, conducted a longitudinal analysis examining the relationship between low self-control, social bonds, delinquent peer association, and drug use. They found that low self-control had a negative association with all four social bonds (i.e., attachment, belief, commitment, and involvement) and deviant peer association; however, its effect on self-reported delinquency was mediated by belief and deviant peer association. Alternatively, Chapple (2005) conducted a longitudinal analy sis and found that low self-control had both a direct and indirect effect on self-reported delinquency through deviant peer association. Similarly, Mason and WindleÂ’s (2002) longitudinal analysis examining the effects of low self-control on delinquency found t hat low self-control had both a direct and indirect (through informal social control) ef fect on subsequent delinquency. Several studies have also observed similar associations between low selfcontrol and self-reported behaviors among samples from different nations. Seipler and E ifler (2010) found that low self-control had an association with self-reported offending am ong a sample of German adults. Baron (2003) found that low self-control was associated wi th self-reported offending (i.e., property, violent, and drug offenses) among a sampl e of Canadian youth. It was also found that individuals with low self-control were somewhat more likely to associate with deviant peers, hold deviant values, be homeless, and be unemployed compared to those with high self-control. Ribeaud and Eisner (2006) found that low self-control predicted self-reported delinquency and drug use among a sa mple of Swiss youth. They also found that low self-control accounted for some of the associa tion between drug-use and crime. Ozbay and Koksoy (2008) found low self-control to have a relationship with self-reported violence net of the effects of competing the oretical
variables (e.g., strain and social bonds) among a sample of university students in Tur key. In contrast, Cheung and Cheung (2008) found that the effects of low self-control on self reported delinquency was reduced to non-significance once other theoretically r elevant variables were included in the model (e.g., labeling, deviant peers association, a nd social bonds) among a sample of Chinese youth. Other studies have examined the relative importance of low self-control in explaining antisocial behaviors for samples from different cultures and na tions. For example, Morris, Wood, and Dunaway (2006) compared the association between self-control and substance use between a group of White high school students and Native American students. They found an association between low self-control and s ubstance use among both Native Americans and Whites, although the effect was slightly hig her for Native Americans. Similarly, Vazsyoni, Wittekind, Belliston, and Van Loh (2004) found an effect of low self-control on various types of self-reported antisocial beha viors (e.g., assault, theft, and school misconduct) and that these associations were similar a cross a sample of Japanese youths and a sample of American youths. Vazsonyi, Pickerin g, Junger, and Hessing (2001) found a relationship between low self-control and self-reported delinquency across samples of youth from four different countries (Hungary Netherlands, Switzerland, and the United States). Rebellon, Straus, and Medeiros ( 2008) also examined the relationship between low self-control and self-reported offendi ng across samples of university students from 32 countries and found the association between low self-control and crime to be significant across nations. Additionall y, it was found that self-control was a better predictor of crime than delinquent peer assoc iation.
n While not exhaustive, the previous review of the literature examining the relationship between self-control and deviant behavior provides some insight into the proposition of self-control theory since Pratt and CullenÂ’s (2000) meta-analysis. In general, the findings across these studies are consistent with the finding s of Pratt and Cullen. It appears that low self-control is an indicator of crime and analogous be haviors. However, this review also reveals that low self-control tends to be only modestl y associated with such behaviors with other theoretical variables from rival explanations remaining important predictors. In sum, self-control appears to be one indicator of crime and analogous behaviors; however, it is not Â“theÂ” individual-level explanation of crime a s claimed by Gottfredson and Hirschi. Where does self-control come from? Recently, research has begun to examine the development and sources of self-control. For the most part, researchers have t ested Gottfredson and HirschiÂ’s proposition that parenting is responsible for the development of self-control. Research in this area has taken two general forms: those t hat have simply examined the association between parenting and levels of self-control (e.g., C ochran et al., 1998) and those that have examined whether low self-control mediates the relationship between parenting and antisocial outcomes (e.g., Hay, 2001; Perrone, Sullivan, Pratt, & Margaryan, 2004; Unnever, Cullen, & Agnew, 2006). Polakowski (1994) conducted one of the first empirical studies that explored the role of parenting in the development of self-control. Consistent with Gottfredson and HirschiÂ’s contention, parenting was found to be associated with level of self-control at ages 8 Â– 10, but was unrelated to level of self-control at ages 12 Â– 14. Cochran et al. (1998) examined the effects of parenting on self-control among a sample of
undergraduate college students. Contrary to expectations, however, parental attac hment was found to be related to level of self-control but parental supervision was not. Phythian, Keane, and Krull (2008) examined the effects of parenting and family str ucture on the development of self-control using a large nationally representative sample of youth. More specifically, the question was whether parental monitoring and di scipline predicted levels of self-control. While it was found that poor and inconsistent discipli ne resulted in low self-control, the relationship between monitoring and self-control wa s found to be somewhat more complicated in that it varied by family structure. Furthermore, in intact families, monitoring was not important in the development of s elfcontrol; however, in single-parent families it was found to be a significant predic tor of self-control. As previously mentioned, other attempts to study the development of self-control have conducted mediation analyses in order to more thoroughly test the claim that t he effect of parenting on crime will no longer exist once self-control is meas ured. Consistent with the theory, these studies find that parenting exerts a positive effect on self-control, suggesting that it is in fact responsible for the development of s elf-control. Additionally, the extant research finds that both parenting and level of self-control are negatively related to antisocial behavior. However, with the exception of one study in which low self-control was found to completely mediate the relationship between parenting and antisocial behavior (Gibbs, Giever, & Martin, 1998), most studies have found that self-control only partially mediates this relationship and that parenti ng continues to have a significant impact on antisocial behavior even when self-control is controlled (Gibbs, Giever, & Higgins, 2003; Hay, 2001; Perrone et al., 2004; Unnever,
Cullen & Pratt, 2003). A partial mediation effect of low self-control on the parentin gdeviance relationship was also found among a Native American sample (Morris e t al. 2007) and samples from nations other than the United States including the Netherlands (Finkenauer, Engels, & Baumeister, 2005; Vaszyoni & Belliston, 2007), Japan, Switzerland, and Hungary (Vaszyoni & Belliston, 2007). Partial mediation was al so found in one study that accounted for constructs from competing explanations (i.e., learning) that were expected to also mediate this relationship (Unnever et al., 2006). In sum, the results from these studies suggest that self-control only partially me diates the relationship between parenting and delinquency. However, these studies relied on cros ssectional designs and a more rigorous test of this would require a longitudinal analy sis in which parenting is measured prior to self-control and both are measured prior to delinquency. Currently, only two studies have employed longitudinal designs in order to examine if self-control mediates the relationship between parenting and devi ance. The findings from these studies are mixed (Burt, Simons, & Simons, 2006; Hope & Chapple, 2005). Hope and Chapple (2005) found some evidence for complete mediation. It was found that the relationships between parenting Â– both parental monitoring and parental attachment Â– and number of sexual partners were completed mediated by low self-control. They also found that the relationships between both parental monitoring and parental attachment and whether the child was in a committed or causal relat ionship were completed mediated by low self-control. They also found self-control to media te the relationship between parental attachment and being sexually active. However low selfcontrol did not mediate the relationship between monitoring and being sexually active
r Alternatively, while Burt et al. (2006) did find parenting to have a somewhat stronger effect on the development of self-control than previous studies, consistent with re sults from cross-sectional studies they found that self-control only partially medi ated the effects of parenting on delinquency. In conjunction with the findings from cross-sectional studies, the general findings suggest that parenting is in fact, at least partially responsible for the development in self-control; however, parenting also continues to be an important factor in behaviors that occur after self-control has presumably become a stable trait. It remains unclear from these studies whether parentin g is the sole source of self-control as the theory so adamantly states. In order to empirically assess Gottfredson and HirschiÂ’s claims that pare nting is the sole source of self-control, several studies have examined the importance of alternative socializing agents in the development of self-control. For exam ple, studies have examined the impact of school (Beaver, Wright, & Maume, 2008; Burt et al., 2006; Meldrum, 2008; Turner, Piquero, & Pratt, 2005), neighborhood context (Gibson, Sullivan, Jones, & Piquero, 2010; Pratt, Turner, & Piquero, 2004; Turner et al., 2005; Teasdale & Silver, 2009), and peer association (Burt et al., 2006; Meldrum, 2008) on the development of self-control. With regards to school socialization, Burt et al. (2006) found that attachment to teachers remained a significant predictor of the level of self-control even w hile controlling for parenting. In a longitudinal analysis, Beaver et al. (2008) f ound that school and classroom context (e.g., extent of misbehavior and crime) continued to explain variation in level of self-control beyond the effects of parenting. Mel drum (2008) also found that school socialization had a direct effect on the development of self-cont rol
net of the effects of parental socialization. It was found that children who att ended schools that effectively monitored their students showed increases in self-cont rol compared to students in schools with poor monitoring practices; these school effects existed above and beyond the effects of parenting. Turner et al. (2005) examined whether the effect of school socialization on level of self-control varied by parenting style. The findings suggested that schools were more effective in developi ng self-control among those experiencing poor parenting compared to those experiencing good parenting. Studies have also examined the effect of neighborhood context on the development of self-control. Consistent with Gottfredson and Hirschi, some researche rs have found that neighborhoods do not influence the development of self-control (Gibson et al. 2010; Meldrum, 2008). However, other studies have found that neighborhoods do influence the development of self-control. For instance, Pratt et al. (2004) found that adverse neighborhood conditions had a direct negative impact on parenting and level of self-control. More importantly, neighborhood context emerged as having the stronges t effect on level of self-control even while accounting for the effects of parent ing. Similarly, Teasdale and Silver (2009) found that neighborhood context (i.e., neighborhood disadvantage and collective efficacy) had both a direct and an indirect effect through parenting on levels of self-control, with those in more advantaged neighborhoods having higher levels of self-control. Alternatively, while Turner e t al. (2005) found that neighborhood context had a direct positive effect on level of self-control, they also found that parenting exerted a significant positive effect on s elf-control regardless of neighborhood context. Given this pattern of findings there are reasons to
doubt Gottfredson and HirschiÂ’s contentions regarding the influence of neighborhoods on the development of self-control. Only a few studies have examined the effect of peer association on self-con trol. Burt et al. (2006) found that increases in deviant peer association were associa ted with decreases in self-control over time. Accordingly, they found that increase s in pro-social peer association were associated with increases in self-control over t ime. While Meldrum (2008) found that deviant peer association had a negative, contemporaneous effect on the development of self-control, they failed to find an effect of deviant peer association on levels of self-control measured at later time points. Similar ly, Wright, Beaver, Delisi, and Vaughn (2008) found that delinquent peer association explained differences in levels of self-control measured both contemporaneously and prospe ctively between pairs of paternal and fraternal twins. However, delinquent peer associa tion did not explain differences in self-control within pairs of twins suggesting that the association found between twin pairs was due to a selection effect in which twin pairs with low se lfcontrol associate with delinquent peers. Studies have also tested Gottfredson and HirschiÂ’s claim against the intergenerational transmission of criminality by examining whether par enting practices mediate the relationship between parental self-control and child self-control (Boutwell & Beaver, 2010; Nofziger, 2008). Nofziger (2008) found that mothers with low self-control produced children with low self-control and that some of this association could be attributed to the type of parenting practices employed. Contrary to the theory however, a significant direct effect of maternal self-control on the childÂ’s self-c ontrol remained. In a similar study, Boutwell and Beaver (2010) examined if both mothersÂ’ and fathersÂ’ level
of self-control were related to the childÂ’s level of self-control beyond the e ffects of parenting, neighborhood context, and neuropsychological deficits. Results from hierarchical regression analysis revealed that parental self-control i n both cases was directly related to the childÂ’s level of self-control and was not mediated by any other factors. Contrary to self-control theory, both of these studies support the intergenerational transmission of self-control beyond parenting. Although Unnever et al. (2003) did not examine the intergenerational transmission of self-control, they did find that being diagnosed with Attention Deficit Hyperactivity Disorder was a si gnificant predictor of low self-control net of the effects of parenting practices, sugge sting support for a biological basis for self-control. These studies indicate that Gottf redson and Hirschi are not correct regarding the role of parenting as the singular source of self -control. Additional studies also indicate that parenting is not the only source of self-control. Several studies have directly examined whether there is a genet ic basis for level of self-control, mainly through the use of twin studies. For the most part these studie s find that once the effects of shared and non-shared environmental factors on levels of self-control are accounted for, levels of self-control are more similar for m onozygotic twins (i.e., those that share 100% of alleles) compared to dizygotic twins (i.e., those who share only 50% of alleles), suggesting that self-control is genetically tr ansferred (Beaver et al., 2009; Wright & Beaver, 2005; Wright et al., 2008). Studies have also found support for a genetic transference of self-control through the use of genotyping (Beaver, DeLisi, Vaughn, & Wright, 2010; Beaver, Ratchford, & Ferguson, 2009). Beaver and colleagues (2009, 2010) examined the effects of genes and neurological deficits on l evels of self-control. In both studies, DNA samples were taken in order to detect the presen ce
of genotypic polymorphisms. Beaver et al.(2009) examined if levels of serotonin, a neurotransmitter that controls impulses in the brain, were associated with l evel of selfcontrol. While they did not reveal a direct significant effect of serotonin on level s of selfcontrol, they did find that serotonin interacted with delinquent peer association in a w ay that predicted lower levels of self-control. Beaver et al. (2010) also examined whether neurobiological deficits (i.e., verbal ability) and genetic effects (i.e., MAOA gene activity) predicted levels of self-control. Because low MAOA gene activity has been found to be related to criminal behavior, it was expected that those with low levels of MAOA would have lower levels of self-control. Similar to the findings of Beaver et al. (2009), they found that MAOA gene activity did not have a direct effect on level of s elfcontrol. However, there was a significant interaction effect between MAOA and neurobiological deficits on self-control. The general consensus of studies examining alternative sources of self-cont rol suggests that there are multiple factors impacting the level of self-cont rol beyond parenting. Studies examining contextual effects (e.g., neighborhood/community) su ggest that parenting does not occur in a vacuum and that such factors impact self-control both directly and indirectly. Additionally, other social institutions (e.g., school) r esponsible for socializing children can be effective sources of self-control. It is also i mportant to note that these studies find sources of self-control, including parenting, to impact s elfcontrol beyond the age at which Gottfredson and Hirschi suggest that it is set. Thus, oneÂ’s level of self-control may be much more susceptible to change and less sta ble than the theory contends. In sum, the findings from these studies question the validity of
Gottfredson and HirschiÂ’s claims regarding social agents responsible for the development of self-control, as well as it malleability beyond ages 8 Â– 10. Empirical status of the stability hypothesis. Recently, studies have begun to examine the stability hypothesis using a variety of methods and analytic te chniques. Research results have been inconsistent; some find support for Gottfredson and Hirs chiÂ’s contention that self-control is stable across the life-course, while others not e that selfcontrol is not as stable as the theory contends. Research examining the absolute stability of self-control has been inconsi stent. Some studies have found that mean levels of self-control do not change. For example, Arneklev et al. (1998) did not find mean differences between self-control measured at the beginning and end of the semester among a group of undergraduates. Yun & Walsh (2010) also did not find any mean differences in self-control assessed yearly over f ive years (ages 14 Â– 18) among a sample of South Korean youth. Similarly, Raffaelli et al. (2005) did not find any mean differences between self-control measured when youth in their sample were ages 8 to 9 and subsequently at ages 12 to 13. Alternatively, some studies have found significant mean-level increases in s elfcontrol over time. Vazsonyi and Huang (2010) found an increase in mean-levels of s elfcontrol measured at three time points when children in their sample were ages 5, 9, and 11. Winfree et al. (2006) also found that mean levels of self-control gradually increas ed over five years among a sample of high school students. Finally, Turner and Piquer o (2002) found gradual mean-level increases in self-control measured at seven diff erent time points from childhood (age 7) through adolescence (age 19).
