Geospatial Patterns of Property Crime in Thailand: A Socioeconomic Perspective for Sustainable Cities

Abstract

Property crime is a pressing issue in maintaining social order and urban sustainability, particularly in regions marked by pronounced socioeconomic disparity. While the link between socioeconomic stress and crime is well established, regional variations in Thailand have not been fully examined. Therefore, the purpose of this research was to examine spatial patterns of property crime and identify the potential associations between property crime and socioeconomic environment across Thailand. Using nationally compiled property-crime data from official sources across all provinces of Thailand, we employed geographic information system (GIS) tools to conduct a spatial cluster analysis at the sub-national level across 76 provinces. Both global and local statistical techniques were applied to identify spatial associations between property-crime rates and neighborhood-level socioeconomic conditions. The results revealed that property-crime clusters are primarily concentrated in the south, while low-crime areas dominate parts of the north and northeast regions. To analyze the spatial dynamics of property crime, we used geospatial statistical models to investigate the influence of socioeconomic variables across provinces. We found that property-crime rates were significantly associated with monthly income, areas experiencing high levels of household debt, migrant populations, working-age populations, an uneducated labor force, and population density. Identifying associated factors and mapping geographic regions with significant spatial clusters is an effective approach for determining where issues concentrate and for deepening understanding of the underlying patterns and drivers of property crime. This study offers actionable insights for enhancing safety, resilience, and urban sustainability in Thailand’s diverse regional contexts by highlighting geographies of vulnerability.

1. Introduction

Crime is a complex and deep-rooted social problem in many societies worldwide, not only affecting individuals but also significantly infiltrating and undermining the social and economic infrastructure [1,2]. For developing nations with marked economic inequality, the issue of crime is even more significant [3]. Crime not only concerns safety but also signifies the instability of the welfare system and the unequal distribution of resources among communities [3,4]. One type of crime that directly affects public sentiment about security, including financial security, at both the individual and national levels, is property crime, such as theft and break-ins involving individuals, homes, and businesses [5]. Property crime has psychological impacts, resulting in feelings of insecurity and lower confidence in the government and the community as a whole, especially in areas where people have unstable incomes or are burdened with high debt [6,7]. Therefore, criminal behavior is increasingly acknowledged not only as antisocial acts by individuals but also as an indicator of underlying structural economic and social challenges.

The relationship between economic and social inequality and criminal behavior has long attracted the attention of researchers, including criminologists, urban planners, and geographers [8,9]. Although previous research has confirmed this link, the topic of spatial dimensions of crime in diverse landscapes and societies remains an area of ongoing investigation [10,11].

In Southeast Asia, Thailand is experiencing rapid economic growth while simultaneously grappling with persistent economic and social inequalities across its regions [12]. This spatial aspect has become increasingly important in addressing property crime in Thailand. Understanding where and why property crime occurs requires a closer examination of economic and social variables and how these factors interact with spatial characteristics such as population density, infrastructure quality, and urban layout [13,14].

In Western criminology, a large body of research documents the link between economic inequality and higher crime rates [15,16,17]. Spatial analyses of urban environments have identified consistent relationships between neighborhood-level poverty indicators and property-crime incidence. For example, a previous study involving spatial analysis in Vancouver highlighted the influence of social disorder and the structure of daily activities on the distribution of crime. The study highlighted how micro-geographic differences in socioeconomic characteristics such as income, education, and housing can predict crime rates within specific census tracts [10].

Similarly, studies conducted in European cities have shown that factors such as unemployment, income inequality, and housing insecurity are among the predictors of theft and robbery. In one study, researchers analyzed the cities of Glasgow and Birmingham in the United Kingdom and discovered that, over a 15-year span, regions characterized by enduring poverty had elevated and more spatially concentrated crime rates [18]. Another recent study in Portugal used spatial modeling to reveal that disadvantaged areas—particularly those with low education, high population density, and persistent poverty—tend to have significantly higher property-crime rates. This spatial clustering of crimes in socioeconomically marginalized neighborhoods reinforces that social inequality is not just an underlying factor but the main driver of the geographic concentration of crime [19].

From this spatial perspective, Sun et al. investigated the role of environmental variables in property crime. The study found crime risk to be higher in areas with poor urban planning, including insufficient lighting, inadequate surveillance, and fragmented public spaces [20]. The findings indicate that social and environmental vulnerabilities interact within particular geographic contexts to influence the geographical distribution of property crime.

The Western-focused literature cannot always be easily extended to Asian contexts. Recent years have seen a growing body of spatial criminology research emerging from Asia, underscoring both similarities and unique regional dynamics compared to Western settings. For example, Chen et al. applied spatial regression to analyze residential burglary in a major Chinese city, highlighting the role of urban form and community structure in crime distribution [21]. They found that densely built-up neighborhoods with fragmented street networks tended to produce higher burglary rates, indicating that the physical layout of a neighborhood may either encourage or prevent criminal activity. In India, Patel and Singh used kernel density estimation to map theft hotspots across Mumbai’s municipal wards, revealing how informal settlement patterns influence crime clustering [22]. According to their research, there were noticeably higher rates of theft in places with unplanned housing and mixed land use, which usually do not have official street lighting. Their findings highlight how infrastructure deficits function as key risk factors in shaping crime distribution. In Malaysia, Ahmad et al. conducted a spatial–temporal investigation of violent crime in Selangor, Kuala Lumpur, and Putrajaya, revealing that the country’s rapid urbanization and migration patterns produced persistent hot spots and uneven crime trends over time [23].

Recent studies within Thailand have begun to illuminate the local spatial dynamics of property and violent crime in that country. Using hotspot analysis in Nakhon Pathom province, Phlicharoenphon and Rober demonstrated that commercial and transportation hubs frequently attract theft and burglary, and that built-environment elements such as sidewalk connectivity and public lighting have a significant deterrent effect on criminal behavior [24]. Another study used partial least squares structural equation modeling (PLS-SEM) to examine how socioeconomic indicators at the provincial level, such as the proportion of migrant labor and the unemployment rate, relate to property-crime rates. However, because the method treats each province as an independent unit, it does not account for spatial clustering or reveal how the strength of these relationships may vary across different regions [25]. A recent homicide study in Thailand employed hotspot analysis on district-level data, revealing strong spatial clustering of incidents and showing that the importance of predictors such as population density and CCTV coverage can vary dramatically from district to district [26]. While these domestic studies offer valuable insights into localized crime dynamics, they remain fragmented in scope and methodology, and they have yet to capture the broader spatial distribution of property crime across Thailand or its connection to national-scale socioeconomic patterns.

