ABSTRACT
To address police misconduct, law enforcement agencies traditionally have used deterrence-based methods–in the form of ‘external controls’, which monitor and punish unacceptable behaviour. Some scholars, however, claim that ‘internal controls’ are more effective for addressing workplace misconduct and these controls are produced when employees perceive a greater degree of organisational justice within their agencies. Using survey data from 15,807 police officers from 101 agencies, this study tests whether (a) organisational justice impacts officers’ attitudinal support for misconduct, (b) organisational commitment is the mechanism that mediates the relationship, and (c) elements of command-and-control enhance or detract from the power of organisational justice to reduce attitudinal support for misconduct. Results suggest that organisational justice has both a direct and indirect (through organisational commitment) effect on officers’ assessments of misconduct and that elements of command-and-control can enhance the power of organisational justice to reduce attitudinal support for misconduct among police officers.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data Availability
The data for the NIJ-funded project are available through the Inter-university Consortium for Political and Social Research.
Notes
1 ‘Deviance’ and ‘misconduct’ are used interchangeably in this article.
2 We are building on the work of Tom Tyler and adopting his ‘command-and-control’ language.
3 In some formulations, ‘interactional justice’ is broken down into ‘informational’ and ‘interpersonal’ justice (e.g., Colquitt et al., Citation2006).
4 This theory is consistent with the proposition that organisational commitment mediates the relationship between organisational justice and rule adherence, but organisational commitment is not necessarily the exclusive potential mediator. Other potential mediators, consistent with this theory, include organisational identification and legitimacy.
5 The Platform was designed to be a long-term project to collect systematic data about individual police officers, supervisors and organizations over time to establish benchmarks for excellence in American policing. Study components of the original Platform included a longitudinal study of recruits, a longitudinal study of supervisors, agency-level data collection, and interviews of executives. The current study used data from the “all-agency surveys” wherein participating agencies agreed to send surveys to the population of their personnel. The current version of the Platform is sponsored by the National Police Foundation and is described at https://www.nationallawenforcementplatform.org/
6 Stratification was based on agency type (police and sheriff), geographic location and agency size. Sheriff offices were included in the frame if they provided policing services.
7 The request included IRB-approved consent language emphasizing that survey completion was voluntary, and responses would be kept confidential.
8 The initial sample size was 23,736 respondents, of which we excluded 5,139 respondents who identified as civilian employees (i.e. those without general arrest powers). This left us with a sample of 18,597 respondents of whom 634 were excluded because their job assignment was not listed; yielding a sample of 17,963. This sample was further reduced by 2,156 because these officers did not operate in capacities that while they had general arrest powers, were unlikely to be in the types of situations depicted in the scenarios that comprised the dependent variable (e.g., jailers, court security, and dispatchers). The final sample included here represents 66.6% of the initial sample and 88% of those sworn officers with job tasks we could identify. Sensitivity analyses suggest that the inclusion of more officers does not substantively change the findings reported here.
9 Only the highest and lowest values had labels: 5 was ’very serious’ and 1 was ’not at all serious’.
10 This determination is made by looking at the level of Θ (on the x-axis) when the probability of endorsing the item (i.e. saying it is not serious) is .50 or greater (shown on the y-axis).
11 The approach to the development of this measure varies slightly from the process described above because, while the IRT models are employed to create a manifest variable consistent with the procedure used for the other variables, the measurement scheme is based on the work of Rosenbaum and McCarty (Citation2017).
12 These relationships are not shown in , as they would make the diagram visually confusing.