Self-Control Theory

III. Empirical Status of the General Theory of Crime

Gottfredson and Hirschi’s (1990) theory makes many testable claims, some of which have weathered the storm of empirical criticism, others of which remain wanting. In general, however, the general theory of crime has remained robustly supported across multiple types of samples, methodologies, and variations in measurement, in terms of its central claim that low self-control predicts criminal and analogous behaviors. In Gottfredson’s (2006) most recent review of the literature, he concluded that low self-control has remained predictive across gender, location, age, race, offense type, offenders, analogous behavior, and time. Lilly et al. (2007) further noted that the theory has received extensive support across individual studies, a comprehensive meta-analysis, and a full narrative literature review. Akers and Sellers’s (2004) review also found substantial support for self-control theory across cultures, explaining anywhere from 3.0% to 19.0% of the variation in criminal behavior.

The most quantitatively sound review of the general theory of crime was conducted by Pratt and Cullen (2000), who used meta-analytic techniques to ascertain the empirical status of self-control theory. The authors’ data came from 21 peer-reviewed published articles, for a total of 126 effect size estimates, across 17 individual data sets, and a total of 49,727 individual cases. To control for measurement effects, the researchers coded for whether studies had used behavioral versus attitudinal measures of self-control; when attitudinal measurements were used, the studies were further coded to control for whether the researchers used Grasmick, Tittle, Bursik, and Arneklev’s (1993) extremely popular self-control scale. Finally, Pratt and Cullen controlled for whether studies had used longitudinal versus cross-sectional research designs. In terms of the predictive value of low self-control as it relates to crime, they found that the effect size of low self-control regularly exceeded 0.20, net of variables specified by other theories and methodological considerations. Furthermore, the meta-analysis results supported the theory’s contention that crime and analogous behavior can be predicted from a single source (low self-control) across race, sex, age, and community. It is important to note that Pratt and Cullen found that similar effect sizes were found across behavioral and attitudinal scales of self-control and regardless of whether attitudinal scales invoked Grasmick et al.’s indicators of low self-control. These findings go a long way toward responding to the criticism that self-control studies that use behavioral measurements are inherently tautological (i.e., that the use of behavioral indicators of low self-control—in short, deviant yet legal behaviors—to predict other deviant yet illegal behaviors merely amounts to bad behavior being related to bad behavior, a finding of little interest to criminologists).

Beyond the low self-control–criminal behavior link, Pratt and Cullen’s (2000) meta-analysis failed to support at least two of Gottfredson and Hirschi’s (1990) contentions, namely, (1) that the self-control–crime relationship would hold across both cross-sectional and longitudinal studies and (2) that low self-control represents a general explanation of behavior. Gottfredson and Hirschi argued that because self-control is invariant across the life course, cross-sectional research designs (which represent the bulk of social science research) were adequate for testing self-control theory. Pratt and Cullen, however, found that the effect sizes for self-control were significantly weaker in longitudinal studies compared to those from cross-sectional studies. Further, Gottfredson and Hirschi explicitly stated the irrelevance of other criminological theories because of their assumption that self-control was a precursor to all other criminogenic factors. Even so, Pratt and Cullen found that variables specified by social learning theory—in particular, deviant peer influences and antisocial attitudes—not only had an impact on crime independent of self-control but also increased the explanatory power of each study’s overall statistical models. This strongly suggests that Gottfredson and Hirschi’s general theory of crime may not be so general.

Indeed, much of the research reviewed by Akers and Sellers (2004) and by Lilly et al. (2007) found Gottfredson and Hirschi’s (1990) claims beyond the low self-control– criminal behavior connection wanting. For example, Akers and Sellers found mixed results for the hypothesis concerning the relative stability of self-control over the life course, with some research suggesting its relative stability, others indicating its variability, and yet others finding it to be at once both stable and variable. They found similar results among the research exploring the extent to which low self-control is a unidimensional factor. Furthermore, they found research suggesting that, in regard to low self-control’s causes, those proposed by Gottfredson and Hirschi were predictive, but not independent other factors, such as fair discipline and parental acceptance. Similarly, Lilly et al. found support for sources of low self-control outside of the family, including neighborhood-level factors as well neurobiological factors—findings that are reviewed at length later in this research paper.

Ultimately, the central proposition of the general theory of crime—that low self-control predicts criminal, delinquent, antisocial, and analogous behaviors—holds across several studies, methodologies, samples, and measurements. Research has not, however, supported Gottfredson and Hirschi’s (1990) claim to having authored a “general” theory. Instead, it appears that their theory of criminal behavior, based on levels of low self-control, specifies a vital predictor of criminal behavior that is necessary for criminological models so as to avoid misspecification but is nevertheless far from the sole predictor of criminal and deviant behavior.