According to data from the Royal Thai Police, property crime has a consistently high reporting rate in Thailand and is not limited to large urban areas such as Bangkok. It also commonly occurs in secondary cities and peripheral areas such as the northern and northeastern regions of the country, which are facing economic changes and a large number of labor movements [27,28]. These conditions have led to household-level instability, which is linked to property crime at the community level [29]. A previous study has found that households facing economic difficulty and family disturbance frequently contribute to reduced social-control mechanisms. As a result of these weakened mechanisms, property crime tends to emerge [29]. Research has also shown that when micro-level vulnerabilities are concentrated in communities, they weaken community cohesiveness and lead to an increase in crime clustering [30].

In addition to economic inequality, spatial context also plays an important role in understanding criminal behavior. Areas with poor infrastructure, such as insufficient lighting, inadequate CCTV coverage, or a lack of community participation in surveillance, experience gaps that facilitate crime [31,32]. Recent research has shown that perceptions of neighborhood disorder, including vandalism, poor lighting, or visible neglect, are negatively associated with collective efficacy and social cohesion. These perceptions increase fear, reduce community trust, and limit residents’ willingness to intervene in crime-related situations [33].

Several empirical studies have demonstrated that specific geographic factors, such as population density, overcrowded housing, and the presence or absence of secure public areas, are strongly correlated with crime incidence at the local level [24,25]. For instance, one study found that in Nakhon Pathom province, areas with higher levels of commercial activity, dense residential patterns, and limited surveillance infrastructure exhibited greater concentrations of property crimes. This spatial analysis underscored how geographic clustering of crime is shaped by the physical and functional layout of urban environments [24]. To comprehensively analyze geographic features of property crime, it is imperative to consider spatial ideas together with social and economic factors to describe criminal behavior in each region and devise contextually suitable crime-prevention strategies.

Section 2 presents a literature review describing major criminological theories and previous studies relevant to spatial crime analysis. Section 3 outlines the data sources and methodology for this study, including spatial statistics and regression models. Section 4 reports our empirical findings and spatial analysis results. Section 5 offers a discussion of key findings and their implications, and it is followed by our conclusions and policy recommendations in Section 6.

2. Literature Review

Theories of crime have evolved over time from individual-focused models to more complex theories that incorporate spatial, structural, and environmental dimensions. In analyzing property crime, which is often determined by both social context and geographic environment, early theories of crime—such as rational choice theory—conceptualized crime as the product of deliberate decisions for which offenders have weighed potential benefits against risks [34]. However, such theories could not help explain why crime is concentrated in certain areas or varies from area to area.

Instead of focusing exclusively on the psychology of offenders, routine activity theory proposed the spatial and temporal conditions under which crime occurs. It held that three primary components—willing offenders, attractive targets, and ineffective policing—must be present for crime to occur [35,36]. This theory shifted the focus of criminology from individuals to places and opportunities. Geographic information systems (GIS) enable researchers to discover and delineate high-risk locations where convergence is more probable, such as densely populated commercial zones with less surveillance [37]. In addition to enhancing the prediction value of routine activity theory, GIS also allows analysts to model offenders’ trajectories and visualize opportunity structures in urban landscapes [38].

Social disorganization theory, which originates in urban sociology, links higher crime rates to structural conditions such as poverty, precarious housing, and the erosion of social control mechanisms in communities [39]. The theory underscores that these environments impede collective efficacy, which is the community’s capacity to regulate behavior and preserve social order [40]. In accordance with this theory, researchers have employed GIS and spatial statistics to quantify indicators of disorganization, including income inequality, derelict housing, and emigration rates, and to examine their spatial correlations with crime. For example, in a study in Portugal, spatial regression models showed that areas with higher youth populations, unemployment rates, and school dropout rates had significantly higher levels of property crime, supporting the theory of social disorder [19]. Similar results were found in South Africa, where spatial autoregressive analysis confirmed that socioeconomic disadvantage tends to be geographically concentrated alongside property crime [41]. This spatial dependency underscores the need for coordinated, regionally sensitive policy interventions, particularly in rapidly urbanizing areas where socioeconomic disparities are deeply entrenched.

Furthermore, crime pattern theory and environmental criminology have been investigated as valuable resources for integrating theoretical frameworks with geographical data [42,43]. Environmental criminology underscores the significance of the physical environment in influencing criminal behavior. For example, Sun et al. found that the spatial distribution of property crime in urban areas is significantly shaped by environmental factors—such as shop density, entertainment venues, house prices, and nighttime light intensity—which interact to influence where crimes are likely to occur within the familiar activity spaces of offenders [20]. According to this concept, criminals typically operate in familiar environments, or cognitive spaces, that are characterized by their regular activities and anchors such as home, work, and leisure [44]. Crime pattern theory has recently been expanded to include temporal dimensions, with research illustrating that offenders adhere to spatial patterns while also displaying time-specific routines that affect the likelihood of crime occurrence [45]. Research has shown that crime patterns occur not only in familiar areas but at predictable times, consistent with daily routines. The integration of temporal and spatial data aligns with the principles of environmental criminology, which highlight place and opportunity as determinants of criminal conduct. Recent spatial analyses of crime have employed a comparable methodological approach, utilizing hotspot mapping to identify concentrations of criminal activity in Thailand [26,46]. This underscores the influence of environmental factors and spatial proximity on the likelihood of crime. By integrating spatial analysis with established criminological frameworks, researchers can develop evidence-based strategies that respond directly to the social and geographic aspects of crime.

Recent years have seen significant efforts to integrate geographical statistics with spatial criminology using models such as geographically weighted regression (GWR) to examine the spatial dependence and non-homogeneity of crime data. Such spatial models differ from traditional linear regression, which assumes the impacts of variables are the same across locations. For example, spatial models can demonstrate potential variations in the relationship between crime and factors such as poverty, education, or unemployment in different regions. Therefore, spatial analysis enables academics to attain a more profound comprehension of crime occurrence across various localities. In one study, Wang et al. compared ordinary least squares (OLS) regression with GWR in an analysis of property crime across 140 neighborhoods in Toronto [47]. Their results demonstrated that while the OLS model provided a general overview, it masked important local variations. In contrast, the GWR model not only improved the overall model fit but also revealed significant spatial non-stationarity, showing that the impact of variables such as material deprivation and residential instability varied considerably across neighborhoods. In another study, Tavares and Costa used GWR to analyze property crime across Portugal at the municipal level, integrating population density, income, and unemployment data. They found that the GWR model is a highly effective tool to explain spatial variation in crime [19].

Building on these insights, this study uses 2021 province-level data to provide a detailed spatial snapshot of property-crime variation in Thailand. Although this design does not capture temporal trends, combining global Poisson regression with GWR uncovers both broad national patterns and localized deviations. Using these methods, we identify crime clusters and region-specific risk factors across all provinces. (Subsequent research would benefit from multi-year analyses to reveal how these crime–socioeconomic relationships evolve.)