Mean-level analyses are somewhat limited in assessing stabilit y. Because meanlevel analyses assess absolute stability at the aggregate level, they may mask any withinindividual change. Therefore, the finding that means do not change over time does not necessarily indicate that change is not occurring at the individual level. I f individuals or groups of individuals are changing in opposite directions they would cancel each other out and such important changes would not be reflected in mean-level analyses. Despite the limitations of mean-level analysis, only two studies have ex amined absolute stability using methods that account for within-individual changes in self -control over time. Both Arneklev et al. (1998) and Hay and Forrest (2006) used hierarchical linear modeling (HLM) to assess within-individual changes in self-control over time. Specifically, Arneklev et al. (1998) examined between-individual differences in selfcontrol (i.e., level two) and within-individual change over time (i.e., level one). Whil e they found significant differences between individuals on self-control, there wa s no evidence of within-individual change. However, in addition to employing a short t ime period between measurements (i.e., 4 5 months), self-control was only assessed at two time points. Using HLM in this way does not allow for the possibility that self-c ontrol changes in a non-linear manner. Alternatively, Hay and Forrest (2006) conducted H LM analysis on self-control measured at five separate time points (i.e., ages 7, 9, 11, 13, 15) Not only did they find within-individual changes in absolute levels self-control, but they also found that changes in parenting were partially responsible for these change s, even past the age that the theory suggests self-control is set. Several researchers, however, have argued that the stability postulate is one of relative stability (i.e., rank-order consistency) as opposed to absolute st ability (see Hay &
n Forrest, 2006; Mitchell & McKenzie, 2006). Studies examining the relative stabilit y of self-control have typically employed the use of stability coefficient s. Stability coefficients are simply correlation coefficients between self-contr ol scores measured at two time points. Research on the development of personality has suggested that correlation coefficients of .60 or higher are indicative of relative stabili ty (Costa & McCrae, 1988). Based on this criterion, some studies have found support for Gottfredson and HirschiÂ’s theoretical proposition that self-control is relatively sta ble, at least over the short term. For example, Arneklev et al. (1998) and Beaver et al. (2008) found scores on a measure of self-control assessed at two separate time points to be highly correlated (i.e., r Â’s = .82 and .64, respectively). Vazsonyi and Huang (2010) found self-control assesse d between ages 8 to 12 to be correlated at .70. Alternatively, several studies have not supported the relative stability of self-control by the same criterion. For ex ample, Burt et al. (2006) found that self-control assessed when youth were ages 10-12 to be correlated with self-control at ages 12-14 at only 48. Using a two-wave panel design, Mit chell and McKenzie (2006) found scores on a measure of self-control to be correlated at .48 over a period of six months. Based on data collected from samples during similar developmental periods (i.e., ages 8 Â– 12), both Polakowski (1994), and Raffaelli et al. (2005) found self-control measured at the beginning of this period to be correlated with self-control measured at the end of this period at .59 and .50, respectively. Other studies have found that as time increases between assessment period s, the relative stability of self-control decreases, suggesting that self-cont rol is less stable over the long-term. Yun and Walsh (2010) reported correlations for self-control measured yearly at five waves. They found that the correlation decreased over time ( wave one and
two, r = .52; wave one and waves three, four, and five r Â’s = .47, .42, and .42, respectively). Winfree et al. (2006) produced similar findings for self-contr ol measured yearly over five waves ( r Â’s = .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. In sum, studies assessing the relative stability of self-control through the use of stability coefficients question Gottfredson and HirschiÂ’s claim that i ndividuals will remain within their rank-order position on self-control over time. The results from these studies tend to produce correlations that do not reach the suggested criterion of .60. Additionally, the few studies that do find coefficients that meet the .60 cutoff tend t o be those that have relied on shorter time periods between data points. This may, in turn, inflate the magnitude of the correlations as study participants are subject t o measurement recall. It is important to note the limitations of using stability coefficients in as sessing the relative stability of self-control. Similar to mean-level analys is, correlations are aggregate measures of stability and, therefore, might conceal individual diffe rences in stability (Lamiell, 1981; Mroczek, 2007). For example, if a large portion of the sam ple had consistent scores on a measure of self-control, a relatively strong coef ficient would be produced even if there were changes for a smaller, yet meaningful, numbe r of individuals within the sample. Another important limitation is that even when self-control is assessed at more than two time points, the use of correlations limits t he analyses in that the resulting stability coefficient only represents correlations between two measurement periods. Thus, correlations do not utilize all the information avail able
and fail to account for nonlinear developmental patterns (i.e., changes; Mroczek, 2007). Additionally, correlation coefficients do not account for the possibility that the re may be smaller, yet non-trivial, groups of individuals who evince change in development of selfcontrol (Blonigen, 2010; Mroczek, 2007). Thus, inconsistent with Gottfredson and HirschiÂ’s claims that self-control is stable for everyone once set at ages 8 Â– 10, it is possible that for a smaller portion of individuals self-control is not stable. Some individuals may begin acquiring self-control at a later age, some may develop se lf-control at a slower rate, and others may actually decrease in self-control over ti me. Research supporting the identification of distinct developmental patterns in offending also provides indirect support for the notion that there may be distinct developmental patterns in the development of self-control. With a few exceptions (Piquero, Blumstein, Brame, Haapanen, Mulvey, & Nagin, 2001; Sampson & Laub, 2003), most studies tend to find that four different groups exist (Blokland & Nieuwbeerta, 2005; Chung et al., 2002; DÂ’Unger et al., 1998; Fergusson et al., 2000; Laub et al., 1998; Nagin et al., 1995; Nagin & Land, 1993) Â– non-offending, moderate offending, adolescent onset, and chronic offending groups. If self-control underlie s offending, then it may be possible that developmental patterns in self-control paral lel those found for offending behavior. For example, it is possible that those in the high-chronic offending group are characterized by persistent low self-control w hile those in the non-offending and moderate offending groups are characterized by rela tively higher, stable levels of self-control. Accordingly, those in the adolescent onset group are more likely characterized by changing levels of self-control. It may be t hat these individuals
r develop self-control later than moderate or low-level offending groups or that the se individuals actually decline in self-control during this period. The invariance of the age crime curve also provides similar support for the notion that individuals may follow unique developmental patterns in self-control. If selfcontrol is as highly related to offending as Gottfredson and Hirschi claim, then it would be expected that the two would covary. In this sense, it may be that developmental patt erns in self-control parallel the age crime curve. It may also be that unique t rajectories in selfcontrol underlie the age-crime curve. For example, those individuals responsible for a large proportion of offending may not begin to develop self-control until late adolescenc e at which point they acquire self-control at a high rate. This would explain the decre ase in aggregate measures of offending typically found. It is important to note that this perspective is contradictory to Gottfredson and HirschiÂ’s theory. They counter this notion by suggesting that it is not self-cont rol that changes, it is the behavior that changes. As previously mentioned, individuals lowes t on self-control will always be the lowest within their cohort according to the t heory. Those low on self-control will always be more likely to engage in antisocial behavior than those high on self-control even as they age. However, the types of behaviors or manif estations of low self-control will change as individuals age. Therefore, while offending de creases for everyone, levels of self-control do not change. The implications of finding groups that follow unique developmental patterns in self-control would contradict the theore tical propositions put forth by Gottfredson and Hirschi. Research on personality development also provides indirect evidence for the possibility that not all individuals follow similar patterns in the development of sel f-
control and that a variety of distinct developmental patterns exist (Blonigen, 2010). Studies examining the stability of personality traits, some that overlap wit h the concept of self-control, support this possibility. Roberts, Caspi, and Moffitt (2001) found that while 73% of their sample maintained the same levels of Negative Emotionality (NE M; e.g., anger, antagonism, and anxiety), 20% decreased on NEM, and 7% (a small, yet meaningful portion) of the sample increased their levels of NEM. While Robins, Frale y, Roberts, and Trzesniewski (2001) found that the majority of their sample followed normative trends in the development of global personality traits (e.g., agreea bleness, conscientiousness, and neuroticism) from late adolescence into young adulthood, the y did identify a small group that decreased on conscientiousness during this period. Johnson, Hicks, McGue, and Iacono (2007), using growth mixture modeling, identified distinct developmental trajectories for each facet of the Multidimensional Personalit y Questionnaire (MPQ; Tellegen, 2006). Most notably, they found three groups for Well-being: one which remained stable over time, one that started low and increased, and one that started moderate and decreased. For the measure Â“Alienation, Aggress ion, and Harm Avoidance,Â” they found four distinct groups; and one group that started high yet decline d at a much faster rate than the other three groups. Similarly, Cote, Tremblay Nagin, Zoccolillo, and Vitaro (2002) found four distinct impulsivity trajectories for a sa mple of young boys over a span of seven years (i.e., ages 6 Â– 12). They identified one group that remained low on impulsivity during this period, another group that remained relatively higher than the other groups, and two groups that evinced considerable change: one that began low on impulsivity and increased over time and one than started with moderate levels of impulsivity and decreased over time. Schaeffer, Petrus, Ialongo, Poduska, and
Kellam (2003) examined trajectories of aggression for youth between the firs t through the seventh grades. They found four distinct developmental patterns that included three groups that remained relatively stable (low, moderate, and high aggression); how ever, they identified one group that began low and increased to levels of aggression that surpassed the high-stable aggressive group. While these studies find distinct developmental trajectories in personality traits related to offending simi lar to self-control, they do not directly examine trajectories in self-control. Only a few studies have examined the possibility that different groups foll ow distinct developmental patterns in self-control. For example, Turner and Piquero (2002) found that males maintained significantly lower levels of self-control compa red to females over seven waves of data collection. They also found that while White and minority youths had similar levels of self-control from ages seven to thirtee n, Whites had significantly higher levels of self-control from ages fifteen to nineteen suggesting that Whites develop self-control faster than non-Whites. Regarding developmental patt erns of self-control by gender, Winfree et al. (2006) also found that males scored signi ficantly lower on self-control over the course of five years compared to females. They also found that developmental patterns differed by race. Specifically, they found that A frican Americans reported significantly lower levels of self-control than Hispa nics during the second year, that Hispanics had higher scores by the fourth year, and that Whites s cored significantly lower on self-control during the last two years compared to Af rican Americans. While these findings suggest that there are different developme ntal patterns of self-control for certain groups, they rely on a-priori identified characte ristics such as demographic characteristics (e.g., gender and race) and offending status to determine
group membership. These studies assume that individuals who are similar on certa in characteristics share similar developmental trajectories in self-c ontrol. While there may be both theoretical and empirical bases for expecting individuals associated wi th such groups to have similar life experiences and, thus, share similar trajectori es in self-control, studies taking this approach may overlook similarities and differences in development not associated with these group characteristics. Group-based trajectory modeling is a semi-parametric method designed to identify groups with distinct developmental trajectories without preconceive d notions about what constitutes a group (see e.g., Jones & Nagin, 2007). However, only two studies have used this technique to identify if there are, in fact, groups of individuals t hat follow distinct developmental patterns in self-control. Both Higgins et al. (2009) and Hay and Forrest (2006) attempted to identify groups of individuals that follow unique developmental patterns of self-control using semi-parametric group-ba sed trajectory modeling. Higgins et al. (2009) assessed the stability of self-control among a nationally representative sample for youth between ages of 12 16. Their group-based trajec tory analysis identified five separate groups with different developmental traj ectories in selfcontrol. Specifically, they identified a group that maintained high levels of self-control over the five years which was made up of 12% of the sample. They identified another group that maintained moderate levels of self-control over the study period tha t made up 35% of the sample and another group that maintained slightly lower mean levels of self-control that accounted for 13% of the sample. They identified a fourth group that continued to evince low levels of self-control over the study period that consisted of 23% of the sample. The within-group and between-group levels of self-control among these
four groups did not change, suggesting both absolute and relative stability for a majorit y of the sample (83%). However, one group consisting of 17% of the sample was identified that started with high levels of self-control and gradually decr eased in their levels self-control over the five waves. The identification of this group sugges ted that self-control is not stable for a small, yet meaningful group of individuals. In addition to identifying trajectories of self-control they also identified traject ories of violent victimization. They identified three trajectory groups for violent victim ization: one that showed no victimization, one with modest levels of victimization, and one that showed a sharp increase in victimization around the age of 14. They then linked trajectories of self-control to trajectories a victimization. The results revealed an ass ociation between trajectories of self-control and trajectories of victimization. It w as found that those groups follow high trajectories of self-control were more likely to be in the tra jectory that evinced no victimization. Accordingly, those that were identified as following trajectories of low levels of self-control were more likely to follow tr ajectories marked by some degree of victimization. In their trajectory analysis, Hay and Forrest (2006) identified eight groups with different trajectories in self-control. Four of the eight groups were chara cterized by stability: a very high-stable group (12%); a high stable group (42%); a medium-stable group (26%); and a low-stable group (4%). However, they also identified four groups that were marked by change. One group started low on self-control and gradually increased over time (i.e., low-increasing group) and accounted for 5% of the sampl e. Two groups showed decreasing levels of self-control over time: one that started high on self-control and one that started with moderate levels of self-control and decr eased over
time (i.e., high-decreasing and medium-decreasing, respectively) and ac counted for 1% and 9% of the sample, respectively. Finally, they identified a group who followed a curvilinear trajectory, starting low and continuing to decrease but then showed an increase near the end of the time period; this group accounting for 1% of the sample Based on these findings, they concluded that self-control is stable for a major ity of individuals (84%). However, they identified a smaller, yet sizeable group of indi viduals for which self-control was not stable in the absolute or relative sense. Two of the groups evinced within-group increases in self-control over time. While it is not complet ely clear from the theory if absolute increases contradict its propositions regarding sta bility, two groups were identified as having within-group losses in self-control. This findi ng is at odds with Gottfredson and HirschiÂ’s theory as they clearly state that self -control cannot be lost. Also, they found that some of the group trajectories intersected with or cr ossed many of the other group trajectories. They interpreted this as showing that s ome groupsÂ’ trajectories change enough to alter their rank-order position, questioning the r elative stability of self-control. In addition to identifying trajectories of self-control, they also exam ined if changes in parenting continued to influence levels of self-control past ages 8 Â– 10 using hierarchical linear modeling. They found that changes in parenting accounted for w ithinindividual changes in self-control such that increases in the quality of parenting re sulting in increases in self-control and decreases in the quality of parenting resulti ng in decreases in self-control. They interpreted these findings as suggesting that change s in parenting are responsible for rank-order changes in levels of self-control.
The findings between these two studies are consistent in that they both find similar proportions of their samples to be marked by both stability and change in sel fcontrol (approximately 85% and 15%, respectively). Alternatively, the number of g roups identified in each of these studies is different, with Higgins et al. (2009) identif ying five groups and Hay and Forrest (2006) identifying 8 groups. These differences in the number of groups may be a function of the sample size, with the former study relyin g on a smaller sample size ( n = 408) and the latter using a larger sample size ( n = 3,793). Also, while both studies identify groups that evince decreases in self-control, a fi nding at odds with Gottfredson and Hirschi, only in one of these studies were groups identifi ed that were marked by increases in self-control (Hay & Forrest, 2006). It is important that these studies be replicated and extended. In addition to t he fact that these are the only two existing studies that attempt to identify unique deve lopmental trajectories, there is a need for replication of these findings among a diffe rent sample in an attempt to reconcile some of the inconsistencies. While both found similar proportions of their samples were stable (or not), the number of groups differed. Thus, replicating these studies should provide greater clarity. There is also a need for research that expands upon those of the prior two studies. While Hay and Forrest (2006) did examine if parenting accounted for changes in self-control, they did not directly link parenting to the trajectories of self-control that they identified. Rese arch is needed that not only examines if parenting distinguishes trajectories of self-control, but al so if other competing factors distinguish trajectory groups of self-control. Currently, t here is no information available that begins to address why there are different traject ories. The general theory does not suggest different groups will be identified. Thus, there are no
n theoretical insights that can be gleaned. Unfortunately, there are als o no empirical studies that examine factors associated with group membership. These gaps i n the literature need to be filled. If factors are found that differentiate betwee n the different trajectories, it might shed light on distinct etiological or developmental pat hways. This, in turn, can be used to refine theory, but also scientifically-inform prevention and intervention strategies. A summary of the empirical status of self-control theory. Many of the propositions contained within Gottfredson and HirschiÂ’s theory have not received consistent empirical support and suggest the validity of many of their claims nee d to be reconsidered. Their claim that self-control is the sole factor in explaini ng antisocial behavior has generally been rejected, with several studies finding that other the oretical constructs continue to explain a significant amount of variance in antisocial outcome s. This is not to dismiss the fact that low self-control continues to have a signifi cant effect on antisocial behavior; in several cases its effect is stronger than other th eoretical factors (Pratt & Cullen, 2000). A modification of Gottfredson and HirschiÂ’s theory may re quire incorporating how low self-control interacts with other variables to better expl ain antisocial behavior. Gottfredson and Hirschi also suggest that parenting, and to some extent schools, are the socializing agents responsible for the development of self-control, cl aiming that self-control will completely mediate the association between parenting (or school) and antisocial behavior. However, much of the literature finds that low self-contr ol does not fully mediate the relationship between parenting and antisocial behavior; in m any cases parenting continues to exert a direct effect on antisocial behavior beyond the effe cts of
self-control, a finding that is inconsistent with Gottfredson and HirschiÂ’s clai m. Additionally, in studies that do find a relationship between parenting and self-control the association tends to be modest and other factors are identified as being importa nt in its development. Thus, parenting appears not to be the only socializing agent that contributes to the development of self-control. Other sources of socialization ex ert a direct effect on self-control such as peer association or biological factors while structural factors such as neighborhood context and socioeconomic status may work indirectly through parenting to influence oneÂ’s level of self-control. In this case, the propos itions of the theory as they are currently stated should be reconsidered to incorporat e those empirically-validated factors that influence the formation and development of selfcontrol. A reformulation may also be necessary for the stability hypothesis. As the previous review of the literature shows, self-control is stable for most individual s. However, there are also findings that suggest it is not stable for everyone. This f inding may likely be due to the fact that not everyone follows the same developmental patt ern and the theory should account for this possibility. While it may be that most individuals follow normative developmental patterns in self-control, other individuals may follow unique developmental patterns. For example, some may begin to develop self-control in later developmental stages only at a much higher rate, while others may begin ea rlier at slower rates. This line of reasoning is in line with theoretical explanations of patterns in offending such as that put forth by Moffitt (1993). MoffittÂ’s taxonomy of offending behavior describes two groups with distinct developmental patterns in offending:
adolescent-limited (ALs) and life-course persistent (LCPs) offenders. S he describes the ALs as having normative offending patterns, engaging in antisocial behavior during adolescence and then quickly desisting in adulthood. Alternatively, the LCPs begi n engaging in antisocial behavior very early in development and persisting into adult hood. Moffitt also suggests that the offending patterns in each of these groups has dis tinct etiologies, with the ALs explained by peer influence and social mimicry whi le the LCPs are characterized as having neurological deficits and poor executive functi oning. MoffittÂ’s taxonomy offers an explanation of distinct patterns in offending that under lie aggregate patterns such as that depicted in the age-crime curve. Applying a s imilar approach to the development of self-control may help to identify and explain distinct developmental patterns in self-control that underlie aggregate patterns in its de velopment. This approach would provide a more precise and thorough explanation for its development across all individuals, even those who do not follow normative patterns in its development. Factors distinguishing developmental patterns of self-control. While no studies have identified whether certain factors can differentiate betwee n distinct developmental trajectories in self-control, research identifying factor s that differentiate between distinct developmental trajectories in offending provides indirect suppor t for those factors that may be important. 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, have experienced poor parenting, and to have parents with criminal histories. Fergusson et al. (2000) found that individuals wit h distinct developmental trajectories in offending (i.e., non-offending, moderate,
r adolescence-limited, and chronic offending groups) had differential levels o f family dysfunction (e.g., parental criminality, alcohol, and drug use), social disadvanta ge, and deviant peer associations. Fergusson et al. (2002) also found that parental criminal ity, gender, and parental conflict were associated with being in the chronic offendi ng group compared to other offending trajectories. Weisner and Capaldi (2003) found that those who were classified as being chronic were more likely to have poor parenting compa red to those in the non-offending or rare-offending groups. They also found that those identified as having decreasing offending trajectories were less like ly to associate with delinquent peers than those in the chronic offending groups. Finally, Chung et al. (2002) found that those youth with desisting trajectories in offending are more likely t o experience better parenting, less family conflict, higher levels of school commitment and attachment, and have fewer delinquent peers compared to those with increasing offen ding trajectories. Research has also identified several factors responsible for the developme nt of self-control, including parenting (e.g., Finkenauer et al., 2005; Gibbs et al., 2003; 1998; Hay, 2001; Morris et al., 2007; Perrone et al., 2004; Unnever et al., 2003; Unnever et al., 2006; Vazsonyi & Belliston, 2007), school (Beaver et al., 2008; Burt et al., 2006; Meldrum, 2008; Turner et al., 2005), neighborhood context (Gibson et al., 2010; Pratt et al., 2004; Turner et al., 2005; Teasdale & Silver, 2009), peer association (Burt et al., 2006; Meldrum, 2008), and biological factors (e.g., Beaver et al., 2009a; Boutwell & Beaver, 2010; Nofziger, 2008; Wright & Beaver, 2005; Wright et al., 2008). It is possible that these factors identified as important in the development of self-cont rol may also be associated with the unique trajectories that individuals follow and, thus, a ble to
differentiate among distinct developmental patterns. However, research has yet to directly examine if certain factors are able to distinguish between dif ferent developmental trajectories of self-control. While Hay and Forrest (2006) did identify disti nct developmental trajectories in self-control and separately found that increase s in parenting resulted in subsequent increases in self-control, they did not examine if parenting could distinguish among the developmental trajectories that they identified in the ir sample. Together, findings from these studies suggest that factors important in distinguishing developmental trajectories in offending such as parental chara cteristics, parenting style, school socialization, and delinquent peer association are simil ar to those that have been found to be responsible for the development of self-control. Thus, it is possible that these factors are also able to distinguish among different developm ental trajectories in self-control; however, studies have yet to directly exam ine if these risk factors are associated with having a certain developmental trajectory o f self-control. Only two studies have attempted to identify unique developmental trajectories of self-control (Hay & Forrest, 2006; Higgins et al., 2009). These two studies suggest t hat for a majority of their samples self-control was stable; however both identifie d groups marked by change. The proportion of those for which self-control was identified as not stable were similar between these two studies. However, the number of groups ide ntified and the patterns for which those groups followed were not similar between these t wo studies. Therefore, replication of these studies with a different sample is ne cessary to confirm those findings that were consistent between these two studies, as wel l as to reconcile some the inconsistent findings. Research is also needed that expands upon the
previous two studies by identifying if factors are directly related to d evelopment trajectories of self-control. The empirical evidence regarding many of the propositions of self-control t heory suggest a need for its reformulation. Similarly, research on the stabilit y hypothesis suggests that while it is stable for a majority of individuals it is not stable f or everyone, and point toward the possibility that some individuals follow unique developmental trajectories even beyond the ages 8 10. This finding is at odds with Gottfredson and HirschiÂ’s claim that self-control is stable for everyone once it is set and t hat everyone will follow the same developmental pattern. This, in turn, suggests a need for modificati ons to this portion of the theory as well. Given that only two studies have directly exami ned this (Hay & Forrest; Higgins et al., 2009), more research is needed that confir ms their findings and provides direction for the modifications that are needed. This requires replication as well as expansion. Replication of these studies can provide confidenc e in the necessary modifications, and identification of risk factors that distinguis h unique trajectories of self-control can provide insight into the underlying causes of unique developmental trajectories.
