Reliability and Validity of a Self-Control Measure

Rejoinder

Published in: Criminology, v. 36, no. 1, Feb. 1998, p. 175-182

Posted on RAND.org on December 31, 1997

by Douglas L. Longshore, Susan Turner, Judith Stein

Read More

Access further information on this document at www.blackwellpublishing.com

This article was published outside of RAND. The full text of the article can be found at the link above.

We used a revised set of items from Grasmick et al. (1993) for our study of self-control and crime in a sample of male and female criminal offenders. Viewed as a unidimensional construct in the general theory of crime, self-control, as we measured it, appeared instead to be multidimensional. Using factor analytic techniques and fit indices in EQS, we found five substantive subfactors, interrelated but not reducible to a single dimension. They were impulsivity/self-centeredness, preference for simple tasks, risk seeking, preference for physical activity, and volatile temper. Moreover, while a five-factor solution fit the data for male offenders, neither a one-factor nor a five-factor solution fit the data adequately for female offenders. Finally, when we regressed two crime variables on the overall self-control measure and (separately) on the five subfactors, beta coefficients for three subfactors-risk seeking, impulsivity/self-centeredness, and volatile temper-were at least as strong as coefficients for the overall measure. Thus, in our analysis of this data set, the self-control measure seemed tenable with male but not female offenders, and the predictive power of self-control was no better than that of existing constructs in criminological theory. Piquero and Rosay (in this volume) took a different approach in analyzing our data set. They avoided use of correlated error residuals by adding items within subscales of the self-control scale and relied on maximum .likelihood methods and goodness-of-fit (GFI) indices in LISREL. They obtained a one-factor solution for men and women and concluded that the data do in fact support a unidimensional view of selfcontrol. Moreover, correlations between their self-control measure and crime were significant and in the expected direction. While we recognize the scientific concerns raised by Piquero and Rosay, we believe their analytic method was inappropriate, and in any event, led to substantive conclusions identical to ours.

This report is part of the RAND Corporation external publication series. Many RAND studies are published in peer-reviewed scholarly journals, as chapters in commercial books, or as documents published by other organizations.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.