Property crime continues to be an issue in Thailand. This concern is closely linked to socioeconomic gaps and urban–rural dynamics on a spatial level. Although national data indicates fluctuations in property-crime rates, conventional studies inadequately address variations in crime between locales or the local determinants that influence such variations. Relying on national data alone may obscure significant local crime trends. This research aims to fill this void by using GIS and spatial statistical techniques to investigate how socioeconomic variables influence the distribution of property crimes across each province of Thailand.

To guide this investigation, the study addresses the following research questions:

• RQ1: What are the spatial patterns and clusters of property crime across Thailand’s provinces?
• RQ2: How do socioeconomic factors—such as income, household debt, population density, and labor-force characteristics—relate to the distribution of property crime?
• RQ3: How do relationships between socioeconomic factors and property crime vary across different geographic regions?

3. Materials and Methods

3.1. Study Area

The study area is Thailand, and the spatial units of analysis are its provinces. Data from the year 2021 were collected from official sources, namely the National Statistical Office of Thailand (NSO) [27]. The year 2021 was the most recent year with available socioeconomic data for all 77 provinces.

Census data in Thailand from 2010 and 2020 indicate a decline in the young population and a shift towards an older population over the ten-year period [48]. More rural areas are experiencing depopulation, especially in the northern and northeastern parts of the country. Conversely, urban areas such as the Bangkok metropolitan region and its neighboring provinces have witnessed substantial population growth. In 2020, NSO estimated the population of Thailand to be 66.2 million people (32.4 million males and 33.8 million females), with the aging trend continuing. The contrast in population density between rural and urban areas is stark. Population density ranged from 19.5 to 5300 inhabitants/km² across the country’s 77 provinces [49]. The lowest population density was in Mae Hong Son province, in northern Thailand, and the highest population density was in Bangkok.

The overall poverty rate in Thailand has been declining, dropping from 17.5% in 2010 to 6.4% in 2020 [50]. However, in 2020 and 2021, during the COVID-19 pandemic, there was a rise in poverty rates, especially affecting regions and informal workers. Furthermore, both public and household debt have increased. In 2010, public debt was 26.3% of the GDP. By 2020, it had risen to 43.5% of the GDP, due to government actions during the pandemic [51]. Household debt has also increased, reaching 89.7% of the GDP in 2020 compared to 59.3% back in 2010, primarily because people borrowed more for housing and other spending [52]. Additionally, there has been a rise in education levels. The basic education consists of six years of primary schooling (Grades 1–6), followed by three years of lower secondary education (Grades 7–9) and another three years of upper secondary education (Grades 10–12). These socioeconomic shifts have had a profound effect on the economic and social landscape in Thailand over the past decade.

3.2. Collected Data

This study used data collected by the Royal Thai Police on property crimes such as robbery, theft, pickpocketing, and burglary [53]. The data was published in the NSO database [27]. For this study, we downloaded property-crime data for 2021 in Excel format and sorted and cleaned it for further analysis.

The socioeconomic indicators analyzed in this study include population density, average household debt, monthly income, immigrant population, labor population, and uneducated labor force. The dataset was acquired from the NSO under Thailand’s Ministry of Digital Economy and Society. The data were collected for each province to explore spatial variations for different areas.

Geographic and boundary data applied in this study were obtained from the Humanitarian Data Exchange [54], an open-access platform administered by the United Nations Office for the Coordination of Humanitarian Affairs, which provides political–administrative boundary datasets for research applications. The GIS boundary data was collected at the province-level administrative scale and was stored in shapefile format for processing.

3.3. Method

3.3.1. Detecting Spatial Patterns and Clusters

Understanding the spatial distribution of property crime is essential for identifying high-risk areas and formulating effective crime prevention strategies. In this study, spatial autocorrelation analysis was used to detect patterns of crime concentration across the study area. The concept of spatial autocorrelation refers to the degree to which crime incidents are clustered, dispersed, or randomly distributed across geographic space.

To assess these spatial patterns, global Moran’s I was applied to determine whether property-crime incidents exhibited a statistically significant clustered or dispersed distribution. This method measures spatial dependence by evaluating the relationship between crime occurrences and their geographic locations. The null hypothesis of Moran’s I states that crime incidents are randomly distributed across the region. The output value ranges from −1 to +1, where a positive value suggests clustering, a negative value indicates dispersion, and a value close to zero implies a random distribution.

Following the global assessment, we used local Moran’s I to pinpoint areas with either a very high or a very low concentration of property crimes. In contrast to the global measure, local Moran’s I assesses spatial dependence at specific locations, facilitating the identification of high-crime and low-crime clusters. The findings classify areas based on five patterns: regions with high crime rates clustered together (HH), isolated regions with high crime rates or low crime rates (HL and LH), clusters of low-crime regions (LL), and areas with no significant patterns identified. These spatial patterns were visually represented to identify high crime areas and analyze how they are spread geographically.

3.3.2. Identifying Socioeconomic Risk Factors

To investigate how socioeconomic factors affect property-crime rates, we specified six independent variables based on prior research and theoretical relevance: population density, average household debt, average monthly income, number of immigrants, total labor population, and uneducated labor force [19,20]. These variables were chosen as risk factors that could influence crime levels, drawing from theories that link income inequality, social unrest, and limited resource availability to increased criminal behavior.

Then, to quantify the associations between these variables and property crime, a generalized linear regression (GLR) analysis was conducted on socioeconomic determinants of property crime. In our study, the Poisson regression model was chosen because it is suitable for modeling the frequency of occurrences in count data while ensuring that predictions remain non-negative and appropriately handle the discrete nature of crime incidents [19,21]. This model assesses the influence of independent variables (such as factors) on the variable (crime rate) while taking spatial relationships into account. The Poisson regression model used in this study is expressed in Equation (1) as follows:

$${\mathrm{y}}_{\mathrm{i}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{x}}_{1\mathrm{i}}+{\mathsf{\beta }}_2{\mathrm{x}}_{2\mathrm{i}}+\dots +{\mathsf{\beta }}_{\mathrm{m}\mathrm{i}}{\mathrm{x}}_{\mathrm{m}\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}},$$

(1)

In this equation, ${\mathrm{y}}_{\mathrm{i}}$ represents the crime count at location i, ${\mathrm{x}}_{\mathrm{m}\mathrm{i}}$ represents the socioeconomic variables, ${\mathsf{\beta }}_{\mathrm{m}\mathrm{i}}$ represents the regression coefficients indicating the strength and direction of relationships, and ${\mathsf{\epsilon }}_{\mathrm{i}}$ represents the residual error term.

The model outcomes of our analysis were understood by considering β coefficients and standard errors while evaluating significance levels at 5%. This stage sheds light on how social and economic factors influence crime rates by highlighting patterns in gaps and their relationships with criminal behavior.