CHAPTER 3: METHODS Data Collection Procedures 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 youth from middle school s and high schools located in the state of Kentucky. The RSVP study was designed to colle ct individual and contextual data about study participants known to influence criminal offending, victimization, and substance abuse. The RSVP used a multi-stage sam pling technique to select study participants. Based on population-based strata, 30 counties located within the state of Kentucky were randomly selected with rural count ies over sampled in order to ensure representativeness. In order to be considered for incl usion schools had to be public and include 7 th graders. Schools within each county were systematically selected to participate i n order to represent both urban and rural areas. Within the 30 counties, 74 middle schools were eligible and asked to participate in the study. Of the 74 schools that were sampled, 9 declined to take part leaving a total of 65 middle schools from which 7 th grade students were recruited to participate. At the start of data collection there wer e a total of 9,488 7 th graders within these schools who were recruited for participation. Since the study design was longitudinal and students were to be followed over multiple waves of data collec tion,
it was necessary to be able to identify youth who participated in the study. Beca use unique identifiers were used to track students over the four waves, it was required by the Institutional Review Board that active consent be obtained from the parents. Pa rental consent was obtained by mailing consent forms to the homes of parents. Parental conse nt was obtained for 4,102 (43%) of the students. Data was collected from multiple sources (i.e., students, teachers, and principals) for those students for which parental consent was granted. The survey was administered over four waves of data collection beginnin g in 2001 when students were in the 7 th grade and ending in 2004 when they were in the 10 th grade. Over the entire four years of data collection there was at le ast one completed survey for 3,976 of the participants; however, at each consecutive wave beginning with the first there were 3,692, 3,638, 3,050, and 3,040 completed surveys, respectively. Participants In the current study only those individuals who had data on self-control at three time points were included in the analyses. Of the 3,976 individuals who had at least one completed survey, 727 (18%) had self-control measurements for only two or fewer time points with valid data on self-control and were therefore omitted from the analys es. This resulted in a final sample size of 3,249 individuals who had data on self-control for at least three time points. Of those 3,249 individuals included in the final sample, 159 (4.9%) are African American, 2791 (85.9%) are White, 59 (1.8%) are mixed race (Black and White), 16 (0.5%) are Native American, 16 (0.5%) are Asian American, 46 (1.4%) reported being Â“otherÂ”, and 162 (5%) has missing information on race. A slight majori ty of the sample was female (n = 1655; 51%) compared to males (n = 1431, 44%). The
racial and gender makeup of the final sample of the current study are consistent with and representative of the Kentucky public school system as reported for the 2009-10 s chool year ( http://www.education.ky.gov/KDE/HomePageRepository/News+Room/Kentucky+ Educ ation+Facts.htm ). Table 1 reports the descriptive statistics for these individuals on all variable s included in the current study. A small proportion of the sample reported that one of theirparents had ever been arrested. Although the maximum and minimum for age suggest a wider range of ages at each grade, only a small proportion of the indi viduals fell outside of the typical age range at each grade. For example, in the 7 th grade there were ten 12 year-olds, seventy-one 15 year-olds, two 16 year-olds, two 17 year-olds, an d one 19 year-old. The majority of the sample (n = 3163; 97%) reported being either 13 or 14 at grade 7. Also reported in Table 1 are the mean levels of self-control for the sampl e at each grade which suggest that self-control is fairly stable across the four time points (see discussion below). Mean levels of types of mother parenting (attachment and supervision) are on the higher end of the continuum with values clustering around the mean. Table 1 also suggests that father parenting, on average, is at the lower end of the potential ra nge and the standard deviations suggest that there is considerable variation in the level of parenting by the father reported in the sample. Also, on average, participants appeare d to report levels of both school commitment and belief in the school rules that fall cl oser to the upper range of the scale. Finally, mean levels of peer delinquency are close r to the scale minimum.
Table 1. Descriptive statistics for study variables Valid n Mean S.D. Minimum Maximum Minority 3087 .10 0.29 0 1 Female 3086 .54 0.50 0 1 Parental Crime 3027 .21 0.41 0 1 Age Grade 7 Grade 8 Grade 9 Grade 10 3249 3249 3249 3249 13.39 14.39 15.39 16.39 0.56 0.56 0.56 0.56 12 13 14 15 19 20 21 22 Self-Control Grade 7 Grade 8 Grade 9 Grade 10 3038 3171 2943 2940 31.51 31.50 32.77 32.75 6.76 6.86 6.72 6.85 10 10 10 10 40 40 40 40 Parenting Mother Attachment Supervision Father 3073 3072 3007 18.98 17.45 36.71 4.41 3.06 10.45 3 4 1 25 20 50 School Bonds Commitment Rules 3072 3089 16.41 20.11 2.97 4.43 3 4 20 28 Deviant Peers 2476 1.47 2.50 0 12
n Comparisons were made between those individuals who were excluded from subsequent analyses to 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 analyse s. According to these analyses, those who were excluded were more likely to be mal e, 2 (1, N = 3,678) = 0.06, p <.001, = .06; more likely to be White, 2 (1, N = 3,676) = 30.60, p <.001, = .09; and less likely to have a parent who had been to jail or prison, 2 (1, N = 3,586) = 53.97, p <.001, = -.12. Those who were excluded also had significantly lower mother attachment, t (3656) = -6.15, p < .001, d = -.20; mother supervision, t (3655) = -5.97, p < .001, d = -.20; father parenting, t (3576) = -7.37, p < .001, d = -.25; school commitment, t (3688) = -6.07, p < .001, d = -.20; and belief in school rules, t (3672) = -5.56, p = .003, d = -.18. Also, those excluded reported that their peers engaged in significantly more delinquency than those who were in the sample t (2926) = 2.43, p = .003, d = .09. As would be expected, it was also found that those who were missing more than one wave of data on self-control had significantly lowe r levels of self-control measured during the 7 th grade, t (3569) = -6.69, p < .001, d = -.22. Because these differences in key variables were found between those included in the analys es and those that were not, there is the potential for systematic attrition. Specifi cally, these comparisons suggest the possibility that individuals discontinued their participation as a result of low self-control. Therefore, the current analyses may be over-repr esenting those with higher levels of self-control and under-representing those with lower self -control, as well as those with more risk factors (e.g., poor parenting and deviant peer associati on). It is important to take these differences and the potential for sampling bias in ac count when
interpreting the results from the current study. Yet, it should also be noted that the e ffect sizes reported above are rather small. Measures: Dependent Variable Self-control. A 10-item measure of self-control was developed for use in the current study. The items used to represent self-control were selected b ased on face validity and assess the extent that respondents are able to control their own behavi or (e.g., When I am angry, I lose control over my actions ) and focus or pay attention (e.g., Little things or distractions/interruptions throw me off ), especially when such behavior would be most important (e.g., I have difficulty remaining seated at school; see Appendix A for a complete list of items). Response sets for items were based on four-point L ikert scales (1 = never true Â– 4 = always true). Each item was reverse coded and a total sel f-control score was produced by summing all the items at each wave. Thus, higher scores represent higher levels of self-control. The 10 items measured at each wave were subject to a principal components analysis. Observation of the component loadings and scree plot identified a one factor solution at waves one, two, and three. Wave four revealed two factors due to a few items double loading; however, the scree plot suggested that a one-factor solution best fit the data. The items at each wave were also subjec t to reliability analysis revealing excellent internal consistency a lphas ranging from .89 .92. A visual inspection of the distribution of scores on the self-control measure appeared t o resemble a normal distribution with some clustering at the minimum and a smalle r cluster at the maximum end of the scale. This is important for specifying the correct m odel in trajectory analysis (see below). Measures: Risk Factors
Data on several theoretically and empirically relevant variables are included in the analyses. Information on all risk factors were collected at wave one, al though it is important to note the specific terminology used (see Appendix A for a complete lis t of items on scales for risk factors). For example, although the wording of items on parenting (e.g., My mother/father seems to understand me and My mother/father knows where I am when I am away from home ) and school bonds (e.g., Everyone knows what the school rules are and The school rules are strictly enforced ) are asked about in the present tense, they actually measure what has occurred in the past. Other ite ms more clearly referenced past events. For example, items inquiring about deviant peer association use past tense (e.g., Thinking about your closest friends Â– How many of them have done any of the following things in the present school year? Â– sold marijuana or other drugs ) and one item inquires about parental criminality, and whether this has occurred at any time in the past. This suggests that such an event occurred prior to wa ve 1 data collection (see appendix for scale items). After the identification of t rajectories of self-control, these variables are included in a multinomial logistic reg ression analysis in order to examine if they are more or less prevalent for certain groups. Parenting. Both mother parenting and father parenting were measured at wave one in the current study. Both sources of parenting were measured based on 10 items in which individuals responded to questions about their motherÂ’s/fatherÂ’s parenting styl e using a 5-point Likert scale (1 = never Â– 5 = always). Responses were summed for items reflecting mother and father parenting to yield overall mother parenting an d overall father parenting with higher scores indicative of better parenting. Items on the sc ale reflect Gottfredson and HirschiÂ’s (1990) notion of monitoring/supervision (e.g., My
r mother/father knows where I am when I am away from home; My mother/father knows who I am with when I am away from home ). While the measures used in the current analyses do not directly assess par ents ability to recognize bad behavior or discipline, these can be inferred from response s to items about monitoring/discipline. For example, the omission of the recognition elem ent has been addressed before (see Gibbs et al.; 1998; Hay, 2001) suggesting that pare nts who are likely to supervise and monitor their children are also likely to recognize bad behavior and respond to it using discipline. Thus, the omission of these elements is not as problematic as originally thought. Additionally, both parenting scales include items measuring parental attachment (e.g., My mother/father makes me feel wanted; I share my thoughts and feelings with my mother/father ). Although Gottfredson and Hirschi suggest that parental attachment does not need to be represented in measures of parenting si nce attachment is implied when parents successfully carry out the other three com ponents, it has been found that measures of parenting that included attachment explain additional variance in self-control compared to measures that strictly follow the monitori ng, recognition, and discipline model (see e.g., Hay, 2001; Perrone et al., 2004). Items reflecting mother parenting and father parenting were each subje ct to principal components analysis with varimax rotation. Observation of the factor l oadings and scree plot for the items from the mother scale revealed a two-factor st ructure with one factor representing supervision (4 items) and the other representing a ttachment (6 items), while the father scale revealed a single-factor structure for all ten items. Thus, there are two scales assessing mother supervision and attachment and one father
parenting scale. Items for both mother and father parenting measures showe d good internal consistency ( Â’s = .86 and .93, respectively). Delinquent peer association. Delinquent peer association was based on items asking respondents to indicate the number of their friends that had engaged in 12 specifi c types of delinquency. Items included acts from smoked marijuana to having physica lly attacked someone (example: punched, slapped, kicked) and for each item respondents were asked, Thinking of your closest friends Â– How many of them have done any of the following things in the present school year? Each item was binary coded so that responses indicating that none of their friends had engaged in any specific type of behavior were coded as Â“0Â” and those indicating that one or more of their friends enga ged any specific type of behavior were coded as Â“1Â”. These items were then su mmed to obtain a variety of peer delinquency scale where higher scores are indicati ve of more variety in peer delinquency. School bonding. School bonding was measured by 12 items in which respondents answer questions about their school based on a 4-point Likert scale (1 = strongly dis agree Â– 4 = strongly agree). In addition, items included in the measure of school socializ ation also capture the extent to which students feel committed to school. The 12 items wer e then subject to principal components analysis using varimax rotation. The factor loadings and scree plot identified two distinct components. Specifically, one compone nt was identified that represented oneÂ’s Â“belief in school rulesÂ” and was compris ed of 7 items (e.g., Everyone knows what the school rules are ; Teachers keep order in the classroom ; The school rules are strictly enforced ). The other component was comprised of 5 items that captured Â“school commitmentÂ” (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 ( Â’s = .77 and .69, respectively). Parental crime. In order to account for the intergenerational transmission of criminality, the current study also includes a measure of parental crime. R espondents were asked about if either one of their parents had ever been in jail or prison (1 = y es, 0 = no). Control variables. Individuals in the current study were asked to indicate their race and gender. Gender was binary coded (1 = male, 0 = female). In order to meas ure race, individuals were asked to identify themselves as one of the following: Afri canAmerican, Asian-American, Native-American, White, White and Black, or Other Responses were then recoded into a binary outcome of White (1) or minority (0). Both race and gender are associated with levels self-control (Gottfredson & Hirschi, 1990) and, thus, it is necessary to statistically control for them.
CHAPTER 4: ANALYTIC PLAN In the current study, several steps were taken to test the stability hypothe sis using both traditional and a more contemporary method for examining the stability of trai ts. First, mean-level differences in levels of self-control across tim e were examined in order to test the absolute stability of self-control. Second, rank-order, or rela tive stability of self-control was examined using correlational analyses. The results f rom both of these analyses allowed for a comparison between the stability of self-contr ol found in the current sample and that found in previous studies. Third, after assessing stabili ty using more traditional methods, the current study applied group-based trajectory m odeling in order to identify relatively homogenous groups that follow distinct patterns in t he development of self-control and assess the extent that more traditional methods m ask the identification of such groups. The fourth step examined if there were mean diff erences across groups on the relevant risk factors. Finally, in order to account for the unique effects of the risk factors in accounting for group membership, mean differenc es in risk factors across groups were examined and a multinomial logistic regressi on was conducted in order to assess the unique association between risk factors and group membership. In sum, the following analyses were conducted in this specific order: 1.) Mean-level differences in self-control across all waves.
2.) Stability coefficients were examined for self-control measured at e ach wave with each other wave. 3.) Group trajectories of developmental patterns of self-control across time wer e examined. 4.) Mean differences in risk factors across groups identified in trajectory modeling were assessed. 5.) The unique ability of each risk factor in distinguishing among groups was examined. Assessing Absolute Stability In order to discover if there are absolute changes in the level of self-control for the entire sample, the means of self-control were computed at each wave. Addit ionally, a repeated-measure ANOVA was conducted in order to examine if, on average, individualsÂ’ levels of self-control varies significantly across the four wa ves. Specifically, the within-factor is age at each of the four waves of data collection and the de pendent variable is level of self-control measured at each age. According to the t heory, it is expected that mean levels of self-control will remain constant. A significant decrease or increase in mean-levels of self-control would contradict the theory. Assessing Relative Stability Stability coefficients were examined in order to assess the extent tha t self-control is relatively stable over time. Spearman rank-order correlations were calculated between self-control measured at each wave. Stability coefficients reaching .60 or higher have typically been considered to represent a high degree of stability (Costa & McCrae, 1988). Correlations found to be equal to or greater than .60 would support the theory, suggesting
a high degree of relative or rank-ordered stability. Correlations less tha n .60 would contradict the theory, suggesting that self-control is not relatively stabl e for everyone and that, for some individuals, self-control changes enough to alter the rank-order pos itioning. Identification of Developmental Trajectories While mean-level analyses and stability coefficients are informative regarding the stability of traits, both are limited in their ability to identify unique develo pmental patterns. More importantly, traditional methods of assessing stability overlook t he possibility that there are homogenous subgroups of individuals that share similar pa tterns in development. In order to identify such groups with distinct trajectories in the development of self-control, the current study employed semi-parametric gr oup-based trajectory modeling. Group-based trajectory modeling enables the identifica tion of homogenous clusters of individuals with similar developmental patterns (Nagin, 2005 ). Background on group-based trajectory modeling. Two more common methods used to examine developmental trajectories Â– hierarchical modeling and l atent growth curve analysis Â– are similar to group-based trajectory modeling in t hat all three can specify the shape of developmental trajectories as a function of age (Nag in, 1999; Nagin & Tremblay, 1999). However, group-based trajectory modeling differs i n two important ways. First, both hierarchical modeling and latent growth-curve ana lysis assume that individuals within the population follow the same general pattern of development (i.e., trajectories are continuously distributed) and are, therefore, not w ellsuited for identifying groups that have unique developmental patterns. Alternativel y, group-based trajectory modeling does not make this assumption and, instead, takes a multinomial modeling approach and assumes that there are groups of individuals that
follow relatively similar patterns of development within the population. Second, group based trajectory analysis does not rely on ad-hoc, subjective identification of groups. Group-based trajectory modeling provides posteriori indices that can be used to determine the identification of groups and the precision of group membership. Other methods require that groups be identified prior to trajectory identification bas ed solely on subjective classification schemes (Nagin, 1999; Nagin & Tremblay, 1999). Thus, gr oupbased trajectory modeling is the most ideal method for identifying groups wit h distinct trajectories of development. The current study utilizes the TRAJ procedure (PROC TRAJ) in SAS which is designed to estimate developmental trajectories using data collected at multiple waves (Jones, Nagin, & Roeder, 2001). There are several steps for estimating models in P ROC TRAJ. First, the procedure allows the user to model trajectories for outcomes ba sed on three different distributions Â– count, binary, and psychometric data. The user must indicate which distribution would be best for modeling their specific outcome. The censored normal distribution is ideal for modeling outcomes based on psychometric scales that have finite ranges of possible scores with clustering occur ring at the scale minimum and maximum. The Poisson distribution is ideal for count data in which values are censored at zero, while the binary logit is specified for data with a binar y outcome (Jones et al., 2001; Nagin, 1999; Nagin & Tremblay, 1999). The current analysis specifies a model based on a censored normal distribution. This decision was base d upon visual inspection of the distribution of the outcome measure (see description of self-control in measures section above).
