Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=udbh20

Deviant Behavior

ISSN: 0163-9625 (Print) 1521-0456 (Online) Journal homepage: http://www.tandfonline.com/loi/udbh20

Does Self-esteem Negatively Impact Crime andDelinquency? A Meta-analytic Review of 25 Yearsof Evidence

Carrie Mier & Roshni T. Ladny

To cite this article: Carrie Mier & Roshni T. Ladny (2018) Does Self-esteem Negatively ImpactCrime and Delinquency? A Meta-analytic Review of 25 Years of Evidence, Deviant Behavior, 39:8,1006-1022, DOI: 10.1080/01639625.2017.1395667

To link to this article: https://doi.org/10.1080/01639625.2017.1395667

Published online: 21 Nov 2017.

Submit your article to this journal

Article views: 341

View Crossmark data

Does Self-esteem Negatively Impact Crime and Delinquency? AMeta-analytic Review of 25 Years of EvidenceCarrie Mier and Roshni T. Ladny

Indiana University East, Richmond, IN, USA

ABSTRACTSelf-esteem has been deemed an important variable in many theories over thepast 100 years, and several interventions are aimed at improving self-esteem inorder to prevent criminal and delinquent behavior. However, no systematicattempt to quantify the collective results of this variable’s effect on criminaland delinquent behavior has been undertaken, and only one narrative litera-ture review has taken place. This meta-analysis seeks to inform the research byfocusing on 48 effect sizes from 42 studies spanning 25 years (1990–2015) and71,130 individuals. Our findings indicate that self-esteem has a small, negativebut significant effect on crime and delinquency.

ARTICLE HISTORYReceived 5 December 2016Accepted 17 January 2017

Self-concept is a decidedly human function, acting upon biological drives and genetic predispositions as aprotective part of our personality (Bouchard 2004). The self-concept represents the proverbial looking-glass self (Matsueda 1992) and sets the foundation fromwhich a sense of worth, competence, confidence,and self-esteem can develop. All of these constructs are inter-related, but self-esteem in particular hasbeen studied as an individual psychological concept for more than 100 years (Leary 1999). Although self-esteem and self-concept are used interchangeably in research and practice, there are importancedifferences. Self-concept is how one understands one’s self, such as an understanding or knowing ofpersonal likes, dislikes, temperament, and morals. Self-esteem refers to the feelings one has toward one’sself and how one may feel positively or negatively about their identity (Campbell 1990). Self-esteem,however, may be influenced by how well one knows their identity; the greater clarity of self-concept, thehigher one’s self-esteem (Campbell 1990; Erikson 1959; Marcia 1980). This makes self-esteem a uniquevariable to study, especially in relation to delinquency in crime, since individuals strive to feel good aboutthemselves regardless of the situation (Leary 1999; Zimmerman et al. 1997).

After the rise of interactionist social psychological theories in the 1950s, researchers began toexplore the impact of self-concept and related constructs alongside traditional sociological factors toexplain deviant behavior. There has been much discussion and research on the importance of self-esteem in particular to explain variation in human behavior, specifically maladaptive behavior andother psychological issues (Leary and Baumeister 2000). Given the interest in the relationshipbetween self-esteem and human behavior, naturally researchers have become more interested inthe relationship between crime and self-esteem. One of the early theoretical rationales for the linkbetween self-esteem and crime, and other maladaptive behavior is the frustration-aggression hypoth-esis (Dollard et al. 1939), in which aggression stems from the blockage of any goal directed activity.The individual differences in tendencies to behave aggressively result from individual differences inresponse to frustrations. One’s response to such frustration is in part a result of an individual’s self-esteem and/or self-concept (Berkowitz 1989; Scheff, Retzinger, and Ryan 1989).

CONTACT Carrie Mier cwarstle@iue.edu School of Humanities and Social Sciences, Tom Raper Hall 363, Indiana UniversityEast, 2325 Chester Boulevard Richmond, Indiana 47374; Roshni Ladny rl11e@my.fsu.edu*This manuscript is an original piece of work that is not currently under review at any other journal or source simultaneously andhas not been published elsewhere© Taylor & Francis Group, LLC

DEVIANT BEHAVIOR2018, VOL. 39, NO. 8, 1006–1022https://doi.org/10.1080/01639625.2017.1395667

The explanatory of role of self-esteem on criminal and delinquent behavior is complex andinconclusive with some studies finding support for a positive relationship between self-esteem andcrime or aggression (Kernis 1993) while others finding support for a negative relationship (Osner2006), and yet additional studies showing very weak support, if any, in either direction (Gendreau,Little, and Goggin 1996). There is also some argument that delinquency and criminal behavior is asign of “normal” behavior and adjustment among adolescents while maladjustment might be moreapparent in those who abstain from crime (Hendrix 2016). Notably, the differences in relationshipsbetween self-esteem and crime result from the variation in types of crimes committed, such asviolent versus property crimes. However, even when one examines the body of studies exploring theimpact of self-esteem on select crime types, findings are often mixed. Depending on the study,findings show that both low self-esteem (Donnellan, Trzensniewski, and Robins 2005; Osner 2006)and high self-esteem (Baumeister and Boden 1998; Papps and O’Carroll 1998) can lead to violence,aggression, and antisocial behavior while other research indicates a protective over a risk effect(Trzesniewski, Donnellan, Moffitt, Robins, Poulton, and Caspi 2006; Boden, Fergusson, and JohnHorwood 2007; Harris 2011; Ostrowsky 2010; Steinke 2012).

Some of these mixed findings may be due to variations in the conceptualization of self-esteem,similarities between high self-esteem and narcissism that are not addressed (Bushman andBaumeister 1998), as well as variations in the use of conditional variables when examining the linkbetween self-esteem and crime. For example, when exploring the effects of esteem stability on therelationship between crime and self-esteem, findings show that individuals with unstable high self-esteem are more likely to respond aggressively to ego threats than those with stable high self-esteem(Kernis 1993). Perhaps studies that do not find significant effects of high self-esteem on aggressionhave not specified the level of esteem stability or other potential moderators in conjunction with self-esteem. The mixed findings of the effects of self-esteem on aggression are only a small portion of thegreater research area in need of further study when examining the link between self-esteem anddelinquency and crime in general, which unfortunately continues to remain under-researched. Therehas to date been only one narrative literature review, conducted by Baumeister et al. (2003), whichexplored the relationship between self-esteem and juvenile delinquency. Their finding of a weakrelationship between self-esteem and juvenile delinquency in their in-depth literature review may bean indicator of some of the methodological issues mentioned above. The current study seeks toexpand the research and explore a more nuanced connection between self-esteem, delinquency, andcrime using quantitative meta-analytical methods.

The need for self-esteem’s effects on crime and delinquency to be adequately researched aremany, as much attention is paid to the concept of self-esteem in areas of rehabilitation and personaldevelopment. Various therapeutic interventions are based upon the premise that improving anindividual’s self-esteem will benefit them in many ways and steer them away from negative outcomeslike delinquency and criminal behavior (Baumeister et al. 2003; Leary 1999). Schools often advocatefor involvement in extra-curricular activities like sports and various after-school programs toincrease self-esteem and general well-being (Faulkner et al. 2007; Stogner and Gibson 2010).Numerous prison programs focus on helping inmates improve self-confidence and self-esteemwith hopes that high self-esteem and high self-confidence can serve as a foundation upon whichinmates can build adaptive coping skills, thus reducing recidivism (Eitle, Taylor, and Pih 2010;Gillespie 2005; Hubbard 2006). Even religious activities for both youths and adults have portions thatfocus on building self-esteem (Harris 2011).

However, if the effect of self-esteem is moderated by other factors such that the high self-esteemdoesn’t always lead to positive outcomes and low self-esteem doesn’t always lead to negative outcomes,then programs meant to increase self-esteem to shift individuals away from negative outcomes may nothave the intended effect. A recent study by Hendrix (2016) highlights this dichotomy by pointing outthat some abstainers from criminal and delinquent activities could actually be considered maladjustedor “loners” which would put them in an at-risk group despite not engaging in crime and delinquency.Similarly, self-esteem’s effects may be more potent at particular points in the life-course like childhood

DEVIANT BEHAVIOR 1007

and adolescence compared to adulthood (Trzesniewski, Brent Donnellan, and Robins 2003). Differentgroups of people, such as those from certain cultures or youth that are considered at risk, may also feelthe benefits or potential harms of self-esteem more than other groups (Bachman et al. 2011). Tounderstand how self-esteem operates in relation to crime and delinquency, it is necessary to isolate theprotective effect of self-esteem while paying heed to its potential risks. This need becomes even morecrucial when considering the amount of financial resources and time that may be put into programs thatdo not work the way they should.

The current meta-analysis focuses on identifying the role of self-esteem on the negative outcome ofdelinquency and crime. As mentioned, Baumeister et al. completed a literature review on this topic in2003 and found a weak, although generally negative influence of self-esteem on delinquency. Although,we also similarly hypothesize an overall negative effect of self-esteem on delinquency and crime will berevealed from the studies analyzed, more than 12 years have passed since the initial 2003 review. Wewonder if the problems present in self-esteem research have been addressed in more recent studies, thuspotentially affecting the conclusions we might reach while conducting meta-analysis and providingmore insight into how self-esteem operates in relation to delinquency and crime.