3.3.3. Local Variability in the Socioeconomics–Crime Relationship

Although generalized regression models offer insights into the link between crime and socioeconomics, they assume a stationary relationship, which means that the association remains constant across the study area. However, crime patterns and their root causes can differ substantially from one area to another. Taking this heterogeneity into consideration, we applied GWR in this study.

GWR calculates localized regression equations (as opposed to global models), which allows the intensity and direction of socioeconomics–crime correlations to vary across locations.

This approach enables the strength and direction of the socioeconomics–crime relationships to differ based on the location. The GWR model for analyzing property crime is shown in Equation (2):

$${\mathrm{y}}_{\mathrm{i}}\left(\mathrm{u}\right)={\mathsf{\beta }}_{0\mathrm{i}}\left(\mathrm{u}\right)+{\mathsf{\beta }}_{1\mathrm{i}}\left(\mathrm{u}\right){\mathrm{x}}_{1\mathrm{i}}+{\mathsf{\beta }}_{2\mathrm{i}}\left(\mathrm{u}\right){\mathrm{x}}_{2\mathrm{i}}+\dots +{\mathsf{\beta }}_{\mathrm{m}\mathrm{i}}\left(\mathrm{u}\right){\mathrm{x}}_{\mathrm{m}\mathrm{i}},$$

(2)

In this equation, $\mathrm{u}$ represents the spatial coordinates of each province, and ${\mathsf{\beta }}_{\mathrm{m}\mathrm{i}}\left(\mathrm{u}\right)$ represents the regression coefficients that vary by location.

The effectiveness of GWR lies in its ability to capture variations in crime patterns within areas and offer in-depth insights that can support targeted crime prevention strategies for those locations. To evaluate the model performance level accurately, Akaike’s information criterion (AIC) was employed as a gauge that compares the local models to assess goodness-of-fit metrics. A lower AIC value indicates that the model can accurately accommodate differences, making it better at reflecting the geographic disparities in crime distribution.

4. Results

All spatial visualizations in this study were generated using ArcGIS Pro 2.6.0 and classified using the natural breaks (Jenks) method. The classification into natural groupings enhances interpretability by minimizing within-class variance and maximizing between-class differences, allowing for clearer identification of meaningful spatial patterns across provinces.

4.1. Spatial Distribution

Figure 1 illustrates the spatial distribution of property-crime rates across Thailand in 2021, classified into five categories ranging from 0 to 176.463 crimes per 1000 people. It uses different shades of blue to indicate the various crime rates. Property-crime rates are presented at the province level, highlighting significant diversity across different geographical locations.

As shown in Figure 1, the provinces with the highest property-crime rates are mostly located in the southern and central regions of the country. The capital, Bangkok, exhibits a notably high property-crime rate, with 176.463 crimes per 1000 people as the upper limit. In the south, several coastal provinces, such as Phuket, Songkhla, and Surat Thani, also show notably high crime rates, with values reaching up to 176.463 per 1000 people. In contrast, in the northern and northeastern regions, crime rates are noticeably lower, as indicated by lighter shades.

The descriptive statistics for the variables examined in this study are presented in

Table 1. Descriptive statistical results of the variables.

Variable	Mean	Min	Max	SD
Rate of property crime	66.25	23.98	176.46	30.92
Population density (person/sq km)	243.43	22.39	3517.94	475.67
Average household debt (THB thousand)	202.95	47.60	370.53	72.91
Average monthly income (THB thousand)	24.67	15.50	41.13	5.72
Number of immigrants (thousand persons)	5.30	0.04	109.83	13.97
Total of labor population (thousand persons)	759.92	148.56	7847.36	921.38
Uneducated labor force (thousand persons)	28.86	1.84	230.22	40.51

. The property-crime rate ranges from 23.98 to 176.46, with a mean value of 66.25. Population density varies widely across the study area, with an average of 243.43 persons per square kilometer, ranging from a minimum of 22.39 to a maximum of 3517.94 (standard deviation (SD) = 475.67).

Economic indicators also exhibit notable variation. The average household debt is 202.95 thousand Thai baht (THB), with a minimum value of 47.60 and a maximum of 370.53 (SD = 72.91). The average monthly income ranges from THB 15.50 to 41.13 thousand, with a mean of 24.67 (SD = 5.72). Demographic factors further highlight differences in labor-force composition. The number of non-citizens applying for work permits averages 5.30 thousand people, with values ranging from 0.04 to 109.83 (SD = 13.97). The total labor population fluctuates between 148.56 and 7847.36 thousand, with a mean of 759.92, signifying disparities in labor-market size.

These findings indicate significant variations in population distribution, economic conditions, and workforce characteristics, all of which contribute to disparities in property-crime rates across the study area. Figure 2 presents the spatial distribution of the raw values for each explanatory variable.

4.2. Spatial Autocorrelation Analysis

The results of the global Moran’s I index of property crime also indicate that crime is spatially clustered, with a Moran’s I value of 0.5763 and z-score of 6.5491, while the p-value of 0.0000 confirms statistical significance at a high confidence level. These results imply that property crime demonstrates significant spatial dependence. The spatial clustering analysis of property crime is presented in Figure 3.

The results indicate that high–high clusters are concentrated in most provinces in the southern region. In contrast, the low–low areas are mainly found in several provinces in the northeastern and northern regions. Chiang Mai is a high–low outlier, with crime rates significantly higher than those in the surrounding low-crime provinces. A few provinces in the southern region exhibit the low–high outlier patterns. These provinces include Nakhon Si Thammarat, Chumphon, Ranong, Pattani, and Narathiwat.

These spatial patterns suggest that the clustering of property crime could be influenced by underlying socioeconomic factors. Therefore, investigating the associations between property crime and these variables is necessary to understand the extent to which factors such as income levels, unemployment rates, and population density contribute to crime distribution.

4.3. Spatial Correlation of Property Crime and Socioeconomic Conditions

4.3.1. Global Correlation Between Property Crime and Socioeconomic Conditions

Table 2. Statistical results of the GLR analysis on socioeconomic determinants of property crime.

Socioeconomic Variables		GLR Results
	β	ε	z-Value	p-Value	VIF
Intercept	6.178560	0.005396	1145.027869	<0.001	--------
Population density	−0.251516	0.009397	−26.765787	<0.001	4.606837
Average household debt	−0.085108	0.005650	−15.064481	<0.001	1.313479
Average monthly income	0.271795	0.006318	43.020719	<0.001	1.987156
Number of immigrants	−0.346915	0.012157	−28.536658	<0.001	5.780759
Labor population	0.778364	0.011265	69.095886	<0.001	5.984708
Uneducated labor force	0.092712	0.005490	16.888614	<0.001	1.786104

outlines the Poisson regression outcomes based on z-score standardized (normalized) socioeconomic variables in relation to property crime across Thailand. With 80.99% of the deviance explained, the model demonstrates a strong ability to capture the variability in property-crime rates based on the chosen variables.