n The next step in the analysis is to identify the best-fitting model. This is done by specifying separate models, each with a different number of groups. Within each of t hese different models (i.e., number of groups specified) the shape of each groupÂ’ s trajectory is identified. This is an iterative process in which the user specifies an order (polynomial) for each group and observes the parameter estimates to identify the best-fi tting trajectory shape for each group. The PROC TRAJ procedure allows the user to specify up to third-order polynomials for time (Jones et al., 2001). Each groupÂ’s trajectory can take on a constant or flat shape specified by a zero-order polynomial, a linear traje ctory specified as a first degree polynomial, a curvilinear shape specified by an second order pol ynomial, or a cubic shape specified by a third degree polynomial. Using NaginÂ’s (2005) appr oach the group that was specified as having the highest order polynomial to characte rize the shape of that groups trajectory that did not reach statistical significan ce (i.e., p > .05) was adjusted (i.e., reduced). For example, if a group was specified as having a cubi c trajectory shape and another was specified as having a linear trajectory yet neither had significant parameter estimates, only the group specified as having a cubic shape would be adjusted. After the order for that group was adjusted the model was then reestimated and this process was continued until a model was specified in which all of the groupsÂ’ trajectory shapes were identified as being significant (i.e., the Â“t rimmedÂ” model). Once the best-fitting model for each different number of groups specified w as identifiedthe Bayesian Information Criterion (BIC) was used to compare which of these models most accurately represented the data. Although the BIC provides an obje ctive statistical approach to model selection, it is not always useful for selecti ng the best model and more subjective means of model selection are necessary (Nagin, 2005). As Nagin
(2005) points out, in model selection a balance is needed between objective approaches that favor parsimony and subjective interpretation that weigh the substantive m eaning of the groups identified. The BIC identifies the correct model when it is maxim ized (i.e., less negative) and also rewards models that include fewer parameters (i.e., number of groups). The PROC TRAJ procedure also provides a metric for evaluating the precision of group membership Â– the posterior probabilities Â– which can be used to determine the probability that an individual belongs to each group. It has been recommended that the mean of these probabilities fall above .70 (Nagin, 2005) indicating that, on average, individuals with in that group have a .70 probability of actually following that traject ory. Statistical Formulas. The formulas presented below are those for modeling outcomes with censored normal (CNORM) distributions using group-based trajec tory modeling and determining model fit. This is done in order to provide a statistical supplement to the conceptual basis for such models. In a model assuming a quadratic relationship between age and the outcome variable for individual iÂ’ s age at time t given their membership in group j, the formula is: y it *j = j 0 + j 1 Age it + j 2 Age 2 it + it (1) where j 0Â…2 are the parameters that determine the shape of the trajectory of each group j and, thus, are not free to vary across groups allowing each group to follow different trajectories. Age it is the age of individual i at time t and Age 2 it is their age squared at time t and it is the error term which is assumed to be normally distributed within the population. In the CNORM model y it *j is a latent variable that can be thought of as a
measure of the potential for engaging in some behavior (e.g., manifestations of low selfcontrol). The BIC and the posterior probabilities are common indices used for model selection the equations for each are presented below. The equation for calculati ng the BIC is: BIC = log( L ) Â– 0.5*log( n )*( k ) (2) where L is the maximized likelihood for the model, n is the sample size, and k is the number of parameters in the equation. Thus, multiplying by k penalizes models with more parameters. Calculation of the posterior probability that individuals belong to a certain group can be calculated by using the following equation: n r r r (3) Where Y i is a vector of individual i Â’s measured behavior at each time period, t and is the estimated probability that individual i belongs in group j given Y i Because it is impossible to calculate from parameter estimates, itÂ’s value is estimated from which represents the probability that individualÂ’s i behavior would have been observed given membership in group j. Also, is the estimated proportion of the population in group j. Because will inherently be inflated for larger sized groups, these values are weighted by The mean of can provide a useful tool for identifying the optimal model (i.e., >.70). Developmental trajectories of self-control
nr The method and procedures described above were used to identify developmental trajectories of self-control. Specifically, self-control is measure d at each wave of data collection and were modeled as a function of the individualsÂ’ grade at each wave For reasons described above, the model was specified as a censored normal distribut ion. Selection of the order of each groupÂ’s trajectory that best represents the dat a is informed by the statistical comparison of the parameters of each model specified as well as the theoretical basis. The approach taken in the current study, began with the highest order possible (i.e., cubic) and to adjusted or Â“trimmedÂ” the order for those groups in w hich the parameter estimates were not significant. Subsequent models were th en specified by increasing the number of groups and trimming the model until the BIC was max imized. In order to ensure that the best fitting model is chosen, the posterior probabilit ies were also examined. The posterior probabilities provide an index of the probability that each individual belongs to each group Â– the higher the probability the more confident one can be that an individualÂ’s trajectory has been correctly identified Â– and can be used to classify individuals into a certain group (i.e., the one for which they have the highest probability of belonging). Therefore, a high mean posterior probability for ea ch group is a good indicator that the model specified best fits the data. Nagin (2005) recommends that the mean posterior probabilities be at least .70, in order to ensure certaint y in group membership. A trajectory model that identifies only one group or several groups that have the same stable pattern in their development would provide support for self-control theory. In the context of examining aggregate levels of self-control over time, the theory would be supported if several groups were identified that underlie and parallel stab le mean-level
n pattern of self-control. On the other hand, if distinct developmental patterns in self-control that evince change are identified or if a single trajectory that declines or increases in self-control over time is identified, then this would be inconsistent with the theory Identification of Risk Factors Once the final, best-fitting model for identifying developmental patter ns in selfcontrol was identified, the group assignment for each individual (i.e., the group with t he highest probability of membership for each individual) was identified. This allowe d for the creation of a variable that reflects each individualÂ’s group assignment whi ch was then used in subsequent analyses. First, an ANOVA was conducted examining mean differences for each of the risk factors across the group assignment varia ble. Then, a sequence of multinomial logistic regression models (MLOGITs) was conduc ted using STATA 11.0 software in order to examine the unique contribution of each risk factor in distinguishing group membership. Robust standard errors were used in these models in order to account for the possibility of clustering by school. First, demographic variables (i.e., gender and race) were included in the model. Second, theoretically and empiri cally relevant risk factors (i.e., parental characteristics, school bonds, and peer a ssociation) were included in the model. The MLOGIT was also computed while accounting for the clustering of dat a within schools by adjusting the standard errors. Clustering refers to the notion tha t individuals within a specific context may be more similar to one another than they ar e to individuals in other contexts. To the extent this occurs, standard errors are infla ted and can lead to inaccurate tests of statistical significance. Because indi viduals in the current data are nested within schools, a clustering effect may occur for those w ithin the same
n schools. To ensure this did not occur in the present analyses, robust standard errors wer e employed. Such an approach takes into account the nested nature of the data and allows for more accurate and proper inferences. One issue with examining differences in risk-factors across groups af ter the groups have been identified (i.e., the Â“classify-analyzeÂ” approach) is that it doe s not account for the fact that there is uncertainty in group membership and can compound the reification problem (Nagin, 2005). Specifically, reification refers to the ide a that identified groups actually exist and the individuals within those groups follow their respective trajectory lock-in-step. Assigning individuals to discret e groups in order to conduct further analyses assumes that the trajectory analysis has accur ately identified the trajectory group to which each individual belongs. In actuality, the groups identified in the trajectory analysis are only statistical approximations of reali ty. However, Roeder, Lynch, and Nagin (1999) suggest that this is not as problematic when the mean posterior probabilities are high enough (i.e., .70). Thus, the mean poster-probabilities were observed to ensure confidence in group membership and to justify including the identified groups as outcomes in subsequent analyses. Because Gottfredson and Hirschi do not make claims about the number or nature of groups that would be identified, linking risk factors to trajectories self-control is exploratory. However, general predictions based on their theoretical claims can be made regarding the risk factors. Because Gottfredson and Hirschi suggest that par enting and to some extent school socialization are influential in the development of self-control it is expected that those groups that follow trajectories with low levels of self-c ontrol will be more likely to experience poorer quality of parenting and have weaker school bonds tha n
n those that follow high trajectories of self-control. Alternatively, Gottfre dson and Hirschi disregard the direct influence of race, gender, parental criminal ity, and deviant peers on the development of self-control. In turn, it is not expected that these risk f actors will be associated with group membership. Hypotheses Given the preceding discussion, several specific hypotheses have emerged within the context of the current study: Hypothesis 1: It is not expected that significant mean-level changes in sel f-control will be observed. Hypothesis 2: It is expected that adjacent measurement periods will be positive ly correlated at or above .60. Hypothesis 3: It is expected that only one trajectory group will be identifie d or that multiple groups will be identified that have constant (i.e., zero-order) shapes and only differ on mean-levels of self-control will be identified. Hypothesis 4: It is expected that gender will not differ among groups. Hypothesis 5: It is expected that race will not differ among groups. Hypothesis 6: It is expected that deviant peer association will not differ among groups. Hypothesis 7: It is expected that groups with low mean-levels of self-control wi ll have poorer quality of mother supervision than those with high mean levels of self-control. Hypothesis 8: It is expected that groups with low mean-levels of self-control wi ll have poorer quality of mother attachment than those with high mean levels of self-control Hypothesis 9: It is expected that groups with low mean-levels of self-control wi ll have poorer quality of father parenting than those with high mean levels of self-control
n Hypothesis 10: It is expected that parental criminality will not differ a mong groups. Hypothesis 11: It is expected that groups with low mean-levels of self-contr ol will have lower school commitment than those with high mean levels of self-control. Hypothesis 12: It is expected that groups with low mean-levels of self-control will have lower belief in school rules than those with high mean levels of self-control. Study Strengths and Potential Limitations This design of the current study has several advantages. The longitudinal des ign over four waves of data allows for an appropriate test of the stability hypothe sis. Because self-control is assessed at four time points, this allows for the identificati on of groups that follow unique trajectories (i.e., constant, linear, quadratic, and cubic shapes). This is an advantage over assessing stability at only two or three time points as these would restrict the extent that different developmental patterns could be modeled. The current da taset also provides information on relevant constructs for a large sample of individuals. Information on theoretically relevant constructs for the development of self-c ontrol are available such as parenting, school bonds, and peer association. Reliance on dataset s that do not account for such factors may lead to mis-specified models by disregarding othe r important, competing factors that may influence the development of self-control In order to increase representativeness, rural counties were over-sampled. T his approach ensures that certain segments of the society (i.e., individuals who come from rural backgrounds) are not underrepresented in the current study. Ensuring that individuals from rural counties are included in the analyses increases the generali zability of the study findings and limits sampling bias. The application of the SPGM is appropriate for assessing the stability of self-control and has advantages over other methods suc h as
n hierarchical linear modeling. The SPGM method is especially appropriate for e xamining change while accounting for the possibility that individuals follow different developmental trajectories. It is also important to note some potential limitations of the current study. Fi rst, there are a few sampling issue that may bias the results and limit their generalizability. The universal sampling frame was over 9,000 students; however, due to the requirement that active parental consent had to be obtained, just over 4,000 students were enrolled. Those youth for whom active parental consent was not obtained may systematically differ from those who participate on the outcome variable of self-control as well as other r isk factors (e.g., parenting). Based on the current data it is not possible to make compa risons between who did and did not participate on the variables under consideration; however, it would be possible to address this limitation if these individuals were identifiable a nd the relevant information was collected at the same time as the data were coll ected from those who are included in the sample. Thus, it will be important to interpret the results w ith caution. The sample was also limited to the state of Kentucky and may not generali ze to the entire population. Another limitation is that information on all potential risk factors is not avail able and may result in misidentified model. For example, neighborhood context has been suggested to be important in the development of self-control (Turner et al., 2005); however, information on neighborhood is unavailable in the current dataset and cannot be accounted for. Another limitation regarding risk factors is the point at which they were measured. All risk factors were measured at wave one. This limits the abili ty to make claims about causality regarding the risk factors predicting trajecto ries. This limitation is
n less applicable for more static variables such as gender, race, and prior par ental criminality. However, the effect of the more dynamic risk factors such as parenting, school bonds, and peer deviance must be interpreted with caution. A final limitation applies to all studies that use SPGM as a way to model change Nagin and Tremblay (2005) have identified three common misconceptions that can lead to a reification of the results produced from SPGM. The common misconceptions include the interpretation that individuals really belong to these groups; the number of gr oups produced is certain; and that those within each group follow the pattern of their respec tive group membership exactly. It is important to be aware of these misconceptions when interpreting the findings from the current study in order to avoid a reificat ion of the groups identified as having a certain developmental pattern in self-control.
nn CHAPTER 5: RESULTS Mean-level analysis (absolute stability) Figure 1 graphically depicts mean levels of self-control across each of t he four waves. These results show that mean levels of self-control remained fairly stable over the four waves of data collection with only a slight increase. Mean levels of se lf-control remained constant between grade 7 ( M = 31.51) and grade 8 ( M = 31.50) and between grade 9 ( M = 32.77) and grade 10 ( M = 32.75). The most notable change occurred between grades 2 and 3; however, if this increase in mean levels of self-contr ol is not significant it would be consistent with Gottfredson and HirschiÂ’s notion regarding absolute changes in self-control. r r r r n r Self-ControlGrade Figure 1. Mean levels of self-control by grade
n Table 2. Repeated measure ANOVAs for self-control 95% CI Comparisons Mean Difference SE Lower Bound Upper Bound Grade 7 vs. Grade 8 0.43 0.13 -0.30 0.38 Grade 7 vs. Grade 9 1.11* 0.14 -1.48 -0.74 Grade 7 vs. Grade 10 Grade 8 vs. Grade 9 Grade 8 vs. Grade 10 Grade 9 vs. Grade 10 1.07* 1.15* 1.14* 0.04 0.15 0.13 0.14 0.12 -1.46 -1.50 -1.50 -0.28 -0.68 -0.81 -0.74 0.36 p < 0.001 In order to assess whether these changes self-control were significant a repeated measure ANOVA was conducted. In the current ANOVA, the assumption of spheric ity was violated (MaulchyÂ’s W = .92, p<.001). Because MaulchyÂ’s W was greater t han .75 (Girden, 1992), the Hyunh-Feldt was used to correct for sphericity and suggested that there were significant differences in mean levels of self-control over t he four time periods ( F = 45.209, p<.001). However, this does not indicate between which measurement periods the significant differences occurred. In order identify where the signi ficant differences occurred, independent t-tests were examined using bonferonni corre ction. Table 2 reports the results from the ANOVA analysis comparing the different t ime points for significance. As shown in the table, self-control measured at Grades 9 and 10 w ere both significantly higher than that at Grade 7 and 8; however, there were no signifi cant differences between self-control at Grade 7 and 8, nor did they differ between G rades 9
n and 10. This finding does not support hypothesis 1 and is inconsistent with Gottfredson and HirschiÂ’s notion regarding stability. Stability Coefficients (relative stability) Table 3. Correlations between self-control measured at each grade Note All correlations significant at the p <.001 level. Table 3 reports the results for the SpearmanÂ’s rank-order correlations be tween self-control measured at each time-point. Overall, the results show moderate t o strong correlations between measures of self-control. However, only one of the coeff icients reached the suggested level of .60 for assessing relative stability (Cos ta & McCrae, 1994). The correlation analyses also show that, consistent with previous research, as time increases between measurement periods the size of the correlations tend to de crease. Also, the correlations between time-points become stronger as individualsÂ’ age. In sum the correlations suggest that there is a considerable amount relative stabilit y in selfcontrol; however, the correlations are not perfect and only one reaches the .60 crite ria for stability suggested by Costa & McCrae (1988). In sum, the findings from the corre lation analysis does not support hypothesis 2 and is inconsistent with Gottfredson and HirschiÂ’ s notion regarding relative stability. Self-Control Grade 7 Grade 8 Grade 9 Grade 7 -Grade 8 .52 -Grade 9 .48 .53 -Grade 10 .43 .48 .61
r There are several potential reasons that the correlations are not as high a s the theory would suggest. It is possible that there is a considerable amount of measur ement error that accounts for this; however, because we do not use disattentuated correla tions it is not clear if this is the reason. Another explanation is that self-control is not r elatively stable for everyone and there may be some individuals who follow unique developmental trajectories of self-control. Because correlations assess relati onship at the aggregate level, they may mask the identification of any such groups. Further analyses ar e necessary in order to test for this possibility. Group-Based Trajectory Modeling Model selection. In order to identify if there is heterogeneity in the development of self-control, group-based trajectory modeling was employed. As previously de scribed, the best fitting model was identified using a number of indices (e.g., BIC, group-membership, post-predicted probabilities). The best fitting model for each model t hat specified a different number of groups (e.g., 1-group model, 2-group model, 3-group model, etc.) was first identified for that which the BIC was maximized (appr oaches zero) after trimming the order of each group (see analytic plan for details on model s election). Table 4 reports the BICs for the trimmed models where 1 Â– 7 groups were specifie d. The BIC is not reported for an 8-group model because this model would not converge. As shown in the table, the BIC continues to be maximized as the number of groups specified increases. Beginning with the 4-group model where the BIC was maximized by trimming the orders, two stable groups were identified: a high-stable group (n = 1581; 48%; Mea n GrpProb = .89) and a moderate-stable group (n = 1261; 38.4%; Mean GrpProb = .76).
