151

Article

The Predictive Policing Challenges of Near

Repeat Armed Street Robberies
Cory P. Haberman and Jerry H. Ratcliffe

Abstract New research methodologies like the near repeat phenomenon provide police with a potentially powerful
predictive technique, if law enforcement possesses the capacity to capitalize on identified patterns in time. The current
study examines armed street robbery data from Philadelphia in order to identify and quantify the existence of
multiple-event near repeat chains. The impact of near repeat chains on the temporal stability of micro-level armed
street robbery hot spots is also explored. The findings demonstrate that near repeat armed street robbery chains tend to
be relatively short in terms of chain length, and the number of days between the initiation and termination of a chain is
rarely longer than 7 days. These results suggest that if police are to proactively address short-term crime event
predictions, a range of complex organizational and analytical capacities have to be in place. Furthermore, despite
the fact that a number of hot spots were found to be primarily derived of near repeat events, the results show that the
temporal stability of armed street robbery hot spots is not associated with the proportion of near repeat events within
the hot spots, a finding supportive of long-term opportunity reduction measures.

Introduction timely and accurate intelligence followed by effect-

A common ethos amongst street-level police com- ive tactics that are deployed rapidly (McDonald,
manders and patrol officers is that the police should 2002). To move beyond a generalized response to
focus on emerging crime problems (Buerger, 2010). crime requires a range of organizational capacities;
It is not surprising then that police departments the ability to monitor crime events frequently, the
have traditionally engaged in ‘whack-a-mole’ analytical skill to identify emerging trends, a
(Ratcliffe, 2008) or ‘fire-brigade’ (Tilley, 2003) decision-making system adept at demanding
policing where calls for service are dealt with in differential responses, and an operational capacity
rapid succession and a high dosage of police pres- to quickly implement new tactics.
ence is assumed to be an effective policing tactic Other strategies, such as problem-oriented poli-
(Eck and Maguire, 2000). The basic principles of cing, also begin with the identification of crime
Compstat (for example) attest to this by including patterns (Eck and Spelman, 1987); however, the


Cory P. Haberman, Department of Criminal Justice, Temple University, Philadelphia, PA, USA.
E-mail: [email protected]

Jerry H. Ratcliffe, Department of Criminal Justice, Temple University, Philadelphia, PA, USA.
E-mail: [email protected]

Advance Access publication: 2 May 2012

Policing, Volume 6, Number 2, pp. 151–166
doi:10.1093/police/pas012
ß The Authors 2012. Published by Oxford University Press. All rights reserved.
For permissions please e-mail: [email protected]
152 Policing Article C. P. Haberman and J. H. Ratcliffe

proliferation of computer crime mapping and from another person on the street. The term does
Compstat has not resulted in a shift from flooding not refer to armed robberies at banks, petrol
areas where crimes have recently occurred with stations, or other locations, and does not include
police personnel to more holistic strategies such robberies that do not involve the use of a weapon.
as problem solving as many policing reformers We are therefore dealing with the most potentially
had hoped (Weisburd et al., 2003, p. 443–444). lethal form of robbery. Specifically, this study quan-
Thus, saturation patrol (or equivalents) remains tifies the extent of armed street robbery in
the first port of call for many Basic Command Philadelphia at the event-level by identifying and
Unit (BCU) commanders focusing on emerging describing how near repeat armed street robberies
crime patterns. This ‘cops on corners’ approach occur in multiple-event chains. Second, the impact
also assumes that crime is spatially and temporally of near repeat armed street robberies on the
concentrated to the extent that a quick response can temporal stability of micro-level armed street
efficiently disrupt crime, even though research robbery hot spots is examined. The study concludes
examining the spatial and temporal concentration with a discussion of the results’ implications for the
of crime is still developing. Therefore, we can add to allocation of police resources. This article adds to
the list of organizational requirements (above) that the existing literature in three ways. First, it is the
the crime type under examination should also first near repeat study of armed street robberies.
display spatial and temporal characteristics that Second, by examining the multiple-event chains,
are amenable to a rapid, saturation patrol preventa- it provides empirical clarity to the temporal char-
tive effort. In other words, this requires the crime in acteristics of emergent robbery patterns. Third and
question to have predictable patterns that are perhaps most importantly, it does all of this within
reasonably constrained spatially (so patrols do not the context of an operational policing response,
have to spread too far) and last long enough in time comparing the quantitative findings with a range
for there to be time to organize a suitable response. of viable police reactions.
Recent research into the near repeat phenom-
enon may provide some of this required analytical
power. It has long been known that some victims
What do we know about near
and places are repeatedly victimized (Bowers et al.,
repeats?
1998; Farrell and Pease, 1993; Polvi et al., 1991).
This understanding has been enhanced by the The near repeat phenomenon originated as an
discovery of a near repeat phenomenon: not only epidemiological concept to study the transmission
are locations at risk of repeat victimization, but of infectious diseases (Knox, 1964). More recently,
nearby locations are also at increased risk of environmental criminologists have incorporated
crime up to a certain distance and for a certain epidemiologists’ space–time concept and method-
time (Bowers and Johnson, 2004; Ratcliffe and ology in the study of crime. The idea is that a pre-
Rengert, 2008; Townsley et al., 2003). vious crime event creates a heightened risk of
The present study examines patterns of armed victimization for spatially proximate targets that
street robberies in Philadelphia (Pennsylvania), decays over time (Johnson et al., 2007a; Johnson
USA, in order to develop a more robust under- and Bowers, 2004b; Ratcliffe and Rengert, 2008).
standing of the extent to which violent crime con- Simply stated, the near repeat phenomenon shows
centrates in space and time simultaneously. For this that crime clusters in space and time. Using
study, armed street robbery refers to crimes known epidemiological methods, recent research has
as ‘stick-ups’: a person with a weapon, predomin- found support for the near repeat crime hypothesis.
antly a firearm, demanding cash or other valuables While researchers proposed hypotheses about why
Near Repeat Armed Street Robberies Article Policing 153