Among the selected predictors, there is a positive association with average monthly income, labor population, uneducated labor force, and rates of property crime. However, population density, the number of immigrants, and household debt are negatively correlated with property crime. All variables show statistical significance (p-value < 0.05), providing strong evidence of their relationships with property-crime rates.

To further assess the reliability of these relationships, z-values were reported for each variable. The consistently high absolute z-values across predictors strengthen the evidence that these socioeconomic factors are significantly associated with variations in property crime across regions. As the variables were standardized, the coefficients represent the change in the log count of property crime per one standard deviation increase in each predictor. Given an AICc value of 9039.0, it is suggested that the model has a good fit. Meanwhile, the joint Wald statistic of 35,857.5 also confirms its statistical significance, reinforcing the reliability of these associations. The assessment of multicollinearity using variance inflation factor (VIF) values suggests that most variables remain within an acceptable range. Despite the VIF values for the number of immigrants (5.78) and labor population (5.98) approaching moderate levels, they do not reach a threshold that would significantly impact the model’s reliability.

4.3.2. Regional Correlation Between Property Crime and Socioeconomic Conditions

As the relationships between socioeconomic factors and property crime may vary across different locations, this study used GWR to assess spatial heterogeneity to ensure that regional variations in crime determinants were properly considered. Based on the confirmed Poisson regression results, the socioeconomic variations listed in Table 2 were further analyzed. The results are presented in

Table 3. Comparative results from GLR (global-level) and GWR (local-level) analyses.

Model	Deviance Explained Model [%]	AICc
GLR	80.99	9039.0
GWR	92.40	3463.3

.

Table 3 shows the comparison results between the global and local models. The higher deviance value of 92.40% in the local model—compared to 80.99% in the global model—indicates a better fit and improved quality of the local statistical model. This 11.41% gain in explanatory power indicates the local model’s ability to better capture spatial differences in how socioeconomic factors influence the occurrence of property crime. The AICc value also shows that the local model, at 3463.3, is significantly lower than the global model at 9039.0, which implies that the model provides a better fit.

As shown in Figure 4, the GWR results display the local percentage deviance across the study area, with values ranging from 0.779 to 0.966. The results are depicted on the map with colors from dark purple to dark green, representing variations in model fit. Higher values indicate areas where the model performs better in explaining property-crime variations.

The areas with strong statistical fit are primarily located in northern and central Thailand, indicating a strong relationship between socioeconomic factors and property crime, as seen in provinces such as Chiang Mai and Bangkok. These areas, with deviance close to 0.966, suggest that local economic and social conditions effectively explain crime variations. In contrast, areas with low deviance (below 0.779, shown in dark purple) are primarily found in several southern provinces and in parts of the central region. These lower values imply that the selected socioeconomic predictors may not fully capture the complexity of crime determinants in these areas. One possible explanation is the influence of unobserved contextual or structural factors specific to these regions, such as political instability, religious tensions, or cross-border security issues, particularly in the southern provinces. Additionally, certain cultural or law-enforcement dynamics that differ regionally might also affect crime patterns and reduce the explanatory power of the model. To further understand these variations, the coefficients of each socioeconomic variable across the study area were explored, providing deeper insights into their spatial relationships with property crime in Figure 5.

The maps in Figure 5 illustrate the local coefficient estimates derived from the GWR model for each socioeconomic variable. Negative coefficients are represented in shades of blue, while positive coefficients are displayed in shades of orange to dark red. This visualization approach enhances interpretability by clearly distinguishing both the direction and strength of associations across provinces. All maps were constructed using manual classification with thresholds centered around zero, allowing for consistent comparison between negative and positive effects.

For population density, provinces such as Ubon Ratchathani, Yasothon, Amnat Charoen, and Si Sa Ket in the northeastern region are displayed in dark red shades, while provinces such as Tak and Loei are shaded in dark blue, reflecting localized variation in coefficient values. This pattern suggests that the relationship between population density and property crime is not uniform across space. While some areas experience increased crime with greater population density, others may benefit from stronger social cohesion or infrastructure that mitigates such effects.

For average monthly income, Chiang Mai, Chiang Rai, and Nan in the northern region show high positive coefficients (dark red), reflecting that increased income levels in these provinces may attract property crime. In contrast, Ubon Ratchathani and nearby areas show strong negative coefficients (dark blue), implying an inverse relationship in those contexts.

Regarding the number of immigrants, high positive coefficients are observed in Mukdahan, Amnat Charoen, and Ubon Ratchathani, while Nong Khai and Loei present high negative associations. These regional differences point to the localized influence of immigration on property-crime trends.

In the southern region, particularly in Surat Thani, Phuket, and Nakhon Si Thammarat, strong positive coefficients for average household debt are evident (shown in dark red), suggesting that rising debt levels may contribute to property crime. However, some northeastern provinces, such as Mukdahan and Amnat Charoen, also show high positive coefficients, reflecting a broader pattern beyond the south. Despite this, the overall model fit in this region remains lower, possibly reflecting the influence of region-specific factors or unmeasured characteristics not included in the current dataset.

The labor population coefficient shows very strong positive relationships in northern provinces such as Chiang Rai, Chiang Mai, and Nan, highlighting that areas with larger working populations may experience higher property-crime rates. These provinces are prominently shaded in dark red.

Lastly, for uneducated labor force, slight positive associations are found in Nakhon Sawan, Kanchanaburi, and Lopburi, all in the central region, with coefficients shaded in light orange tones. In contrast, northeastern provinces such as Udon Thani and Loei show strong negative coefficients, indicating that a higher proportion of uneducated labor may not be linked to increased property crime in those areas.

5. Discussion

This study investigated the spatial distribution of property crime and analyzed the spatial association between socioeconomic conditions across Thailand. These spatial geographical analyses broadened our understanding of property crime based on spatial patterns and variables that are specific to a given region. The results show both local clustering and statistically significant global clustering within the study area. This work represents an initial effort to employ a combination of spatial analysis methodologies and socioeconomic data to map spatial clusters of property crime at the province level throughout Thailand.

Despite the widespread acceptance of spatial techniques such as Moran’s I, Poisson regression, and GWR in spatial crime research, the use of these techniques in Thailand is still limited. However, the main contribution of this work is not in methodological innovation but in the contextual application of these models to an empirically underexplored setting. To date, very few national-scale studies have examined the spatial structure of property crime in Thailand using quantitative spatial models. Therefore, while the analytical framework follows conventional spatial procedures, its application provides valuable insights for both research and policy in a setting where such methods are rarely implemented. Moreover, the use of province-level data reflects the most granular level of data currently available consistently across the country. We acknowledge that this scale may obscure intra-provincial variation; this limitation is discussed in detail later in the manuscript, and future research directions are proposed accordingly.

This study situates conventional spatial techniques within a novel geographic and empirical context, drawing attention to methodological decisions that affect local interpretation and model performance. In particular, the GWR model is known to be sensitive to specific analytical choices such as kernel type, bandwidth selection method, and variable specification. These modeling decisions can influence the estimation of local coefficients and thus affect the interpretability of spatial patterns. In this study, the bandwidth parameter was optimized using the golden-section search algorithm, which iteratively identifies the bandwidth that minimizes the corrected AICc. While this approach ensures a statistically efficient fit, we recognize that alternative settings—such as different kernel functions or fixed bandwidths—could yield different spatial patterns. Further exploration of model sensitivity would help strengthen the interpretability and reproducibility of local spatial results in future GWR applications.

Although property-crime rates show regional disparities—with higher crime rates in the central and southern regions, particularly in Bangkok and the coastal provinces—this study further analyzes the data and identifies spatial clustering and regional crime patterns. We identified a strong positive geographical global autocorrelation, indicating that property-crime occurrence was not randomly distributed within the research area. The global Moran’s I value demonstrated the existence of spatial dependency in property crime in Thailand; the value ranges from −1 to +1 (fully dispersed to completely clustered, respectively). The global Moran’s I index discovered in this investigation was 0.5763.

Furthermore, the local Moran’s I evaluation was employed to identify spatial patterns of property crime within the research boundaries. The local Moran’s I analysis also identified high–high (HH) crime clusters in almost every province in southern Thailand, potentially influenced by urbanization and economic conditions. However, Chumphon, Ranong, Nakhon Si Thammarat, Pattani, and Narathiwat were classified as low–high (LH) outliers in this region. In the northeastern region, it is notable that many provinces reveal low–low clustering, indicating consistently low crime rates. In contrast, in northern Thailand, Chiang Mai is identified as a high–low outlier; while the crime rate is high in this province, neighboring provinces have significantly lower crime rates.

In addition to the observed spatial clusters, a large number of provinces were classified as non-significant in the local Moran’s I results (as shown in Figure 3), indicating no statistically significant spatial autocorrelation in property-crime rates. These non-clustered areas do not exhibit discernible spatial patterns, which could be attributed to several theoretical and contextual reasons. First, property crime in these provinces may be shaped by highly localized dynamics that differ from those in neighboring areas, leading to an absence of spatial dependence. For instance, provinces such as Nakhon Ratchasima, Saraburi, and Nakhon Sawan—located along major transport routes or economic corridors—act as transitional zones that combine both urban and rural characteristics. Their diverse economic bases, patterns of migration, and social structures might disrupt consistent clustering effects. Additionally, varying levels of law-enforcement efficiency, public reporting behavior, and data quality may also contribute to the lack of spatial association in these areas.

To understand how different socioeconomic factors influence crime distribution, the study also examined correlations between diverse factors and property crime across Thailand. While our spatial and statistical analysis provides valuable insights into structural determinants of property crime, it may not fully capture the individual and community-level perceptions of insecurity, inequality, and vulnerability that often drive criminal behavior. These subjective dimensions can influence social cohesion and collective efficacy and thus deserve further exploration. This variation is reflected in both the spatial mapping, which graphically depicts regional disparities in crime incidents, and the statistical analysis. The GLR findings revealed that property-crime incidence is positively linked with average monthly income and is slightly positively associated with an uneducated labor force and the labor population.

A previous study found that areas with higher income levels and wealth concentrations often experience higher rates of property crime [55]. Our study also aligned with the concept that wealthier areas may be attractive targets for property crime due to the presence of more valuable assets, increasing the motivation for criminal activity [56]. Previous research has also shown a significant association among unemployment, urbanization, and crime rates. The study revealed that the role of education and training is crucial in mitigating crime incidence, as higher educational attainment and skill development programs contribute to reducing unemployment and improving labor-market conditions [57,58]. Our research findings are consistent with Yang’s study of 50 U.S. states and the District of Columbia, which found that lower levels of education are associated with higher crime rates [58]. These findings emphasize the interconnected relationship among income, education, and the labor force, suggesting that higher income levels may attract property crime, while lower education levels and an unskilled labor force contribute to increased crime rates. To reduce the incidence of crime, targeted policies that support economic opportunities, education, and skill development are needed.

Although conventional criminological theories often associate higher population density with increased crime, several studies suggest that densely populated areas can have lower crime rates due to increased social surveillance and law-enforcement presence. Our results align with the perspective indicating that high-density areas may benefit from improved crime-prevention mechanisms. A study in Nigeria also supports this finding, revealing that greater population density is associated with improved social cohesion, better lighting in public spaces, and increased law-enforcement presence [59]. Rapid urbanization in Thailand has resulted in a high concentration of people in locations such as Bangkok, Chiang Mai, and Pattaya, increasing the need for sophisticated surveillance systems and more robust law enforcement to protect the public. Prior research has shown that some urbanized areas in Thailand have improved security by expanding their CCTV networks [60]. Several additional advancements in crime prevention have also been implemented to enhance urban safety, including AI-driven crime detection and real-time monitoring systems [61].

Based on routine activity theory, our GWR analysis suggests that population density is positively associated with crime in provinces such as Ubon Ratchathani, Yasothon, Amnat Charoen, and Si Sa Ket. This finding is consistent with the theory that high population density increases the availability of suitable targets and enhances opportunities for crime. Particularly in places where social control systems are inadequate or insufficient to discourage motivated offenders, the link between population density and property crime can be explained by the lack of effective guardianship, both official and informal. In areas such as Ubon Ratchathani, Yasothon, Amnat Charoen, and Si Sa Ket, elevated population density may lead to a greater presence of potential targets due to higher levels of human interaction and routine activities [62,63]. However, these locations are more susceptible to criminal behavior when social control systems—such as law-enforcement presence, community surveillance, and robust social networks—are absent or inadequate. The neighborhood becomes more vulnerable to criminal activity [64].

While these provinces are not major metropolitan centers, they often exhibit population clustering in specific districts, particularly in urban municipal areas. Such micro-urbanized zones may lack sufficient public safety infrastructure, and the limited distribution of law enforcement or surveillance coverage can weaken crime deterrence. This semi-urban pattern, coupled with demographic concentration, creates conditions in which motivated offenders may act with reduced risk of apprehension, thereby increasing property-crime incidence.

This finding appears to contrast with the results of the global Poisson regression model, which show a negative overall association between population density and property crime. This discrepancy highlights the differing assumptions and analytical scopes of the two models. While the Poisson regression provides a single, global estimate across all provinces—assuming spatial stationarity—the GWR model accommodates local variation and identifies spatially non-uniform effects. In other words, population density may have a deterrent effect in some provinces while increasing crime risk in others, depending on the presence or absence of social control mechanisms, urban infrastructure, or enforcement capacity.

In northern regions, particularly Chiang Rai and Mae Hong Son, the GWR analysis shows a strong positive correlation between average monthly income and property crime. According to routine activity theory, higher income levels in these areas attract motivated offenders due to the availability of valuable assets and increased economic activities that make these locations prime targets for property-related crimes. Previous reports revealed that these regions have experienced economic growth and increasing tourism, contributing to a greater availability of suitable targets [65,66].

In the southern part of the country, regions such as Nakhon Si Thammarat and Phuket face higher rates of property crimes, and Surat Thani province shows similar patterns, according to analysis using the GWR technique. A link between property crime and household debt is observed in these regions, possibly as a result of economic pressures faced by local residents. Rapid economic growth fueled by tourism and business activities has led to an increase in household debt in tourist spots such as Phuket and Surat Thani provinces. In regions with less cohesive or more transient communities, such as tourist spots, individuals dealing with money problems might turn to theft to cope. Moreover, heavily touristed areas draw temporary residents and opportunistic criminals, establishing conditions described by routine activity theory: crime occurs when vulnerable targets and motivated offenders come together in settings with minimal supervision [67,68]. Research has shown patterns whereby economic expansion linked to tourism in developed coastal regions leads to higher crime as a result of income inequalities and opportunities for exploitation [69,70].

In central regions such as Nakhon Sawan, Kanchanaburi, and Lopburi provinces, there is a notable link between property crimes and the presence of uneducated laborers. These provinces frequently have a high concentration of migrant and low-skilled laborers. Kanchanaburi, situated along the Myanmar border, encounters obstacles as a hub for cross-border migration. The arrival of workers brings advantages but can also stretch local resources and make social integration more challenging. Together with a lack of law-enforcement resources and disjointed community institutions, this can create an environment where crime is more likely to occur. Prior research into car theft in the border regions of Kanchanaburi has highlighted barriers to enforcement, such as national policy guidelines, poor coordination across different agencies, limited community understanding, and legal gaps that impede effective crime prevention measures. It can be difficult to manage border criminal activity due to challenges in sharing information among agencies and conducting surveillance in complex geographical areas [71]. These difficulties are often compounded by social instability and the significant number of workers in the region. These factors point to vulnerabilities within the system that align with the principles of social disorganization theory, where weakened institutions and a lack of effectiveness can heighten the risk of property crime.

While this study draws primarily from routine activity theory and social disorganization theory, future research could consider integrating and comparing alternative theoretical frameworks such as rational choice theory. Routine activity theory effectively explains crime risks in tourism-driven areas such as Phuket and Surat Thani, where motivated offenders and suitable targets converge in the absence of capable guardianship. In contrast, social disorganization theory is more suitable for interpreting crime dynamics in peripheral provinces where weak institutions and social fragmentation prevail. However, neither theory alone can fully explain complex spatial crime patterns in transitional or hybrid urban–rural zones. Rational choice theory, which assumes deliberate offender decision-making based on cost-benefit analysis, could further enrich interpretations—particularly in areas where property crimes reflect calculated responses to perceived opportunity structures. However, rational choice theory may under-emphasize structural and community-level drivers, while social disorganization theory may overlook individual decision-making processes. In addition, spatially contextual factors not directly captured in the datasets—such as religious diversity in southern provinces where both Buddhist and Muslim communities co-exist—may influence levels of social cohesion, trust, and collective efficacy, which are linked to crime incidence. Likewise, the distribution and accessibility of transportation networks, including interprovincial highways and ferry routes, could facilitate the mobility of offenders and access to potential targets, further shaping spatial patterns of property crime. A comparative theoretical approach would offer a more holistic understanding of crime geographies and guide the selection of policy responses tailored to local criminogenic conditions.

Table 4. Socioeconomic and spatial conditions contributing to property crime in Thailand.

Region	Economic Development Trend	Social Structure Impact	Crime Risk Mechanism
Northern and Eastern	Increased income and tourism	More valuable targets, less informal control	Target-rich, low guardianship
Western and Central	Labor migration from border	Weak integration, strained enforcement	Social fragmentation
Southern	Rapid tourism-driven expansion	Temporary residents, widening inequality	Offender–target convergence
Northeastern	Moderate density, low infrastructure	Weak guardianship, poor planning	Target accumulation

consolidates and summarizes the key socioeconomic and spatial characteristics of each region discussed and their associated risk pathways for property crime.

The diverse spatial patterns observed in this research show the location-specific forces that impact property crime in parts of Thailand. Depending on the local context, the impact of socioeconomic factors, including income levels, household debt, labor mix, and population density, varies greatly. The use of GWR has underscored the significance of analyzing data at a localized level, demonstrating that crime trends cannot be comprehensively grasped through crime statistics alone. In addition to structural and spatial factors, future research should also incorporate community-level insights and residents’ perceptions regarding safety and social conditions. Understanding how people interpret and respond to their neighborhood environments—including perceptions of risk, trust in law enforcement, and sense of social cohesion—can significantly inform targeted interventions. This bottom-up knowledge is particularly valuable for designing community-driven crime prevention programs that align with local realities. Moreover, incorporating subjective well-being indicators and perceived inequalities into spatial crime analysis may help to bridge the gap between objective data and lived experiences, contributing to more holistic approaches that address both social and health disparities within vulnerable communities. These results highlight the necessity of region-specific crime prevention plans that take into consideration the various institutional, social, and economic circumstances in different areas. Understanding how routine activities, economic stress, and social disorder interact spatially offers useful information for policymakers attempting to reduce crime through more focused, context-sensitive initiatives.

It is also important to recognize that the data used in this study were collected during the COVID-19 pandemic in 2021. The pandemic may have influenced both socioeconomic conditions and property crime patterns through multiple channels. For instance, lockdown-induced mobility restrictions may have reduced certain types of crime because potential criminals and their potential victims were not moving about the community, while rising unemployment and economic stress could have exacerbated criminal motivations in other contexts.

Additionally, shifts in law-enforcement priorities, reduced public interaction, and changes in reporting behavior during the pandemic may have affected the observed crime statistics. While our analysis did not explicitly account for these pandemic-related dynamics, they represent an important contextual factor that may have shaped the spatial patterns found in this study. Future research could expand on this dimension by incorporating time-series data across multiple years to isolate and evaluate the effects of pandemic-related disruptions.

Despite the strengths of GLR and GWR in identifying spatial associations and capturing local variability, it is important to acknowledge the underlying assumptions of these models, which may not always hold true in real-world settings. In the context of property crime in Thailand, assumptions such as the linearity of relationships and the statistical consistency across space could require careful consideration when applied to provinces with diverse social, economic, and institutional dynamics. For example, the relationship between income and crime or between population density and crime may be non-linear in certain provinces, particularly in areas with unique socioeconomic structures, such as tourist-dependent regions like Phuket or Surat Thani. In such locations, seasonal population fluctuations, structural inequality, and informal economic activities could distort linear trends and introduce local complexities that traditional regression models may struggle to capture. Furthermore, as GWR results can be sensitive to choices of kernel type, bandwidth, and variable specification, future work should include formal model comparisons (e.g., comparing AIC values across alternative local and global models) and sensitivity analyses (e.g., testing different variable sets or data transformations) to ensure the robustness and reliability of the spatial findings.

Additionally, variations in crime reporting practices and differences in law-enforcement capacity across provinces can further affect the consistency and reliability of the dependent variable. In rural or underserved communities, property crimes may be systematically underreported due to limited access to law-enforcement services or a lack of trust in formal institutions [72,73]. For instance, a 2023 public opinion survey in Thailand reported that nearly 70% of respondents had little or no confidence in police integrity, particularly in rural areas [73]. These factors highlight the need for cautious interpretation of model results and suggest that future research should consider more flexible modeling approaches and incorporate qualitative data to better account for micro-level dynamics influencing spatial crime patterns.

The study has several limitations despite its contributions. First, due to data constraints, we focused on a core set of socioeconomic variables; we did not include measures of urban infrastructure quality (e.g., street lighting, CCTV coverage), detailed education levels, crime-reporting practices, or law-enforcement presence. Our examination concentrated on population density, household debt, monthly income, number of immigrants, labor population, and uneducated labor force. Moreover, our province-level spatial and statistical models cannot capture micro-level factors such as community cohesion, informal social controls, and individual behavior patterns. Additionally, this research utilized cross-sectional data from 2021, which may not fully represent temporal dynamics or long-term trends in property-crime patterns.

Another notable constraint of this research is its spatial resolution, which was limited to the provincial level and could hide significant local variations within provinces. This level of resolution has the potential to conceal high-crime or vulnerable locations within specific districts or sub-districts. While this resolution enabled consistent nationwide comparison, it limits the ability to capture finer-grained spatial dynamics. Future studies could address this limitation by integrating higher-resolution data sources, such as sub-district statistics, mobility data, social media content, or crowdsourced platforms. These data types could provide more spatial granularity and potentially enable real-time or near-real-time crime monitoring. Furthermore, using multi-scale spatial frameworks could help uncover contextual dynamics and complex crime patterns that remain hidden at the provincial level. Even though this study’s analytical approach is based on well-established spatial methods, its value lies in applying these techniques to an underexplored national context. We recognize, however, that future research could benefit from incorporating emerging spatial methodologies to further enhance analytical depth and innovation.

Future research could also incorporate qualitative methods—such as semi-structured interviews with community leaders, focus groups with residents, and standardized surveys on collective efficacy—to complement quantitative analyses and provide deeper insights into the social processes driving property-crime clusters. Such approaches would allow researchers to capture neighborhood-level narratives and lived experiences that are often invisible in spatial datasets, thereby supporting more community-responsive and evidence-informed policy design. Finally, future research should consider comparing spatial modeling approaches—such as GWR, multiscale GWR, and spatial Bayesian models—to evaluate their performance and underlying assumptions. Such comparisons would contribute to methodological innovation and clarify the contexts in which each model performs best, particularly for detecting non-stationary spatial relationships in criminological data. In addition, future studies should also consider integrating additional variables, including urban infrastructure, education levels, crime reporting practices, and law-enforcement capacity, to gain a more comprehensive understanding of crime determinants.

To improve spatial precision, future work could adopt more refined units of analysis, such as districts or sub-districts, enabling more localized policy interventions. Moreover, longitudinal studies using panel data or time-series data would provide valuable insights into the evolution of property crime and its socioeconomic correlates over time, allowing for the identification of long-term patterns, delayed effects, or causal relationships. To strengthen causal inference, researchers could also employ quasi-experimental approaches, including instrumental variable analysis, natural experiments, or regression discontinuity designs. For instance, if a new public safety policy is piloted in selected provinces, it may serve as a natural experiment that enables pre- and post-intervention comparisons between treated and untreated areas. Integrating resident-generated insights such as perceptions of social safety, inequality, and well-being into spatial crime models would further enhance the policy relevance and equity focus of future research.

6. Conclusions

This study explored how property crime is distributed across space and how it relates to varying socioeconomic conditions throughout Thailand. As a country with distinct disparities between regions, Thailand presents a valuable case for understanding how local context shapes crime patterns. Our results underscored both global and local geographical patterns of property crime, emphasizing the spatial variety in the influence of socioeconomic variables on criminal conduct across different geographical regions. The findings of spatial autocorrelation indicate that crime is not randomly distributed but rather clusters in specific areas, especially in the southern part of the country, where high–high clusters were identified, and in the north and northeast regions, where low–low clusters were identified. This research offers key insights on how spatial technologies can efficiently pinpoint areas with heightened risk, providing actionable data for targeted crime reduction initiatives. By applying both GLR and GWR, the analysis revealed that the impact of factors such as household debt, population density, and labor characteristics varied spatially. These findings align with the core principles of routine activity theory and social disorganization theory, which emphasize contextual variations in crime opportunities and social control. These regional differences suggest that the drivers of property crime are shaped by complex interactions among local economic structures, migration patterns, and institutional capacities, which must be carefully considered in the design of effective crime prevention strategies. Ultimately, this study underscores the importance of customizing urban development and crime response techniques to the distinct social and economic environment of each region. Future research could benefit from incorporating participatory approaches and subjective indicators to better align spatial crime prevention strategies with community needs and perceptions.

Beyond its academic implications, this study is particularly relevant to local policymakers, urban planners, and public safety officials, as it offers concrete insights for enhancing urban safety and promoting sustainable development. In tourism-driven hotspots such as Phuket and Surat Thani—where household debt and seasonal population shifts are strongly correlated with property crime—local governments could integrate spatial crime maps into tourism safety plans, invest in street lighting, and expand CCTV networks around high-risk transit zones. Conversely, in rural and peripheral districts where underreporting may obscure actual crime patterns, mobile reporting units and community-based crime perception surveys could help surface hidden vulnerabilities and inform more inclusive safety strategies. In central provinces with large migrant labor populations, strategies such as improving worker protections and fostering social integration could address the underlying causes of crime, while in northeastern provinces with limited infrastructure, policies that strengthen social capital and service accessibility may be more effective. These examples illustrate how place-specific evidence can inform targeted, community-responsive interventions that promote both immediate safety and longer-term resilience, in alignment with Thailand’s sustainable development objectives.