Table 4. Adjusted and Unadjusted Bayesian Information Criteria (BIC) by number of groups specified There was also a group that started out with moderate levels of self-control but decl ined over time (n = 194; 6.2%; Mean GrpProb = .76) and a group that started with low levels of self-control but began to increase in level of self-control in the 10 th grade (n = 213; 6.7%; Mean GrpProb = .81). In the trimmed 5-group model there were three stable groups: a high-stable group (n=1328; 40.2%; Mean GrpProb = .85), a moderate-high stable group (n = 1260; 38%; Mean GrpProb = .77), and a low stable group (n = 126; 4.1%; Mean GrpProb = .84). Additionally, a low-increasing group (n = 212; 7.2%; Mean GrpProb = .71) and a moderate decreasing group (n = 323; 10.5%; Mean GrpProb = .75) were identified. In addition to the BIC for the 5-group model being maximized over the 4-group model (see Table 4), a unique group is identified that follows a low-stable # of Groups BIC (N = 3249) BIC (N=12092) 1 -38123.20 -38125.19 2 -36774.89 -36779.49 3 -36463.65 -36469.56 4 -36394.08 -36402.63 5 -36329.33 -36343.17 6 -36306.96 -36321.42 7 -36301.77 -36316.77
pattern of development which is not identified in the 4-group model. The 5-group model appeared to be the better-fitting model compared to the 4-group model. As shown in Table 4, compared to the 5-group model the BIC in the 6-group model was larger suggesting a better fit. More importantly, the 6-group model i dentified two declining groups, one of which was overlooked in the 5-group model. The addition of this group provided substantive meaning to the model and the research question at hand (i.e. stability of self-control). Specifically, two groups were identified in t he 6group model as starting with moderate levels of self-control and declining over time. However, one group declines at a much higher rate than the other and actually swit ches places with those who started out lowest on self-control (see description of the 6-g roup model below). Although a low-stable group is not identified in the 6-group model, the groups identified in this model are substantively important as they highlight the lac k of rank-order stability in self-control overlooked in previous models and when using more traditional methods. The 6-group model provides a much more insightful and nuanced picture of the stability of self-control, identifying trajectories that intersect or converge with one another not captured in the prior models. Given the current research question, the trajectories identified in the 6-group model provide substantive meaning above and beyond those identified in prior models as trajectories were identified that refl ect changes in both absolute levels and the rank-order of self-control not captured in prior models. Finally, even though the BIC is maximized in the 7-group model, the 6-group model was selected because parsimony was given priority over the BIC. In the 7-group model, an additional group was identified that followed a high-stable trajectory of selfcontrol, yet was only marginally higher than the high-stable group identifie d in the 6-
group model. Thus, the 7-group model offered no substantive contributions beyond that of the 6-group model. In turn, the 6-group model was selected as the best-fitting t hat identified the most meaningful groups in terms of the stability of self-control Figure 2 presents the trajectories for each of the six different groups along with the percentage of the total sample comprising each one. Group 1 is composed of 3.5% of the sample. The trajectory for this group was identified as taking on a quadra tic form, starting out low on self-control and gradually increasing over time to moderate-low levels of self-control. This group was identified as the Â“low-increasing groupÂ”. Group 2 m akes up 2.3% of the sample. This group was also characterized as having a trajectory of quadratic form, beginning with moderately-low levels of self-control and then decl ining over time. This group was identified as the Â“moderate-decreasing groupÂ”. Group 3 i s r r r r n r nnr Figure 2: Trajectories of self-control G1 (n = 110, 3.5%) G2 (n = 70, 2.3%) G3 (n = 244, 8.3%) G4 (n = 1307, 38.5%) G5 (n = 360, 12.0%) G6 (n = 1158, 35.4%)
made of 8.3% of the sample. Individuals in this group followed a trajectory with a quadratic form, starting out with moderate levels of self-control and increa sing over time. This group was identified as the Â“moderate-increasing groupÂ”. Group 4 made up 38.5% of the sample. This group was also identified having a trajectory that was quadr atic in form. Individuals in group 4 started with moderately-high levels of self-control and only slightly increased over time; thus, this group was identified as the Â“moderatehigh stable groupÂ”. Twelve percent of the sample was identified as being in Group 5. This group followed a linear pattern in development moving from moderate-high levels of self-control to moderate-low levels of self-control and was identified as the Â“modera te-high decreasing groupÂ”. Finally, group 6 made up 35.4% of the sample and also followed a linear pattern. This group started high in self-control and slightly increased This group was identified as the Â“high-stable groupÂ”. Although a majority of the sample (73% ) was identified as having stable trajectories, the results from the traject ory analysis are not supportive of hypothesis 3 since four of the six groups identified followed trajectori es that were marked by change. The model organizes individuals in the sample into groups using an approach that maximizes group-membership probability based on model coefficient estimate s. This approach does not guarantee perfect assignment of an individual to a specific group (i.e., a probability of 1.0), but identifies the group to which individuals have the highest possibility of belonging. The model for which all individuals in the sample have been assigned to the group for which they have the highest probability is the Â“maximize dÂ” model (see Nagin, 2005). Table 5 reports the maximum mean posterior probabilities f or the six-group model along with the number of individuals in each group. As shown, the
Table 5. Mean and median posterior probabilities for six-group model Note Median posterior probabilities are reported in parentheses. mean posterior probabilities for each of the six groups are high, ranging from .71 .83, providing confidence that individuals were assigned to the trajectory group that mos t accurately captures their actual trajectory. The high mean posterior probabi lities also justifies the Â“classify-analyzeÂ” approach used in the current study to exa mine group differences in risk-factors by avoiding uncertainty in group membership (Roeder Lynch, & Nagin, 1999). Group Differences in Self-Control In addition to identifying developmental trajectories of self-control it was a lso examined if the identified groups differed significantly on self-control at ea ch time point. Table 6 reports results from one-way ANOVAs assessing mean-level diff erences in selfcontrol at each time-point across groups. There were significant differences found in levels of self-control at every wave. TukeyÂ’s b was used to determine which groups Group Prob G1 Prob G2 Prob G3 Prob G4 Prob G5 Prob G6 1 (n = 110) .81 (.86) .10 (.04) .04 (.00) .00 (.00) .02 (.00) .00 (.00) 2 (n = 70) .06 (.02) .76 (.79) .01 (.00) .00 (.00) .03 (.00) .00 (.00) 3 (n = 244) .08 (.03) .02 (.00) .71 (.71) .04 (.01) .09 (.03) .00 (.00) 4 (n = 1307) .00 (.00) .00 (.00) .12 (.07) .75 (77) .14 (.08) .17 (.13) 5 (n = 360) .05 (.01) .12 (.08) .12 (.07) .07 (.02) .72 (.73) .00 (.00) 6 (n = 1158) .00 (.00) .00 (.00) .00 (.00) .15 (.08 ) .00 (.00) .83 (.87)
Table 6. One-way ANOVAs for self-control across groups Note. = p<. 001; G1 = Group 1 (Low-increasing); G2 = Group 2 (m oderate-decreasing); G3 = Group 3 (moderate-increas ing); G4 = Group 4 (moderate-high stable); G5 = Group 5 (moderate-high decreasing); G6 = Group 6 (high-stable) Self-Control G1 G2 G3 G4 G5 G6 F TukeyÂ’s b Grade 7 Grade 8 Grade 9 Grade 10 15.85 16.47 20.11 22.16 25.81 19.96 15.80 14.80 20.80 22.96 28.97 31.91 31.39 31.47 32.81 32.81 28.71 26.90 25.25 22.70 36.55 36.79 37.96 37.85 865.96* 931.02* 978.41* 1151.81* G6>G1 Â– G5; G4 > G1,G2,G3,G5; G5>G1,G2,G3; G2>G1,G 3; G3>G1 G6>G1 Â– G5; G4 > G1,G2,G3,G5; G5>G1,G2,G3; G3>G1,G2 ; G2>G1 G6>G1 Â– G5; G4 > G1,G2,G3,G5; G3>G1,G2,G5; G5>G1,G2 ; G1>G2 G6>G1 Â– G5; G4 > G1,G2,G5; G3>G1,G2,G5; G5>G2; G1>G 2
n differed significantly at each wave. The high-stable group showed consider able relative stability maintaining significantly higher levels of self-control tha n any other group over the entire period. However, there were several changes in group differences on selfcontrol over the four grades among the other five groups (see Table 6), sugges ting rankorder changes in self-control. Most notable, the moderate-high stable group had significantly higher levels of self-control than group 2 in the 7 th grade; however, at wave 4 these two groups shared similar levels of self-control. Also, the low-increas ing group had lower self-control than the moderate-decreasing group in the 7 th grade but this relationship was reversed in 10 th grade. Several other changes in the rank order positioning of groups on level of self-control are captured in Table 6. In sum, the ANOVAs revealed that a considerable portion of the sample had relatively stable (i.e., the high-stable and moderate-high stable groups) l evels of selfcontrol. However, several of the groups showed distinct, changing patterns in the development of self-control. For these groups, their levels of self-control chang ed enough to alter their rank-order position. These findings fail to support hypothesis 3, in that a sizeable portion of the sample evidenced considerable change in both the absol ute and relative sense. Risk Factors Mean differences in risk-factors. The first step in examining if risk factors differed across groups was to conduct a series of one-way ANOVAs. Table 7 report s the results from ANOVAs for each risk factor separately. The results re garding gender and minority status confirm hypothesis 4 and 5, respectively. There was no significa nt variation in gender across groups and while there were significant differenc es in minority
Table 7. One way ANOVAs examining gender, minority status, peer association, parenting, parental crimi nality, school bonds, and self-control by group Note. *p<.05, **p<.001; G1 = Group 1 (Low-increasing); G2 = Group 2 (moderate-decreasing); G3 = Group 3 (m oderateincreasing); G4 = Group 4 (moderate-high stable); G5 = Group 5 (moderate-high dec reasing); G6 = Group 6 (high-stable) G1 G2 G3 G4 G5 G6 F TukeyÂ’s b Gender .49 .54 .47 .53 .51 .56 1.89 None Minority Status .86 .87 .88 .90 .89 .92 2.33* None Peer Association 3.40 2.16 2.70 1.48 1.86 .81 34.39 ** G1>G2,G3,G4,G5,G6; G3>G4,G5,G6; G2>G6; G5>G6 Parenting Mother Supervision Attachment Father 15.57 15.41 30.32 16.17 17.32 32.53 16.39 17.59 32.97 17.44 18.79 36.12 16.77 17.78 34.99 18.16 20.30 39.52 31.62** 47.32** 34.60** G6>G1,G2,G3,G5; G4>G1,G2,G3; G5>G1 G6>G1,G2,G3,G4,G5; G4>G1,G2,G3; G5>G1; G3>G1; G2>G1 G6>G1,G2,G3,G4,G5; G4>G1,G2,G3; G5>G1 Parental Criminality .44 .37 .37 .22 .29 .13 27.61* G1>G4,G5,G6; G3>G4,G6; G2>G4,G6; G5>G6 School Bonds Commitment Rules 13.60 17.67 14.62 18.75 14.95 18.52 16.37 19.91 15.66 19.48 17.35 21.18 65.11** 28.92** G6>G1,G2,G3,G4,G5; G4>G1,G2,G3; G5>G1,G2; G3>G1 G2>G1; G6>G1,G2,G3,G4,G5; G4>G1,G3; G5>G1
status, the post-hoc tests did not identify for which groups there were significant differences. Because of this, minority status could not be further examined a nd interpreted. As shown in Table 7, the ANOVAs reveal significant variation in deviant peer association across groups. The findings suggest that those groups that are increasing in levels of self-control (i.e., the low-increasing and moderate-increasing groups) tend to have more deviant peer association than those that follow stable and decreasing patte rns. While it is not surprising that the increasing groups have more deviant peer as sociation than the stable groups, it is somewhat counterintuitive that they have more deviant peers association than those with decreasing patterns. However, this finding makes mor e sense given that both increasing groups had lower levels of self-control in 7 th grade than the decreasing groups. Thus, this finding may be reflecting the association betwee n levels of self-control and deviant peer association at grade seven only. Somewhat mor e intuitive is the finding that the moderate-high decreasing group had more deviant peer associ ation than the high-stable group, suggesting that deviant peer association may be an im portant risk factor in differentiating those that lose self-control from those that ma intain high levels of self-control. These findings lead to the rejection of hypothesis 6 that deviant peer association will not be associated with group membership. Table 7 also shows significant differences among groups for parenting varia bles. In general, those with higher, stable trajectories show higher quality of parent al socialization for all of its sources (father, mother supervision, and mother atta chment) than those groups with lower levels of self-control with less stable trajectorie s. This finding is consistent with hypotheses 7 through 9; however inconsistent with Gottfreds on
90 and HirschiÂ’s claims, mother attachment was the only form of parenting that differentiated the low increasing group from the moderate-increasing and mode ratedecreasing groups. The former finding is not surprising; however, the latter f inding is somewhat perplexing. Similar to deviant peer association, this may reflect this relationship at the 7 th grade assessments given that the increasing had lower levels of selfcontrol at this time point. As shown in Table 7, parental criminality also varied significantly across groups. With few exceptions, the findings from the ANOVAS generally suggest that t hose with lower, less stable trajectories in self-control report a higher degree o f parental criminality than those with higher, stable trajectories. This finding is not in line with hypothes is 10 and is inconsistent with the theory indirectly suggesting support for the intergener ational or genetic transmission of self-control. Finally, Table 7 shows that both school commitment and belief in school rules varied significantly across groups. Once again, the more stable groups wit h higher mean levels of self-control tended to report higher levels of school commitment and belief in school rules than those groups with lower, less stable trajectories. However, t he moderate-high stable group was more similar to the less stable groups than the highstable group on both commitment and belief in rules. These findings are consistent and supportive of both hypotheses regarding school commitment and belief in school rules (hypotheses 11 and 12). They are also in line with Gottfredson and HirschiÂ’s (2003) more recent acknowledgment of school socialization as important in the development of self-control.
91 In sum, the ANOVAs suggest that the risk factors tend to differentiate those groups with higher, more stable patterns in self-control from those with lower, les s stable patterns. In general, those with less stable self-control tend to be charac terized by more risk-factors than those with more stable trajectories. There are some sig nificant differences in risk factors among the less stable groups and while it appears t hat these differences seem to reflect the association between mean-levels of se lf-control and risk factors measured in the 7 th grade, the veracity of this claim cannot be unequivocally verified with this data. Multinomial Logistic Regression Relative differences in risk-factors across the six identified tr ajectory groups. In order to assess the relative importance of risk factors, three multinomial l ogistic regression models (MLOGIT) were conducted. In the first model, group membership i s regressed onto race and gender and in the second model group membership is regressed onto race, gender, deviant peer association, parenting, parental criminality, and s chool bonds. Table 8 reports the results from the multinomial logistic regression analys is. Within the context of the current research question, it is important to identify the r elative importance of factors that differentiate those with stable trajectories f rom those with changing trajectories. Additionally, the ANOVA analyses suggested m ost of the differences in risk factors are between those groups lacking stability in s elf-control and
92 Table 8 Multinomial logistic regression: group membership by risk factors Variable Model 1 Model 2 B SE exp(B) B SE exp(B) Gender Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing -.28 -.11 -.39** -.12 -.21 .22 .23 .15 .08 .11 .76 .90 .68 .89 .81 .21 .27 .01 -.04 .05 .32 .35 .17 .12 .15 1.23 1.31 1.01 .96 1.05 Minority Status Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing -.60 -.48 -.41* -.05 -.34* .31 .36 .19 .11 .15 .55 .62 .66 .95 .71 -.38 -.38 -.31 .06 -.34 .24 .39 .18 .13 .18 .68 .68 .73 1.06 .71 Deviant Peers Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing ---------------.29*** .18*** .24*** .13*** .17*** .04 .05 .03 .02 .03 1.37 1.20 1.27 1.14 1.19 Mother Attachment Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing ----------------.14*** -.06 -.04 -.04* -.07** .03 .04 .02 .02 .02 .87 .94 .96 .96 .93 Mother Supervision Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing ----------------.01 -.07 -.05 -.02 -.04 .04 .07 .04 .02 .03 .99 .93 .95 .98 .97 Father Parenting Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing ----------------.04** -.03 -.03*** -.02** -.02 .01 .02 .01 .01 .01 .97 .98 .97 .98 .98 Parental Criminality Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing ---------------.44** .22 .27* .20 .26* .13 .16 .12 .11 .12 1.55 1.24 1.31 1.22 1.30 School Commitment Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing ----------------.29*** -.27*** -.18*** -.09** -.17*** .05 .08 .04 .03 .03 .75 .77 .84 .92 .84 School Rules Low-increasing Moderate-decreasing Moderate-increasing Moderate-high stable Moderate-high decreasing ----------------.00 .02 -.02 -.02 .01 .03 .05 .02 .02 .02 1.00 1.02 .98 .98 1.01 Wald 2 (R 2) 26.43** (.00) 2885.33***(.07) Note. = p<.05, ** = p<.01, *** = p<.001; Group 6 is the high-stable group and served as the reference group for all comparisons.
93 those with higher, stable trajectories (most differences occur ring with high-stable group). Therefore, the high-stable group is designated as the comparis on group. This enables the identification of the relative importance of risk factors in diff erentiating those with high, stable levels of self-control from those with lower, less stable levels of se lf-control. Model 1 of Table 8 reports differences in minority status and gender between the high-stable group and all other groups. As shown, those in the moderate-increasing group are more likely to be male and of minority status than those in the high-stable group. Additionally, the moderate-high decreasing group was more likely to be of minority status compared to those in the high-stable group. In general, however, neit her gender nor minority status were useful in distinguish group membership. This finding confirms those regarding minority status and gender found in the ANOVAs while also confirming hypotheses 4 and 5. Model 2 of Table 8 reports differences in parenting, school bonds, parental criminality, and deviant peers controlling for minority status and gender. The differences in minority status and gender found in the first model are washed out in the second model once other variables are taken into account. In general, the final model supports the findings from the ANOVAs that those groups with higher, more stable levels of sel fcontrol have fewer risk factors than those with lower, less stable trajectori es. However, once other factors are held constant, mother supervision and belief in school rules no longer differentiate those in the high stable group from any of the other groups. The former finding is inconsistent with Gottfredson and Hirschi and fails to find support f or hypothesis 7 that mother supervision should differ across groups. The latter finding is
94 also inconsistent with their more recent claims that school socialization is i mportant in the development of self-control and fails to find support for hypothesis 12. Deviant peer association and school commitment are the most consistent in differentiating the groups, with the high-stable group having less deviant peer as sociation and higher levels of school commitment than all other groups. Additionally, both deviant peer association and school commitment have the strongest coefficients compa red to those of other variables. The finding that deviant peer association distinguishes among these groups is inconsistent with Gottfredson and HirschiÂ’s notion that peers do not influence oneÂ’s level of self-control and, also does not support hypothesis 6. Counter to the findings regarding school rules, the finding that school commitment distingui shed groups is consistent with Gottfredson and HirschiÂ’s (2003) more recent claims about schools and their role in the development of self-control. Thus, we fail to reject hypothesis 11. Although mother attachment, father parenting, and parental criminality were less consistent in terms of which specific groups were significantly differe nt from the highstable group and the magnitude of the coefficients were small, the findings stil l suggest that, in general, those groups with lower, less stable trajectories are a ssociated with more risk factors than those in the high-stable group. For the most part, the lower, less s table groups tended to have lower father parenting and mother attachment and higher parenta l criminality than the high-stable group. However, the coefficients for mother attachment and father parenting were relatively small, increasing the likelihood that signi ficant comparisons for these risk factors are due to chance. On the other hand, the coeffici ents
95 for parental criminality were higher, suggesting that it is might be i mportant in the development of self-control. Not surprisingly, the group that most consistently differed from the high-s table group across risk factors was the low-increasing group. Alternatively, the moderatedecreasing group was the only group that differed from the high-stable group on onl y two risk factors: school commitment and deviant peers. With the addition of having higher parental criminality, the other decreasing group (i.e., the moderate-high dec reasing group) also differed on these two risk factors. Thus, it may be that these two fact ors (deviant peer association and school commitment) are most important in identifyin g those individuals that will follow decreasing trajectories in self-control. How ever, the finding that both increasing groups also differed from the high-stable group on deviant peer association and school commitment makes the utility of these risk factors i n identifying trajectories less clear. Although it cannot be discerned from the current d ata, it might be the case that there are qualitative differences in the peer associations and school commitments between these groups. In sum, the findings from the MLOGIT analysis generally support those of the ANOVAs in that several factors are identified as important in differenti ating among those groups with lower, less stable trajectories of self-control from those with hi gher, more stable trajectories. Although these general patterns do emerge, the number of groups and risk factors examined make the results a bit unwieldy and hard to interpret. Additionally, it limits the ability to make comparisons between those groups with increasing trajectories from those with decreasing trajectories.
96 Relative differences in risk-factors for stable, increasing, and decre asing groups. To limit the number comparisons and allow for the examination of what risk factors differentiate those groups with increasing trajectories from thos e with decreasing trajectories, an additional MLOGIT was conducted. In order to do so, three groups were created by collapsing those with increasing trajectories into a single group, those with decreasing trajectories into a single group, and those with stable traje ctories into a single group, resulting in an increasing group, a decreasing group, and a stable group, respectively. In order to maintain consistency and make comparisons across mode ls, the stable group will be specified as the comparison group. This will allow for compari sons in risk factors between the stable and decreasing groups Â– do risk factors differ entiate those who decrease in self-control from those who maintain high levels of self-cont rol?; and those between the stable and increasing groups Â– do risk factors differentia te those with stable levels of self-control from those with increasing levels? The results for the MLOGIT models examining differences in risk-fact ors across the groups that were decreasing, increasing, and stable in self-control are p resented in Table 9. Essentially, minority status, deviant peers, mother attachment, and school commitment differentiated those groups that were stable on self-control fr om those with increasing and decreasing trajectories. Specifically, those who wer e stable were less likely to be of minority status, have fewer deviant peers, report higher levels of mother attachment, and school commitment compared to those with trajectories marked by changing levels of self-control. Although gender differentiated the increas ing from decreasing groups, this effect was no longer significant when statistica lly controlling for
97 other variables. Additionally, parental criminality was significant, but only in differentiating those who had increasing trajectories from those with sta ble trajectories. Table 9. Multinomial logistic regression examining differences in risk factors across increasing, decreasing, stable groups. Note. = p<.05, ** = p<.01, *** = p<.00.1 The increasing group was more likely to report parental criminality compared t o the stable group. In general, these findings support those of prior analyses in the cur rent study, suggesting that there are important risk factors that differentiat e those who have Variable Model 1 Model 2 B SE exp(B) B SE exp(B) Gender Decreasing Increasing -.12 -.29* .09 .13 1.13 .75 .10 .10 .12 .16 1.11 1.11 Minority Status Decreasing Increasing -.34* -.45* .16 .19 1.40 .64 -.37* -.32* .18 .14 .69 .73 Deviant Peers Decreasing Increasing ------.01* .08** .01 .03 .99 1.08 Mother Attachment Decreasing Increasing -------.04* -.05* .02 .02 .96 .95 Mother Supervision Decreasing Increasing -------.03 -.03 .03 .03 .97 .97 Father Parenting Decreasing Increasing -------.01 -.02 .01 .01 .99 .98 Parental Criminality Decreasing Increasing ------.12 .19** .07 .06 1.13 1.21 School Commitment Decreasing Increasing -------.13*** -.16*** .02 .03 .88 .85 School Rules Decreasing Increasing -------.02 -.00 .01 .02 .98 1.00 Wald 2 (R 2) 13.11* *(.00) 258.70***(.08)
98 higher, stable trajectories of self-control from those with lower, less sta ble levels of selfcontrol.
99 CHAPTER 6: DISCUSSION The current study tested one component of Gottfredson and HirschiÂ’s self-control theory Â– the stability hypothesis. The theory suggests that once set, around the ag es 8 Â– 10, oneÂ’s level of self-control will remain stable throughout the life-course. Alt hough the stability hypothesis has received empirical attention (Arneklev et al. 1998; Burt et al., 2006; Mitchell & MacKenzie, 2006; Winfree et al., 2006; Turner & Piquero, 2002), most studies have applied methods that overlook the possibility that not all individuals follow the same developmental patterns. In the current study, both the absolute and relati ve stability of self-control were tested using traditional approaches by apply ing semiparametric group-based trajectory modeling. This approach accounts for the possibility that not all individuals follow the same patterns in development and allows for the identification of distinct developmental trajectories (Nagin, 2005). Only two studie s have applied this method to examine the stability of self-control (Hay & Forrest, 2006; Higgins et al., 2009). Their findings were consistent with those of the current study, identif ying stability for the majority of their samples, yet, also identifying s maller groups that were marked by change. The current study extends prior research on this issue by examining whether important risk-factors associated with the development of self-control (i.e ., race, gender, deviant peer association, parenting, parental criminality, and school bonding) var ied
100 across the different trajectory groups that were identified. In general stability in selfcontrol characterized a large majority of the sample; however, several groups were identified that followed trajectories lacking stability. It was also f ound that risk factors other than parenting were associated with group-membership. These findings from the current study are discussed in more detail below. First, a review and summary of t he findings from the different analyses are presented. Second, the implications that these findings have on Gottfredson and HirschiÂ’s theory are discussed. Third, some practic al implications stemming from the current study are offered. Finally, the limi tations of the current study and directions for future research are addressed. Summary of Findings Traditional approaches for assessing the stability of self-control in the current study identify a lack of stability. Mean-level analyses (i.e., repeat ed measure ANOVAs) revealed that there was a significant increase in self-control between t he grades 9 and 10. These significant increases in self-control are inconsistent with Gottfre dson and HirschiÂ’s notions of stability. Although, Gottfredson and Hirschi may allow for a slight gradu al increase in self-control is consistent with their notion that everyone will inc rease in selfcontrol over time due to continued socialization, it would not be expected that these increases would be significant. Contrary to Hirschi and GottfredsonÂ’s position, most researchers have interpre ted the stability hypothesis as one of relative stability, which cannot be asses sed with meanlevel analyses, and instead require the use of stability coefficients (e.g., M itchell & McKenzie, 2006; Winfree et al., 2006; Yun & Walsh, 2010). This study also applied the use of stability coefficients in the current sample, which were fairly consistent with prior
101 research. As with prior research, the majority of the coefficients ( rÂ’ s = .43 .61) did not reach the magnitude suggested to be indicative of relative stability (Costa & McCrae, 1988). Based on these results, it is unclear why the stability coefficients we re not higher. One possibility is that measurement error accounts for a lack of relative sta bility reflected in the correlations. Another possible explanation is that the theory is wrong in that se lfcontrol is not entirely stable. However, another possibility that is at odds with Gottfredson and HirschiÂ’s claims about stability is that there is heteroge neity in developmental patterns of self-control. Despite being a widely used method for assessing stability, both ANOVAs a nd stability coefficients are aggregate measures of stability and, there fore, overlook the possibility that not all individuals follow the same developmental patterns in selfcontrol. The current study extended prior research assessing the stability of sel f-control by using a semi-parametric group-based modeling (SPGM) strategy to account for t he possibility that there is heterogeneity in developmental patterns of self-control. For the most part, the results from the SPGM are consistent with those of the ANOVAs and correlations. Self-control appeared to remain stable for the two l argest groups which represented a majority of the sample (n=2465, 73.9%). However, several groups were also identified that followed developmental trajectories for whic h selfcontrol was not stable. While these groups combined to make up a smaller portion of the sample, they were not trivial in size (n = 704, 26.1%). In all, six distinct developmental trajectories of self-control were ide ntified Â– two groups that were stable, two groups that increased, and two groups that decreased in self-control over the four year period. The two stable groups were higher on self-control tha n
102 the other four groups, and both stable groups changed at the same rate; however, one group had a higher mean-level than the other. Both of the increasing groups started out having the lowest and second lowest mean levels of self-control than the other groups a nd increased at the same rate as each other. The decreasing groups both started out ha ving moderate levels, one slightly lower than the other, of self-control, yet, one dec reased at a higher rate than the other. The SPGM models also revealed that several of the trajectories associ ated with the identified groups intersected with one another. This finding is also consistent w ith those of Hay and Forrest (2006) and Higgins et al. (2009) and provides evidence for rank order changes in levels of self-control over time for some of the identified g roups. That is, while most of the sample maintained their respective rank-order, there we re some groups that did not. This interpretation is supported by the results from the ANOVA analyses examining mean differences in self-control across groups over ti me. From the ANOVAs, it was observed that mean differences in self-control between groups we re not consistent over the four grades. Some groups were actually higher on self-control tha n others at grade seven assessments, but were lower than the same group/s at grade ten assessments. These changes in mean-level differences in self-control between groups over the four year period are evidence that self-control is not relatively s table for some groups and that rank-order changes do occur. The next step of the analyses was to identify what factors might be associate d with fluctuations and/or stability in self-control. This was carried out using AN OVAs and MLOGITs. The results from the ANOVAs identified several unique differe nces across the groups. The general pattern of findings suggest that those groups wi th less
103 stable trajectories tend to have more risk factors compared to groups that have more stable trajectories. The findings from the ANOVA also underscore the uniqueness of the groups identified in the trajectory models. For example, significant difference s were found between the low-increasing and moderate-increasing groups on both school commitme nt and mother attachment. Similarly, the high-stable and moderate-high stable groups di ffer significantly on several risk factors, including mother attachment, school com mitment, school rules, and father parenting. The moderate-high decreasing group was also found to have significantly higher levels of school commitment than the moderate-decre asing groups. These differences in risk factors suggest that the groups identified in the trajectory analysis are distinct in that they have unique relationships with t he risk factors. In order to identify the relative importance of risk factors, an MLOGIT wa s conducted. The results from the MLOGIT were consistent with the ANOVA, reveal ing that those groups with lower, less stable trajectories were more likely to e xperience inadequate parenting, have poor school commitment, be exposed to peers who engage in a variety of delinquent behavior, and have parents with a criminal history, compared to those in the high stable group. In order to examine if risk factors distinguished more general patterns of development, those groups that followed similar patterns were collapsed into three different groups Â– increasing, decreasing, and stable. Using MLOGIT analysi s, it was examined if the same risk-factors distinguished these groups. The result s suggested that minority youths with moderate levels of self-control who associate with devia nt peers, lack commitment to school, and feel unattached to their mother are more likely to
104 experience instability in self-control compared to those who maintain high le vels of selfcontrol, a result underscored in the prior MLOGIT analysis. Implications for Theory and Research The results of the current study have important theoretical implications for both theory and research regarding the stability of self-control. The findings s uggest that Gottfredson and HirschiÂ’s contentions are incorrect and suggest that the theory ne eds to be modified to account for the empirical finding that self-control is not stable in t he absolute or relative sense. The findings of the current study are consistent with prior research identifyi ng a lack of stability in self-control (Burt et al., 2006; Mitchell & McKenzie, 2006; Wi nfree et al., 2006). More importantly, they are consistent with those of prior research employi ng group-based methods to assess the stability of self-control (Hay & Forrest 2006; Higgins et al., 2009) and similar traits such as impulsivity (Cote et al., 2002) and aggression (Johnson et al., 2007; Schaeffer et al., 2003). Similar to the current study, these studies found that the majority of individuals in their samples followed normative patterns of development (i.e., decreases in aggression and impulsivity and increases in self-c ontrol). However, these studies consistently identified smaller, yet non-trivial, groups that followed unique developmental trajectories Â– with both increasing and decreasi ng trajectories identified. Based on the current study and prior research it can be concluded that the theory needs to be modified in a way that accounts for the possibility that some individuals do not follow the same developmental patterns in self-control. The fact that there a re any individuals that demonstrate unique developmental trajectories is inconsistent w ith what
105 the theory would suggest. The empirical evidence tends to suggest that some individuals may begin to acquire self-control much later, as evidenced by the low-increa sing group found in the current study. The current study and the two prior studies (Hay & Forre st, 2006; Higgins et al., 2009) that employed trajectory analysis to model the stabilit y of self-control find similar portions of individuals who evince patterns of change in the development of self-control. Together, these studies suggests that some may begin developing self-control as early as other groups, yet do so at a much slower rate as evidenced by the moderate-increasing group. However, it is important to note that the time frame of the current analysis was only four years, which may not be suffic ient to fully capture developmental trajectories. As such, the notion of difference s in developmental patterns is only speculative at this point. More research is needed tha t begins examining the development of self-control much earlier and ends much later in t he life course. It is also of importance to understand why some individuals might not begin to develop self-control until much later or develop self-control at a much slower rate. Drawing from theoretical explanations of offending trajectories, it may be that these unique developmental patterns are marked by differences in neurological or psychological deficits that occur early in the life course (Moffitt, 1993). Res earch has also suggested that differences in the development of self-control can be linked to ge netic factors (e.g., Beaver et al., 2008). Such deficits may hinder, but not completely ha lt, the development of self-control. More research is needed that directly examines the l ink between the early existing deficits and differences in developmental tr ajectories of selfcontrol.
106 It is also important to address the finding that some individuals experienced loss es in self-control. This finding is somewhat surprising given that Gottfredson a nd Hirschi were fairly adamant that self-control could not be lost once acquired. Although thi s finding may be partially attributable to measurement error, the size of the groups decreasing in self-control are large enough to suggest that this not a trivial finding. Additionally, decreasing trajectories have been corroborated in prior rese arch using group-based methods (Hay & Forrest, 2006; Higgins et al., 2009) and the percentages of individuals evincing decreases in the development of self-control (12 Â– 16% of the samples) are similar across these studies providing more confidence in this finding. One possible explanation can be drawn from Baumeister, Heatherton, and Tice (1994). They developed the self-regulatory strength model, suggesting that the self-regula tory process, a concept similar to self-control, is analogous to a muscle and can become fatig ued after periods of continued use. While this has been tested and supported in laboratory experiments, it has not been examined in field experiments. However, this concept may be extended to the idea that long-term stressful environmental circumstances ( e.g., family, neighborhood, and school) can place undue strain on ones capacity for self-control and result in declines in self-control over longer periods of time. Although the exact nature (e.g., number, shape, and size) of the trajectory groups identified vary across studies employing group-based methods, they consist ently identify groups with distinct developmental trajectories of self-control (Hay & Fo rrest, 2006; Higgins et al., 2009). These findings underscore the potential for more traditional methods of assessing stability to overlook heterogeneity in developmental patter ns of self-control. Thus, it may be necessary for research examining the stabil ity of self-
107 control, and research on self-control in general, to employ methods that can account for this heterogeneity (e.g., Blonigen, 2010). The findings of the SPGM are also in direct contrast to Gottfredson and HirschiÂ’ s claim that those individuals lowest on self-control will remain the lowest and, in tur n, will always be responsible for a disproportionate amount of crime. This notion suggest s that there are no typologies of offenders (e.g., chronic, late-onset, non-offender s) and that there are only two distinct groups: offenders (those low on self-control) and non-offenders (those high on self-control). However, the findings of the current study a re more consistent with research that has identified unique offending trajectori es (e.g., Blokland & Nieuwbeerta, 2005; Chung et al., 2002). In general, this research identifies four to six different offending groups (e.g., non-offending, adolescent-limited/la te-onset, chronic). It is possible that the trajectories of self-control map onto these gr oups that are typically identified. For example, groups that follow high-stable trajector ies of selfcontrol may be consistent with the non-offending groups. The groups decreasing in se lfcontrol may also underlie the adolescent-limited or late-onset offending gr oups and the increasing self-control groups are consistent with groups characterize d by desistance. It is less clear what self-control trajectory might parallel the chronic offending group. More research, however, is needed that directly links developmental trajectories of self-control to offending trajectories to better understand how trajectories of self-contr ol are associated with offending trajectories. The results from the MLOGITs also have important implications for both theory and research. These findings suggest that factors other than parenting disting uish among different developmental trajectories of self-control. Although causal order is not
108 established in the current study, these findings do suggest that deviant peer assoc iation and school commitment are associated with the development of self-control. Additionally, the findings suggest that these factors have an impact on levels of se lfcontrol past the age that Gottfredson and Hirschi suggest that they no longer would. The finding that sources of socialization other than parenting are responsible for the development of self-control provides support for the social causation model. Briefly, the social causation model suggests that social factors can influence offending and development. However, this finding is in direct contrast to Gottfredson and HirschiÂ’ s notion of parenting as the primary source of self-control and that self-control is unresponsive to any socialization past the ages 8 Â– 10. The findings of the current study suggest that both school commitment and deviant peer association are associated w ith certain developmental patters. These findings are in line with prior resear ch indentifying socializing agents, including those other than parenting, that impact oneÂ’s level of selfcontrol beyond ages 8 Â– 10 (e.g., Burt et al., 2006; Hay & Forrest, 2006; Meldrum, 2008). Thus, it is quite possible that these risk factors do influence the development of sel fcontrol. While this may be possible, the potential dynamic nature of these risk factor s limits the ability to make claims about their relationship with levels of sel f-control over time. As previously mentioned, the current analyses do not allow for making causal statements about the relationship between group membership and risk factors. Thus, an alternative explanation, and one more consistent with the theory, is the social selec tion model. In contrast to the social causation model, the social selection model sugges ts that an individualÂ’s criminal propensity influences their social relationships. As suc h, any
109 relationship between social factors and offending is spurious. It may be that bec ause individuals have lower levels of self-control, they are more likely to associa te with deviant peers and have fewer concerns about long-term educational goals. Previous research offers support for social selection and causation models (Wright, Cas pi, Moffitt, & Silva, 1999); it may be the case here that both processes are occurring as well. Another more static risk factor found to be important in differentiating among the identified groups was parental criminality. It was found that the low-increasi ng group, moderate-increasing group, and moderate-high decreasing group were more l ikely to have a parent with a criminal history than the high-stable group. It was also f ound that having a parent who had been to jail or prison distinguishes those groups with lower yet increasing levels of self-control from the high-stable group, but did not distingui sh between those groups with decreasing levels of self-control from the stable gr oup. This finding is somewhat perplexing in that it suggests that having a parent who had been to jail or prison is likely to result in an increase in levels of self-control over ti me. For the most part, previous research suggests that parental imprisonment has a negative i mpact on children (see e.g., Parke & Clarke-Stewart, 2003); however, one possibility is t hat the relationship between the child and the parent before legal intervention was poor a nd the incarceration of the parent resulted in a more stable environment, conducive to the prope r development of the child (Katz, 2002). This may be the case if the children were place d with others who provided better parenting upon the incarceration of the parent, or that the incarceration resulted in the removal of a parent from the home who had a negative impact on the childÂ’s development. Although parental criminality is more static than the other risk-factors included in the analyses, it is important to keep in mind that this
110 variable does not capture undetected parental criminality and does not reflect t hose that may have had criminal justice contact after the first measurement peri od. Furthermore, it could also be the case that the participants were unaware of their parentÂ’s pre vious incarceration. These caveats should be considered. It was also found that deviant peer association distinguished those with unstable trajectories from those with higher, more stable trajectories of self-c ontrol. This finding is inconsistent with Gottfredson and HirschiÂ’s claim that oneÂ’s development of sel fcontrol is unaffected by the types of peers that individuals associate with. It should be noted that the current analyses do not allow for causal statements to be made, and, therefore, it is unclear if deviant peers are responsible for changes in self -control or if oneÂ’s level of self-control influences with whom one chooses to associate wit h. However, prior research has found evidence that peer association does, in fact, infl uence the development of self-control (e.g., Burt et al., 2006). In sum, the current study findings refute many of the theoretical claims as stated by Gottfredson and Hirschi regarding the stability of self-control. They al so hint at the idea that there are different sources that influence the development of self-c ontrol, which, again, is inconsistent with Gottfredson and Hirschi. Future efforts that more direc tly assess the sources of self-control, and the influences on its development, need to be undertaken. Implications for Policy and Practice The current study also has several important implications for policy and pract ice. First, they provide insight into the timing of intervention. Second, they address the uti lity of self-control as a risk factor itself for later offending. Third, the identif ication of risk
111 factors that distinguish different trajectories of self-control provides a use ful tool for identifying which youth may be most at risk for later offending. Also, the ident ification of these risk factors provides direction for programs regarding what life do mains might make effective targets for programs designed help youth develop self-cont rol. Each of these implications is elaborated below. In general, the results suggest that self-control is malleable for some i ndividuals beyond the ages of 8 Â– 10, especially for those lowest on self-control. These findings suggest that programs similar to those that have been found to be successful in promoting self-control and reducing later delinquency among children ages ten younger (Pi quero et al., 2010) may also be effective in promoting the development of self-control and reducing offending among even older adolescents. This also appears to be the case for those individuals who are of most concern for the criminal justice system Â– those low est on self-control. More specifically, the findings suggest that those with more m oderate levels of self-control are at higher risk for late-onset offending, at least to the extent that their offending is due to a lack of self-control. Therefore, while it may be that programs have success in promoting self-control among even those with the lowest self-contr ol, it is important for these programs to not overlook those with more moderate levels of sel fcontrol that are at-risk of decreasing in self-control over time (see disc ussion above). As a result of the current study finding that self-control is not relativel y stable (i.e., changes in rank-order positioning), the utility of self-control as a risk fa ctor for offending is somewhat limited for specific subgroups. For example, those lowest on selfcontrol at age 13 may not be the lowest at age 16 and, in turn, may no longer be the most prone to offending during the later years of adolescence. Accordingly, those youth who
112 evince higher levels of self-control early in adolescence may end up being the lowes t on self-control in later years. Thus, it may not be completely accurate to ident ify youth as being at high risk for later offending based on their level of self-control at ag e ten. Risk factors found in the current study to distinguish between those with higher, more stable trajectories from those with lower, less stable trajectorie s may be useful to those responsible for treatment and prevention in the identification of youth who might be most at-risk for offending. The current study also offers some direction for pr ograms by identifying factors that might be associated with fluctuations and/or stab ility in selfcontrol. The results identify several factors associated with traject ories of self-control suggesting that there are multi-determinates of how oneÂ’s self-control dev elops. This suggests that a multisystemic approach may be most effective in promoting the positive development of self-control. A multisystemic program would account for the various potential factors affecting the development of self-control and target multiple life domains in order to increase chances for success (Borduin et al., 1995). Although not empirically addressed in the current study, another possibility is tha t these risk factors interact with self-control to either promote or prevent offendi ng. Previous research has suggested that prosocial and antisocial ties have the greatest impact in reducing and increasing offending, respectively, among those low on self-control (Wright et al., 1999). Additionally, crime prevention efforts, such as deterrence, are most effective in reducing offending among those with low self-control (Wright, C aspi, Moffitt, & Paternoster, 2004). Thus, programs that create buffers from offending ( e.g., prosocial ties) and limit antisocial influence (e.g., deviant peers) may be e ffective in reducing offending behavior among those lowest in self-control.
113 Whether programs are designed to reduce offending by promoting the development of self-control or by creating buffers from offending, the current s tudy suggests that interventions that focus on promoting proper parenting, creating school-based programs that increase educational investment, and creating prosocial a ctivities for peer groups may be most effective in reducing offending. Limitations The findings from the current study must be considered in light of several important limitations. A few limitations of the current study may limit the representativeness of the sample. First, the target sample was much large r than the final sample size due to a lack of active parental consent. While it is impossible to determine the characteristics of those that did not participate in the study, it is concern ing that they may have not participated for reasons that could bias the results. Most relevant t o the current study is if non-participation was due to levels of self-control. It is q uite possible that parents did not supply consent due to lacking self-control. However, it may also be that parents did not consent because they were concerned about their childÂ’s involvement in the study (e.g., taking away from academics, psychological well-being, etc.) which would reflect high levels of self-control. In either event, it is impossible to empirically state why some parents did not provide consent. More concerning, however, is that those individuals not included in the current analyses due to having too few measurements of self-control were found to differ significantly on several key study variabl es. It is especially troubling that those not included were lower on self-control. Thus, the res ults from the current study may be somewhat biased. However, if these individuals wer e included in the sample it is likely that they would have contributed to those groups lower
114 on self-control, either increasing the size of the currently identified groups or revealing more variation in the number of groups. A second limitation contributing to a lack of representativeness in the sample concerns its racial, ethnic, and social class/economic status. While the current sample may be reflective of the population in Kentucky, the findings may not generalize to other geographic regions or the entire adolescent population. For example, the development of self-control has been found to vary across categories of race and ethnicity ( Turner & Piquero, 2002; Winfree et al., 2006); thus, the current study findings may not generalize to the entire population given the racial, ethnic, and class homogenity of the sample used in the current study. Additionally, the current study dichotomized race into two categories and did not inquire about ethnicity (e.g. Hispanic/Latino). Assessing ra ce in this way may overlook distinct racial and ethnic differences in self-control tha t may exist. However, given the racial homogeneity of the current sample, more nuanced racial categories may have been too small to capture much variation in self-control. The current study is also limited in the measurement of several key concepts and variables included in the analyses. All measures are based on self-report and ar e subject to all the limitations that self-reports suffer (e.g., recall, honesty, e tc.). Other sources may provide more validity in measurement than relying solely on self-repor ts Â– especially in measuring a construct like self-control in which individuals who lack self-cont rol may have trouble providing accurate insight into their own behaviors and attitudes. Another measurement limitation is in regards to measuring race/ethnicity (discuss ed above). Another limitation of the current study is that self-control was observed over a somewhat homogenous developmental period Â– adolescence. The current study does not
115 capture important developmental transitions such as those from early childhood to adolescence or adolescence to young adulthood. Extending trajectories over a l onger period of time may reveal more important transitions in the development of self-cont rol and provide a better picture of how stable (or not) self-control is over different developmental periods. Additionally, the theory contends that levels of self-control should vary considerably among the childhood years of development and that the socialization occurring during this time period is of key importance in shaping it s development (see e.g., Vazsonyi & Huang, 2010). Failure to account for how self-c ontrol develops during this key period of life my overlook how early developmental patterns influence developmental patterns in adolescence. Although several key factors relevant to the development of self-control were identified, there may be several that were not specified in the current study. F or example, one potential factor that may vary across different developmental trajector ies of selfcontrol is neighborhood context. Previous research suggests that neighborhood context has an impact on the development of self-control (Pratt, Turner, & Piquero, 2004; Teasdale & Silver, 2009; c.f., Gibson et al., 2010). Unfortunately, neighborhood variables were not available in the current study. Several other relevant r isk factors may be important in differentiating groups with distinct developmental trajectorie s of selfcontrol not included in the current study (e.g., victimization, abuse, neurological defic its). Future research is needed that examines how these other factors impact the deve lopment of self-control. Another limitation of the current study is that trajectori es of self-control were not linked to offending patterns. Given the proposed link between self-control and
116 offending behavior, future research should also consider the link between developmenta l patterns of self-control and offending patterns. Another limitation is that aside from race, gender, and parental criminality causal order cannot be established between risk factors and developmental trajectories of selfcontrol in the current study. The risk factors in the current study are also ass umed to be static. However, several of the risk factors are likely dynamic in nature, s uch as peer association, parenting, and school bonds. This potential variation in risk factors is not taken into account in the current study. Future research should examine the development of self-control as a process and account for the interaction, mediating, and rec iprocal effects among these factors in order to truly understand the development of selfcontrol. Future research, however, is needed that accounts for the dynamic nature of peer associations in order to establish if deviant peer association fluctuates with levels of selfcontrol. Research is also needed that establishes causal order between self -control and deviant peer association. From the current study it cannot be stated whether i ndividuals gain/lose self-control and then select out of/into deviant peer groups or if deviant pe er associations precede losses in self-control. It is also likely that this proce ss is much more complex. For example, it may be that parenting influences deviant peer associa tion through self-control or that parenting and deviant peer association both have direc t effects on self-control (Brown, Mounts, Lamborn, & Steinberg, 1993). Nonetheless, these factors may be important areas of focus for both future research and policy. Rese arch is needed that untangles how each of these factors are related to the development of s elfcontrol and how each of these contribute to offending. Such research may inform the
117 development of more precise interventions aimed at increasing levels of sel f-control that may, in turn, reduce offending behavior. Another limitation is in regards to the approach taken in the final MLOGIT in which the six identified groups were collapsed into three groups. This approach potentially undermines the results of the SPGM in that it suggests that a 3-group model is a more appropriate fit than the identified 6-group model. This approach was taken in order to create substantively meaningful contrasts not available from the MLO GIT that compared all groups to the high-stable group. It would be possible to run an MLOGIT using each group as the comparison; however, this would produce a somewhat unmanageable set of results. Thus, the current approach in which the groups were collapsed was done for the sake of parsimony and ease of interpretation. Additional ly, the results across the 6-group and 3-group MLOGITs were similar, and, therefore provided justification for taking this approach. It is also important to note the limitations of the current measure of self-con trol. Mainly, the items included in the current measure do not capture all six components of self-control as outlined by Gottfredson and Hirschi. The items used in the current s tudy appear to capture only the impulsive and/or temper elements of self-control. Additionally, the use of self-reports to measure self-control may be biased as i ndividuals lowest on self-control may be more likely to respond inconsistently or inaccurate ly. A multi-method approach may provide a more valid indicator of self-control, such as using mother and/or teacher reports along with self-reports. Finally, it is important to address the issue of Â“reificationÂ” when using gr oupbased trajectory modeling. Essentially, reification is the inclination t o refer to the
118 identified groups as something that actually exist and to view those individuals ident ified as belonging to groups to follow the trajectories lock-in-step. This issue has been raised as a concern on several occasions within the literature (e.g., Nagin, 2000; Nagin, 2005; Nagin & Tremblay, 2005; Raudenbush, 2005; Sampson & Laub, 2005). While the method itself presents a clear concern regarding the interpretations of wha t is a group and the implications that these groups have for both theory and policy, it is less a methodological issue than one of interpretation by the researcher. Thus, it rests i n the hands of the researcher using this method to be aware of this issue and explicitly p rovide caution to readers that these groups are only statistical approximations of t he heterogeneity in development that might actually exist. While the poste rior probabilities reported in the current study suggest a considerable degree of confidence that individuals have been correctly identified as belonging to their respective groups, it must be kept in mind that there may be unidentified variation within groups and that these groups only roughly estimate the degree of heterogeneity that exists.
119 CHAPTER 7: CONCLUSION The current study was only the third study to account for the heterogeneity in t he development of self-control. More importantly, this study expanded on the prior two studies through the examination of certain risk factors and whether or not they differentiated those who followed one trajectory from those who followed another. Despite the limitations of the current study, the findings provide important insight s into the development of self-control. Essentially, the currents study findings invali date the stability hypothesis set forth by Gottfredson and Hirschi. It was found that self-control is not stable for everyone and this is the case for both relative and absolute stability It was also found that certain factors identified in the current study lead to differe nt developmental patterns of self-control. Although it is not clear from the curre nt study what the actual process is regarding the relationship between these factor s and the development of self-control it is clear that these risk factors vary across the different trajectories Â– factors other than parenting. As reviewed above, prior research has been not favorable toward the theory as a whole and its major propositions. In general, research does not support the claims that the theory makes with regards to the unidimensionality of the construct, the fact ors responsible for the development of self-control, its association with crime and delinquency Â– both its explanatory power and as the sole explanation of crime, as well as
120 the stability of the construct. The current study findings are in line with the ex isting body of literature suggesting that self-control is not stable beyond the ages that Gottf redson and Hirschi contend it will become set. Combined with the existing body of litera ture on self-control, the current study findings suggest that, at the very least, the theory needs to be reformulated. Additionally, the current study findings combined with the empi rical status of Gottfredson and HirschiÂ’s theory have implications for criminology as a whole. Based on empirical tests of other propositions of the theory, a reformulation mig ht begin with reevaluating the contention that individuals are born with hedonistic values and that everyone must acquire self-control in order to become law-abiding citizens We do not have to look far to find alternative perspectives regarding the natural tendenci es of humans from birth (e.g., Kilpinen, 1999; Mansbridge, 1990; Miller, 1999; Sen, 1977; Veblen, 1898). Similarly, the factors responsible for the development of self-contr ol may need to be reconsidered. Research continues to suggest that other social factors (e .g., Turner et al., 2005; Burt et al., 2006) and biological factors (e.g., Beaver et al ., 2008Â’ Beaver et al., 2009) influence oneÂ’s level of self-control. Research also sugge sts that self-control is not the sole explanation of crime, that other variables matte r, and that it may interact with other theoretical constructs such as social learning va riables to explain antisocial behavior (Pratt & Cullen, 2000). Research needs to continue to examine how these unique constructs work together to result in crime in order to gain a more thorough understanding and better explain its existence. From there the theory can be re worked to make more precise statements about the phenomenon of crime. The current body of literature, including the current study, also suggests a mor e critical approach be taken toward the general theory. The theory has, at best received
121 only Â“partial support.Â” Scientific principles suggest that there is no such thi ng as partial support. If a theory does not receive full support it should be rejected. By these cri teria, self-control and its many propositions should be discarded. This has wider implicat ions for the discipline of criminology as a whole. While the idea of a single, simple explanation of crime and delinquency is attractive from both a theoretical and prac tical standpoint, criminologists need to recognize that the phenomenon of crime is much mor e complicated than self-control theory contends. Although the construct of self-contr ol appears to be one factor in explaining crime (Pratt & Cullen, 2000), its role in cr ime is not in line with the theoretical propositions laid out by Gottfredson and Hirschi. Des pite a continued lack of support for the theory, researchers continue to exert time and energy into testing the propositions of the theory as laid out by Gottfredson and Hirschi. This is a major set-back to the discipline of criminology as a whole. In order to advance the discipline, researchers must let go of the propositions made by many theories that have time and time again been disproved, take from the existing theories what has been found to be fruitful, and build from there.
122 CHAPTER 8: LIST OF REFERENCES Akers, R. L. (1991). Self-control as a general theory of crime. Journal of Quantitative Criminology, 7(2) 201Â–11. Arneklev, B. J., Cochran, J. K., & Gainey, R. R. (1998). Testing Gottfredson and HirschiÂ’s Â“low self-controlÂ” stability hypothesis: An exploratory study American Journal of Criminal Justice, 23(1), 107 Â– 127. Arneklev, B. J., Grasmick, H. G., & Bursik, R. J. (1999). Evaluating the dimensionality and invariance of Â“low self-controlÂ”. Journal of Quantitative Criminology, 15(3), 307 Â– 331. Arneklev, B. J., Grasmick, H. G., Tittle, C. R., & Bursik, R. J. (1993). Low self-control and imprudent behavior. Journal of Quantitative Criminology, 9(3), 225 Â– 247. Baron, S. W. (2003). Self-control, social consequences, and criminal behavior: Street youth and the general theory of crime. Journal of Research in Crime and Delinquency, 40, 403 Â– 425. Baumeister, R. F., Heatherton, T. F., & Dianne M. T. (1994). Losing Control: How and Why People Fail at Self-Regulation San Diego, CA: Academic Press. Beaver, K. M., DeLisi, M., 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 Â– 42. Beaver, K. M., Ratchford, M., & Ferguson, C. J. (2009). Evidence of genetic and environmental effects on the development of low self-control. Criminal Justice and Behavior, 36, 1158 Â– 1172. Beaver, K. M., Schutt, J. E., Boutwell, B. B., Ratchford, M., Roberts, K., & 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 Â– 60.
123 Beaver, K. M., Wright, J. P., DeLisi, M., & Vaughn, M. G. (2008). Genetic influences on the stability of low self-control: Results from a longitudinal sample of twins Journal of Criminal Justice, 36, 478 Â– 485. 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 Â– 181. Blokland, A. A. J. & Nieuwbeerta, P. (2005). The effects of life-circumstances on longitudinal trajectories of offending. Criminology, 43, 1203 Â– 1240. Blonigen, D. M. (2010). Explaining the relationship between age and crime: Contributions from the developmental literature on personality. Clinical Psychology Review, 30, 89 Â– 100. Borduin, C. M., Mann, B. J., Cone, L. T., Henggeler, S. W., Fucci, B. R., Blaske, D. M., & Williams, R. A., Multisystemic treatment of serious juvenile offenders: Lon gterm prevention of criminality and violence. Journal of Consulting and Clinical Psychology, 63, 569 Â– 578. Boutwell, B. B. & Beaver, K. M. (2010). The intergenerational transmission of low s elfcontrol. Journal of Research in Crime and Delinquency, 47, 174 Â– 209. Brown, B. B., Mounts, N., Lamborn, S. D., & Steinberg, L. (1993). Parenting practices and peer group affiliation in adolescence. Child Development, 64, 467 Â– 482. 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 ge neral theory of crime. Criminology, 44(2), 353 Â– 396. Cairns, R. B., Leung, M., Buchanan, L., & Cairns, B. D. (1995). Friendships and social networks in childhood and adolescence: Fluidity, reliability, and interrelations Child Development, 66(5), 1330 Â– 1345. Carmines, E. G., & Zeller, R. A. (1979). Reliability and validity assessment. Beverly Hills, CA: Sage. Caspi, A., Robert, B. W., & Shiner, R. L. (2005). Personality development: Stability and change. Annual Review of Psychology, 56, 453-484. Chapple, C. L. (2005). Self-control, peer relations, and delinquency. Justice Quarterly, 1, 89 Â– 106.
124 Cheung, N. W. T. & Cheung, Y. W. (2008). Self-control, social factors, and delinquency: A test of the general theory of crime among adolescents in Hong Kong. Journal of Youth and Adolescence, 37, 412 Â– 430. Chung, I. Hill, K. G., Hawkins, D., Gilchrist, L. D., & Nagin, D. S. (2002). Childhood predictors of offense trajectories. Journal of Research in Crime and Delinquency, 39, 60 Â– 90. Cochran, J. K., Aleska, V., & Chamlin, M. B. (2006). Self-restraint: A study on the capacity and desire for self-control. Western Criminology Review, 7, 27 Â– 40. Cochran, J. K., Wood, P. B., Sellers, C. S., Wilkerson, W., & Chamlin, M. B. (1998). Academic dishonesty and low self-control: An empirical test of a general theory of crime. Deviant Behavior, 3, 227 Â– 255. Cohen, L. E. & Vila, B. J. (1996). Self-control and social control: An exposition of the Gottredson-Hirschi/Sampson-Laub debate. Studies on Crime and Crime Prevention, 5, 125 Â– 150. Cohn, E. G. & Farrington, D. P. (1999). Changes in the most-cited scholars in twenty criminology and criminal justice journals between 1990 and 1995. Journal of Criminal Justice 27 345-359. Cohn, E. G. & Farrington, D. P. (1998). Changes in the most-cited scholars in major criminology and criminal justice journals between 1986 1990 and 1991 1995. Journal of Criminal Justice 26(2) 99-116. Costa, P. T., & McCrae, R. R. (1988). Personality in adulthood: A six-year longitudinal study of self-reports and spouse ratings on the NEO Personality Inventory. Journal of Personality and Social Psychology, 54 853Â– 863 Cote, S., Tremblay, R. E., Nagin, D., Zoccolillo, M., & Vitaro, F. (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 Â– 618. Cretacci, M. A. (2008). A general test of self-control theory: Has its importa nce been exaggerated? International Journal of Offender Therapy and Comparative Criminology, 52, 538 Â– 553. DÂ’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 regre ssion analyses. The American Journal of Sociology, 103, 1593 Â– 1630. De Kemp, R. A. T., Vermulst, A. A., Finkenauer, C., Scholte, R. H. J., Overbeek, G., Rommes, E. W. M., & Engels, R. (2009). Self-control and early adolescent anti-
125 social behavior: A longitudinal analysis. Journal of Early Adolescence, 29, 497 Â– 517. DeLisi, M. (2001). It's all in the record: Assessing self-control theory w ith an offender sample. Criminal Justice Review, 26, 1-16. DeLisi, M., Hochstetler, A., & Murphy, D. S. (2003). Self-control behind bars: A validation study of the Grasmick et al. scale. Justice Quarterly, 20, 241 -263. Fergusson, D. M., Horwood, L. J., & Nagin, D. S. (2000). Offending trajectories in a New Zealand Cohort. Criminology, 38, 525 Â– 552. Fergusson, D. M. & Horwood, L. J. (2002) Male and female offending trajectorie s. Development and Psychopathology, 14 159-177. Finkenauer, C., Engels, R. C. M. E., Baumeister, R. F. (2005). Parenting and adolescent behavioral and emotional problems: The role of self-control. International journal of behavioral development, 29, 58 Â– 69. Gibbs, J. J., Giever, D. & Higgins, G. E. (2003). A test of Gottfredson and HirschiÂ’s general theory using structural equation modeling. Criminal Justice and Behavior, 30, 441 Â– 458. Gibbs, J. J., Giever, D., & Martin, J. S. (1998). Parental management and self-control: An empirical test of Gottfredson and HirschiÂ’s general theory. Journal of Research in Crime and Delinquency, 35 40Â–70. Gibson, C., Sullivan, C. J., Jones, S., & Piquero, A. R. (2010). Â“Does it take a village?Â” Assessing neighborhood influences on childrenÂ’s self-control. Journal of Research in Crime and Delinquency, 47, 31 Â– 47. Gibson, C. & Wright, J. (2001). Low self-control and co-worker delinquency: A researc h note. Journal of Criminal Justice, 29, 483 Â– 492. Gottfredson, M. & Hirschi, T. (1990). A General Theory of Crime Stanford, CA: Stanford University Press. Gottfredson M. & Hirschi, T. (2003). Self-control and opportunity. In C. Britt & M. Gottfredson (Eds.), Advances in Criminology Volume 12: Control Theories of Crime and Delinquency (pg. 5 Â– 20). New Brunswick, NJ: Transaction Publishers. Grasmick, H. G., Tittle, C. R., Bursik, R. J., & Arneklev, B. J. (1993). Testing the core empirical implications of Gottfredson and HirschiÂ’s general theory of crime. Journal of Research in Crime and Delinquency, 30(5), 5-29.