near repeat crime patterns would exist, early concentrated among those targets regardless of
research focused on simply determining whether whether it is the work of the same offender.
near repeat crime patterns existed. Townsley and Most studies have tested the boost hypothesis. A
colleagues (2003) demonstrated that a greater study using data from Merseyside, UK, found that
number of burglary event pairs in Queensland, burglaries occurring close in space and time (near
Australia, were clustered within a distance of repeat pairs) were more likely to have been carried
200 m and 2 months than would be expected on out with the same modus operandi than burglaries
the basis of chance. A similar near repeat burglary at greater distances in space and time (Bowers and
pattern was also found in Merseyside, UK, at dis- Johnson, 2004). This finding suggested that the
tances of 400 m and 2 months (Johnson and same offender(s) returned to the area to use a pre-
Bowers, 2004a). Furthermore, an analysis of data viously successful burglary tactic on other houses,
from 10 separate cities from five different countries perhaps because the original victim implemented
(Australia, Netherlands, New Zealand, UK, and crime prevention measures. Similarly, near repeat
US) revealed near repeat burglary patterns were burglaries from the Netherlands were found to be
ubiquitous. The results showed that near repeat cleared to the same offender more often than
burglary patterns were consistent across all studied non-near repeat burglaries (Bernasco, 2008).
locations; burglary point pairs were clustered Finally, Johnson et al. (2009b) found that pairs of
within distances of about 100 m and 2 weeks burglaries as well as thefts from motor vehicles were
(Johnson et al., 2007a). Finally, a near repeat ana- more likely to be cleared to the same offender(s)
lysis conducted using burglary and theft from when the events occurred closer in space and time
motor vehicles (TFMV) data from Bournemouth, than events that occurred farther apart in space and
UK, found a near repeat pattern for both the burg- time. On the other hand, support for the flag hy-
lary events (400 m and 6 weeks) and TFMV events pothesis was found in Queensland, Australia, by
(400 m and 6 weeks) separately, but a modified comparing the level of near repeat burglary in sub-
analysis that examined burglary and TFMV simul- urbs with homogenous and heterogeneous housing
taneously failed to find any evidence of a bivariate stocks, and it was determined that a greater number
space–time concentration (Johnson et al., 2009). of near repeat events occurred in the suburbs with a
After determining that near repeat property crime homogenous housing stock than the suburbs with a
patterns existed, more recent studies have focused heterogeneous housing stocks (Townsely et al.,
on directly testing why. Two hypotheses have been 2003).
presented: 1. the ‘boost’ and 2. ‘flag’ hypotheses. The There have been only a few studies conducted to
‘boost’ hypothesis suggests that past victimization determine if violent crimes follow a near repeat
boosts the likelihood of future victimization pattern. Ratcliffe and Rengert (2008) applied the
(Farrell et al., 1995; Johnson, 2008; Pease, 1998). near repeat phenomenon to shootings in
This event-dependency hypothesis is an offender- Philadelphia. While the authors did not have data
based dynamic where it is argued that the same of- to directly test their theory, they used a body of
fender (and/or colleagues) returns to the area of a literature suggesting that inner city violence often
previous offence to capitalize on the opportunities results in retaliation to frame their rationale for
the offender learned about during the previous expecting a near repeat shooting pattern. Using
offence. Conversely, the ‘flag’ hypothesis argues operational knowledge of the nature of gun
that target risk factors concentrate opportunity violence from the Philadelphia Police Department
and, as a result, concentrate crime (Farrell et al., to guide the spatial and temporal parameters of
1995; Johnson, 2008; Pease, 1998). In other words, their analysis, a near repeat pattern was found at
the attractiveness of targets means crime will be the distances of about one block and 2 weeks.
154 Policing Article C. P. Haberman and J. H. Ratcliffe

Wyant and colleagues (in press) examined a Near repeats and street robbery
bivariate near repeat phenomenon for illegal fire- Adapted versions of the boost and flag hypotheses
arm carrying (Violations of the Uniform Firearms (Farrell et al., 1995; Johnson, 2008; Pease, 1998) for
Act (VUFA)) and shootings from 2004–07 in armed street robbery underpin the present study.
Philadelphia. They worked from two hypotheses: First, it is plausible that near repeat armed street
1. arresting people for illegally carrying firearms robbery can also be explained by an offender-based
would suppress later shootings and 2. shooting boost process. In short, the success of a previous
events would result in increased police presence event most likely teaches the offender that a general
and thereby increase VUFA arrests in an area. In location provides quick escape routes and suitable
concordance with their hypotheses, shootings were: targets that lack adequate guardianship; traits
1. found to significantly decrease anywhere from 28 reported in ethnographic research by street robbers
to 47% after a VUFA arrest (the effect varied by as necessary for successful street robbery (St. Jean,
police district) and 2. VUFA arrests were found to 2007; Wright and Decker, 1997). Because active
increase for about a week and up to a distance of street robbers typically live the ‘fast life’, which
about 0.2 miles after a shooting. includes partying, drug use, and gambling, their
Finally, although not a traditional street crime, a lifestyle creates a constant need for cash and likely
near repeat pattern has also been demonstrated for drives continuous offending (Wright and Decker,
improvised explosive device (IED) attacks in Iraq 1997). When an offender decides to hunt for rob-
(Townsley et al., 2008). Townsley and colleagues bery victims in the area where he or she was previ-
(2008) framed their study in the context of rational ously successful, these continuous acts result in the
choice, assuming that terrorists would carry out space–time concentration of armed street robbery.
attacks close in space and time in order to minimize Alternatively, the flag hypothesis can also be reit-
the amount of effort exerted to carry out an attack. erated for armed street robbery. Simply stated, the
After analysing the locations and dates of 916 IED characteristics of certain areas may provide
attacks occurring in a 3-month time frame, it was increased levels of street robbery opportunities
revealed that the greatest risk for a future IED attack within a specific temporal rhythm. Wright and
was within the distances of about 1 km and 2 weeks Decker’s (1997) interviews with active street
after a previous attack. The authors concluded that robbers found that the interviewees preferred
this supported their assumption that insurgent locations where people were likely to be carrying
attacks involved rational planning. cash and provided access to quick escape routes.
For all this research, the practical policing Specifically, the street robbers noted preferences
benefits from this area of study are still embryonic. for areas with automatic teller machines (ATM),
Studies exploring the near repeat nature of violent check cashing businesses, supermarkets, and
crimes are scarce. Since recent research has shown shopping malls. Similarly, ethnographic research
that the near repeat nature of crime can be used for in Chicago’s Wentworth neighbourhood with
short-term crime forecasting (Bowers et al., 2004; both offenders and police officers found that areas
Johnson et al., 2007b, 2009a), understanding the with distracted persons carrying cash were the most
near repeat nature of violent crime may have likely place for a robbery to take place (St. Jean,
value for proactive policing and crime prevention. 2007). St. Jean (2007) argued that these areas have
Therefore, in this study we focus on identifying a ‘ecological (dis)advantage’, while environmental
near repeat armed street robbery pattern and criminologists would describe these locations as
developing a more robust understanding of the crime generators or crime attractors (Brantingham
extent to which it occurs. and Brantingham, 1993).
Near Repeat Armed Street Robberies Article Policing 155

