Quantifying Crime Prevention Potential of Near-Repeat Burglary

Abstract

The space-time risk window associated with near-repeat burglary patterns would seem to present a natural opportunity for burglary prevention efforts. However, constraints associated with the reporting of, police response to, and space-time patterning of burglaries can reduce the crime prevention potential of such efforts. To better estimate the crime prevention potential of focusing on near-repeat burglaries, we studied burglary patterns in 10 U.S. cities. Descriptive aspatial and spatial statistics were used to answer the research questions. Significant space-time clustering does not necessarily indicate an actionable near-repeat problem. Police analysts and researchers should also consider the crime prevention potential of focusing on near repeats—in other words, the proportion of burglaries that are preventable. The results of this test provide new information to guide the implementation and evaluation of crime prevention efforts focused on near-repeat events.

Keywords

microlevel near repeat residential burglary crime prevention

Background

Translational criminology, evidence-based policing, and crime science are three transformative ideas in policing. Translational criminology emphasizes the conversion of research evidence into policies and programs that can be rigorously tested for how well they reduce or prevent crime (Laub, 2011). It goes beyond scientific discovery and simple dissemination to demanding subsequent testing and evaluation about how and when something works. Evidence-based policing involves using evidence to create guidelines and evaluate programs in the real world and then using the results of those evaluations to improve the guidelines and, subsequently, practice (Sherman, 1998). Evidence-based policing emphasizes a cycle of continual improvement; new research findings inform policies and programs, programs implemented by agencies are evaluated to determine whether they are successful, and the results of those evaluations inform improvements to guidelines and programs. Crime science originated among adherents of situational crime prevention. It focuses on applying scientific methods to prevent and reduce crime (Laycock, 2005). These ideas inspired and informed the present examination of studying near-repeat burglary patterns to prevent future residential burglaries.

A large and growing body of evidence indicates that once a burglary has occurred, nearby homes are at higher risk for also being the victim of a burglary (Johnson, Davies, Murray, Ditta, Belur, & Bowers, 2017; Townsley, Homel, & Chaseling, 2000). Studies conducted in Newark, NJ; Houston, TX; Indianapolis, IN; Jacksonville, FL; Long Beach, CA; Philadelphia, PA; and Pompano Beach, FL, found significant microlevel, space-time clustering of near-repeat burglaries over a range of distances from 100 m (328 ft) to 400 m (1,328 ft) and at 14 days or less (Johnson et al., 2007; Moreto, Piza, & Caplan, 2014; Piza & Carter, 2017; Short, D’Orsogna, Brantingham, & Tita, 2009; Zhang, Zhao, Ren, & Hoover, 2015). International studies have found that the space and time profile varies but is most often within 200 to 400 m (656–1,328 ft) and 2 to 4 weeks of the initial burglary incident, after which the risk declines to its preburglary level (Bowers & Johnson, 2016; Johnson et al., 2007; Townsley, Homel, & Chaseling, 2003). The relatively small and predictable space-time window provides an excellent target for crime prevention efforts. Most field studies—many of which were conducted outside the United States—have translated this knowledge into successful crime reduction programs in large housing complexes and neighborhoods (Anderson, Chenery, & Pease, 1995; Chenery, Holt, & Pease 1997; Forrester, Chatterton, Pease, & Brown, 1988; Johnson et al., 2017). Sounding a cautionary note, one recent study found that the proportion of burglary events that were near repeats varied by area examined, suggesting that there may be variation in the expected benefits from focusing on near repeats (Chainey, Curtis-Ham, Evans, & Burns, 2018).

Existing literature points out that focusing on near-repeat crime offers several potential advantages to police. First, it allows targeting of scarce police resources (Johnson et al., 2007). Most law enforcement agencies operate under intense budgetary pressure, and improving efficiency is one way to maintain public safety in a challenging environment. Second, the incidence of near-repeat burglary can be used as a performance indicator to highlight police effectiveness (Ratcliffe & Rengert, 2008). Third, many crime prevention programs targeting near repeats emphasize the importance of neighborhood residents in an effective response. Efforts to increase the involvement of neighbors around burglary are likely to provide the basis for improved police–community partnerships (Groff & Taniguchi, 2019; Johnson et al., 2017).

Several microlevel studies have examined whether notification of increased risk and provision of crime prevention information, tools, or both could interrupt near-repeat patterns. Wellsmith and Birks (2008) investigated whether some combination of information about preventing burglary and tools to implement target-hardening measures could reduce near repeats. Because of limitations in the data, the authors could conclude only that the program had been a success compared with the rest of the jurisdiction. A randomized controlled experiment focusing on the high-risk space-time windows for residential burglary tested whether resident notification by uniformed volunteers could interrupt near-repeat patterns in two cities (Groff & Taniguchi, 2019). Individual cities showed no significant difference between treatment and control sites, but across both there was a slight but significant reduction. In contrast, several studies have found a significant reduction in near-repeat burglaries when police provided a pamphlet of crime prevention information to the victimized house as well as immediate neighbors within 48 hours of the original burglary report. The number of immediate neighbors varied across studies from a low of 8 (Stokes & Clare, 2019), to 26 (Weems, 2014), to all others on same block (Thompson, Townsley, & Pease, 2008). Another recent study, which conducted interventions at the microlevel but analyzed results at the neighborhood level, provides strong evidence that providing crime prevention information to the eight nearest neighbors on either side of the victimized dwelling significantly reduces crime (Johnson et al., 2017).

Drawing from the evidence described earlier and from the success of hot spot policing strategies generally (Braga, 2007), several police-centric microlevel studies have examined whether increased police patrol would reduce near-repeat burglary. A field experiment conducted in Amstelveen, The Netherlands, examined the use of increased police patrol to interrupt near-repeat patterns by putting police in a better position to make arrests, reduce near-repeat victimizations, or both (Elffers, Peeters, van der Kemp, & Beijers, 2018). In a 250-m (820-ft) radius around the burglary location, they increased patrols between 7 a.m. and 11 p.m. They found no significant crime prevention effect. In contrast, Fielding and Jones (2012) used 400-m buffers around burglaries and calculated the times during which burglary was highest. The resulting spatiotemporally focused policing reduced residential burglary by 26.6%. Other interventions in which increased patrol was sent to “microtime hot spots”—defined as areas where two or more events occur close in time (within 14 days) and space (within 0.79 sq mi) (Santos & Santos, 2015a, 2015b)—reduced residential burglaries by 20.76%.

Several factors might have contributed to findings of no significant effects (Elffers et al., 2018; Groff & Taniguchi, 2019). One clear candidate is the low base rates of near-repeat burglaries in target and control areas. In addition, follow-on burglaries often occurred the same day as the originator, and thus, police cannot respond quickly enough to prevent near-repeat burglary.¹ Residential burglary often is not discovered immediately; such a lag makes temporal ordering difficult to establish and delays police deployment.

Researchers (Groff & Taniguchi, 2019) emphasized that the program they tested had focused on preventable burglaries (i.e., a burglary that occurs after a potential originator has been reported to the police and within the near-repeat space-time window). Near repeats that occur before the police become aware of a problem are not preventable by reactive strategies. Existing tools identify whether statistically significant space–time interaction is present but do not provide any information about the size of the preventable near-repeat problem. A scientific approach to evaluating a crime reduction program should include a measure of the crime prevention potential (Laycock, 2005; Sherman, 1998).