This article will proceed as follows: first, study acquisition and scope of studies included will bediscussed. Second, relevant issues that arose during a literature review of these studies will beilluminated. Third, methods for conducting statistical analysis through meta-analysis will be pre-sented. Fourth, results from meta-analysis in terms of relevant effect sizes for main and moderatinghypotheses will be explored. Lastly, conclusions concerning the protective, negative effect of self-esteem on delinquency will be assessed along with areas of future research.

Methods

The present study

The goal of this project is to conduct a quantitative meta-analysis of the effect of self-esteem ondelinquency and crime. To date, there has not been a systematic meta-analysis completed and onlyone prior literature review exists. Meta-analysis as a whole can be a complicated and burdensomeundertaking rife with issues of bias or comparing studies that differ vastly in quality (Cohen 1988; LipseyandWilson 2001). However, when undertaken carefully, meta-analysis can aid in the overall synthesis ofa large body of work with often inconsistent findings in an objective manner (Cooper 2010; Pigott 2012).This function is an important one as many researchers often use different models, methods, measures,and theories in their studies that can substantially affect their results. Using meta-analytical methods canhelp standardize and find precise estimates of relationships among these studies to then be used forcomparison among otherwise incomparable groups (Cohen 1988; Cooper 2010). Our main hypothesisfollows the prior literature and findings of Baumeister et al. (2003): we expect to see a small but negativeimpact of self-esteem on delinquency and crime. Additionally, we will be examining several relevantissues within the literature as moderators to see if the effect of self-esteem changes substantially. Lastly, afail-safe n measure will be conducted to address the “file-drawer problem” or the tendency for peer-review journals to pay attention to only significant findings and to neglect null ones (Rosenthal 1979).This sort of sensitivity analysis will provide additional evidence that suggests the estimates are reliableassessments of the relationship between self-esteem and delinquency/crime.

Search criteria and study acquisition

To begin the meta-analysis, we first had to acquire relevant studies that focused on our researchquestion. To be included, studies had to have self-esteem as a predictor variable and some measureof delinquent behavior or criminal involvement as a dependent variable. Self-esteem was defined ashaving a good view of yourself, seeing yourself as worthwhile, and in a generally positive light. Toachieve this, a global dimension rather than a dimensional scale was used for consistency and to

1008 C. MIER AND R. LADNY

avoid measuring a concept other than self-esteem (such as feelings on one’s physical attractiveness orpopularity). Global self-esteem is typically measured with the Rosenberg Self-Esteem Scale (RSES) asused by Rosenberg (1985) which we discuss later, but studies having any measure of overall self-worth or self-concept were included. As indicated in the Introduction, there is a substantivedifference between self-esteem and self-concept; however, we analyzed the use of the term “self-concept” to see how it matched up to self-esteem. To be included in this meta-analysis, the measurehad to be a global one related to sense of self and feelings of self-worth which many studies that used“self-concept” were actually measuring. Self-esteem also did not have to be the main independentvariable to be included in this meta-analysis, and we searched for studies where self-esteem was usedas a control variable.

We limited our search of dependent variables to just delinquency and crime. Delinquency wasdefined as problematic and antisocial behaviors that could be viewed as illegal when committed byjuveniles, and crime was anything that is against the law and for which someone could beprosecuted, punished, or imprisoned. Some studies would use measures of externalizing behaviorslike “acting out, talking back, or rule breaking,” use status offenses like running away or truancy, orfocus on difficult personalities like conduct disorder or Attention Deficit Hyperactivity Disorder(ADHD); these studies were not included in the meta-analysis. Delinquency is typically measuredusing a frequency scale like the Self-Reported Delinquency scale created by Elliott, Huizinga, andAgeton (1985), but any study that included measures of crime and delinquency including count,dichotomous measures, and other frequencies were also included in this meta-analysis.

To be included in this meta-analysis, studies also had to test a bivariate statistical relationshipbetween self-esteem and delinquency/crime at a minimum. This requirement was to ensure that wecould gather a relevant effect size which was Pearson’s correlation coefficient r for our meta-analysis.Using Pearson’s correlation coefficient allows for an effect size that is easy to interpret and that hasmany formulas available for converting other sample statistics (i.e. t, f, χ2) into r (Cooper 2010; Mullin1989; Rosenthal 1984). If bivariate correlations were not reported but a suitable sample statistic wasavailable to convert into r, these results would be included in the study. However, many studies did notreport correlation coefficients and instead merely reported multivariate, conditional results (i.e.regression coefficients); these studies were not included in our meta-analysis due to these coefficientsbeing generally considered inappropriate for meta-analytical comparison (Peterson and Brown 2005).1

Additionally, we wanted to maintain independence within our findings so we did not allow overlapwithin samples or large datasets (i.e. Add Health). To remain objective about which findings toinclude, we used the most recent study using overlapping samples for an effect size. Maintainingindependence also applied to multiple findings from the same study in different subgroups (male vs.female; Black vs. White) and types of criminal behavior (property vs. violent crime). If a study hadmultiple findings from the same sample,2 we averaged the results into one effect size.

We did not place a geographic restriction on the studies included in our meta-analysis so thereare study results that occur outside of the United States, but included studies did have to be inEnglish so we could more easily read and interpret relevant findings. We did place a restriction onthe earliest year of publication we included in our study which was 1990. The reason for this timeframe restriction was that Rosenberg’s initial self-esteem and delinquency paper was published in1989, so we wanted to take advantage of articles published after this point. The time frame concludedwith 2015 being the most recent year of publication. Lastly, we made an attempt to get at non-published works like dissertations and theses using Dissertation Abstracts International and did notlimit our meta-analysis to just published results in order to address the “file-drawer problem.”

1This decision lead to 12 studies being excluded from our analyses. These studies include Hoffmann and Cerbone (1999), Broidy(2001), Fagan and Western (2003), Gillespie (2005), Gallupe (2006), Hubbard (2006), Baron (2007), Papadakaki et al. (2009), Eitle,Taylor, and Pih (2010), Harris (2011), Van De Schoot and Wong (2012), and Peck (2013).

2A few studies included in our meta-analysis used different samples, dependent variables, and methods within the samepublication (i.e. Brent et al. (2005); in these cases, multiple effect sizes were collected from the same study and included inthe final meta-analysis.

DEVIANT BEHAVIOR 1009

We endeavored to find as many articles as we could that looked at the relationship between self-esteem and crime/delinquency. We used several different combinations of repeated search termsacross many different online databases including ProQuest Research Library, Web of Science,PsycInfo, Criminal Justice Abstracts, Dissertation Abstracts International, Social Science CitationIndex, WorldCat, and Google Scholar. The search terms included the following: self-esteem, self-worth, self-concept, self-derogation, delinquency, crime, criminal involvement, and several combina-tions of these terms. We also used the references of studies included in our meta-analysis to bring inother studies that we might have missed in our initial search. After applying our search criteria, weisolated 42 studies containing 48 effect sizes over 73,130 individuals from 1990–2015.3 The studiesincluded in this meta-analysis are marked with an asterisk in the references section.

It is important to note that although we developed and used a clear classification for whichstudies would be used in the meta-analysis, we did not use an actual ranking scale (such as theUniversity of Maryland Scientific Scoring Procedure which ranks studies on a scale of one to fivebased on quality) to assign numerical values to the quality of each study. Our rationale for not usinga similar ranking procedure is that (1) to compare our findings to Baumeister et al. (2003) literaturereview findings, we needed to include a variety of studies as their literature review contained, evenones with certain methodological flaws that may have been given a lower ranking if a scale wereused, and ultimately excluded and (2) if certain methodological shortcomings are indeed moderatingthe differences in the effect of self-esteem on crime, then they can be isolated and looked atindividually as opposed to the impact of the overall quality of the study. For example, a study canbe assigned a level 5 on a scale of 1 to 5 because of the use of an experimental design, but have anextremely small sample size. By simply looking at the numerical value assigned to a study, it can bedifficult to isolate the effects of one particular methodological issue, such as sample size.

Issues in the literature review

More nuanced studies that examine the relationship between self-esteem and delinquency in con-junction with potential moderating factors are needed; however, there are numerous methodologicalproblems that arise when attempting to study self-esteem. One common problem is figuring out howto conceptualize self-esteem for study. Some researchers see this trait as stable and predictive of goodoutcomes like healthy relationships, overall well-being, academic achievement, and positive peerrelationships in high doses while attributing to substance abuse, delinquency, health problems,depression, and anti-social behavior in low quantities (Rosenberg, Schooler, and Schoenbach 1989;Trzesniewski, Brent Donnellan, and Robins 2003; Zimmerman et al. 1997). Other studies find thatself-esteem acts as a “psychological sociometer” that fluctuates and shifts depending on the socialcontext one finds themselves in Leary (1999)and Leary and Baumeister (2000). This finding meansthat self-esteem will be high when things are going well and low when things are not so good. Theability of self-esteem to change in this way and remain stable at the same time has led to differentscale constructions to measure this trait (Trzesniewski, Brent Donnellan, and Robins 2003). Globalself-esteem scales capture how an individual generally feels about themselves and their self-worth,while specific self-esteem scales capture how an individual feels about themselves in relation to aparticular life dimension, such as academic performance or physical attractiveness. Both scales canbe used to assess change in self-esteem over time, such as changes with developmental stages.

