ABSTRACT
Policing by consent has long been viewed as a fundamental feature of modern policing. Police need citizens to report crime and suspicious activity and to assist police with their enquiries. The procedural justice model is commonly employed to explain cooperation with police, yet few studies consider how social context informs cooperation. In this study we examine the role of contextual factors in developing a better understanding of the procedural justice model of cooperation with police. To do so we compare results in two contexts: St Louis County (US) and Brisbane (Australia). We find similarities and differences in the way contextual factors (including feelings of insecurity and social cohesion and trust) impact the willingness to assist police across our two research sites.
Acknowledgement
The authors would also like to thank Professor Lorraine Mazerolle for her assistance in facilitating this collaboration.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 The area units in Brisbane are larger than in St Louis County, which may mean that the residents within hot spots have more homogenous views than residents in suburbs. However, the larger sample sizes in the Brisbane sample may help with this. Furthermore, we account for the nesting of residents in hot spots/suburbs by modeling the random effects (although see footnote 3).
2 Like others who have surveyed high crime areas, the St Louis County response rate is not enviable (e.g. Ferguson and Mindel (Citation2007) had a 33% response rate, Chermak et al. (Citation2001) had a 31% response rate and a 49% cooperation rate, while Hinkle et al. (Citation2013) had a 46.1% cooperation rate). See Pashea and Kochel (Citation2016) for an explication of the difficulties of conducting surveys in high crime areas. Nonetheless, it is an improvement over at least one foundational study about the antecedents to cooperation with police (Sunshine and Tyler Citation2003 reported 22% at baseline).
3 Liklihood-ratio tests for each site indicate that whilst the data are clustered, multilevel models were not a significant improvement over the OLS models.
4 Cohen et al. (Citation1999) provide a detailed discussion of POMP as a meaningful unit of measurement for the social sciences. For each variable, we averaged across non-missing indicators for each case and applied the following POMP score formula:
For POMP scores the range will be 0–100. However, it should be noted that the original scale length affects the POMP scores in that smaller scales will have slightly larger gaps between scores such that a 3 on a 4-point scale becomes .75 whereas a 3 on a 5-point scale becomes .60. Each score on a 4-point scale covers .25 points whereas each score on a 5-point scale covers .20 points. However, in the Brisbane sample, which has the 5-point scale, ‘3’ typically reflects a neutral measure and so scoring an 80 on a POMP score in Brisbane is like scoring 75 in St Louis County. Nevertheless, POMP allows for a like comparison even when different scales are used.
5 We ran the St Louis County model with only demographic information (excluding assessments about police procedural justice and legitimacy). What we find is that only 12% of the variance in cooperation is explained by the demographics only model (versus 34% when procedural justice and legitimacy judgments are included). More interestingly though, the relationship between minority status and cooperation is statistically significant (b = −.098, p = .001) and minority status is the second strongest demographic predictor of cooperation after age. We find in St Louis County that African Americans are less willing to cooperate with the police, but Model 1 shows that this effect is mediated by procedural justice. The implication is that a perceived lack of procedural justice delivered to African Americans explains less cooperative attitudes toward police for this group.
6 Furthermore, we note that our measure of legitimacy, on face value, has some differences across the two contexts. To explore this point of measurement further we ran parallel post hoc analyses of the final model using a single-item measure of legitimacy that is comparable across both contexts. The item reflects the moral obligation to obey. We found that substituting this single-item measure produces the same substantive findings as the full latent measure in each sample. The exception is that this less comprehensive (inferior) measure of legitimacy is not statistically significant in the St. Louis County analysis. As the key findings of interest are consistent with the initial analyses, we retained the more reliable multi-item measure of legitimacy in our final models presented in this paper.