126 Hay, C. (2001). Parenting, self-control, and delinquency: A test of self-control theory Criminology, 39 707Â–36. Hay, C. & Forrest, W. (2006). The development of self-control: Examining self-control theoryÂ’s stability hypothesis. Criminology, 44(4), 739 Â– 774. Higgins, G. E. (2005). Can low self-control help with the understanding of the soft-war e piracy problem? Deviant Behavior, 26, 1 Â– 24. Higgins, G. E., & Makin, D. A. (2004). Does social learning theory condition the effects of low self-control on college studentsÂ’ software piracy? Journal of Economic Crime Management 2 1Â–22. Higgins, G. E., Fell, B. D., & Wilson, A. L. (2007). Low self-control and social learnin g in understanding studentsÂ’ intentions to pirate movies in the United States. Social Science Computer Review, 25, 339 Â– 359. Higgins, G. E., Jennings, W. G., Tewksbury, R., & Gibson, C. L. (2009). Exploring the link between low self-control and violent victimization trajectories in adolesc ents. Criminal Justice and Behavior, 36, 1070 Â– 1084. Hirschi, T. & Gottfredson, M. (1993). Commentary: Testing the general theory of cri me. Journal of Research in Crime and Delinquency, 30, 47 Â– 54. Holtfreter, K., Reisig, M .D., Piquero, N. L., & Piquero, A. R. (2010). Low self-control and fraud: Offending, Victimization, and their overlap. Criminal Justice and Behavior, 37, 188 Â– 203. Hope, T. L. & Chapple, C. L. (2004). Maternal characteristics, parenting, and adoles cent sexual behavior: The role of self-control. Deviant Behavior, 26, 25 Â– 45. Johnson, W., Hicks, B. M., McGue, M., & Iacono, W. G. (2007). Most of the girls are alright, but some arenÂ’t: Personality trajectory groups from ages 14 to 24 and some associations with outcomes. Journal of Personality and Social Psychology, 93, 266 Â– 284. Jones, B. L., & Nagin, D. S. (2007). Advances in group-based trajectory modeling and a SAS procedure for estimating them. Sociological Methods and Research, 35, 542Â–572. Jones, B. L., Nagin, D. S., & Roeder, K. (2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods & Research, 29, 374 Â– 393.
127 Junger, M., West, R., & Timman, R. (2001). Crime and risky behavior in traffic: An example of cross-sectional consistency. Journal of Research in Crime and Delinquency, 38, 439 Â– 459. Katz, A. (2002). Parents in Prison. Criminal Justice Matters, 50, 18 Â– 19. Kerley, K. R., Hochstetler, A., & Copes, H. (2009). Self-control, prison victimization, and prison infractions. Criminal Justice Review, 34, 553 Â– 568. Kilpinen, E. (1999). What is rationality? A new reading of VeblenÂ’s critique of util itarian hedonism. International Studies of Politics, Culture, and Society, 13(2) 187 Â– 206. Lamiell, J. T. (1981). Toward an idiothetic psychology of personality. American Psychologist, 36(3) 276 Â– 289. Land, K., McCall, P.. & Nagin, D. (1996). A comparison of Poisson, negative binomial, and semiparametric mixed Poisson regression models with empirical applic ations to criminal careers data. Sociological Methods & Research, 24, 387 440. Langton, L. (2006). Low self-control and parole failure: An assessment of ris k from a theoretical perspective. Journal of Criminal Justice, 34, 469 Â– 478. 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 Â– 238. Lipsey, M. (1992). Juvenile delinquency treatment: A meta-analytic inquiry into the variability of effects. In T.D. Cook, H. Cooper, D.S. Cordray, H. Hartmann, L.V. Hedges, R.J. Light, T.A. Louis, and F. Mosteller (Eds.), Meta-Analysis for Explanation: A Casebook (83 Â– 127). New York, NY: Russell Sage Foundation. Lipsey, M. 1999. Can intervention rehabilitate serious delinquents? Annals of the American Academy of Political and Social Science 564:142Â–166. Longshore, D., Chang, E., Hsieh, S., & Messina, N. (2004). Self-control and social bonds: A combined control perspective on deviance. Crime & Delinquency, 50, 542 Â– 564. Longshore, D., Turner, S., & Stein, J. A. (1996). Self-control in a criminal sample: An examination of construct validity. Criminology, 34(2), 209-228. Mansbridge, J. J. (1990). Beyond Self Interest Chicago, IL: Chicago University Press. Marcus, B. (2004). Self-control in the general theory of crime: Theoretical implications of a measurement problems. Theoretical Criminology, 8(1), 33 Â– 55.
128 Mason, W. A. & Windle, M. (2002). Gender, self-control, and informal social control in adolescence: A test of three models of the continuity of delinquency behavior. Youth & Society, 33, 479 Â– 514. Meldrum, R. C. (2008). Beyond parenting: An examination of the etiology of selfcontrol. Journal of Criminal Justice, 36, 244 Â– 251. Miller, D. T. (1999). The norm of self-interest. The American Psychologist, 54(12), 1053 Â– 1060. Mitchell, O. & Mackenzie, D. L. (2006). The stability and resiliency of selfcontrol in a sample of incarcerated offenders. Crime & Delinquency, 52, 432 Â– 449. Moffitt, T. E. (1993). Adolescence-limited and life-course persistent antisoci al behavior: A developmental taxonomy. Psychological Review, 100 674-701. Morris, G. D., Wood, P. B., & Dunaway, R. G. (2006). Self-control, native traditionalism, and Native American substance use: Testing the cultural invariance of a general theory of crime. Crime and Delinquency, 52, 572 Â– 598. Morris, G. D., Wood, P. B., & Dunaway, R. G. (2007). Testing the cultural invariance of parenting and self-control as predictors of American Indian delinquency. Western Criminology Review, 8(1) 32 Â– 47. Mroczek, D. K. (2007). The analysis of longitudinal data in personality research. In R .W. Robins, R.C. Fraley & R.F. Krueger (Eds.), Handbook of research methods in personality psychology (pp. 543-556). New York: Guilford Press. Nagin, D. (1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4(2), 139 Â– 157. Nagin, D. (2000). Response to Â“methodological sensitivities to latent class analys is of long-term criminal trajectoriesÂ”. (2000). Journal of Quantitative Criminology, 20, 27 Â– 35. Nagin, D. S. (2005). Group-Based Modeling of Development Cambridge, Massachusetts: Harvard University Press. Nagin, D. S., Farrington, D. P., & Moffitt, T. E. (1995). Life-course trajecto ries of different types of offenders. Criminology, 33 ll-40. 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.
129 Nagin, D. & Paternoster, R. (1993). Enduring individual differences and rational choice theories of crime. Law & Society Review, 27(3), 467 Â– 496. Nagin, D. & Paternoster, R. (2000). Population heterogeneity and state dependence: Sta te of the evidence and directions for future research. Journal of Quantitative Criminology, 16(2), 117 Â– 144. Nagin, D. S. & Tremblay, R. E. (1999). Trajectories of boys' physical aggression, opposition, and hyperactivity on the path to physically violent and nonviolent juvenile delinquency. Child Development, 70, 1181 1196. Nagin, D. S., & Tremblay, R. E. (2005). Developmental trajectory groups: Fact or use ful statistical fiction? Criminology, 43, 873 Â– 904. Nofziger, S. (2008). The Â“causeÂ” of low self-control: The influence of maternal se lfcontrol. Journal of Research in Crime and Delinquency, 45, 191 Â– 224. Ozbay, O. & Koksoy, O. (2010). Is low self-control associated with violence among youths in Turkey. International Journal of Offender Therapy and Comparative Criminology, 53, 145 Â– 167. Parke, R. D. & Clarke-Stewart, K. A. (2003). The effects of parental incarcera tion on children: Perspectives, Policies, and Promises. In J. Travis & M. Waul (Eds.) Prisoners Once Removed: The Impact of Incarceration and Reentry on Children, Families, and Communities (pp. 189 Â– 131). Washington D.C.: Urban Institute Press. Perrone, D., Sullivan, C. J., Pratt, T. C., Margaryan, S. (2004). Parental efficacy, selfcontrol, 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 Â– 312. Piquero, A. R., Jennings, W. G., Farrington, D. P. (2010). On the malleability of selfcontrol: Theoretical and policy implications regarding a general theory of cr ime. Justice Quarterly, 27 803Â–83. Piquero, A. R., MacDonald, J., Dobrin, A., Daigle, L. E., & Cullen, F. T. (2005). Selfcontrol, violent offending, and homicide victimization: Assessing the general theory of crime. Journal of Quantitative Criminology, 21, 55 71. Piquero, A. R., Macintosh, R., & Hickman, M. (2000). Does self-control affect survey response? Applying exploratory, confirmatory, and item response theory analys is to Grasmick et al.'s self-control scale. Criminology, 38, 897-930. Piquero, A. R., Moffitt, T. E., & Wright, B. E. (2007). Self-control and criminal career dimensions. Journal of Contemporary Criminal Justice, 23, 72 Â– 89.
130 Piquero, A. R., and Rosay, A. B. (1998). The reliability and validity of Grasmick et al .Â’s self-control scale: A comment on Longshore et al. Criminology 36: 157Â–174. Piquero, A. & Tibbets, S. (1996). Specifying the direct and indirect effects of low s elfcontrol and situational factors in offenderÂ’s decision making: Toward a more complete model of rational offending. Justice Quarterly, 13, 481 Â– 510. Perrone, D., Sullivan, C. J., Pratt, T. C., Margaryan, S. (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 Â– 312. Phythian, K., Keane, C., & Krull, C. (2008). Family structure and parental behavior: Identifying the sources of adolescent self-control. Western Criminology Review, 9, 73 Â– 87. Piquero, A. R., Blumstein, A., Brame, R., Haapanen, R., 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 Â– 74. Piquero, A. R. & Rosay, A. B. (1998). The reliability and validity of Grasmick et al.Â’s self-control scale: A comment of Longshore et al. Criminology, 36(1) 157-173. Polakowski, M. (1994). Linking selfand social control with deviance: Illuminating t he structure underlying a general theory of crime and its relation to deviant activi ty. Journal of Quantitative Criminology 10 41-78. Pratt, T. C. & Cullen, F. T. (2000). The empirical status of Gottfredson and HirschiÂ’ s general theory of crime: A meta-analysis. Criminology, 38 931-64. Pratt, T. C., Turner, M. G., & Piquero, A. R. (2004). Parental socialization and community context: A longitudinal analysis of the structural sources of low self-control. Journal of Research in Crime and Delinquency, 41 219Â–243. Raffaelli, M., Crockett, L. J., & Shen, Y. (2005). Developmental stability and change i n self-regulation from childhood to adolescence. The Journal of Genetic Psychology, 166 54Â–75. Raudenbush, S. W. (2005). How do we study Â‘what happens next?Â’ Annals of the American Academy of Political and Social Science, 602, 131 Â– 144.
131 Rebellon, C. J., Straus, M. A., & Medeiros, R. (2008). Self-control in global perspective : An empirical assessment of Gottfredson and HirschiÂ’s general theory wi thin 32 national settings. European Journal of Criminology, 5, 331 Â– 361. Ribeaud, D. & Eisner, M. (2006). The Â‘drug-crime linkÂ’ from a self-control perspecti ve: An empirical test in a Swiss youth sample. European Journal of Criminology, 3. 1477 Â– 3708. Roberts, B. W., Caspi, A., & 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 Â– 683. 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 Â– 640. Roeder, K., Lynch, K., & Nagin, D. (1999). Modeling uncertainty in latent class membership: A case study in criminology. Journal of the American Statistical Association, 94, 766 Â– 776. Sampson, R. J. & Laub, J. H. (2003). Life-course desisters? Trajectories of cri me among delinquent boys followed to age 70. Criminology, 41, 301 Â– 340. Sampson. R. J. & Laub, J. H. (2005). Seductions of method: Rejoinder to Nagin and TremblayÂ’s Â“ Developmental trajectory groups: Fact or fictionÂ”. Criminology, 43, 905 Â– 913. Schaeffer, C. M., Petras, H., Ialongo, N., Poduska, J., & Kellam, S. (2003). Modeling growth in boysÂ’ aggressive behavior across elementary school: Links to lat er criminal involvement, conduct disorder, and antisocial personality disorder. Developmental Psychology, 39, 1020 Â– 1035. Sen, A. K. (1977). Rational fools: A critique of the behavioral foundations of economic theory. Philosophy & Public Affairs, 6(4), 317 Â– 344. Seipler, C. & Eifler, S. (2010). Opportunities, rational choice, and self-control: On t he interaction of person and situation in a general theory of crime. Crime & Delinquency, 56, 167 Â– 197. Smith, T. R. (2004). Low self-control, staged opportunity, and subsequent fraudulent behavior. Criminal Justice and Behavior, 31, 542 Â– 562. Teasdale, B. & Silver, E. (2009). Neighborhoods and self-control: Toward and expanded view of socialization. Social Problems, 56, 205 Â– 222.
132 Tellegen, A. (2006). Manual for the Multidimensional Personality Questionnaire. Minneapolis: University of Minnesota Press. Thaler, R. H. & Shefrin, H. M., 1981. An economic theory of self-control Journal of Political Economy 89(2), 392-406. Tittle, C. R., Ward, D. A., & Grasmick, H. G. (2003). Self-control and crime/deviance: Cognitive vs. behavioral measures. Journal of Quantitative Criminology, 19(4), 333 Â– 365. Tittle, C. R., Ward, D. A., & Grasmick, H. G. (2004). Capacity for self-control and individualÂ’s interest in exercising self-control. Journal of Quantitative Criminology, 20(2), 143-172 Turner, M. G. & Piquero, A. R. (2002). The stability of self-control. Journal of Criminal Justice, 30, 457 Â– 471 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 Â– 339. Unnever, J. D. & Cornell, D. G. (2003). Bullying, Self-control, and Adhd. Journal of Interpersonal Violence, 18, 129 Â– 147. Unnever, J. D., Cullen, F. T., & Agnew, R. (2006). Why is Â“bad: parenting criminogenic? Implications from rival theories. Youth Violence and Juvenile Justice, 4, 3 Â– 33. Unnever, J. D., Cullen, F. T., & Pratt, T. C. (2003). Parental Management, ADHD, and delinquent involvement: Reassessing Gottfredson and HirschiÂ’s general theor y. Justice Quarterly, 20(3), 471-500. Vazsonyi A. & Belliston, L .M. (2007). The family Â– Low self-control Â– Deviance : A cross-cultural and cross-national test of self-control theory. Criminal Justice and Behavior, 34, 505 Â– 530. Vazsonyi, A. & Huang, L. (2010). Where self-control comes from: On the development of self-control its relationship to deviance over time. Developmental Psychology, 46(1), 245 Â– 257. Vazsonyi, A. T., Pickering, L. E., Junger, M., & Hessing, D. (2001). An empirical test of a general theory of crime: A four-nation comparative study of self-control and the prediction of deviance. Journal of Research in Crime and Delinquency, 38, 91 Â– 131. Vazsonyi, A. T., Wittekind, J. E. C., Belliston, L. M., & van Loh, T. D. (2004). Extending the general theory of crime to Â“the EastÂ”: Low self-control in Japanes e adolescents. Journal of Quantitative Criminology, 20 (3), 189-201.
133 Veblen, T. (1898). Why is economics not an evolutionary science? The Quarterly Journal of Economics, 12(4), 373 Â– 397. Weisner, M. & 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 Â– 262. Winfree, L. T., Taylor, T. J., He, N., & Esbensen, F. (2006). Self-control and variability over time: Multivariate results using a 5-year, multisite panel of youths. Crime & Delinquency, 52, 253 Â– 286. Wood, P., Pfefferbaum, B., & Arneklev, B. (1993). Risk-taking and self-control: Soc ial psychological correlates of delinquency. Journal of Crime and Justice, 16 11130. Wright, J. & Beaver, K. (2005). Do parents matter in creating self-control in thei r children? A genetically informed test of Gottfredson and HirschiÂ’s theory of l ow self-control. Criminology, 43, 1169 Â– 1202. Wright, J., Beaver, K., DeLisi, M., & Vaughn, M. (2008). Evidence of negligible parenting influences on self-control, delinquent peers, and delinquency in a sample of twins. Justice Quarterly, 25, 544 Â– 569. Wright, B. R. E., Caspi, A., Moffitt, T. E., & Paternoster, R. (2004). Does the perceived risk of punishment deter criminally prone individuals? Rational choice, self-control, and crime. Journal of Research in Crime and Delinquency, 41, 180 Â– 213. Wright, B. R. E., Caspi, A., Moffitt, T. E., & Silva, P. A. (1999). Low self-control, social bonds, and crime: Social causation, social selection, or both? Criminology, 37(3) 479 Â– 514. Yun, I. & Walsh, A. (2010). The stability of self-control among South Korean adolescents. International Journal of Offender Therapy of Comparative Criminology, xx(x), 1 Â– 15.
APPENDIX A: LIST OF MEASURES Self-control 1. I have difficulty remaining seated at school. 2. I get very restless after a few minutes if I am supposed to sit still. 3. When I am angry, I lose control over my actions. 4. I have difficulty keeping attention on tasks. 5. I get so frustrated that I feel like a bomb ready to explode. 6. Little things or distractions/interruptions throw me off. 7. IÂ’m nervous or on edge. 8. I canÂ’t seem to stop moving. 9. I donÂ’t pay attention to what I am doing. 10. I am afraid I will lose control of my feelings. Mother Attachment 1. My mother seems to understand me. 2. My mother helps me with my homework. 3. My mother makes me feel wanted. 4. I share my thoughts and feelings with my mother. 5. I do things (example: watch TV, go to sports events, go to dinner, and so on) with my mother. Mother Supervision 1. My mother knows where I am when I am away from home. 2. My mother is concerned with how I am doing in school. 3. My mother knows who I am with when I am away from home. 4. My mother sets a time for me to be home at night. 5. My mother makes rules that seem fair to me. Father Parenting 1. My father seems to understand me. 2. My father helps me with my homework. 3. My father makes me feel wanted. 4. I father my thoughts and feelings with my mother. 5. I do things (example: watch TV, go to sports events, go to dinner, and so on) with my father. 6. My father knows where I am when I am away from home. 2. My father is concerned with how I am doing in school. 3. My father knows who I am with when I am away from home.
135 4. My father sets a time for me to be home at night. 5. My father makes rules that seem fair to me. Belief in School Rules 1. Students have a say in making school rules. 2. Everyone knows what the school rules are. 3. The school rules are fair. 4. The punishment for breaking school rules is the same for all students no matter who you are. 5. The school rules are strictly enforced. 6. If a school rule is broken, students know what kind of punishment will follow. 7. The teachers keep order in the classrooms. School Commitment 1. I care a lot what my teachers think of me. 2. Getting an education is important to me. 3. I look forward to coming to school most mornings. 4. I would quit school now if I could. 5. Most of my classes are a waste of time. Deviant Peer Association Out of those closest friends Â– How many of them have done any of the following things in the present school year? 1. Smoked marijuana. 2. Cut school completely. 3. Driven after drinking. 4. Been suspended from school. 5. Taken a gun to school. 6. Taken an explosive to school. 7. Taken a weapon to school (knife, brass knuckles, and so on, other than gun or explosive). 8. Gotten arrested. 9. Sold marijuana or other drugs. 10. Stolen someoneÂ’s money or property when they were not around. 11. Physically attacked someone (e.g., punched, slapped, kicked). 12. Vandalized public or private property (e.g., destroyed property, graffiti, and s o on).