The current literature establishes that, for some crime prevention efforts, such as redesigning the
crimes, there are more point-pairs close in space physical environment and a greater allocation of
and time than would be expected on the basis of crime prevention resources than compared to
chance in official crime data, but little is known temporally unstable hot spots, or in other words,
about the extent to which these patterns occur a short-term crime outbreak (Johnson et al., 2008).
(see, Townsley, 2007, for a notable exception). A hot spot created by a multiple-event near repeat
Understanding the extent to which near repeat chain actually only represents a short-term, geo-
victimization occurs at the event level may be graphically concentrated risk even though it may
useful for the allocation of crime prevention appear as an area in need of police resources
resources. If near repeat victimization occurs in during retrospective data analysis. Because the tem-
extensive multiple-event chains then allocating re- poral instability of hot spots is one of the more
sources to disrupt these chains will be a useful crime serious critiques against hot spots policing
prevention strategy. On the other hand, if these (Rosenbaum, 2006), understanding how the near
chains are short-lived then it is probably more repeat process impacts the temporal stability of
beneficial and practically feasible to focus on hot spots has significant import for the allocation
long-term crime problems rather than chasing of police resources.
crime outbreaks. By quantifying the extent of the In summary, the research literature is still sparse
armed street robbery near repeat phenomenon at regarding patterns of near repeats in violent, or
the event level, this study will begin to shed light on potentially violent, crime. Furthermore, there is
whether focusing on multiple-event near repeat next-to-nothing known about the temporal length
armed street robbery chains can be an efficient of near repeat chains, and whether they persist long
crime prevention strategy. enough for there to be a viable policing response.
Finally, if the near repeat process results in These questions are addressed in the remainder of
multiple-event near repeat chains, then it is possible this article.
that near repeat chains contribute to the formation
of crime hot spots. Crime hot spots are abstract
entities that are difficult to parsimoniously define
Methodology
(Buerger et al., 1995; Taylor, 2010), but have gen- Data
erally been thought of as places—’addresses, build- The present study uses 2009 armed street robbery
ings, block faces, street segments, or clusters of event data from the City of Philadelphia.
addresses’ (Mastrofski et al., 2010, p. 251)—with Philadelphia is located in the north-eastern region
a greater than average concentration of crime of the US, and the nearly 1.5 million residents of
(Chainey and Ratcliffe, 2005, pp. 241–245; Eck Philadelphia make it the fifth most-populated city
and Weisburd, 1995; Sherman et al., 1989). While in the country.1 Philadelphia’s population is pre-
an accumulating body of successful hot spot poli- dominantly made up of African American (43.5 %)
cing evaluations has led to significant policy discus- and white (42.5%) residents with the remaining
sions (Mastrofski et al., 2010), scholars have 14% of citizens consisting of a fair proportion of
discussed the importance of understanding the Hispanic residents. The city’s median household
temporal stability of crime hot spots before design- income of $36,222 is almost $16,000 below the na-
ing strategies focused on hot spots (Johnson et al., tional level (US Census Bureau, 2006–08).
2008; Ratcliffe, 2004b). In short, hot spots that are The Philadelphia Police Department (PPD)
stable over time will likely require more complex polices an area of roughly 150 square miles with
1
According to the 2010 US Census.
156 Policing Article C. P. Haberman and J. H. Ratcliffe

an authorized force of 6,600 sworn officers new location using a random number generator.
(Philadelphia Police Department, 2010). In 2009, After each simulation, the space–time distances
Philadelphia reported a total of 9,037 robberies for all points in the simulated dataset are
for a rate of 584 robberies per 100,000 residents re-measured and the cell frequencies in the contin-
compared to the national rate of 133 robberies gency table are recalculated. Statistical significance
per 100,000 (Federal Bureau of Investigation, can then be determined by computing the number
2009). The present study, however, only focuses of times the observed cell frequency exceeded the
on armed street robberies or robberies in which expected cell frequency values for all of the simula-
the offender used a deadly weapon (most frequently tions. The likelihood of near repeat victimization
a firearm) to forcefully take someone else’s property for each space–time distance pair can be calculated
on a city street. Of the 3,611 total armed robberies by dividing the observed cell frequency by the mean
recorded in 2009, 3,556 contained adequate data for of the expected cell frequencies. Values below 1
geocoding (98.5% geocoding hit rate) and are the indicate repeat victimization is less likely than
subject of this study (Ratcliffe, 2004a). expected on the basis of chance and values above
1 indicate repeat victimization is more likely than
Identifying near repeat patterns expected on the basis of chance. By subtracting the
Near repeat scholars have employed virtually the Knox ratio from 1, the effect size can be interpreted
same methodology,2 the Knox method (Knox, as an increased/decreased percentage of near repeat
1964), to identify the space–time clustering of victimization likelihood, similar to the interpret-
events. The first step of the Knox method is to ation of an odds ratio (for alternative explan-
measure spatial and temporal distances between ations, see Johnson et al., 2007a; Ratcliffe and
each event and every other event within the dataset. Rengert, 2008).
The total number of space–time distance measure- A free computer program that automates the
ments will equal n(n-1)/2, where n is the total methodologically and computationally rigorous
number events in the dataset. Next, the researcher Knox method is currently available online
specifies the spatial and temporal bandwidths that (Ratcliffe, 2009). In order to test for a near repeat
will be used construct a contingency table. These pattern, the program requires the input of a cor-
parameters are placed on the X and Y axis of the rectly formatted data set with the XY-coordinates
contingency table and the total number of and date of occurrence for the events of interest.
point-pairs within each cell of the contingency The user then specifies the spatial and temporal
table, a space–time distance interaction, is calcu- bandwidths, the geographic distance measurement
lated. A Monte Carlo simulation is then used to technique to be used, and the statistical significance
create an expected distribution of cell frequencies level (which determines the number of Monte
in order to determine if the observed cell frequen- Carlo simulations used). Manhattan distance and
cies are greater than would be expected on the basis a P-value < 0.001 (999 Monte Carlo simulations) is
of chance. For each Monte Carlo simulation, the used in the present study.
spatial locations of the events are held constant The spatial bandwidth for this analysis is 400 ft,
while the dates of the events are reassigned to a the average length of a city street block in
2
Earlier studies (Townsely et al., 2003; Johnson and Bowers, 2004a) did not utilize the Monte Carlo simulation technique to
create an expected distribution of cell frequencies and determine statistical significance. Johnson et al. (2007) introduced this
modification to avoid the fact that the original Knox method violated the assumption of independent observations by
assuming ‘that in the absence of contagion, the statistical distribution of the expected values for the cells of the Knox
[contingency] table would conform to a Poisson distribution, and can be computed using the marginal totals of the table’
(p. 208).
Near Repeat Armed Street Robberies Article Policing 157

