Skip to main content
SearchLoginLogin or Signup

Patrolling the largest drug market on the eastern seaboard: A synthetic control analysis on the impact of a police bicycle unit

Published onJul 20, 2023
Patrolling the largest drug market on the eastern seaboard: A synthetic control analysis on the impact of a police bicycle unit


Research Summary: This study employed a microsynthetic control method to evaluate the impact of the newly introduced bicycle patrol unit in the Kensington neighborhood of Philadelphia, which is well-known as a major drug market. The findings reveal that the bicycle patrol unit led to a notable reduction in social disorder crimes, which was one of the primary objectives of the patrol officers. However, it also resulted in a significant increase in the number of narcotic crimes, violent person crimes, and shooting offenses. The amount of total crimes and property crimes remained unchanged. Policy Implications: Bicycle patrols can effectively reduce street-level disorder and increase police efforts to arrest drug offenders and remove drugs from the streets. However, the introduction of this police activity may disrupt the normal operation of drug markets, which can lead to increased violence from instability in the street-level drug business. Therefore, departments should conduct detailed crime analyses and gather intelligence alongside directed patrols to better understand and respond to the potential consequences of their interventions.

Daniel S. Lawrence, PhD, ORC ID: 0000-0002-0777-2965

CNA Corporation

Center for Justice Research and Innovation

Arlington, VA

Author Bios:

Daniel S. Lawrence, PhD is a research scientist in the Center for Justice Research and Innovation at the CNA Corporation. He previously worked at other non-profit research institutes, including the Urban Institute and RTI International. His research interests include criminal justice technology, police legitimacy and procedural justice, and community policing. He received his MA and PhD from the University of Illinois at Chicago.


Direct correspondence to Daniel S. Lawrence, CNA Corporation, 3003 Washington Blvd., Arlington, VA, 22201 (email: [email protected])


I would like to express my gratitude to the following individuals whose invaluable contributions have helped shape this paper into its final form: Jerry Ratcliffe, whose insightful comments, edits, and critiques greatly enhanced the quality of this work; Scott Decker, for his review of the second version of the paper; and the editors at Criminology and Public Policy, as well as the reviewers of the manuscript, for their thoughtful and constructive feedback that led to significant improvements in the paper.


Police, Bicycle patrol, drug market, Kensington, Philadelphia


In the post-Ferguson, post-George Floyd world, police departments in the United States are faced with two pressing needs. First is the imperative to improve public trust and police legitimacy. At the same time, there has also been a substantial increase in violent crime, with murder increasing in 2020 by the largest one-year increase in history (Asher, 2022). In urban environments, police bicycle patrols have the potential, in theory, to help address both concerns. The introduction of cars as the central mode of patrol shifted policing from preventative to reactive (Bekiempis, 2015), and officers—driving down the road at 40 mph locked in boxes of metal and glass—lost contact with the public. The re-establishment of community policing requires a “reorientation of patrol in order to facilitate two-way communication between police and the public” (Skogan & Harnett, 1997, p. 428). The openness and slower speed of bicycle patrol is one reorientation with potential to increase the opportunities for interaction with the public. Equally, while foot patrol has the potential to reduce violence (Ratcliffe et al., 2011), more mobile units inevitably respond to the majority of calls for service (Groff et al., 2013). Bicycle patrol has the potential to bridge the mobility gap that can enhance community contact over cars, but still respond to calls for service in a timelier manner than foot beats.

According to the most recently collected Law Enforcement Management and Administrative Statistics (LEMAS) that included an item on bicycle patrols, 33.4% of local law enforcement agencies in 2016 relied on bicycle patrols “as needed,” while another 6.1% reported using them on a regular basis (DOJ, 2020). This is a modest increase from 2007, when the LEMAS program reported roughly 27% of local law enforcement agencies used bicycle patrols on a regularly scheduled basis (DOJ, 2011). Many environments and climates will preclude patrol on bicycle, so these numbers are hard to contextualize; however, given about a third of police departments may use bicycle patrols at some point, the lack of research around their effectiveness is surprising. Here the author reviews the existing literature before reporting on a quantitative study of the impact of bicycle patrols on various crime types in the city of Philadelphia, Pennsylvania. The author concludes with a summary of the findings and discussion of their potential policy implications.

Effectiveness of Police Bicycle Patrols

Little research has directly examined the impact of police bicycle patrols on reducing crime, instead focusing on their contribution to a community and social support role. Bicycle patrols have predominately been promoted to improve police-community relations because officers are more accessible on a bicycle than in their vehicle. What literature exists suggests they are better able to respond to community-initiated requests and calls for service, as well as conduct informal, personal, and friendly conversations when on a bicycle patrol compared to driving around a beat. These types of activities align with the President’s Task Force on 21st Century Policing (2015) report pillar on community policing and crime reduction. Specifically, bicycle patrols align with the recommendation to use “patrol deployment practices [that] allow sufficient time for patrol officers to participate in problem solving and community engagement activities.” For example, Menton’s (2008) examination of 1,105 police-community contacts in five cities found that bicycle patrols resulted in over twice as many contacts with the public (both in the number of contacts and the amount of community members involved) than colleagues in cars, as well as being rated more positively than car patrol contacts. He also observed a greater willingness by community members to stop or approach officers on a bicycle and provide information on possible illicit activity.

More recently, Sytsma and Piza (2018) surveyed 178 officers from the Toronto Police Service. They also found that bicycle patrols, which were part of the department’s community response unit, were associated with more contacts with the public and higher levels of proactive policing, although they did not find differences in the quality of contact between the types of patrol. In addition to those domains, Sytsma and Piza examined measures of officers’ self-reported perceptions of crime control performance that included four items: perceptions of how well they do in their job to supply information to public to reduce crime, enforce the laws, ensure citizen safety, and prevent crime. Across all four of these measures, bicycle patrol officers were more likely to be rated higher compared to officers who patrolled by motor vehicles.

In a more direct—although dated—examination of the impact bicycle patrols may have on crime, Barclay et al. (1996) assessed a bicycle-mounted security patrol that was implemented in a commuter parking garage in Vancouver that experienced high levels of motor vehicle thefts. For this experiment, four non-law enforcement security guards irregularly patrolled a garage during workday hours. Prior to the patrols, the treatment garage accounted for 30% of motor vehicle thefts in garages among the city’s four major hot spots. During the patrols, this percentage dropped to 4% and only increased to 13% in the months after the patrols—a substantially lower contribution than before the patrols. The author found no conclusive evidence of temporal or spatial displacement associated with this reduction in motor vehicle thefts.

It is worth noting that none of the above cited literature relied on experimental or quasi-experimental designs to assess the impact of bicycle patrol units. The methodology that Barclay et al. (1996) used was a direct comparison between the one commuter parking garage with a bicycle patrol compared to others without. Menton’s (2008) approach used a participant observation design through ride alongs with bicycle and automobile patrols. And finally, Sytsma & Piza (2018) applied a purposive sampling technique to select police divisions exhibiting both the highest rates of bicycle use and lowest rates and then surveyed officers to compare. The lack of an experimental, or even quasi-experimental, approach makes it hard to attribute the reported results to the bicycle patrols as opposed to other factors, somewhat diminishing the confidence in the above reported results. More rigorous designs are severely needed to assess the impact that bicycle patrol units may have on criminal activities.

