1. Introduction
As a branch of environmental criminology research, crime pattern research has attracted the attention of many researchers due to its potential in helping people understand the dimensions of risk and optimize the efficiency of crime reduction effort [
1,
2]. As an important discovery in crime pattern research, the repeat and near-repeat (RNR) phenomenon provides insight into a more precise understanding of crime distribution. The RNR phenomenon suggests that crimes are spatially and temporally correlated. The information can be transferred from one crime to another crime within a certain distance and a certain time period [
3,
4,
5]. According to the crime types, research of RNR can be separated into two categories: RNR within a single crime type and RNR across different crime types.
RNR is primarily examined within a single crime type. Bowers and Johnson suggested that a recorded burglary victim, as a starting point in RNR, is a variant of risk predictors for burglaries [
3]. The offenders would become more familiar with the modus operandi on a victimized place or object and, subsequently, repeat similar offense types in nearby households. Street robbery and gun violence share similar RNR patterns according to other research [
4,
6].
Compared with classic research on the risk after each crime, other research has considered spatial-temporal hotspots with RNR calculation [
7,
8,
9]. Hotspots can be classified into spatial hotspots, temporal hotspots, and spatial-temporal hotspots [
7]. In this study, a ‘spatial-temporal hotspot’ is referred to as a ‘hotspot.’ The information dissemination mechanism among offenders exists among crime types.
Compared with the distinct relationship among the same type crimes, the research of correlations among different types of crimes need to be investigated. Johnson et al. attempted to detect the existence of RNR patterns among different crime types [
10,
11,
12]. Using burglary and theft from motor bicycle data, the RNR pattern was investigated. The results indicate a lack of support for this hypothesis. However, further research is “clearly warranted to see if such patterns exist” among other crime types [
10]. The observed phenomenon contributed additional knowledge of crime distribution to research and is expected to benefit the patrol efficacy of practitioners. Existing research provides a new opportunity to gain insight into the impact of hotspots on the crime rate of different types (
Figure 1).
In this paper, we will examine the spatial-temporal pattern of two crime types: Pocket-picking (PP) and vehicle, motor vehicle theft (VMVT) using crime data collected by the police department in a large Chinese city. First, we will review previous research and outline the theoretical framework of the tested questions. We will present the RNR patterns after hotspots with different crime types after the RNR test within and across the two listed crime types. This research will determine whether risk can be transferred within crime type. In the third part, we will consider the risk undulation from the results and determine if risk can be transferred among crime types by concluding the crime contagion pattern. This analysis can help optimize crime prevention strategies and improve patrol efficacy.
PP is defined as the theft of property carried by somebody else on ‘public area’ or ‘public transports.’ ‘Public area’ and “carried by somebody else’ are two core components of the definition of VMVT. Corresponding to the definition of VMVT, the VMVT can be defined as the theft of a bicycle, motorcycle, or car that cannot carried by somebody else in a ‘public area.’
2. Literature Review
The crime propagation phenomena, which is known as RNR, have attracted a substantial amount of attention in criminological research. Two hypotheses account for the RNR phenomenon: The boost hypothesis and the flag hypothesis. The boost account suggests that short-run space-time clustering is boosted by the initial crime [
13,
14]. Previous offenders will transfer information or experience of the victims to future offenders. Future offenders return to the targeted victims and commit another crime by experience. The following crimes seem to be ‘boosted’ by the initial crime. The notion that most clustered crimes are committed by serial offenders or gang members strongly supports the ‘boost’ hypothesis [
10,
15,
16]. The flag account argues that most repeatedly victimized places (or locations) are flagged by opportunistic offenders due to their special properties. Offenders are assumed to be attracted by the targets and to commit similar crimes on the same (or nearby) victims in a period of time [
13]. Both boost accounts and flag accounts have been employed to interpret RNR victimizations.
Many crime theories can account for the RNR phenomenon. Crime pattern theory recognizes place as a core factor that affects the crime distribution [
17]. For example, places with mixed land use (e.g., commercial and civic institutional recreation land use) are usually subjected to high risk because they are comparatively more attractive to offenders than other people [
18]. By this theory, potential offenders that are attracted by particular objects (or locations) are assumed to repeatedly commit similar crimes or inform other group members to commit similar crimes by sharing the learned experience/modus operandi, such as what boost account is hypothesized. Routine activity theory argues that people’s routine activity determines crime distribution [
19]. In this theory, potential offenders repeat daily routine activities and commit a crime until sufficient conditions (copresented unprotected target and motivated offenders) are fulfilled [
20]. From the viewpoint of routine activity theory, the RNR is derived from people’s routine activities. For many of the offenders, committing crimes is a part of routine activities. Vulnerable objects are more likely to satisfy the conditions and are easily violated by potential offenders. In this case, vulnerable objects have roles as information transmitters which can distribute risk information to similar potential offenders. All aforementioned theories provide explanations of RNR and accept that the experience of victimization (modus operandi and object’s characteristics) can be spatially and temporally transformed within the same crime type.
Though the RNR phenomenon informs of the increased risk after each crime victimization, the phenomena of crime spatiotemporal displacement focus on the decreased crime risk after hotspots, according to Nakaya’s and Zengli’s research [
8,
9]. According to their research, the risk will not increase immediately after hotspots but will immediately decrease in the proximity area and then “return” to the same locations after a short period. The phenomena of crime transformation can be interpreted as a risk-avoiding activity of offenders. When the crime risk within an area is excessive, the potential offenders can feel the increased risk and then displace it to other places or other crime types. When the risk disappeared, potential offenders may return and search for more crime opportunities at the same locations. In the ‘return’ activity, potential offenders are attracted by objects after the risk is decreased or even disappears [
18]. In this case, the ‘avoiding’ and ‘return’ activities indicate that the risk status within hotspots can be ‘felt’ by the potential offenders. The existence of experience transformation between two crimes is further confirmed. However, similar to RNR research, crime displacement research remains within a single crime type.
Beyond single crime type research, crimes of different types are also hypothesized to be related based on several common facts. For example, some breaking and entering offenses may change to rape after the male potential offender sees a single female in a burglary [
21]. A VMVT may change to robbery if the offender was felt/caught at the moment of stealing [
22]. Other hypotheses support the existence of a relationship between two different types. For example, some offenders may displace their modus operandi if they determine that their current modus operandi is outdated or at high risk of being caught by a policeman. Several kinds of displacement exist [
23]. The change in the modus operandi is a classic displacement type. Offenders can displace to other crime types after they determine that a targeted area (or object) is subject to high risk. However, the current literature on the relationship between two crime types yields negative results. Burglary and theft from motor vehicle (TFMV) are two significantly different crime types. Most burglaries occur in victims’ homes, while TFMV primarily (37 percent) occur in public areas, such as streets and parking lots [
11]. Furthermore, many of the techniques (e.g., car opening) associated with auto theft are distinct to those of burglary [
11,
12]. It is difficult for an offender to easily duplicate the technique or experience learned in a burglary to a TFMV and vice versa. This difficulty can cause a lack of information exchange between the offenders of two crimes and objectively lead to the non-existence of the RNR pattern between two crimes. Further research about the relationship between two distinct crime types remains vague.
In current research, PP and VMVT are selected as research objects. VMVT victimizations are spatially clustered according to the literature [
24,
25,
26,
27]. Obviously, the parking lots and garages where many VMVT are situated with little guardianship are usually identified as VMVT hot areas [
24,
28,
29]. Further research indicates that a large proportion of VMVTs occur at residential locations [
30]. Research about PP crime indicates that PPs are concentrated in some special places [
31,
32,
33,
34]. A great number of PPs occurs at bus stops [
31] and tourist attractions [
32]. In short, these two crime types have been investigated and have been reportedly clustered in space. However, further research about the relationship between distinct crime types is still vague.
Inspired by the current gap, three questions are proposed in this study:
Question 1: Are the PPs and VMVTs spatially and temporally correlated within one crime type in a large Chinese city?
Many empirical and experimental studies have verified the existence of the RNR phenomenon with burglary data of Chinese cities [
8,
15,
35,
36,
37]. However, this research may be the first attempt to investigate the existence of the RNR phenomenon among PP and VMVT records in a large Chinese city. This experiment can determine whether crimes are spatially and temporally correlated within each crime type.
Question 2: Does the interaction effect exist among different crime types?
Though current research has denied the spatial-temporal relationship between two crime types, further research about the RNR relationship among other crime types is warranted [
10]. The results of this experiment will heighten our understanding of the relationship among crime types.
Question 3: If hypothesis 2 is confirmed, then another hypothesis will be proposed; that is, whether or not a significant undulation of crime risk occurs after hotspots with the heterogenous crime type.
Though this research has acknowledged that hotspots can affect the crime risk within the proximity area for a period within the same crime type, the impact on other crime types remains to be investigated. Based on this research, two distinct risk statuses exist immediately after hotspots of another crime type: Increased risk and decreased risk. After a hotspot, the potential offenders are hypothesized to change their modus operandi and even commit other kinds of crimes. The crime risk of another crime type will increase immediately after hotspots. Correspondingly, the high crime risk caused by the hotspots can also be hypothesized to transform the risk information to other crime types offenders and deter them from this area for a period. Additional experiments are needed to determine which of these hypotheses is correct.
6. Discussion
In this research, the relationship between the two crime types was further investigated. The three questions about this relationship were answered by several experiments.
First, the phenomenon of RNR in PP and bicycle and motor bicycle theft were experimentally suggested—probably the first attempt—to exist in a large Chinese city. The existence of RNR has been experimentally examined with many types of crimes in current research [
4,
10,
46,
47,
48,
49]. Few studies have investigated RNR with the two types of crimes. The research results indicate that both types of crimes tend to be spatially clustered. When one crime victimization occurs, the same victimization will probably occur at the same location or locations in the vicinity in the following period. The boost and flag hypothesis may account for this cluster of crimes [
3,
40]. Based on the boost hypothesis, experienced offenders are more inclined to commit serial crimes within the adjacent spatial-temporal extent. The flag hypothesis argues that some particular attributes of vulnerable objects will be flagged to attract potential offenders. Both hypotheses can be employed to interpret the existence of an RNR pattern in the two crime types in a public area.
For the second question, interaction occurs between the two types of crimes. The spatial extent with increased risk after another type of crime occurs is limited within 100 m. Combined with the results from
Figure 3, the findings concluded that the two types of crimes are spatially clustered. Note that the narrow spatial extent deserves more attention due to its potential ability to benefit crime prediction and policing patrol. This result indicates a strong interaction between the two crime types. The possibility of another VMVT victimization after PP may be higher than that after a previous VMVT. Subsequently, PP victimization may have a better ability to predict VMVT victimizations. The results can be interpreted as a displacement of modus operandi between two types. The interaction between the two crime types can be reasonably determined. Both PP and vehicle/motor vehicle theft are usually committed in crowded public places. A dense crowd provides a large number of objects for potential pocket-pickers and other thieves. Most of these public areas are crowded with people and bicycle/motor bicycles and provide numerous targets for both potential offenders. Consider bus stops as an example. A large number of people take buses every day. A high percentage of these people will ride/drive the distance from their homes to the bus stops. Therefore, these bus stops in a large city become gathering places for a large number of passengers and their different means of transport [
26]. The spatially overlapped objects cause a strong interaction among potential offenders and a large proportion of overlapped hotspots of the crime types. The spatially limited interaction may increase the interaction and transformation between the two crimes types. Since the objects of two crime types are similar, the information or experience acquired from one crime type can easily be learned by the potential offenders of another crime type. The previously mentioned reasons produce a strong interaction between the two types of crimes. This result is different from that of the existing research [
10]. This gap is attributed to the different crime types in the study. Burglary and theft from motor bicycle are two distinct crime types in terms of modus operandi and their location. PP and VMVT are similar for crime locations and acquired experience. Thus, further research is “warranted to determine the existence of the interaction” among other crime types [
10].
Following to the third question, the risk undulation of one crime type after another crime type was also investigated. The impact of the two crime types on each other are similar; when a hotspot of one crime type appears, the possibility of another crime will increase. This finding supports the interaction between the two crime types. In addition, the increase in risk after hotspots of another crime type can be interpreted as modus operandi displacement. The offenders can change their targets after experiencing the high crime risk, even of another type. Some differences are observed in the impact of hotspots of one crime type on the risk of other crime types. After the hotspots of VMVT, the risk area of PP spreads to distanced locations (>300 m) from the limited area (<100 m). Conversely, the risk area of VMVT after PP hotspots converge from a large area (<100 m and >300 m) to a limited area (<100 m). The reason for this difference should be caused by the different distributions of targets of the two types of potential offenders. The bicycle and motor bicycles are usually parked at specified areas in the large Chinese city. The crimes cannot spatially displace a large amount. Conversely, the potential offenders of PP usually search for targets in pedestrians or crowd, which can distribute in a wider range of area than a parking area. Risk-avoiding activities occur within the area between 100 m and 300 m. Therefore, the risk can be transferred across crime types.