Philadelphia (McCord and Ratcliffe, 2009; Ratcliffe assigned to a chain with a unique identifier, the
and Rengert, 2008). The street block is considered number of originating points and near repeat
the unit of analysis that best depicts variations in points within each chain were summed and com-
crime across the urban landscape (Taylor, 1997). pared. Since each near repeat should have a preced-
Practically speaking, the term ‘street block’ is in ing originator, the totals for these two categories for
the vocabulary of many and provides a simple heur- each chain should be equal. This was confirmed.
istic for interpreting and applying the results. The Descriptive statistics describing the extent of the
temporal bandwidth used in the present analysis is 7 near repeat chains were then calculated.
days. The 7-day bandwidth supplies both a short
timeframe as well as a conceptually simple measure Near repeat chains and the temporal
for interpreting, understanding, and applying the stability of armed street robbery hotspots
results. Since police commanders are typically The final phase of this analysis was designed to
focused on emerging crime problems (Buerger, examine the extent to which near repeat armed
2010), a short temporal bandwidth will make the street robberies impact the temporal stability of
results practically appealing, especially for depart- armed street robbery hot spots. The hierarchical
ments with Compstat-type organizational responses nearest neighbour (HNN) clustering routine avail-
to crime (Weisburd et al., 2003). able in CrimeStat v3.2 (Levine, 2009) was used to
identify micro-level armed street robbery hot spots
Identifying near repeat armed street in the 2009 Philadelphia armed street robbery data.
robbery chains HNN creates an output of clusters at different
The second phase of the analysis involved connect- orders. Only first-order clusters are examined in
ing near repeat armed street robbery events into the present analysis because first-order clusters are
multiple-event chains. Near repeat events were event-based or ‘hot spots’ of events; whereas,
considered part of the same chain if the temporal higher-order clusters (i.e. second-order or third-
and spatial distances between events were within order) are clusters of clusters formed by aggregating
the statistically significant spatial and temporal par- clusters at a lower order into larger clusters (Levine,
ameters established by the Knox analysis. The Near 2009). First-order clusters are also a more micro-
Repeat Calculator provides an additional feature to level hot spot and, more theoretically, representative
examine which events from the data set are points of the variation in crime levels typically found
within a near repeat pattern (Ratcliffe, 2009). These within larger areal units (Andersen and Malleson,
results are provided in a new data file that reports 2011). A researcher must specify the minimum
each event’s XY-coordinates, date of occurrence, number of events that each first-order cluster
and counts for the number times the point was an must encompass. Events are then grouped on the
originator in a near repeat pair, or a repeat in a near basis that the spatial distance between each event
repeat pair (i.e. the second incident). A total of 888 and its nearest neighbour is shorter than would be
(25% of all geocoded armed street robberies) near expected under the assumption of complete spatial
repeat armed street robbery points were identified randomness (Eck et al., 2005). In the present
as being part of an event pair. The near repeat data analysis, first-order clusters are specified to have at
file was displayed in a Geographic Information least 10 armed street robbery events.
System (GIS) and the related near repeat points In order to examine how near repeat events
were linked. In the present analysis, any events impact the temporal stability of micro-level hot
occurring within 7 days and 1200 feet of another spots, a homogeneity index, commonly used in
near repeat event were aggregated into the same social science research to summarize the
chain. After each near repeat event had been distribution of data across nominal categories
158 Policing Article C. P. Haberman and J. H. Ratcliffe

(Blau, 1977; Chainey and Ratcliffe, 2005; Gibbs and near repeat analysis are displayed in Table 1. As
Martin, 1962), was calculated. The homogeneity expected, the increased chance for a near repeat
P
index was computed using the formula: h = 1- pi2 armed street robbery to occur after an originating
where pi is the proportion of data within each cat- robbery diminishes as the distance and time from
egory of the nominal variable of interest. The an originating event increases. Specifically, the
homogeneity statistic is bounded by a maximum Knox ratio of 1.80 for the 0 to 7 days and 1 to
of 1-1/ni, where ni is the total number of observed 400 feet cell in Table 1 indicates that a subsequent
categories, and a minimum of 0. A value close to the armed street robbery is 80% more likely to occur
maximum indicates the data are heterogeneous or within one block and 1 week after an initial armed
equally dispersed across the observed categories. A street robbery than if a near repeat pattern was not
value of 0 indicates complete homogeneity or that identified. The Knox ratio of 1.31 within the same 0
the data are entirely concentrated within just one to 7 days time frame and 401 to 800 feet cell indi-
category. cates that a near repeat armed street robbery is 31%
In the present study, each hot spot’s homogeneity more likely on the second block away and within
index was calculated using the count of armed street the week following the originating event. Finally,
robberies that occurred within each of the thirteen the increased likelihood for a subsequent armed
(n = 13) 28-day intervals that made up the 2009 cal- street robbery to occur after an originating event
endar year. A hot spot with the maximum value decreases to only about 16% more likely than if a
(0.923) would indicate the events within the hot near repeat pattern was not identified at the dis-
spot were equally dispersed across all of the 28-day tance of about 2 to 3 blocks (801 to 1200 feet)
intervals for 2009 and a value of zero would indicate and within the same 0 to 7 day time frame after
that robberies within the hot spot occurred within the originating event (though still statistically sig-
just one of the 28-day intervals for 2009. The rela- nificant). In total, there is a greater likelihood for a
tionship between the temporal stability statistic subsequent armed street robbery event to occur
(homogeneity index) and the proportion of near within 1200 feet and 7 days of an initial armed
repeat events within each hot spot is reported street robbery event than would be expected if a
using a Pearson’s correlation coefficient. space–time interaction for armed street robberies
was not identified within the data.3

Results The extent of near repeat armed street

The near repeat nature of armed street robbery chains
robbery The second phase of this analysis focused on aggre-
A statistically significant near repeat armed street gating close pairs of armed street robberies into
robbery pattern was identified. The results from the multiple-event chains in order to quantify the
3
Additional analyses were performed using slightly different spatial and temporal parameters: (1) 400 ft and 14 days; (2)
800 ft and 7 days; and (3) 800 ft and 14 days. In sum, the results were not substantively different, but less sharp than the
analysis reported above. In all three analyses, it was found that increasing the temporal and spatial parameters slightly
increased the geographic extent and temporal length of the near repeat pattern, but only in the contingency table cells
that also included the events from the 400 ft and 7 days analysis. For example, in the 400 ft and 14 days analysis, the cell for
events 400 ft and 14 days was significant, but this cell also contained all events from the statistically significant 400 ft and 7
days cell in the analysis discussed above. More importantly, the spatial and temporal parameters of 400 ft and 7 days
(discussed in the findings) provided the most robust results: all Knox likelihood ratios were larger than those from the
other analyses. Because the strongest effects were found in the 400 ft and 7 day analysis and those parameters provide the most
conceptually simple and practically useful results, they are presented above. Nonetheless, results from the sensitivity analyses
can be provided upon request.
Near Repeat Armed Street Robberies Article Policing 159

Table 1: Armed street robbery near repeat analysis: Knox ratios

Time 0 to 7 days 8 to 14 days 14 to 21 days 22 to 28 days More than 28 days

Distance
Same location 2.49** 1.89** 1.18 0.89 0.90
1 to 400 feet 1.80** 0.89 0.85 1.11 0.97
401 to 800 feet 1.31** 1.01 0.86 0.99 0.99
801 to 1200 feet 1.16* 1.03 0.95 1.00 0.99
1201 to 1600 feet 1.09 1.02 1.01 0.94 1.00
More than 1600 feet 1.00 1.00 1.00 1.00 1.00**

*P < 0.05; **P < 0.001. Source: Philadelphia Police Department, 2009.

extent to which near repeat armed street robbery Table 2: Descriptive statistics for near repeat armed
occurred at the event level. The 888 individual street robbery chains
armed street robbery events identified as part of Number of days chain
the near repeat armed street robbery pattern in persisted (Risk Time)a
the first part of the analysis were aggregated into a
Chain Number Total Min. Max. Mean SD
total of 363 near repeat chains. Table 2 displays event of events in
descriptive statistics for the near repeat armed length chains all chains

street robbery chains. For the most part, near 2 251 502 1 6 3.18 1.64
repeat armed street robberies occurred in close 3 79 237 1 12 5.47 2.57
4 21 84 3 14 8.29 2.95
pairs. In fact, roughly 57% (n = 502) of the individ-
5 8 40 4 17 10.75 5.04
ual near repeat armed street robbery events identi- 6 3 18 8 13 11.00 2.65
fied were either an originator or near repeat event in 7 1 7 —b — — —
one of 251 close pairs identified. The second most All chains 363 888 1 17 4.20 2.77
frequently identified chain length was 3-event near Note that each event in a chain is within 1200 feet and 7 days of at
repeat armed street robbery chains. In total, least one other event in the chain. aRisk time is total number of days
that a near repeat chain persisted, counting the first event day as day
roughly 27% (n = 237) of the individual near one, and continuing to the last event day (inclusive); bOnly one chain
repeat armed street robbery events identified 7 events long was identified, so descriptive statistics are not displayed.
The risk time for the 7-event chain was 11 days. Source: Philadelphia
made-up 79 3-event near repeat armed street Police Department, 2009.
robbery chains. About 17% (n = 149) of the indi-
vidual near repeat armed street robbery events
identified were part of the 33 chains identified generally increased as the chain event length
having 4 or more events, distributed thus: 4-event increased. Nonetheless, the average risk time for
chains (n = 21), 5-event chains (n = 8), 6-event all chains was still only 4.2 days, and 89.5% of all
chains (n = 3), and a chain with 7 events (n = 1). chains, regardless of event length, expired within 7
In Table 2, what we refer to as the risk time is days or less. Simply put, the risk time variable
determined as the period in days during which a reflects the fact that most chains contained very
near repeat chain persisted. In other words, risk few events that, by definition of the near repeat
time counts the first event day as day one, and con- phenomenon, occurred within close temporal
tinues to the last event day inclusively. The short proximity to each other. While armed street
event length of the near repeat chains was reiterated robberies occurred close in both space and time
in terms of risk time. As depicted in Table 2, when more often than would be expected on the basis
comparing the mean risk time for chains of varying of chance, the extent of the heightened risk in the
event lengths individually, the average risk time nearby area is minimal at the event level.
160 Policing Article C. P. Haberman and J. H. Ratcliffe

Near repeats and the temporal stability of street robbery events within the armed street
armed street robbery hot spots robbery hot spots represented only about 9% of
A preliminary examination of maps showing the all geocoded armed street robberies but about
distribution of all armed street robbery events 36% of all near repeat armed street robbery events
against all near repeat armed street events suggested that occurred in Philadelphia during 2009. The
that the distribution of near repeat armed street locations of the robbery hotspots are shown in
robbery events closely mirrored the spatial distri- Fig. 1.
bution of all armed street robbery events in the city. Table 3 shows that the proportion of near repeat
Thus, the final phase of this study was designed to events to total armed street robberies within each
examine how near repeat armed street robbery hotspot ranged from 0 to 0.8. The majority (&60%;
events contributed to the development and subse- n = 31) of armed street robbery hotspots contained
quent temporal stability of armed street robbery a proportion of near repeat armed street robberies
hotspots. HNN clustering was employed to identify to total armed street robberies of .5 or less. On the
micro-level armed street robbery hot spots (Levine, other hand, a total of 21 (&40%) armed street
2009). A total of 52 first-order clusters were identi- robbery hotspots were identified where >50% of
fied (Fig. 1). Descriptive statistics for the 52 the total events within the cluster were part of a
first-order hotspots are displayed in Table 3. The near repeat process. The temporal stability statistics
52 first-order clusters contained 20.75% (n = 738) ranged from 0.72 to 0.90 with the average equalling
of all geocoded 2009 armed street robberies, but 0.84 and a standard deviation of 0.04. The consist-
merely 2.73% (n = 589) of Philadelphia street inter- ently high values across all 52 hot spots indicated
sections. Roughly 43% (n = 317) of the total armed that the robbery events that formed the hot spots
street robberies within the 52 hot spots were part of were similarly dispersed over multiple 28-day inter-
a near repeat chain. The 317 near repeat armed vals throughout the year. The scatter plot in Fig. 2

Figure 1: Armed street robbery hotspots, Philadelphia, PA, 2009.

Near Repeat Armed Street Robberies Article Policing 161

depicts the relationship between the proportion of derived mostly from near repeat events experienced
total armed street robberies that were near repeat the same temporal pattern as the hot spots derived
events within each hotspot and the temporal stabil- of only a few or no near repeat events. This finding
ity statistics (r = -0.16; P-value non-significant). suggests that even if a hot spot was driven by near
The scatter plot visually demonstrates the tempor- repeat events, those near repeat events were spread
ally stable patterns across the hot spots, regardless out over the duration of 2009 rather than the result
of near repeat composition. In short, the hot spots of one short armed street robbery outbreak.

Table 3: Descriptive statistics for first-order armed Discussion

street robbery clusters (n = 52)
This study applied the near repeat phenomenon to
Min. Max. Mean Median Standard
deviation
a previously unexamined crime type, armed street
robbery, and focused on quantifying the extent at
Robberies 10 29 14.19 12.00 4.17
Near repeat 0 17 6.10 5.50 3.86 which the near repeat phenomenon occurs. At least
robberies in Philadelphia, it is statistically more likely for an
Proportion of 0 0.8 41.30 44.13 19.80 armed street robbery to occur within about three
near repeat
robberies city blocks (1200 feet) and 1 week (7 days) of a
Temporal stability 0.72 0.90 0.84 0.86 0.04 previous armed street robbery than compared to a
Area (sq. miles) 0.012 0.043 0.026 0.026 0.007 random spatio–temporal distribution. Linking near
Intersections 4 22 11.33 11 4.16 repeat armed street robberies into chains of
Miles of street 0.917 3.08 1.76 1.74 0.50
multiple near repeat events, however, revealed
Source: Philadelphia Police Department, 2009. that the heightened risk to nearby targets created

Figure 2: Relationship between the temporal stability statistic and the proportion of near repeat events for each
armed street robbery hot spot (n = 52).
162 Policing Article C. P. Haberman and J. H. Ratcliffe

by a previous armed street robbery event was short Furthermore, the crime type in question must
lived. In fact, roughly 90% of near repeat armed have a near repeat pattern that persists with
street robbery chains terminated after just 2 or 3 sufficient temporal length to still be viable once
events and 7 days or less. The short extent of the the organization has mustered a response.
near repeat armed street robbery chains also trans- In the case of near repeat armed street robbery,
lated to the fact that individual armed street the vast majority of the chains lasted less than 7 days
robbery outbreaks never led directly to the forma- and only 38 of the 363 chains persisted beyond
tion of any armed street robbery hot spots. The 7 days.4 Of the 154 events in these 38 chains, only
majority of armed street robbery hot spots were 58 actual repeat events occurred beyond 7 days
determined to be predominantly comprised of iso- (from an initial sample of 3,556). This provides
lated armed street robbery incidents, but even those the police with a considerable intelligence and
hot spots primarily consisting of near repeat events organizational challenge, at least with regard to
(21 of the 52) were found to be temporally stable armed street robberies. In the case of Philadelphia,
across 2009. In other words, those near repeat 30% (112/363) of new near repeat pairs will
chains occurred in multiple instances throughout continue to have another event, which is a reason-
2009 rather than in one long chain of events or able crime prevention opportunity; however, it
concentrated chains of events. must be recognized quickly as the overall pattern
The statistically significant near repeat pattern rarely lasts beyond a week. This draws into question
suggests that past armed street robberies can be the value of Compstat-type meetings that are held
useful for forecasting the occurrence of future weekly and biweekly, lacking as they do the currency
armed street robberies. This provides police with to identify and react with sufficient flexibility. Many
two options; a tactical short-term response and a Compstat meetings are conducted with data that are
strategic long-term response. Unfortunately, the at least a day or two out of date in order to allow
evidence suggests a tactical short-term response crime analysts time to prepare the (often compen-
would be difficult to muster within the time frame dious) books of statistics that accompany the
of even the longest near-repeat chains. Consider the meeting. This delay, while inevitable, further adds
analytical and organizational capabilities that have to the argument that the responses that emanate
to occur before police can capitalize on the near from a Compstat meeting may be starting too late
repeat strategy. The organization must have: to be effective.
 a surveillance mechanism adequate enough to The evidence in this article instead promotes the
monitor crime events with sufficient necessity for a rapid assessment of crime patterns
frequency; conducted at the local level. Given the nascent
state of ‘predictive policing’ (Johnson et al., 2009a;
 an analytical regime capable of recognizing a
Tompson and Townsley, 2010), a 30% chance that a
chain of events quickly and against a back-
near repeat pair will have a follow-on within a week
ground noise of unrelated crimes;
and within 1,200 feet is a strong finding with real
 a decision-making framework capable of iden- proactive potential; however, to capitalize on this
tifying the need for, and coordinating, a suit- opportunity requires a real-time analytical capacity,
able tactical response; and vigilant local mid-level command staff, or sophisti-
 the operational flexibility to adapt to changing cated automated systems capable of alerting com-
conditions and implement a new tactic. manders to an emergent crime fighting possibility.
4
These 38 chains included 3-event chains (n = 17), 4-event chains (n = 12), 5-event chains (n = 6), 6-event chains (n = 3), and
a 7-event chain (n = 1).
Near Repeat Armed Street Robberies Article Policing 163

Given this context, prospective hot spotting The finding that some temporally stable hot
techniques could be applied to armed street rob- spots experience a relatively high number of near
bery. One prospective hot spotting technique repeat events also provides new insight for design-
(ProMap), that uses a revised kernel density equa- ing strategies to address hot spots in general. Clarke
tion to give more weight to recent crime events, has and Eck (2003) explained the importance of under-
been demonstrated to be more predictive of future standing the underlying process fuelling a hot spot
crime events than traditional kernel density estima- before attempting to design and implement a
tion maps (see Bowers et al., 2004; Johnson et al., crime-prevention strategy and it is well-docu-
2007b, 2009a). Further prospective mapping mented that problem-solving efforts commonly
techniques that combine long-term risk heterogen- fail because the involved parties fail to fully analyse
eity with near repeat patterns are underway by and understand the problem (Bullock et al., 2006).
researchers at Temple University’s Center for Because the processes driving hot spots with an
Security and Crime Science. The implementation abundance of near repeat events are likely different
and use of these tools to effectively prevent crime, than the processes driving hot spots of mostly iso-
however, will likely need to be accompanied by lated events, the current findings suggest that crime
fundamental organizational changes. prevention planners should be aware of the targeted
The finding that the armed street robbery hot hot spot’s near repeat composition before design-
spots were temporally stable regardless of their ing a crime-prevention strategy. In short, the crime
near repeat composition throughout 2009 suggests prevention strategy undertaken to address a hot
that the underlying processes contributing to the spot of isolated events might be different than a
heightened opportunity for armed street robbery strategy designed to address a hot spot of predom-
within each of the hot spots also remained constant inantly near repeat events.
across the study period. Though the chains were
short-lived and would be difficult to address Limitations
unless the organizational capacities in the previous Although this study has answered a number of
paragraph existed, more holistic strategies designed research questions, it has also illustrated the need
to reduce crime opportunities in the more tempor- for additional research examining the near repeat
ally stable armed street robbery hot spots are likely phenomenon. Although this study was framed by
to be more fruitful than chasing short-term crime two hypotheses (boost and flag), available data did
outbreaks. There is preliminary evidence to suggest not permit directly testing them. Testing these
that the same places within a city experience high hypotheses will likely require using a number of
levels of crime over the long periods of time (Braga different methodologies and access to related
et al., 2011; Weisburd et al., 2004). Therefore, con- offender information. For example, the boost
sidering that it has been shown that changes in hypothesis might be supported by using official
crime levels at small geographies can impact arrest data to determine if strings of near repeat
city-wide crime levels (Ratcliffe, 2010) and the crimes are more likely to be cleared to the same
growing body of empirical research demonstrating persons than non-near repeat crimes (Bernasco,
the effectiveness of different hot spot policing tac- 2008; Bowers and Johnson, 2004; Johnson et al.,
tics in crime hot spots (for general summaries, see, 2009b). In addition, interviews with street robbers
Braga, 2005; Lum et al., 2011; for a specific ex- about target selection and offending frequency
ample, see, Ratcliffe et al., in Press), it seems to might also provide important insight on the boost
make more sense for police commanders to focus hypothesis. Alternatively, data on the physical
resources at the hot spots-level versus the environment might be used to predict the spatial
event-level. patterning of near repeat events in order to support
164 Policing Article C. P. Haberman and J. H. Ratcliffe

