Open research questions in Policing Practices and Perceptions

What the literature leaves open

• However, if our analysis of the CPD’s front-stage reputation management efforts spurs further work on the back-stage production of those claims, or even their public reception, then maybe that could count as progress, too.

  Making Progress in the Chicago Police Department, 1862–2024 · 2026 · DOI

• While this survey experiment did not find evidence that media narratives shift attitudes or behaviors, that does not necessarily mean that it is impossible in every context. This could be due to other factors such as (1) design elements specific to this survey experiment, (2) the use of a survey experiment in general, or (3) the timing of fielding. First, the news story that respondents read in the experiment is ambiguous, leaving room for diverse perspectives in terms of who is at fault. While this was a conscious decision, as we wanted to hold that ambiguity constant and then shift the additional information people read in a way that we expected would fill in the rest of the story, it does amount to a lighter treatment. If the treatment was more clear, such as if it indicated that the officer was unnecessarily escalating the situation or even that they were later charged for their actions during the traffic stop, it may have more effectively shaped respondents’ attitudes toward the police and planned political engagement. Similarly, it could be that the particular dependent variables that we chose were not affected by the media narratives, but others would be. While we chose dependent variables that are well established in the literature (Ang and Tebes 2024; Burch 2022; Christiani and Shoub 2022; Morris and Shoub 2024; Walker 2019; 2020), there is always the possibility that other, unasked dependent variables would have shown some movement. In this case, it may have been that even more specific outcome measures, like belief in the existence of racial bias in policing or policy preferences on training, body cams, and the use of force, could have shown some movement. Future work may consider expanding this analysis to additional dependent variables. It could also simply be that a survey experiment is not the appropriate design to assess whether media narratives can shift attitudes or behaviors. The kind of media attention that came to police violence in 2014 and 2020 was much more encompassing than a brief interaction on a survey experiment. Further, it was not a media narrative in isolation. Instead, it was coupled with a mass movement— something that is bigger, more meaningful, and more impactful than a brief snippet of a news story. Thus, media coverage and associated narratives in the context of broader social movements may be able to make changes that are not easily detectable in a survey experimental context. Recent work has demonstrated that the prevalence of the Black Lives Matter movement, in particular, makes it difficult to detect smaller or more localized shifts like what would occur in a survey experiment. Sances (2025) in attitudes, demonstrates, by applying a synthetic control design, that the BLM movement did polarize attitudes toward the police by partisanship. He finds polarizing attitudinal shifts when he compares across policy opinions because, he argues, the ubiquity of the BLM movement means that essentially every American has already received the “treatment” of learning about BLM, police violence, and associated narratives. Finally, it could be elements specific to the current time period that are blocking any effects. It may be that before attitudes become too crystallized, some shift in opinion or political engagement would result from narratives contextualizing police https://doi.org/10.1017/rep.2025.10049 Published online by Cambridge University Press The Journal of Race, Ethnicity, and Politics 23 violence. Now, in an era of ossification of attitudes, though, such effects may be impossible.

  Examining How Narratives of Events Change Attitudes or Motivate Political Participation: The Case of Police Violence · 2026 · DOI

