Police Activities and Community Views of Police in Crime Hot Spots

Abstract

Evidence on how hot spot policing affects community members’ views of police is very limited and inconclusive. Scholars have thus called for further study of community attitudes in hot spots to guide police in the formulation of hot spot strategies—an issue that is especially salient given recent public controversy surrounding policing, particularly in the United States. Using survey responses collected in 2018 from more than 1,000 community members living or working in more than 100 hot spots across 2 mid-sized cities in the United States, this study examines how community members’ perceptions of police activities in hot spots relate to their wider attitudes about police. Bivariate and multivariate analyses indicate that community members in hot spots in both cities exhibit more positive attitudes towards police along several dimensions (e.g., trust and confidence in police, views of police legitimacy, and perceptions of police responsiveness and procedural justice) when they see more frequent patrol and when they see positive police-community interactions. They have more negative views of police when they witness higher levels of investigative and enforcement activity. The findings support hot spot policing strategies that emphasize regular, systematic patrol in hot spots, complemented by positive community engagement efforts and problem-solving work. In contrast, they imply that enhanced enforcement activity in hot spots should be used judiciously.

Keywords:

• policing
• hot spots
• community attitudes
• survey

Introduction

“Hot spot” policing—i.e., policing focused on small geographic places or areas where crime is concentrated—has been one of the most important policing innovations of recent decades (Weisburd & Braga, 2019). Police in the United States and elsewhere commonly use crime mapping and targeted hot spot strategies to address a variety of crime problems (Braga et al., 2019; Burch, 2012; Koper, 2014; Reaves, 2010; Sherman, 2013; Weir & Bangs, 2007), and numerous evaluations support the effectiveness of these approaches in reducing crime and disorder (for reviews, see Braga et al., 2019; Braga & Weisburd, 2022; Lum & Koper, 2017; National Academies of Sciences, Engineering, and Medicine [NAS], 2018; National Research Council [NRC], 2004; Telep & Weisburd, 2012).

However, there is still considerable concern and debate regarding the impact of hot spot policing strategies on other aspects of community well-being, particularly police-community relations. Although preventing crime remains a core responsibility of the police, recent public backlash against aggressive policing, police use of force, and racially disparate policing, particularly in the United States, shows that there is a need to assess police strategies not only for their crime prevention benefits but also for their impacts on other community outcomes including perceptions of police legitimacy and procedural justice, police-community relations and cooperation, and community members’ confidence and trust in police. In addition to their intrinsic value in a democratic society, police strategies that enhance police-community relationships may also bolster crime prevention efforts, as some evidence suggests that people are more likely to obey the law, offer cooperation, and support the police when they view the police as legitimate (e.g., NRC, 2004; Peyton et al., 2019; Sunshine & Tyler, 2003; Tyler, 1990; 2004; Tyler & Fagan, 2008). While these issues have been longstanding concerns in policing, they have recently been reemphasized by the President’s Task Force on 21st Century Policing (2015) as well as by many police leaders and scholars (e.g., Lum & Nagin, 2017; NAS, 2018; NRC, 2004; Rosenbaum, 2019; Tyler, 2004).

Evidence on how hot spot policing affects community members’ views of police is very limited and inconclusive, and scholars have noted the need for further study of community attitudes in hot spots to guide police in the formulation of hot spot strategies that can prevent crime while also improving police-community relations (e.g., Haberman et al., 2016). Our study addresses these concerns by examining how community members’ perceptions of police activity in hot spots relate to their wider attitudes about police, including their views of police effectiveness, fairness, and legitimacy. As described below, we examine these issues using survey responses collected in 2018 from more than 1,000 community members living or working in more than 100 hot spots across 2 mid-sized cities on the East Coast of the United States. We attempt to extend prior community survey research on policing in three ways: 1) by adding to the very limited literature that specifically examines the views of residents and workers in micro hot spots of crime and disorder; 2) by making direct linkages between respondents’ exposure to police activities and their attitudes towards police (albeit in an exploratory, cross-sectional manner); and 3) by providing a view of community members’ attitudes in a contemporary context that has been characterized by growing public concern, criticism, and controversy surrounding policing. We conclude by considering the implications of the findings for designing optimal hot spot policing strategies.

Policing and Community Views in High-Crime Locations

In brief, hot spot policing strategies focus police activity on specific addresses, street blocks, intersections, and clusters of blocks where crime is disproportionately concentrated. These approaches are grounded in research showing that about half of crime in a jurisdiction generally occurs at 5% or less of its street blocks (for a review, see Weisburd, 2015). Further, this concentration is stable over time, due in large measure to chronic problem locations that have social and environmental features that cause, attract, or facilitate criminal and disorder behavior (e.g., Andresen et al., 2017; Brantingham & Brantingham, 1993; Curman et al., 2015; Eck & Weisburd, 1995; Groff & Lockwood, 2014; Groff et al., 2010; Koper et al., 2015; Sherman et al., 1989; Weisburd et al., 2004; 2012).

Although police have had considerable success in reducing crime and disorder with hot spot policing, the effects of these efforts on community members’ views of police and their communities are less clear. On the one hand, hot spot policing might promote positive views of the police. Many members of the public support greater police presence in their communities, particularly in high-crime areas (e.g., Chermak et al., 2001; Hawdon et al., 2003; Saad, 2020; Shaw, 1995). For those who live and work in hot spots, greater police presence might enhance their feelings of safety as well as their sense of receiving a needed and fair allocation of police resources (thus improving their sense of distributive justice—see NRC, 2004).¹ Police might also develop and sustain good relations with community members in hot spots if they use community-oriented policing approaches (Gill et al., 2014; NAS, 2018; Skogan, 2019; Weisburd, 2016) and interact with community members in a respectful, empathetic, and fair manner (Mazerolle et al., 2013; Walters & Bolger, 2019; Weisburd et al., 2022). Reductions in crime and disorder caused by hot spot policing might also enhance public views of the police (e.g., Haberman et al., 2016; Wheeler et al., 2020).

On the other hand, some have raised concerns that hot spot policing approaches may have a ‘backfire effect,’ eroding police-community relations and even increasing fear and undermining collective efficacy, particularly if police emphasize aggressive enforcement-oriented strategies (e.g., Kochel, 2011; Rosenbaum, 2019; also see Hinkle & Weisburd, 2008; Weisburd, 2016). For example, heavy police presence and enforcement in hot spots might increase negative interactions between police and community members, increase public fear (of crime and/or police encounters), and lead community members to view police negatively as an occupying force. Such backfire effects might also weaken the ability of hot spot interventions to produce long-term reductions in crime if people become less likely to view the police as legitimate authorities and less likely to cooperate with them in reporting and preventing crime.

Knowing more about how different hot spot policing approaches affect public perceptions and attitudes would greatly inform research on best strategies to use in hot spots for enhancing crime prevention as well as police-community relations (Braga et al., 2019; Braga & Bond, 2008; Groff et al., 2015; Kochel et al., 2015; Koper, 2014; Taylor et al., 2011). Moreover, the issue of public reaction to hot spot policing is particularly salient in the current social context. In the United States, public concern regarding overly aggressive policing, particularly in high-crime and predominantly non-white communities, has been rising since the 2014-2015 timeframe, when a number of high-profile police use of force cases (including those in Ferguson, Missouri, Baltimore, and Chicago) led to significant protests, riots, and calls for police reform, as well as the formation of the President’s Task Force on 21st Century Policing (2015). Anti-policing protests and calls for police reform have intensified further since the highly egregious and publicized killing of George Floyd by police in Minneapolis in 2020, which sparked protests in thousands of communities across America and around the world (Bliss, 2020; Buchanan et al., 2020). These events have raised concerns about a legitimacy crisis in policing that may be worsening attitudes towards police (Kochel, 2019; Reny & Newman, 2021; Tesler, 2020), undermining community cooperation with police (Ang et al., 2021; Desmond et al., 2016), and even fueling increases in violence, particularly in Black communities, as more people become alienated from legitimate methods of social control and turn to informal and violent methods of dispute resolution (Rosenfeld, 2016; Rosenfeld et al., 2021). Against this backdrop, it is crucial for police to calibrate hot spot policing strategies in ways that balance crime prevention with community sentiment.

At present, however, there is very limited evidence that specifically addresses how police activities affect the views and attitudes of residents and workers in micro hot spot locations that are the focus of place-based policing strategies. Only a handful of survey studies have investigated this issue, testing how different variations of hot spot policing (including directed and foot patrol, order maintenance/disorder reduction, problem-solving, offender-focused tactics, and community-building) affect varying views about police legitimacy (i.e., the belief that police are legitimate authorities whose decisions should be respected), procedural justice (i.e., the extent to which police treat people respectfully and fairly), satisfaction with police, and willingness to cooperate with police. In sum, experiments in Philadelphia (Ratcliffe et al., 2015), Brooklyn Park (Minnesota) (Weisburd et al., 2021), and the California cities of Redlands, Ontario, and Colton (Weisburd et al., 2011) produced null results, suggesting that hot spot policing does not significantly affect community members’ attitudes towards police for better or worse. In contrast, an experiment in St. Louis County, Missouri found that directed patrol may produce negative effects on community views of procedural justice, police legitimacy, and police abuse, at least temporarily (Kochel & Weisburd, 2017; also see Kochel et al., 2015). However, the same study also found that both directed patrol and problem-solving interventions were linked to long-term positive effects on the willingness of respondents to cooperate with police, though these changes did not become evident until several months after the interventions ended.^2,3

Only one of these studies was conducted after 2014 (Weisburd et al., 2021), so their results may not be indicative of community sentiment and attitudes in the current context. Another ambiguity in these studies is that they focused on overall contrasts of respondents in intervention and non-intervention locations, but it seems that many respondents in the intervention areas were unaware of the enhanced police activities. Where measured, studies have shown no differences in perceptions of police activity across intervention and non-intervention locations or otherwise shown relatively low levels of program awareness and exposure (and little or no change therein) among respondents in intervention areas (Kochel et al., 2015; Weisburd et al., 2011; 2021). It is unclear how attitudes changed among people who had greater awareness of the hot spot policing efforts and, if so, how those changes varied depending on the nature of those efforts.

Beyond these hot spot studies, there is a larger body of survey research showing that community policing strategies have generally positive impacts on community members’ satisfaction with police and their views of police legitimacy, including in high-crime areas (e.g., see review and meta-analysis in Gill et al., 2014). This research might also usefully inform hot spot strategies, but with caveats. Community policing strategies encompass a variety of efforts to work more closely with citizens to co-produce public safety and develop solutions to local concerns (e.g., NAS, 2018; Skogan, 2006). In practice, these efforts may include one or a combination of strategies like foot patrol, neighborhood watch, community meetings, citizen contact patrols, community sub-stations, problem-solving (when done collaboratively with citizens), disorder reduction, and efforts to improve police accountability (e.g., citizen review boards). However, this makes it difficult to draw clear inferences about which aspects of community policing efforts have the most meaningful impacts on public views (for an exception, see Peyton et al., 2019). Some research suggests, for example, that patrol visibility may have stronger positive effects on attitudes towards police than community policing efforts (Hawdon et al., 2003). Furthermore, community policing studies generally focus on larger areas such as patrol beats and neighborhoods; hence, they may not accurately reflect the views of residents and workers in micro hot spots within these areas. Finally, most studies of community policing were conducted during the 1980s, 1990s, and early 2000s (see Gill et al., 2014), which arguably raises questions about how strongly their results can be generalized to the present context.

Our study attempts to address these gaps by examining the views of residents and workers in micro hot spots in a contemporary context (post-Ferguson but pre-Floyd) and by making direct connections between their awareness of police activities and their views of police. We build on Haberman et al.’s (2016) recommendation that police use community surveys when possible to guide their selection of hot spots strategies. Based on their study of public attitudes in Philadelphia hot spots, Haberman and colleagues argued that strategies that enhance perceptions of procedural justice while reducing social disorder and fear of crime are most likely to improve community satisfaction with police. However, their study did not examine the effects of specific police activities as we do here. Our study also directly complements the work of Hawdon et al. (2003) whose cross-sectional study of 41 South Carolina neighborhoods in 2001 showed that higher levels of patrol as perceived by residents were more strongly linked to positive perceptions of police effectiveness and trustworthiness than were knowledge of community policing tactics or police contacts. We take a similar approach using more recent data from respondents in micro hot spots in two cities.

Data and Methods

This study was conducted in two geographically proximate cities, referenced anonymously as Avalon and Willowdale, located within the same state in the South Atlantic region of the United States. Both have populations in the range of 100,000 to 200,000 people and geographical areas between 50 and 70 square miles. Avalon’s population is about half Black and roughly 40% White, while Willowdale’s population reflects nearly the opposite distribution. Hispanics account for between 5% and 10% of the population in each city. The median household income is 10%-15% below the national average in both cities, and 14%-15% of the population lives in poverty, a rate somewhat higher than the national poverty rate.⁴ In 2017, the year prior to our study, Avalon experienced more than 400 UCR Part 1 violent crimes and more than 4,000 UCR Part I property crimes. The city’s overall violent crime rate is considerably lower than that of other similar-sized cities nationwide, though its murder and property crime rates are considerably higher. Willowdale, in contrast, reported more than 900 UCR Part 1 violent crimes in 2017 and more than 5,000 UCR Part 1 property crimes. The city’s violent and property crime rates are very similar to the overall national rates for cities of its size, but, like Avalon, the city has a considerably higher murder rate.⁵

The survey reported here was conducted during the fall of 2018 and early winter of 2019 (September 2018-January 2019) as part of a study of hot spots that our team conducted with the police agencies in Avalon and Willowdale.⁶ At the time of the survey, the police departments in both cities had geographically focused initiatives, but their systematic efforts were generally targeted on neighborhood areas and administrative patrol areas rather than micro hot spots identified through highly precise crime analysis. Both departments also had considerable experience with community- and problem-oriented policing practices.

Identification of Hot Spots

Working with law enforcement personnel in both cities, we identified a total of 102 hot spots across the two cities (60 in Avalon and 42 in Willowdale). To do this, our team analyzed violent crime incidents during the preceding two years, first using kernel density analysis with Robert Cross and Getis-Ord Gi* statistics and then assessing crime counts at street locations to identify the highest crime concentrations in single and multifamily residential and commercial/business areas. Patrol commanders, sergeants, and patrol officers from the agencies then assisted us in vetting and delineating potential hot spots based on their extensive working knowledge of the crime problems and geographic features of the identified locations. The final selected hot spots were small areas of crime concentration (i.e., clusters of street blocks) that averaged 0.1 and 0.28 square miles in Avalon and Willowdale, respectively (and were separated by buffers averaging 606 and 911 feet, respectively).⁷ The hot spots were largely residential locations, with businesses accounting for 5%-6% of their addresses in both cities.⁸ Although the hot spots were selected based primarily on their violent crime levels, they also had high concentrations of property offending. At the time of our survey, the locations averaged about 3 violent and 26 property offenses per year in Avalon and about 13 violent and 67 property offenses per year in Willowdale.

Sample Selection and Survey Procedures

We surveyed on average 10 to 11 community residents and workers in each of the 102 hot spots for a total of 1,082 completed surveys. We first selected addresses using an area-based stratified random sampling approach to sample residential and commercial addresses based on hot spot size and the residential/commercial composition of the locations. Using the U.S. Postal Service (USPS) delivery sequence file (DSF) as a sampling frame, we randomly selected about 30 addresses in each hot spot (totaling 3,065 addresses across all hot spots) and attempted to contact people at these addresses in person. Out of the 1,928 addresses that were deemed eligible for the study (i.e., addresses that were not vacant and that had adults who had been living or working in the neighborhood for at least six months and who understood and spoke English), surveys were completed at 1,082, for a response rate of 56%. This response rate compares favorably to other community surveys in high crime areas.⁹

For residential households, an adult household member age 18 or older responded to the survey. For business addresses, we surveyed the owner, the manager, or another adult employee (age 18 or older) who was in charge for the day, if neither the manager nor the owner was available. Trained interviewers conducted the surveys in person under the supervision of field managers. The team attempted survey interviews from 9 AM to about 7 PM for the household surveys and from 9 AM to about 9 PM for businesses, inclusive of weekdays and weekends. The team implemented an extensive follow-up program to recruit hard-to-reach participants. The team made at least six attempts at visiting each potential respondent until a survey was completed, the team received a refusal to participate from the respondent, or the address was categorized as “non-responsive.”

The final sample of respondents included 562 individuals from Avalon and 520 from Willowdale. As shown in Table 1, 84%–86% of the sample consisted of hot spot residents. In both cities, two-thirds to three-fourths of the respondents were non-white (primarily Black), most were female (55%-56%), roughly two-thirds were between the ages of 21 and 59, and nearly two-thirds had a high school or limited college education. Three-quarters had lived or worked in the location for at least a year, with a plurality (41%-48%) having done so for one to five years.

Table 1. Descriptive statistics.

	“Avalon” avg (s.d.)	Range	N available	“Willowdale” avg (s.d.)	Range	N available
Outcome variables
Trust	3.9 (0.8)	1 − 5	559	4.0 (0.9)	1 − 5	518
Legitimacy	3.8 (1.0)	1 − 5	560	3.8 (1.0)	1 − 5	515
Reporting	3.9 (1.1)	1 − 5	558	3.9 (1.0)	1 − 5	518
Capability	3.9 (0.8)	1 − 5	554	3.9 (0.9)	1 − 5	515
Responsive	2.7 (1.4)	1 − 5	525	2.7 (1.4)	1 − 5	476
Satisfaction	4.0 (1.2)	1 − 5	552	4.1 (1.1)	1 − 5	514
Procedural justice	3.7 (0.9)	1 − 5	534	3.7 (1.0)	1 − 5	494
Predictors
Perceived patrol	2.4 (0.9)	1 − 5	561	2.7 (0.9)	1 − 5	519
Perceived enforcement	1.5 (0.8)	1 − 5	549	1.6 (0.9)	1 − 5	514
Perceived community engagement	1.4 (0.8)	1 − 5	546	1.5 (0.8)	1 − 5	508
Age	2.7 (0.9)	1 − 4	555	2.6 (1.0)	1 − 4	517
1 = 18 to 25 (11% / 15%)
2 = 26 to 39 (34% / 31%)
3 = 40 to 59 (34% / 33%)
4 = 60+ (21% / 21%)
Race	1.7 (0.5)	1 − 2	555	1.7 (0.5)	1 − 2	518
1 = White (28% / 32%)
2 = Non-white (72% / 68%)
Gender	1.4 (0.5)	1 − 2	561	1.4 (0.5)	1 − 2	520
1 = Female (56% / 55%)
2= Male (44% / 45%)
Time living/working in area	2.1 (0.8)	1 − 3	562	2.1 (0.7)	1 − 3	518
1 = 6 months-1 yr (24% / 20%)
2 = 1-5 yrs (41% / 48%)
3 = 5+ yrs (35% / 32%)
Education	3.2 (1.3)	1 − 6	554	3.1 (1.3)	1 − 6	516
1 = Less than high school (5% / 6%)
2 = high school graduate (31% / 35%)
3 = some college (29% / 29%)
4 = associate degree (14% / 10%)
5 = bachelor’s degree (14% / 13%)
6 = graduate or prof. degree (6% / 7%)
Residential or commercial respondent	0.9 (0.3)	0 − 1	562	0.8 (0.4)	0 − 1	520
0 = commercial (14% / 16%)
1 = residential (86% / 84%)
Property crime victimization	1.1 (0.2)	1 − 2.5	561	1.1 (0.2)	1 − 3.13	520
Violent crime victimization	1.0 (0.1)	1 − 2.25	562	1.0 (0.1)	1 − 2.5	520
Fear of crime	1.9 (1.0)	1 − 5	555	1.9 (1.0)	1 − 5	517
Perceived crime/disorder	1.8 (0.8)	1 − 5	559	2.0 (1.0)	1 − 5	517
Citizen-initiated police contact	2.4 (0.9)	2 − 8	559	2.5 (1.0)	2 − 9	519
Proactive police contact	2.2 (0.7)	2 − 8	553	2.3 (0.9)	2 − 9	517
Recent police stop	1.2 (0.5)	1 − 5	559	1.2 (0.6)	1 − 5	519
Police misconduct	1.2 (0.6)	1 − 5	554	1.2 (0.6)	1 − 4	517
N of hot spots	60			42
N of respondents	562			520

Notes: See text and Appendix Table A1 for descriptions of the outcome variables and measures of perceived patrol, enforcement, and engagement. Detailed breakdowns of demographic items show percentage values for Avalon and Willowdale, respectively. The violent and property crime measures are based on items from the National Crime Victimization Survey and indicate respondents’ average victimization from several forms of violent and property crime during the prior six months on a scale ranging from 1 = never to 5 = 5 or more times. The fear of crime scale is taken from Weisburd et al. (2010) and reflects respondents’ average fear of experiencing several forms of violent and property victimization on a scale ranging from 1 = very unlikely to 5 = very likely. The perceived crime and disorder scale, adapted from Ratcliffe et al. (2015) and Weisburd et al. (2010), measures respondents’ concern about numerous crime and disorder problems on a scale ranging from 1 = not a problem to 5 = a big problem. The police contact and stop items (see text) were measured on scales ranging from 1 = never to 5 = 5 or more times. The police contact and proactive police contact measures each represent the sum of two items (with minimum of 2 = never for both measures). The police misconduct measure (see text) is based on a multi-item scale taken from Rosenbaum et al. (2007) with item responses ranging from 1 = never to 5 = always.

Survey Measures

We measured respondents’ perceptions of police activity in the hot spots, our key predictor variables, through questions asking how often they had seen police engaging in several patrol, enforcement, investigation, and community engagement activities during the prior six months within their respective locations. Each respondent was shown a map of the boundaries of his or her local hot spot to make it easier to respond to the survey items with reference to that specific location. All items were measured on a 5-point scale with responses of never, once a month or less, a few times a month, a few times a week, and daily.¹⁰ Using these items, we created combined average measures of perceived patrol frequency (based on measures of motor and foot patrol),¹¹ perceived enforcement and investigative activity (based on measures of searching for, stopping/questioning, and arresting people), and perceived community engagement (based on measures of informal/friendly conversations with community members and solving community problems).

For our dependent variables, we measured several dimensions of respondents’ attitudes towards police using multi-item psychometric scales measuring police legitimacy (e.g., feeling that one should obey police directives because it is the right thing to do), trust in police (e.g., trusting the police and being unafraid around them), procedural justice in police behavior (e.g., the extent to which police treat citizens with respect and professionalism), police responsiveness to the community (e.g., the extent to which police share information with residents and work with them to solve local problems), police capability in handling crime, disorder, and emergencies, and the willingness of respondents to report crimes, problems (e.g., disorder and suspicious activity), and information (e.g., to solve crimes) to police. The scale items and reliabilities (which ranged from 0.72 to 0.98) are reported in Appendix Table A1, along with their sources (all scales were taken or adapted from prior studies). In addition, we measured respondents’ satisfaction with police presence in the area using a single item (see Table A1). All scales and items were measured on a 1 to 5 scale with higher scores indicating more positive views (items were reverse-coded as needed).

For multivariate analyses, we also measured several covariates reflecting background characteristics and experiences of the respondents that may condition their views of police and relate to their varying likelihoods of exposure to different types of police activity. These include the respondents’ demographic and socioeconomic characteristics (as described above), their recent experiences with violent and property victimization, their fear of crime, and their overall perception of crime and disorder in the location (e.g., Bolger et al., 2021; Brown & Benedict, 2002; Haberman et al., 2016; Maxson et al., 2003; Peck, 2015; Skogan, 2005; Sunshine & Tyler, 2003; Zahnow et al., 2017).¹²

In addition, we controlled for the number and nature of respondents’ recent contacts with police, which may both reflect and shape their views of police (Brown & Benedict, 2002; Maxson et al., 2003; Skogan, 2005; Sunshine & Tyler, 2003). Specifically, we used separate measures of how often during the last six months the respondents had: 1) initiated contact with police to discuss or address community problems (e.g., attending police-community meetings or working with police on problem-solving projects); 2) been contacted by police to discuss or address community problems; and 3) been stopped by police in the area. Finally, the analysis also included a measure of how often respondents had witnessed abusive police behavior in the location during the previous six months (e.g., stopping people without just cause and being physically abusive, profane, threatening, or discourteous).¹³ We note that the levels and types of police activity employed in a location seem likely to influence the probability that residents and workers in the location have different types of contacts with police and potentially their likelihood of witnessing police misconduct. For that reason, the police contact and misbehavior measures are also important for evaluating policing strategies. However, our primary focus is on how police activities affect community views more generally, including among people who have not had recent direct contacts with police.^14,15

Statistical Methods

We estimated the relationships between respondents’ perceptions of police activities and their attitudes towards police using bivariate correlations and multivariate ordinary least squares models that accounted for the respondents’ characteristics and experiences described above. The multivariate models were estimated with standard errors clustered by hot spot to allow for dependence between individuals from the same location.¹⁶ We judge relationships to be statistically significant if their probability levels were estimated to be .05 or less.

All analyses were first run separately for each city to assess commonalities and differences across locations. We did not have a priori reasons to believe that differences in the nature of policing or police-community relations between the cities would cause different patterns of results. However, similarities across the two jurisdictions could suggest common patterns that are more generalizable to other settings. Based on these initial results, which were quite comparable across cities, we also estimated models with survey responses combined from both cities and examined both unstandardized and standardized coefficients from these models.¹⁷

As a final step, we tested whether the impacts of police activities on community attitudes were invariant across White and Black respondents in the multi-city models by interacting the police activity measures with a race indicator.¹⁸ These tests add to a growing body of research that has examined whether the effects of police activities, behaviors, and performance on civilians’ evaluations of police vary across different groups (e.g., based on demographics, victimization, or prior experience with police) and social contexts (e.g., based on local levels of crime and disorder) (e.g., Fox et al., 2021; Reisig et al., 2021; Sunshine & Tyler, 2003; Wolfe et al., 2016; Zahnow et al., 2021). We focused on possible racial differences (which have not been apparent in several of the aforementioned studies) based on three considerations: the recent and dynamic social context surrounding race and policing in the United States (as discussed above); the more negative views of police held by Black citizens in general (Bolger et al., 2021; Brown & Benedict, 2002), which were also apparent in our data (see below); and our focus on urban hot spots where racial differences might be particularly salient due to the concentration of crime, disorder, and police activity in these locations.¹⁹

Results

Respondents had views of police that were very similar across both cities and generally more positive than negative. Scores for trust in police (“trust”), police legitimacy (“legitimacy”), willingness to cooperate with police (“reporting”), police capability (“capability”), satisfaction with police (“satisfaction”), and police use of procedural justice (“procedural justice”) averaged at or near 4.0 on a 5-point scale (with higher values indicating more positive views of police) in both sites (Table 1). Views were somewhat more negative about police responsiveness to the community (“responsive”), with scores averaging 2.7. With regard to police activity, nearly two-thirds of overall respondents reported seeing officers drive through their location daily or multiple times a week, though only about one in five had seen foot patrols (results not shown). One to two-fifths had seen other types of enforcement or engagement activities in their location (depending on the city and type of activity), with 8% to 17% having seen them more than once per month.

Preliminary bivariate correlations (Appendix Tables A2 and A3) showed that perceptions of patrol and community engagement had statistically significant positive correlations with several attitudes towards police in one or both cities, while perceived enforcement had significant negative correlations with most attitudes towards police in both cities (an exception was that perceived enforcement had a significant positive correlation with perceived police responsiveness in Avalon). The significant correlations between perceived police activities and attitudes towards police were generally small to medium in magnitude (ranging from |0.10| to almost |0.30|), with the exception that community engagement had correlations with police responsiveness that were between medium and large (|0.37-0.43|) (see Cohen, 1988: 77-81).

Table 2. Attitudes towards police regressed on perceived police activities, controlling for respondents’ characteristics and experiences (“Avalon” City).

Variable	Trust	Legitimacy	Reporting	Capability	Responsive	Satisfaction	Procedural Justice
Perceived patrol	.164**	.183**	.069	.142**	-.053	.467**	.164**
Perceived enforcement	-.092	-.039	-.113	-.135*	.182*	-.148	-.157*
Perceived community engagement	.083	-.019	-.047	.144**	.751**	.133	.212**
Citizen-initiated contacts	-.039	-.074	.025	-.031	.045	-.087	-.065
Police-initiated contacts	.085*	.049	.022	.048	.276*	.081	.146*
Police stops	-.019	-.068	.074	-.026	-.140	-.067	.009
Police misconduct	-.247**	-.152	-.295**	-.404**	-.501**	-.224*	-.464**
Age (26-39)	-.287*	-.358*	.038	-.175	-.009	-.135	-.335*
Age (40-59)	-.191	-.151	.215	-.251	.001	-.327	-.394**
Age (60+)	-.075	-.164	.507**	-.158	.464*	-.533**	-.357**
Non-white	-.310**	-.308**	-.325**	-.235**	-.368**	-.312**	-.358**
Male	.030	.012	-.053	-.013	-.073	-.080	.031
1-5 years in area	.028	-.104	.043	.065	.232	.103	.074
5+ years in area	.093	.004	.254	.146	.177	.022	.241*
Education	-.020	.003	.051	-.013	.040	-.047	-.030
Resident	-.184	-.132	-.069	-.001	.000	-.013	.004
Property victimization	-.256	-.796**	-.230	-.325	-.290	-.268	-.211
Violent victimization	-.332	.203	-.338	-.493	.222	-.636	-.316
Fear of crime	-.045	.005	-.020	-.013	.059	-.099	-.073
Perceived crime / disorder	-.088	-.040	-.042	-.136	-.188	-.345**	-.058
Constant	5.20**	5.12**	4.72**	5.49**	1.792**	5.70**	4.95**
F-statistic p-value	.000	.000	.000	.000	.000	.000	.000
R-squared	.221	.125	.147	.246	.299	.283	.300
N	501	501	500	495	470	497	478

* p< =.05, ** p< =.01. OLS estimates with standard errors clustered by hot spot. Age indicators are interpreted relative to respondents 18-25 years of age, and indicators for time in the location are interpreted relative to respondents living or working in the location for 6 months to 1 year.

Table 3. Attitudes towards police regressed on perceived police activities, controlling for respondents’ characteristics and experiences (“Willowdale” City).

Variable	Trust	Legitimacy	Reporting	Capability	Responsiveness	Satisfaction	Procedural Justice
Perceived patrol	.124**	.102	.109*	.193**	.021	.317**	.121*
Perceived enforcement	-.203**	-.106	-.213**	-.252**	-.051	-.399**	-.175*
Perceived community engagement	.142**	.115	.072	.123*	.638**	.224**	.180**
Citizen-initiated contacts	-.040	-.016	.116**	-.046	-.067	-.118*	-.030
Police-initiated contacts	.055	-.068	.002	.028	.078	.079	.051
Police stops	-.119	0.250*	-.071	-.039	-.165*	-.034	-.047
Police misconduct	-.430**	-.257*	-.153*	-.403**	-.350**	-.091	-.580**
Age (26-39)	-.048	-.131	.300	-.042	-.072	.001	-.066
Age (40-59)	.075	-.005	.656**	-.153	-.165	-.018	-.066
Age (60+)	.241	.095	.686**	.051	.106	.032	-.061
Non-white	-.08	-.268**	-.257**	-.055	-.554**	-.093	-.161
Male	-.044	.048	-.112	.012	-.111	.239**	.034
1-5 years in area	-.056	-.56	-.047	-.079	.137	-.078	.141
5+ years in area	.084	-.102	-.005	.032	.277	-.027	.251
Education	.002	.043	-.050	.007	.068	-.036	-.024
Resident	.027	-.120	-.112	-.052	-.545**	.053	-.067
Property victimization	.038	.066	.273	-.018	.481*	-.032	.090
Violent victimization	-.244	-.809**	-.426	-.044	.409	.037	-.238
Fear of crime	-.109*	-.032	.027	-.063	-.176*	-.132	-.166**
Perceived crime / disorder	-.040	.083	.011	-.062	-.134	-.037	-.056
Constant	4.88**	4.58**	3.97**	4.68**	2.31**	4.24**	4.78**
F-statistic p-value	.000	.000	.000	.000	.000	.000	.000
R-squared	.298	.091	.166	.226	.329	.183	.301
N	487	485	488	484	454	485	464

* p< =.05, ** p< =.01. OLS estimates with standard errors clustered by hot spot. Age indicators are interpreted relative to respondents 18-25 years of age, and indicators for time in the location are interpreted relative to respondents living or working in the location for 6 months to 1 year.

Multivariate analyses affirmed many of these general patterns. Controlling for the respondents’ background characteristics and experiences, higher levels of perceived patrol and community engagement were again linked to more positive views of police on several dimensions in both cities (Tables 2 and 3). Most notably, higher levels of perceived patrol were significantly associated with higher scores on the trust, capability, procedural justice, and satisfaction measures in both cities. Perceived patrol was also linked to higher levels of police legitimacy in Avalon and willingness to report in Willowdale. Similarly, higher levels of perceived community engagement were significantly associated with higher scores on the capability, responsiveness, and procedural justice measures in both cities as well as higher scores on the trust and satisfaction measures in Willowdale. Perceived patrol and community engagement had mostly positive relationships with views towards police even when not statistically significant, and they had no significant negative relationships with the attitudinal measures.

In contrast, higher levels of perceived enforcement were generally linked to more negative views of police across cities in the multivariate analyses. These associations were statistically significant for perceived capability and procedural justice in both cities and for the measures of trust, reporting, and satisfaction in Willowdale. Other non-significant coefficients for perceived enforcement were generally negative, and this measure had no significant positive relationships with views of police in either city.

In models combining respondents across cities (Table 4), patrol visibility was significantly and positively related to all of the attitudinal measures except police responsiveness, community engagement was significantly and positively associated with all but the legitimacy and reporting measures, and enforcement was significantly and negatively related to all but the legitimacy and responsiveness measures. Further, the relationships between police activities and attitudes towards police did not vary significantly between White and Black respondents (Table 5), despite the more negative views of police held by non-white respondents in general (Tables 2–4). Standardized coefficients for the police activity measures generally varied between |0.1| and |0.2| in the multi-city models (Table 6), with the exceptions that they were larger for the relationships between community engagement and responsiveness |0.37| and between patrol and satisfaction |0.3|. However, no particular type of police activity stood out overall as being most closely associated with community views, as the coefficients were not consistently larger for any one activity measure.

Table 4. Attitudes towards police regressed on perceived police activities, controlling for respondents’ characteristics and experiences (“Avalon” and “Willowdale” responses combined).

Variable	Trust	Legitimacy	Reporting	Capability	Responsiveness	Satisfaction	Procedural Justice
Perceived patrol	.135**	.152**	.100*	.172**	-.003	.397**	.140**
Perceived enforcement	-.162**	-.062	-.155**	-.201**	.037	-.276**	-.176**
Perceived community engagement	.128**	.040	.002	.132**	.675**	.176**	.202**
Citizen-initiated contacts	-.042	-.040	.072*	-.039	-.020	-.102**	-.045
Police-initiated contacts	.065	-.023	.014	.035	.143*	.086	.086*
Police stops	-.091	.105	-.005	-.038	-.146*	-.060	-.030
Police misconduct	-.344**	-.199*	-.231**	-.399**	-.437**	-.188*	-.517**
Age (26-39)	-.160	-.270**	.209	-.102	-.072	-.044	-.183
Age (40-59)	-.048	-.111	.468**	-.194*	-.110	-.166	-.210*
Age (60+)	.088	-.051	.637**	-.054	.256	-.250	-.199
Non-white	-.204**	-.283**	-.265**	-.147**	-.460**	-.194**	-.263**
Male	-.002	.028	-.074	.003	-.093	.080	.030
1-5 years in area	-.009	-.079	.007	-.005	.178	-.013	.095
5+ years in area	.088	-.045	.132	.093	.237	-.009	.235*
Education	-.013	.030	.001	-.007	.058	-.052*	-.029
Resident	-.076	-.133	-.087	-.039	-.351**	.010	-.043
Property victimization	-.100	-.385*	-.050	-.179	.121	-.020	-.052
Violent victimization	-.221	-.444	-.275	-.313	.197	-.321	-.254
Fear of crime	-.075*	-.016	-.004	-.040	-.052	-.124*	-.114**
Perceived crime / disorder	-.065	.018	-.003	-.093**	-.160**	-.168**	-.062
Willowdale	.032	-.016	-.004	.029	-.047	.043	-.058
Constant	5.00**	5.05**	4.23**	5.14**	2.53**	5.06**	4.91**
F-statistic p-value	.000	.000	.000	.000	.000	.000	.000
R-squared	.244	.083	.135	.228	.290	.210	.290
N	988	986	988	979	924	982	942

* p< =.05, ** p< =.01. OLS estimates with standard errors clustered by hot spot. Age indicators are interpreted relative to respondents 18-25 years of age, and indicators for time in the location are interpreted relative to respondents living or working in the location for 6 months to 1 year.

Table 5. Tests for racial invariance in regressions of attitudes towards police on perceived police activities (“Avalon” and “Willowdale” responses combined).

Variable	Trust	Legitimacy	Reporting	Capability	Responsiveness	Satisfaction	Procedural Justice
Perceived patrol	.131*	.207**	.119	.137*	.005	.519**	.148*
Perceived enforcement	-.161*	-.111	-.174**	-.202*	-.130	-.272*	-.215**
Perceived community engagement	.137*	.113	-.075	.101	.648**	.229**	.191**
Perceived patrol * Black	.000	-.034	-.027	.043	-.035	-.139	-.005
Perceived enforcement * Black	-.004	.055	.034	.002	.190	-.003	.022
Perceived community engagement * Black	-.022	-.095	.090	.023	.009	-.042	.003

* p< =.05, ** p< =.01. OLS estimates with standard errors clustered by hot spot. Black and white respondents only. Models control for same respondent characteristics and experiences shown in Tables 3–5 (estimates not shown).

Table 6. Standardized coefficients from regressions of attitudes towards police on perceived police activities: key policing measures (“Avalon” and “Willowdale” responses combined).

Variable	Trust	Legitimacy	Reporting	Capability	Responsiveness	Satisfaction	Procedural Justice
Perceived patrol	.146	.131	.087	.172	n/s	.302	.130
Perceived enforcement	-.157	n/s	-.122	-.183	n/s	-.188	-.152
Perceived community engagement	.119	n/s	n/s	.115	.366	.115	.166
Police misconduct	-.241	-.112	-.229	-.258	-.170	-.090	-.321

Coefficients shown only for variables that were statistically significant at p< =.05 (n/s = not significant). Models control for same respondent characteristics and experiences shown in Tables 3–5 (estimates not shown).

A final note is that recent police contacts were not significantly related to most attitudes towards police. Further, the significant associations that did emerge for the contact indicators were inconsistent within and across cities.²⁰ In contrast, respondents’ perceptions of police misconduct were clearly linked to worse attitudes towards police in both cities. Police misconduct was significantly related to all of the attitudinal measures in the multi-city models (Table 4), and its standardized coefficients (which ranged from −.090 to −.321) were comparable to or larger than those of the police activity measures in several of the models (Table 6).²¹

Discussion

This survey study of hot spot residents and workers in two cities suggests that hot spot policing has the potential to improve or worsen community attitudes towards police, depending on the types of strategies police utilize. Community members in hot spots who see more frequent patrol and positive police-community interactions also exhibit more positive attitudes towards police along a number of dimensions (e.g., trust and confidence in police, views of police legitimacy, and perceptions of procedural justice). In contrast, those who witness higher levels of investigative and enforcement activity also have more negative views of police. These perceptions persist independently of respondents’ background characteristics, their recent contacts with police (both positive and adversarial), and their awareness of police misconduct. Further, similar patterns emerged in two cities and across White and Black respondents, which strengthens confidence in the findings.

This study was cross-sectional and does not provide a strong basis for conclusions about cause and effect. As such, the associations found in this study are only meant to inform discussions about the types of policing strategies and tactics that may be best to employ in hot spots. Notwithstanding this caveat, the results suggest that seeing more police presence in hot spots is reassuring to community members and may promote a sense that police are providing the community with the attention and service that it warrants (Hawdon et al., 2003). Seeing positive police-citizen interactions and engagement may instill a further sense that police are trying to serve the community in positive, collaborative, and respectful ways, consistent with the philosophy of community policing. On the other hand, witnessing higher levels of investigative and enforcement activity may create or intensify a sense that police are treating community members harshly and unfairly, while also perhaps fostering a belief that police have not been effective in preventing crime and disorder, thus necessitating heavier enforcement (e.g., Kochel, 2011; Rosenbaum, 2019).

Our study focused on two cities within the same region of one state. Caution should thus be exercised in generalizing the findings to other jurisdictions, especially ones that have higher crime rates and a more troubled history of policy-community relations. Our community surveys were conducted only a few years ago, thus providing insights into contemporary public attitudes towards police in the post-Ferguson era in the United States. However, the data were collected prior to the killing of George Floyd in 2020, which may have shifted attitudes towards police in a more negative direction in some communities. The COVID-19 pandemic may have also made some people more apprehensive about personal contact of any sort with police. On the other hand, concern about rising crime rates could be pushing public attitudes in the direction of wanting more police presence and action. Further studies of hot spot community views in the present context and in a wider range of jurisdictions might strengthen or weaken confidence about the generalizability of our findings.

Another caution is that the associations between perceived police activities and attitudes towards police in our survey were generally small to medium in magnitude. Community members’ views of police are undoubtedly shaped by many factors, and they may be strongly influenced by their long-term direct and vicarious experiences with police (e.g., Rosenbaum et al., 2005), which were not measured here. Nor is it clear from these results how much police activity of different sorts would be needed to change public attitudes to a substantial degree (positively or negatively). The null results of prior hot spot survey studies suggest that changing community attitudes in these locations may require considerable changes in police activity. Nevertheless, our study locations were not using formal hot spot policing programs at the time of our survey, which raises the possibility that responses might have differed under circumstances involving more intensive hot spot policing.

Limitations of our survey methodology should also be noted. Although our survey response rate compared favorably to those of other similar studies, our survey could have potentially missed some important segment of the local adult population in these locations. Given the size of the study hot spots, for example, the results may not as accurately reflect the views of those living or working at the most problematic street blocks and intersections within these locations. As with all self-report survey studies, survey participants may have suffered from recall and social desirability biases. Further, respondents’ interpretations of some police-community contacts as positive community engagement or enforcement-oriented activity could have been influenced by their preexisting views of police.²²

Despite these limitations, our study has clear and direct implications for police practice. The findings support hot spot policing strategies that emphasize regular, systematic patrol in hot spots, complemented by positive community engagement efforts and problem-solving work (e.g., see recommendations by Lum & Koper, 2017; also see Haberman et al., 2016). Such efforts, for example, can include: patrolling hot spots daily or several times per week for brief but sustained periods that improve visibility (like 15 minutes per visit—see Koper, 1995); increasing the quantity and quality of police-community contacts using strategies like foot patrol, citizen contact (door-to-door) patrols, attendance at community meetings and events, and procedural justice training; and institutionalizing regular and formal efforts to work with residents and workers on crime prevention and other problem-solving projects. Although further studies will be needed to show more definitively that these practices improve community views of police at hot spots, our results suggest that they have the potential to achieve such effects—and at the least should not worsen views of police—when implemented well and at levels sufficient to raise community awareness.

At the same time, our results suggest that enhanced enforcement activity in hot spots, such as expanded use of traffic and pedestrian stops, searches, and arrests for minor offending, should be used judiciously, preferably with some public civic education around the necessity for some of these policing activities. Strategies that carefully target known offenders in hot spots, for example, may provide effective ways of reducing crime without negatively impacting community perceptions (Groff et al., 2015; Ratcliffe et al., 2015). Likewise, when addressing disorder in hot spots, police should emphasize community problem-solving approaches to address physical and social disorder conditions rather than misdemeanor arrests, which are less effective (Braga et al., 2015) and have greater potential to exacerbate police-community tensions. If enhanced enforcement is needed to address particular problems, such as gun carrying in gun violence hot spots (Koper & Mayo-Wilson, 2012; Rosenfeld et al., 2014), our results imply that police should attempt to target their actions as carefully as possible on suspicious persons, activities, and locations (e.g., McGarrell et al., 2001) and consider notifying and consulting with community members about their strategies in advance through means such as community meetings and citizen contact patrols (McGarrell et al., 2001; Shaw, 1995). Otherwise, strategies emphasizing aggressive crackdown enforcement tactics (i.e., zero tolerance approaches) may have deleterious effects on police-community relations that offset any crime prevention benefits they may produce.

Though incidental to our primary focus, the strong links between perceptions of police misconduct and more negative views of police also have important practical implications, as they highlight the potential for police misconduct to offset the impacts of positive policing efforts. For one, they reinforce concerns about the potential problems of excessive enforcement approaches that may increase adversarial police-community encounters and possibly raise perceptions of police misconduct. Moreover, the misconduct findings underscore the value of implementing procedural justice and de-escalation training as ways to minimize use of force, arrests, and abusive police behavior in hot spot encounters (Engel et al., 2022; Owens et al., 2016; Weisburd et al., 2022). Further research and policy development efforts with technologies like body-worn cameras and early warning systems (e.g., see reviews by Lum et al., 2019; 2020; Shjarback, 2021) may also help police to implement hot spot policing strategies in ways that minimize abusive police behavior and adverse community reactions.

A final consideration is the relevance of our results to contemporary debates about defunding police agencies as a form of justice reform (e.g., Eaglin, 2021; Lum et al., 2021). A full discussion of these policy options and their pros and cons is beyond the scope of this paper. Yet, other recent survey findings show that most Americans, including most Black Americans, want police presence to remain the same or increase in their communities (Saad, 2020). Our study suggests that similar sentiments may prevail in locations where crime problems are most serious; hence, research and debate over defunding should give specific consideration as to how these policies might impact police presence, police-community engagement, and public perceptions in such locations.

In closing, this study adds to a very limited body of research examining community attitudes in crime hot spots, and it underscores the value of community surveys for refining hot spot policing approaches and tailoring them to local conditions (also see Haberman et al., 2016). We encourage survey researchers and funders to put more emphasis on studies that assess public attitudes in crime spots where public safety problems and police-community relations are most challenging. We also encourage researchers and police to expand their efforts to gauge community opinion and engage the public, leveraging internet and social media technologies that can potentially make it more feasible and less cost-prohibitive for police to monitor community sentiment in hot spots on an ongoing basis (e.g., Graziano et al., 2014; Grimmelikhuijsen & Meijer, 2015; Rosenbaum et al., 2015). Even in the current political environment, people in troubled locations desire police presence and positive engagement with their police.

Acknowledgments

The authors thank the participating police agencies (identified anonymously) for their cooperation in conducting this study. The authors also thank Jackie Sheridan and William Johnson for research assistance.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by grant 2017-R2-CX-0017 from the National Institute of Justice (U.S. Department of Justice) to NORC and George Mason University. The views expressed are those of the authors. The data for this study has been deposited with the National Archive of Criminal Justice Data (NACJD).

Notes on contributors

Christopher S. Koper

Christopher S. Koper, Ph.D., is a Professor in the Department of Criminology, Law and Society at George Mason University and the Principal Fellow of George Mason's Center for Evidence-Based Crime Policy. He specializes in issues related to policing, firearms policy, and program evaluation.

Bruce G. Taylor

Bruce G. Taylor, Ph.D. is a Senior Fellow at NORC at the University of Chicago and a criminologist studying the intersecting areas of violence, health and justice. His research in law enforcement includes work on officer safety, health, officer use-of-force and violence against officers and proactive policing strategies.

Weiwei Liu

Weiwei Liu, Ph.D. is a Senior Research Scientist at NORC at the University of Chicago. She is a Criminologist specializing in risk and protective factors of violence, firearm violence prevention, evidence-based policing, police officer use-of-force, and officer safety and wellness.

Xiaoyun Wu

Xiaoyun Wu, Ph.D., is a Research Data Scientist at the National Policing Institute. Her areas of interest include police technologies, public safety, and data science and development.

Notes

1 At the same time, focusing on hot spots can potentially reduce over-policing of lower risk areas (Weisburd, 2016).

2 Patterns in community views across these studies were also not consistently related to whether the interventions reduced crime and disorder.

3 Along similar lines, experiments in Tucson, Houston, and Cambridge (MA) showed that augmenting hot spot policing with procedural justice training for officers reduced arrests and improved some aspects of how community members viewed police (Weisburd et al., 2022). This suggests that an emphasis on procedural justice can enhance the benefits of hot spot policing or mitigate its negative consequences. These experiments compared different modes of hot spot policing (with and without procedural justice training), thus precluding an assessment of whether hot spot policing improved or worsened community attitudes in general (relative to conditions without hot spot policing).

4 All population statistics are based on figures for 2019 reported in the U.S. Census Bureau’s QuickFacts.

5 These calculations are based on figures reported in the Federal Bureau of Investigations’ Uniform Crime Reports for 2017 (https://ucr.fbi.gov/crime-in-the-u.s/2017/crime-in-the-u.s.-2017).

6 The study was approved by the Institutional Review Board of the National Opinion Research Center at the University of Chicago (IRB 00000967), which has a multiple project assurance with the U.S. Department of Health and Human Services for research involving human subjects.

7 The locations were larger than the types of micro places used in some hot spot studies (e.g., specific blocks or intersection areas), but they were within the range of area types and sizes commonly used in hot spot research and practice (see Braga et al., 2019; Eck, 2005; Koper, 2014), and they provided more precise areas of focus relative to neighborhoods and other police areal units (which is a key consideration in hot spot policing—see Braga et al., 2019: 6). We designed the hot spots to optimize operational areas and maximize base rates of crime for a long-term police intervention and study that were being planned for the locations.

8 The hot spots contained more than 21,000 addresses in Avalon and more than 36,000 in Willowdale.

9 For example, the hot spot community survey studies discussed above reported response or cooperation rates ranging from 9% to 61%, with most being less than 50% (Haberman et al., 2016; Ratcliffe et al., 2015; Kochel et al., 2015; Weisburd et al., 2011; 2021). Our response rate also compares favorably to the 50% average (approximately) which has been found in survey studies across a variety of fields including psychology, communications, and organizational research (Baruch & Holtom, 2008; Macias et al., 2008; Van Horn et al., 2009). However, there is no agreed upon standard for minimally acceptable response rates for surveys (Fowler, 2014) due in part to the fact that response rates are poor predictors of non-response bias (Groves & Peytcheva, 2008).

10 These items were taken from Rosenbaum et al. (2007) and Weisburd et al. (2010).

11 The police agencies in these cities employed foot patrols on an ad hoc and discretionary basis but did not have formal foot patrol programs.

12 See Table 1 for further description of these measures.

13 See Table 1 for further description of these additional police measures.

14 Depending on the measure, 11% to 24% of respondents reported recent police contact stemming from a police-initiated or citizen-initiated encounter of the sort described above.

15 Missing data for the items and scales used in the analysis was very low (almost always under 3%, as shown in Table 1). Given the rarity of missing values and the large samples available in each city, we treated missing data as random and did not impute values for missing data.

16 We considered estimating multi-level models that also incorporated characteristics of the hot spots. However, intraclass correlation coefficients (ICCs) estimated for the models were no greater than 10% (and nearly all were below 10%), indicating that hot spot characteristics contributed little or only very modestly to variation in the dependent measures. In practice, multilevel models offer few benefits when ICCs are smaller than 5% (Dyer et al., 2005). While ICCs at around 10% may be suggestive of a medium group-level variance (Hox et al., 2017), ICCs were below that level in most of our models (particularly in Avalon). Given the small ICCs and our primary interest in estimating individual-level rather than group-level effects, we controlled for hot spot-level variation using clustered standard errors.

17 The standardized regression coefficients should be viewed cautiously, as the use of clustered data may affect estimates of the variables’ standard deviations. However, the key results from the multi-city models were not sensitive to the use of clustered errors, suggesting that the clustering effects were not substantial,

18 White and Black respondents accounted for 91%-93% of the respondents in each city; hence, we excluded other race and ethnic groups from these analyses.

19 The focus on respondents from urban hot spots from two similar cities also minimizes other contextual differences (e.g., area crime and disorder levels) that can cause variation in the effects of policing on attitudes (e.g., Reisig et al., 2021; Zahnow et al., 2021).

20 To illustrate, the police stop and citizen-initiated contact measures had no significant relationships to attitudes in Avalon and varying positive and negative associations with a few attitudes in Willowdale. The police-initiated positive contact measure had positive relationships to some attitudes in Avalon but not in Willowdale. Differences in the nature and quality of these contacts may help to explain the inconsistent results.

21 Other model covariates had relationships with police-related attitudes that were generally inconsistent across outcomes and cities. To varying degrees, attitudes towards police were associated with race, age, gender, fear of crime, residential status, time living or working in the location, crime victimization, and perceived crime and disorder. The most consistent of these findings was that non-whites had more negative views of police. Respondents also tended to have worse views of police, particularly in the multi-city models, when they perceived higher levels of crime and disorder and, to a lesser extent, when they had greater fear of crime.

22 Our analysis assumes that people can reasonably differentiate between different types of police-community contacts based on visual and verbal cues. To some degree, however, people may interpret ambiguous police-community interactions as positive or adversarial based on whether they have positive or negative views of police, respectively. Although our model covariates may control for some of these tendencies, this form of reverse causality could account for some of the associations found in this study between different types of police-civilian contacts and views of police (this issue would not seem to apply to analyses of the patrol visibility measure). Interpreted in this manner, the policy implications are arguably the same (i.e., positive community engagement may improve perceptions of police while overuse of adversarial, enforcement actions may worsen them). However, the reverse causality interpretation implies that changing views about police in a positive manner will be more challenging, as police will have to overcome the proclivity of some community members to interpret their actions negatively.

Appendix Table A1.

Measures of attitudes towards police

Table

Measure	Items (abbreviated)	α	Source(s)
Trust in police scale (“trust”)	Proud of the police Afraid of the police Trusting of the police Uneasy around the police Feeling confident in the abilities of the police Responses: 1 = strongly disagree to 5 = strongly agree	0.79	Rosenbaum et al. (2007)
Police legitimacy scale (“legitimacy”)	Should do what the police tell me to, even if I disagree Should accept decisions made by area police, even if I do not understand the reasons for their decisions Should obey police directives because that is the proper or right thing to do Responses: 1 = strongly disagree to 5 = strongly agree	0.81	Kochel et al. (2015)
Willingness to report to police scale (“reporting”)	A crime committed against you A non-crime problem, e.g., excessive noise, on your block Suspicious activity on your block Information that may help solve a crime Responses: 1 = very unlikely to 5 = very likely	0.72	Weisburd et al. (2010); Kochel et al. (2015)
Police capability scale (“capability”)	Well-trained Knowledgeable about police procedures Respond quickly to emergency calls Satisfied with overall performance Capable of maintaining order in the streets Capable of responding to emergencies Capable of solving crimes Capable of preventing future crime Responses: 1 = strongly disagree to 5 = strongly agree	0.94	Rosenbaum et al. (2007); Kochel et al. (2015)
Police responsiveness scale (“responsive”)	Sharing information with residents Taking residents’ suggestions on local crime and non-crime related problems Working with residents to solve local problems Dealing with problems that really concern residents Providing crime prevention tips to residents Available to talk to residents Responses: 1 = very poor job to 5 = very good job	0.98	Rosenbaum et al. (2007)
Procedural justice scale (“procedural justice”)	Treat citizens with respect and professionalism Neutral and transparent Courteous to residents Provide the same quality of service to all citizens in the area Take the time to listen to people Fair and impartial Treat all people with dignity and respect Explain their actions to people Respectful of the youth Respectful of minorities Responses: 1 = strongly disagree to 5 = strongly agree	0.97	Rosenbaum et al. (2007); Kochel et al. (2015)
Satisfaction with police item (“satisfaction”)	In general, how satisfied are you with the frequency of police presence? Responses: 1 = very dissatisfied to 5 = very satisfied	N/A	N/A

α = Cronbach’s alpha estimated with full multi-city sample (maximum n = 1,082). All scales and items were measured on a 1 to 5 scale with higher scores indicating more positive views (items were reverse-coded as needed). Some scales combine items from multiple listed sources.

Appendix Table A2.

Correlations between perceived police activities and attitudes towards police (“Avalon” city)

Table

	Satisfaction	Trust	Legitimacy	Reporting	Capability	Responsive	Procedural justice	Perceived patrol	Perceived enforcement	Perceived community engagement
Satisfaction	1
Trust	0.4184	1
Legitimacy	0.2873	0.3292	1
Reporting	0.2076	0.3343	0.2442	1
Capability	0.511	0.6753	0.4044	0.3529	1
Responsive	0.1136	0.2627	0.1002	0.1861	0.2891	1
Procedural justice	0.4891	0.6351	0.4151	0.3328	0.7231	0.3695	1
Perceived patrol	0.2599	0.1299	0.1067	−0.0077	0.0977	0.1967	0.1596	1
Perceived enforcement	-0.0976	-0.0964	−0.0702	-0.169	-0.151	0.1942	-0.1002	0.396	1
Perceived community engagement	0.1215	0.0956	0.019	−0.0151	0.0848	0.4305	0.1732	0.4433	0.477	1

Bolded coefficients statistically significant at p< =.05. Maximum n = 562.

Appendix Table A3.

Correlations between perceived police activities and attitudes towards police (“Willowdale” city)

Table

	Satisfaction	Trust	Legitimacy	Reporting	Capability	Responsive	Procedural justice	Perceived patrol	Perceived enforcement	Perceived community engagement
Satisfaction	1
Trust	0.4294	1
Legitimacy	0.2042	0.284	1
Reporting	0.2027	0.4213	0.1918	1
Capability	0.477	0.6983	0.3785	0.3044	1
Responsive	0.1554	0.3031	0.0945	0.1268	0.2858	1
Procedural justice	0.4142	0.6536	0.3237	0.3219	0.7184	0.3901	1
Perceived patrol	0.1648	0.0092	0.0633	0.0466	0.0426	0.0736	0.0186	1
Perceived enforcement	-0.1735	-0.2518	−0.071	-0.1025	-0.2568	0.0347	-0.2234	0.4308	1
Perceived community engagement	0.1132	0.0907	0.0623	0.0666	0.0496	0.369	0.115	0.3927	0.4975	1

Bolded coefficients statistically significant at p< =.05. Maximum n = 520.