Show Summary Details

Page of

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: null; date: 13 May 2021

Crime Location Choice: State of the Art and Avenues for Future Research

Abstract and Keywords

Crime is unevenly distributed in space. This chapter discusses the uneven spatial patterns in crime from an offender decision-making perspective. It describes the main theoretical perspectives in environmental criminology (the rational choice perspective, routine activity approach, and crime pattern theory) and reviews the empirical research with an emphasis on studies that have used a discrete spatial choice framework for analyzing individual crime location choices. The strength of the discrete spatial choice framework, several of its assumptions, and its link with random utility maximization theory are discussed. The chapter concludes with several challenges for future crime location choice research, including challenges regarding temporal aspects of criminal decision making, planned versus opportunistic crimes, and solved versus unsolved crimes.

Keywords: crime location choice, environmental criminology, random utility maximization, discrete choice models, criminal decision making

I. Overview

Offenders make numerous choices when they commit crimes. Geographical and environmental criminologists are mainly interested in those choices that affect the spatial and temporal patterns in crime. Studies on the uneven spatial distribution of crime abound. The study of seasonal patterns in crime also has a long tradition. Crime patterning by time of day and day of week has received far less attention, but clearly crime not only varies over space but also varies in cyclic patterns at different scales of temporal resolution (Andresen and Malleson 2013; Andresen and Malleson 2015). This chapter mainly discusses how offenders decide where to commit crime, but it also stresses the importance of more research into temporal aspects of target choice. Although the choices of potential victims and those of the people and agencies that try to prevent crime from happening will certainly have some impact on offenders’ decisions on where and when to strike, it is ultimately the offender who decides on the location and time of the offense. Therefore, this chapter discusses the uneven spatial patterns in crime from an offender decision-making perspective. It describes the main theoretical perspectives and reviews empirical research with an emphasis on studies that use the discrete choice framework (Ben-Akiva and Lerman 1985) for analyzing individual crime location choices as introduced to the field of criminology by Bernasco and Nieuwbeerta (2003, 2005).

Before discussing the main environmental criminological theories for crime location choices and the often-ignored role of time, it is important to stress that this review mainly discusses these choices for crime types with a clear geography. The crimes should have specific geographical locations and should be committed at distinct times. So far, most crime location choice research has studied residential burglaries (Bernasco and Nieuwbeerta 2003, 2005; Bernasco 2006, 2010a, 2010b; Clare, Fernandez, and Morgan 2009; Bernasco, Johnson, and Ruiter 2015; Townsley et al. 2015; Vandeviver et al. 2015; (p. 399) Townsley et al. 2016). However, the discrete choice framework has also been applied to street (Bernasco and Block 2009; Bernasco 2010b; Bernasco, Block, and Ruiter 2013) and commercial (Bernasco and Kooistra 2010) robberies, theft from vehicles (Bernasco 2010b; Johnson and Summers 2015), rioting (Baudains, Braithwaite, and Johnson 2013), assaults (Bernasco 2010b), drug dealing (Bernasco and Jacques 2015), and any type of crime (Lammers et al. 2015).

II. Environmental Criminological Theories

Research into the geography of crime has a long history in criminology, initially sparked by the moral statisticians of the early 1800s (Balbi and Guerry 1829) who already used maps to show that crimes were not uniformly distributed across the different regions of France. However, most criminological research addresses etiological questions about criminality. Criminal inclinations, the distinction between the people involved in criminal behavior and those who are not, and the conditions and interventions that affect onset and desistance from crime are key topics in this line of criminological research. There is a plethora of criminological theories aimed at understanding criminality. However, from the 1970s onward, several scholars shifted focus from the study of who commits crime and why to the study of criminal events. With their rational choice perspective (Clarke and Cornish 1985; Cornish and Clarke 1986), routine activity approach (Cohen and Felson 1979), and crime pattern theory (P. J. Brantingham and Brantingham 1978, 1984, 2008; P. L. Brantingham and Brantingham 1981, 1993), they shaped the field of environmental criminology with its strong emphasis on criminal opportunities and situational crime prevention (Clarke 1983). This field of criminology is primarily concerned with the circumstances in which crimes occur—when and where, how, and against what targets or victims—and in answering these questions, it generally treats criminal motivation as given.

Cohen and Felson (1979) provided a macro-level approach for analyzing how ordinary legal activities of people shape where and when crimes occur. When in the course of their daily routines, people who are motivated to commit crime converge with suitable victims or targets in the absence of capable guardians, crime is likely to happen. Because these convergence settings are not uniformly distributed over space and time, spatiotemporal crime clusters exist. The approach provides a framework for understanding such clustering, and it also predicts crime-level trends when macro-level shifts in daily routines occur. Its emphasis on the temporal aspects of people’s routines proves highly valuable for understanding the often-ignored temporal patterning of crime. However, by emphasizing that crimes occur during the daily routines, it suggests that most crimes are opportunistic, and the routine activity approach cannot really account for the goal-oriented behavior of many offenders. The approach lacks an individual-level theory for offenders’ crime location choices.

(p. 400) The rational choice perspective (Cornish and Clarke 1986) provides an informal rational choice theory for understanding offender behavior as goal-oriented decision making. Weighing the costs and benefits of different behavior alternatives, offenders would make those choices that are assumed to bring them closer to their goals. Clearly, the rational choice perspective is rather abstract, and it requires auxiliary assumptions and empirical regularities regarding the relevant goals and choice situations of offenders in order to arrive at testable hypotheses for understanding crime location choices. Because of the analogy between environmental criminological questions concerning the spatiotemporal aspects of target selection and questions regarding animals’ choices of diet, foraging territory, and time spent in the territory, several authors have used optimal foraging theory from behavioral ecology as a supplement to the rational choice perspective (Johnson and Bowers 2004; Felson 2006; Morselli and Royer 2008; Bernasco 2009; Johnson, Summers, and Pease 2009; Pires and Clarke 2011; Johnson 2014). Both theories start from a neoclassical microeconomics concept of goal-oriented behavior and, as such, optimal foraging theory lends itself to arrive at testable hypotheses about crime specialization, the distance offenders are willing to travel, the choice of target area, the time spent during the commission of crime, and the profits obtained through crime (Bernasco 2009). For example, Morselli and Royer (2008) show that mobile offenders generally have higher financial earnings, which is in line with predictions from optimal foraging and rational choice theory because higher gains should offset the costs of further travel.

With their crime pattern theory, Brantingham and Brantingham (P. J. Brantingham and Brantingham 1978, 1984, 2008; P. L. Brantingham and Brantingham 1981, 1993) provided a comprehensive explanation of crime that combines insights from behavioral geography with the rational choice perspective and the routine activity approach. Their geometry of crime (P. L. Brantingham and Brantingham 1981) is central to their crime pattern theory, and it asserts that offenders search for suitable targets or victims at places that emit cues that fit their learned templates of the characteristics of a “good” crime site (P. L. Brantingham and Brantingham 1993, p. 5). Their search is not random in space and time, but it involves looking for targets or victims inside the offender’s awareness space. All people, including offenders, have routine activities that shape their spatial knowledge. Activity spaces consist of the routine activity nodes, such as homes, workplaces, schools, shopping areas, and leisure locations, and the usual travel paths between these. P. J. Brantingham and Brantingham (2008, p. 84) define the awareness space as “the area normally within visual range of the activity space,” and people will generally have limited knowledge of areas outside their awareness spaces. Crime pattern theory therefore asserts that crime occurs at locations where attractive opportunities for crime overlap with awareness spaces of individuals motivated to commit crime. Clearly, this does not predict that a particular offender will commit a burglary in exactly the same area as he or she would commit a robbery or any other type of crime because what is viewed as attractive opportunity varies by type of crime. This does not make crime pattern theory a crime-specific theory. On the contrary, it only stresses the importance of specifying what makes targets attractive for the specific type of crime under study.