Focusing on preventable burglaries when formulating crime prevention programs helps law enforcement agencies set realistic expectations for the potential crime reduction that might occur because of the intervention. It also provides police management with a more accurate measure of the likely benefit from deploying a crime prevention program targeted at near repeats. Knowing the likely benefit is critical for setting appropriate benchmarks for accurate evaluations of program efficacy. This information is essential for accurately estimating a priori effect size and designing a sufficiently powerful research protocol for detecting change. Relatedly, this approach helps keep effective programs from being discontinued because they are seen as underperforming when in reality they are being judged against an unrealistic goal. To illustrate, consider the following scenario. Two cities each have 1,000 burglaries a year. They both are aware of and want to use the research evidence that indicates that focusing on near repeats can significantly reduce burglary. Both implement an intervention, and both see burglaries go down by 50 incidents or 5% (50/1,000). The chief of Police in City A discontinues the program because of low return on investment. City B staff, however, had further analyzed the numbers before implementation to identify 100 near repeats in the previous period. Because City B measures the amount of crime reduction against the 100 near repeats, it reports a 50% reduction in near repeats (50/100) and expands the program because of its success.

Technology plays a key role in supporting translational and evidence-based criminology. In the case of near repeats, it is very difficult and time-consuming to manually identify events that are near repeats. However, it is relatively straightforward programmatically because a software program iterates through a file, keeps track of each record it reads, and assigns it a status based on the decision rules specified by the user. The creation of the near-repeat calculator (NRC; Ratcliffe, 2009) enabled researchers and practitioners to quantify their near-repeat crime problems. Two additional software packages now exist to facilitate the examination of near repeats, but neither of them allow the user to control which events are counted as near repeats.² To remedy this situation, we created a new software program, the Near Repeat Crime Prevention Potential Calculator (NR-CPPC), to calculate the percentage of burglary events that would have been preventable near repeats in a historic data set (Groff & Taniguchi, 2019).³ The software provides an easy-to-use tool that allows crime analysts to estimate the likely future size of their preventable near-repeat burglary problem.

This article investigates practice-based questions that agencies should ask when considering whether to implement a crime prevention program targeting near repeats. First, does the crime reduction potential of disrupting near-repeat patterns vary across places? Second, how can agencies decide whether the size of the near-repeat problem is large enough to warrant an expenditure of time and resources? Third, what additional analyses of near-repeat patterns can be used to inform the deployment of resources? Finally, what is the utility of near-repeat patterns for deploying burglary prevention initiatives specifically and crime prevention initiatives generally?

Methods

Study Cities and Data

The NR-CPPC software was used to examine the size of the preventable near-repeat burglary problem across 10 jurisdictions in the United States. Data for two sites—Baltimore County, MD, and Redlands, CA—were obtained directly from those police departments. Data for the remaining cities were located through the Police Foundation’s Public Safety Open Data Portal and then downloaded from each agency’s open data portal.⁴ To be included in the analysis, a jurisdiction had to provide incident-level, geocoded burglary data from some time period during the mid-2010s (Table 1). The cities were geographically dispersed, with five East Coast sites, three West Coast sites, and two in the middle of the country. The sites offered a wide range of burglary volume, from under 400 to almost 9,000 per year. Population and housing units also varied widely, from small cities like Redlands (about 70,000 people and 26,000 units) to major cities such as Philadelphia (about 1.5 million people and 670,000 units).

Table 1.

Descriptive Data for Study Cities.

Reporting agency	Number of burglaries	Data period	HUs^a	Burglary rate per 1,000 HUs	Population^b	Type of street network^c	Geocoding method^d	Geocoding method source^e
Baltimore Co., MD^f	611	9/8/2013 – 9/7/2014	335,622	1.82	831,026	Organic	SCO	Agency representative
Denver, CO	2,495	2015	285,797	8.73	693,060	Grid	P, SCO	Manual inspection
Durham, NC	2,580	2016	103,221	25.00	263,016	Organic	P, SCO	Agency representative
Fayetteville, NC	1,958	2015	87,005	22.51	204,759	Organic	P, SCO	Manual inspection
Orlando, FL	2,122	2015	121,254	17.50	277,173	Mixed	SBC	Shapefile metadata
Philadelphia, PA	8,716	2013	670,171	13.01	1,567,872	Mixed	SCO	Agency representative
Redlands, CA	361	2013	26,634	13.55	71,288	Organic	P	Agency representative
Santa Rosa, CA	522	2016	67,396	7.75	175,155	Mixed	NI	Manual inspection
Seattle, WA	7,567	2016	308,516	24.53	704,352	Mixed	BC	Manual inspection
St. Louis, MO	3,670	2015	176,002	20.85	311,404	Mixed	P	Manual inspection

^aApril 1, 2010, U.S. Census Bureau, 2010 Census of Population, P94-171 Redistricting Data File. Updated every 10 years. American Factfinder. July 1, 2016 housing unit estimates are available for counties but not cities. For consistency, all housing unit figures are April 1, 2010. HU = housing units.

^bJuly 1, 2016, U.S. Census Bureau, Population Estimates Program, updated annually. Population and Housing Unit Estimates.

^cValues in this column describe the street network. Grid: if predominantly a grid network with only a few smaller areas with organic streets; Organic: if predominantly organic with only a few smaller areas with gridded streets; and Mixed: if large portion of the downtown area is a grid but suburban areas are organic.

^dSCO, Street centerline with offset; P, Parcel centroid; SBC, Street block centroid; NI, Nearest intersection. If two methods are noted, the first is the primary method. The second is the fallback method if the first geocoding type is not successful.

^eDocumentation of geocoding methodology was omitted in many of the open datasets used in this analysis. Determining the method of geocoding required a number of different techniques. Where possible, we reached out to agency representatives to request information on geocoding process. When a point of contact was unavailable, manual inspection was conducted. The authors mapped the burglary points, street centerline file, and parcel data to determine geocoding strategy.

^fIncludes only burglary events for the study area used in the Microlevel Near-Repeat Burglary Experiment.

Methodological Issues for Measuring Space-Time Thresholds

This section describes methodological issues important when establishing space-time patterns for preventable near repeats. These issues include decisions related to distance measurement, time measurement, identification of originators and repeats, and assignment of repeats to originators. Each of these decisions influences whether an event is counted as a near repeat.

Distance measurement

Distance is typically quantified using one of three different methods: Euclidean, Manhattan, or Street (also known as network distance). Euclidean distance is measured as the crow flies. Manhattan distance (also called taxicab distance) is measured by combining two straight lines connected at a right angle. Street distance is measured along a street network. The method used to measure distance determines the size of the spatial buffer around an originator, which in turn affects the likelihood that events will be identified as part of a near pair. Figure 1 shows how the distance between the same two crime events varies depending on whether Euclidean (1,297 ft), Manhattan (1,643 ft), or Street distance (1,682 ft) is used. If the spatial threshold is 1,300 ft, the events would be identified as near repeats only for Euclidean distance. Measurements using Euclidean distance typically identify the highest number of near-repeat pairs, followed by measurements made using Manhattan distance and Street distance. This variance has implications both for identifying the strength of the near-repeat pattern and for accurately identifying the geographic extent of higher risk. This study explores the differences in the number of burglary incidents identified as near repeats when using Manhattan and Street distance calculations.

Figure 1.

Methods of measuring distance.

Time measurement

The measurement of time is also important to the identification of high-risk space-time windows in a few different ways. First, many burglary events are time span crimes; burglaries typically occur sometime in the interval between when a person leaves home and returns. Thus, it is often impossible to know when the burglaries occurred. Cases in which days or weeks elapse before discovery pose two potential problems. If the burglary not discovered for days or weeks is the first in a series, related burglaries may occur before the original burglary event is discovered and reported to police. Those related burglaries were not preventable because the law enforcement agency did not know about the original burglary. Thus, when considering preventable burglaries, it is important to begin counting burglaries from the date reported.

Another factor related to time is the temporal bandwidth used to define near repeats. Most previous near-repeat studies have found significant near-repeat risk remaining for a month after the originator. A common strategy when examining a near-repeat pattern is to define the spatial bandwidth as 7 days (1 week) and then examine patterns across an array of temporal bands. The number of temporal bands depends on how long the near-repeat pattern is expected to last. Analysts typically experiment with a variety of different numbers of temporal bands (Ratcliffe, 2009). In Example A, Event Pair I (Table 2 and Figure 2), Burglaries 4 and 9 are within the spatial bandwidth and do not occur on the same day but are outside the temporal bandwidth, so they do not qualify as a near-repeat pair.

Figure 2.

Graphical examples in Table 2.

Table 2.

Identifying Preventable Near Repeats.

Event pair ID	Event identifier	Event identifier	Within distance?	Different day?	Within time?	Events not previously assigned as near repeat?	Near repeat?
Example A
I	4	9	Yes	Yes	No	Yes	Not a near repeat
Example B
II	2	3	Yes	No	Yes	Yes	Not a near repeat
Example C
III	5	6	Yes	Yes	Yes	Yes	Near-repeat pair
IV	5	7	Yes	Yes	Yes	Yes	Near-repeat pair
V	5	8	Yes	Yes	No	Yes	Not a near repeat
VI	6	7	Yes	Yes	Yes	No	Not a near repeat
VII	6	8	Yes	Yes	No	No	Not a near repeat
VIII	7	8	Yes	Yes	No	Yes	Not a near repeat
Example D
IX	1	10	No	Yes	No	Yes	Not a near repeat

Temporal bandwidths are also important when discussing crime prevention strategies to interrupt near-repeat patterns. If near-repeat patterns display temporal bandwidths of several days, swift action will be necessary to interrupt them. On the other hand, patterns that emerge over longer periods of time allow more opportunity for the delivery of crime prevention programs while there are still likely to be preventable burglaries. Operationally, there are limits to how quickly an agency can respond, write a report, geocode, and make an event available for spatial analysis. The utility of interrupting near-repeat patterns as a crime prevention strategy must be considered in light of practical limits on how long it takes for the event to become known to the police and available for analysis.

Identifying originators and assigning follow-on events

This study applied a variety of spatial and temporal bandwidths to identify near repeats (both originators and follow-on events) in historical crime data. The temporal ordering of events was the primary mechanism for determining both originators and follow-on events. Originators had a subsequent burglary within a specified temporal and spatial bandwidth. Follow-on events occurred after an originator and within the specified space-time window.⁵ The NR-CPPC was used to read the retrospective data files in the order in which burglaries were reported to police. The program evaluated each burglary sequentially. Table 3 provides an overview of the distance measurement functionality in the NRC and the NR-CPPC.

Table 3.

Comparison of Functionality: NRC and NR-CPPC.

Parameter	NRC	NR-CPPC
Distance
Euclidean	Yes	Yes
Manhattan	Yes	Yes
Street	No	Yes^a
Time
Days	Yes	Yes
Identifying status
Allow events to be repeats for multiple events	Always	User controlled
Allow events to be both originators and repeats	Always	User controlled
Allow events on the same day to count as a repeat	Always^b	User controlled
Statistical significance
In-application significance testing	Yes^c	No

Note. NRC = near-repeat calculator; NR-CPPC = Near Repeat Crime Prevention Potential Calculator.

^aNetwork analysis can be on a user-defined network shapefile or on data provided by OpenStreetMap.

^bThe NRC counts events occurring on the same day, but the impact of those events is constrained to a single column in the space-time contingency table.

^cStatistical significance testing is conducted by a user-selectable number of Monte Carlo simulations.

The focus on preventable near-repeat events requires several additional constraints not present when calculating global near-repeat risk (Table 2 and Figure 2). First, events occurring on the same day were not counted. If two events occur on the same day, the earlier one becomes the originator and the other is not counted. In Example B (Event Pair II), both Burglaries 2 and 3 were reported on January 18, so no near-repeat burglary is counted (Table 2 and Figure 2).

Second, events could be associated with only one originator. In Example C, Burglary 7 was within the spatiotemporal window for Burglary 5 and counted as a near repeat (Event Pair IV). When Burglary 6 was evaluated as a potential originator for Burglary 7 (Event Pair VI), it did not count as another near repeat because Burglary 7 had already been allocated as a near-repeat event of Burglary 5.

Third, events could not count as both originators and repeats (Table 2 and Figure 2). In Example C, Burglaries 6 and 7 (Event Pair VI) would not constitute a near-repeat pattern because they were designated as part of an earlier near-repeat pattern. These choices allow each event to count only one time toward the quantification of the near-repeat problem. By implementing these three constraints, the NR-CPPC produces the most conservative estimate of near-repeat crime prevention potential. These constraints are user selectable in the NR-CPPC; the most permissive settings produce results that mirror the evaluation criteria of the NRC.

Analytic approach

The NR-CPPC was used to automate the process of quantifying the potential impact of implementing a near-repeat burglary prevention program for each jurisdiction. Findings from past literature indicate that the strongest near-repeat patterns have generally been identified within two blocks and one month. To provide consistency with the literature and with the space-time window used in the microlevel near-repeat burglary prevention experiment, we examined the proportions of burglaries that were near repeats using Manhattan distance for two blocks (244 m/800 ft), three blocks (366 m/1,200 ft), and four blocks (488 m/1,600 ft) (Table 4). To identify the extent to which the identification of near-repeat burglaries was sensitive to Manhattan compared with Street distance measures, we compared the two using a two-block (244 m/800 ft, approximately two city blocks on a typical grid street pattern) spatial window and a variety of 7-day windows up to 1 month (0–7, 0–14, 0–21, 0–28, and 0–35). The percentage of the total events for each threshold combination was calculated and used to estimate the crime prevention potential (Table 5).

Table 4.

Comparison of Near-Repeat Crime Prevention Potential Using Manhattan Distance.

Reporting agency	Burglaries	Two blocks, 28 days, % (number)	Two-block, rank	Three blocks, 28 days, % (number)	Three-block rank	Four blocks, 28 days, % (number)	Four-block rank
Philadelphia	8,716	13.38 (1,166)	3	24.70 (2,153)	3	38.31 (3,339)	1
Seattle	7,567	12.59 (953)	4	19.98 (1,512)	4	31.02 (2,347)	4
St. Louis	3,670	13.68 (502)	2	24.80 (910)	2	37.71 (1,384)	2
Durham	2,580	10.08 (260)	6	18.49 (477)	5	28.84 (744)	5
Denver	2,495	9.71 (286)	7	17.62 (519)	6	27.10 (798)	6
Orlando	2,122	21.11 (448)	1	28.79 (611)	1	37.13 (788)	3
Fayetteville	1,958	10.37 (203)	5	16.65 (326)	7	25.89 (507)	7
Baltimore Co.	611	4.26 (26)	10	7.36 (45)	10	10.15 (62)	10
Santa Rosa	522	8.05 (42)	8	12.07 (63)	8	17.62 (92)	8
Redlands	361	6.65 (24)	9	11.08 (40)	9	16.34 (59)	9

Note. Cities are ranked by number of burglaries.

Table 5.

Comparison of Manhattan and Street Network Distance for Two Blocks (244 m/800 ft) From 7 to 35 Days.

	0–7 days		0–14 days		0–21 days		0–28 days		0–35 days
	Manhattan	Network	Manhattan	Network	Manhattan	Network	Manhattan	Network	Manhattan	Network
Baltimore Co.	1.47 (9)	1.15 (7)	1.96 (12)	1.47 (9)	3.27 (20)	2.78 (17)	4.26 (26)	3.60 (22)	5.24 (32)	4.42 (27)
Denver	2.75 (81)	1.97 (58)	5.23 (154)	3.77 (111)	7.37 (217)	5.50 (162)	9.71 (286)	7.27 (214)	11.71 (345)	8.59 (253)
Durham	2.83 (73)	1.98 (51)	5.39 (139)	4.22 (109)	7.95 (205)	6.05 (156)	10.08 (260)	7.75 (200)	12.44 (321)	9.73 (251)
Fayetteville	3.58 (70)	3.17 (62)	5.67 (111)	4.90 (96)	8.12 (159)	7.05 (138)	10.37 (203)	8.99 (176)	12.97 (254)	10.78 (211)
Orlando	6.64 (141)	4.29 (91)	12.21 (259)	7.63 (162)	16.73 (355)	10.79 (229)	21.11 (448)	13.90 (295)	25.49 (541)	17.62 (374)
Philadelphia	3.77 (329)	2.77 (241)	6.95 (626)	5.31 (463)	10.53 (918)	8.17 (712)	13.38 (1166)	10.46 (912)	16.45 (1434)	12.82 (1117)
Redlands	0.83 (3)	0.28 (1)	3.88 (14)	3.05 (11)	5.26 (19)	3.88 (14)	6.65 (24)	4.43 (16)	9.70 (35)	6.09 (22)
Santa Rosa	2.68 (14)	3.07 (16)	5.36 (28)	6.13 (32)	6.90 (36)	7.28 (38)	8.05 (42)	9.00 (47)	9.20 (48)	10.34 (54)
Seattle	4.48 (339)	2.55 (193)	6.78 (513)	3.24 (245)	9.59 (726)	4.78 (362)	12.59 (953)	6.40 (484)	14.02 (1061)	7.18 (543)
St. Louis	3.87 (142)	2.81 (103)	6.95 (255)	4.55 (167)	10.22 (375)	6.70 (246)	13.68 (502)	8.80 (323)	16.84 (618)	10.60 (389)

Note. This table shows the strength of the near-repeat pattern at two blocks over five different time periods. The fields show the percentage of near-repeat burglaries, with the total number in parentheses. The percentage represents the proportion all burglaries in the jurisdiction that occurred within the space-time window. All totals produced using street information from OpenStreetMap rather than a user-supplied shapefile. Results for other distances are available from first author.

To deploy operational programs, agencies need more analysis to identify where a problem exists. As a natural first step, simple descriptive analyses, including computing the concentration of near repeats, were used to evaluate which operational units accounted for most of the problem. Spatial analysis was used to drill down further to answer the question of where law enforcement action might yield the greatest crime prevention value.

Spatial patterns of the near-repeat events were examined using ArcMap desktop v 10.5. A variety of methods are available to examine the spatial patterns and map clusters of near-repeat burglary events (Eck, Chainey, Cameron, Leitner, & Wilson, 2005). A kernel density surface was used here because it is one of the most accessible, visually appealing, and frequently used by crime analysts. Creating a kernel density surface allows operational personnel to visualize the areas with the higher densities of near-repeat events (the darker shaded areas on the map).

Strengths and weaknesses of the approach

There are several strengths to the approach taken here. First, the process is a computer-assisted approach to identify near-repeat patterns. It is very time-consuming to manually identify near-repeat events in official data because both spatial and temporal conditions must be satisfied for an event to qualify as a near repeat. Computer programs have a number of benefits (Groff, 2013) over manual analysis, including significantly reduced analysis time, increased accuracy and repeatability, and greater transparency. Second, the influence of the distance measurement calculation methodology on the number of near repeats identified was explicitly examined. This area of study has not had much attention in the literature. Third, the external validity of the findings was increased by using data from a variety of jurisdictions rather than a single case study. These jurisdictions varied on important aspects such as the number of burglaries and the number and density of housing units.

Most of the weaknesses of this approach involve issues with using official crime data; these issues are well known or are related to the reporting issues inherent in time span crimes such as burglaries (Chainey & Ratcliffe, 2005). It is rarely possible to identify exactly when a time span crime occurred. The victims can say with certainty only when they saw their property last and when they discovered it missing. The best-case scenario is when the amount of time between those two known moments is minutes or hours rather than days or weeks. From the perspective of identifying near repeats, time span crimes can lead to errors in temporal ordering. This potential problem with temporal ordering is well known and impossible to solve using official data or victimization data. In our study, because we were interested in preventable near repeats, we used the date the burglary was reported to the police.

The study used a convenience sample of agencies that made spatially referenced burglary data available online. These agencies cannot be presumed to be representative of all law enforcement agencies. At the same time, using open-source data allows immediate replication of the study. Future research should examine more sites and use a stratified random sample of jurisdictions with different amounts of burglary and different types of housing styles.

Results

Manhattan Distance Results

Table 4 shows the size of the preventable near-repeat burglary problem as quantified by the NR-CPPC using Manhattan distance across three spatial thresholds (approximations for two, three, and four city blocks) and one temporal threshold (28 days). Orlando, FL, had the highest percentage of burglaries that were preventable near repeats (21.1%) in a two-block/28-day space-time window. Half the sites (St. Louis, MO; Philadelphia; Seattle, WA; and Fayetteville and Durham, NC) had 10% to 14% near repeats; remaining sites had less than 10%.

Using a three-block/28-day window, we once again found that Orlando had the highest percentage of preventable near repeats, at 28.8%. Two sites, Philadelphia and St. Louis, had almost a quarter of their burglaries falling within that space-time window. Two other sites had close to 20% (Seattle and Durham) and two sites had approximately 17% (Denver, CO, and Fayetteville). Santa Rosa, CA; Redlands; and Baltimore County had the lowest proportions.

At four blocks and 28 days, Philadelphia had the highest percentage (38.3%) of preventable near-repeat burglaries. St. Louis and Orlando both had approximately 37%. Seattle, Durham, Denver, and Fayetteville all had 26% to 31%. Across all spatial bandwidths, there was consistency in the ranking of cities. Near-repeat crime prevention potential is highest in Orlando, St. Louis, Philadelphia, and Seattle and lowest in Santa Rosa, Redlands, and Baltimore County. Interestingly, the three sites with the lowest number of burglaries also had the lowest rates of near repeat, but the fifth smallest population site, Orlando, had the highest rate of near-repeat burglaries at both two and three blocks. Except for Orlando, the sites with the most burglaries also had the highest proportion of preventable near repeats. Ironically, the two cities ranked 9th and 10th at all distances, Baltimore County and Redlands, were the two sites used in the Microlevel Near Repeat Experiment. Those researchers had hypothesized that the low rates of near repeat in the experiment likely played a large role in the finding of no significant differences between treatment and control sites within each city (Groff & Taniguchi, 2019).

Comparison of Manhattan and Street Network Distances

The Microlevel Near Repeat Experiment used Street distance to calculate the incidence of near-repeat burglary; this approach was hypothesized to have reduced the number of near repeats. Table 5 compares the strength of the near-repeat pattern using Manhattan distance with that using Street distance (calculated using data from OpenStreetMap). As mentioned earlier, Street distance reflects travel patterns using the shortest path along road networks and thus provides a more accurate estimate of risk area when the locations of potential targets are also related to the road network (Rosser, Davies, Bowers, & Cheng, 2017). In addition, Street distance buffers are usually more compact. Thus, it was not surprising that the number of near repeats quantified through Street/network distance was generally lower than the number discovered through Manhattan distance.⁶

The one exception to this relationship was Santa Rosa, where Manhattan distance counted fewer near-repeat events relative to Street distance. We note that Santa Rosa was the only jurisdiction in this analysis that provided events geocoded to the nearest intersection. Data for all other agencies were geocoded to the parcel, street centerline, or the centroid of the street centerline.

We make two observations about these results. First, the type of distance measurement affects the estimation of the size of the near-repeat problem, and the size of that effect varies by city. Within each distance measure method, the difference in the proportion of crimes identified as near repeats varies considerably across the 10 sites. Second, results from Santa Rosa suggest that how events are geocoded may affect the relationship between distance measurement techniques and near-repeat counts.

We examined how much the method of distance measurement mattered by looking at the increase or decrease factor associated with using Manhattan distance compared with that associated with Street distance. Table 6 summarizes 25 different space-time windows.⁷ For 9 of the 10 jurisdictions, Manhattan distance was associated with increases of 22% to 102%, on average, over Street distance. In most sites, the percent difference between the two measures is larger at shorter distances than longer distances because the numbers of events are smaller.

Table 6.

Increase/Decrease Associated With Manhattan or Street Network Distance.

Reporting Agency	% Δ Between distance measurement techniques
Reporting Agency	Average^a	Minimum^b	Maximum^c	SD ^d	Space/Time Window of Minimum^e	Space/Time Window of Maximum^f
Baltimore Co.	23.05	6.90	42.11	7.27	610/7	366/14
Denver	25.21	8.56	50.00	14.70	610/21	122/21
Durham	25.18	10.71	43.14	7.29	122/21	244/7
Fayetteville	21.54	4.04	34.29	8.56	366/7	122/28
Orlando	43.70	5.28	104.00	33.09	610/7	122/7
Philadelphia	26.64	−2.38	109.59	31.44	610/7	122/7
Redlands	46.10	−40.00	200.00	55.15	488/7	244/7
Santa Rosa	−1.15	−12.50	11.11	7.58	244/7 and 14	610/7
Seattle	102.29	6.13	404.62	128.39	610/7	122/21
St. Louis	52.03	5.56	208.70	61.69	610/21	122/7

Note. All statistics were calculated across all 25 space-time threshold combinations.

^aThe average percentage difference between Street/network distance and Manhattan distance. Positive values indicate larger counts of events using Manhattan distance.

^bThe minimum percentage difference between Street/network distance and Manhattan distance. Positive values indicate larger counts of events using Manhattan distance.

^cThe maximum percentage difference between Street/network distance and Manhattan distance. Positive values indicate larger counts of events using Manhattan distance.

^dThe standard deviation percentage difference between street/network distance and Manhattan distance.

^eThe space-time window with the smallest difference between Street/network distance and Manhattan distance (e.g., 610/7 indicates that of all spatiotemporal bandwidths considered, 610 m and 7 days had the smallest difference between Street/network distance and Manhattan distance).

^fThe space-time window with the largest difference between Street/network distance and Manhattan distance.

Our findings suggest that the characteristics of the site where the program will be deployed matter when an agency is deciding which distance measurement to use. In cities such as St. Louis, using a 244-m (800-ft)/28-day threshold would indicate a potential to reduce burglary by 13.7% (502 burglaries) if using Manhattan distance but only 8.8% (323 burglaries) if using Street distance—a difference of 179 burglaries. However, in Fayetteville, the same space-time bandwidth would have a differential of only 1.4% (27 burglaries). This is true regardless of the scale of the intervention, because the actual differences between the crime prevention potentials measured by Manhattan and Street distances will also vary within each city. The safest course of action would be to test all three methods before deciding on the most appropriate one for the situation.

Identifying the size of the potentially preventable problem with near-repeat crime is the first step in analyzing near-repeat crime. The next step is investigating the characteristics of near-repeat burglaries and thus where and when crime prevention programs may be most effectively implemented. Some relevant factors are the absolute number of preventable near repeats, the proportion of all burglaries that are near repeats, and the amount of concentration in the geographic locations of near repeats.⁸ A final step is assessing the potential costs and benefits of a crime prevention program targeting near repeats before implementation and then again after implementation. The next section uses Philadelphia data for an example of in-depth analysis.

Further Analysis Using Philadelphia as Example

Quantifying the size of the near-repeat problem

Philadelphia data were used to provide an example of the workflow that an analyst might use. First, output from the NR-CPPC reveals that 13.4% (n = 1,166) of burglaries were preventable (Table 4). Given the scope of the near-repeat problem, an agency may conclude that a near-repeat–based intervention is worthwhile. An examination of the timing of preventable repeats citywide revealed 819 originators for the spatial bandwidth of two blocks and temporal bandwidth of 28 days (Table 7). Of those originators, 69.4% had one near repeat and 22.2% had two near repeats associated with them, suggesting that focused enforcement in 91.6% of near-repeat high-risk areas would prevent only one or two follow-on events. The next section examines whether focusing on areas with higher near-repeat rates might help increase the yield of crime prevention efforts.

Table 7.

Length of Near-Repeat Chains.

	Originators
Near repeats	Number	Percentage
1	568	69.35
2	182	22.22
3	57	6.96
4	8	0.98
5	2	0.24
6	1	0.12
7	1	0.12
Total	819	100.00

Variation in near-repeat patterns across operational units

One obvious method of moving down the cone of resolution is to explore differences by police operational units—police districts in Philadelphia. Residential burglary was unevenly distributed across Philadelphia’s 21 police districts. The top four districts (15, 2, 12, and 22) accounted for 33% of all burglaries in Philadelphia (Table 8) and 27% of all near-repeat burglaries. The same districts that had most of Philadelphia’s burglaries also had most of Philadelphia’s near-repeat burglaries, with only minor changes in ranking (to 15, 12, 22, and 2).

Table 8.

Burglary Across Districts in Philadelphia, PA, Sorted by the Percentage of Overall Burglaries Per District.

District	Total burglaries	Near-repeat burglaries (two blocks/28 days)	Percentage of all Philadelphia burglaries occurring in district	Percentage of Philadelphia near-repeat burglaries occurring in district	Percentage of district burglaries that are near repeats
15	837	126	9.60	10.81	15.05
2	705	82	8.09	7.03	11.63
12	699	115	8.02	9.86	16.45
22	620	90	7.11	7.72	14.52
14	557	67	6.39	5.75	12.03
19	549	67	6.30	5.75	12.20
24	487	72	5.59	6.17	14.78
18	450	50	5.16	4.29	11.11
39	446	57	5.12	4.89	12.78
35	438	53	5.03	4.55	12.10
26	403	60	4.62	5.15	14.89
7	378	45	4.34	3.86	11.90
25	378	43	4.34	3.69	11.38
17	346	57	3.97	4.89	16.47
3	314	45	3.60	3.86	14.33
16	243	31	2.79	2.66	12.76
8	238	18	2.73	1.54	7.56
9	211	30	2.42	2.57	14.22
5	162	21	1.86	1.80	12.96
1	129	15	1.48	1.29	11.63
6	126	22	1.45	1.89	17.46

Another way to quantify the size of the near-repeat problem is to examine the extent to which the burglary problem is associated with near-repeat patterns. Table 9 shows the same data, sorted by the percentage of district burglaries that are near repeats. In the top four districts (6, 17, 12, and 15), more than 15% of burglaries were involved in near repeats. Comparing Tables 8 and 9 illustrates the differences between ranking by the number of burglaries and ranking by the proportion of burglaries that are near repeats. For example, District 6 has a low number of burglaries (n = 22) but a high rate of near repeat among those burglaries (17.5%). The fact that high numbers of burglaries are not always associated with high numbers of near-repeat burglaries makes it prudent to examine more than one method for identifying where to implement prevention programs targeting near-repeat burglary.

Table 9.

District-Level Analysis, Philadelphia, PA, Sorted by the Percentage of That District’s Burglaries That Are Near Repeats.

District	Total burglaries	Near-repeat burglaries (two blocks/28 days)	Percentage of all Philadelphia burglaries occurring in district	Percentage of Philadelphia near-repeat Burglaries occurring in district	Percentage of district burglaries that are near repeats
6	126	22	1.45	1.89	17.46
17	346	57	3.97	4.89	16.47
12	699	115	8.02	9.86	16.45
15	837	126	9.60	10.81	15.05
26	403	60	4.62	5.15	14.89
24	487	72	5.59	6.17	14.78
22	620	90	7.11	7.72	14.52
3	314	45	3.60	3.86	14.33
9	211	30	2.42	2.57	14.22
5	162	21	1.86	1.80	12.96
39	446	57	5.12	4.89	12.78
16	243	31	2.79	2.66	12.76
19	549	67	6.30	5.75	12.20
35	438	53	5.03	4.55	12.10
14	557	67	6.39	5.75	12.03
7	378	45	4.34	3.86	11.90
2	705	82	8.09	7.03	11.63
1	129	15	1.48	1.29	11.63
25	378	43	4.34	3.69	11.38
18	450	50	5.16	4.29	11.11
8	238	18	2.73	1.54	7.56

In summary, near-repeat burglary was a significant problem across Philadelphia. All but one district had an 11% or higher rate of preventable near-repeat burglaries. Districts 15 and 12 appear in each analysis and represent the districts with the strongest pattern of near repeats. Together, they account for 17.6% of all burglaries and 20.7% of all near-repeat burglary in Philadelphia. Of course, police management may also decide to focus their efforts to prevent near-repeat burglary in particular areas within districts (i.e., those where it is most prevalent) or where areas of concentration cross police district boundaries. The next section addresses how the output from the NR-CPPC can be used to visualize the spatial distribution of near-repeat high-risk places using District 15 (15d) as an example.

Local spatial concentrations of near repeats

The spatial distribution of near-repeat events within 15d was examined using the 126 preventable near-repeat burglaries that took place in 15d plus the 13 additional preventable near-repeat burglaries that fell within 244 m/800 ft of 15d’s boundary. Including burglaries in the areas adjacent to 15d was important to understand the district burglary patterns. Examination of the spatial pattern revealed four specific areas of highest concentration that would be candidates for more in-depth near-repeat burglary prevention analysis (Figure 3(a)).⁹

Figure 3.

Comparison of near-repeat and all burglaries in one district.

Because patterns in official data are a sample of actual crime that has occurred, the next analysis included all 837 residential burglaries in 15d as well as the 106 burglaries that occurred within 800 feet of the district boundary. The map of all burglaries produced much larger hot spot areas that covered a greater proportion of the district (Figure 3(b)).¹⁰ Although the two distributions share many of the same hot spots, the map of only near repeats provides a more precise indication of where near-repeat burglary is a problem.

Additional analysis would be necessary to identify the type of near-repeat prevention program with the best chance of success. Such an analysis would require collecting additional data to examine the characteristics of burglary events (both originators and near repeats) and the social and environmental landscape of where they occurred. Previous studies have provided guidance regarding the characteristics of places that are associated with residential burglary and the characteristics of burglaries that deserve special attention. This information could be used in a problem-solving framework to develop appropriate burglary prevention activities (Clarke & Eck, 2005).

Considering the cost–benefit of a near-repeat crime prevention program

Determining the crime prevention potential also provides a basis for calculating cost–benefit analysis (Ekblom, Law, Sutton, Crisp, & Wiggins, 1996). For burglary cost, we used the national average of monetary loss to represent potential cost to the victim. Multiplying the average dollar loss by the number of preventable burglaries provides a rough estimate of potential savings. For example, in 2016, the national average loss in the United States was $2,361 per burglary (Federal Bureau of Investigation, 2017). Preventing 75 burglaries would translate to $177,075 in losses prevented (75 × $2,361 = $177,075).¹¹ The cost of a program to the department depends on the staffing (both sworn and civilian) and any equipment or other expenses required for implementation. Comparing potential costs and benefits (savings of preventing near-repeat burglaries) provides an estimate of potential savings (if outcome is positive) or potential costs (if it is negative).

Researchers have suggested enhancing this preliminary approach by developing cost–benefit estimates across varied scenarios reflecting the crime prevention intensities of various potential strategies (Bowers, Johnson, & Hirschfield, 2003, 2004; Ekblom et al., 1996). Intensity is measured as the cost of an intervention per household. If possible, costs should be estimated for a range of different interventions at different levels of intensity (e.g., volunteer-based information delivery compared with information plus directed patrol) and thus different costs to implement. Potential benefit could also be calculated for different outcomes (i.e., 0%, 25%, 50%, 75%, and 100% reductions in near repeats).

Data from Philadelphia provide an example. The top four police districts in Philadelphia experienced 2,008 residential burglaries, of which 320 were preventable near-repeat events. Using the national average loss of $2,361, it is straightforward to estimate the projected savings (benefit) from burglaries prevented across several program effectiveness levels. If a new program targeting near repeats reduced them by 25% (n = 80), the cost savings would be $188,880. If instead the program reduced near-repeat burglary by 50% or 75%, the savings would be $377,760 and $566,640, respectively. Of course, program costs reduce these figures. For example, if staff time is the only cost and it comes to $100,000, then the net savings of implementing a program that reduces near repeats by 25% is $88,880, by 50% is $277,760, and by 75% is $466,640. Having this information enables police managers to evaluate whether the size of the preventable near-repeat problem is large enough to warrant that particular expenditure of time and resources.

Discussion

Repeat and near-repeat victimization are excellent examples of research findings that have been translated into several types of crime prevention programs and delivered at different-size geographies. Recent evidence suggests that the usefulness of focusing on near repeats varies by area (Chainey et al., 2018). Specifically, at the microgeographic level of the high-risk space-time window, the rates of repeat events might be low even when there is significant space-time clustering at the jurisdiction level (Elffers et al., 2018; Groff & Taniguchi, 2019). Our research attempts to address this new aspect of near repeats in three ways. First, we proposed a new way of conceptualizing near repeats that focused on quantifying the number of near repeats that are actionable by police and how they could be used to determine the crime prevention potential of focusing on near-repeat burglary. Second, we examined whether crime prevention potential varied among U.S. cities. Third, we drilled down the geographic cone of resolution in one city to investigate the extent to which a near-repeat problem at the jurisdiction level translated into microlevel crime concentrations that could inform police interventions (Brantingham, Dyreson, & Brantingham, 1976). The quantification and use of the crime prevention potential to guide the formulation and deployment of crime prevention programs is consistent with the three transformative ideas in policing—translational criminology, evidence-based policing, and crime science.

To operationalize our measure of crime prevention potential, we used the NR-CPPC as an automated method for identifying preventable burglaries. We used the most restrictive definition possible (excluding the originator, including only those burglaries that occurred after an initial burglary was reported to the police, and allowing burglaries to be classified as only an originator or a repeat, not both), which resulted in lower counts of near-repeat burglaries.¹² This lower number has important implications for assessing whether an intervention was an effective use of police resources because it more accurately reflects actionable near repeats and thus sets more realistic expectations for the potential crime reduction that might follow a crime prevention program. We found that the percentage of actionable near-repeat burglaries varied significantly among the 10 cities. This supports our suggestion that both researchers and practitioners examine the crime prevention potential in addition to checking for significant space-time clustering using the NRC. We also found that the method of distance measurement affected the size of the actionable near-repeat problem revealed. In 9 of the 10 cities, using Street distance decreased the number of actionable near repeats within the space-time window. The one exception to this finding, Santa Rosa, used a different geocoding method, suggesting that geocoding strategy may affect the calculation of near-repeat patterns (Table 1). More research using data from other jurisdictions would be needed to confirm this finding.

In addition, we found the size of the preventable near-repeat problem varies by and within jurisdictions. This finding is consistent with Chainey et al.’s (2018) research in New Zealand, even though their study used a less conservative definition of near-repeat events. It also supports the need for examining the size of the problem before implementing a crime reduction program. Cities with high near-repeat crime prevention potential and greater numbers of near-repeat events per originator offer greater potential for crime reduction.

Taken together, these results suggest that quantifying the global level of space-time clustering as revealed by the NRC is a necessary first step to understanding near-repeat patterns. However, the presence of global clustering is not sufficient to justify a crime prevention program targeting near-repeat crime. It is possible to have significant space-time clustering at the jurisdiction level (as revealed by the NRC) but not have enough near repeats to justify deployment of police resources.

An example can clarify. The NRC showed that both Redlands and Philadelphia (Table 10; Groff & Taniguchi, 2019) had strong and significant near-repeat burglary patterns. The pattern in Redlands was higher in the 0- to 7-day and 1- to 400-ft bandwidth than it was in Philadelphia. However, reviewing the percentage of near-repeat events tells a different story. In Redlands, 6.7% of events were involved in near-repeat patterns. In Philadelphia, that percentage was twice as high, 13.4% (Table 4). This detail suggests that the strength of the near-repeat pattern alone should not be the determining factor in implementing a near-repeat–based crime prevention intervention.

Table 10.

Near-Repeat Calculator Results for Philadelphia and Redlands.

Days
	0–7	8–14	15–21	22–28	29–35
Redlands^a
Same location	3.52	0.88	1.08	2.90	3.72
1–400 ft	2.29	0.58	0.94	0.94	0.53
401–800 ft	1.09	1.07	0.89	1.46	1.29
801–1,200 ft	1.52	1.23	1.03	0.73	1.10
Philadelphia^b
Same location	5.88	2.18	1.41	1.24	1.45
1–400 ft	1.72	1.19	1.18	1.11	1.13
401–800 ft	1.09	1.04	1.05	1.03	1.03
801–1,200 ft	1.12	1.07	1.03	1.05	1.02

Note. Values in shaded cells were significant at p < .05.

^aCalculated with 2013 and 2014 residential burglary data.

^bCalculated with 2013 residential burglary data.

The NRC is based on the Knox statistic, which is a global statistic. Global statistics provide a summary across a population but do not necessarily apply to specific instances. Thus, the significant space-time threshold identified by the NRC exists for the entire study area; identifying particular places within that study area with significant near-repeat patterns is not possible. The NR-CPPC, or some other analysis of the crime prevention potential, offers a valuable second step in the analysis that enables identification of actionable microlevel concentrations.¹³

Considering crime prevention potential before developing, implementing, and evaluating programs to address near repeats has several advantages. First, such analysis can identify where near repeats are concentrated, leading to a better understanding of the characteristics associated with the near-repeat phenomenon. Second, calculating crime prevention potential more accurately grounds agency and citizen expectations for program impact in empirical reality. Third, decision makers require easy-to-understand and robust estimates from which to make decisions and evaluate programs (Figure 4). Programs with low costs can have low to moderate crime reductions and be considered successful but higher cost programs require greater crime reductions. For agencies using a problem-oriented policing framework, this workflow easily fits into the scanning or analysis stage.

Figure 4.

Likely outcome of cost and benefit interaction on perception of crime prevention program.

Finally, we suggested a simple approach to cost–benefit analysis. It did not consider the other costs or benefits that crime prevention efforts may experience. For example, the Microlevel Near Repeat Experiment reported considerable positive impact on both community–police and citizen volunteer–agency relationships (Groff & Taniguchi, 2019). Quantifying these kinds of intangible benefits is more challenging. We also ignored opportunity costs for agencies engaging in these programs. If agencies are deciding between conducting a near-repeat–based program or implementing an alternative program, there may be other impacts on crime to consider. Nevertheless, routine use of a cost-based approach to selecting crime prevention interventions is overdue (Groff & Taniguchi, 2019). Given the difficulty of quantifying these other factors, a more simplistic approach is warranted as a first important step in establishing a mindset of development, implementation, and evaluation.

These evidence-based estimates provide a realistic metric for evaluating program success. By comparing the program cost with the program benefit, decision makers can easily identify the potential for success. For example, a program that costs little but provides little crime prevention value is likely to be discontinued. At the other end of the spectrum, a program that costs a great deal and offers a great deal of crime prevention value may be considered a success but be unsustainable.

Implications for Future Research

For researchers, understanding the crime prevention potential allows for the development of research protocols that are more appropriately powered or designed. Future researchers may want to use the crime prevention potential to influence the design of microlevel interventions. Instead of targeting intervention and evaluation efforts based on the space-time bandwidths with significant global near-repeat crime patterns, researchers and practitioners can instead target areas with the highest concentration of near-repeat events. This approach was used successfully by Santos and Santos (2015a, 2015b) to focus on microlevel flare-ups of residential burglary. Others may want to use recently developed space-time tools to proactively identify microlevel intervention areas before an initial burglary takes place.¹⁴

Finally, future work is needed to tailor projects even further on the basis of results from ongoing evaluations. It may be possible, for example, to improve the efficiency of treatment delivery by identifying the events that are most likely to result in near-repeat events or, even better, the events that are most likely to result in multiple near-repeat events. Event characteristics (e.g., how the burglary was carried out) or environmental factors (e.g., housing density, housing type, other land use patterns) may help us better understand which events will eventually turn into originators. Evidence from a recent study finds that emergence of near-repeat patterns is positively related to social disorganization (as measured by concentrated disadvantage, mobility, population density, and racial heterogeneity) and to crime generators (such as ATMs and banks; Piza & Carter, 2017).

Conclusion

The findings of this research illustrate the need for researchers and practitioners to carefully quantify the likely crime prevention value of near-repeat interventions targeted at the microlevel. Identifying that a global near-repeat pattern exists is necessary but not sufficient to quantify the likely crime prevention value. Once the problem is quantified, an informed decision can be made about whether focusing on near-repeat crime is likely to yield enough benefits to justify the cost. If so, an appropriate intervention should be developed, conducted, and evaluated. Having a clear idea of the size of the problem will help with developing a realistic evaluation because it gives the police department and the community a good idea of the maximum potential crime reduction. The examples in this article demonstrate how this process could be applied to residential burglary to establish how much an agency should spend given different levels of intervention effectiveness. Previous research has demonstrated that near-repeat patterns exist for a wide variety of crime types; thus, consideration of crime prevention potential is likely to be relevant when assessing and evaluating prevention aimed at other crime types. Within the near-repeat crime framework, calculation of the crime prevention potential provides police agencies with an alternative measure for use in analyzing the problem and in designing and evaluating the success of reactive intervention programs.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grant no. 2012-IJ-CX-0039 to the Police Foundation from the National Institute of Justice.

Notes

Author Biographies

Elizabeth Groff is a professor in the Department of Criminal Justice at Temple University where she is also associated faculty in Center for Security and Crime Science. Dr. Groff has spent the last 25 years applying geographic theory and methodology to the study of crime-related issues. Her research interests include geographic criminology, police practices, and the use of technology in policing. One area of focus has been on developing evidence to improve police practice. She is a fellow of the Academy of Experimental Criminology and the Chair of the American Society of Criminology’s Division of Communities and Place.

Travis Taniguchi is a research criminologist in Policing Research Program at RTI International. His research interests include crime and place, police effectiveness, and strategies to improve the health and wellness of law enforcement officers. He currently serves as an executive counselor for the American Society of Criminology, Division of Policing.

References

Anderson

Chenery

Pease

(1995). Biting back: Tackling repeat burglary and car crime. London, England: Home Office Police Research.

Bowers

K. J.

Johnson

S. D.

(2016). Situational Prevention. In Weisburd

Farrington

D. P.

Gill

(Eds.), What works in crime prevention and rehabilitation: Lessons from systematic reviews (pp. 111–135). New York, NY: Springer New York.

Bowers

K. J.

Johnson

S. D.

Hirschfield

(2003). Pushing back the boundaries: New techniques for assessing the impact of burglary schemes. London, England: The Home Office.

Bowers

K. J.

Johnson

S. D.

Hirschfield

(2004). The measurement of crime prevention intensity and its impact on levels of crime. British Journal of Criminology, 44, 419–440. doi:10.1093/bjc/azh020

Braga

A. A.

(2007). Effects of hot spots policing on crime. Campbell Systematic Reviews, 2007(1). doi:10.4073/csr.2007.1

Brantingham

P. J.

Dyreson

D. A.

Brantingham

P. L.

(1976). Crime seen through a cone of resolution. American Behavioral Scientist, 20(2), 261–273. doi:10.1177/000276427602000207

Chainey

Curtis-Ham

S. J.

Evans

R. M.

Burns

G. J.

(2018). Examining the extent to which repeat and near repeat patterns can prevent crime. Policing: An International Journal of Police Strategies & Management, 41(5), 608--622. doi:10.1108/PIJPSM-12-2016-0172

Chainey

Ratcliffe

J. H.

(2005). GIS and crime mapping. London, England: John Wiley and Sons.

Chenery

Holt

Pease

(1997). Biting back II: Reducing repeat victimisation in Huddersfield ( Vol.2). London, England: Home Office Police Research Group.

10.

Clarke

R. V.

Eck

(2005). Crime analysis for problem solvers in 60 steps. Washington, DC: U.S. Department of Justice, Office of Community Oriented Policing Services.

11.

Eck

J. E.

Chainey

Cameron

J. G.

Leitner

Wilson

R. E.

(2005). Mapping crime: Understanding hotspots. Washington, DC: National Institute of Justice.

12.

Ekblom

Law

Sutton

Crisp

Wiggins

(1996). Safer cities and domestic burglary. London, England: Home Office, Research and Statistics Directorate.

13.

Elffers

Peeters

M. P.

van der Kemp

Beijers

(2018). Quasi-experimental evaluation of near repeat patrolling: The Amstelveen experiment. Amsterdam, The Netherlands: Netherlands Institute for the Study of Crime and Law Enforcement. doi:10.13140/RG.2.2.33339.31524

14.

Federal Bureau of Investigation. (2017). 2016 crime in the United States: Burglary. Retrieved from https://ucr.fbi.gov/crime-in-the-u.s/2016/crime-in-the-u.s.-2016/topic-pages/burglary

15.

Fielding

Jones

(2012). ‘Disrupting the optimal forager’: Predictive risk mapping and domestic burglary reduction in Trafford, Greater Manchester. International Journal of Police Science & Management, 14, 30–41. doi:10.1350/ijps.2012.14.1.260

16.

Forrester

Chatterton

Pease

Brown

(1988). The Kirkholt burglary prevention project, Rochdale. London, England: Home Office.

17.

Groff

E. R.

(2013). Measuring a place’s exposure to facilities using geoprocessing models: An illustration using drinking places and crime. In Leitner

(Ed.), Crime modeling and mapping using geospatial technologies ( Vol.8, pp. 269–295). New York, NY: Springer.

18.

Groff

E. R.

Taniguchi

(2019). Using citizen notification to interrupt near repeat residential burglary patterns: The micro-level near repeat experiment. Journal of Experimental Criminology. doi:10.1007/s11292-018-09350-1

19.

Johnson

S. D.

Bernasco

Bowers

K. J.

Elffers

Ratcliffe

J. H.

Rengert

Townsley

(2007). Space–time patterns of risk: A cross national assessment of residential burglary victimization. Journal of Quantitative Criminology, 23, 201–219. doi:10.1007/s10940-007-9025-3

20.

Johnson

S. D.

Bowers

K. J.

(2004). The burglary as clue to the future. European Journal of Criminology, 1, 237–255. doi:10.1177/1477370804041252

21.

Johnson

S. D.

Bowers

K. J.

Birks

D. J.

Pease

(2009). Predictive mapping of crime by ProMap: Accuracy, units of analysis, and the environmental backcloth. In Weisburd

Bernasco

Bruinsma

G. J.

(Eds.), Putting crime in its place (pp. 171–198). New York, NY: Springer.

22.

Johnson

S. D.

Davies

Murray

Ditta

Belur

Bowers

K. J.

(2017). Evaluation of Operation Swordfish: A near-repeat target-hardening strategy. Journal of Experimental Criminology, 13, 505–525. doi:10.1007/s11292-017-9301-7

23.

Laub

J. H.

(2011). Strengthening NIJ: Mission, science and process. NIJ Journal, 268, 1–21.

24.

Laycock

(2005). Defining crime science. In Smith

M. J.

Tilley

(Eds.), Crime science: New approaches to preventing and detecting crime (pp. 3–24). Devon, England: Routledge.

25.

Manning

Vorsina

Wong

Johnson

Tilley

(2017). Cost benefit tool. Coventry, England: College of Policing. Retrieved from http://whatworks.college.police.uk/Research/Pages/Cost_Benefit_Tool.aspx

26.

Moreto

W. D.

Piza

E. L.

Caplan

J. M.

(2014). “A plague on both your houses?”: Risks, repeats and reconsiderations of urban residential burglary. Justice Quarterly, 31(6), 1102–1126. doi:10.1080/07418825.2012.754921

27.

Piza

E. L.

Carter

J. G.

(2017). Predicting initiator and near repeat events in spatiotemporal crime patterns: An analysis of residential burglary and motor vehicle theft. Justice Quarterly, 35, 842–870. doi:10.1080/07418825.2017.1342854

28.

Ratcliffe

J. H.

(2009). Near repeat calculator (version 1.3). Philadelphia, PA: Temple University.

29.

Ratcliffe

J. H.

Rengert

G. F.

(2008). Near-repeat patterns in Philadelphia shootings. Security Journal, 21, 58–76. doi:10.1057/palgrave.sj.8350068

30.

Rosser

Davies

Bowers

Johnson

Cheng

(2017). Predictive crime mapping: Arbitrary grids or street networks? Journal of Quantitative Criminology, 33, 569–594. doi:10.1007/s10940-016-9321-x

31.

Santos

R. G.

Santos

R. B.

(2015a). Practice-based research: Ex post facto evaluation of evidence-based police practices implemented in residential burglary micro-time hot spots. Evaluation Review, 39, 451–479. doi:10.1177%2F0193841X15602818

32.

Santos

R. B.

Santos

R. G.

(2015b). Examination of police dosage in residential burglary and residential theft from vehicle micro-time hot spots. Crime Science, 4(1), 27–39.

33.

Sherman

L. W.

(1998). Evidence-based policing. Washington, DC: Police Foundation. Retrieved from http://www.policefoundation.org/wp-content/uploads/2015/06/Sherman-1998-Evidence-Based-Policing.pdf

34.

Short

M. B.

D’Orsogna

M. R.

Brantingham

P. J.

Tita

G. E.

(2009). Measuring and modeling repeat and near-repeat burglary effects. Journal of Quantitative Criminology, 25(3), 325–339. doi:10.1007/s10940-009-9068-8

35.

Stokes

Clare

(2019). Preventing near-repeat residential burglary through cocooning: Post hoc evaluation of a targeted police-led pilot intervention. Security Journal. 32(1), 45–62

36.

Thompson

Townsley

Pease

(2008). Repeat burglary victimisation: Analysis of a partial failure. Irish Journal of Psychology, 29, 131–139. doi:10.1080/03033910.2008.10446279

37.

Townsley

Homel

Chaseling

(2000). Repeat burglary victimisation: Spatial and temporal patterns. Australian & New Zealand Journal of Criminology, 33, 37–63. doi:10.1177/000486580003300104

38.

Townsley

Homel

Chaseling

(2003). Infectious Burglaries. A Test of the Near Repeat Hypothesis. British Journal of Criminology, 43(3), 615–633. doi:10.1093/bjc/43.3.615

39.

Weems

J. R.

(2014). Testing PCSO cocooning of near repeat burglary locations (Unpublished master’s thesis). Wolfson College, Cambridge, UK. Retrieved from https://www.crim.cam.ac.uk/global/docs/theses/jim-weems.pdf/at_download/file

40.

Wellsmith

Birks

D. J.

(2008). Research on target: A collaboration between researchers and practitioners for a target hardening scheme. International Review of Law, Computers & Technology, 22, 181–189.

41.

Zhang

Zhao

Ren

Hoover

(2015). Space–time clustering of crime events and neighborhood characteristics in Houston. Criminal Justice Review, 40(3), 340–360. doi:10.1177/0734016815573309