the flag hypothesis. Finally, simulation and math- stability of hot spots suggests that chasing
ematical modelling might be used to test both the- short-term outbreaks may not be the best use of
ories individually and simultaneously (Groff, 2007, police resources for addressing armed street rob-
2008; Johnson, 2008; Pitcher and Johnson, 2011). bery. In the future, crime scientists may be able to
This study is also limited in that only one crime develop predictive models to aid police depart-
type, armed street robbery, was examined; albeit ments in formulating crime prevention strategies
reducing armed street robbery is of considerable focused on short-term changes in crime, but the
concern to most police organizations. Examining current body of literature on the stability of spatial
near repeat chains for other crime types may pro- crime patterns, the empirical evaluations of poli-
duce different results with different implications for cing tactics, and the findings of this study still pro-
crime prevention and policing. The present analysis vide overall support for more holistic strategies that
should be replicated using data from other locations focus on reducing armed street robbery opportu-
before any police departments begin using the near nities in long-term armed street robbery hot spots.
repeat nature of armed street robbery to make op- Predictive policing, while remaining largely
erational decisions. Furthermore, the temporal sta- undefined, is still mostly an analytical challenge
bility statistic in this study suffers from a modifiable and curiosity for crime analysts and computational
temporal unit problem that is similar to the modi- scientists. It currently lacks integration with oper-
fiable areal unit problem (MAUP) that many geo- ational policing as found with more holistic and
graphic studies face. While using different temporal established frameworks such as problem-oriented
bounding units, such as months, did not substan- policing and intelligence-led policing. The evidence
tively change the results of this study, future re- from this article has one clear implication: to
search should continue to develop new statistics capitalize on the opportunities provided by predict-
for assessing temporal stability. ive analytics requires a range of complex organiza-
tional capabilities to be in place. Without
simultaneously and explicitly considering the
Conclusion analytic and organizational structures necessary to
implement the findings of predictive regimes,
While the present study identified a near repeat police agencies will not be able to effectively utilize
gunpoint robbery pattern, the results suggest that predictive research, and frustration and disillusion-
a range of challenging organizational capacities are ment will result.
required to capitalize on this pattern. These organ-
izational challenges include possessing the analyt-
ical capacity to identify a pattern, the leadership References
mechanism to direct a new strategy, and the oper- Andersen, M. A. and Malleson, N. (2011). ‘Testing the
ational flexibility to rapidly respond, all within a Stability of Crime Patterns: Implications for Theory and
timeframe of usually less than 1 week from the Policy.’ Journal of Research in Crime and Delinquency
48(1): 58–82.
start to the end of the crime series. With these capa-
Bernasco, W. (2008). ‘Them Again? Same-Offender
cities in place, the nearly one-third of near repeat Involvement in Repeat and Near Repeat Burglary.’
armed street robbery pairs that will have a subse- European Journal of Criminology 5(4): 411–431.
quent event within a week and about three city Blau, P. M. (1977). Inequality and Heterogeneity:
blocks has the potential to aid proactive policing A Primitive Theory of Social Structure. New York:
Free Press.
efforts.
Bowers, K. J., Hirschfield, A. and Johnson, S. D. (1998).
That being said, exploring the influence of near ‘Victimization Revisited.’ British Journal of Criminology
repeat armed street robbery chains on the temporal 38(3): 429–452.
Near Repeat Armed Street Robberies Article Policing 165

Bowers, K. J. and Johnson, S. D. (2004). ‘Who Commits Gibbs, J. P. and Martin, W. T. (1962). ‘Urbanization,
Near Repeats? A Test of the Boost Explanation.’ Western Technology, and the Division of Labor: International
Criminology Review 5(3): 12–24. Patterns.’ American Sociological Review 27(5): 667–677.
Bowers, K. J., Johnson, S. D. and Pease, K. (2004). Groff, E. R. (2007). ‘Simulation for Theory Testing and
‘Prospective Hot-Spotting: The Future of Crime Experimentation: An Example Using Routine Activity
Mapping.’ British Journal of Criminology 44(5): 641–658. Theory and Street Robbery.’ Journal of Quantitative
Braga, A. A. (2005). ‘Hot Spots Policing and Crime Prevention: Criminology 23(2): 75–103.
A Systematic Review of Randomized Controlled Trials.’ Groff, E. R. (2008). ‘Adding the Temporal and Spatial
Journal of Experimental Criminology 1: 317–342. Aspects of Routine Activities: A Further Test of Routine
Braga, A. A., Hureau, D. M. and Papachristos, A. V. (2011). Activity Theory.’ Security Journal 21: 95–116.
‘The Relevance of Micro Places to Citywide Robbery Johnson, S. D. (2008). ‘Repeat Burglary Victimisation: A
Trends: A Longitudinal Analysis of Robbery Incidents at Tale of Two Theories.’ Journal of Experimental
Street Corners and Block Faces in Boston.’ Journal of Criminology 4(1): 215–240.
Research in Crime and Delinquency 48(1): 7–32. Johnson, S. D., Bernasco, W., Bowers, K. J., Elffers, H.,
Brantingham, P. L. and Brantingham, P. (1993). Ratcliffe, J. H., Rengert, G. and Townsley, M. (2007a).
‘Criminality of Place: Crime Generators and Crime ‘Space–Time Patterns of Risk: A Cross National
Attractors.’ European Journal of Criminal Policy and Assessment of Residential Burglary Victimization.’
Research 3: 1–26. Journal of Quantitative Criminology 23: 201–219.
Buerger, M. E. (2010). ‘Policing and Research: Two Cultures Johnson, S. D., Birks, D. J., McLaughlin, L., Bowers, K. J.
Separated By an Almost-Common Language.’ Police and Pease, K. (2007b). Prospective Crime Mapping in
Practice and Research 11(2): 135–143. Operational Context. London: Home Office.
Buerger, M. E., Conn, E. G. and Petrosino, A. J. (1995). Johnson, S. D. and Bowers, K. J. (2004a). ‘The Burglary as
‘Defining the “Hot Spots of Crime”: Operationalizing Clue to the Future: The Beginning of Prospective Hot
Theoretical Concepts for Field Research.’ In Eck, J. E. Spotting.’ European Journal of Criminology 1(2): 237–255.
and Weisburd, D. L. (eds), Crime and Place. Mosney, Johnson, S. D. and Bowers, K. J. (2004b). ‘The Stability of
NY: Willow Tree Press, pp. 237–257. Space–Time Clusters of Burglary.’ British Journal of
Bullock, K., Erol, R. and Tilley, N. (2006). Problem-Oriented Criminology 44(1): 55–65.
Policing and Partnerships: Implementing an Evidence- Johnson, S. D., Bowers, K. J., Birks, D. J. and Pease, K. (2009a).
Based Approach to Crime Reduction. Collumpton, UK: ‘Predictive Mapping of Crime by Promap: Accuracy, Units
Willan. of Analysis, and the Environmental Backcloth.’ In
Chainey, S. and Ratcliffe, J. H. (2005). GIS and Crime Weisburd, D., Bernasco, W. and Bruinsma, G. (eds),
Mapping. London, UK: Wiley. Putting Crime in its Place: Units of Analysis in Geographic
Eck, J. E., Chainey, S., Cameron, J. G., Leitner, M. and Criminology. New York, NY: Springer, pp. 171–198.
Wilson, R. E. (2005). Mapping Crime: Understanding Johnson, S. D., Lab, S. P. and Bowers, K. J. (2008). ‘Stable
Hot Spots. Washington, DC: National Institute of Justice. and Fluid Hot Spots of Crime: Differentiation and
Eck, J. E. and Maguire, E. (2000). ‘Have Changes in Policing Identification.’ Built Environment 34(1): 32–45.
Reduced Violent Crime? An Assessment of the Evidence.’ Johnson, S. D., Summers, L. and Pease, K. (2009b). ‘Offender
In Blumstein, A. and Wallman, J. (eds), The Crime Drop as Forager? A Direct Test of the Boost Account of
in America. New York: Cambridge University Press, Victimization.’ Journal of Quantitative Criminology 25:
pp. 521–565. 181–200.
Eck, J. E. and Spelman, W. (1987). Problem Solving: Knox, E. G. (1964). ‘Epidemiology of Childhood Leukaemia
Problem-oriented Policing in Newport News. Washington in Northumberland and Durham.’ British Journal of
DC: Police Executive Research Forum. Preventive and Social Medicine 18: 18–24.
Eck, J. E. and Weisburd, D. L. (1995). ‘Crime Places in Levine, N. (2009). CrimeStat v3.2. Washington, D.C.:
Crime Theory.’ In Eck, J. E. and Weisburd, D. L. (eds), National Institute of Justice.
Crime and Place. Monsey, NY: Criminal Justice Press, Lum, C., Koper, C. S. and Telep, C. W. (2011). ‘The
pp. 1–33. Evidence-Based Policing Matrix.’ Journal of Experimental
Farrell, G. and Pease, K. (1993). ‘Once Bitten, Twice Bitten: Criminology 7: 3–26.
Repeat Victimisation and its Implications for Crime Mastrofski, S. D., Weisburd, D. and Braga, A. A. (2010).
Prevention.’ Police Research Group: Crime Prevention ‘Rethinking Policing: The Policy Implications of Hot
Unit Series Paper 46, 32. Spots of Crime.’ In Frost, N. A., Freilich, J. D. and
Farrell, G., Philips, C. and Pease, K. (1995). ‘Like Taking Clear, T. R. (eds), Contemporary Issues in Criminal
Candy: Why Does Repeat Victimization Occur?.’ British Justice Policy. Belmont, CA: Wadsworth, Cengage
Journal of Criminology 35(3): 384–399. Learning, pp. 251–264.
166 Policing Article C. P. Haberman and J. H. Ratcliffe

McCord, E. S. and Ratcliffe, J. H. (2009). ‘Intensity Value Taylor, R. B. (1997). ‘Order and Disorder of Streetblocks
Analysis and the Criminogenic Effects of Land Use and Neighborhoods: Ecology, Microecology and the
Features on Local Crime Patterns.’ Crime Patterns and Systemic Model of Social Organization.’ Journal of
Analysis 2(1): 17–30. Research in Crime and Delinquency 34: 113–115.
McDonald, P. P. (2002). Managing Police Operations: Taylor, R. B. (2010). ‘Hot Spots Do Not Exist, and Four
Implementing the New York Crime Control Model - Other Fundamental Concerns About Hot Spots
CompStat. Belmont, CA: Wadsworth. Policing.’ In Frost, N. A., Freilich, J. D. and Clear, T. R.
Pease, K. (1998). Repeat Victimisation: Taking Stock. Crime (eds), Contemporary Issues in Criminal Justice Policy.
Detection and Prevention Series Paper no. 90. London: Belmont, CA: Wadsworth, Cengage Learning,
Police Research Group. pp. 271–278.
Pitcher, A. B. and Johnson, S. D. (2011). ‘Exploring Theories Tilley, N. (2003). ‘Community Policing, Problem-Oriented
of Victimization Using a Mathematical Model of Policing and Intelligence-Led Policing.’ In Newburn, T.
Burglary.’ Journal of Research in Crime and Delinquency (ed.), Handbook of Policing. Cullompton, Devon: Willan,
48(1): 83–109. pp. 311–339.
Polvi, N., Looman, T., Humphries, C. and Pease, K. (1991). Tompson, L. and Townsley, M. (2010). ‘(Looking) Back to
‘The Time Course of Repeat Burglary Victimization.’ The Future: Using Space–Time Patterns to Better Predict
British Journal of Criminology 31(4): 411–414. the Location of Street Crime.’ International Journal of
Ratcliffe, J. H. (2004a). ‘Geocoding Crime and a First Police Science & Management 12(1): 23–40.
Estimate of a Minimum Acceptable Hit Rate.’ Townsley, M. (2007). Near Repeat Burglary Chains:
International Journal of Geographical Information Science Describing the Physical and Network Properties of a
18(1): 61–72. Network of Close Burglary Pairs. Paper presented at the
Ratcliffe, J. H. (2004b). ‘The Hotspot Matrix: A Framework Institute for Pure and Applied Mathematics Conference
for the Spatio-Temporal Targeting of Crime Reduction.’ on Crime Hot Spots: Behavioral, Computational and
Police Practice and Research 5(1): 5–23. Mathematical Models.
Ratcliffe, J. H. (2008). Intelligence-Led Policing. Collumpton, Townsley, M., Homel, R. and Chaseling, J. (2003).
UK: Willan. ‘Infectious Burglaries: A Test of the Near Repeat
Ratcliffe, J. H. (2009). Near Repeat Calculator (version 1.3). Hypothesis.’ British Journal of Criminology 43(3):
Philadelphia, PA and Washington, D.C.: Temple 615–633.
University and National Institute of Justice. Townsley, M., Johnson, S. D. and Ratcliffe, J. H. (2008).
Ratcliffe, J. H. (2010). ‘The Spatial Dependency of Crime ‘Space Time Dynamics of Insurgent Activity in Iraq.’
Increase Dispersion.’ Security Journal 23(1): 18–36. Security Journal 21: 139–146.
Ratcliffe, J. H. and Rengert, G. F. (2008). ‘Near Repeat Weisburd, D., Bushway, S., Lum, C. and Yang, S -M. (2004).
Patterns in Philadelphia Shootings.’ Security Journal ‘Trajectories of Crime at Places: A Longitudinal Study of
21(1–2): 58–76. Street Segments in the City Of Seattle.’ Criminology 42(2):
Ratcliffe, J. H., Taniguchi, T., Groff, E. R. and Wood, J. D. 283–321.
(2011). ‘The Philadelphia Foot Patrol Experiment: A Weisburd, D., Mastrofski, S. D., McNally, A. M.,
Randomized Controlled Trial of Police Patrol Greenspan, R. and Willis, J. J. (2003). ‘Reforming To
Effectiveness in Violent Crime Hotspots.’ Criminology. Preserve: Compstat and Strategic Problem Solving in
49(3): 795–831. American Policing.’ Criminology & Public Policy 2(3):
Rosenbaum, D. P. (2006). ‘The Limits of Hot Spots 421–456.
Policing.’ In Weisburd, D. and Brag, A. A. (eds), Police Wright, R. T. and Decker, S. H. (1997). Armed Robbers in
Innovation: Contrasting Perspectives. New York: Action: Stickups and Street Culture. Boston, MA:
Cambridge University Press, pp. 245–263. Northeastern University Press.
Sherman, L. W., Gartin, P. R. and Buerger, M. E. (1989). Wyant, B. R., Taylor, R. B., Ratcliffe, J. H. and Wood, J. (in
‘Hot Spots of Predatory Crime: Routine Activities and the press). ‘Deterrence, Firearm Arrests, and Subsequent
Criminology of Place.’ Criminology 27(1): 27–55. Shootings: A Micro-Level Spatio-Temporal Analysis.’
St. Jean, P. K. B. (2007). Pockets of Crime: Broken Windows, Justice Quarterly. doi:10.1080/07418825.2011.576689.
Collective Efficacy, and the Criminal Point of View.
Chicago, IL: The University of Chicago Press.