(p. 401) III. The Discrete Choice Framework in Crime Location Choice Research

Before Bernasco and Nieuwbeerta (2003, 2005) introduced the discrete choice framework in the geography of crime, three other approaches to study crime location choices were used (Bernasco and Ruiter 2014). The offender-based approach as used in journey-to-crime research uses either offenders or offenses as units of analysis and studies the distribution of distances (Townsley and Sidebottom 2010) and directions in which the offenders traveled (Van Daele and Bernasco 2012) and how these vary with characteristics of the offender and the offense. The target-based approach uses potential targets as units of analysis and studies how victimization rates vary with characteristics of the targets (Haberman and Ratcliffe 2015). The mobility approach uses pairs of geographical locations—all potential areas of departure (home areas of offenders) in combination with all potential target areas—as the unit of analysis to study how the frequency of crime trips between these dyads varies with characteristics of the origin and the destination of the trips (Reynald et al. 2008). As described in much greater detail by Bernasco and Nieuwbeerta (2005) and Bernasco and Ruiter (2014), all three approaches have their own strengths and weaknesses, and the discrete choice framework improves upon all three approaches and thus provides the best approach to the study of crime location choices. It explicitly starts from a decision-making perspective and uses the outcome of a crime location choice, a particular target location out of a set of potential target location alternatives, as the dependent variable, and the unit of analysis is the individual decision maker.

Although relatively new in the geography of crime, the discrete choice framework has been used in many disciplines to theorize and statistically model individual choice behavior (Ben-Akiva and Lerman 1985). The framework is firmly rooted in the microeconomic random utility maximization (RUM) theory, a formal rational choice theory that adds a random component to the utility function. This random component only reflects incomplete information on the part of the analyst, not on the part of the decision maker. With some assumptions regarding the distribution of this random component, the statistical model can be directly derived from the theory (McFadden 1973). This tight connection between theory and model makes it easy for researchers to rigorously test new hypotheses. The discrete choice framework specifies four key elements to any choice situation:

  1. 1. Decision makers: These are the agents who make the choice.

  2. 2. Alternatives: Decision makers must choose one alternative from the choice set—a set of countable, mutually exclusive, and collectively exhaustive alternatives.

  3. 3. Attributes: All alternatives have characteristics that affect the utility a decision maker would derive from choosing the alternative. The decision maker evaluates the utilities of all alternatives. All decision makers have characteristics that potentially also affect the utility they derive from the alternatives.

  4. (p. 402) 4. Decision rule: RUM theory predicts that decision makers choose the alternative from which they expect to derive maximum utility. Utility can be any gain, profit, or satisfaction.

Following the notation of Bernasco and Nieuwbeerta (2005), the utility for decision maker i from choosing alternative j is given by the following equation:

Uij=βzij+eij

where zij is the matrix of attributes that vary across alternatives and possibly across decision makers; β is the vector of coefficients that need to be estimated empirically; and eij is the random error term, which contains all unmeasured aspects of the utility derived by decision maker i from choosing alternative j. These include unmeasured attributes that are actually relevant to the decision as well as measurement error.

When the discrete choice framework is applied to the study of crime location choices, it is obvious that offenders are the decision makers. The alternatives are all separate potential targets (generally target areas) in a study area. From the environmental criminological theories described previously, hypotheses regarding the relevant attributes of alternatives and offenders can be derived. The general decision rule translates into the prediction that offenders will target those areas where they expect to obtain highest rewards (RE) with least effort (EF) and minimal risks (RI). Entering these abstract choice criteria into the general utility function yields the following:

Uij=βRERE+βEFEF+βRIRI+eij

where RE would be some variable that measures expected rewards, EF is a measure of effort, and RI is a variable that measures risk. Obviously, this linear function could be extended to include multiple measures for the three relevant choice criteria simply by adding another additive term. From the decision rule follows the expectation that βRE should be positive because alternatives that yield higher rewards are more likely to be chosen, βEF should be negative because alternatives that require more effort are less likely targeted, and βRE should also be negative because offenders should favor less risky alternatives over more risky ones. A formal test of these expectations requires, of course, a translation of this formal model into a statistical model. McFadden (1973) shows that under specific assumptions, the formal theoretical model directly translates into the conditional logit model. In this statistical model, the probability that an offender i targets area j is given by the following formula:

Prob(Yi=j)=eβzijj=1Jeβzij

where Yi is the choice actually made by offender i, and zij are all hypothesized attributes relevant to the decision. Conditional logit model estimates eβ are interpreted as multiplicative (p. 403) effects of a unit increase in the independent variable on the odds a target alternative is chosen. Although there is no room to discuss all assumptions underlying the conditional logit model here, in light of recent developments in crime location choice studies, it is important to discuss one specific assumption of the conditional logit model—that is, the assumption of the independence of irrelevant alternatives (IIA). This assumption states that adding or removing alternatives from the choice set will not change the relative odds associated with any existing alternatives. Bernasco (2010a) already argued that it is difficult to maintain the IIA assumption when modeling spatial decision making because nearby alternatives are generally very similar. That is why he used a different statistical model (the competing destinations model) that is consistent with RUM theory but does not rely on the overly restrictive IIA assumption. Nevertheless, most empirical crime location choice studies since then have used the conditional logit model (but see Townsley et al. 2016).

Although the discrete choice framework provides a superior approach to the study of crime location choices, it shows some statistical resemblance to the target-based approach. A target-based approach that uses a Poisson model to estimate how crime counts vary with target characteristics returns identical parameter estimates as a conditional logit model without offender-specific regressors and the same targets as choice set because the likelihood functions are equivalent (Guimarães, Figueirdo, and Woodward 2003). However, target-based approaches generally use all crimes reported to police, whereas the discrete spatial choice models purposefully use only the subsample of cleared crimes because that allows for the inclusion of offender-specific regressors such as distance. This obviously breaks the statistical equivalence. The question then arises which statistical model for crime location choice is to be preferred. Because distance is always one of the most important variables in the discrete crime location choice models (discussed later) and it could well be correlated with target characteristics, the gain of using more crime data in a Poisson model is probably offset by the possibility that its estimates will be biased. The degree of bias, however, could be assessed by comparing results from three different models: a Poisson model with all crimes, a Poisson model with cleared crimes only, and a conditional logit model with offender-specific regressors.

IV. Applications of the Discrete Choice Model

All 17 published crime location choice studies that used the discrete choice approach either implicitly or explicitly started from the general decision rule that offenders target those areas where they expect to obtain the highest rewards with least effort and minimal risks. The bulk of these studies analyzed burglaries, but the framework has also been applied to street and commercial robbery, theft from a vehicle, assault, rioting, drug dealing, and even any type of crime (table 19.1). In the remainder of this section, the main findings of all studies are discussed. (p. 404) (p. 405)

Table 19.1 Crime Location Choice Studies That Use the Discrete Choice Framework

Study

Crime Type(s)

Study Area(s)

Spatial Units

Main Findings (+/– for Positive/Negative Effects)

Bernasco and Nieuwbeerta (2003, 2005)

Residential burglary

The Hague, Netherlands

Neighborhoods

Proximity (+); number of residential units (+); proportion of single-family dwellings (+); ethnic heterogeneity (+) stronger for non-natives than for natives

Bernasco (2006)

Residential burglary

The Hague, Netherlands

Neighborhoods

Proximity (+); number of residential units (+); physical accessibility (+); choice criteria the same for solo burglars and co-offending groups

Clare et al. (2009)

Residential burglary

Perth, Australia

Suburbs

Proximity (+); number of residential units (+); river and major road between suburb and offender home (–); train connecting suburb with offender home area (+); ethnic heterogeneity, percentage indigenous (+) stronger for indigenous than for non-indigenous offenders; percentage rental properties (+)

Bernasco and Block (2009)

