Gender and Self‐Control Theory: Are There Differences in the Measures and the Theory's Causal Model?

Abstract

This study examined the distribution differences across genders in key measures from self‐control theory and differences in its central model. The study used responses to multiple measures of self‐control, parental management, and a single measure of deviance from 425 college students. The findings show that all the measures are different across genders. In addition, using structural equation modeling and multiple group structural equation modeling, the findings show the central model for males and females were similar. The findings suggest that self‐control theory is general in the sense that its central model is usable for male and female deviance.

Keywords:

• Gender
• Self‐control theory
• Structural equation modeling

Notes

George E. Higgins is Assistant Professor in the department of Justice Administration at the University of Louisville. He received his PhD from Indiana University of Pennsylvania in 2001. His current research focuses on criminological theory testing and quantitative methods. Correspondence to: George E. Higgins, 208 Brigman Hall, Department of Justice Administration, University of Louisville, Louisville, KY 40292, USA; Tel: (502) 852‐0331; Fax: (502) 852‐0065; E‐mail: [email protected]

This view of Gottfredson and Hirschi's (1990) theory does not discount that an increase in supervision will in all likelihood reduce opportunities for females to commit deviance. However, this view of the theory remains consistent with the overall causal model of Gottfredson and Hirschi's theory in the literature (see Tittle, Ward, and Grasmick, 2003, pp. 429–430; Gibbs et al., 1998).

Information regarding the principal components factor analysis and scree tests for all of the measures are available from the author on request.

Since the deviance items are general in nature, each participant has had an opportunity to commit them at some point in their lifetime. The general nature of the items represents one way of holding opportunity constant. Gottfredson and Hirschi (1990) argue:

• Opportunity can be held constant through the assignment to conditions of varying opportunity, by natural variation, or restricting the attention to people sharing identical crime‐relevant characteristics. Differences in criminal activity can be ascribed to differences in tendency, since there are no differences in opportunity (p.220).

This measure of deviance carries out this task, thus reducing the need to assess opportunity. In addition, empirical information reduces the need to assess opportunity. Further, Pratt and Cullen (2000) show opportunity behaves erratically in the theory.

Using a participation measure rather than a frequency measure has some benefits; first, the participation measure reduces the fault of the estimates for the measure. Second, the participation measure helps avoid the criticism that the frequency of the responses is artificially inflating the overall score on the deviance measure. Unfortunately, the measure does reduce the variation in the deviance measure. However, the decrease in variation should not justify discounting the measure or the results of this study.

Popular studies of low self‐control assume that attitude measures are the best method of capturing self‐control information (see Pratt & Cullen, 2000; Unnever et al., 2003). However, Piquero, MacIntosh, and Hickman (2000) argued for caution when using attitude measures of low self‐control because low self‐control may affect the individual's survey responses. Therefore, the scenario‐based technique becomes attractive in assessing low self‐control because it creates a unique opportunity to capture phenomenological self‐control information. This measure has a possible limit that is noteworthy. The responses given to this scale may not reflect the individual's real behavior or real attitude.

These two measures (PM1 and PM2) may produce three problems. First, the students may present their parents in a positive light (e.g., social desirability). Second, the validity of the response may be a question, especially for the third grade measure – given the age of the sample. Third, precise information may be cloudy because parents tell stories about themselves and their children that become part of the family's stories. The current study does not use estimates for precision about parental management tasks; instead, the items capture the broad application of the tasks is important. In addition, McCrae and Costa (1988) suggest recollections of childhood do contain some pieces of truth and that retrospective studies of childhood are useful. Other criminologists (Gibbs et al., 1998; Sims, 2000; Higgins, 2002; Gibbs et al., 2003) agree with this assertion.

Kenny and Rice (1995) argue parenting practices change as the child ages. This implies, as the child ages, supervision and discipline may not be at the forefront of a parent's parenting agenda. Instead, the parental role may change to more age specific equivalents such as the following: adviser, assistant, discussant, and problem solver. In addition, other issues may arose needing parents to provide more respect, autonomy, understanding, and treatment as an adult. Parents of older children may be successful if they can effectively and consistently perform these tasks. In sum, these tasks represent a shift from supervision and discipline of younger children to more attachment with older children.

For the second model, Byrne (1998) provides guidelines for the constraints of MSEM that include constraining the factor loadings and structural paths to find out the robust qualities of the causal model of self‐control theory across genders.

I conducted more analysis on the third grade and ninth grade parental management measures. This consists of splitting the measures into supervision and discipline scales. The findings coalesce with LaGrange and Silverman's (1999) findings on supervision differences. If readers want the psychometric properties of these scales (coefficient alphas and factor analyses) they are available on request from the author.

For a complete discussion of these goodness‐of‐fit indexes see Gibbs et al. (2003).

When interpreting the chi‐square ratio in MSEM the interpretation is comparative. That is, the chi‐square ratio compares the fit of the matrices under study. So the interpretation is different when comparing the baseline and the constrained models.

The vertical paths contain the standardized coefficients, in parentheses, and the unstandardized loadings are above the paths. For the horizontal paths between latent measures, the values in parentheses are product‐moment correlation coefficients and values above them are covariances. The values near the top left of the each variable show their variance.

The horizontal paths contain the standardized coefficients where the values above the paths are for males and values below the paths are for females. The first model is the unconstrained model and the second model is the constrained model.

Additional information

Notes on contributors

George E. Higgins

George E. Higgins is Assistant Professor in the department of Justice Administration at the University of Louisville. He received his PhD from Indiana University of Pennsylvania in 2001. His current research focuses on criminological theory testing and quantitative methods. Correspondence to: George E. Higgins, 208 Brigman Hall, Department of Justice Administration, University of Louisville, Louisville, KY 40292, USA; Tel: (502) 852‐0331; Fax: (502) 852‐0065; E‐mail: [email protected]