Bicycle patrol as a crime control mechanism

Many studies have found success in targeted, place-based strategies in locations with high crime. Braga et al.’s (2019) systematic review to assess the impact of hot spots policing on crime found that 57 of 73 tests (78.1%) from 62 studies favored treatment conditions over control conditions. While the overall effect size for these studies was small, at .132, this increased for specific crime types; notably drug offenses (.244) and disorder offenses (.161). Focused police interventions, such as directed patrols, proactive arrests, and problem-oriented policing, have the potential to yield substantial crime prevention gains at crime hot spots; however, how place-based bicycle patrol units fit into this crime reduction space is sorely missing.

But how could a bicycle patrol unit prevent crime? When discussing the effect of the criminal justice system on crime control, criminologists tend to differentiate between incarceration, general deterrence—a response to the general threat of punishment for crime—and specific deterrence, a reference to the ‘aftermath’ of having experienced punitive action (Chalfin & McCrary, 2017; Nagin, 2013). The fundamental concept with deterrence theory is people are influenced by the certainty, severity, and celerity of punishment. If it is true, as Nagin argues, that “the certainty effect stems primarily from police functioning in their official guardian role rather than in their apprehension agent role” then street crime may be deterred by increased police visibility rather than increased police action (Nagin, 2013, p. 253). Likewise, the rational choice theory considers that a potential offender's decision to commit a crime may factor their risk of being detected (Cornish & Clarke, 1987). It could be expected that street crimes would decrease as officers on bicycles are able to quickly and silently catch offenders in the act of criminal activity. Lastly, according to the routine activity theory, the presence of a motivated offender and a suitable target, combined with the absence of a capable guardian, can significantly increase the likelihood of crime occurring (Felson, 2016). Bicycle patrol officers who have become a common occurrence in a local geospatial area can act as the capable guardians to help reduce crime.

The rapidity by which cars speed down a street, combined with the isolation created by the vehicle’s glass and metal constraints, likely negates much increased sense of visibility and increased perceived risk to offenders from police vehicle patrols. This may explain the results from the Kansas City Preventative Patrol Experiment (Kelling et al., 1974). At the other end of the speed spectrum, static patrols can create “‘bubbles’ of relative safety near officer locations” (Lawton et al., 2005, p. 446), and foot patrol has been shown to be effective at reducing violence (Ratcliffe et al., 2011); however these no/slow speed intervention wins come at the cost of limited spatial effect. Bicycle patrols may function as an ideal intermediate step, providing improved contact with the public unhindered by a vehicle’s glass or engine noise, as well as improved range over walking. In the right kind of urban environment, bicycles may also allow more rapid access to less accessible locations, improved pursuit capacity (over foot patrol), and better situational awareness of the environment (over vehicle patrol).

This view is shared by Philadelphia police officers who responded to a 2010 survey. When asked what tactics would be effective at combating violence, 55% thought foot patrol would be effective or very effective but 73% thought that bicycle patrol would be effective or very effective (Ratcliffe & Sorg, 2017). Therefore, while there are hypothetical mechanisms that suggest the potential to affect crime and disorder, the majority of previous studies has focused on community relations and few researchers have directly explored any crime or disorder impact. Consequently, this study aims to better understand how a bicycle patrol unit implemented in a well-established major drug market may affect crime levels. To do this, a synthetic control method approach was employed to best assess the effects of the introduction of a Philadelphia police substation dedicated to bicycle patrols.

Current Study

The current study explores the impact that a bicycle patrol unit may have on crime levels through a synthetic control method analysis of the Philadelphia Police Department’s bicycle patrol substation in the Kensington neighborhood. Philadelphia is the largest city in Pennsylvania, with a diverse population of 1.6 million in 2020, including 34% white residents, 38% Black residents, and 15% Hispanic residents (U.S. Census Bureau, 2022). The city has historically been plagued by high rates of crime. In 2020, Philadelphia’s violent crime was 956.92 per 100,000 people, making it the 18th most violent city with a population of 300,000 or more in the country: 2.4 times higher than the national average of 398.5 violent crimes per 100,000 people (Open Data Philly, 2022; FBI, 2022; U.S. Census Bureau, 2022).

The area under examination in this study is a 0.36 square mile subsection of the Kensington neighborhood, notoriously known as the largest open-air drug market in eastern United States (Johnson et al., 2020). Police sources cited by the city’s newspaper claim that open-air drug dealing extends to more than 80 of the neighborhood’s block corners (Newall, 2021). Crime analyses by the author found a disproportionate amount of crime happens in this area, which covers only 0.25 percent of the city but accounted for 2.47 of total crime, including 3.23 percent of the city’s violent person crimes, 4.23 percent of its homicides, 5.76 percent of its shootings, and 12.57 percent of its drug crimes in 2020.

This area has not always faced such high levels of crime as described by local Philadelphia police officers who patrol the district. They advised in informal conversations that the past four to five years have been exceeding worse than times before that. They note four major events that increased criminal activities in the area. First, Conrail—the city’s train rail operator—worked with the city in 2017 to improve subway lines that resulted in pushing 400-500 people experiencing homelessness out from underground living settlements back onto the street (Wood & Farr, 2017). Ratcliffe and Wight (2022) report a transit officer who works in Kensington saying “At least when the homeless encampments were in the tracks near Lehigh, it was tucked away. Then they cleared them…And now everyone’s out here on the street.” Second, around the same time, a documentary was released practically advertising where heroin could be purchased, resulting in a huge influx of “drug tourism;” with many people staying once they found how easy it was to get drugs. Third, the increase in accessibility of the opioid overdose treatment Narcan (Naloxone) made it much easier for dealers and users to rapidly reverse overdoses. Social workers and users have been known to plant Narcan and overdose kits strategically in the neighborhood, so that those near people who are overdosing can quickly get the medication into their systems. Every police officer within the substation is equipped with Narcan and it is used on a daily basis to prevent overdose deaths. And lastly, the COVID-19 pandemic shut down a number of shelters forcing many more people onto the streets. As shelters began opening up again, they reinstated a requirement to have a clean drug screening, limiting the amount of people able to get shelter when needed. These events created a perfect storm of homelessness and recurring drug use. Further complicating matters was the illegal drug trade, which operated by “renting” out street corners for dealers to sell their drugs; often renting out a corner to multiple dealers at once, resulting in fighting and violence between dealers. Transit officers from Kensington have reported that the permissive attitude to drugs in the area had harmed the neighborhood, the political environment both locally and nationally has increased scrutiny and removed tools that helped them mitigate the drug problem, and overall they report a sense of fatigue with the challenges in the area (Ratcliffe & Wight, 2022).

To respond to these mounting issues in the area, the Philadelphia Police Department implemented a satellite police substation within the 24th District, which had priority jurisdiction in the 0.36 square-mile area (see Figure 1). This substation became operational on January 25, 2021 with 38 officers assigned to the substation. These officers operate under a hybrid deployment model with each officer having two days with 12-hour shifts, plus 16 additional hours spread out across the rest of the week (40 hours total each week). One-hundred twenty officers were originally requested, and thus the substation does not operate 24-hours a day due to only receiving a third of that amount. Instead, the unit prioritizes having ten to fourteen officers operate on a single 12-hour shift each day, covering hours in the late morning into the late night. These officers patrol on bicycles, either with a single partner or a larger group of officers remaining together. They also work with community members to manage block parties and set up impromptu football games on the streets known to be prominent areas of the drug market. Such community activities are designed to both disrupt drug dealings as well as reassure residents not involved in the drug market that drug-related activities are not welcome. To be eligible for a bicycle patrol, officers undergo a one-week (40 hour) course of instruction in addition to their normal recruit and in-service trainings. Being on a bicycle is generally a sought-after position as it allows for a different style of patrol and a more comfortable uniform in summer months, and in Philadelphia there is a waiting list to get trained.

Figure 1. Kensington neighborhood substation target area


The collected data used in the study—all of which are publicly available—included Philadelphia centerlines (hereafter referred to as street segments) and property assessments from the City of Philadelphia, demographic information from the U.S. Census at the block-group level, and reported crime events from the Philadelphia Police Department. The street segment, property assessments, and crime data were available on Philadelphia’s opensource website (Open Data Philly, 2022). Census data were collected from its online tool (U.S. Census Bureau, 2022). All linkages between these data were conducted using ArcMap 10.8.1.

The unit of analysis in the models was the Philadelphia street segment. As such, individual crime events and demographic information (detailed below) were connected based on proximity to each street segment. The opensource street segment shapefile geographically displayed the network of Philadelphia streets, where two endpoints of a street were connected by a line with no intersecting streets. Where an intersection occurred, a new street segment was detailed. A total of 41,110 street segments were included in the data for the full city.

Because crime is attracted to different property types (Bernasco & Block, 2011; Brantingham & Brantingham, 1995; McCord et al., 2007; Tillyer & Walter, 2019), property assessment data were included from the city’s opensource website, which provides property characteristics and assessment history for all 582,012 properties in Philadelphia (Open Data Philly, 2022). The assessments were conducted recently, with roughly 87.9 percent conducted in 2021, and another 11.1 percent conducted more recently. The author coded the provided 15 property characteristics into five broader categories. Properties marked as residential were those that were assessed to be either apartments with four or more units, multi-family, separate garages that were part of a residence, and single-family properties. Properties assessed to be mixed use or special purpose were categorized into a mixed-use category. Commercial properties included those assessed as commercial, separate garages for commercial properties, hotels, offices, and retail properties. Properties assessed as industrial remained their own category. Finally, vacant properties included those that were assessed as vacant land, and vacant land for both residential and non-residential properties.

Crime data were collected from the city’s opensource website, which provided the date and time of the dispatch, geographic location by latitude and longitude, and the uniform crime report code and description associated with each type of crime (Open Data Philly, 2022). Crime data were collected for the entire city from approximately two years prior to the start of the bicycle patrol substation to approximately two years after (February 5, 2019, to January 14, 2023). This examined time period was based on the creation of 24 thirty-day periods on either side of the intervention start date of January 25, 2021. A total of 577,665 crimes were included in the data, covering 34 different event types. These crimes were connected to the street segments based on their proximity and a dataset was created at the street segment level that summated the criminal events into the 34 crime categories.

For the purposes of this study, the crime data were aggregated into six different categories, used as outcomes in the synthetic control models. The first, total crimes, counted all 34 criminal events.1 The second, social disorder crimes, counted crimes that primarily affect social cohesion and disorder, including disorderly conduct, driving under the influence, gambling violations, prostitution and commercialized vice, public drunkenness, vagrancy/loitering, and vandalism/criminal mischief. The third outcome was crimes associated with narcotic crimes and drug law violations. This measure was a single variable included in the opensource database and included any crimes associated with drugs (generally possession, use, selling, or manufacturing). The violent person crimes outcome counted all aggravated assaults with and without a firearm, criminal homicides, rapes, and robberies with and without a firearm. The property crimes outcome included arson, residential and non-residential burglary, motor vehicle theft, theft from vehicle, and theft. The shootings outcome included all recorded shootings where an individual was hit by gunfire. Shootings included all non-fatal and fatal shooting events, which likely overlap to some degree with the homicides included in the violent person crime category. The author purposely examined all shootings because shootings that end in a death often come down to a matter of millimeters of how the weapon was angled when discharged or where the bullet damages the body, or the caliber used (Braga & Cook, 2018). Furthermore, not all shootings result in a homicide charge and not all homicides occurred as the result of a shooting.

Community demographic information for each Philadelphia block group was collected from the United States Census online tool pertaining to the 5-year 2020 American Community Survey (U.S. Census Bureau, 2022). A total of 1,335 block groups were identified and used. Theories relating to social disorganization posit that structural factors lead to the disruption of community social organization, accounting for increases of crime and delinquency (Sampson & Groves, 1989; Shaw & McKay, 1942). For each block group, the collected data included the total population of residents, the percentages of residents who were white, Black, Hispanic (Table P2), between 15 to 29 years old (Table B01001), under the poverty line (Table B17021), and the percentage of individuals who are unemployed (Table B23025). These measures are consistent with prior research on measures of social disorganization (Sampson et al., 1997). The census values of the block groups were connected to each street segment based on the block group that the segment most fell within.

Analytic Strategy

To evaluate whether the implementation of the bicycle patrol unit affected crime levels, the author examined the data using the microsynthetic control method, which allowed the author to draw a statistically equivalent control area that matches pre-intervention crime trends of the treatment area, while being similar in its demographics and aspects of the built environment (Robbins et al., 2017; Saunders et al., 2015). Recent criminological research has applied the microsynthetic control method to evaluate the effect that place-based interventions have on crime levels, including examinations of marijuana dispensaries in Denver, Colorado (Connealy et al., 2019); a police substation in Newark, New Jersey (Piza et al., 2020); a neighborhood-specific crime intervention in Roanoke, Virginia (Robbins et al, 2017); a police-directed patrol in Flint, Michigan (Rydberg et al., 2018); and a drug market intervention in High Point, North Carolina (Saunders et al., 2015). These scholars have advocated that the microsynthetic control method is superior to propensity score matching (Apel & Sweeten, 2010; Dehejia & Wahba, 2002), another prominent quasi-experimental approach, because the microsynthetic control method creates a weighted vector of control units that can then be aggregated to create a single control unit that matches the pre-intervention trends of the cases within the area covered by the intervention as close as possible, most often with comparison units that exactly match the intervention unit’s pre-intervention characteristics. Furthermore, the microsynthetic control method allows for the inclusion of both time-invariant and time-variant covariates among micro-level units of analysis (Robbins & Davenport, 2019), which was not possible using earlier synthetic control methods (Abadie & Gardeazabal, 2003).

For the purposes of the study, the microsynthetic control procedure was applied separately for each examined crime outcome. All analyses were conducted in R Studio with R v.4.2.2 and the microsynth package (Robbins et al., 2017). Cases in the final dataset included the associated crime tallies and census information for every examined street segment by time period. A total of 48 time periods were included in the data, each being 30-days long, with period 25 onward marked (= 1) as the intervention period for the 287 street segments within the bicycle patrol jurisdiction. As such, the final dataset included 1,973,280 cases (41,110 street segments x 48 time periods). Each time-variant outcome model included the covariates detailed in the data section, along with a time-invariant total count of the other crime incidents experienced during the 2-year pre-intervention period. These crime-count variables ensured that the control street segments experienced comparable crime levels as the treatment street segments. For instance, the model on social disorder offenses included time-invariant total counts for the other five crime categories, along with the other time-invariant covariates. Table 1 displays the equivalency between the treatment unit and the calculated weighted control area, which perfectly matched for each model examining the six different crime measures.