Street robbery

Chicago, IL

Census tracts

Proximity (+); population (+); racial and ethnic dissimilarity (–); gang territory dissimilarity (–); collective efficacy (–); illegal markets (drugs and prostitution) (+); retail employment levels (+); high school presence (+)

Bernasco (2010a)

Residential burglary

The Hague, Netherlands

Six-digit postal code areas

Proximity (+) more important for juveniles than for adults; number of properties (+); population aged 15–25 years (+); real estate values (+); population non-native (–) for native offenders

Bernasco (2010b)

Residential burglary, theft from vehicle, robbery, assault

Greater The Hague area, Netherlands

Four-digit postal code areas

Proximity (+); proximity to previous home location (+); lived in the area until recently (+); lived in the area for longer period of time (+)

Bernasco and Kooistra (2010)

Commercial robbery

Netherlands

Four-digit postal code areas

Proximity (+); proximity to previous home location (+); lived in the area until recently (+); lived in the area for longer period of time (+)

Bernasco et al. (2013)

Street robbery

Chicago, IL

Census blocks

Proximity (+); total population (+); connected by elevated train station or main street (+) also in adjacent block; legal cash economies (+), some also in adjacent blocks; illegal cash economies (+); high schools (+); ethnic group is majority (+)

Baudains et al. (2013)

Rioting

Greater London

Lower super output areas

Proximity (+) stronger for juveniles than for adults; presence of schools (+) stronger for juveniles than for adults; same side of river Thames (+); retail businesses (+); target area the previous days (+)

Johnson and Summers (2015)

Theft from vehicle

Dorset, UK

Lower super output areas

proximity (+) stronger for juveniles than for adults; number of cars and vans (+); presence of school (+) only for juveniles; presence of train station (+) only for adults; contains major road (+) only for adults; population turnover (+); socioeconomic heterogeneity (+)

Townsley et al. (2015)

Residential burglary

The Hague, Netherlands; Birmingham, UK; Brisbane, Australia

Neighborhoods; super output areas; statistical local areas

Proximity (+) all three study areas, and juveniles more strongly affected in Birmingham and Brisbane; proportion accessible targets (+) all three study areas; number of potential targets (+) all three study areas

Townsley et al. (2015)

Residential burglary

Brisbane, Australia

Statistical local areas

Proximity (+); proximity to the city center (+); number of households (+); percentage single families (+); proximity effect strongest for juveniles, decreases until stabilizes in adulthood (+)

Lammers et al. (2015)

All crimes

Greater The Hague area, Netherlands

Four-digit postal code areas

Proximity (+); lived in the area until recently (+); lived in the area for longer period of time (+); previously targeted area (+) stronger effects for more recent offenses and same crime type; proximity to previous offense (+); number of prior offenses in same area (+); proportion of non-Western residents (+); number of employees in target area and number of several types of facilities (+)

Bernasco et al. (2015)

Residential burglary

West Midlands, UK

Lower super output areas

Proximity (+);previously targeted area (+) stronger effects for more recent offenses and closer to previous offenses; train stations (+); proximity to city center (+); number of households (+); mean house price (+); ethnic diversity (+); population turnover (+)

Vandeviver et al. (2015)

Residential burglary

East Flanders, Belgium

Individual properties

Proximity (+); semi-detached less likely than terraced (–); garage present (–); central heating or air conditioning present (–)

Bernasco and Jacques (2015)

Drug dealing

Amsterdam city center, Netherlands

Street segments

Tourist-attracting facilities (+); police activity (+); alleyways (+); public toilets (+)

(p. 406) In their groundbreaking burglary target choice study, Bernasco and Nieuwbeerta (2003, 2005) were the first to use the discrete choice framework in the geography of crime. They analyzed 548 burglaries committed by solitary offenders who lived in the city of The Hague, Netherlands. They used the conditional logit model to estimate why these burglars chose to commit their burglaries in one of the 89 potential neighborhoods of the city of The Hague. They hypothesized that burglars would favor affluent neighborhoods (measured by average property values), with many residential units, better accessible targets (measured by proportion of single-family dwellings), and in which it was less likely that they would be disturbed by the residents and other guardians (indicated by high residential mobility and high ethnic heterogeneity).

In line with both crime pattern theory and the rational choice approach, Bernasco and Nieuwbeerta (2003, 2005) also argued that burglars would prefer to target neighborhoods closer to home because they would be more familiar with these areas and committing offenses farther away would require more time and effort. Also for reasons of familiarity, they hypothesized that offenders would be more likely to target neighborhoods closer to the city center. They further acknowledged that choice criteria need not be equally applicable to all burglars. They reasoned that the proximity effect would be stronger for juveniles than for adults because juveniles are more constrained in their mobility and consequently have smaller awareness spaces. They also hypothesized that the effect of ethnic heterogeneity would be stronger for non-natives because non-natives would be more easily identified as outsiders in homogeneous (i.e., native white) neighborhoods than natives would in ethnically mixed neighborhoods. Their results showed that burglars indeed target areas that are closer to home, have more residential units, have more single-family dwellings, and are ethnically heterogeneous. Indeed, ethnic heterogeneity was more important for non-native than for native burglars, and although the different proximity effects did not statistically differ between juveniles and adults, the effects were in the expected direction. In a later study that used much smaller units of analysis and data that covered a longer period, Bernasco (2010a) also found statistical support for the hypothesized age-specific proximity effects.

Because the discrete choice model requires the researcher to define a single decision maker, in their original study, Bernasco and Nieuwbeerta (2003, 2005) simplified their analysis by excluding all multiple-offender burglaries. However, burglaries are clearly not always committed by solo offenders. Bernasco (2006) therefore extended the initial study by adding multiple-offender burglaries, and he tested whether the choice criteria were different for solo offenders and co-offending groups. The only difference he found was that solo offenders were more likely to target their own neighborhoods compared to co-offending groups.

In one of the few replication studies in criminology, Townsley et al. (2015) compared the original The Hague findings of Bernasco and Nieuwbeerta (2003, 2005) to the target choice criteria of burglars in Birmingham, United Kingdom, and Brisbane, Australia. The results showed consistent effects for proximity to the home of the offender, the proportion of easily accessible targets in the area, and the total number of targets available. Distance appeared to impede juveniles more than adults, although the difference only (p. 407) reached statistical significance in Birmingham and Brisbane. For two of the three consistent findings, the effects appeared to differ in size between the study areas. According to Townsley et al. (2015), the different proximity effects could be attributed to population density differences, although more replication studies are needed to test this more rigorously.

In a subsequent study using only the Brisbane burglary data, Townsley et al. (2016) used the mixed logit model, a generalization of the conditional logit model, to study the extent to which the choice criteria actually vary between burglars. The mixed logit model relaxes several overly restrictive assumptions of the conditional logit model, such as the IIA assumption, but also—and probably more important—the assumption that the choice criteria affect all offenders in the same way. The conditional logit model simply estimates a single parameter for each choice criterion for all offenders, and the only way to address the issue that effects might differ between offenders is by including different effects in the structural part of the model. The conditional logit model therefore implicitly assumes that all offenders weigh the costs and benefits equally. However, Bernasco and Nieuwbeerta (2005) already provided two examples of varying effects when they estimated age-specific proximity effects and separate effects of ethnic heterogeneity for natives and non-natives. The mixed logit model allows for the estimation of random effects or offender-specific parameters. This opens several new avenues for research because it allows the researcher further scrutiny of these random effects. Townsley et al. (2016) first show that the results from a conditional logit and a mixed logit model differ considerably. The effect of residential mobility even switches signs. They subsequently show that distance is much more important for juveniles than for adults, and the effect decreases until adulthood, during which it remains stable. These findings are in line with those from previous journey-to-crime research, but they also demonstrate that age explains only 3 percent of the variance in the proximity effect. Clearly, the mixed logit model provides new research opportunities. For instance, it enables studies on which offenders comply with the theorized effects, and it calls upon new theory to explain why some offenders weigh specific choice criteria more than others.

A recent burglary target choice study estimated how characteristics of individual houses affected their likelihood of being burglarized (Vandeviver et al. 2015). Using an unprecedentedly large choice set of more than 500,000 individual residences, the study showed that controlling for the distance from the home of the offender, burglars were less likely to target semidetached houses compared to terraced dwellings, and houses with garages and central heating were also less likely to be burglarized.

V. Barriers and Connectors

Although the finding that burglars prefer to target areas closer to home corroborated previous journey-to-crime research that consistently showed distance decay, Euclidean distance as used in the first discrete choice models actually provides a rather crude measure for the (p. 408) impedance people encounter when traveling. Clare et al. (2009) convincingly argued that travel in a city can be hampered by physical barriers such as major roads and rivers. They hypothesized that burglars will be less likely to cross such barriers when traveling to their burglary target area. Conversely, they argued that potential target areas that are connected by major rail tracks are more likely to be targeted. Their analysis of 1,761 burglaries committed in 292 suburbs of Perth, Australia, provided support for both hypotheses. They also showed that barriers have stronger effects the closer they are to the burglars’ homes.

Although physical barriers clearly limit the possibilities of travel, Bernasco and Block (2009) argued that the same applies to social barriers. These barriers would deter offenders from committing crime in areas that are socially (economically, culturally, and ethnically) different from the areas in which offenders live. In their study of 12,872 robberies committed in the 844 census tracts of Chicago, Bernasco and Block showed that this indeed applies. Robbers are more likely to commit a robbery in a census tract in which the majority of the population is of their own ethnic background. This finding can be explained in two ways. First, it could be that they simply spend more time in those areas and consequently also commit their robberies there. If this is the case, their routine activities alone could explain this effect. Second, they could purposefully target those areas because they may be less conspicuous in those areas, which limits the risk of being identified as an outsider who is there to commit a crime. Distinguishing between these two explanations is difficult because it requires detailed information on offenders’ routine activities. Bernasco and Block also identified a social barrier related to gang territory. Chicago robbers appeared to be less likely to commit a robbery in an area that had gang-related crime of a different gang than the gangs active in their home area. The same Chicago data were used to study in more detail how cash economies affect where street robberies are committed (Bernasco et al. 2013). The presence of both legal (a wide variety of small businesses) and illegal (drug, prostitution, and gambling areas) cash economies in a census block attracts street robbers, and some of these effects spill over into the adjacent census blocks, but not beyond. The social barrier effect was also tested more thoroughly, and the results showed that robbers of specific ethnic groups are much more likely to target census blocks in which their ethnic group is the majority, and Hispanic robbers are much less likely to target a census block with an African American majority. Furthermore, census blocks that were better connected by main streets and public transport were also more likely to be targeted.

Just like with robberies in Chicago, the presence of small businesses also attracts drug dealers to particular street segments in the Red Light District of the city of Amsterdam (Bernasco and Jacques 2015). Although their study was too small to statistically test for effects and to use a multivariate design, and because they did not include any offender-specific characteristics, their discrete choice model is similar to a target-based approach, their drug-dealing study is a unique crime location choice study because it did not use police-recorded data and it simultaneously studied multiple crime location choices of the same offenders. They studied where drug dealers solicit for customers separate from where they close the deal. The dealers were mainly attracted to street segments with the facilities that generally attract tourists (e.g., coffee shops, liquor stores, smartshops, tobacco shops, hotels and hostels, and clubs), and they were not at all deterred by police (p. 409) activity. On the contrary, they were much more likely to solicit for customers and to close the deal in street segments that were under some police surveillance. Alleyways and public toilets also stood out as drug dealer attractors.

VI. Awareness Space

Although some of the earlier crime location choice studies touched upon crime pattern theory’s prediction that offenders mainly target areas within their awareness spaces, most crime location choice studies only include information about the home areas of offenders at the time they committed the offenses. However, the home is not the only reference point, of course. Many crimes are committed by homeless people, and they obviously have no home. Rengert (2004) argues that some of these people are homeless because of the addiction to drugs and their crimes are clustered about drug sales areas.

Canter and Larkin (1993) make a distinction between marauders and commuters—in other words, those who use the home base as a start of their crime journeys versus those who commit crimes outside their home range. They show that the marauder model best predicts the crime locations of their sample of serial rapists, which stresses the importance of the home base in crime location choice. Nevertheless, Costello and Wiles (2001) argue that the journey-to-crime distances are sometimes overestimated because it is almost always assumed that offenders started their crime journeys from their own homes, but it happens that they committed the crime when they were, for instance, staying with a friend.

Bernasco (2010b) and Bernasco and Kooistra (2010) were the first to explicitly theorize about the dynamics of awareness spaces in a discrete spatial choice framework. They tested whether offenders were more likely to target not only areas near their current homes but also areas close to where they used to live. This clearly was the case. They also showed that areas in which the offender had lived for a longer period of time and until more recently had stronger effects.

Obviously, other important nodes, such as where offenders went to school, their workplaces, the residential areas of family members and friends, as well as where they spend leisure time, could be included in crime location choice models in order to further test crime pattern theory. Information concerning these activity nodes is often missing and can only be collected using offender-based research designs (Bernasco 2013). Because crime location choice studies generally use police data, these studies simply do not contain systematically collected information concerning these other activity nodes of offenders. Baudains et al. (2013) and Johnson and Summers (2015) show, however, that it is not necessarily required to measure individual-level activity nodes to test predictions derived from crime pattern theory. Baudains et al. studied the crime location choices of 2,299 rioters during the 2011 London riots. In addition to a strong effect of proximity, they showed that rioters were more likely to target areas with schools, areas that were on the same side of the river Thames, and areas close to retail businesses. All these areas were of course more likely to be known to the offender than otherwise comparable (p. 410) areas. As predicted, the effects of school presence and distance appeared to be stronger for juveniles than for adults. They further showed that areas that were targeted the previous day (not necessarily by the same rioter) were more likely to be targeted again.

Johnson and Summers (2015) further scrutinized the age-specific effects for offenders who committed thefts from vehicles. Several findings were in line with crime pattern theory. Both juveniles and adults favored areas close to where they lived, but juvenile offenders committed their offenses closer to home compared to adult offenders, and only juveniles were also more likely to commit these offenses in areas where schools were located. The other variables included in their model again provide support for the idea that offenders prefer to commit crime where they are less likely to be disturbed by the residents and other guardians (indicated by population turnover and high socioeconomic heterogeneity).

Two recent studies (Bernasco et al. 2015; Lammers et al. 2015) used police data to test whether offenders also return to previously targeted areas, as was suggested in the literature on near repeat victimization (Bernasco 2008; Bowers and Johnson 2004). The effects of previous crime locations were more pronounced than those for previous residential areas. Offenders are very likely to return to areas of their previous offenses, especially when the time between a previous offense and a new one is short.

Although awareness spaces are defined to include both the area around activity nodes and the travel paths between these, research on the impact of the usual travel routes of offenders on their spatial decision making is virtually nonexistent. Rengert and Wasilchick (2000) showed that Philadelphia burglars were more likely to target homes that were along the routes to their workplaces and recreation sites, and Frank et al. (2011) showed directional consistency toward shopping malls in the city of Coquitlam, Canada. These findings suggest that offenders are indeed more likely to target areas that are along the paths between their activity nodes. It is not necessary to extensively measure the travel behavior of offenders in order to test for such effects in a discrete choice model. If some activity nodes are known, the likely routes between these can be estimated using metrics derived from graph theory (Davies and Johnson 2015), and their effects on crime location choices can subsequently be tested. The current approaches that use only distance from a particular activity node (e.g., distance from former home location) assume that the likelihood of targeting a particular area decays in a circular symmetric way around the activity node. However, it is more probable that the likelihood of targeting is especially increased in the direction of the offender’s other activity nodes.

VII. Spatial Units of Analysis: Computational and Theoretical Challenges

In geographical criminology, there is a trend toward analyzing crime at increasingly smaller spatial units of analysis (Weisburd, Bernasco, and Bruinsma 2008). The (p. 411) crime location choice modeling literature has followed suit. It started with residential neighborhoods (Bernasco and Nieuwbeerta 2003, 2005) and census tracts (Bernasco and Block 2009) as spatial units of analysis, but it moved to smaller units such as postal code areas (Bernasco 2010a), census blocks (Bernasco et al. 2013), and lower super output areas (Baudains et al. 2013; Bernasco et al. 2015). This trend culminated in the recent burglary target choice study of Vandeviver et al. (2015), who used individual residences in East Flanders, Belgium. Although the studies at different scales of spatial resolution have not yet led to contradictory findings, it is unclear to what extent crime location studies are impacted by what geographers call the modifiable areal unit problem (Openshaw 1984). This is the problem that study results are often highly dependent on the size and shape of the spatial units of analysis. Although smaller spatial units of analysis seem to better fit the theoretical models of crime location choice, the trend toward smaller units also leads to computational and theoretical challenges.

The computational challenge is caused by the fact that the likelihood function needs to be computed for each offense-by-alternative combination during the iterative estimation procedure of the discrete choice model. For example, Bernasco and Block (2009) used a discrete choice model to study 12,872 street robberies in 844 census tracts of Chicago. For the estimation, they created a data set of more than 10 million rows (the number of robberies multiplied by the number of alternatives). Decreasing the size of the spatial unit of analysis increases the number of alternatives and thereby the computational complexity. This is immediately evident from the study of Bernasco et al. (2013), in which they analyzed the same data at the census block level. With 24,594 census blocks in Chicago, the required full data set would have contained more than 300 million rows. However, McFadden’s (1978) sampling-of-alternatives method provides a solution to this computational problem because it yields consistent estimators for the parameters of the conditional logit model. This solution, however, entirely rests on the stringent IIA assumption, which is likely violated in crime location choice modeling (Bernasco 2010a; Townsley et al. 2016). Nevertheless, several studies have used this solution for estimating the models on very large data sets (Bernasco 2010a; Bernasco et al. 2013; Vandeviver et al. 2015). Sampling-of-alternatives methods for more complex models that relax the IIA assumption (e.g., nested logit and mixed logit models) have recently been developed (Guevara and Ben-Akiva 2013; von Haefen and Domanski 2013), but these have not yet been used in criminological research.

The theoretical challenge is twofold. The first relates to spatial spillover effects as recently discussed by Bernasco and Ruiter (2014) and empirically addressed by Bernasco et al. (2013). In general, the smaller the spatial units of analysis, the more likely it is that spatial spillover effects are in play, especially if these effects rapidly decay in space. For example, Bernasco et al. (2013) describe how a street robber might follow a customer from a store in a particular block to another block nearby that provides a better location for attack. Their empirical results provide evidence that the effects of robbery attractors indeed spill over to adjacent blocks (first-order spatial spillover) but not beyond (second-order spatial spillover effects were not statistically significant). (p. 412) The fact that these spillover effects decay so rapidly shows that it is especially important to address spillover effects when using small spatial units of analysis. In analyses with larger units (e.g., neighborhoods or census tracts), these effects will probably be negligible because short-distance spillover occurs for the most part within the boundaries of the larger units. The second theoretical challenge is related to one of the core assumptions of RUM theory. The decision maker is assumed to have complete information about the alternatives—an assumption that becomes increasingly untenable when the number of alternatives increases into the hundreds or even thousands. No offenders will be able to determine the difference between all those alternatives, so how could they then weigh the costs and benefits in their crime location choice decision? They simply cannot. Nevertheless, by using the entire set of alternatives in the study area, all crime location choice studies so far have implicitly assumed that offenders can determine the difference. Before the discrete spatial choice approach gained traction in criminology, Elffers (2004) argued that “a sophisticated version of spatial rational choice theory does not assume that a prospective burglar evaluates all available targets in the same way as his information about some targets might be better than about others. … His evaluation of these [unknown] targets is nonexistent” (p. 189). Crime pattern theory actually provides such a sophisticated model because it predicts that offenders will commit their crimes in areas where their activity spaces overlap with attractive opportunities for crime. The challenge is how to translate this into empirical crime location choice research because it requires extensive data on offenders’ activity spaces and possibly even the development of discrete spatial choice models with offender-specific choice sets.

VIII. Avenues for Future Research

Although the past decade has experienced a steady increase in interest in the discrete choice approach to study crime location choices, clearly more research is needed. This section discusses several potential avenues for future research, which can be summarized with the following questions:

  • Crime location choice studies have so far focused on only the spatial aspects of criminal decision making, but how do temporal aspects affect these choices?

  • In most crime location choice studies, only the home location of offenders was known, but are better data on awareness spaces and offenders’ routine activities required?

  • The discrete choice approach for crime location choices seems to be a model for planned crimes, but what to do with opportunistic crimes?

  • Discrete choice models for crime location choice include only solved crimes, but can the findings be generalized to unsolved crimes? What about cybercrimes?

(p. 413) A. Temporal Aspects, Routine Activities, and Dynamic Awareness Spaces

Although most environmental criminological research is devoted to spatial patterns in crime and temporal aspects are often ignored, there are four reasons why future crime location choice studies should also consider temporal aspects of criminal decision making. First, offenders’ daily routines directly affect where and when they are able to travel. Based on arguments from time geography (Hägerstrand 1970), Ratcliffe (2006) developed a temporal constraint theory for explaining the spatial patterns in opportunistic crimes. He argues that the nondiscretionary routine activities such as work and school strongly affect offenders’ discretionary time, which limits the possibilities of travel and directly affects their awareness spaces. As such, the temporal constraints directly relate to the spatial constraints as already described in the geometry of crime (P. L. Brantingham and Brantingham, 1981). Because of the temporal constraints, Ratcliffe claims, offending patterns cluster around the activity nodes of offenders. When these temporal constraints are ignored, it is implicitly assumed that offenders are able to commit offenses at any time of the day and in all places—a clearly unrealistic assumption. Future crime location choice studies could incorporate these ideas by collecting data on the daily activities of offenders. Although many crimes are indeed opportunistic (Wiles and Costello 2000) and committed while offenders are on their way to regular legal activities, temporal considerations appear to be equally important for offenders who have much more discretionary time. In fact, Rengert and Wasilchick (2000) describe professional burglars who had purposefully quit their day jobs because the jobs did not allow them to burglarize homes when they were vacant during the day. This directly ties into the second reason why temporal aspects deserve more attention in crime location choice studies: Opportunities for crime are time specific, which leads to time-varying target attractiveness (Haberman and Ratcliffe 2015). For example, the best opportunities for burglary are when homes are unguarded. Rengert and Wasilchick argue that the burglars they interviewed knew when the properties were vacant because the spatial behavior of the homeowners is remarkably predictable. Although vacancy makes most suburban properties especially vulnerable during the daytime, Coupe and Blake (2006) stress the importance of studying what makes some properties attractive daytime targets and others good nighttime burglary opportunities. Haberman and Ratcliffe studied whether the effects of potentially criminogenic places for street robbery vary by time of day. Combining information on macro-level human activity patterns with information on the hours of operation of specific facilities in their target-based approach, they were able to show that certain types of facilities attract robberies only at specific times of the day, whereas others attract robbery all day. Although the arguments for such time-varying effects seem trivial from a routine activity perspective—because it is not known why retail businesses would attract robbers when they are closed—all crime location choice studies so far have treated the effects of crime attractors and generators as time stable.

(p. 414) The third reason why future crime location choice studies should address temporal aspects has to do with seasonal variation in routine activities. Andresen and Malleson (2013) show for the city of Vancouver that all crime types exhibit seasonal variation, but more important, the seasonal variation is different in different areas of the city. They argue that such seasonal variation is linked with changes in leisure activities throughout the year. Most cities show seasonal patterns in leisure activities, with more indoor activities in wintertime and outdoor activities during the summer. Although almost all crime location choice studies have used police-recorded crime data for a year or more, none of them have addressed seasonal variation in the characteristics of the alternative target areas. However, it seems reasonable to assume that just like with burglars who know the daily rhythm of home owners, offenders will incorporate seasonal variation with regard to where people spend their leisure activities into their decision on where to offend.

Finally, the dynamics of awareness spaces provide another reason to devote more attention to temporal aspects in crime location choice studies. Because most empirical environmental criminological research is cross-sectional, awareness spaces are generally operationalized as static. Often, only the home locations of the offenders are known. However, in reality, awareness spaces vary over time (Bernasco 2010b). P. L. Brantingham and Brantingham (1981) argued that it is more reasonable to conceptualize awareness spaces as dynamic. They claimed that novice offenders will generally start with relatively small awareness spaces developed through noncriminal activities, but as they continue to commit crimes, their awareness spaces will expand into areas adjacent to their pre-novice awareness spaces. Bernasco (2010b) and Lammers et al. (2015) generalized these claims about the dynamics of awareness spaces. Bernasco redefined awareness space as “a person’s current activity space as well as his or her activity spaces in the recent past, including the area normally within visual range of these activity spaces” (p. 393). Wiles and Costello (2000) already provided evidence from their interviews with offenders that their awareness spaces were linked to where they had lived. Bernasco argued that although people’s routine activities are quite stable across days, weeks, and months, their activity nodes eventually change. Consequently, not only do people acquire new activity nodes but also old activity nodes disappear. According to Bernasco, these old activity nodes gradually fade out of the dynamic awareness space because memory fades and the environments change. From this follows his recency hypothesis that offenders will be more likely to commit offenses in areas where they had activities more recently. Because people’s image of particular areas will be more accurate if they have visited the areas over a longer period of time, Bernasco also hypothesized that the duration of exposure to a particular area is positively related to the likelihood that offenders target the area. Lammers et al. built on these claims, and they hypothesized that offenders are more likely to target areas they have visited more frequently. As already discussed, all three hypotheses concerning these dynamics of awareness spaces have been corroborated in research that used residential location histories (Bernasco 2010b; Bernasco and Kooistra 2010; Lammers et al. 2015) and crime location histories (Bernasco et al. 2015; Lammers et al. 2015) in the prediction of crime location choices. Although these crime location choice studies thus incorporated the dynamics of awareness spaces, these only (p. 415) concerned long-term changes in awareness spaces. However, awareness spaces develop during people’s routine activities, and people generally spend specific times of the day and days of the week at some activity nodes (e.g., during office hours at workplaces), and only few activity nodes will be visited at many different times (e.g., the home). Why, then, would an offender have an accurate image of a particular area at a time of day during which he or she never visits the area? It seems more reasonable to assume that people in fact only have time-specific knowledge of their environment. If so, these short-term dynamics of awareness spaces would lead to time-specific predictions about where offenders commit crime. This again calls for crime location choice studies that address the different times at which offenses are committed and also for better measures of offenders’ routine activities.

Using a discrete spatial choice model, Kitamura, Chen, and Narayanan (1998) showed how time of day can be incorporated into models of traveler destination choices. Bernasco and Nieuwbeerta (2003, 2005) introduced the discrete choice approach as it was developed by transportation researchers and economists to the field of criminology. It seems that criminologists should once again consider the more recent developments in those fields to learn about how to incorporate temporal aspects in the geography of crime.

B. A Framework Only for the Scientific Study of Crimes That Were Planned, Solved, and Have a Clear Geography?

Although none of the applications of the discrete choice framework have made the distinction between crimes that were committed after careful planning and opportunistic crimes, the name “crime location choice model” implicitly paints an image of an offender who rationally weighs the potential benefits against the effort and risks. RUM theory actually assumes that the decision maker evaluates the utilities of all alternatives and then chooses the alternative with the highest expected utility. Of course, RUM theory provides only an abstract model of human decision making. It nevertheless cannot be said that someone who is sufficiently motivated to commit crime and who happens upon an opportunity on his or her way to work really evaluates the utility of the other crime location alternatives. Crime journeys are often not driven by plans to offend but, rather, the opportunities for crime simply present themselves during normal routine activities (Bernasco 2014). For such opportunistic crimes, the only real decision offenders make is to seize the opportunity—a yes/no decision, not a location choice decision. However, if crime location choice studies do not differentiate between the two, their model results will reflect a combination of the two processes that lead offenders to commit crimes where they do. Suppose that half the offenders commit crimes while they encounter opportunities on their way to work and the other half of offenders make a rational crime location choice at home before they leave on their crime journeys. The (p. 416) results of a crime location choice model that combines the two will in large part reflect job location choice instead of crime location choice. Future crime location choice studies should therefore investigate the motives for the crime journeys and preferably only apply the discrete choice approach to those journeys with criminal intent.

Because the discrete choice approach allows for the simultaneous assessment of how potential target area characteristics as well as offender characteristics affect crime location choices, the approach is applied to solved crimes only. Although dark number issues are common to criminological research and recent research suggests that the spatial behavior of arrested offenders is remarkably similar to that of non-arrested offenders (Lammers 2014), it would be interesting to test whether the research findings concerning the effect of previous crime locations on subsequent crime location choices (Bernasco et al. 2015; Lammers et al. 2015) are identical for non-arrested offenders. In fact, the findings are somewhat counterintuitive because why would someone return to an area to commit an offense if he or she was unsuccessful (got arrested) the previous time? It is plausible to assume that offenders who were not arrested for a particular crime are much more likely to return to the same area. Obviously, this cannot be studied using police data and again calls for an offender-based research design.

Most crime location choice studies have examined burglary target choices, but as described in this chapter, the discrete choice framework has been applied to a wide variety of crimes. At the beginning of this chapter, it was argued that crime location choice studies apply to crime types with a clear geography. That is of course true for crime location choice studies, but it is not a requirement per se for crime target choice studies. The statement was only about the current state of affairs. The discrete choice framework does not require a geographical focus, and in fact, it would be interesting to apply it to cybercrimes such as DDoS attacks and malware. Tajalizadehkhoob et al. (2014) presented an innovative study on cybercrime target choice using the instruction files that were sent by the Zeus botnet to infected machines. Although it is largely a descriptive study that uses a target-based approach on cybercrime target choice, it shows how new ways of data collection could help address criminological research questions regarding target choice in cyberspace.

Although the discrete choice framework for crime location choice modeling has only been used with the scientific aim to test hypotheses from environmental criminological theories, environmental criminology itself is very much involved with practical questions of how to reduce crime. That is why this chapter concludes with a description of how crime location choice studies can also have practical value.

All crime location choice studies use observed preferences designs in which the behavioral decision-making model is derived from the actual choices made by offenders. This is a clearly observational design, and as such it lacks the possibility of experimental control. Although the research is firmly rooted in theory, it should be acknowledged that the research findings are mainly correlational, and it is difficult to make any causal claims, let alone predict change when implementing some crime reduction strategy based on the research findings. At best, the discrete choice approach allows for natural experiments—for instance, when a sudden external shock (p. 417) changed the opportunity structure or the connectivity of the street network and crime location choices before and after the shock can be compared. Unfortunately, no studies have used such natural experiments to test for truly causal effects. However, there is another way in which the discrete choice approach can potentially help to fight crime. It has the ability to improve geographic offender profiling techniques. Current techniques rely on several strong assumptions about the crime location choice mechanisms of offenders, such as circular symmetry and distance decay (van Koppen, Elffers, and Ruiter 2011). However, the crime location choice literature as described in this chapter clearly shows that target characteristics, which are generally not uniformly distributed, are very important in the decision where to offend. In a simulation study, Bernasco (2007) shows that a geographic profiling technique based on a reversal of the discrete choice model outperforms standard techniques that take only distance decay into account. Whether an improved geographic profiling technique based on the discrete choice framework also has practical value has yet to be evaluated in empirical research.

References

Andresen, M. A., and N. Malleson. 2013. “Crime Seasonality and Its Variations Across Space.” Applied Geography 43:25–35.Find this resource:

    Andresen, M. A., and N. Malleson. 2015. “Intra-week Spatial–Temporal Patterns of Crime.” Crime Science 4(1): 1–11.Find this resource:

      Balbi, A., and A.-M. Guerry. 1829. Statistique comparée de l’état de l’instruction et du nombre des crimes dans les divers arrondissements des Académies et des Cours Royales de France. Paris: Jules Renouard.Find this resource:

        Baudains, P., A. Braithwaite, and S. D. Johnson. 2013. “Target Choice During Extreme Events: A Discrete Spatial Choice Model of the 2011 London Riots.” Criminology 51(2): 251–85.Find this resource:

          Ben-Akiva, M. E., and S. R. Lerman. 1985. Discrete Choice Analysis: Theory and Application to Travel Demand. Cambridge, MA: MIT Press.Find this resource:

            Bernasco, W. 2006. “Co-offending and the Choice of Target Areas in Burglary.” Journal of Investigative Psychology and Offender Profiling 3(3): 139–55.Find this resource:

              Bernasco, W. 2007. “The Usefulness of Measuring Spatial Opportunity Structures for Tracking Down Offenders: A Theoretical Analysis of Geographic Offender Profiling Using Simulation Studies.” Psychology, Crime and Law 13(2): 155–71.Find this resource:

                Bernasco, W. 2008. “Them Again? Same-Offender Involvement in Repeat and Near Repeat Burglaries.” European Journal of Criminology 5(4): 411–31.Find this resource:

                  Bernasco, W. 2009. “Foraging Strategies of Homo Criminalis: Lessons from Behavioral Ecology.” Crime Patterns and Analysis 2(1): 5–16.Find this resource:

                    Bernasco, W. 2010a. “Modeling Micro-Level Crime Location Choice: Application of the Discrete Choice Framework to Crime at Places.” Journal of Quantitative Criminology 26(1): 113–38.Find this resource:

                      Bernasco, W. 2010b. “A Sentimental Journey to Crime: Effects of Residential History on Crime Location Choice.” Criminology 48:389–416.Find this resource:

                        Bernasco, W. 2013. Offenders on Offending: Learning About Crime from Criminals. New York: Routledge.Find this resource:

                          (p. 418) Bernasco, W. 2014. “Crime Journeys: Patterns of Offender Mobility.” In Oxford Handbooks Online in Criminology and Criminal Justice, edited by M. Tonry, pp. 1–31. Oxford: Oxford University Press.Find this resource:

                            Bernasco, W., and R. Block. 2009. “Where Offenders Choose to Attack: A Discrete Choice Model of Robberies in Chicago.” Criminology 47(1): 93–130.Find this resource:

                              Bernasco, W., R. Block, and S. Ruiter. 2013. “Go Where the Money Is: Modeling Street Robbers’ Location Choices.” Journal of Economic Geography 13(1): 119–43.Find this resource:

                                Bernasco, W., and S. Jacques. 2015. “Where Do Dealers Solicit Customers and Sell Them Drugs? A Micro-Level Multiple Method Study.” Journal of Contemporary Criminal Justice 31(4): 376–408.Find this resource:

                                  Bernasco, W., S. D. Johnson, and S. Ruiter. 2015. “Learning Where to Offend: Effects of Past on Future Burglary Location.” Applied Geography 60:120–29.Find this resource:

                                    Bernasco, W., and T. Kooistra. 2010. “Effects of Residential History on Commercial Robbers’ Crime Location Choices.” European Journal of Criminology 7(4): 251–65.Find this resource:

                                      Bernasco, W., and P. Nieuwbeerta. 2003. “Hoe kiezen inbrekers een pleegbuurt? Een nieuwe benadering voor de studie van criminele doelwitselectie.” Tijdschrift voor Criminologie 45(3): 254–70.Find this resource:

                                        Bernasco, W., and P. Nieuwbeerta. 2005. “How Do Residential Burglars Select Target Areas? A New Approach to the Analysis of Criminal Location Choice.” British Journal of Criminology 45(3): 296–315.Find this resource:

                                          Bernasco, W., and S. Ruiter. 2014. “Crime Location Choice.” In Encyclopedia of Criminology and Criminal Justice, edited by G. Bruinsma and D. Weisburd, pp. 691–99. New York: Springer Verlag.Find this resource:

                                            Bowers, K. J., and S. D. Johnson. 2004. “Who Commits Near Repeats? A Test of the Boost Explanation.” Western Criminology Review 5(3): 12–24.Find this resource:

                                              Brantingham, P. J., and P. L. Brantingham. 1978. “A Theoretical Model of Crime Site Selection.” In Crime, Law and Sanction, edited by M. Krohn and R. L. Akers, pp. 105–18. Beverly Hills, CA: Sage.Find this resource:

                                                Brantingham, P. J., and P. L. Brantingham. 1984. Patterns in Crime. New York: Macmillan.Find this resource:

                                                  Brantingham, P. J., and P. L. Brantingham. 2008. “Crime Pattern Theory.” In Environmental Criminology and Crime Analysis, edited by R. Wortley and L. Mazarolle, pp. 78–93. Devon, UK: Willan.Find this resource:

                                                    Brantingham, P. L., and P. J. Brantingham. 1981. “Notes on the Geometry of Crime.” In Environmental Criminology, edited by P. L. Brantingham and P. J. Brantingham, pp. 27–54. Beverly Hills, CA: Sage.Find this resource:

                                                      Brantingham, P. L., and P. J. Brantingham. 1993. “Nodes, Paths and Edges: Considerations on the Complexity of Crime and the Physical Environment.” Journal of Environmental Psychology 13(1): 3–28.Find this resource:

                                                        Canter, D., and P. Larkin. 1993. “The Environmental Range of Serial Rapists.” Journal of Environmental Psychology 13(1): 63–69.Find this resource:

                                                          Clare, J., J. Fernandez, and F. Morgan. 2009. “Formal Evaluation of the Impact of Barriers and Connectors on Residential Burglars’ Macro-Level Offending Location Choices.” Australian and New Zealand Journal of Criminology 42(2): 139–58.Find this resource:

                                                            Clarke, R. V. 1983. “Situational Crime Prevention: Its Theoretical Basis and Practical Scope.” Crime and Justice 4:225–56.Find this resource:

                                                              Clarke, R. V., and D. B. Cornish. 1985. “Modelling Offenders’ Decisions: A Framework for Research and Policy.” Crime and Justice 6:147–85.Find this resource:

                                                                (p. 419) Cohen, L. E., and M. Felson. 1979. “Social Change and Crime Rate Trends: A Routine Activity Approach.” American Sociological Review 44:588–608.Find this resource:

                                                                  Cornish, D. B., and R. V. Clarke. 1986. Reasoning Criminal: Rational Choice Perspectives on Offending. New York: Springer-Verlag.Find this resource:

                                                                    Costello, A., and P. Wiles. 2001. “GIS and the Journey to Crime: An Analysis of Patterns in South Yorkshire.” In Mapping and Analysing Crime Data: Lessons from Research and Practice, edited by A. Hirschfield and K. J. Bowers, pp. 27–60. London: Taylor and Francis.Find this resource:

                                                                      Coupe, T., and L. Blake. 2006. “Daylight and Darkness Targeting Strategies and the Risks of Being Seen at Residential Burglaries.” Criminology 44(2): 431–64.Find this resource:

                                                                        Davies, T., and S. D. Johnson. 2015. “Examining the Relationship Between Road Structure and Burglary Risk via Quantitative Network Analysis.” Journal of Quantitative Criminology 31(3): 481–507.Find this resource:

                                                                          Elffers, H. 2004. “Decision Models Underlying the Journey to Crime.” In Punishment, Places and Perpetrators: Developments in Criminology and Criminal Justice Research, edited by G. Bruinsma, H. Elffers and J. De Keijser, pp. 182–97. Portland, OR: Willan.Find this resource:

                                                                            Felson, M. 2006. Crime and Nature. Thousand Oaks, CA: Sage.Find this resource:

                                                                              Frank, R., V. Dabbaghian, A. Reid, S. Singh, J. Cinnamon, and P. Brantingham. 2011. “Power of Criminal Attractors: Modeling the Pull of Activity Nodes.” Journal of Artificial Societies and Social Simulation 14(1): 6.Find this resource:

                                                                                Guevara, C. A., and M. E. Ben-Akiva. 2013. “Sampling of Alternatives in Multivariate Extreme Value (MEV) Models.” Transportation Research Part B: Methodological 48:31–52.Find this resource:

                                                                                  Guimarães, P., O. Figueirdo, and D. Woodward. 2003. “A Tractable Approach to the Firm Location Decision Problem.” Review of Economics and Statistics 85(1): 201–4.Find this resource:

                                                                                    Haberman, C. P., and J. H. Ratcliffe. 2015. “Testing for Temporally Differentiated Relationships Among Potentially Criminogenic Places and Census Block Street Robbery Counts.” Criminology 53(3): 457–83.Find this resource:

                                                                                      Hägerstrand, T. 1970. “What About People in Regional Science?” Papers of the Regional Science Association 24:7–21.Find this resource:

                                                                                        Johnson, S. D. 2014. “How Do Offenders Choose Where to Offend? Perspectives from Animal Foraging.” Legal and Criminological Psychology 19:193–210.Find this resource:

                                                                                          Johnson, S. D., and K. J. Bowers. 2004. “The Stability of Space–Time Clusters of Burglary. British Journal of Criminology 44(1): 55–65.Find this resource:

                                                                                            Johnson, S. D., and L. Summers. 2015. “Testing Ecological Theories of Offender Spatial Decision Making Using a Discrete Choice Model.” Crime and Delinquency 61(3): 454–80.Find this resource:

                                                                                              Johnson, S. D., L. Summers, and K. Pease. 2009. “Offender as Forager? A Direct Test of the Boost Account of Victimization.” Journal of Quantitative Criminology 25(2): 181–200.Find this resource:

                                                                                                Kitamura, R., C. Chen, and R. Narayanan. 1998. “Traveler Destination Choice Behavior: Effects of Time of Day, Activity Duration, and Home Location.” Transportation Research Record 1645:76–81.Find this resource:

                                                                                                  Lammers, M. 2014. “Catch Me if You Can: Using DNA Traces to Study the Influence of Offending Behaviour on the Probability of Arrest.” Amsterdam: Vrije Universiteit.Find this resource:

                                                                                                    Lammers, M., B. Menting, S. Ruiter, and W. I. M. Bernasco. 2015. “Biting Once, Twice: The Influence of Prior on Subsequent Crime Location Choice.” Criminology 53(3): 309–29.Find this resource:

                                                                                                      McFadden, D. 1973. “Conditional Logit Analysis of Qualitative Choice Behavior.” In Frontiers in Econometrics, edited by P. Zarembka, pp. 105–42. New York: Academic Press.Find this resource:

                                                                                                        (p. 420) McFadden, D. 1978. “Modeling the Choice of Residential Location.” In Spatial Interaction Theory and Planning Models, edited by A. Karlkvist, L. Lundkvist, F. Snikars, and J. Weibull, pp. 75–96. Amsterdam: North Holland.Find this resource:

                                                                                                          Morselli, C., and M.-N. Royer. 2008. “Criminal Mobility and Criminal Achievement.” Journal of Research in Crime and Delinquency 45(1): 4–21.Find this resource:

                                                                                                            Openshaw, S. 1984. Concept and Techniques in Modern Geography, Number 38: The Modifiable Areal Unit Problem. Norwich, CT: Geo Books.Find this resource:

                                                                                                              Pires, S. F., and R. V. Clarke. 2011. “Sequential Foraging, Itinerant Fences and Parrot Poaching in Bolivia.” British Journal of Criminology 51(2): 314–35.Find this resource:

                                                                                                                Ratcliffe, J. H. 2006. “A Temporal Constraint Theory to Explain Opportunity-Based Spatial Offending Patterns.” Journal of Research in Crime and Delinquency 43(3): 261–91.Find this resource:

                                                                                                                  Rengert, G. F. 2004. “The Journey to Crime. In Punishment, Places and Perpetrators: Developments in Criminology and Criminal Justice Research, edited by G. Bruinsma, H. Elffers, and J. De Keijser, pp. 169–81. Portland, OR: Willan.Find this resource:

                                                                                                                    Rengert, G. F., and J. Wasilchick. 2000. Suburban Burglary: A Tale of Two Suburbs. Springfield, IL: Charles C Thomas.Find this resource:

                                                                                                                      Reynald, D., M. Averdijk, H. Elffers, and W. Bernasco. 2008. “Do Social Barriers Affect Urban Crime Trips? The Effects of Ethnic and Economic Neighbourhood Compositions on the Flow of Crime in The Hague, The Netherlands.” Built Environment 34(1): 21–31.Find this resource:

                                                                                                                        Tajalizadehkhoob, S., H. Asghari, C. Gañán, and M. van Eeten. 2014. “Why Them? Extracting Intelligence About Target Selection from Zeus Financial Malware.” Paper presented at the Proceedings of the 13th Annual Workshop on the Economics of Information Security, WEIS 2014, State College, PA, June 23–24, 2014.Find this resource:

                                                                                                                          Townsley, M., D. Birks, W. Bernasco, S. Ruiter, S. D. Johnson, G. White, and S. Baum. 2015. “Burglar Target Selection. A Cross-national Comparison.” Journal of Research in Crime and Delinquency 52(1): 3–31.Find this resource:

                                                                                                                            Townsley, M., D. Birks, S. Ruiter, W. Bernasco, and G. White. 2016. “Target Selection Models with Preference Variation Between Offenders.” Journal of Quantitative Criminology 32:283.Find this resource:

                                                                                                                              Townsley, M., and A. Sidebottom. 2010. “All Offenders Are Equal, But Some Are More Equal than Others: Variation in Journeys to Crime Between Offenders.” Criminology 48(3): 897–917.Find this resource:

                                                                                                                                van Daele, S., and W. Bernasco. 2012. “Exploring Directional Consistency in Offending: The Case of Residential Burglary in The Hague.” Journal of Investigative Psychology and Offender Profiling 9(2): 135–48.Find this resource:

                                                                                                                                  van Koppen, M. V., H. Elffers, and S. Ruiter. 2011. “When to Refrain from Using Likelihood Surface Methods for Geographic Offender Profiling: An Ex Ante Test of Assumptions.” Journal of Investigative Psychology and Offender Profiling 8(3): 242–56.Find this resource:

                                                                                                                                    Vandeviver, C., T. Neutens, S. van Daele, D. Geurts, and T. Vander Beken. 2015. “A Discrete Spatial Choice Model of Burglary Target Selection at the House-Level.” Applied Geography 64:24–34.Find this resource:

                                                                                                                                      von Haefen, R. H., and A. Domanski. 2013. “Estimating Mixed Logit Models with Large Choice Sets.” Paper presented at the Third International Choice Modelling Conference, Sydney.Find this resource:

                                                                                                                                        Weisburd, D., W. Bernasco, and G. Bruinsma. 2008. Putting Crime in Its Place. New York: Springer.Find this resource:

                                                                                                                                          Wiles, P., and A. Costello. 2000. The ‘Road to Nowhere’: The Evidence for Travelling Criminals, vol. 207. London: Research, Development and Statistics Directorate, Home Office.Find this resource: