Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
The Concept and Technologies of Quality of Geographic Information Service: Improving User Experience of GIServices in a Distributed Computing Environment
Previous Article in Journal
A Novel Dynamic Dispatching Method for Bicycle-Sharing System
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
IJGI
Volume 8
Issue 3
10.3390/ijgi8030119
ijgi-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

An Examination of Spatial Differences between Migrant and Native Offenders in Committing Violent Crimes in a Large Chinese City

by
Jiaxin Feng
1,
Lin Liu
2,3,*,
Dongping Long
2,* and
Weiwei Liao
1
1
Center of Integrated Geographic Information Analysis, School of Geography and Planning, Sun Yat-sen University, Guangzhou 510275, China
2
School of Geographical Sciences, Center of GeoInformatics for Public Security, Guangzhou University, Guangzhou 510006, China
3
Department of Geography, University of Cincinnati, Cincinnati, OH 45221-0131, USA
*
Authors to whom correspondence should be addressed.
ISPRS Int. J. Geo-Inf. 2019, 8(3), 119; https://doi.org/10.3390/ijgi8030119
Submission received: 25 December 2018 / Revised: 7 February 2019 / Accepted: 25 February 2019 / Published: 1 March 2019
Browse Figures
Figure 1
<p>Conceptual framework for comparing migrant and native violent crimes.</p> "> Figure 2
<p>Box maps for different violent crimes at the neighborhood level in the whole city and in the old town. (<b>a</b>) Migrant violent crimes; (<b>b</b>) Native violent crimes.</p> ">
Versions Notes

Abstract

:
Immigrants and natives are generally comparable in committing violent crimes in many Western cities. However, little is known about spatial differences between internal migrant offenders and native offenders in committing violence in contemporary urban China. To address this gap, this research aims to explore spatial variation in violent crimes committed by migrant and native offenders, and examine different effects of ambient population, crime attractors, crime generators, and offender anchor points on these crimes. Offender data, mobile phone data, and points-of-interest (POI) data are combined to explain the crime patterns of these offenders who committed offenses and were arrested from 2012 to 2016 in a large Chinese city by using box maps and negative binomial regression models. It is demonstrated that migrant and native violent crimes vary enormously across space. Ambient population is only positively related to migrant violent crimes. Crime attractors and generators have more significant and stronger correlations with migrant violent crimes, while offender anchor points have a stronger association with native violent crimes. The results reveal that migrant offenders tend to be attracted by larger amounts of people and more affected by crime attractors and generators than native offenders.

1. Introduction

Violent crime, one of the serious social ills, is generally comparable between immigrant and native offenders in many Western cities [1,2,3,4]. There is no consensus as to whether immigrants are more prone to the commission of violence, because it depends on the origin of country, ethnicity, generation, or location [5,6,7]. Interestingly, second generation immigrants became more involved in crimes than the children of native-born parents in the US and Netherlands [8,9]. The main reasons for this are culture conflict, family-related factors, an undesirable neighborhood environment, and involvement in gangs [6,10,11,12]. However, immigrants do not always have a higher rate of offenses than natives. Indeed, immigrants are ascribed to fewer crimes than natives as a number of studies point out [13,14,15,16,17,18,19,20]. The explanation for the low crime rates of immigrants is straightforward: they came to a new country from far away to seek economic opportunities and to improve their life, and therefore behaved themselves [9].
Comparatively, little is known about differences in involvement in violent crimes between internal migrants and natives. The lack of research might reflect the almost complete urbanization of Western countries [21]. In the US, internal migration reached its peak around 1980 and then presented a steady downward trend [22,23]. Moreover, an international comparison shows that the internal mobility rate is even lower in European countries and other developed countries than in the US. However, internal migration is still a significant demographic factor promoting the process of urbanization in many developing countries [24,25,26]. According to the United Nations Population Fund, 244 million people were international immigrants, taking up 3.3 percent of the world’s population in 2015. In the same year, there were 247 million internal migrants in China announced by the Chinese Health and Family Planning Commission. The volume of Chinese internal migration is roughly the same as that of international immigration around the globe. Notably, one of the negative consequences of a large influx of internal migration to a city is an increase in urban crime [27,28,29]. The non-Western context of China contributes to addressing the aforementioned research gap by exploring violent offending among internal migrants and natives.
In urban China, internal migrants arrive at another domestic city with the intention to obtain employment, unlike international immigrants who move to another country for the main purpose of settlement. It is not that Chinese migrants do not want to settle down. In fact, they are held back by the institutional barriers of the household registration system [30,31]. The Chinese household registration system, also known as Hukou, was officially established in the late 1950s to strictly prevent the geographic mobility of internal migration in the country. It defines the welfare of residents in a specific place by classifying residents by the place of household registration. Since the 1980s the system has gradually loosened its control by conditionally providing rural residents with access to urban residency [32]. A significant migration population has moved from rural to urban areas or from underdeveloped to developed areas in response to the demand for employment opportunities in big cities or special economic zones. It has been argued that this floating unemployed population contributes to an increase in crime and is one of the social problems brought about by freedom of movement [28]. Conversely, native residents are seldom mentioned in crime statistics. In the mid-1990s, 80% of solved crimes were committed by migrants in some districts in Shanghai [33]. Meanwhile, migrant offenders took up 90% among prostitutes and drug traffickers in Guangdong province. In the 2000s, 80% to 90% of police arrests were migrants in Shenzhen and ZG city [34,35]. Recent research proves that the proportion of migrants is criminogenic in Chinese coastal provinces where most of the big cities are located [36].
To date, research on the pattern of crimes committed by non-native or native offenders remains to be excavated in both Western and non-Western contexts regarding the interaction between potential victims, motivated offenders, and the environment. Social and physical environment factors are inherent in the commission of crimes [37]. Encouragingly, Johnson has examined the spatial pattern and volume of European Union (EU) migrant criminal activity at the English Police Force level in England, and suggests that more detailed microscale analysis concerning the opportunity and spatial behaviors of offenders is of great benefit to gain insight into the location of offending [38]. Bernasco and Block have verified the spatial influence of crime generators, crime attractors, and offender anchor points on robberies ascribed to all kinds of robbers at the block level in Chicago [39]. Studies in ZG city, China show that ambient population, resident’s daily activities, and the locations of amenities affect the pattern of theft from the person across time and space [40,41,42]. As Brantingham and Brantingham point out, crimes do not take place randomly or uniformly across social groups [43]. The patterns of non-native and native criminal activities and their comparison is intriguing.
This article aims to assess the spatial differences between migrant and native violent crimes in contemporary urban China to gain a better understanding of the patterns of violent crimes ascribed to migrant and native offenders. It pays specific attention to migrant offenders who register in other domestic cities but live in the city of ZG, and native offenders who register and live in this city. These two residential groups are differentiated by the household registration system according to the place of household registration. Different from prior research, the article compares migrant and native violent crimes in a non-Western context, and explores their spatial patterns to unveil the spatial behaviors of offenders by using invaluable offender data and multi-source big data like mobile phone data and points-of-interest (POI) data. It is a valuable empirical study that addresses significant research gaps and supplements the scarce literature in environmental criminology, crime geography, and urban safety.

2. Theoretical Foundation, Research Questions, and Conceptual Framework

To uncover the crime patterns of migrants and natives, opportunity theories such as routine activity theory and crime pattern theory undoubtedly provide a useful analytical framework for the explanation in crime events. Under the guidance of these two theories, research questions are put forward and a conceptual framework is made to be examined in this paper.

2.1. Routine Activity Theory and Crime Pattern Theory

Routine activity theory highlights the convergence of three elements in time and space to promote crimes: likely offenders, suitable targets, and capable guardians [44]. Offenders complete criminal activities under the cover of legal activities of everyday life. This crime event triangle was successively revised by Eck and Felson [45,46]. Inside of the triangle includes the target, the offender and the place, and outside of the triangle includes the guardian for the target, the handler for the offender, and the manager for the place.
Crime pattern theory explains how activity nodes, travel paths, and networks of people influence the spatiotemporal concentration of crimes [43,47]. It takes full account of the routine activity space of mobile offenders and mobile targets, as well as the locations of stationary targets and networks of offenders. The pattern is dynamic, complex, and comprehensive, embedded with the social and physical environment in the city.
On the basis of the two aforementioned theories, the interaction between offenders, targets and guardians, and the interaction between people and the environment are of great importance to the occurrence of crime. Accordingly, some research sheds light on various crime patterns by analyzing the influence of the ambient population, crime attractors, crime generators, and offender anchor points. Generally, ambient population indicates potential victims. Crime attractors and crime generators are places or larger areas that trigger crime events, while offender anchor points represent motivated offenders who are likely to commit crimes near their homes due to restricted mobility.

2.1.1. Ambient Population

For violent crimes, the targets will be persons and the size of the potential victim population is dynamic in space and time. The same holds true for the guardians. Knowing where people actually are located becomes critical. An alternative measurement of the population at risk is the ambient population in the existing research [48]. To capture the ambient population, diverse data sources have been applied, including remote sensing dataset, travel survey data, workday census data, Twitter data, mobile phone data, taxi ridership data, and subway ridership data [40,49,50,51,52,53]. For example, population density and population change are estimated from mobile phone users’ locations in a recent study to discern the effect of ambient population on snatch-and-run offenses [54]. Since most people carry their mobile phones in everyday life, it is reliable to calculate ambient population from mobile phone data with higher spatiotemporal resolution.

2.1.2. Crime Attractors and Crime Generators

Crime attractors are some particular places that provide well known criminal opportunities thus pulling in motivated offenders [55]. These places do not necessarily experience a large number of people, but they facilitate motivated offenders to find victims or targets with ease. Typical examples include but are not limited to bars, drug markets, prostitution areas, and parking lots. A handful of studies verify that the numbers of crime attractors such as bars and clubs, check-cashing stores, drug incidents, prostitution incidents, and gambling incidents have a positive impact on robbery counts per block [39,56].
Different from crime attractors, crime generators are featured with a large number of people passing by [55]. These places produce crimes because a concentration of people activates the criminal motivation of some potential offenders who had no intention of committing a criminal offense in the beginning. In this regard, shopping and entertainment districts, sports stadiums, workplaces, public transport stations, and high schools are some typical examples. Supportive evidence suggests that the overall level of crime elevates around the stadium because the facility generates crime with large numbers of ambient population [57].
In reality, pure crime attractors and pure crime generators seldom exist as most areas are mixed [43]. Treating commercial activity as both a crime attractor and generator, a study of land use and urban crime opportunities reveals that the assault rate is extremely high at regional shopping centers, followed by neighborhood pubs, community shopping centers, motels and auto courts, and fast-food restaurants [58]. Another study finds that more neighborhood disorder and crimes are perceived by residents who live in the vicinity of more crime-generating land uses (subway stations, expressway off-ramps, and high schools) and more crime-attracting land uses (halfway houses, homeless shelters, pawnbrokers, check-cashing stores, drug-treatment centers, beer establishments, and liquor clubs) [59]. Both crime attractors and crime generators are criminogenic places where concentrations of crime occur.

2.1.3. Offender Anchor Points

Offender anchor points generally refer to the residences of offenders. The residential history has an effect on a criminal’s crime location choice. There is strong evidence that offenders who commit robberies and assaults are more likely to re-offend near their current and prior residences rather than other similar areas [60]. Further, violent crimes are inclined to occur in close proximity to offender anchor points than property crimes [61]. In Chicago, many violent crimes are executed at or very near the offenders’ home addresses [62]. It has been tested that the concentration of offender anchor points is related to the spatial distribution of street robbery across Chicago census blocks [39]. The activity space of individuals is limited, and the presence and concentration of offender anchor points probably promote the occurrence of violent crimes.

2.2. Research Questions and Conceptual Framework

Although no prior research has compared migrant violent crimes with native violent crimes, some studies have investigated general crime patterns for specific crime types in different cities and thus provide references for the current study. To examine migrant and native violent crimes, two research questions and a conceptual framework are put forward.
The first research question is: Where do migrant and native offenders commit violent crimes respectively? The research on crime began with its geography and the scientific interest in the crime geography has long existed [63]. The spatial patterns of crimes vary across social groups. Migrant and native residents are quite different in terms of demographic, economic and housing characteristics. More specifically, Chinese migrant residents tend to be younger and less educated, gaining lower income and mostly residing in urban villages [64,65]. Previous studies point out that individual characteristics and home locations are significant factors that determine the characteristics of activity space [66,67,68]. In this regard, socioeconomic differentiation between these two groups may result in diverse daily movement patterns, and may further lead to distinct spatial patterns of violent crimes.
The second research question to be addressed is: What is the extent to which ambient population, crime attractors, crime generators, and offender anchor points affect the occurrence of violent crimes committed by migrant and native offenders respectively? The roles of these influencing factors are highly illustrated in routine activity theory and crime pattern theory [39,49,56]. However, whether they can fully explain offending in China remains unknown. The applicability of these Western theories needs to be further tested in the context of China [69].
A conceptual framework is designed to illustrate the relationship between two research questions and the comparison between migrant and native violent crimes in Figure 1. We suppose that violent crimes committed by migrant and native offenders occur at different locations. Further, they are hypothesized to be significantly impacted by some or all factors from routine activity theory and crime pattern theory in varying degrees, which are labeled with influence A and influence B respectively. To test these hypotheses, initially, spatial distributions of migrant and native crimes are demonstrated and then spatial differences are described. Afterwards, influencing factors are examined for a possible explanation for diverse spatial patterns, and the effects of influencing factors (influence A and B) are compared between two groups.

3. Data and Methods

This section introduces data sources and analytical methods used to address and compare migrant and native violent crimes. First, data sources and procedures of case selection are described. Second, operationalization of variables and application of negative binomial models are explained.

3.1. Data

This article uses several sources of data collected from different organizations in the city of ZG, a large coastal city in southern China. Due to a confidentiality agreement, the real name of this city is concealed. Despite the requirement of anonymity, the context of ZG city makes it ideal for research on migrant and native violent crimes. As one of the most prosperous metropolitan cities in China, ZG city has witnessed a rapid growth of resident population due to the large influx of migrants. Meanwhile, migrant residents become the predominant group in committing violence, according to a crime database that records offenses since 2012. Therefore, ZG city is chosen as a representative city in China. Because of the availability of mobile phone data, only nine districts of ZG city are included for the current study. In line with the ZG city administrative division in 2010 when the census was conducted, the study area is composed of 1973 neighborhoods, which have an average size of 1.62 square kilometers and an average population of 5480.
The first source of data is offender data recorded by the ZG Municipal Public Security Bureau. It consists of offending and personal information on offenders arrested by the police for criminal offenses. All offenders were arrested in the years 2012–2016, and violent offenders are extracted for the current study. Their violent offenses include simple assault, aggravated assault, robbery, and rape. To distinguish between migrant and native offenders, the location of household registration is used for each violent offender. Subsequently, offenders registering in ZG city are classified as native offenders, while the others registering in other internal cities or villages are considered as migrant offenders. The number of offenses, crime types, addresses of violent offenses, and residence addresses of violent offenders are used for statistical analysis. The gender, age, household registration, education, occupation, drug taking, number of offenders and their offenses are described to reveal the basics of the offender sample.
The second data source is anonymized and aggregated mobile phone data obtained from one of the three largest telecommunication operators in China. ZG city is divided into numerous cells by the telecommunication operator, but the actual shape and location of each cell remain confidential. Instead, the latitudes and longitudes of cell towers were provided. We are suggested by the telecommunication operator to create Thiessen polygons to represent the service areas of cells. In sum, there are 52,026 cell towers. For each cell tower, an estimated number of mobile phone users whose devices communicate with it are calculated on a per hour basis for the period of 12–18 May 2016. These statistics were computed by the telecommunication operator according to mobile phone activities such as making or receiving a call, sending or receiving a message, using mobile data, etc. Though mobile phone count is a proxy for population, it is one of the practical measures of ambient population at risk [49].
The third source is POI data collected from a cartographic information company for ZG city in 2016. It contains various types of amenities with their own names and geographic locations. These amenities cover lodging, dining, recreational, financial, commercial, retail, educational, medical, cultural, transport facilities and others. Some of them are reasonable representations of crime generators and crime attractors. Besides, police stations are included and classified as one of the political-legal organs.
The fourth source is census data provided by the ZG Municipal Bureau of Statistics in 2010. It includes diverse demographic, socioeconomic, and housing attributes of residents, such as household registration, age, education, occupation, type of accommodation, marital status, housing area, rent, etc. Residential district and residential neighborhood are the two kinds of census units in the census data in China. In general, a residential district contains several or a dozen residential neighborhoods. The residential neighborhood is a proxy for the community and therefore it is chosen as the analysis unit in the article.

3.2. Procedures of Case Selection

There are three main procedures for the case selection in offender data. The first step is to identify migrant offending incidents and native offending incidents. Co-offending by both natives and migrants are excluded in the sample. The second step is to select cases committed by migrant and native offenders who live in ZG city according to their residence addresses. All remained cases took place in the ZG city in the years 2012–2016. The third step is to remove cases that happened in six residential neighborhoods without records of housing attributes in census data. These neighborhoods are not added to models due to a lack of data, and migrant and native violent crimes in these neighborhoods are subsequently removed. As a result, 17,118 cases of violent crimes committed by 26,054 offenders in ZG city are selected for the following analysis. Specifically, 13,929 cases (82.9%) are attributed to 21,601 migrant offenders (81.4%), whilst 3189 cases (17.1%) are attributed to 4453 native offenders (18.6%).
Table 1 presents some statistics of offenders and crimes for migrant and native residents respectively. In general, the indicators of migrants are much higher than those of natives regardless of the count or rate. The offender data reveals that migrant residents have a much higher rate of offending than native residents in ZG city. Migrant and native violent crimes follow a similar pattern. The count or rate of migrant violent crime is four times that of native violent crime. Moreover, the number of offenders is much larger than that of offenses. A major reason is that co-offending takes up 34.5% of all violent offenses.
Table 2 provides an overview of migrant and native violent offenders who are responsible for selected cases. Most migrant and native offenders are male but the percentage of females is slightly higher among native offenders. Migrant offenders are younger than native ones as the percentages of age groups 12–18, 19–30, and 31–45 are higher for migrant offenders. All the native offenders and 69.0% migrant offenders register in the Central South of China, whereas 31.0% of migrant offenders come from other parts of China. Basically, migrant offenders have a lower level of education, but they present a lower percentage of unemployment (73.9%) compared with native offenders (78.5%). Two groups of employed violent offenders are mainly occupied in sales and services, manufacturing industry, primary industry, business, finance, and management. More native offenders are found to have problems with drug taking.

3.3. Measurement of Variables

In order to examine migrant and native violent offending, two dependent variables are constructed for migrant and native violent crimes. Each one is measured by a 5-year total number of crimes at the neighborhood level. As shown in Table 3, for example, all neighborhoods experience a mean of 7.1 migrant violent crimes that range from 0 to 262 and a mean of 1.6 native violent crimes that range from 0 to 22 in the years 2012–2016. Both mean and standard deviation indicate that overdispersion exists in two independent variables. In accordance with the Mann–Whitney test, two independent variables vary significantly (z = 23.748, Prob > |z| = 0.0000).
Three subsets of independent variables are constructed to measure ambient population, crime attractors, crime generators, and offender anchor points. To estimate potential victims, ambient population per neighborhood is measured by using mobile phone data. First, Thiessen polygons are created based on the locations of cell towers derived from mobile phone data. Second, Thiessen polygons are clipped according to the shape of the study area. Third, a geometric intersection of Thiessen polygons and neighborhoods is computed. Fourth, daily mobile phone users are distributed into each intersection by weighting the percentage of intersection area in a Thiessen polygon. At last, daily mobile phone users are aggregated to neighborhoods due to a one-to-one correspondence between a neighborhood and its intersections. Thus, the ambient population can be calculated by aggregating the average daily mobile phone users in a week per neighborhood to indicate potential victims.
Before measuring crime attractors and crime generators, we first analyze the preference of target choices of migrant and native offenders. In light of the offender data, approximately 69% of migrant offenders and 68% of native offenders would tell their targeted places to the police. Our statistics show that both migrant and native offenders pay particular attention to places like bars, trade markets, parks and squares, restaurants, groceries, bus stops, parking lots, and hotels. Based on this, counts of these places for each neighborhood are calculated using POI data to represent crime attractors and crime generators. Since some crime attractors and crime generators may experience similar spatial patterns, a correlation analysis is conducted. As a result, the number of groceries is highly correlated with the number of restaurants, and so are the number of bus stops and parking lots. We finally select the number of bars, trade markets, parks and squares, restaurants, and hotels for the following analysis. Notably, trade markets have a huge difference from gambling markets, narcotics markets, and prostitution markets; instead, they refer to farmers’ markets and wholesale markets. The presence of businesses like shops and markets, no matter legal or illegal, is relevant with the frequent occurrence of crimes [39,56,59,70].
In line with the measurement of offender anchor points in the existing literature [39,71], the number of offender residence addresses per neighborhood is calculated for migrant and native offenders respectively based on offender data. These offenders reside in ZG city and are responsible for selected cases. A residence address will be recorded repeatedly for an offender on the condition that he/she had committed different kinds of violent crimes. For instance, the number will be four if an offender commits simple assault, aggravated assault, robbery, and rape respectively. However, the number will be one if an offender commits the same type of violence more than once.
Police and several social environment variables serve as control variables. Police is indicated by the distance to the nearest police station using POI data. The distance is measured in kilometers between the centroid of the neighborhood and the location of the nearest police station. Social environment characteristics including registration heterogeneity, percentage rental, divorce rate, and population are revealed by census data at the neighborhood level. Ethnic heterogeneity has been widely tested in many Western countries, but registration heterogeneity is chosen in the context of China because of the significant impact of household registration systems on population mobility and population heterogeneity [32,72], and the diverse influence of various registered population groups on urban crime [73]. Specifically, registration heterogeneity is measured by an index of qualitative variation using one minus the sum of the squared percentages of three household registration categories: (1) registration in ZG city, (2) registration outside ZG city but in the same province, and (3) registration in other provinces. In addition, percentage rental is the percentage of rental properties among all house units in the neighborhood. The divorce rate is the number of divorced residents divided by surveyed residents in the neighborhood. Population is the total number of migrant and native residents in the neighborhood.

3.4. Negative Binomial Models

To model violent crime counts, negative binomial regression models are used with regard to overdispersion. The formula for the negative binomial regression model is shown as follows:
ln ( λ i ) = ∑ k = 0 k β k χ i k + ε i
P ( Y i = y i ) = Γ ( y i + τ ) y i ! Γ ( τ ) τ τ λ i y i ( τ + λ i ) τ − y i
Equation (1) is a regression equation calculating the natural logarithm of the expected number of migrant or native violent crime counts λi for the i neighborhood. It is a sum of each independent variable xik multiplied by a regression coefficient βk, and a random variable representing the unobserved part of the conditional expectation function or neighborhood heterogeneity. Equation (2) means that the probability of the observed number of migrant or native violent crime counts yi for the i neighborhood follows a negative binomial distribution. Additionally, Γ is the gamma function and τ is a fuzzy parameter indicating the degree of overdispersion [74].
Separate models will be estimated for migrant and native violent crime counts to address their differences. To examine these violent offenses with police and social environment variables controlled, all neighborhoods are selected for inclusion at first but six neighborhoods are removed because of lack of housing attributes. In this regard, models for migrant and native violent crimes only apply to 1967 neighborhoods in ZG city, whose population ranges from 134 to 51,447. Rather than include population size as an exposure measure in the model, we treat it as a control variable in tune with previous studies [39,56].
We conduct collinearity diagnostics by calculating variation inflation factors (VIF) of independent variables prior to the estimation. High VIF value means a serious multicollinearity problem and a benchmark value of 10 is commonly used. The highest and the mean VIF values are reported for each model. Furthermore, we report overdispersion parameter alpha and its likelihood ratio test result, spatial autocorrelation Pearson residual, and Akaike information criterion (AIC) for each model after estimation. Alpha significantly larger than zero suggests that the negative binomial model is more appropriate than the Poisson model. To test residual spatial autocorrelation, we calculate the residual by subtracting the predicted count from the observed count and compute Moran’s I for it using queen contiguity weighted average. To assess goodness-of-fit, we adopt AIC whose lower value indicates better goodness-of-fit. Both coefficients and incident rate ratios (IRRs) are presented in the models that explain the relationships between violent crimes and ambient population, crime attractors, crime generators, and offender anchor points for migrants and natives. The difference between IRR and one multiplied by 100 indicates a percentage change of violent crimes when one unit increases in an independent variable [75].

4. Results

4.1. Spatial Patterns of Violent Crimes Committed by Migrant and Native Offenders

Box maps are drawn to illustrate the spatial patterns of migrant and native violent crimes in Figure 2. The box map is an extension of quantile choropleth map with particular attention to spatial outliers [76]. Its classification is consistent with that used in the box plot, including lower outlier, less than 25%, 25% to 50%, 50% to 75%, more than 75%, and an upper outlier based on the number of violent crimes. In a box map, numbers and locations of classified neighborhoods, especially those more than 75% and upper outliers, indicate the spatial pattern of crimes.
As can be seen in Figure 2, violent crimes committed by migrant offenders are more clustered in the whole city, while violent crimes committed by native offenders are more spatially scattered. For instance, migrant violent crimes concentrate around the city center in the middle of the study area, and also gather in suburban areas. Comparatively, native violent crimes mainly spread among suburban areas. Moreover, the old town, which is depicted in the rectangular region next to the whole study area in Figure 2, includes many neighborhoods with a relatively large number of native violent crimes, but only includes limited neighborhoods with lots of migrant violent crimes. In short, violent crimes committed by migrant and native offenders vary enormously across space.

4.2. Influencing Factors of Violent Crimes Committed by Migrant and Native Offenders

To analyze how the spatial patterns of migrant and native violent crimes are shaped, negative binomial models are estimated to examine the influence of ambient population, crime attractors, crime generators, and offender anchor points, controlling for police represented by the distance to the nearest police station and social environment characteristics including registration heterogeneity, percentage rental, divorce rate, and population. Model results are reported in Table 4, providing coefficients and incidence rate ratios for violent crimes committed by migrant and native offenders respectively, and diagnostic information is provided at the bottom of the table.
Interestingly, the effect of daily mobile phone users remains significantly positive on migrant violent crimes, but is not significant on native violent crimes. In the migrant group, with a one-unit increase in mobile phone users in the neighborhood, the number of violent crimes is expected to increase by 16.0%.
The number of bars, hotels, trade markets, restaurants, parks and squares significantly increase both migrant and native violent crime counts, but the significant level and the extent of influence have subtle differences. Specifically, numbers of hotels and trade markets are associated with the count of migrant violent crimes at the most significant level, but are only linked with the count of native violent crimes at the least significant level. Moreover, numbers of hotels, trade markets, parks and squares have a greater impact on migrant violent crimes than native violent crimes as IRRs suggest, while the effects of bars and restaurants are on the opposite.
Residences of offenders are associated with violent crimes committed by migrant and native offenders at the most significant level. However, the impact of residences of native offenders on native violent crime is much larger than the impact of residences of migrant offenders on migrant violent crimes (16.6% vs. 1.7%). Adding one native offender residing in the neighborhood will result in a 16.6% increase in the expected number of native violent crimes, 14.9% larger than that for migrant violent crimes.
Between two independent variables, migrant violent crimes have the most significant and strongest association with mobile phone users, while native violent crimes have the most significant and strongest association with residences of offenders.
Effects of control variables reveal the relationship between police and violent crimes, and the relationship between the social environment and violent crimes in the neighborhood. Interestingly, the distance to the nearest police station has a negative influence on different violent crimes, perhaps because victims report more crimes to the police at short distances. Increasing registration heterogeneity in the neighborhood is expected to bring more migrant violent crimes, while the higher divorce rate is negatively associated with the occurrence of migrant violent crimes. The percentage of rental only significantly reduces native violent crimes.
Diagnostic statistics are informative. The significance of two alpha values means that considerable overdispersion exists in the data and the negative binomial model performs better than a Poison model. Both the largest and mean VIFs are low, which indicate that these models are free of the collinearity problem. Spatial autocorrelation Pearson residuals of two violent crimes are quite low, indicating that they can be generally explained by independent variables, especially migrant violent crimes. AIC value in the migrant model is much larger than that in the native model, as there are much more migrant violent crimes.

5. Discussion

In the current study, migrant violent crimes do outnumber native violent crimes, which is generally consistent with previous studies. The prevalence of violent crimes among migrant offenders can be interpreted in a socioeconomic background. Looking back, income inequality inevitably elevates alongside the great success in economic revitalization in China [77,78,79]. Massive migration into the city because of income disparities makes it easier for internal migration to commit crimes [80]. Moreover, violence is regarded as the traditions connected with rural life and is brought into the city by the migrants from rural areas [81]. Nevertheless, migrant residents tend to lead a normal life in urban areas on the condition that more secure jobs and social welfare are accessible. In recent years, a series of household registration system reform programs have been carried out in China to accelerate migrants transforming into new citizens in urban areas. A significant step was the establishment of residence permit system in 2016. A migrant who possesses a residence permit will have some equal rights as a native citizen in terms of employment, basic public education, basic medical service, etc. The reform has been playing an active part because volume of internal migration shows a downward trend according to the Chinese Health and Family Planning Commission.
The differences between migrant and native offenders in committing violent crimes are majorly assessed by the spatial diversity of offenses and the influence of environmental factors. In particular, we examine ambient population, crime attractors, crime generators, and offender anchor points to uncover how diverse spatial patterns are shaped. The influence of these environmental factors on migrant and native violent crime counts will be further discussed to reveal the spatial behaviors of migrant and native offenders.
Ambient population serves as a contributing factor for the occurrence of migrant violent crimes instead of native violent crimes. Before testing simultaneous effects of all environmental factors, the only inclusion of ambient population makes the residual autocorrelation close to zero in the model of migrant violent crimes, but the residual autocorrelation is larger than zero in the model of native violent crimes. The simultaneous regression results further support the positive effect of ambient population on migrant violent crimes. Treating all offenses as equal, ambient population is proved to better assess the population at risk for person-targeted crimes such as violence and theft from the person [49,82]. The distribution of ambient population reveals where people conduct activities in a day. These people are likely to be potential victims that are targeted by motivated violent offenders. Comparatively, migrant offenders are more likely to be attracted to places that provide more suitable targets in spite of a higher risk of being detected or captured, while native offenders may pay less attention to a larger amount of ambient population. In space, migrant violent crimes mainly concentrate in urban areas that include a large amount of ambient population, whereas more outliers of native violent crimes distribute in less populated suburban areas. Meanwhile, it is found that migrant offenders mainly reside in urban areas, while many native offenders live in suburban areas. That is to say that distributions of violent crimes are similar with locations of residence for both groups. Offenders have their normal movement patterns and where they offend is close to their activity and awareness space [43]. Therefore, compared with native offenders, migrant offenders are likely to carry out daily routines and commit crimes in those areas with more people.
Generally, more crime attractors and crime generators have more significant positive relationships with migrant violent crimes than native violent crimes. The locations of significant crime attractors and crime generators indicate where offenders choose targets on purpose and come across suitable targets in the course of their daily lives, respectively. Bars, hotels, restaurants, parks and squares are regarded as criminogenic facilities or risky land-use sites for crimes [39,55,56,58,83]. They act as parts of many people’s activity space and awareness space. A time use survey for Chinese residents in 2008 suggests public places like parks, hotels, and bars are popular sites for residents to relax, receives services, entertain, and socialize [84]. Defining these land use types as crime attractors and crime generators is a comprehensive task. For example, bars and hotels at quiet time attract motivated offenders acting like crime attractors, while restaurants, parks and squares at peak time are daily activity nodes of many people acting like crime generators. Conversely, bars at peak time can be regarded as crime generators for violence, and parks and squares at quiet time can be considered as crime attractors [85]. Besides, trade markets with people buying and selling can become crime attractors or generators when crowd density is intermediate, since there are both potential victims and motivated offenders. This is especially true for robbery. Although it is hard to indicate aforementioned land use types as either crime attractors or crime generators, examining their relationships with violent crimes can further reveal different patterns of migrant and native offenders. Hotels, trade markets, parks and squares have a greater influence on migrant violent crimes, suggesting that migrant offenders are more likely to search for criminal opportunities or conduct daily activities in these places. It is known that migrant residents actively participate in trade markets in China, since a higher proportion of migrants engage in commerce [65]. A great number of well-established migrant settlements are famous for outstanding wholesale markets, such as Zhejiangcun in Beijing and Xiaohubei in Guangzhou [86,87]. In contrast, the influence of bars and restaurants is greater on native violent crimes. This indicates native offenders are more likely to search for criminal opportunities or came across potential victims at bars or restaurants.
Offender anchor points have a greater impact on native violent crimes than on migrant violent crimes. The close connection between residences of offenders and counts of violent crimes implies that home locations and offense locations of offenders have similar spatial patterns. Prior research suggests that it is most likely for predatory violence to occur in neighborhoods that accommodate many offenders [62]. In this scenario, one possible explanation for the difference between two groups is that native offenders are more likely to take their homes as anchor points in order to find suitable targets in surrounding areas. As is shown in the regression models, neighborhoods with more native violent offenders are likely to be prone to more native violent crimes. To better understand aggregate-level results, future work should further examine the influence of offender anchor points on offense locations at the individual level, for example, calculating distance to crime. Another possible explanation is that the effect of ambient population on native violent crimes is insignificant, and the effects of crime attractors and crime generators on native violent crimes are limited. As a result, the effect of offender anchor points becomes noticeable. In comparison, migrant violent crimes are also strongly associated with ambient population and many crime attractors and crime generators, although highly related to offender anchor points.
Certain caveats and limitations are important to mention in this study. First, offender data has some issues with accuracy and bias that limit us to reflect on the true underlying offenders. It is less inclusive than crime data to present the total volume of crimes [88], since not all offenders can be caught red-handed or afterwards. Besides, it only reflects those offenders who draw the attention of police officers [89]. As migrants are deemed to have a higher propensity for crimes, it is doubtful whether migrants are more likely to be reported by victims and obtain the attention of police officers. In this regard, further investigation is warranted. In spite of these issues, we still consider offender data as an invaluable source of data that is worthy of mining. It reveals diverse crime templates of migrant and native offenders. Second, mobile phone data also has its limitations in accuracy. Mobile phone count does not represent all persons in space and time; instead, it is an approximate estimation of mobile phone users who receive services from a telecommunication operator. Nevertheless, it could be regarded as random sampling to indicate whether a large amount of persons are present in places. Considering that service areas of a cell tower are estimated by ourselves and the data is only available for one week, mobile phone data with higher spatial resolution and a longer period of time is expected to more reliably capture ambient population. Finally, the interaction between potential guardians, motivated offenders, and suitable victims should be better tested in the future, if three elements can be estimated exactly in time and space in line with their refined definitions [90,91]. Notably, empirical research has confirmed that victims of violence are also criminal offenders and both share similar characteristics in terms of demography, structure, and behavior [92,93,94]. However, the victim-offender overlap cannot be assessed by using the current dataset. More efforts must be made to link victims and offenders in future research.

6. Conclusions

Making an investigation into violent crimes attributed to migrant and native offenders, this article first uncovers their spatial diversity in a Chinese context not studied previously. It is demonstrated that offense locations of migrant offenders are more clustered outside the old town, whereas those of native offenders are more scattered in the whole study area.
Furthermore, the differences in crime patterns of migrant and native offenders are found based on routine activity theory and crime pattern theory. Ambient population, crime attractors, crime generators, and offender anchor points do have an influence on violent crime counts, but these impacts differ in two groups. Ambient population is only positively related to migrant violent crimes, while other factors are positively related to both migrant and native violent crimes. Moreover, more crime attractors and crime generators have more significant and greater correlations with migrant violent crimes, while offender anchor points have a stronger association with native violent crimes. Migrant and native offenders potentially develop different activity space and awareness space. While migrant offenders tend to be attracted by larger amounts of people and more affected by crime attractors and crime generators, native offenders are on the opposite side and their violent crimes are well predicted by offender anchor points.
This research represents the very first study in comparing migrant and native offenses in a Chinese city. It contributes to filling the vacant area in previous research on crimes and applying Western theories to contemporary urban China. The findings may help local police department take targeted measures to discourage violence committed by migrants and natives in different locations and land use types. Given that migrant crimes are predominant, more effective allocation of police resources for them may introduce substantial decline in the overall crime rate.
Future research, especially spatio-temporal crime patterns, may further address the distinction between migrants and natives in committing violence. It is likely that spatio-temporal patterns for migrants are much more concentrated than for natives. Analysis of spatio-temporal factors could promote an understanding of how dynamic patterns are shaped. There is also an opportunity for a natural experiment to compare migrant and native crime rates before and after 2016 when the residence permit system came into existence. Changes in living conditions for migrants in the city are likely to discourage their crime rates, thus reducing the disparity between migrant and native offending. Last but not least, a comparison could be conducted with a focus on international migrants in China, as they tend to be different from internal migrants and natives in terms of activity space and awareness space. Accordingly, variation in committing crimes is expected across space and time.

Author Contributions

J.F. and L.L. designed the experiments; J.F. and W.L. processed the data; J.F. analyzed the data and prepared the figures and tables; J.F. wrote the paper; L.L. revised the paper; D.L. gave suggestions for the revision of the paper.

Funding

This research was supported by the Natural Science Foundation of China (No. 41531178), the Guangzhou Science and Technology Program Key Projects (No. 201804020016), and the Natural Science Foundation of Guangdong Province (No. 2014A030312010).

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Bersani, B.E.; Loughran, T.A.; Piquero, A.R. Comparing patterns and predictors of immigrant offending among a sample of adjudicated youth. J. Youth Adolesc. 2014, 43, 1914–1933. [Google Scholar] [CrossRef] [PubMed]
  2. Jennings, W.G.; Zgoba, K.M.; Piquero, A.R.; Reingle, J.M. Offending trajectories among native-born and foreign-born Hispanics to late middle age. Sociol. Inq. 2013, 83, 622–647. [Google Scholar] [CrossRef]
  3. Sampson, R.J. Rethinking crime and immigration. Contexts 2008, 7, 28–33. [Google Scholar] [CrossRef]
  4. Ramiro Martinez, J.; Lee, M.T. On immigration and crime. Crim. Justice 2000, 1, 485–524. [Google Scholar]
  5. Rosenwaike, I.; Hempstead, K. Mortality among three Puerto Rican populations: Residents of Puerto Rico and migrants in New York city and in the balance of the United States, 1979–1981. Int. Migr. Rev. 1990, 24, 684–702. [Google Scholar] [CrossRef] [PubMed]
  6. Ferracuti, F. European migration and crime. In Crime and Culture: Essays in Honor of Thorsten Sellin; Wolfgang, M.E., Ed.; John Wiley & Sons: New York, NY, USA, 1968. [Google Scholar]
  7. Ignatans, D.; Roebuck, T. Do more immigrants equal more crime? Drawing a bridge between first generation immigrant concentration and recorded crime rates. Crime Secur. Soc. 2018, 1. [Google Scholar] [CrossRef]
  8. Vinkers, D.J.; Duits, N. Crime and mental disorders among native Dutch and ethnic minority juvenile defendants in the Netherlands. Int. J. Law Psychiatry 2011, 34, 131–134. [Google Scholar] [CrossRef] [PubMed]
  9. Tonry, M. Ethnicity, crime, and immigration. Crime Justice 1997, 21, 1–29. [Google Scholar] [CrossRef]
  10. Hällsten, M.; Sarnecki, J.; Szulkin, R. Crime as a price of inequality? The delinquency gap between children of immigrants and children of native Swedes. Br. J. Criminol. 2013, 53, 456–481. [Google Scholar] [CrossRef]
  11. McNulty, T.L.; Bellair, P.E. Explaining racial and ethnic differences in serious adolescent violent behavior. Criminology 2003, 41, 709–747. [Google Scholar] [CrossRef]
  12. Ross, H. Crime and the native born sons of European immigrants. J. Crim. Law Criminol. (1931–1951) 1937, 28, 202–209. [Google Scholar] [CrossRef]
  13. Piquero, A.R.; Bersani, B.E.; Loughran, T.A.; Fagan, J. Longitudinal patterns of legal socialization in first-generation immigrants, second-generation immigrants, and native-born serious youthful offenders. Crime Delinq. 2016, 62, 1403–1425. [Google Scholar] [CrossRef]
  14. Bersani, B.E. An examination of first and second generation immigrant offending trajectories. Justice Q. 2014, 31, 315–343. [Google Scholar] [CrossRef]
  15. Ramiro Martinez, J.; Lee, M.T. Comparing the context of immigrant homicides in miami: Haitians, Jamaicans and Mariels. Int. Migr. Rev. 2000, 34, 794–812. [Google Scholar] [CrossRef]
  16. Hagan, J.; Palloni, A. Immigration and crime in the United States. In The Immigration Debate; Smith, J.P., Edmonston, B., Eds.; National Academy Press: Washington, DC, USA, 1998. [Google Scholar]
  17. Sellin, T. Culture conflict and crime. Am. J. Sociol. 1938, 44, 97–103. [Google Scholar] [CrossRef]
  18. Taft, D.R. Nationality and crime. Am. Sociol. Rev. 1936, 1, 724–736. [Google Scholar] [CrossRef]
  19. Lind, A.W. Some ecological patterns of community disorganization in Honolulu. Am. J. Sociol. 1930, 36, 206–220. [Google Scholar] [CrossRef]
  20. Abbott, G. Immigration and crime. J. Am. Inst. Crim. Law Criminol. 1915, 6, 522–532. [Google Scholar] [CrossRef]
  21. Zhong, H.; Xu, J.; Piquero, A.R. Internal migration, social exclusion, and victimization: An analysis of Chinese rural-to-urban migrants. J. Res. Crime Delinq. 2017, 54, 479–514. [Google Scholar] [CrossRef]
  22. Cooke, T.J. Internal migration in decline. Prof. Geogr. 2013, 65, 664–675. [Google Scholar] [CrossRef]
  23. Molloy, R.; Smith, C.L.; Wozniak, A. Internal migration in the United States. J. Econ. Perspect. 2011, 25, 173–196. [Google Scholar] [CrossRef]
  24. Chan, K.W. China: Internal migration. In The Encyclopedia of Global Human Migration; Ness, I., Ed.; Wiley-Blackwell: Hoboken, NJ, USA, 2013. [Google Scholar]
  25. Charles, A.; Denis, M. Internal migration in Ghana: Determinants and welfare impacts. Int. J.Soc. Econ. 2012, 39, 764–784. [Google Scholar] [CrossRef]
  26. Lusome, R.; Bhagat, R. Trends and patterns of internal migration in India, 1971–2001. In Proceedings of the Annual Conference of Indian Association for the Study of Population (IASP), Thiruvananthapuram, India, 7–9 June 2006. [Google Scholar]
  27. Zhang, K.H.; Song, S. Rural–urban migration and urbanization in China: Evidence from time-series and cross-section analyses. China Econ. Rev. 2003, 14, 386–400. [Google Scholar] [CrossRef]
  28. Curran, D.J. Economic reform, the floating population, and crime: The transformation of social control in China. J. Contemp. Crim. Justice 1998, 14, 262–280. [Google Scholar] [CrossRef]
  29. Davin, D. Images of the migrant in the Chinese media. In Proceedings of the Workshop on European—Chinese and Chinese Domestic Migration, Oxford, UK, 3–5 July 1996. [Google Scholar]
  30. Chan, K.W.; Buckingham, W. Is China abolishing the hukou system? China Q. 2008, 195, 582–606. [Google Scholar] [CrossRef]
  31. Wang, F.; Zuo, X. Inside China’s cities: Institutional barriers and opportunities for urban migrants. Am. Econ. Rev. 1999, 89, 276–280. [Google Scholar] [CrossRef]
  32. Chan, K.W.; Zhang, L. The hukou system and rural-urban migration in China: Processes and changes. China Q. 1999, 160, 818–855. [Google Scholar] [CrossRef]
  33. Xu, L. Guanhao liudong renkou wending shehui [managing transient population, stabilizing society]. Leg. Dly. 1995, 1. [Google Scholar]
  34. Liu, L.; Feng, J.; Ren, F.; Xiao, L. Examining the relationship between neighborhood environment and residential locations of juvenile and adult migrant burglars in China. Cities 2018, in press. [Google Scholar] [CrossRef]
  35. Wang, Z. Crime Research in Three Big Economy Regions in Contemporary China; China People’s Public Security University Press: Beijing, China, 2006. [Google Scholar]
  36. Yijing, L.I. Spatio-temporal change of crime at provincial scale in China—Since the economic reform. Asian J. Criminol. 2017, 12, 303–340. [Google Scholar] [CrossRef]
  37. Chainey, S.; Ratcliffe, J. GIS and Crime Mapping; John Wiley & Sons, Ltd.: Chichester, UK, 2005. [Google Scholar]
  38. Johnson, D.E.U. Migrant criminal activity: Exploring spatial diversity and volume of criminal activity attributed to inter EU migrants in England. Appl. Geogr. 2014, 50, 48–62. [Google Scholar] [CrossRef]
  39. Bernasco, W.; Block, R. Robberies in Chicago: A block-level analysis of the influence of crime generators, crime attractors, and offender anchor points. J. Res. Crime Delinq. 2011, 48, 33–57. [Google Scholar] [CrossRef]
  40. Song, G.; Liu, L.; Bernasco, W.; Xiao, L.; Zhou, S.; Liao, W. Testing indicators of risk populations for theft from the person across space and time: The significance of mobility and outdoor activity. Ann. Am. Assoc. Geogr. 2018, 108, 1–19. [Google Scholar] [CrossRef]
  41. Long, D.; Liu, L.; Feng, J.; Zhou, S.; Jing, F. Assessing the influence of prior on subsequent street robbery location choices: A case study in ZG city, China. Sustainability 2018, 10, 1818. [Google Scholar] [CrossRef]
  42. Song, G.; Xiao, L.; Zhou, S.; Long, D.; Zhou, S.; Liu, K. Impact of residents’ routine activities on the spatial-temporal pattern of theft from person. Acta Geogr. Sin. 2017, 72, 356–367. [Google Scholar] [CrossRef]
  43. Brantingham, P.J.; Brantingham, P.L. Crime pattern theory. In Environmental Criminology and Crime Analysis; Wortley, R., Rombouts, M.L.S., Eds.; Willan: Devon, UK, 2008; pp. 78–93. [Google Scholar]
  44. Cohen, L.E.; Felson, M. Social change and crime rate trends: A routine activity approach. Am. Sociol. Rev. 1979, 44, 588–608. [Google Scholar] [CrossRef]
  45. Felson, M. Those who discourage crime. In Crime and Place; Eck, J.E., Weisburd, D., Eds.; Criminal Justice Press: Monsey, NY, USA, 1995; pp. 53–66. [Google Scholar]
  46. Eck, J.E. Drug Markets and Drug Places. Ph.D. Thesis, University of Maryland, College Park, MD, USA, 1994. [Google Scholar]
  47. Brantingham, P.J.; Brantingham, P.L. Patterns in Crime; Macmillan: New York, NY USA, 1984. [Google Scholar]
  48. Andresen, M.A.; Jenion, G.W. Ambient populations and the calculation of crime rates and risk. Secur. J. 2010, 23, 114–133. [Google Scholar] [CrossRef]
  49. Malleson, N.; Andresen, M.A. Exploring the impact of ambient population measures on London crime hotspots. J. Crim. Justice 2016, 46, 52–63. [Google Scholar] [CrossRef] [Green Version]
  50. Stults, B.J.; Hasbrouck, M. The effect of commuting on city-level crime rates. J. Quant. Criminol. 2015, 31, 331–350. [Google Scholar] [CrossRef]
  51. Malleson, N.; Andresen, M.A. The impact of using social media data in crime rate calculations: Shifting hot spots and changing spatial patterns. Cartogr. Geogr. Inf. Sci. 2015, 42, 112–121. [Google Scholar] [CrossRef]
  52. Felson, M.; Boivin, R. Daily crime flows within a city. Crime Sci. 2015, 4, 31. [Google Scholar] [CrossRef] [Green Version]
  53. Andresen, M.A. The ambient population and crime analysis. Prof. Geogr. 2011, 63, 193–212. [Google Scholar] [CrossRef]
  54. Hanaoka, K. New insights on relationships between street crimes and ambient population: Use of hourly population data estimated from mobile phone users locations. Environ. Plan. B Plan. Des. 2016, 45, 295–311. [Google Scholar] [CrossRef]
  55. Brantingham, P.; Brantingham, P. Criminality of place. Eur. J. Crim. Policy Res. 1995, 3, 5–26. [Google Scholar] [CrossRef]
  56. Haberman, C.P.; Ratcliffe, J.H. Testing for temporally differentiated relationships among potentially criminogenic places and census block street robbery counts. Criminology 2015, 53, 457–483. [Google Scholar] [CrossRef]
  57. Kurland, J.; Johnson, S.D.; Tilley, N. Offenses around stadiums:A natural experiment on crime attraction and generation. J. Res. Crime Delinq. 2014, 51, 5–28. [Google Scholar] [CrossRef]
  58. Kinney, J.B.; Brantingham, P.L.; Wuschke, K.; Kirk, M.G.; Brantingham, P.J. Crime attractors, generators and detractors: Land use and urban crime opportunities. Built Environ. 2008, 34, 62–74. [Google Scholar] [CrossRef]
  59. McCord, E.S.; Ratcliffe, J.H.; Garcia, R.M.; Taylor, R.B. Nonresidential crime attractors and generators elevate perceived neighborhood crime and incivilities. J. Res. Crime Delinq. 2007, 44, 295–320. [Google Scholar] [CrossRef]
  60. Bernasco, W. A sentimental journey to crime: Effects of residential history on crime location choice. Criminology 2010, 48, 389–416. [Google Scholar] [CrossRef]
  61. Schmitz, P.; Cooper, A.K. Using mobile phone data records to determine criminal activity space. In Proceedings of the International GIS Crime Mapping Conference, Brussels, Belgium, 25–26 September 2007. [Google Scholar]
  62. Block, R.; Galary, A.; Brice, D. The journey to crime: Victims and offenders converge in violent index offences in Chicago. Secur. J. 2007, 20, 123–137. [Google Scholar] [CrossRef]
  63. Vandeviver, C.; Bernasco, W. The geography of crime and crime control. Appl. Geogr. 2017, 86, 220–225. [Google Scholar] [CrossRef]
  64. Zheng, S.; Long, F.; Fan, C.C.; Gu, Y. Urban villages in China: A 2008 survey of migrant settlements in Beijing. Eurasian Geogr. Econ. 2009, 50, 425–446. [Google Scholar] [CrossRef]
  65. Fan, C.C. The elite, the natives, and the outsiders: Migration and labor market segmentation in urban China. Ann. Assoc. Am. Geogr. 2002, 92, 103–124. [Google Scholar] [CrossRef]
  66. Perchoux, C.; Chaix, B.; Cummins, S.; Kestens, Y. Conceptualization and measurement of environmental exposure in epidemiology: Accounting for activity space related to daily mobility. Health Place 2013, 21, 86–93. [Google Scholar] [CrossRef] [PubMed]
  67. Rainham, D.; McDowell, I.; Krewski, D.; Sawada, M. Conceptualizing the healthscape: Contributions of time geography, location technologies and spatial ecology to place and health research. Soc. Sci. Med. 2010, 70, 668–676. [Google Scholar] [CrossRef] [PubMed]
  68. Newsome, T.H.; Walcott, W.A.; Smith, P.D. Urban activity spaces: Illustrations and application of a conceptual model for integrating the time and space dimensions. Transportation 1998, 25, 357–377. [Google Scholar] [CrossRef]
  69. Liu, L.; Li, J.C.M. Progress and future directions of crime research in China with selected case studies. J. Res.Crime Delinq. 2017, 54, 447–453. [Google Scholar] [CrossRef]
  70. St. Jean, P.K.B. Pockets of Crime: Broken Windows, Collective Efficacy, and the Criminal Point of View; University of Chicago Press: Chicago, IL, USA, 2007. [Google Scholar]
  71. Bernasco, W.I.M.; Luykx, F. Effects of attractiveness, opportunity and accessibility to burglars on residential burglary rates of urban neighborhoods. Criminology 2003, 41, 981–1002. [Google Scholar] [CrossRef]
  72. Cheng, T.; Selden, M. The origins and social consequences of China’s hukou system. China Q. 1994, 139, 644–668. [Google Scholar] [CrossRef]
  73. Chen, J.; Liu, L.; Zhou, S.; Xiao, L.; Jiang, C. Spatial variation relationship between floating population and residential burglary: A case study from ZG, China. ISPRS Int. J. Geo Inf. 2017, 6, 246. [Google Scholar] [CrossRef]
  74. Gardner, W.; Mulvey, E.P.; Shaw, E.C. Regression analyses of counts and rates: Poisson, overdispersed Poisson, and negative binomial models. Psychol. Bull. 1995, 118, 392–404. [Google Scholar] [CrossRef] [PubMed]
  75. Long, S.J.; Freese, J. Regression Models for Categorical Dependent Variables Using Stata, 3rd ed.; Stata Press Publication, StataCorp LP: College Station, TX, USA, 2014; pp. 80–81. [Google Scholar]
  76. Anselin, L. Interactive techniques and exploratory spatial data analysis. In Geographical Information Systems: Principles, Techniques, Management and Applications; Longley, P.A., Maguire, D.J., Rhind, D.W., Eds.; Wiley: New York, NY, USA, 1999; pp. 251–264. [Google Scholar]
  77. Lo, T.W.; Jiang, G. Inequality, crime and the floating population in China. Asian J. Criminol. 2006, 1, 103–118. [Google Scholar] [CrossRef]
  78. Gustafsson, B.; Shi, L. The anatomy of rising earnings inequality in urban China. J. Comp. Econ. 2001, 29, 118–135. [Google Scholar] [CrossRef]
  79. Khan, A.R.; Riskin, C. Income and inequality in China: Composition, distribution and growth of household income, 1988 to 1995. China Q. 1998, 154, 221–253. [Google Scholar] [CrossRef]
  80. Cao, L.; Dai, Y. Inequality and crime in China. In Crime and Social Control in a Changing China; Liu, J.H., Zhang, L.N., Messner, S.F., Eds.; Greenwood Press: Westport, CT, USA, 2001; pp. 73–88. [Google Scholar]
  81. Liu, J. Crime patterns during the market transition in China. Br. J. Criminol. 2005, 45, 613–633. [Google Scholar] [CrossRef]
  82. Andresen, M.A. Crime measures and the spatial analysis of criminal activity. Br. J. Criminol. 2006, 46, 258–285. [Google Scholar] [CrossRef]
  83. Groff, E.; Weisburd, D.; Morris, N.A. Where the action is at places: Examining spatio-temporal patterns of juvenile crime at places using trajectory analysis and GIS. In Putting Crime in Its Place: Units of Analysis in Geographic Criminology; Weisburd, D., Bernasco, W., Bruinsma, G.J.N., Eds.; Springer: New York, NY, USA, 2009; pp. 61–86. [Google Scholar]
  84. NBS. Time Use Patterns in China; Department of Social Science and Technology Statistics, National Beau of Statistics: Beijing, China, 2010.
  85. Newton, A. Macro-level generators of crime, including parks, stadiums, and transit stations. In The Oxford Handbook of Environmental Criminology; Bruinsma, G.J.N., Johnson, S.D., Eds.; Oxford University Press: Oxford, UK, 2018. [Google Scholar]
  86. Liu, Y.; Li, Z.; Liu, Y.; Chen, H. Growth of rural migrant enclaves in Guangzhou, China: Agency, everyday practice and social mobility. Urban Stud. 2015, 52, 3086–3105. [Google Scholar] [CrossRef]
  87. Zhang, L. Migration and privatization of space and power in late socialist China. Am. Ethnol. 2001, 28, 179–205. [Google Scholar] [CrossRef]
  88. Weisburd, D. The law of crime concentration and the criminology of place. Criminology 2015, 53, 133–157. [Google Scholar] [CrossRef]
  89. Johnson, D. The space/time behaviour of dwelling burglars: Finding near repeat patterns in serial offender data. Appl. Geogr. 2013, 41, 139–146. [Google Scholar] [CrossRef]
  90. Hipp, J.R. General theory of spatial crime patterns. Criminology 2016, 54, 653–679. [Google Scholar] [CrossRef]
  91. Hollis, M.E.; Felson, M.; Welsh, B.C. The capable guardian in routine activities theory: A theoretical and conceptual reappraisal. Crime Prev. Commu. Saf. 2013, 15, 65–79. [Google Scholar] [CrossRef]
  92. Broidy, L.M.; Daday, J.K.; Crandall, C.S.; Sklar, D.P.; Jost, P.F. Exploring demographic, structural, and behavioral overlap among homicide offenders and victims. Homicide Stud. 2006, 10, 155–180. [Google Scholar] [CrossRef]
  93. Miller, S.L. Victims as Offenders: The Paradox of Women’s Violence in Relationships; Rutgers University Press: Piscataway, NJ, USA, 2005. [Google Scholar]
  94. Klevens, J.; Duque, L.F.; Ramírez, C. The victim-perpetrator overlap and routine activities: Results from a cross-sectional study in Bogotá, Colombia. J. Int. Violence 2002, 17, 206–216. [Google Scholar] [CrossRef]
Ijgi 08 00119 g001 550
Figure 1. Conceptual framework for comparing migrant and native violent crimes.
Figure 1. Conceptual framework for comparing migrant and native violent crimes.
Ijgi 08 00119 g001
Ijgi 08 00119 g002 550
Figure 2. Box maps for different violent crimes at the neighborhood level in the whole city and in the old town. (a) Migrant violent crimes; (b) Native violent crimes.
Figure 2. Box maps for different violent crimes at the neighborhood level in the whole city and in the old town. (a) Migrant violent crimes; (b) Native violent crimes.
Ijgi 08 00119 g002
Table
Table 1. Descriptive statistics of crimes and offenders by migrant and native residents.
Table 1. Descriptive statistics of crimes and offenders by migrant and native residents.
MigrantNative
CountRate (*10000)CountRate (*10000)
Violent crimes21,60149.944536.9
 Simple assault835719.315682.4
 Aggravated assault614314.215992.5
 Robbery558112.99571.5
 Rape15203.53290.5
Violent offenders13,92912.931892.9
 Simple assaulters47974.411411.1
 Aggravated assaulters40883.811401.1
 Robbers36203.35970.6
 Rapists14241.33110.3
Table
Table 2. Characteristics of migrant and native violent offenders.
Table 2. Characteristics of migrant and native violent offenders.
MigrantNative
Count%Count%
Gender
Male19,84491.9404290.8
Female17578.14119.2
Age (By 2016)
12–188433.91633.7
19–30977145.2135830.5
31–45821838.0151033.9
46–60256711.9119526.8
61–861880.92225.0
Unknown140.150.1
Household registration
Central South14,90369.04453100.0
South West369617.100.0
East18988.800.0
North2081.000.0
North East4942.300.0
North West4021.900.0
Education
Illiteracy5062.3501.1
Primary school390818.146110.4
Junior high school10,78549.9190042.7
Senior high school255211.897021.8
University7603.54369.8
Postgraduate210.1290.7
Unknown306914.260713.6
Occupation
Primary industry6102.81312.9
Manufacturing industry18838.71794.0
Sales and services242411.245410.2
Government service110.1200.4
Business, finance and Management3571.7571.3
Health130.150.1
Education200.1260.6
Art and culture50.020.0
Religion10.000.0
Out of employment15,95473.9349478.5
Unknown3231.5851.9
Drug taking
No20,51895.0418994.1
Yes2241.0922.1
Unknown8594.01723.9
Table
Table 3. Descriptive statistics of dependent variables and independent variables.
Table 3. Descriptive statistics of dependent variables and independent variables.
VariablesMinMaxMeanSD
Dependent variables
 Migrant violent crimes0.0262.07.114.0
 Native violent crimes0.022.01.62.2
Independent variables
 Mobile phone users (/10,000)0.018.40.91.3
 Bars0.013.00.30.9
 Hotels0.024.00.51.3
 Trade markets0.024.01.52.5
 Restaurants0.054.03.95.7
 Parks and squares0.016.00.30.9
 Residences of migrant violent offenders0.0527.011.028.7
 Residences of native violent offenders0.033.02.33.0
Control variables
 Distance to the nearest police station0.07.70.70.9
 Registration heterogeneity (×10)0.06.73.91.8
 Percentage rental0.01.00.30.3
 Divorce rate0.00.20.10.0
 Population (/10,000)0.05.10.60.5
Table
Table 4. Negative binomial regression models of migrant and native violent crimes per neighborhood. Both coefficients (Coef.) and incidence rate ratios (IRR) are presented.
Table 4. Negative binomial regression models of migrant and native violent crimes per neighborhood. Both coefficients (Coef.) and incidence rate ratios (IRR) are presented.
VariablesMigrant Violent CrimesNative Violent Crimes
Coef.IRRCoef.IRR
Mobile phone users (/10,000)0.148 ***1.160−0.0130.987
Bars0.089 ***1.0930.102 ***1.107
Hotels0.080 ***1.0830.039 *1.039
Trade markets0.058 ***1.0600.025 *1.025
Restaurants0.023 ***1.0230.030 ***1.030
Parks and squares0.081 **1.0840.073 **1.076
Residences of offenders0.016 ***1.0170.154 ***1.166
Distance to the nearest police station−0.230 ***0.795−0.082 **0.921
Registration heterogeneity (×10)0.111 ***1.1180.0131.013
Percentage rental0.0141.014−0.313 **0.731
Divorce rate−6.020 ***0.0020.3401.405
Population (/10,000)−0.0150.9850.1011.107
Alpha0.539 ***0.411***
Largest variance inflation factor2.3902.330
Mean variance inflation factor1.5801.560
Spatial autocorrelation Pearson residual0.0060.024
AIC9976.8636129.172
N19671967
Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

Share and Cite

MDPI and ACS Style

Feng, J.; Liu, L.; Long, D.; Liao, W. An Examination of Spatial Differences between Migrant and Native Offenders in Committing Violent Crimes in a Large Chinese City. ISPRS Int. J. Geo-Inf. 2019, 8, 119. https://doi.org/10.3390/ijgi8030119

AMA Style

Feng J, Liu L, Long D, Liao W. An Examination of Spatial Differences between Migrant and Native Offenders in Committing Violent Crimes in a Large Chinese City. ISPRS International Journal of Geo-Information. 2019; 8(3):119. https://doi.org/10.3390/ijgi8030119

Chicago/Turabian Style

Feng, Jiaxin, Lin Liu, Dongping Long, and Weiwei Liao. 2019. "An Examination of Spatial Differences between Migrant and Native Offenders in Committing Violent Crimes in a Large Chinese City" ISPRS International Journal of Geo-Information 8, no. 3: 119. https://doi.org/10.3390/ijgi8030119

APA Style

Feng, J., Liu, L., Long, D., & Liao, W. (2019). An Examination of Spatial Differences between Migrant and Native Offenders in Committing Violent Crimes in a Large Chinese City. ISPRS International Journal of Geo-Information, 8(3), 119. https://doi.org/10.3390/ijgi8030119

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop