Does Self‐control Affect Survey Response? Applying Exploratory, Confirmatory, and Item Response Theory Analysis to Grasmick Et Al.'s Self‐control Scale*

Abstract

A General Theory of Crime has sparked a great deal of theoretical debate and empirical investigation. Tests of the theory have focused on measuring the core element, the latent trait of self-control. The majority of this research has used the 24-item scale developed by Grasmick et al. (1993), and a great deal of attention has been directed at the validity of this scale. Empirical debate revolves around the unidimensionality of the scale as established using conventional factor analytic techniques [exploratory factor analysis (EFA) and confirmatory factor analysis (CFA)]. In this paper, we provide the first application of an item response theory (IRT) Rasch model to the validation of the Grasmick et al. scale. IRT models focus on the interaction between the human subject and survey items, and the extent to which cumulative scales fail to provide fundamental measurement. Our results suggest that although conventional factor analyses yield results similar to those previously reported, IR T analysis reveals that one's level of self-control influences self-report responses, a finding consistent with Hirschi and Gottfredson. One of the most influential theoretical developments of the past decade is Gottfredson and Hirschi's (1990) A General Theory of Crime.Their theory has attracted much attention, with a growing body of literature spanning both critiques (Akers, 1991; Barlow, 1991; Polk, 1991; Reed and *We would like to thank the three reviewers for their valuable suggestions. A portion of this paper was written when the first two authors were Fellows participating in the "Violence Over the Life Course" Institute at the Center for Advanced Study in the Behavioral Sciences at Stanford University in the Summer of 1998. CRIMINOLOGY VOLUME 38 NUMBER 3 2000 897 ~~ 1. Since the focus of this paper concerns measurement issues associated with selfcontrol, we forego a detailed review of Gottfredson and Hirschi's (19%) theory.

Self-control behind bars: A validation study of the Grasmick et al. scale

“…However, like previous researchers (e.g., Piquero et al, 2000), we found that the six-factor model was the most accurate model. Following modification indices, we used the following procedures to refine the six-factor model.…”

“…Exploratory factor analysis determines the number of factors that account for the covariation between the 24 items in the scale. Prior research Grasmick et al, 1993;Piquero et al, 2000;Vazsonyi et al, 2001) suggested that six factors will emerge with an eigenvalue greater than 1. These factors represent the six theorized dimensions of self-control.…”

“…Although previous researchers (e.g., Vazsonyi et al, 2001) have used RMSEA < .1 as a good fit, others (e.g., Steiger, 1998Steiger, , 2000 have suggested that RMSEA values of .05 or less are obviously more acceptable. The six-factor RMSEA = .06 suggests a poor fit (but see Hu & Bentler, 1999;Piquero et al, 2000). Three final goodness-of-fit statistics were examined.…”

Measuring Collective Efficacy

“…This is not an issue for our models, which estimate single population models, and, accordingly, leave intercepts unconstrained. For applications of IRT to self-reported delinquency, see Osgood et al (2002) and Piquero et al (2000). 2 Much earlier, Maccoby, Johnson, and Church (1958:39) hypothesized that in disorganized neighborhoods, "individual adults will feel less responsible for guiding other people's children,"…”

“…Recent attention to measurement in the social sciences covers a range of substantive and theoretical topics, including behavioral expectations (e.g., Manski 2004), social capital (e.g., Paxton 2002), democracy (Coppedge and Gerring, et al 2011), employment careers (Manzoni et al (2010), time use data from diaries (Kan and Pudney 2008), and family activities (Broege et al 2007). In criminology, the measurement of delinquent behavior (e.g., Osgood et al 2002;Piquero et al 2002) and self-control (e.g., Piquero et al 2000) has been examined using models for survey response. While measurement has been discussed in detail for individual-level psychometric models of test scores and survey responses, it is no less an issue for models of macro-ecological units, such as neighborhoods, communities, and schools.…”