3Studies included in the meta-analysis are Luengo et al. (1994); Martinez (1995); Ross (1995); Wilson (1995); Edwards (1996);Heaven (1996); Peiser and Heaven (1996); Rigby and Cox (1996); Katims, Yin, and Zapata (1997); Jang and Thornberry (1998);Brendgen, Vitaro, and Bukowski (1998); Shah (1999); Ellickson and McGuigan (2000); Min and Kim (2000); Caldwell (2000); Stiles,Liu, and Kaplan (2000); Levy (2001); Mason (2001); Jajoura et al. (2002); Moran and DuBois (2002); Tiêt and Huizinga (2002);Costello and Dunaway (2003); Jang and Johnson (2003); Parker and Benson (2004); Brent et al. (2005); Holsinger and Holsinger(2005); Owens-Sabir (2005); Thornton, Beech, and Marshall (2005); Walker and Gudjonsson (2006); Barry et al. (2007); Boden,Fergusson, and John Horwood (2007); Faulkner et al. (2007); Gonzalez (2008); Johnson and Morris (2008); Barry, Pickard, andAnsel (2009); Ferris et al. (2009); Lau et al. (2011); Church et al. (2012); Lee and Lee (2012); Steinke (2012); Cheng (2014); andCheng and Cheng (2015).

1010 C. MIER AND R. LADNY

(Baumeister et al. 2003; Rosenberg, Schooler, and Schoenbach 1989; Trzesniewski, Brent Donnellan,and Robins 2003; Van De Schoot and Wong 2012).

Despite the creation of two scales that are able to handle the fluctuation of self-esteem and thusaddress some of the conceptualization problem, measurement issues arise since self-esteem can beboth a cause and an effect of various correlates. Rosenberg et al. noted in 1989 that the problem ofcausal order presented an issue that researchers needed to be aware of when trying to decipher self-esteem’s impact. Though it makes sense that since self-esteem is a personality trait that is biologicallyinfluenced it would be first in the causal chain (Bouchard 2004), there still is a concern aboutreciprocal causation. This concern becomes even more important when looking at delinquency as anoutcome. For example, engaging in delinquency can act as a mechanism to increase self-esteemrather than high self-esteem acting as a protective factor against delinquency (Mason 2001; Faulkneret al. 2007; Lee and Lee 2012; Owens 1994).

A review of prior literature indicated four main areas where self-esteem research suffers: havingsmall, unrepresentative samples, the lack of longitudinal designs, and the validation level of both theself-esteem measure and delinquency/crime measure used (Boden, Fergusson, and John Horwood2007; Ostrowsky 2010; Trzesniewski, Brent Donnellan, and Robins 2003). Unfortunately, each of thestudies included in our meta-analysis suffers from these problems to some extent.

The first methodological area of critique is the lack of large, representative samples. Prior researchstates that many samples used in self-esteem and delinquency research are at-risk or are delinquent,clinical samples (Brent et al. 2005; Trzesniewski et al. 2006). These individuals could have inflated high orartificially low levels of self-esteem due to their situations in life and being in a center of treatment, thusrisk level of sample might actually moderate the effects of self-esteem on delinquency. Additionally,many studies make use of cherry-picked convenience samples or school-based samples (Boden,Fergusson, and John Horwood 2007; Broidy 2001) which makes it hard to establish external validityor how these results would apply to the real world beyond adolescence. The size of these samples alsocreates a concern of how stable the regression results are between self-esteem and delinquency/crime.

The second area of methodological critique is the lack of a longitudinal design. Many studies inself-esteem research are cross-sectional or short-term longitudinal with perhaps one year betweentime points, creating difficulty in assessing growth over time and in some cases causal orderimpossible to assess (Rosenberg, Schooler, and Schoenbach 1989; Trzesniewski et al. 2006;Trzesniewski, Brent Donnellan, and Robins 2003; Zimmerman et al. 1997). Similarly, early self-esteem research also indicated that self-esteem and delinquency grew together over time using latentgrowth modeling as a strategy to isolate this effect (Mason 2001). The failure to use long periods oflongitudinal analysis undermines the ability of studies to isolate the long-term effects of self-esteemon negative outcomes like delinquency. Technically better studies may take advantage of longerstudy periods or longitudinal designs in general to address causal order.

The third methodological area of critique involves measuring the independent variable ade-quately. Prior research advocates using scales to tap into the different dimensions of self-esteem orinsist that multiple measures of self-esteem be used to make sure no dimensions were missed(Boden, Fergusson, and John Horwood 2007; Leary 1999; Leary and Baumeister 2000;Trzesniewski, Brent Donnellan, and Robins 2003). The most well-known and accepted of thesescales is the RSES. The use of this particular scale is meant to tap into global or overall self-esteem sostudies that use this particular 10-item scale may be stronger methodologically than studies that usea self-esteem measure not validated in research.

The fourth methodological area of critique involves how the dependent variable is measured. Onevery specific problem is that there is a chance that the self-report nature of self-esteem shares acommon component with other self-report measures like delinquency or crime (Baumeister et al.2003). When an individual reports a high level of self-esteem, there is a desire to remain consistentwithin other areas. This consistency can lead to negative self-report measures not getting reportedaccurately (Trzesniewski et al. 2006). However, Baumeister et al. (2003) noted that frequencymeasures of delinquency were more reliable than count measures as respondents had to think

DEVIANT BEHAVIOR 1011

about the actual amounts committed instead of answering either yes or no. So studies that measuredelinquency and crime as a frequency measure may also be technically better than those that usecount measures. Delinquency and criminal behavior might also have differing relationships with self-esteem. Prior research points out that areas of violence in particular could be driven by self-esteemor more importantly by egotism or high self-esteem (Boden, Fergusson, and John Horwood 2007;Ostrowsky 2010), so studies that use more serious criminal behavior as an outcome may havedifferent results than those that focus only on delinquency.

These four methodological issues in self-esteem research pose a threat to the validity of theresearch findings. To study the impact of these issues on the nature of self-esteem research findings,we make an effort to assess the level of heterogeneity among the samples, and then identify whatrelevant factors produce this heterogeneity, thus allowing for the identification of factors thatmoderate the effects of self-esteem on delinquency and crime.

Statistical analysis

The effect size calculated across the studies was the correlation coefficient or Pearson’s r (Hunter andSchmidt 2004). Studies used in this meta-analysis had to either have a bivariate relationship reportedor had to have a sample statistic that could be converted into r. If a study only reported multivariatecoefficients, it was not included in the meta-analysis due to multivariate results being highlydependent on what variables are included in the regression model which could introduce bias ifcompared to non-beta correlation coefficients (Peterson and Brown 2005). Regardless of how r wasderived, all of the correlation coefficients were transformed into Fisher’s Z due to the skeweddistribution of the standard errors of the correlation coefficient (Lipsey and Wilson 2001; Wolf1986). This standardized transformation allows the sampling distribution to approach normality(Blalock 1972) and allows for synthesis of results that would otherwise be incomparable. Aftersynthesis, the Fisher’s Z statistic is transformed back into r for interpretation.

The quantitative technique of meta-analysis was used to combine each study’s effect size (Cooper2010; Pigott 2012). We used the inverse-variance, random-effects model since we assumed that anyheterogeneity present is due to between study variability (true heterogeneity) and within-studyvariability (sampling error) (Huedo-Medina et al. 2006). Under this inverse-variance random effectsmodel, studies with larger samples provided finer precision and thus were given greater weight in thecalculation of the average weighted effect size. The equation is as follows:

rj ¼wi� riwi

where r̄j represents the average correlation for the jth correlation (i.e. outcome for delinquency/crime), wi represents the inverse-variance, random-effects weight for study i, and ri represents theaverage correlation for study i. Confidence-intervals and hypothesis testing can be accomplishedusing the standard error of the average weighted effect size (Borenstein et al. 2005; Lipsey andWilson 2001).

Additionally, our studies span 25 years and many different measures and methods, so it isimportant for us to see whether our effect sizes vary significantly according to study characteristics.An analysis of homogeneity or the Q statistic which follows the χ2 distribution was calculated for theoverall sample and followed by a hypothesis test.4 We expect for significant heterogeneity to befound in our sample, so moderator analyses will also be conducted to see if they account for some ofthe variation seen in our effect sizes. We hypothesize that effect sizes will vary by type of research

4The equation for calculating Q is as follows: Σ(w*ES²) – [Σ(w*ES)]²/Σw where w = inverse-variance weight and ES = effect size. Todetermine the significance of moderator effects, a Qw or variance-within groups Q value was calculated for each group in themoderator analysis, added together and then subtracted from the overall Q to create a Qb or a variance-between groups Q value(Cooper 2010). This value is reported in Table 2, and significant values indicate that the moderator has a significant impact oneffect sizes for self-esteem and delinquency/crime.

1012 C. MIER AND R. LADNY

design (longitudinal vs. cross-sectional), use of a validated self-esteem scale (RSES), measurement ofcrime and delinquency as a frequency, and the population used in the study (representative vs. non-representative). These moderators are those cited most often as important issues in the literaturereview. We also study whether the country the study took place in (international vs. domestic) andthe year of publication (1990–2003 vs. 2004–present) matter as context and time period could impactthe relationship of self-esteem with delinquency/crime (Gentile, Twenge, and Keith Campbell 2010).The year 2003 was chosen as the cut-off year for publication date due to Baumeister et al.’s (2003)narrative literature review being published then, so we wanted a before-after assessment of potentialfield improvement after its publication.

Lastly, a fail-safe n is calculated for the overall sample and each of the moderators.5 This valueindicates how many null findings would be needed to overturn the results and make them merely“chance findings” in order to correct for publication bias or the “file-drawer problem” (Rosenthal1979). This number is given to offer some context for our findings and show the robustness of ourresults; however, it should be noted that the fail-safe n is not a perfect measure and relies upon itsown assumptions of what studies are being left out of the meta-analysis (Fragkos, Tsagrias, andFrangos 2014).

Results

Table 1 indicates the outcome for self-esteem’s overall effect on delinquency and crime for the entiresample of 42 studies which produced 48 independent effect sizes as well as individual outcomes forboth crime and delinquency. The average weighted effect size is –.10 which is small in magnitude(Cohen 1988) but in the hypothesized direction and statistically significant (p < .001, 95% CI [–.11,–.09]). 46% of the findings were also in the hypothesized direction and significant. Though this effectsize does not seem that large, it does fit in with other meta-analyses of criminal behavior where testsof large, well-established theories like self-control and social learning theory result in modest effectsizes of around .20 (see Pratt and Cullen 2000; Pratt et al. 2010). Self-esteem is also traditionally usedas a conditional, social support variable when analyzing criminal and delinquent behaviors (Agnew1992), so finding a smaller effect size is not that surprising. The fail-safe n indicates that 358 nullfindings would be needed to overturn these results which supports the robustness of this effect size.However, there is also a high degree of heterogeneity within our sample (Q = 5967, p < .001), so it isnecessary for us to conduct moderator analyses to understand the potential causes of heterogeneity.

Average weighted effect sizes were derived separately for both delinquency (33 effect sizes) andcrime (15 effect sizes). The effect size for delinquency alone is stronger than the overall effect size(wES = -.12, p < .001, 95% CI [–.13, –.11]), and its fail-safe n is also quite robust (FSN = 528). Theeffect size for crime alone though is quite small (wES = –.04, p < .001, 95% CI [–.05, –.03]), though itis still statistically significant and in the hypothesized direction. This difference in crime groupscould point to delinquency as having a stronger relationship with self-esteem over crime. Thepercentage of significant findings also lends some support to this conclusion as 61% of delinquency

Table 1. The effect of self-esteem on overall delinquency/crime and delinquency and crime separately.

wES % Significant CI (95%) FSN Q

Overall (48) −.10 46.00% −.11 - -.09 358 5967Delinquency (33) −.12 61.00% −.13 - -.11 528 4749Crime (15) −.04 20.00% −.05 - -.03 8 1109

=p < .001; wES = average weighted effect size; CI = confidence interval (95%), FSN = fail-safe n; () = the number of findings pereffect size is in parentheses

5The equation for the fail-safe n is reported as follows for significance at the .05 significance level: FSN = (Σz-scores/1.645)² – Nwhere N is the number of findings.

DEVIANT BEHAVIOR 1013

findings were in the hypothesized direction compared to 20% of crime findings. Interestingly, about33% of findings for crime as the outcome variable pointed to the opposite hypothesis where self-esteem had a significant positive impact on criminal involvement instead of a negative one. Thisfinding supports Baumeister et al.’s (1996) “dark side of self-esteem” argument where high self-esteem could create a narcissism effect increasing criminal and delinquent behavior instead of aprotective, social support reduction (Baumeister, Smart, and Boden 1996). This possibility could alsoexplain why the Q scores still illustrate high levels of heterogeneity for both the delinquency and thecrime groups.

Moderator analysis

Table 2 indicates the moderators we assessed for varying effects in the relationship between self-esteemand delinquency/crime. The first moderator we examine is the type of research design used—long-itudinal vs. cross-sectional studies. The effect sizes were similar to one another with cross-sectionalstudies being about as strong as the overall effect size (wES = –.10, p < .001, 95% CI [–.12, –.08] for cross-sectional studies and wES = -.09, p < .001, 95% CI [-.10, -.08] for longitudinal studies). Cross-sectionalstudies had a slightly higher percentage of findings in the significant, negative direction, and the fail-safen’s for both are stable despite being at a lower threshold for longitudinal studies. Though both effect sizesare significant and in the hypothesized direction, their close proximity is backed up by the insignificantQvalue which indicates that study design is not a significant moderator.

The second moderator we examine is the type of population used—representative vs. non-representative. The significant Q value (Q = 21, p < .001) indicates that there is a significantdifference between results taken from differing populations. Though both of the results are in thedesired and hypothesized direction (wES = –.08, 95% CI [–.09, –.07], p < .001 for representativestudies and wES = -.12, 95% CI [–.13, –.11], p < .001 for non-representative studies), the effect size ofnon-representative studies is much stronger and larger than the overall effect size. This effect sizecould be related to the fact that many of these studies are also cross-sectional which also had a strongeffect size. In terms of strength though, representative studies tended to produce more significantresults in the hypothesized direction compared to non-representative studies. The fail-safe n’s forboth groups are stable and similar in size.

The third moderator we explore is the use of the validated RSES and whether studies used thismeasure compared to a non-validated scale or single measure when gathering information aboutself-esteem. The Q value (Q = 29, p < .001) indicates that there is a statistically significant differencebetween the studies where this scale was used and not used, and the effect sizes seem to reflect this.While the results are both in the predicted direction (wES = –.07, 95% CI [–.08, –.06], p < .001 forstudies using the RSES and wES = –.11, 95% CI [–.12, –.10], p < .001 for studies not using the scale),

Table 2. The effect of moderators on the self-esteem and delinquency/crime relationship.

wES % significant CI (95%) FSN Qb

Cross-sectional (25) −.10 48.00% −.12 - -.08 123 2Longitudinal (23) −.09 43.00% −.10 - -.08 34Representative (17) −.08 52.00% −.09 - -.07 69 21Non-representative (31) −.12 42.00% −.13 - -.11 62Validated SE Scale (25) −.07 48.00% −.08 - -.06 91 29Non-validated SE scale (23) −.11 43.00% −.12 - -.10 43Frequency scale D/C (27) −.07 52.00% −.08 - -.06 138 31Non-frequency scale D/C (21) −.11 38.00% −.12 - -.10 16Domestic studies (32) −.11 47.00% −.12 - -.10 138 71International studies (16) −.03 44.00% −.05 - -.01 19Studies up to 2003 (25) −.04 44.00% −.05 - -.03 82 156Studies 2004 and beyond (23) −.13 48.00% −.14 - -.12 51

=p < .01, = p < .001; wES = average weighted effect size; CI = confidence interval (95%), FSN = fail-safe n; Qb = variance betweengroups; () = the number of findings per effect size is in parentheses

1014 C. MIER AND R. LADNY

the results are larger for studies not using the validated scale and its effect size is greater than theoverall effect size. Much of this effect could be driven by the fact that many of these studies were alsocomposed from non-representative populations which had a larger effect size as well. Studies usingthe RSES were more likely to produce significant results in the hypothesized direction, and the fail-safe n’s are stable though lower for the studies not using a validated self-esteem scale.

The fourth moderator we look at is how the dependent variable of delinquency or crime wascaptured, notably whether this measure was calculated through the use of a frequency scale. Thesignificant Q value (Q = 31, p < .001) indicates that there is a significant difference between studiesbased whether they use a frequency scale or not to measure delinquency and crime, and the effectsizes are relatively similar to those seen comparing validated to non-validated self-esteem scale use(wES = –.07, p < .001, 95% CI [–.08, –.06] for studies using a frequency scale for delinquency andcrime and wES = –.11, p < .001, 95% CI [–.12, –.10] for studies not using a frequency scale). Theseeffects closely mirror the results for the use of a validated self-esteem scale because many studies whodo not use a validated self-esteem scale also do not employ a frequency scale for measuringdelinquency and crime. Studies who use a frequency scale to measure delinquency and crime aremuch more likely to find significant results that support the hypothesized relationship between self-esteem and delinquency/crime. The fail-safe n’s are also stable, but the threshold is substantiallylower for the studies that do not use a frequency scale to measure delinquency and crime.

The fifth moderator we consider is the country in which the study was conducted in—domestic vs.international. Though this factor was not raised as a moderator during our literature review, contextcould be very important when determining the effect of self-esteem (Gentile, Twenge, and KeithCampbell 2010). The significant Q value (Q = 71, p < .001) indicates that there is a significant differencein the effect sizes depending upon the country where the study was conducted. While both the effectsizes were significant and in the hypothesized direction, the average effect size for domestic studies(wES = –.11; p < .001; 95% CI [–.12, –.10]) is closer in strength to the overall average effect size than thesmaller effect size for international studies (wES = –.03; p < .01; 95% CI [–.05, –.01]). The strength ofstatistically significant results was similar among both groups, though slightly stronger for domesticstudies, and the fail-safe n for both groups was stable though lower for international studies.

The final moderator we examine is the year of publication which we set at 2003 to reflect theearlier narrative meta-analysis done by Baumeister et al. and the studies done since then. Thesignificant Q value (Q = 156, p < .001) reflects that there is a significant difference in effect sizesbased upon the time period that the study was conducted in. Though both effect sizes are significantand in the hypothesized direction, the effect size for studies 2004 and beyond (wES = –.13, p < .001,95% CI [–.14, –.12]) is substantially larger and the biggest effect size presented in this meta-analysiscompared to studies prior to 2004 (wES = –.03, p < .001, 95% CI [–.05, –.03]). It appears that therehas been some improvement in study quality since the narrative literature review conducted byBaumeister et al. (2003) with 48% of studies producing results in the hypothesized direction, but it isimportant to note that many newer studies still suffer from the methodological concerns presentedin the literature review. The fail-safe n’s are stable for both groups though lower for studiesconducted 2004 and beyond.

Discussion and conclusions

The present study sought to examine the relationship between self-esteem and delinquency/crime.We hypothesized that we would see an overall negative relationship between these variables, and thatthis relationship would be significantly moderated by several variables of theoretical importance: thetype of research design, the type of population used, the use of a validated self-esteem scale, the useof a frequency scale to measure delinquency/crime, the year that the study was published, and thecountry that the study took place in. Our results validated our hypotheses and show that there is asmall, but significant and negative relationship between self-esteem and delinquency/crime. Thiseffect size increases when studying delinquency alone. Additionally, almost all of our moderators

DEVIANT BEHAVIOR 1015

were significant predictors of differences in effect sizes for this relationship with some moderators—notably non-representative populations, use of a non-validated self-esteem scale, not using afrequency delinquency or crime scale, domestic studies, and studies conducted 2004 and beyond—producing effect sizes that were larger than the overall average effect size. The only moderator thatfailed to reach significance was research design where there was not much difference between resultsfrom cross-sectional and longitudinal studies.

Our study is not without its limitations though. While our study search was quite thorough andsought to include a number of non-published resources like theses and dissertations, there is achance that we may have overlooked some results especially in the cases of self-esteem being used acontrol variable. We did include many null results in our meta-analysis which helps address thisconcern, and the fail-safe n’s are relatively stable for our groups which indicates that it is relativelyunlikely that our results would be overturned due to missing studies. Given that fact that we assessed25 years of research though and in many instances self-esteem was not highlighted as a key term ormain variable of interest but was instead buried as a control variable, it is likely we may have missedsome findings for inclusion in our study. Though our overall effect size is not a large one by anymeans, our results should still be viewed conservatively. Our methods of operationalization for bothself-esteem and delinquency/crime could also have introduced some bias into our study as differingscales may be composed of differing items and thus be different in quality. We tried to be asthorough as possible in terms of using a global self-esteem scale and focusing just on delinquent orcriminal behavior for the dependent variable, but concerns in quality and measurement are alwaysan issue when conducting meta-analysis (Cooper 2010; Lipsey and Wilson 2001).

We also address methodological issues and issues of study quality by looking at potential moderatorsusing Q. By conducting our moderating analysis with Q, we are exploring and controlling for whatvariables may condition the relationship between self-esteem and delinquency/crime. These results alsohelp pinpoint what factors make stronger studies and which ones are part of studies that are not asstrong. One potential issue with using an overall of Q for each groups is that there may be clustering ofmethodological issues among studies (i.e. studies that are non-representative are also cross-sectionaland do not use the RSES), so it would be hard to distinguish what factor exactly was driving the effectsize. Though some of these factors were explained in the results, it is still a factor worth consideringwhen viewing the moderators even though an overall Q test is often used to assess heterogeneity inmeta-analysis. Another issue withQ is that it simply relays the presence or absence of heterogeneity, notthe extent of it. An I2 index has been proposed as a way of calculating the actual extent of heterogeneitywhen present and is recommended when examining a small set of studies. Perhaps using the I2 testmight produce different results as it calculates the extent of true heterogeneity (due to between studydifferences), thereby separating heterogeneity due to sampling error from true heterogeneity. However,it is important to note that in a comparison of the performances of the Q test to the I2 index, it wasfound that overall the I2 confidence interval performs similarly to the Q test and the shortcomings thatare usually emphasized in using Q test are similar to the shortcomings when using I2 index when thenumber of studies are low (Huedo-Medina et al. 2006). Although the current meta-analysis by nomeansincludes a large number of different studies, 42 studies would not be considered small. Given that thereare still certain advantages to using the I2 index (Huedo-Medina et al. 2006), we recommend doing sofor future meta-analyses to compare outcomes to Q test results.

Future research and programming

Although not conducted in this meta-analysis, it may be beneficial to examine the effect of self-esteem on crime as it may vary with different types of crime. Though the effect size for crime wasmuch smaller than that for delinquency, results for this outcome tended to favor the opposinghypothesis where high self-esteem actually produces an increase in criminal and delinquent beha-vior. Prior research points out that areas of violence in particular could be driven by self-esteem ormore importantly by egotism or high self-esteem (Baumeister, Smart, and Boden 1996; Boden,

1016 C. MIER AND R. LADNY

Fergusson, and John Horwood 2007; Ostrowsky 2010). It may be that individuals with high self-esteem are more likely to engage in violent crimes due to anger resulting from a threat to one’s viewof self and threat to ego (Baumeister, Smart, and Boden 1996; Bushman and Baumeister 1998;Baumeister, Boden and Campbell 2000). Thus the use of violence in particular as a dependentoutcome could alter the nature of the relationship between “crime/delinquency” and self-esteem asmight other studies that categorize delinquency and crime into frequency scales. Future meta-analyses might consider breaking down analysis by crime type, and even addressing aggressiveand externalizing behaviors that are not criminal in nature.

In contrast, perhaps those with low self-esteem are more prone to substance abuse in attempts tohelp cope with life issues and stressors (McGee and Williams 2000; Taylor and Joe 1992), thusplacing them at a greater risk for committing drug crimes. Assessing the impact of self-esteem ondifferent types of crime is crucial to understanding if and how self-esteem development techniquesshould be implemented in treatment and prevention programs.

Prior research has noted that self-esteem might vary at different points in the life-course(Trzesniewski, Brent Donnellan, and Robins 2003), indicating that the importance of self-esteemmight vary between adolescent and adult populations. There unfortunately was not enough variationin our meta-analysis to explore comparing adult populations to adolescent ones, but the majority offindings were taken from adolescent groups. Future studies should specifically examine the influenceof self-esteem on crime at varying stages of the life-course, especially since life and crime extendsbeyond adolescence (Cullen 2011). Given the biological and psychological complexities of theteenage years, perhaps low self-esteem holds more weight in influencing various behaviors inadolescents compared to adults. As mentioned previously, delinquency might be a sign of regular,normal development among adolescents while abstainers might be more maladjusted in terms ofpersonality traits like low self-esteem but not be engaged in delinquency (Hendrix 2016). It might behelpful for future research to focus on exploring the effects of self-esteem on more serious proble-matic behavior like repetitive violence in adolescents and on the relationship between self-esteem oncrime for adults, given that crime is not a “normalized” behavior in adulthood as some level ofdelinquency is normalized for adolescents (Hendrix 2016). Studying self-esteem this way wouldallow researchers to focus on the effect of self-esteem for truly maladaptive, harmful behaviors.Though prior research notes that the majority of populations focus on at-risk or delinquent juvenilegroups, it is important to view the effect of self-esteem in both the general population and in at-riskgroups to assess its value as an insulating, protective, social support variable.

The large amount of positive, non-significant findings also raises the question of an unseenreciprocal relationship between delinquency and self-esteem. Rosenberg et al. (1989) pointed out theimportance of using reciprocal models for these variables, and many other researchers have followedsuit (see Mason 2001; Jang and Thornberry 1998; Owens 1994). Some research proposed thatdelinquency could be used as a way to increase self-esteem or as a way keep self-esteem consistentlyhigh as delinquency makes the individual feel good. It could be that the positive findings found inthe studies in this meta-analysis could be tapping into this effect. Future research should consider thereverse order of these two variables in order to determine its effect. Overall, the results of this meta-analysis support prior research and situates self-esteem as a variable that should be considered inrelation to reducing crime and delinquency. Despite our conclusion that self-esteem should continueto be studied as a variable that has the potential to influence crime, delinquency, and othermaladaptive behavior, it is important for treatment and prevention programs that target self-esteemin attempts to decrease maladaptive behaviors and crime to become more individualized based onthe nature of the behavior and co-exiting negative emotionality traits. Nuanced treatment plans thatpersonalize exactly how self-esteem should be addressed depending on the type of crime ormaladaptive behavior exhibited would require further research into exactly which variables matterthe most when looking at the effects of self-esteem on crime. Some questions to ask might be: Whatother variables in addition to threatened egotism and stability of esteem influence how self-esteemwill operate in relation to crime? Are at-risk youth more susceptible to low self-esteem or high self-

DEVIANT BEHAVIOR 1017

esteem compared to youth not at risk? How does self-esteem operate differently between adults andchildren in relation to crime? Answers to such questions may guide future program developmentresearch. Although the current research has identified ego strength, and stability level of self-esteemas influential factors, perhaps more personal factors exist. To fully inform self-esteem programing byindividualizing treatment and prevention programs for both youth and adults, further research isneeded on this subject.

Notes on contributors

CARRIE MIER, Ph.D. is an assistant professor of Criminal Justice at Indiana University East. She has her doctorateand MS degree in criminology and criminal justice from Florida State University. Mier has worked extensively withdistance learning education and undergraduate advising at Florida State University. Her main research interestsinclude drug use, abuse, and dependence; drug policy; and the impact of drugs on the criminal justice system. She isalso interested in risk, protection, and resilience factors for violence and victimization, media influences on crime anddrug use, and criminal justice education. Mier’s prior work has appeared in Crime and Delinquency.

ROSHNI T. LADNY is a Ph.D. candidate in the College of Criminology and Criminal Justice at Florida StateUniversity. Ladny has an MS degree in clinical psychology and an MS degree in criminal justice and criminology.She has worked as a mental health therapist with youth involved in the criminal justice system, victims of familyviolence, and clients with substance abuse treatment needs. Her current research interests include family violencepathways, psychosocial factors of violence, empathy, procedural justice, and methods of offender treatment. Ladny haspublished in The Journal of Criminal Justice Education and Aggressive Behavior.

References

*Studies included within the meta-analysis will be marked with an asteriskAgnew, Robert. 1992. “Foundation for a General Strain Theory of Crime and Delinquency.” Criminology 30(1):47–88.

doi:10.1111/crim.1992.30.issue-1.Bachman, Jerald G., Patrick M. O’Malley, Peter Freedman-Doan, Kali H. Trzesniewski, and M. Brent Donnellan. 2011.

“Adolescent Self-Esteem: Differences by Race/Ethnicity, Gender, and Age.” Self and Identity 10(4):445–73.doi:10.1080/15298861003794538.

Baron, Stephen W. 2007. “Street Youth, Gender, Financial Strain, and Crime: Exploring Broidy and Agnew’s Extensionto General Strain Theory.” Deviant Behavior 28(3):273–302. doi:10.1080/01639620701233217.

* Barry, Christopher T., Sarah J. Grafeman, Kristy K. Adler, and Jessica D. Pickard. 2007. “The Relations amongNarcissism, Self-Esteem, and Delinquency in a Sample of At-Risk Adolescents.” Journal of Adolescence 30(6):933–42. doi:10.1016/j.adolescence.2006.12.003.

* Barry, Christopher T., Jessica D. Pickard, and Lisa L. Ansel. 2009. “The Associations of Adolescent Invulnerabilityand Narcissism with Problem Behaviors.” Personality and Individual Differences 47(6):577–82. doi:10.1016/j.paid.2009.05.022.

Baumeister, Roy F. and Joseph M. Boden. 1998. “Aggression and the Self: High Self-Esteem, Low Self-Control, and EgoThreat.” Pp. 111–37 in Human Aggression: Theories, Research and Implications for Social Policy, edited by R. G.Geen. San Diego, CA: Academic Press.

Baumeister, Roy F., Brad J. Bushman, and W. Keith Campbell. 2000. “Self-Esteem, Narcissism, and Aggression DoesViolence Result from Low Self-Esteem or from Threatened Egotism?” Current Directions in Psychological Science 9(1):26–29. doi:10.1111/1467-8721.00053.

Baumeister, Roy F., Jennifer D. Campbell, Joachim I. Krueger, and Kathleen D. Vohs. 2003. “Does High Self-EsteemCause Better Performance, Interpersonal Success, Happiness, or Healthier Lifestyles?” Psychological Science in thePublic Interest 4(1):1–44. doi:10.1111/1529-1006.01431.

Baumeister, Roy F., Laura Smart, and Joseph M. Boden. 1996. “Relation of Threatened Egotism to Violence andAggression: The Dark Side of High Self-Esteem.” Psychological Review 103(1):5–33. doi:10.1037/0033-295X.103.1.5.

Berkowitz, Leonard. 1989. “Frustration-Aggression Hypothesis: Examination and Reformulation.” PsychologicalBulletin 106(1):59–73. doi:10.1037/0033-2909.106.1.59.

Blalock, Herbert M. 1972. Social Statistics, 2nd ed., New York, NY: McGraw-Hill.* Boden, Joseph M., David M. Fergusson, and L. John Horwood. 2007. “Self-Esteem and Violence: Testing Links

between Adolescent Self-Esteem and Later Hostility and Violent Behavior.” Social Psychiatry and PsychiatricEpidemiology 42(11):881–91. doi:10.1007/s00127-007-0251-7.

1018 C. MIER AND R. LADNY

Borenstein, Michael, Larry Hedges, Julian Higgins, and Hannah Rothstein. 2005. Comprehensive Meta- Analysis:Version 2. Englewood, NJ: Biostat.

Bouchard, Thomas J. 2004. “Genetic Influence on Human Psychological Traits: A Survey.” Current Directions inPsychological Science 13(4):148–51. doi:10.1111/j.0963-7214.2004.00295.x.

* Brendgen, Mara, Frank Vitaro, and William M. Bukowski. 1998. “Affiliation with Delinquent Friends: Contributionsof Parents, Self-Esteem, Delinquent Behavior, and Rejection by Peers.” The Journal of Early Adolescence 18(3):244–65. doi:10.1177/0272431698018003002.

Broidy, Lisa M. 2001. “A Test of General Strain Theory.” Criminology 39(1):9–36. doi:10.1111/j.1745-9125.2001.tb00915.x.

Bushman, Brad J. and Roy F. Baumeister. 1998. “Threatened Egotism, Narcissism, Self-Esteem, and Direct andDisplaced Aggression: Does Self-Love or Self-Hate Lead to Violence?” Journal of Personality and SocialPsychology 75(1):219–29. doi:10.1037/0022-3514.75.1.219.

* Caldwell, Roslyn M. 2000. “Family versus Peer Involvement: The Role of Self-Esteem, Sex and Cognitive Style asPredictors of Delinquency among High-Risk Adolescents.” Unpublished dissertation, University of California,Santa Barbara.

Campbell, Jennifer D. 1990. “Self-Esteem and Clarity of the Self-Concept.” Journal of Personality and Social Psychology59(3):538–49. doi:10.1037/0022-3514.59.3.538.

* Cheng, Christopher. 2014. “The Predictive Effects of Self-Esteem, Moral Self, and Moral Reasoning on DelinquentBehaviors of Hong Kong Young People.” International Journal of Criminology and Sociology 3:133–46. doi:10.6000/1929-4409.2014.03.12.

* Cheng, Po F. and Christopher H.K Cheng. 2015. “Effects of Moral Self, Self-Esteem, and Parental Bonding onDelinquency among Young People in Hong Kong.” International Journal of Criminology and Sociology 4:119–27.doi:10.6000/1929-4409.2015.04.12.

* Church, Wesley T., Sara Tomek, Kathleen A. Bolland, Lisa M. Hooper, Jeremiah Jaggers, and John M. Bolland. 2012.“A Longitudinal Examination of Predictors of Delinquency: An Analysis of Data from the Mobile Youth Survey.”Children and Youth Services Review 34(12):2400–08. doi:10.1016/j.childyouth.2012.09.007.

Cohen, Jacob. 1988. Statistical Power Analysis for the Behavior Sciences, 2nd ed., New York, NY: Academic Press.Cooper, Harris. 2010. Research Synthesis and Meta-Analysis: A Step-by-Step Approach, 4th ed., Thousand Oaks, CA:

Sage Publications, Inc.* Costello, Barbara J. and R. Gregory Dunaway. 2003. “Egotism and Delinquent Behavior.” Journal of Interpersonal

Violence 18(5):572–90. doi:10.1177/0886260503251128.Cullen, Francis T. 2011. “Beyond Adolescence-Limited Criminology: Choosing Our Future – the American Society of

Criminology 2010 Sutherland Address.” Criminology 49(2):287–330. doi:10.1111/j.1745-9125.2011.00224.x.Dollard, John, Neal E. Miller, Leonard W. Doob, O. H. Mowrer, and Robert R. Sears. 1939. Frustration and Aggression.

New Haven, CT: Yale University Press.* Donnellan, M. Brent, Kali H. Trzesniewski, Richard W. Robins, Terrie E. Moffitt, and Avshalom Caspi. 2005. “Low

Self-Esteem Is Related to Aggression, Antisocial Behavior, and Delinquency.” Psychological Science 16(4):328–35.doi:10.1111/j.0956-7976.2005.01535.x.

* Edwards, Willie J. 1996. “A Measurement of Delinquency Differences between A Delinquent and NondelinquentSample: What are the Implications?” Adolescence 31(124):973–89.

Eitle, David, John Taylor, and Kay K. Pih. 2010. “Extending the Life-Course Interdependence Model: Life Transitionsand the Enduring Consequences of Early Self-Derogation for Young Adult Crime.” Youth Society 41:519–45.doi:10.1177/0044118X09337500.

* Ellickson, Phyllis L. and Kimberly A. McGuigan. 2000. “Early Predictors of Adolescent Violence.” American Journalof Public Health 90(4):566–72. doi:10.2105/AJPH.90.4.566.

Elliott, Delbert S., David Huizinga, and Suzanne S. Ageton. 1985. Explaining Delinquency and Drug Use. ThousandOaks, CA: Sage Publications.

Erikson, Erik H. 1959. “Identity and the Life Cycle.” Pp. 1–171 in Psychological Issues, edited by G. S. Klein. New York:International Universities Press.

Fagan, Abigail A. and John Western. 2003. “Gender Differences in the Relationship between Offending, Self-Harm andDepression in Adolescence and Young Adulthood.” Australian & New Zealand Journal of Criminology 36(3):320–37. doi:10.1375/acri.36.3.320.

* Faulkner, Guy E., Edward M. Adlaf, Hyacinth M. Irving, Kenneth R. Allison, John J. Dwyer, and Jack Goodman.2007. “The Relationship between Vigorous Physical Activity and Juvenile Delinquency: A Mediating Role for Self-Esteem?” Journal of Behavioral Medicine 30(2):155–63. doi:10.1007/s10865-006-9091-2.

* Ferris, D. Lance, Douglas J. Brown, Huiwan Lian, and Lisa M. Keeping. 2009. “When Does Self-Esteem Relate toDeviant Behavior? The Role of Contingencies of Self-Worth.” Journal of Applied Psychology 94(5):1345–53.doi:10.1037/a0016115.

Fragkos, Konstantinos C., Michail Tsagris, and Christos C. Frangos. 2014. “Publication Bias in Meta-Analysis:Confidence Intervals for Rosenthal’s Fail-Safe Number.” International Scholarly Research Notices 1–17.doi:10.1155/2014/825383.

DEVIANT BEHAVIOR 1019

Gallupe, Owen. 2006. “Relational Strain, Gender, Anger, and Delinquency among Street Youth.” Unpublished thesis,Queen’s University, Kingston, Ontario, Canada.

Gendreau, Paul, Tracy Little, and Claire Goggin. 1996. “A Meta-Analysis of the Predictors of Adult Recidivism: WhatWorks!” Criminology 34(4):575–608. doi:10.1111/j.1745-9125.1996.tb01220.x.

Gentile, Brittany, Jean M. Twenge, and W. Keith Campbell. 2010. “Birth Cohort Differences in Self-Esteem, 1988-2008:A Cross-Temporal Meta-Analysis.” Review of General Psychology 14(3):261–68. doi:10.1037/a0019919.

Gillespie, Wayne. 2005. “Racial Differences in Violence and Self-Esteem among Prison Inmates.” American Journal ofCriminal Justice 29(2):161–85. doi:10.1007/BF02885734.

* Gonzalez, Jorge E. 2008. “Depression and Offending in a Sample of Mexican American Adolescents.” Unpublishedthesis, University of Texas at San Antonio, San Antonio, Texas.

Harris, Mildred M. 2011. “Exploring Potential Mediators of the Relationship between Adolescent Religiosity andDelinquency using the Risk and Resilience Framework.” Unpublished dissertation, The Ohio State University,Columbus, Ohio, USA.

* Heaven, Patrick C. 1996. “Personality and Self-Reported Delinquency: A Longitudinal Analysis.” Journal of ChildPsychology and Psychiatry 37(6):747–51. doi:10.1111/jcpp.1996.37.issue-6.

Hendrix, Joshua A. 2016. “Angels and Loners: An Examination of Abstainer Subtypes.” Deviant Behavior 37(12):1361–79. doi:10.1080/01639625.2016.1177391.

Hoffmann, John P. and Felicia G. Cerbone. 1999. “Stressful Life Events and Delinquency Escalation in EarlyAdolescence.” Criminology 37(2):343–74. doi:10.1111/j.1745-9125.1999.tb00489.x.

* Holsinger, Kristi and Alexander M. Holsinger. 2005. “Differential Pathways to Violence and Self-Injurious Behavior:African American and White Girls in the Juvenile Justice System.” Journal of Research in Crime and Delinquency 42(2):211–42. doi:10.1177/0022427804271938.

Hubbard, Dana J. 2006. “Should We Be Targeting Self-Esteem in Treatment for Offenders: Do Gender and RaceMatter in whether Self-Esteem Matters?” Journal of Offender Rehabilitation 44(1):39–57. doi:10.1300/J076v44n01_03.

Huedo-Medina, Tania B., Julio Sánchez-Meca, Fulgencio Marín-Martínez, and J. Juan Botella. 2006. “AssessingHeterogeneity in Meta-Analysis: Q Statistic or I2 Index?” Psychological Methods 11(2):193–231. doi:10.1037/1082-989X.11.2.193.

Hunter, John E. and Frank L. Schmidt. 2004. Methods of Meta-Analysis: Correcting Error and Bias in ResearchFindings, 2nd ed., Thousand Oaks, CA: Sage Publications, Inc.

Jang, Sung Joon and Byron R. Johnson. 2003. “Strain, Negative Emotions, and Deviant Coping among AfricanAmericans: A Test of General Strain Theory.” Journal of Quantitative Criminology 19(1):79–105. doi:10.1023/A:1022570729068.

* Jang, Sung Joon and Terrance P. Thornberry. 1998. “Self-Esteem, Delinquent Peers, and Delinquency: A Test of theSelf-Enhancement Thesis.” American Sociological Review 63:586–98. doi:10.2307/2657269.

* Jarjoura, G. Roger., Ruth A. Triplett, and Gregory P. Brinker. 2002. “Growing up Poor: Examining the Link betweenPersistent Childhood Poverty and Delinquency.” Journal of Quantitative Criminology 18(2):159–87. doi:10.1023/A:1015206715838.

* Johnson, Matthew C. and Robert G. Morris. 2008. “The Moderating Effects of Religiosity on the Relationshipbetween Stressful Life Events and Delinquent Behavior.” Journal of Criminal Justice 36(6):486–93. doi:10.1016/j.jcrimjus.2008.09.001.

* Katims, David S., Zenong Yin, and Jesse T. Zapata. 1997. “A Longitudinal Analysis of the Link between LearningDisabilities and Juvenile Delinquency in Mexican American School- Age Youth.” Juvenile & Family Court Judges48:23–34. doi:10.1111/j.1755-6988.1997.tb00762.x.

Kernis, Michael H. 1993. “The Roles of Stability and Level of Self-Esteem in Psychological Functioning.” Pp. 167–82 inSelf-Esteem: The Puzzle of Low Self-Regard, edited by R. F. Baumeister. New York, NY: Springer.

* Lau, Katherine S., Monica A. Marsee, Melissa M. Kunimatsu, and Gregory M. Fassnacht. 2011. “ExaminingAssociations between Narcissism, Behavior Problems, and Anxiety in Non-Referred Adolescents.” Child & YouthCare Forum 40(3):163–76. doi:10.1007/s10566-010-9135-1.

Leary, Mark R. 1999. “Making Sense of Self-Esteem.” Current Directions in Psychological Science 8(1):32–35.doi:10.1111/1467-8721.00008.

Leary, Mark R. and Roy F. Baumeister. 2000. “The Nature and Function of Self-Esteem: Sociometer Theory.” Advancesin Experimental Social Psychology 32:1–62.

* Lee, Kyungeun and Julie Lee. 2012. “Self-Esteem and Delinquency in South Korean Adolescents: Latent GrowthModeling.” School Psychology International 33:54–68. doi:10.1177/0143034311409856.

* Levy, Kenneth S. C. 2001. “The Relationship between Adolescent Attitudes Towards Authority, Self-Concept, andDelinquency.” Adolescence 36(142):333–46.

Lipsey, Mark W. and David B. Wilson. 2001. Practical Meta-Analysis. Thousand Oaks, CA: Sage Publications.* Luengo, M. A., J. M. Otero, M. T. Carrillo De La Pena, and L. Miron. 1994. “Dimensions of Antisocial Behaviour in

Juvenile Delinquency: A Study of Personality Variables.” Psychology, Crime and Law 1(1):27–37. doi:10.1080/10683169408411934.

1020 C. MIER AND R. LADNY

Marcia, James E. 1980. “Identity in Adolescence.” Pp. 159–87 in Handbook of Adolescent Psychology, edited by J.Adelson. New York: Wiley & Sons.

* Martinez, Lori T. 1995. “The Relationship between Stress and Delinquency among Hispanic Youth Living in a RuralEnvironment: The Role of Moderator Variables.” Unpublished dissertation, Chicago School of ProfessionalPsychology, Chicago, Illinois.

* Mason, W. Alex. 2001. “Self-Esteem and Delinquency Revisited (Again): A Test of Kaplan’s Self-Derogation Theoryof Delinquency Using Latent Growth Curve Modeling.” Journal of Youth and Adolescence 30(1):83–102.doi:10.1023/A:1005276905961.

Matsueda, Ross L. 1992. “Reflected Appraisals, Parental Labeling, and Delinquency: Specifying a SymbolicInteractionist Theory.” American Journal of Sociology 1577–611. doi:10.1086/229940.

McGee, R. O. B. and Sheila Williams. 2000. “Does Low Self-Esteem Predict Health Compromising Behaviours amongAdolescents?” Journal of Adolescence 23(5):569–82. doi:10.1006/jado.2000.0344.

* Min, Moosuk and Insoon Kim. 2000. “A Study of Delinquency of Girl Students and Corrective Measures Taken bySchools.” Women Studies Forum 16:49–67.

* Moran, Benjamin L. and David L. DuBois. 2002. “Relation of Social Support and Self-Esteem to Problem Behavior:Investigation of Differing Models.” The Journal of Early Adolescence 22(4):407–35. doi:10.1177/027243102237190.

Mullin, Brian. 1989. Advanced BASIC Meta-Analysis. New York, NY: Lawrence Erlbaum Associates.Osner, Carrie B. 2006. “The Criminal Offending–Self-Esteem Nexus Which Version of the Self-Esteem Theory Is

Supported?” The Prison Journal 86(3):344–63. doi:10.1177/0032885506291024.Ostrowsky, Michael K. 2010. “Are Violent People More Likely to Have Low Self-Esteem or High Self-Esteem?”

Aggression and Violent Behavior 15(1):69–75. doi:10.1016/j.avb.2009.08.004.* Owens, Timothy J. 1994. “Two Dimensions of Self-Esteem: Reciprocal Effects of Positive Self- Worth and Self-

Deprecation on Adolescent Problems.” American Sociological Review 59(3):391–407. doi:10.2307/2095940.Owens-Sabir, Cecilia M. 2005. “The Continuum of Control and Self-Esteem: A Study of Delinquency.” Unpublished

dissertation, Mississippi State University, Starkville, Mississippi, USA.Papadakaki, Maria, Georgia S. Tzamalouka, Sevasti Chatzifotiou, and Joannes Chliaoutakis. 2009. “Seeking for Risk

Factors of Intimate Partner Violence (IPV) in a Greek National Sample: The Role of Self-Esteem.” Journal ofInterpersonal Violence 24(5):732–50. doi:10.1177/0886260508317181.

Papps, Benjamin P. and Ronan E. O’Carroll. 1998. “Extremes of Self-Esteem and Narcissism and the Experience andExpression of Anger and Aggression.” Aggressive Behavior 24(6):421–38. doi:10.1002/(SICI)1098-2337(1998)24:6<421::AID-AB3>3.0.CO;2-3.

* Parker, Jennifer S. and Mark J. Benson. 2004. “Parent-Adolescent Relations and Adolescent Functioning: Self-Esteem, Substance Abuse, and Delinquency.” Adolescence 39(155):519–30.

Peck, Jennifer H. 2013. “Examining Race and Ethnicity in the Context of General Strain Theory, Depression, andDelinquency.” Deviant Behavior 34(9):706–26. doi:10.1080/01639625.2012.759050.

* Peiser, Nadine C. and Patrick C. L. Heaven. 1996. “Family Influences on Self-Reported Delinquency among HighSchool Students.” Journal of Adolescence 19(6):557–68. doi:10.1006/jado.1996.0054.

Peterson, Robert A. and Steven P. Brown. 2005. “On the Use of Beta Coefficients in Meta- Analysis.” Journal ofApplied Psychology 90(1):175–81. doi:10.1037/0021-9010.90.1.175.

Pigott, Terri. 2012. Advances in Meta-Analysis. Statistics for Social and Behavioral Sciences. New York, NY: Springer.Pratt, Travis C. and Francis T. Cullen. 2000. “The Empirical Status of Gottfredson and Hirschi’s General Theory of

Crime: A Meta-Analysis.” Criminology 38(3):931–64. doi:10.1111/j.1745-9125.2000.tb00911.x.Pratt, Travis C., Francis T. Cullen, Christine S. Sellers, L. Thomas Winfree Jr., Tamara D. Madensen, Leah E. Daigle,

Noelle E. Fearn, and Jacinta M. Gau. 2010. “The Empirical Status of Social Learning Theory: A Meta-Analysis.”Justice Quarterly 27(6):765–802. doi:10.1080/07418820903379610.

* Rigby, Ken and Ian Cox. 1996. “The Contribution of Bullying at School and Low Self-Esteem to Acts of Delinquencyamong Australian Teenagers.” Personality and Individual Differences 21(4):609–12. doi:10.1016/0191-8869(96)00105-5.

Rosenberg, Morris. 1985. “Self-Concept and Psychological Well-Being in Adolescence.” Pp. 206–46 in TheDevelopment of the Self, edited by R. L. Leahy. Orlando, FL: Academic Press.

Rosenberg, Morris, Carmi Schooler, and Carrie Schoenbach. 1989. “Self-Esteem and Adolescent Problems: ModelingReciprocal Effects.” American Sociological Review 54(6):1004–18. doi:10.2307/2095720.

Rosenthal, Robert. 1979. “The “File Drawer” Problem and Tolerance for Null Results.” Psychological Bulletin 86(3):638–41. doi:10.1037/0033-2909.86.3.638.

Rosenthal, Robert. 1984. Meta-Analytic Procedures for Social Research. Beverly Hills, CA: Sage Publications, Inc.* Ross, Lee E. 1995. “School Environment, Self-Esteem, and Delinquency.” Journal of Criminal Justice 23(6):555–67.

doi:10.1016/0047-2352(95)00046-1.Scheff, Thomas J., Suzanne M. Retzinger, and Michael T. Ryan. 1989. “Crime, Violence, and Self-Esteem: Review and

Proposals.” Pp. 165–99 in The Social Importance of Self- Esteem, edited by A. M. Mecca, N.J. Smelser, and J.Vasconcellos. Los Angeles, CA: University of California Press.

DEVIANT BEHAVIOR 1021

* Shah, Vaishali. 1999. “Self-Esteem and Delinquency among At-Risk Adolescents.” Unpublished thesis, CaliforniaState University, Long Beach.

* Steinke, Camila M. 2012. “The Role of Self-Esteem in the Treatment of Youth in Residential Treatment Centers: AnExamination of Delinquency and Rearrest after Discharge from Treatment.” Unpublished dissertation, University atAlbany, State University of New York, Albany, New York, USA.

Stiles, Beverly L., Xiaoru Liu, and Howard B. Kaplan. 2000. “Relative Deprivation and Deviant Adaptations: TheMediating Effects of Negative Self-Feelings.” Journal of Research in Crime and Delinquency 37(1):64–90.doi:10.1177/0022427800037001003.

Stogner, John and Chris L. Gibson. 2010. “Healthy, Wealthy, and Wise: Incorporating Health Issues as a Source ofStrain in Agnew’s General Strain Theory.” Journal of Criminal Justice 38(6):1150–59. doi:10.1016/j.jcrimjus.2010.09.003.

Taylor, Douglas N. and Del Pilar. Joe. 1992. “Self-Esteem, Anxiety, and Drug Use.” Psychological Reports 71(3):896–98.doi:10.2466/pr0.1992.71.3.896.

* Thornton, David, Anthony Beech, and William L. Marshall. 2005. “Pretreatment Self-Esteem and PosttreatmentSexual Recidivism.” International Journal of Offender Therapy and Comparative Criminology 48(5):587–99.doi:10.1177/0306624X04265286.

* Tiêt, Quyen Q. and David Huizinga. 2002. “Dimensions of the Construct of Resilience and Adaptation among Inner-City Youth.” Journal of Adolescent Research 17(3):260–76. doi:10.1177/0743558402173003.

Trzesniewski, Kali H., M. Brent Donnellan, Terrie E. Moffitt, Richard W. Robins, Richie Poulton, and AvshalomCaspi. 2006. “Low Self-Esteem during Adolescence Predicts Poor Health, Criminal Behavior, and Limited EconomicProspects during Adulthood.” Developmental Psychology 42(2):381–90. doi:10.1037/0012-1649.42.2.381.

Trzesniewski, Kali H., M. Brent Donnellan, and Richard W. Robins. 2003. “Stability of Self- Esteem across the LifeSpan.” Journal of Personality and Social Psychology 84(1):205–20. doi:10.1037/0022-3514.84.1.205.

Van De Schoot, Rens and Thessa M. L. Wong. 2012. “Do Delinquent Young Adults Have a High or a Low Level ofSelf-Concept?” Self and Identity 11:148–69. doi:10.1080/15298868.2010.517713.

* Walker, Julian S. and Gisli H. Gudjonsson. 2006. “The Maudsley Violence Questionnaire: Relationship to Personalityand Self-Reported Offending.” Personality and Individual Differences 40:795–806. doi:10.1016/j.paid.2005.09.009.

* Wilson, Marcia R.F. 1995. “An Integrated Theoretical Framework of Juvenile and Gang Delinquency Causation: AStructural Equation Model.” Unpublished dissertation, University of Southern California, Los Angeles, California.

Wolf, Fredric M. 1986. Meta-Analysis: Quantitative Methods for Research Synthesis. Newbury Park, CA: SagePublications, Inc.

Zimmerman, Marc A., Laurel A. Copeland, Jean T. Shope, and Terry E. Dielman. 1997. “A Longitudinal Study of Self-Esteem: Implications for Adolescent Development.” Journal of Youth and Adolescence 26(2):117–41. doi:10.1023/A:1024596313925.

1022 C. MIER AND R. LADNY