Table 1. Pre-Implementation Treated and Weighted Control Equivalency

Bicycle Patrol

Weighted Control Area

Number of street segments (n)



Time-invariant covariates

Population a



Percent white b



Percent Black b



Percent Hispanic b



Percent 15-29 year olds b



Percent under poverty b



Percent unemployed b



Count of commercial properties



Count of industrial properties



Count of mixed-use properties



Count of residential properties



Count of vacant properties



Count of total crimes



Count of total social disorder crimes



Count of total drug / narcotic crimes



Count of total violent person crimes



Count of total property crimes



Count of shootings



Outcome-specific time-variant covariates

Avg. Monthly Total Crime c



Avg. Monthly Social Disorder Offenses c



Avg. Monthly Drug / Narcotics Offenses c



Avg. Monthly Person Offenses c



Avg. Monthly Property Offenses c



Avg. Monthly Shootings c



a: Total population across block groups; b: Fixed units expressed as an average over each, not the cumulative sum; c: Crime counts were equivalent across groups for each month. Average pre-implementation counts are presented here.

The author was interested in assessing the gradual cumulative impact the bicycle unit may have on the crime outcomes and when exactly that impact may become significantly different from the synthetic control. As such, the impact of the intervention was assessed through effect estimates of the linear difference-in-differences estimates of the cumulative crime counts in the treated area compared to the weighted control for each post-implementation time period. The differences in the first six-months after the initial implementation (periods 25 to 30) were not examined to allow time for the new unit to get its bearing in the field. Post-implementation estimates are presented as percent differences in graphical form, which detail the percent difference in the amount of crimes in the bicycle unit compared to the weighted control areas.

Trendlines of the raw crime amounts in the treated and control areas are provided in the appendix, along with permutation models. A total of 99 placebo tests were conducted with jackknife replication groups in order to evaluate the statistical significance of the synthetic control results. This involved randomly permuting the treatment status among the original street units, which resulted in a random set of 287 street units being assigned to treatment. The same synthetic control analysis was then conducted, which created a standard error of the treatment effect estimate. It is expected that under random permutation, the placebo-treated units would not have any impact on crimes post-treatment, but there would be some variation around zero. This variation under the null hypothesis allows estimation of a standard error around the treatment effect estimate, without relying on any specific distributional assumptions of the treatment estimate.


Table 2 displays the estimated crime counts in the street segments under the bicycle patrol jurisdiction and the weighted control area for each outcome at one-year and two-year post-implementation time periods. Raw values are displayed as well as corresponding percent differences, along with 95 percent confidence intervals for the differences between the two groupings. Upon initial review of this table, it appears that during the first year of operation, the bicycle patrol unit did not produce significant changes in any of the analyzed crimes. However, significant changes in crime rates were observed after two years of operation. This suggests the need for more comprehensive analyses that evaluate the cumulative impact of the intervention at each time period. Such an approach would better identify the specific time periods when significant changes occurred, as described in the subsequent sections for each crime category.

Table 2. Crime type change estimates, by post-implementation time period

Bicycle Patrol

Weighted Control



Lower 95% CI

Upper 95% CI

Total Crimes




276.8 (13.1%)

-42.4 (-2.0%)

644.5 (30.4%)




367.6 (8.5%)

-86.8 (-2.0%)

872.2 (20.1%)

Social Disorder Crimes




-36.8 (-13.8%)

-95.8 (-35.9%)

42.4 (15.9%)




-134.7 (-24.8%) *

-214.9 (-39.6%)

-35.3 (-6.5%)

Drug / Narcotics Crimes




85.8 (20.5%)

-35.5 (-8.5%)

245.5 (58.7%)




198.5 (30.1%) **

34.2 (5.2%)

401.7 (61.0%)

Violent Person Crimes




9.5 (2.1%)

-102.4 (-23.3%)

158.9 (36.0%)




172.3 (24.3%) *

12.8 (1.8%)

367.1 (51.8%)

Property Crimes




6.8 (1.6%)

-80.7 (-17.0%)

98.7 (24.3%)




-69.0 (-7.2%)

-180.0 (-18.8%)

58.4 (6.1%)





22.9 (27.2%)

-15.7 (-18.7%)

83.3 (99.1%)




88.6 (63.6%) **

12.0 (8.6%)

203.9 (146.3%)

* p < .05; ** p < .01, *** p < .001; CI: Confidence Interval

The 12 post-implementation periods correspond to 01/25/2021 to 01/19/2022, and the 24 post-implementation periods correspond to 01/25/2021 to 01/14/2023

Total Crimes

Results indicate that the bicycle patrol unit had no effect on the total amount of crimes after one or two years. As detailed in Table 2, the amount of total crimes in the bicycle patrol area was 13.1 percent higher than the control area after the first year, however, that lowered to 8.5 percent higher at the two-year post-implementation mark, approximately 368 crimes more than the amount of crimes in the control area. Examination of the percent differences if the amount of the cumulative total of all crimes between the bicycle patrol area and the synthetic control are detailed in Figure 2. For each time period, which include post-implementation periods covering approximately 7 months after the bicycle patrol unit was deployed (period 31) to 24 months after (period 48), the horizontal line indicates the estimate of the percent difference between the bicycle patrol unit and its control area, the vertical line indicates the 95 percent confidence internal of this estimate, and an asterisk indicates the difference being significant at p < .05.

Figure 2. Percent difference of cumulative total crimes between areas, by post-implementation periods 31 to 48

Results indicate that the cumulative counts of all crimes were significantly higher by the seventh post-implementation month (period 31) and remained significantly higher by the end of the nineth post-implementation month (period 33). Although a few other periods were found with significantly greater cumulative amounts of total crime (periods 37, 38, and 40), there was a gradual reduction in the cumulative differences over time, except in periods from October 17, 2022 to January 14, 2023 (periods 46 to 48, 22 to 24 months post implementation) where the cumulative amounts of total crimes began to increase, although not reaching a significant difference. In comparing the first-year post-implementation to the second year, the bicycle patrol unit experienced 276.80 more crimes in the first year compared to the control (23.07 more per month on average), but only 90.82 more crimes in the second year compared to the control (7.57 more per month on average).

Social Disorder Crimes

Unlike the other examined crime types, social disorder crimes were the only category where a reduction was observed. After the first year the bicycle patrol unit was deployed, social disorder crimes in the bicycle patrol jurisdiction were 13.8 percent lower when compared to those crimes observed in the control area; although this difference was not significant (see Table 2). A larger reduction was observed after two years, where 24.8 percent fewer social disorder crimes, or approximately 135 offenses, were noted in the bicycle patrol area compared to the control (p < .05). In examining the trend line of the cumulative amounts of social disorder crimes, detailed in Figure 3, it is noted that such crimes were more commonly below the expected amount observed in the control unit resulting in an expanding decrease in the difference of cumulative amounts, eventually reaching significance in the 20th post-implementation month onward (periods 44 to 48). In the first year, the bicycle patrol unit experienced 37 fewer social disorder crimes compared to the control (3.07 fewer per month on average), and that increased to 97.88 fewer social disorder crimes during the second year compared to the control (8.16 fewer on average).

Figure 3. Percent difference of cumulative social disorder crimes between areas, by post-implementation periods 31 to 48

Drug / Narcotic Crimes

Results on the effect the bicycle patrol unit had on drug and narcotic crimes indicate that these crimes significantly increased after the unit was implemented by the end of the second year (Table 2). In the first-year post-implementation, drug crimes were observed to be 20.5 percent higher than observed in the control area; and this amount increased to 30.1 percent, or approximately 199 drug offenses, after the two-year deployment (p < .01). Examination of the cumulative drug crime trend line, detailed in Figure 4, shows that the percent difference of cumulative drug crimes in the bicycle patrol unit was consistently about 25 percent higher than its synthetic control. An uptick in drug crimes in the bicycle patrol area after October 17, 2022 onward (periods 46 to 48) increased the percent difference into significance compared to the control area. In the first year, the bicycle patrol unit experienced 85.80 more drug crimes compared to the control (7.15 more per month on average) and that increased to 112.69 more drug crimes during the second year compared to the control (9.39 more on average).

Figure 4. Percent difference of cumulative drug / narcotic crimes between areas, by post-implementation periods 31 to 48

Violent Person Crimes

The model examining the change in violent person crimes indicate that these crimes were significantly higher in the bicycle patrol area after two years compared the values observed in the control area. As shown in Table 2, after the first year the bicycle patrol unit was deployed, violent person crimes in the bicycle patrol jurisdiction were just 2.1 percent higher when compared to those crimes observed in the control area; however, this difference significantly increased to 24.3 percent more violent person crimes after two years, or approximately 172 offenses, in the bicycle patrol area compared to the control (p < .05). Examination of the trend line on the percent difference in the cumulative amounts of violent crimes between the two areas, detailed in Figure 5, show that violent crimes were trending downward from July 24, 2021, to November 20, 2021 (periods 31 to 34, 7 to 10 months post implementation). During the full first-year post-implementation, only 9.48 more violent person crimes in the bicycle patrol area (0.79 more on average); however, beginning on November 21, 2021 (period 35, 11 months post implementation) onward, the bicycle patrol area observed a steady increase of violent person crimes compared to the expected amount observed in the control area. In the second year, the bicycle patrol unit experienced 162.79 more crimes compared to the control (13.57 more per month on average). This difference in the amount of violent person crimes compared to the control area year-to-year is 17.16 times higher, indicating that the vast majority of these violent person crimes occurred in the second year after the bicycle patrol implementation, eventually resulting in significant differences in the amounts of cumulative violent person crimes beginning September 17, 2022 (period 45, 21 months post implementation) onward.

Figure 5. Percent difference of cumulative violent person crimes between areas, by post-implementation periods 31 to 48

Property Crimes

Results on property crimes indicate no significant differences between the bicycle patrol area and the synthetic control area after one or two years. After one year, the bike patrol unit had approximately 6.8 additional property crimes (2.1%) compared to the control, but 69 fewer property crimes (-7.2%) by the end of the second year (see Table 2). When examining just the second year, property crimes in the bicycle patrol area were 75.75 fewer than in the control (6.31 fewer each month, on average). This second-year crime decline primarily occurred after July 19, 2022 (period 43, 19 months post implementation) when far fewer property crimes occurred in the bicycle patrol area compared to the control. Figure 6 details the trendline of the percent difference of the cumulative amounts of property crimes between the two areas, which shows non-significant estimates ranging from -7.2 percent to 14.1 percent.

Figure 6. Percent difference of cumulative property crimes between areas, by post-implementation periods 31 to 48


The relative changes in shooting crimes were the largest observed between the bicycle patrol unit and the synthetic control compared to the other crime categories. After one year of the bicycle patrol’s implementation, shootings increased by approximately 23 offenses, or 27.2 percent, than what was expected based on the control area values; although this difference was not significant (see Table 2). After two years, approximately 89 more shootings (63.6%) were observed in the bicycle patrol unit compared to the control area (p < .01). Figure 7 details the percent differences in the cumulative amounts of shooting events between the bicycle patrol area and its synthetic control. Much wider confidence intervals are observed resulting from the rarity of shooting events. Nonetheless, the majority of post-implementation months (16 of the 24) observed more shootings in the bicycle patrol area than the control area. Percent differences of the cumulative amounts indicate that a steady increase in shootings beginning around period 38 (February 19, 2022, approximately 9 months post implementation), eventually reaching a significantly higher cumulative amount of shootings compared to the synthetic control by period 40 (April 20, 2022, approximately 16 months post implementation). In the second year of the bicycle patrol program, there was an increase in the number of shootings in the patrol area compared to the control, similar to the observed trend in violent person crimes. In the first year, the bicycle patrol unit experienced 22.90 more shootings compared to the control (3.69 more per month on average), and that increased to 65.72 more shootings during the second year compared to the control (6.11 more each month on average). As such, the increase of shootings in the second year was 2.87 times higher than the increase in the first year.

Figure 7. Percent difference of cumulative shootings between areas, by post-implementation periods 31 to 48

Discussion and policy implications

The goal of this study was to better understand how a bicycle patrol unit implemented in a well-established major drug market may affect crime levels. The analysis also functions as a quantitative interrogation into the value of a police substation dedicated to addressing a seemingly intractable group of social problems.

Results show that any significant changes of crime levels as a result of the bicycle patrol unit occurred gradually, after the officers had been deployed in the jurisdiction for 16, 20, 21, and 22 months for shootings, social disorder crimes, violent person crimes, and drug crimes, respectively. Much of this may be due to how the COVID-19 pandemic impacted policing operations in Philadelphia. First, vaccines for the COVID-19 pandemic in the United States were initially released in mid-December 2020 and only 0.24 percent of the Philadelphia population was fully vaccinated when the bicycle patrol unit was deployed on January 25, 2021 (Statesman Journal, 2023). Levels for fully vaccinated people did not increase much by the end of the bicycle patrol unit’s first year, reaching only 13.9 percent in Philadelphia on January 24, 2022. As such, it may be likely that the officers of the unit were still operating under Philadelphia Police Department COVID-19 procedures, which required them to change their approaches to face-to-face public interactions with residents and community members. For example, a department protocol from April 3, 2020 advised that officers “will not personally engage any individual or group” when requesting residents to adhere to the Mayor’s stay at home order (PPD, 2020). This protocol, along with other COVID-related protocols, likely stymied the effect the bicycle patrol unit could have had in comparison to the control group, which was operating under the same procedures to reduce public contact. It was not until June 1, 2022, approximately 17 months after the bicycle patrol unit was deployed, when the majority of the city (64.6%) was fully vaccinated (Statesman Journal, 2023). By this point, if not much sooner due to being vaccinated, officers were much more likely to fully embrace their directives in the bike patrol jurisdiction, resulting in the significant effects observed in the second year.

Results indicate that the total amount of crimes and property crimes were not affected by the bicycle patrol unit. The trendlines for both the streets under the bicycle patrol jurisdiction and the matched comparison streets showed similar declines in the total amount of crimes in the first year, followed by similar increases in the second year (see Figure A-1 in the appendix). The trendlines for property crimes for the two groups both showed gradual and statistically similar increases in property crimes after the bicycle patrol unit was implemented (Figure A-2). On the other hand, social disorder crimes were observed to significantly declined as a result of the bicycle patrol unit, with a 24.8 percent reduction in social disorder crimes by the end of the second year. These findings likely relate to the directives of the bicycle patrol unit. The officers assigned to this unit were primarily tasked to improve the social structure of the neighborhoods, namely by improving social cohesion and order. While the officers responded to calls for service regarding other minor offense types and property crimes, they did not make such offenses a priority when proactively patrolling the jurisdiction. Instead, they were tasked with disrupting the drug market through frequent and sudden interactions with unsuspecting offenders or users. And it is very likely that the increased presence of officers in the streets increased offenders’ belief that they could be caught when committing a public social disorder crime, such as disorderly conduct, gambling, loitering, or vandalism. Similar reductions in disorder crimes have been observed in other place-based, hot-spot policing studies (Braga et al. 2019).

Which leads us to the perhaps paradoxical findings associated with the impact the bicycle patrol unit had on drug and narcotic offenses, violent person offenses, and shootings, all of which observed significant increases in the bike patrol jurisdiction compared to the expected amount observed in the control area. By the end of the two-year post-deployment period, drug and narcotic crimes, violent person crimes, and shootings were found to be 30.1, 24.3, and 63.6 percent higher than observed in the control area, respectively.

The primary directive of the bicycle patrol unit was to combat and reduce drug crimes in a well-establish drug market. The officers prioritize disrupting drug deals on the streets by arresting dealers and apprehending the narcotics they were selling, resulting in an increase in official documentation of drug crimes. And it is likely because of this drastic disruption of the drug market that violent person crimes and shootings simultaneously increased, an effect that has been observed in other studies. Werb et al. (2011) conducted a systematic review that examined the impacts of law enforcement activities focusing on drug crimes and markets and found that 14 of the 15 identified studies reported increases in violent crimes. For example, Benson et al. (1998, 2001) found that increases in the rate of drug arrests in Florida were associated with violent and property crimes increasing by a factor between 2.20 to 4.63. They argue, in part, that the trade-off when police departments prioritize efforts to reduce drug-related crimes is a reduction to police resources that are focused on controlling violent crimes, leading to notable increases in such offenses (Benson et al., 1998). Other scholars posit that this is likely a result of officers’ successful efforts to disrupt well-established drug markets, allowing for new and inexperienced individuals to fill the void, subsequently increasing the likelihood of conflict as individuals attempt to stake their claim in the illicit business (Maher & Dixon, 1999; Rasmussen et al., 1993). Levitt and Venkatesh’s (2000) observational study of Chicago gangs found that 25 percent of the gang’s activities were responding violently to conflict created by law enforcement pressure on the drug market. In the current study, it is possible that as officers actively patrolled the area and use the bicycle patrols to better approach drug sellers to catch them in the act of a deal, the drug market responded by introducing new, inexperienced sellers who are willing to except the high-risk position to sell drugs (as observed in Maher & Dixon, 1999) and this led to increased violence as dealers tried to establish themselves and the areas they wanted to control.

The behaviors of people who use drugs and live within disrupted drug markets could also be a contributing factor to the observed escalation of violence. Ray et al. (2023) examined law enforcement efforts in 2020 and 2021 in Marion County, Indiana that disrupted local drug markets by seizing opioids. The authors found increased spatiotemporal clustering of overdose events within 500 meters and 21 days following opioid enforcement activities, including fatal overdoses, nonfatal overdoses, and naloxone administrations. The authors suggest that individuals with opioid use disorder who no longer have access to their regular supply because of dealer arrests may face a combination of reduced tolerance and withdrawal symptoms, leading them to actively seek out a new supply, often disregarding the risks associated with the varying potency levels in the illicit opioid market. As a result, they are exposed to unknown levels of tolerance, uncertainty regarding safe dosages, and an elevated risk of overdose. These behaviors likely create additional demand for the illicit drug market, leading drug dealers and cartels to identify new areas with novice personnel who establish themselves among new clientele and often employ violence as a means to gain a competitive advantage.

The unintended outcomes associated with Philadelphia’s bicycle patrol unit provide valuable insights for agencies seeking to address drug crime. The primary finding is that bicycle patrol can have a beneficial impact on social disorder crime; however, as previously shown, such police activities can have unforeseen consequences. In this case, the intervention site saw an increase in drug offenses, violent person crimes, and shootings. The author hypothesizes here that the interruption in the normal operation of the Kensington area drug markets by the introduction of a significant change to police activity also introduced instability to the street-level drug business with the corollary of increased violence. But other possible benefits of a focused bike program need to be weighed against the potential for increased market instability and violence. Past research has identified that bicycle patrol deployments increase face-to-face contact with community members, which could lead to greater effectiveness in problem-solving and community engagement, especially when supported by trainings on procedural justice and operated within community-oriented policing mandates (Sytsma and Piza, 2018). Implementing a geographically targeted bicycle patrol strategy that prioritizes positive community interactions has the potential to generate enduring positive effects on community relations, as well as contribute to the reduction of both minor and more severe criminal activities. For example, del Pozo et al. (2021) advocate for considerable discretion by officers when engaging with individuals who use drugs, particularly in the enforcement of drug-related misdemeanors or nonviolent felonies. This approach aims to minimize potential harms that may arise from disrupting an individual's drug supply. Such activities could strengthen views of police legitimacy, especially in such challenged areas where social services may be lacking and improved crime control is sorely needed.

The findings further emphasize that a comprehensive approach to bicycle patrols extends beyond mere presence, highlighting the importance of tailoring these programs to address specific local concerns and factors. For instance, departments should consider detailed crime analyses and intelligence gathering that go alongside patrols that are designated to breakup drug markets. Such information could include identifying the key individuals managing the drug market and their respective corners that they rent out to sellers. With this knowledge, district and substation commanders could be better equipped to direct patrols in areas where corners are newly available for “rent” after a dealer is arrested and have a higher potential for conflict to occur. In fact, officers in the substation informed the research team that newly rented or open corners are where the majority of violence occurs. This happens when new and inexperienced individuals believe they can take over a corner from a gang or when a gang rents the corner out to multiple individuals at the same time. Directed patrol in these areas may benefit from localized support with foot patrol operations, which have been successful reducing violent crime in hotspots elsewhere in the city in 2009 (Ratcliffe et al., 2011; Wood et al., 2015). These past foot patrol efforts allowed wide discretion in officer approaches—from direct enforcement activities to more community-centered methods—and when combined with faster responding bicycle patrols, could provide a well-balanced approach to addressing violence in these areas experiencing a high potential for violence as a result in drug market disruptions. Given the potential for both benefits and negative consequences, ongoing evaluation and tracking of policing initiatives is an imperative (Sherman, 2013).

Many of the above policies are not possible without the necessary workforce to support the operations. The policing occupation is currently facing significant staffing challenges (see Adams et al., 2023) and it may be difficult for an agency to implement a bicycle patrol unit given staffing shortages and budget concerns. The bicycle patrol unit in Philadelphia is a perfect example. The substation requested 120 officers for the unit and subsequently received only a third of that amount (38 officers). The department implemented the substation to directly respond to a problem known by officers, officials, and residents; and yet it still could not be adequately staffed. As departments implement targeted interventions in specific locations, they must ensure that the units are adequately staffed not only to comply with priority directives but also to respond to any unforeseen consequences that may arise.

Limitations and conclusion

A number of limitations in this study present opportunities for future research on bicycle patrols. First, because the results of this study are limited to experiences from a single, large jurisdiction, the generalizability of these findings is an empirical question that should be explored in other cities. Of the estimated 33 percent of law enforcement agencies that have implemented bicycle patrol units (DOJ, 2020), few have done so in the same type of environment, for the same reasons, or by the same process as the Philadelphia Police Department. For example, the bicycle patrol substation was implemented in response to a significant need to address violent and drug crime problems that developed in a section of the Kensington neighborhood as a result of multiple external factors. Other cities may decide to implement bicycle patrols for different reasons.

Second, the reliance on a quasi-experimental design that was conducted post-hoc the bicycle patrol unit’s implementation is not nearly as strong a design that could be implemented through a randomized controlled trial (RCT). In this context, an RCT could be implemented within the Kensington drug market where randomly selected officers patrol on bicycles and control officers continue business-as-usual. This approach would allow more confidence in the results being directly caused by the bicycle patrol unit. While an RCT was not possible in this setting, this study instead relies on a novel approach to create equivalent place-based, experimental groupings to assess the bicycle patrol unit's impact.

Third, there are likely many other factors that influence crime levels. For example, specific police missions and operations were not accessible and therefore excluded from the models, as were community activities and events organized by community-based organizations, all of which may affect criminal activities. Furthermore, it is unknown if these types of activities differed across the experimental groupings.

Fourth, while the author attempted to include a wide array of criminal events in the evaluation, the study was limited to what was contained within official police data sources, which undercounts actual figures of crime (Coleman & Moynihan, 1996). This is especially salient in the finding on the increase of drug crimes after the bicycle patrol unit’s implementation. It would likely be inaccurate to posit that drug crime increased because of the bicycle unit, who’s primarily directive was to arrest drug dealers. Instead, it is much more likely that drug crime was unaccounted prior to the unit being deployed and the unit’s actions better accounted for the criminal drug events. While this is an issue, it is likely an issue that is similar across areas of the city and the use of the synthetic control method allowed models that created equivalent matches to the pre-period drug crime levels.

Finally, the author cannot say a great deal about the quality of the intervention. While the study was able to assess the impact, the author was not able to accompany the officers or examine the types of calls for service they attended, or activity they generated.

These limitations notwithstanding, there remains value in the results, given that while the total amount of crime remained unchanged, streets covered by the bicycle patrol unit observed a reduction in street-level disorder such as disorderly conduct, prostitution, public drunkenness, vagrancy/loitering, along with increased police-activities to arrest drug offenders and remove drugs from the streets. But the potential for iatrogenic consequences on other important crime types is possible, such as increases in violent person crimes and shootings. As Philadelphia continues to address high crime levels, divergence from traditional patrol mechanisms may provide considerable value if managed within a strategic framework of community safety. More research is needed to maximize the value of bicycle patrol units while simultaneously minimize potential consequences, not only to support community policing efforts but to respond to and address crime levels.


Adams, I. T., Mourtgos, S. M., & Nix, J. (2023, March 1). Turnover in large US police agencies following the George Floyd protests.

Asher, J. (February 1, 2022). We need to understand America’s spike in murders, CNN. Retrieved from:

Apel, R. J., & Sweeten, G. (2010). Propensity score matching in criminology and criminal justice. In D. Weisburd & A. Piquero (Eds.), Handbook of Quantitative Criminology (pp. 543-562). New York: Springer.

Barclay, P., Buckley, J., Brantingham, P. J., Brantingham, P. L., & Whinn-Yates, T. (1996). Preventing auto theft in suburban Vancouver commuter lots: Effects of a bike patrol. In R.V. Clarke (ed.), Preventing Mass Transit Crime. Crime Prevention Studies, vol. 6. (pp. 133-161). Monsey, NY: Criminal Justice Press.

Bekiempis, V. (May 17, 2015). Foot Patrol: A Catch-22 of community policing. Newsweek.

Benson, B. L., Leburn, I. S., & Rasmussen, D. W. (2001). The impact of drug enforcement on crime: An investigation of the opportunity cost of police resources. Journal of Drug Issues, 31(4), 989-1006.

Benson, B. L., Rasmussen, D. W., & Kim. I. (1998). Deterrence and public policy: Trade-offs in the allocation of police resources. International Review of Law and Economics, 18, 77-100.

Bernasco, W., & Block, R. (2011). Robberies in Chicago: A Block-Level Analysis of the Influence of Crime Generators, Crime Attractors, and Offender Anchor Points. Journal of Research in Crime and Delinquency, 48(1), 33–57.

Braga, A. A. & Cook, P. J. (2018). The association of firearm caliber with likelihood of death from gunshot injury in criminal assaults. JAMA Network Open, 1(3).

Braga, A. A., Papachristos, A. V., & Hureau, D. M. (2010). The concentration and stability of gun violence at micro places in Boston, 1980–2008. Journal of Quantitative Criminology, 26, 33–53.

Braga, A. A., Turchan, B. S., Papachristos, A. V., & Hureau, D. M. (2019). Hot spots policing and crime reduction: An update of an ongoing systematic review and meta-analysis. Journal of Experimental Criminology, 15(3), 289-311.

Brantingham, P. & Brantingham, P. (1995). Criminality of place: Crime generators and crime attractors. European Journal on Criminal Policy and Research, 3, 5-26.

Chalfin, A., & McCrary, J. (2017). Criminal deterrence: A review of the literature. Journal of Economic Literature, 55(1), 5-48.

Cohen, J. (1962). The statistical power of abnormal-social psychological research: A review. The Journal of Abnormal and Social Psychology, 65(3), 145–153.

Coleman, C. & Moynihan, J. (1996). Understanding crime data: Haunted by the dark figure. England, UK: Open University Press.

Connealy, N., Piza, E., & Hatten, D. (2019). The Criminogenic Effect of Marijuana Dispensaries in Denver, Colorado: A Microsynthetic Control Quasi-Experiment and Cost-Benefit Analysis. Justice Evaluation Journal, 3(1), 69-93.

Cornish, D. B., & Clarke, R. V. (1987). Understanding crime displacement: An application of rational choice theory. Criminology, 25(4), 933-948.

Dehejia, R. H., & Wahba, S. (2002). Propensity score-matching methods for nonexperiment causal studies. The Review of Economics and Statistics, 84(1), 151-161.

del Pozo, B., Beletsky, L., Goulka, J., & Kleinig, J. (2021). Beyond decriminalization: ending the war on drugs requires recasting police discretion through the lens of a public health ethic. American Journal of Bioethics, 21(4), 41–44.

Federal Bureau of Investigation (FBI). (2022). Crime data explorer. Washington, DC: Federal Bureau of Investigation, Uniform Crime Reporting Program. Retrieve at:

Felson, M. (2016). The routine activty approach. In Wortley, R., & Townsley, M. (eds.) Environmental criminology and crime analysis. London, UK: Routledge.

Groff, E. R., Johnson, L., Ratcliffe, J. H., & Wood, J. D. (2013). Exploring the relationship between foot and car patrol in violent crime areas. Policing: An International Journal of Police Strategies and Management, 36(1), 119-139.

Johnson, N. J., Roman, C. G., Mendlein, A. K., Harding, C., Francis, M., & Hendrick, L. (2020). Exploring the influence of drug trafficking gangs on overdose deaths in the largest narcotics market in the eastern United States. Social Sciences, 9(11), 202-222.

Kelling, G. L., Pate, T., Dieckman, D., & Brown, C. E. (1974). The Kansas City Preventative Patrol Experiment: A Summary Report. Washington DC: Police Foundation.

Lawton, B. A., Taylor, R. B., & Luongo, A. J. (2005). Police officers on drug corners in Philadelphia, drug crime, and violent crime: Intended, diffusion, and displacement Impacts. Justice Quarterly, 22(4), 427-451.

Levitt, S. D., & Venkatesh, S. A. (2000). An economic analysis of a drug- selling gang’s finances. Quarterly Journal of Economics, 115(August), 755– 789.

Maher, L., & Dixon, D. (1999). Policing and public health: Law enforcement and harm minimization in a street-level drug market. The British Journal of Criminology, 39(4), 488.

McCord, E. S., Ratcliffe, J. H., Garcia, R. M., & Taylor, R. B. (2007). Nonresidential Crime Attractors and Generators Elevate Perceived Neighborhood Crime and Incivilities. Journal of Research in Crime and Delinquency, 44(3), 295–320.

Menton, C. (2008). Bicycle patrols: An underutilized resource. Policing: An International Journal of Police Strategies and Management, 31(1), 93-108.

Nagin, D. S. (2013). Deterrence in the Twenty-First Century. Crime and Justice, 42(1), 199-263.

Newall, M. (May 21, 2021). Business and bloodshed. Philadelphia Inquirer. Retrieved from:

Open Data Philly. (2022). Databases on Crime incidents, Properties and Assessment History, Shooting Victims, Street Centerlines. Philadelphia, PA: Open Data Philly. Data retrieved on March 9, 2023 from:

Philadelphia Police Department (PPD). (2020). COVID-19 enforcement of Mayor’s stay at home order (04-03-2020). Philadelphia, PA: Philadelphia Police Department.

Piza, E. L., Wheeler, A. P., Connealy, N. T., and Feng, S. Q. (2020). Crime control effects of a police substation within a business improvement district: A quasi-experimental synthetic control evaluation. Criminology & Public Policy, 19, 653-684.

President’s Task Force on 21st Century Policing. (2015). Final Report of the President’s Task Force on 21st Century Policing. Washington, DC: Office of Community Oriented Policing Services.

Rasmussen, D. W., Benson, B. L., & Sollars, D. L. (1993). Spatial competition in illicit drug markets: The consequences of increased drug war enforcement. Review of Regional Studies, 23(3), 219-236.

Ratcliffe, J. H., & Sorg, E. T. (2017). Foot Patrol: Rethinking the Cornerstone of Policing. New York: Springer (CriminologyBriefs).

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(3), 795-831.

Ratcliffe, J. H., & Wight, H. (2022). Policing the largest drug market on the eastern seaboard: Officer perspectives on enforcement and community safety. Policing: An International Journal.

Ray, B., Korzeniewski, S. J., Mohler, G., Carroll, J. J., del Pozo, B., Victor, G., Huynh, P., & Hedden, B. J. (2023). Spatiotemporal analysis exploring the effect of law enforcement drug market disruptions on overdose, Indianapolis, Indiana, 2020–2021. American Journal of Public Health, 113(7), 750-758.

Robbins, M., Saunders, J. & Kilmer, B. (2017). A framework for synthetic control methods with high-dimensional, micro-level data: Evaluating a neighborhood-specific crime intervention. Journal of the American Statistical Association, 112(517), 109-126.

Rydberg, J., McGarrell, E., Norris, A., & Circo, G. (2018). A quasi-experimental synthetic control evaluation of a placebased police-directed patrol intervention on violent crime. Journal of Experimental Criminology, 14, 83–109.

Sampson, R. J., & Groves, W. B. (1989). Community structure and crime: Testing social-disorganization theory. American Journal of Sociology, 94, 774–802.

Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277(5328), 918–924.

Saunders, J., Lundberg, R., Braga, A., Ridgeway, G. & Miles, J. (2015). A synthetic control approach to evaluating place-based crime interventions. Journal of Quantitative Criminology, 31(3), 413-434.

Shaw, C., & McKay, H. (1942). Juvenile delinquency and urban areas. Chicago, IL: University of Chicago Press.

Sherman, L. W. (2013). The rise of evidence-based policing: Targeting, testing and tracking. In M. Tonry (Ed.), Crime and Justice in America, 1975-2025 (Vol. Crime and Justice 43). Chicago: University of Chicago Press.

Skogan, W. G., & Hartnett, S. M. (1997). Community Policing, Chicago Style. New York: Oxford University Press.

Statesman Journal. (2023). Philadelphia County, PA COVID-19 Vaccine Tracker. Statesman Journal. Data retrieved on April 13, 2023 from:

Sytsma, V. A., & Piza, E. L. (2018). The influence of job assignment on community engagement: Bicycle patrol and community-oriented policing. Police Practice and Research, 19(4), 347-364.

Tillyer, M. S., & Walter, R. J. (2019). Busy Businesses and Busy Contexts: The Distribution and Sources of Crime at Commercial Properties. Journal of Research in Crime and Delinquency, 56(6), 816–850.

U.S. Census Bureau. (2022). Tables P2, B01001, B17021, B23025, and H1 by block group from 5-year 2020 American Community Survey. Retrieved from

United States Department of Justice (DOJ). (2011). Law Enforcement Management and Administrative Statistics (LEMAS), 2007. Office of Justice Programs. Bureau of Justice Statistics. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [producer and distributor].

United States Department of Justice (DOJ). (2020). Law Enforcement Management and Administrative Statistics (LEMAS), 2016. Office of Justice Programs. Bureau of Justice Statistics. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [producer and distributor].

Werb, D., Rowell, G., Guyatt, G., Kerr, T., Montaner, J., & Wood, E. (2011). Effect of drug law enforcement on drug market violence: A systematic review. International Journal of Drug Policy, 22(2), 87-94.

Wood, J. D., Taylor, C. J., Groff, E. R., & Ratcliffe, J. H. (2015). Aligning policing and public health promotion: Insights from the world of foot patrol. Police Practice and Research, 16(3), 211-223.

Wood, S. & Farr, S. (June 15, 2017). Philly and Conrail to clean up “heroin hellscape.” Philadelphia Inquirer. Retrieved from:


Figure A-1. Synthetic control effect estimates for total crimes

Figure A-2. Synthetic control effect estimates for social disorder crimes

Figure A-3. Synthetic control effect estimates for Drug / Narcotic crimes

Figure A-4. Synthetic control effect estimates for Violent Person crimes

Figure A-5. Synthetic control effect estimates for Property crimes

Figure A-6. Synthetic control effect estimates for Shooting Crimes

No comments here
Why not start the discussion?