![Sample our Law journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5942/TOC-Subject-Sample-2021-Law.jpg"})
Abstract
Given the reliance on meta-analyses to produce criminal justice policy recommendations, it is important to think critically about how this method is being applied in practice. In this study, we use data from a meta-analysis of corporate crime deterrence to demonstrate that applying meta-analytic methods to conceptually ambiguous research domains is problematic. Although meta-analysis is capable of modeling methodological variations in different research projects examining the same construct, analysts should not assume that meta-analytic methods are always appropriate; methodological differences may reflect underlying conceptual dissimilarities – this violates an assumption of meta-analysis. We also offer a critique of the corporate crime field for failing to clearly define its outcome, a critique that can be extended to other areas of criminological study.
Acknowledgements
Many researchers and students, in various capacities, have assisted with this review. We gratefully acknowledge the contributions of Patricia Joseph, Rachael Powers, Kerry Richmond, Megan Bears Augustyn, Katheryn Decelles, Cliff Akiyama, Alex Bob, and Nathan Bernstein. We also appreciate the support of the University of Maryland and the University of Pennsylvania. Finally, we benefitted from the input of anonymous reviewers, whose suggestions greatly improved the article.
Notes
2. Of course, variations in operational definitions do not necessarily indicate conceptual variation – it is very possible to measure the same concept a variety of ways and conceptual variation is likely not an issue in many meta-analyses. Our argument is that meta-analysts should not automatically assume that variations in effect sizes are solely resulting from methodological differences.
3. A reviewer noted that treatment effects from ‘completely different classes of dependent variables’ can be handled together in meta-analysis (see Heinsman & Shadish, Citation1996) and that these solutions could be used to address the comparison of varying operational definitions, units of analysis, and sanction types as discussed in the present paper. Heinsman and Shadish purposefully combined studies with completely different research questions/subject areas (e.g. the impact of pre-operation education on post-operation outcomes, the impact of SAT coaching on scores) to determine whether randomized experiments produced different effect sizes than nonrandom experiments. Importantly, in addition to looking at overall effect sizes, they examined the impact of randomization specifically for conceptually different categories. The current paper argues that this is what needs to be done, not only when outcomes are obviously measuring different things, but also when the dependent variable is labeled by the same name (but might be measuring different things).
4. It is important to note that – due to the data available – we were only able to examine the impact of defining corporate crimes in terms of non-profit vs. for-profit organizations, corporate vs. individual-level processes, and regulatory sanctions vs. criminal law enforcement. We were unable to examine whether management level of offenders (e.g. upper managers only) impacted methodological decisions or results.
5. Due to the lack of readily available and accessible data on corporate offending there is a lack of systematic program or policy evaluations. Thus, while our initial search was exhaustive, most of the literature in this area is not quantitative. Of the original 241,014 hits, only 4664 of those sources were potentially eligible, meaning they were both quantitative and relevant to corporate crime.
6. We also conducted a forward search in Web of Knowledge and sent a preliminary list of studies to experts to look for additional studies.
7. Under each of the five treatment categories discussed above, we had individual samples, corporate samples, geographic areas and other units of analysis from both cross-sectional and longitudinal studies.
8. We were unable to run moderating analyses on groups that contained little variation on the constructs of interest – which was generally due to a small number of studies in the overall category. It’s worth noting that a preference for certain methods or variables in the literature may result in a form of sampling bias.
9. ‘Self-report’ data includes reported intentions to offend as well as actual offending behaviors.
10. The use of such broad outcomes may be less problematic in traditional crime research. In these cases there is general consensus regarding definitions and behaviors are clearly proscribed by law. However, scholars should verify that included studies meet the assumption of conceptual homogeneity.
11. Interestingly, though, strategies that incorporate multiple types of interventions were found to consistently deter corporate crime (broadly defined; see Schell-Busey et al., Citation2016. It may be that, when an outcome involves such heterogeneity (e.g. in motivations, offender characteristics, etc.), the use of multiple strategies at once can be effective in addressing a wide variety of different behaviors. Such a strategy is promulgated in the ‘Pulling Levers’ approach (Kennedy, Citation1996) or ‘Responsive Regulation’ (Ayres & Braithwaite, Citation1992).
12. A reviewer noted that ambiguity in the dependent variable is not the only issue requiring attention – conceptual ambiguity in independent variables is likely to cause similar problems for meta-analyses.
13. It should be noted that meta-analyses may be more practicable if original manuscripts contained the data needed for meta-analysis. With more data available, meta-analysts could include more studies and thus may be better able to analyze more distinct and cohesive categories of research.
14. This is actually demonstrated in the current project; in Schell-Busey et al., Citation2016, we were able to draw more definitive conclusions on the impact of ‘multiple treatments’ on corporate crime (when corporate crime is measured at the individual level and at the corporate level). This is likely because the studies included in this category were defined similarly and there were enough of them to make analysis possible.
15. A reviewer argues that it is unrealistic for the field to settle on a common conceptual definition of white-collar crime. Although we recommend a specific definition here, our main goal is not necessarily to decree one absolute conceptualization and ignore others. Our goal is to demonstrate the benefits of breaking broad outcomes into narrower categories to develop a more convincing (and hopefully consistent) set of knowledge that we can employ more effectively for theoretical and practical purposes.
16. Although small businesses may differ in their reasons for compliance/noncompliance and the likelihood of offending (e.g. see Herbert, Green, & Larragoite, Citation1998; Lepoutre & Heene, Citation2006), we feel that all profit-seeking corporations warrant study and the impact of company size is an important variable to consider.