Abstract
Borrowing from intersectionality theory, this study aims to understand how experiences of arrest – alone or in combination with victimization and criminal justice ties – inform students’ attitudes towards systems of justice, the major, and career motivations. Using a survey of 80 questions on a sample of students from urban colleges with large minority representation, the authors rely on ordered logit regressions to find: students with experiences of arrest hold more negative attitudes towards systems of justice; find the CJ major a more relevant subject matter; and are more motivated to make a community difference than students with no experiences of arrest. Differences are enhanced for students with experiences of arrest and victimization but reduced for students who know someone working in CJ fields. Experiences of arrest remain largely unvaried by demographic indicators. Results suggest student experiences of trauma are important dimensions of students’ decisions about the major and profession.
Keywords:
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Sampling students from day and night classes as well as weekday and weekend sections was not possible. At times, the instructors were giving an exam. In other instances, part-time instructors could not be reached. For these reasons, we could not rely on random sampling but use convenience sampling, a commonly used strategy in the research literature (see Gabbidon et al., Citation2003; Krimmel & Tartaro, Citation1999).
2 After being read a consent script in class, students were given the choice to fill out a paper survey which contained no identifying information to ensure anonymity. Almost all students who received the survey completed them, with fewer than five returning blank surveys.
3 We created multiple categories of the corollary independent variables. When assessing involvement, victimization, and ties to the CJ system, students may fall in one of the following categories: 1 – no ties, no victimization, no involvement; 2 – no ties, no victimization, involvement; 3 – no ties, victimization, no involvement; 4 – no ties, victimization, involvement; 5 – ties, no victimization, no involvement; 6 – ties, no victimization, involvement; 7 – ties, victimization, no involvement; 8 – ties, victimization, involvement. Comparing such a large number of categories in regression models adds complexity and, in a sense, renders the comparisons meaningless. For parsimony, we decided to compare only four categories (first, fourth, fifth, and eighth) when exploring the effect of multiple experiences in some variants of regression models.
4 “The Mann-Whitney test is of considerable interest especially in situations in which outcomes are measured on scales that either are ordinal or have arbitrary measurement units” (Conroy, Citation2012: 189).
5 Mann-Whitney U is similar to Wilcoxon W. Both indicators are associated to the same z-score and p-value in tests of significance. For parsimony, we report only one, the Mann-Whitney U statistic (see, Sawilowsky, Citation2007).
6 We use bivariate analyses in a limited scope; only to show which reasons towards the major or motivations towards the career are significant. Then, we run and report ordered logistic models for the indicators that reach bivariate statistical significance.
7 The ordered logit models assume the effect of variables of interest is the same across categories of the dependent variable meaning, “the odds ratios will stay the same regardless of which of the collapsed logistic regressions is estimated” (Williams, Citation2016: 9). In our tables, we report the parallel line test coefficient which, if rejected, indicates the effect of the independent variables remains the same across outcome categories.
8 The regression models are based only on significant bivariate statistics (see Appendices A, B, & C). We run regression models on all the reasons behind a CJ major and all the motivations behind a CJ career but report and analyze only the significant indicators. The tests of parallel lines are rejected in our reports. Therefore, proportional odds models are adopted. Regression models for all indicators are available upon request.