• Several limitations constrain the interpretation of these findings. First, the use of real headlines increases ecological realism but introduces content beyond the intended thematic frame. For example, several National Politics headlines reference national political figures such as President Trump, President Biden, and Vice-President Harris, 1 3S. M. Mourtgos individuals who may evoke reactions unrelated to policing. Although random assign- ment ensures that headline-specific content is balanced across conditions in expecta- tion, the design cannot isolate which specific linguistic elements within a theme drive the observed effects. Future research using constructed stimuli could more precisely identify which content features are responsible for the observed differences. Second, the between-subjects design measures differences between groups rather than changes within individuals. I cannot determine whether exposure caused respon- dents to revise their views; I can only observe that groups exposed to different headline themes reported different average levels of approval. The design also does not assess whether the observed differences persist beyond the immediate post-exposure mea- surement or how effects might accumulate through repeated exposure. Panel designs with repeated measurements would be required to evaluate persistence and dosage. Third, the stimuli consist exclusively of text-only newspaper headlines, which represent only one component of a much richer media environment. Contemporary news consumption often involves video, images, social media commentary, and algorithmically curated feeds (Brady et al., 2021; Prior, 2007). The present design removes these elements to isolate thematic content, but the findings may not general- ize to multimedia formats or social media platforms. Whether similar effects would be larger, smaller, or differently structured under more complex media conditions remains an open question. Fourth, the outcome measure captures general approval of police use-of-force rather than evaluations of specific incidents. As discussed earlier, prior research indi- cates that general attitudes tend to be less responsive to short-term informational cues than incident-specific judgments (Boivin et al., 2017). The relatively small effect sizes observed here are consistent with this distinction. Studies employing incident- specific outcomes might therefore observe larger shifts in opinion. Fifth, although Bayesian model averaging reduces the risk of relying on a single misspecified ordinal model, it cannot eliminate misspecification entirely. The ensem- ble weights favor the stopping-ratio specification for most hypotheses, suggesting that a sequential decision process may approximate how respondents arrive at their approval scores. If respondents instead evaluate all three scenarios as part of a single holistic judgment, some residual bias could remain in the ensemble estimates. Sixth, the survey did not include measures of prior police contact, victimization, fear of crime, or neighborhood conditions, all of which prior research has linked to attitudes toward police. Because treatment assignment was randomized, omitting these variables does not confound the treatment–control contrasts. However, their absence limits the ability to assess whether headline effects vary across respondents with different prior experiences with police or crime, or to explain baseline differ- ences in approval of police use-of-force. Future research should incorporate these measures to examine treatment-effect heterogeneity across experiential and neigh- borhood contexts.

  News headline themes and approval of police use-of-force: evidence from a national survey experiment · 2026 · DOI

• First, the video data did not allow us to determine which party initiated the use of force, as the encounters often escalated too quickly to distinguish whether the civilian or the officer acted first. In some cases, the footage suggested that civilians initiated the confrontation, but this could not be established with certainty across all incidents. What this note highlights—as did Dai et al. (2011)—is that future research should move beyond asking whether officers “apply” procedural justice and instead examine whether encounters meaningfully afford it. A promising contribution would be to conceptualise “procedural justice opportunity” as a property of the situation rather than of the officer. In this approach, opportunity would be defined in terms of situ- ational affordances for the four elements of procedural justice—voice, neutrality and explanation, respect, and trustworthy motives—rather than as officer motivation, dis- position, or skill. Empirically, this concept could be operationalised through observable markers that shape what is realistically possible in an encounter, such as available time, number of parties present, weapon cues, digital devices, immediate threats, active resistance, medical urgency, and crowd behaviour. Pre-registering how these markers affect the feasible procedural justice bandwidth would allow researchers to distinguish situa- tions in which procedural justice could reasonably be expected from those in which it could not. Crucially, this framework separates opportunity from delivery by mod- elling officers’ procedural justice behaviour as a function of situational opportunity, thereby avoiding the misclassification of high-risk, high-urgency encounters as fail- ures of procedural justice. This approach aligns with emerging body-worn video cod- ing strategies and offers a stronger basis for evaluating both officer behaviour and the conditions under which procedural justice can realistically be enacted (Nawaz & Tankebe, 2018). Second, neither officer group was required to provide oral notification of the camera. However, we were unable to verify from the footage whether officers in fact refrained from giving such notification as intended, because the audio acti- vates approximately 30 s after the interaction begins. Any initial notification—if it occurred—would therefore not have been recorded. While random assignment sug- gests that any such notifications, if they occurred, were likely distributed evenly across groups, this remains something the video footage could not reveal. It is also worth noting that the absence of oral notification is uncommon among many police departments, where officers are encouraged or required to inform civilians that they are being recorded. This design feature, therefore, creates a somewhat artificial sce- nario that may not fully reflect real-world practice. Third, the analysis was based on only 48 use-of-force cases—a relatively small sample. As a result, many comparisons lack statistical power and significant results, and the findings should be interpreted as exploratory patterns rather than definitive Body-worn camera footage analysis of use-of-force encounters–a…1 3 conclusions. Future research should rely on larger samples and include a broader range of encounters, as all cases in our study involved force and thus represent only a narrow subset of police–public interactions. These incidents are inherently high-con- flict (as evidenced by the 91% rate of civilian resistance). Thus, our findings apply specifically to force encounters and should not be generalised to routine stops or everyday interactions, where the dynamics may be substantially different.

  Body-worn camera footage analysis of use-of-force encounters–a closer look at the paradoxical effect of visual warnings · 2026 · DOI

• Despite the contributions of this study, several limitations should be acknowledged. First, it is important to note that at the time of data collection, AI-assisted policing models had not yet been fully implemented in routine law enforcement practice in the study site. Consequently, officers’ evaluations were based on hypothetical sce- narios rather than direct, lived experience with operational AI systems. In addition, the vignette described the use of “an AI model to predict criminal behavior” without providing detailed information about the specific type of technology or its opera- tional features. Without a more detailed description, respondents may have drawn on their own assumptions or prior knowledge of AI technologies, potentially leading to variation in how the scenario was interpreted. Future research should provide more specific descriptions of AI systems or explicitly vary the type of AI application to better isolate their distinct effects. Second, the recruitment strategy relied on internal police WeChat groups, which introduces the possibility of selection bias. Within China’s centralized and hierar- chical policing system, where technological modernization is an institutional prior- ity, officers who are more active on digital communication platforms and willing to participate in surveys about AI may constitute a self-selected subgroup that is more receptive to technological innovation or more aligned with organizational expecta- tions than the broader police population. This dynamic may result in an overestima- tion of officers’ support for AI-assisted policing. More specifically, such selection Contextual determinants of police attitudes toward AI-assisted…1 3 processes may lead to upwardly biased estimates of AI instrumental legitimacy, while potentially understating officers’ ethical concerns about AI if more skeptical or criti- cal individuals were less likely to participate. Future research should therefore adopt more systematic and representative sampling approaches, such as stratified sampling or department-level recruitment, to mitigate potential selection bias and enhance the robustness of findings. Third, the experimental manipulations were not designed to directly target spe- cific dimensions of ethical concern related to AI, such as privacy, transparency, or algorithmic bias. As a result, the absence of significant effects on ethical concerns may reflect that these attitudes are not responsive to the types of situational manipula- tions employed in this study. Future research should examine these dimensions using designs that directly manipulate such features. Finally, the vignette described the scenario using “suspected” criminal activity rather than confirmed wrongdoing. This wording may introduce ambiguity in how respondents interpret the situation, as participants may differ in whether they per- ceive the scenario as involving actual criminal behavior or merely the possibility of wrongdoing. While our use of this wording was intentional and grounded in the realities of policing, where officers typically operate based on suspicion rather than confirmed criminality at the time of intervention, it limits our ability to distinguish responses to confirmed versus suspected criminal activity. Future research could more explicitly manipulate the level of certainty to better isolate its effects on per- ceptions of AI-assisted policing.

  Contextual determinants of police attitudes toward AI-assisted policing: a vignette experiment · 2026 · DOI

• Although the present study executed a survey experiment with a sample size large enough to detect small effects, providing enhanced internal validity, there are some limitations worth mentioning to aid in future research. First, although there was a high degree of internal validity, not all of the observable characteristics of the sam- ple are nationally representative. While the sample is representative of sex, White respondents, and a balanced political sample, Black respondents were over-sampled. While not representative of the rest of the US, this over-representation may have been beneficial to the current study due to the disparity in medical errors occur- ring to Black Americans and the over-representation of Black Americans in police- caused deaths. Nonetheless, readers should take caution when extrapolating these findings to the rest of the US. Public concern and medical and police policy may be shaped by factors beyond national estimates of related deaths; therefore, researchers should continue to examine which factors influence the public to encourage pub- lic policy. While we were powered to detect a small effect, we did not pre-register heterogeneity tests across racial groups because even with the oversampling of Black Concerning, but do we need policy reform? An experimental analysis…1 3 respondents during Prolific’s distribution, we did not project a large enough number of Black respondents within each condition to run a powered heterogeneity test. Therefore, future research should assess whether these treatments were heterogeneous across racial groups. Additionally, we did not embed a manipulation/attention check in the survey. The inability to remove potentially inattentive respondents may have altered results. However, empirical research has shown that removing respondents due to failed atten- tion checks introduces bias (Aronow et al., 2019; Montgomery et al., 2018). Finally, although Prolific states that it handles low-performing and AI-generated responses (Pro- lific, 2025), we did not implement AI checks (e.g., reCAPTCHA) in the survey to help minimize AI-based responses.

  Concerning, but do we need policy reform? An experimental analysis of public reactions to accurate statistics on deaths caused by police officers and medical professionals · 2026 · DOI

• From a methodological perspective – as one reviewer pointed out – there is a question about how to appropriately analyse multi-scenario vignette studies and the sensitivity of results to corrections for multiple testing. Should we pool and adjusted for multiple hypothesis testing, or independently analyse each scenario (as we have done)? At one extreme, it could be argued there are 42 hypotheses being tested (six Are Police vetting decisions free from bias? An experimental vignette…1 3 ethnic groups X seven scenarios). If we accept that, and apply the (conservative) Bonferroni correction, then the Black Caribbean result for scenario four still stands: the 5% alpha threshold would be 0.05/42 = 0.00119, whereas the p-value for that result was 0.00091. Similarly, if we apply the false-discovery rate (FDR) correction (Benjamini and Hochberg, (1995) then the result for Black versus White British in scenario 4 would again hold (the q-value is 0.0012 but the p-value is p = 0.0009). In the language of Greenland’s (2019), a p-value of 0.0009 is the equivalent of flipping a coin and getting heads ten times in a row. But as described above, each ‘round’ of the testing was randomised and there was no read-across from the different scenarios. That is, the information in scenario 1 is irrelevant to scenario 2, meaning that in each scenario there are six hypotheses being tested. If we assume that adjustment is needed by scenario, then the results for scenario four still stand following FDR correction. One limitation we have only partially addressed is serial correlation by participant. However, beyond using robust standard errors, we cannot correct for that further because we do not have identifiers for participants. Although we know robust SEs are not intended to deal directly with multiplicity and will not fully account for clustering, they will be larger than naïve SEs, and thus partly address both issues. Use of an online environment meant we were not able to exactly replicate all facets of the process in terms of being able to check additional (fictional) databases. This affects external validity, but vetting decision-makers perform most of their checks without meeting candidates, so the setup is true to life in that respect. We also sought to match the look and layout of the existing vetting environment and some steps in the process, including the ability to update decisions based on new information, which does mirror real-life decision-making (see Appendix). Separately, as we acknowledge, we were unable to include the population of vetting decision-makers. This affects generalisability, but with holidays, sick leave, other absences, and workload pressure, having 63% of all vetting decision-makers is substantively useful for understanding decision-making.

  Are Police vetting decisions free from bias? An experimental vignette study with vetting decision-makers · 2026 · DOI

• Future research should focus on systematic evaluation of existing interventions, including both short-term and long-term outcomes, to identify which programs effectively improve mental wellbeing. This lack of a standardized approach means that the level of support police personnel receive varies widely, resulting in them having to navigate limited interventions to gain comprehensive mental health support, rather than benefiting from a cohesive mental framework of training and support.

  A review exploring wellbeing interventions for police personnel in the UK: what support is available? · 2026 · DOI

• Black and Hispanic emerging adults disproportionately experience racism and police violence, yet little is known about how racism-based police violence (RPV) affects those living with disabilities.

  Intersecting Inequities: Racism-Based Police Violence and Mental Health Correlates Among Black and Hispanic Emerging Adults With Disabilities · 2026 · DOI

• This study makes five key recommendations to improve South Africa's response to kidnapping. First, SAPS and PSCs should establish a formal Joint Anti-Kidnapping Operations Centre with clearly defined roles, shared protocols, and mutual accountability. Colombia's experience, where coordinated public-private efforts reduced kidnapping by over 90% between 2000 and 2015, shows this approach can work (Miraglia, Ochoa & Briscoe, 2012). Second, both sectors should invest in advanced investigative tools such as digital forensics, mobile tracking, and link analysis software. This can be supported by joint training programmes delivered on equal terms between SAPS and PSC personnel. Thirdly, a centralised kidnapping data repository drawing on reports from SAPS, PSCs, and community organisations should be established to address the widespread underreporting and misclassification identified in this study. Fourth, victim support services including trauma counselling, legal assistance, and post- recovery care should be state-funded and formally integrated into the response framework, rather than remaining fee-based services available only to those who can afford them. Finally, policymakers must guard against over-reliance on private security. PSC services are largely inaccessible to lower-income communities, creating an unequal protection system. Without stronger oversight, expanded PSC involvement also risks surveillance abuse and rights violations. Nigeria's experience warns against an unregulated, fragmented private security landscape, while the United Kingdom and Australia show that well-regulated public-private partnerships can succeed without these problems (Button, 2019; Hounmenou & O'Grady, 2023).

