The Efficacy of Foot Patrol in Violent Places
Technical Appendix
Research Design
The inability to randomize locations by target and control represents a limitation to this study. The major concern in the absence of randomization is selection bias, however, examinations of pre-existing crime and call for service conditions indicate beats did not differ significantly across conditions. Drawing on three years of data before the start of the project, independent sample t-tests revealed nonsignificant differences between treatment and control for armed robberies (t = 0.919, p > 0.05), strong armed robberies (t = 0.459, p > 0.05) offenses; and robberies (t = 0.749, p > 0.5), burglaries (t = 1.026, p > 0.05) calls for service. Non-significant differences across areas were also observed for homicides, aggravated assaults offenses, and calls for service for aggravated assaults, stealing, auto thefts, and overall calls for service for 2008, 2009 and 2010. However notably target areas experience significantly higher calls for service for property damage. Given the totality of crime and call for service data we concluded the target and control areas were comparable given our focus on aggravated assaults and robberies.
Treatment Integrity
We examined treatment integrity prior toassessingtheresultsof theoutcomeanalysis.Inordertodeterminewhether foot patrolreducedviolentcrimeinmicro-places,wefirstneededtoknow thatfootpatrolwas, in fact, being executedinthetargetareas. Although it was not possible for the embedded researchstaff toobserve directlyallfacetsof the treatment during the study,stepsweretakentomonitorprogramintegrity. These steps included observing officers in the field, monitoring internal workload logs submitted by officers, and reviewing official police data.
The officerswerenottold exactlywhattodo whileon footpatrol, rather, theywere instructedto spendthe majorityof theirdiscretionarytimewithintheboundariesof theirassignedfootbeats. For the duration of their foot beat assignments,centraldispatch did not include them incallsforservice. Members of the research teamwere assigned to differentfootbeats with differentofficersthroughouttheintervention.Theyconductedaseries of"walk-alongs," takingfieldnotesongeneralactivities. Those notes wereusedtodetermine that officers (a) were actually engaginginfootpatrol and (b) were operating within the designated areas.
Qualitativeinformation was gatheredfrom the officers. Theyremarkedthattheyunderstood theirassignmentsand indicatedtoobserversthat they had to " “stay in their area”. Observations and feedback from the officers suggested a high degree of boundary adherence.Although maintaining boundary adherence can be a concern in foot patrol (Sorg, Wood, Groff and Ratcliffe, 2014), observations suggested that these new, probationary officers, many of whom had been on the hiring list for KCPD for several years, feared that not adhering to the integrity of the assignment as ordered by senior police leadership would jeopardize their jobs. As a result, the officers assigned to foot beats spent the overwhelming majority of their time in their respective beats. We utilize results presented here as a measure of boundary adherence. Several officers assigned to foot patrol lamentedthattheyweremissing out on the moreexcitingcallsforservice,but thatthey had to remain in their foot beat. It is worth noting that foot beat officers remarked that more senior patrol officers sympathized with the new officers who were assigned to foot patrol. Themes emerged that suggested that senior officers encouraged foot beat officers to "just do their 90 days" so that they could "get in a car and do real police work" (e.g., calls for service). A foot patrol assignment for rookie officers had never before been implemented in this police department, and it was not a normal part of officer culture or professional development, therefore senior officers viewed the entire approach with a great deal of skepticism.
Datasuggested officers on foot patrol weremoreactive and accessibletocitizens, andthedosageof policepresence increased inthefootbeatsduringthe90-day implementationperiod.An examinationof thenumberof offenses recorded(excludingaggravated assaultsandrobberies)duringtheimplementationperiodrevealedasignificant increasein such offensesinthetreatmentarea (see table 1). An averageof 52.7 offenseswerereportedinthetwo-weektimeperiodleadingup toimplementationinthe treatmentarea,butduringthedeploymentof foot patroltheaveragenumberofrecorded offensesspikedto75.57.At theconclusionoftheimplementation,thisratedroppedsuddenly to47.3 pertwo-weektimeperiod.No suchchangeinreportedoffenseswasobservedinthe controlareas(36.8,39.7 and 33.1). Theincreasedpolicepresenceandavailabilityto citizens inthetreatmentareas,sincethe officerswereon foot, may explaintheincrease inreported offenses. Thischangebetween the pre-implementation period andduring the treatment period is statisticallysignificant,and is especiallynotablewhenconsideringthe number of reportedoffenses (otherthanaggravatedassaultsandrobberies)inthetargetand controlcatchmentareas.In the targetcatchmentarea,anaverageof 55.9 offenseswas reportedpriortotreatment,65.7 duringtreatment,and49.7 post-treatment.
---- Insert Table 1 about here ----
Reports filed for offenses in the treatment area increased significantly during the foot patrol project, while no such change was observed in the control or catchment areas. We attribute this to the saturation of foot patrol officers in the treatment area, which provided more opportunities to proactively or reactively record offenses—there were more personnel available for reporting, rather than an outcome of the strategy. Most of those reports were for minor offenses such as disorder or nuisances. This observation provides further evidence that foot beats experienced an increase in officer dosage and activity during the treatment period.
Repeated measure t-tests
Bivariate analyses consisting of repeatedmeasures one-tail directionalt-testswereselectedduetomultiplemeasurements that were takenfor eacharea,betweenthreedifferent timeperiods.Repeatedmeasurest-testswereselectedbecausetheeffectivenessof the treatment (footpatrol)wasmeasuredbycomparingwhetherthemeannumberofaggravated assaultsandrobberieswasreduced across the pre-treatment,during-treatment,andpost-treatment periods, withinbothtargetandtargetcatchment areas.Additionally,the number of observationswas thesamefor eachtypeof area.A left-tailed alternativehypothesiswasusedduetotheresearcherexpectingtreatmenttoreducethenumberof reportedaggravatedassaultsandrobberies,basedon the treatment-related reductionfound byRatcliffeetal.(2011).
Initial analyses involved conducting a series of repeated measure t-tests across the various research conditions over time. Of particular interest is whether the average number of aggravated assaults and robberies changed in the targeted (treatment) areas from before treatment (T0) to during treatment (T1), and whether any observed change persisted after foot beats were removed post-treatment (T2). Table 2 displays the average violent crimes per biweek period for the targeted foot patrol area, control areas and the catchment/displacement area.
The estimates in Table 2 represent the average number of aggravated assaults and robberies in each of the four target areas over time. As can be seen from Model A, the target areas experienced a reduction from an average of 1.85 violent crimes per biweek before treatment to 1.36 violent crimes per biweek during treatment. This 26.4% reduction represents a statistically significant change in violent crime in the time period during which officers were deployed in the four foot beat areas (t=-2.63, p< .05). This suggests that the use of foot patrol was effective in reducing aggravated assaults and robberies in the micro-place areas. Model B further indicates that once foot patrol treatment was removed, the average number of violent crimes returned to pre-treatment levels (t=-2.03, p< .05). Next, models were estimated for violent crime during the same period for the control areas. Of particular importance is the examination of aggravated assault and robbery rates in the control areas relative to the changes in target areas. Model C compares T0 with T1, revealing that violent crime increased only slightly and non-significantly during the time periods that foot patrol was being implemented in the target areas. Further, Model D shows that violent crime declined slightly from T1 to T2; this change also was not statistically significant. These observations indicated that the reductions in crime prevention detected in Model A were not experienced in comparable control areas, supporting the conclusion that the foot patrol treatment resulted in at least short-term crime prevention benefits.
---- Insert Table 2 about here ----
Violent crime rates were then examined for each of the 650' catchment areas surrounding the foot beat areas to determine whether crime was being spatially displaced to those surrounding areas. Model E indicates that robberies and aggravated assaults increased modestly in the catchment areas from T0 to T1 (from 1.67 to 1.75 per biweek); however, that increase was not statistically significant (t=0.18, p > .05). Further, when comparing T1 with T2, Model F shows that violent crime decreased between treatment and post-treatment periods (1.75 to 1.32); that change also was not statistically significant. In summary, the analyses indicated that foot patrol was associated with significant declines in violent crimes between the pre-treatment and treatment periods, and then crime prevention benefits dissipated once the foot patrols were removed in the post-treatment period. Bivarate analysis indicates that declines were achieved with neither spatial displacement nor diffusion of benefits of violent crime to contiguous areas.
Power Analysis
Given the four treatment and four control areas, a power analysis was conducted to ensure sufficient power to detect an effect. The analysis showed that this approach will detect a treatment effect size of 0.7 or more 79.5% of the time. Once the treatment period is broken up into two sub-periods, the statistical power decreases. The PSAR approach will detect an effect size of 0.7 in the 30-day treatment period only 38.98% of the time. It will detect an effect size of 1 67.07% of the time. We see that we sacrifice a great deal of statistical power by breaking the treatment period into two segments. However, even with low statistical power, we are able to detect a significant effect.
Future Research
There are a number of additional lines of inquiry future research should consider regarding the efficacy and sustainability of foot patrol that could benefit both the academic and practitioner communities. First, deliberate attention should address the influence of dosage and duration of foot patrol implementation. The current study examined pairs of officers across two shifts per day for 90 days. Analyses presented suggest crime prevention benefits within violent hot spots decay after about 30 days. This appears to be a reasonable starting point for future research examining foot patrol duration, and strategies should be developed to identify the optimal duration (along the same lines as Koper (1995) or Telp, Mitchell & Weisburd (2014) for hot spots). Relatedly, it is unclear whether two-officer teams, twice per day are ideal. Therefore, ideal dosage levels also are unclear. Future research should examine the effectiveness of varying deployments (e.g., 1 vs. 2 officer assignments; one vs. two shifts per day, etc.) so that policy makers can better understand the return on foot patrol investment. These questions could, and should, be addressed through a randomized experimental design. These are also critical practical questions because foot patrol represents a significant investment in personnel resources - nearly 8,200 hours in the current study. Specification of the ideal dosage and duration is necessary to accomplish effective crime control in the most efficient manner.
Second, future research should more fully examine other dimensions of crime displacement. The current study failed to find evidence of harmful immediate spatial displacement from foot patrol deployment (c.f., Andresen and Malleson 2014; Piza and O’Hara 2012). However, displacement can also include temporal, target, tactical and offense displacement (Guerette 2009), or spatial displacement across greater distances (Weisburd and Telp 2014), and these dimensions of displacement have yet to be fully unpacked within the context of foot patrol. A more comprehensive examination of the variety of displacement dimensions is deserved.
Third, further examination of officers’ activities is deserved. As Ratcliffe et al. (2011) and Groff, Johnson, Ratcliffe and Wood (2013) noted, the impact of foot patrol varied by officers’ focus (e.g., problem oriented policing, mere presence, etc.). Officers in the current study were given only general directives of what they should be doing while on foot patrol, so they enjoyed a considerable amount of discretion regarding their workload. Varying and measuring officers’ activities (as was done in Philadelphia) will contribute to understanding best practices for foot patrol assignments in violent crime areas.
Finally, understanding the subjective indicators of impact and officers’ perceptions of foot patrol remains important. Foundational foot patrol research strongly suggested that citizens’ fear of crime and attitudes toward the police were greatly impacted by foot patrol. Connections between the police and the public are important considerations for promoting police legitimacy and procedural fairness (President’s Task Force on 21st Century Policing 2015), and should be included as an important consideration when evaluating the efficacy of foot patrol. While it is possible that foot patrol can accomplish the dual goals of violent crime prevention and enhancing police legitimacy simultaneously, this remains unknown.
Relatedly, it is unclear exactly how receptive officers are to foot patrol. Subjective information gathered as part of the process evaluation within the current study suggested officers were at best ‘tolerant’ of their foot patrol assignment. Foot patrol was originally conceived in Kansas City to accomplish two general goals – one was to impact violent crime, but a secondary goal was to influence rookie officers’ professional development. It was hypothesized that the unique introduction to policing that foot patrol afforded new officers would impact their orientation toward the job making them more community oriented than law enforcement oriented. This empirical question remains unanswered. Changing the focus and activities of police organizations is an uphill battle when the proposed innovations are counter to the prevailing organizational culture. Over twenty years ago Roberg (1994) noted that officers have been “recruited, trained and socialized in the traditional law enforcement orientation” (p. 251-252) and, therefore, may have a stake in conformity and preserving the status quo. If officers view foot patrol as an ancillary public-relations effort that has no true crime prevention benefit, then it is likely to be rejected among officers who are assigned to perform such tasks. Foot patrol risks running counter to traditional officer and organizational culture, which could further impede sustainability, irrespective of the strategy’s effectiveness.
References
Andresen, M. A. & Malleson, N. (2014). Police foot patrol and crime displacement: A local analysis. Journal of Contemporary Criminal Justice, 30, 186-199.
Groff, E. R., Johnson, L., Ratcliffe, J. H., & Wood, J. (2013). Exploring the relationship between foot and car patrol in violent crime areas. Policing: An International Journal of Police Strategies and Management, 36, 119-139.
Guerette, R. T. (2009). Analyzing crime displacement and diffusion. Problem-Oriented Guides for Police Problem-Solving Tools Series No. 10. Washington, DC: Office of Community Oriented Policing Services.
Koper, C. S. (1995). Just enough police presence: Reducing crime and disorderly behavior by optimizing patrol time in crime hot spots.Justice Quarterly, 12, 649-672.
Piza, E. L.,B. A. O’Hara. (2012). Saturation foot-patrol in a high-violence area: A quasi-experiment evaluation.Justice Quarterly, 31, 693-718.
President’s Task Force on 21st Century Policing. (2015). Final report of the President’s task force on 21stcentury policing. Washington, DC: Office of Community Oriented Policing Services.
Ratcliffe, J. H., Taniguchi, T., Groff, E. R. & Wood, J. D. (2011). The Philadelphia foot patrol experiment: A randomized controlled trial of police patrol effectiveness in violent crime hotspots." Criminology, 49, 795-831.
Roberg, R. R. (1994). Can Today’s Police Organizations Effectively Implement Community Policing? In D. P. Rosenbaum (Ed.),The challenge of community policing: Testing the promises (pp. 249-257). Thousand Oaks, CA: Sage.
Telp, C. W., Mitchell R. J., & Weisburd, D. (2014). How much time should the police spend at crime hot spots? Answers from a police agency directed randomized field trial in Sacramento, California. Justice Quarterly, 31, 905-933.
Weisburd, D. & Telp, C. W. (2014). Hot spots policing: What we know and what we need to know. Journal of Contemporary Criminal Justice, 30, 200-220.
1