  Guardians against abduction: The role of private security companies in combating kidnapping in South Africa · 2026 · DOI

• This study has several limitations. The findings reflect a specific point in time and may not capture rapidly evolving kidnapping patterns. Second, the study focused on urban centres which were primarily Gauteng and the Western Cape. This limits its applicability to rural and peri-urban contexts. Most significantly, the study relied exclusively on interviews with private security professionals, meaning victim perspectives and SAPS viewpoints are entirely absent. This reveals a gap that future multi-stakeholder research should address. The use of snowball sampling can introduce selection bias, since participants from shared professional networks are likely to hold similar pro-PSC views, potentially underrepresenting critical perspectives. Finally, the exclusion of community organisations and civil society voices limits the study's ability to address the equity dimensions of private security provision for the 513 Pillay, International Journal of Research in Business & Social Science 15(3) (2026), 504-515 most vulnerable populations. These limitations define the boundaries of the study rather than invalidating its findings and collectively point toward a future research agenda that is more representative, multi-stakeholder, and longitudinal in design.

  Guardians against abduction: The role of private security companies in combating kidnapping in South Africa · 2026 · DOI

• The matters reported below are limited to the significant internal control deficiencies that resulted in the findings on compliance with legislation included in this report. The conclusion is, however, based on and limited to the available Internal Audit results throughout the 2018/19 financial year which was assessed in terms of the Committee of Sponsoring Organisations of the Treadway Commission (COSO) Internal Control Framework.

  Data-Driven Crime-Fighting: A Case Study of the City of Cape Town, South Africa · 2026 · DOI

• Originality/value Limited research has been conducted in relation to the College of Policing licensed higher education programme, the PPD, equipping aspiring police officers to successfully join the service and influence the cultural change.

  Police misconduct and social media: perceptions of aspiring future police officers · 2024 · DOI

• BWC footage did not account for the visual information and the behaviors of the suspect, potential threats, and bystanders who influence the officers' decision-making during the use of force encounters.

  The eyes have it! Functional field of view differences between visual search behavior and body-worn camera during a use of force response in active-duty police officers · 2024 · DOI

• While scholars have long recognized that the organizational composition of collaborative forums can have a dramatic impact on the efficacy and equity of associated forum outputs, little is known about whether such composition impacts the manner in which everyday citizens perceive forum processes and/or participating organizations.

  Manifesting symbolic representation through collaborative policymaking · 2024 · DOI

• However, when the public authority’s surveillance of security is complemented by individual and collective forms of social care and the market services of property protection, the interpretation of the law is insufficient for exploration, and it is worthwhile to turn to the methodologies offered by sociology.

  Komplementer rendészet · 2023 · DOI

• Though the imminence requirement’s impacts on policing and public safety require ongoing study, the principle of equality before the law and the Doctrine of Doing and Allowing both provide moral grounds for this requirement, especially given the lack of evidence that the status quo helps protect life.

  Applying the Imminence Requirement to Police · 2023 · DOI

• Value: The professional-historical review ranges from the stabilization of the neighbourhood police officers’ service after the initial period to the regime change, so it is limited to the period between 1960 and 1990.

  A körzeti megbízotti szolgálat története 2. A kezdetektől a rendszerváltozásig · 2022 · DOI

• It examines the findings from 46 interviews and offers insight into a hitherto under-researched area regarding policing of domestic violence by officers at the start of their policing career.

  ‘Bread and butter’ policing: first-year constables and domestic violence · 2022 · DOI

• Despite both approaches being widely advocated and implemented, presently little is known about police practitioner understanding of the relationship between POP and EBP, both in principle and in practice.

  Police perceptions of problem-oriented policing and evidence-based policing: evidence from England and Wales · 2022 · DOI

• Abstract Research Summary This study uses interviews with 23 police officers from a small police department to conduct an in‐depth examination of their perceptions of three critical but understudied areas related to body‐worn camera programs: the implementation and policy‐making process, supervision, and training.

  “Culture eats strategy for breakfast”: An in‐depth examination of police officer perceptions of body‐worn camera implementation and their relationship to policy, supervision, and training · 2022 · DOI

• While a significant amount of research has been conducted in regards to public opinion of police, limited information can be found that reflects college students' opinions toward the subject, and even fewer studies have specifically used students on a college campus as a sample to study such opinions.

  Reevaluating factors associated with negative attitudes toward police: capturing information from college students · 2021 · DOI

• A strong empirical relationship exists between severity of abuse and police help-seeking among WOC survivors more generally; however, scarce literature exists examining the nature of interactions between officers and WOC survivors during life-threatening abuse situations.

  Interactions between law enforcement and women of color at high-risk of lethal intimate partner violence: An application of interpersonal justice theory · 2021 · DOI

• This study addresses this gap in the literature by analyzing the effect that perceiving four types of neighborhood crime as a problem had on levels of police legitimacy among 1773 respondents from a city in the Appalachians.

  Correlates of perceptions of police legitimacy: Do perceptions of neighborhood crime matter? · 2021 · DOI

• This article examines the hidden and under-researched area of bullying and harassment of Black and Asian Minority Ethnic (BAME) women in the police service in England.

  Racist bullying of BAME (Black and Asian Minority Ethnic) women within police services in England: Race, gender and police culture · 2021 · DOI

Most-cited papers in Policing Practices and Perceptions

• Racially Biased Policing: Determinants of Citizen Perceptions · Social Forces · 2005 · 396 citations
• A large-scale analysis of racial disparities in police stops across the United States · Nature Human Behaviour · 2020 · 368 citations
• Production of Crime Rates · American Sociological Review · 1970 · 333 citations
• The Opinion-Mobilizing Effect of Social Protest against Police Violence: Evidence from the 2020 George Floyd Protests · American Political Science Review · 2021 · 280 citations
• Race, Class, and Perceptions of Discrimination by the Police · Crime & Delinquency · 1999 · 265 citations
• Burnout, job stress and violent behaviour among Dutch police officers · Work & Stress · 1999 · 253 citations
• An Institutional Perspective of Policing · The Journal of Criminal Law and Criminology (1973-) · 1992 · 238 citations
• Subjective and Objective Measures of Police Service Delivery · Public Administration Review · 1983 · 222 citations
• ENOUGH: COVID-19, Structural Racism, Police Brutality, Plutocracy, Climate Change—and Time for Health Justice, Democratic Governance, and an Equitable, Sustainable Future · American Journal of Public Health · 2020 · 219 citations
• Administrative Records Mask Racially Biased Policing · American Political Science Review · 2020 · 211 citations

Most recent work

• Facial recognition technology in law enforcement- a scoping review of existing empirical studies · Police Practice and Research · 2026
• Anger About Police-Civilian Relations and Support for Policing Reforms · Journal of Research in Crime and Delinquency · 2026
• ‘It could be the straw that breaks the camel’s back’: the impacts of anti-police sentiment on police officer identity in Australia · Police Practice and Research · 2026
• A review exploring wellbeing interventions for police personnel in the UK: what support is available? · Police Practice and Research · 2026
• Conducting online randomized experiments: lessons from a policing survey using AI-Edited videos · Journal of Experimental Criminology · 2026
• Body-worn camera footage analysis of use-of-force encounters–a closer look at the paradoxical effect of visual warnings · Journal of Experimental Criminology · 2026
• News headline themes and approval of police use-of-force: evidence from a national survey experiment · Journal of Experimental Criminology · 2026
• An analysis of absorptive capacity within a public security institution · Journal of Workplace Learning · 2026
• Left Realism, Right Realism and police community relations. · International Journal of Creative and Open Research in Engineering and Management · 2026
• Police Self-Image in Relation to Public Trust in the Second Congressional District of Davao Oriental · Davao Research Journal · 2026

Related topics in Social Sciences