Introduction

The relationship between poverty and crime has always been an important topic of interest in criminology and economics (Allen & Stone, 1999; Stretesky et al., 2004; Ross, 2020). According to Merton’s strain theory of deviance, which argues that crime occurs when there aren’t enough legitimate opportunities for people to achieve society’s normal success goals, there is a "strain" between the goals and the means to achieve them, leading people to turn to crime to attain success. Economists also believe that there is a connection between poverty and crime, and they take crime as a obstructive factor of economic development (Mehlum et al., 2005).

Based on the theory above, a question arises: if governments implement poverty alleviation projects to improve the income level of a region, will it effectively reduce the crime rate in that area? If so, what are the primary mechanisms? Are these effects effective for all types of crimes? Although some studies have discussed the relationship between poverty and crime (Meloni, 2014; Chambru, 2020), most of them have not adequately addressed the endogeneity issue between poverty and crime (Song et al., 2020; Heilmann et al., 2021). Moreover, there is a lack of research specifically focusing on China in this context. As the world’s largest developing country, China serves as an ideal laboratory for studying the aforementioned issues.

Reducing or eliminating poverty is a shared global challenge today, especially considering that the majority of impoverished individuals are concentrated in developing countries (Masron & Subramaniam, 2018). As the largest developing country in the world, China has achieved remarkable accomplishments in poverty alleviation in recent years (Song, 2012). According to United Nations data, China has contributed 70% to global poverty reduction since the 1990s. By international poverty standards(individuals living on less than $2.15 per day), China has reduced its rural impoverished population from 770 to 30 million over the past 40 years, significantly improving people’s living standards.

Based on these reasons above, this study utilizes the poverty alleviation projects implemented by China in 592 poverty counties nationwide by the end of 2015 as the context to examine whether and why a significant reduction in poverty rates would lead to a notable decrease in crime rates (details of the project will be given in the later section). The remaining sections of this paper are organized as follows. First, we provide a literature review on past work related to our research. Second, We propose questions this study will answer. Third, a description of the data, research design and key variables will be given. Fourth, we will analyze and discuss the main results. Fifth, we conclude and provide policy implications, also acknowledge the limitations and talk about the directions that future researcher could proceed.

Literature Review

This paper contributes to the existing literature in the following ways. Firstly, we supplemented the relevant literature on poverty and poverty alleviation. Scholars have not only focused on the origins (Duong & Flaherty, 2023), causes (Singh & Jha, 2023), and measurement of poverty (Ogwang, 2022; Dang and Verme, 2023), but also on the social impacts it brings (Jalal et al., 2022). In addition, numerous studies have aimed to identify factors that can reduce the number of people living in poverty and improve the living standards of those in underdeveloped regions, such as infrastructure development (Zhang et al., 2023), digital transformation (Ding et al., 2023), and climate change (Chalise & Naranpanawa, 2023). Specifically, as the world’s largest developing country, many studies have focused on China’s poverty issues and poverty alleviation efforts (Han & Si, 2023; Su et al., 2023). Some relevant literature that closely related to this study has also assessed the social impacts of poverty alleviation policies (Khosla & Jena, 2022; Lou et al., 2023; Ma et al., 2022; Silva et al., 2023). But previous studies have not evaluated and analyzed the impacts of large-scale poverty alleviation projects in China, and we will fill this gap and offers insights into the social outcomes of such project.

Secondly, this study also contributes to the existing literature on crime. Previous research has discussed the causes (Ajzenman et al., 2023; Baryshnikova et al., 2022; Charlton et al., 2022; Eren et al., 2022), impacts (Mitra & Shajahan, 2022; Muller & Schwarz, 2023; Pak & Gannon, 2023), and governance of crime (Arora, 2023; Castro & Tirso, 2023), as well as factors that can reduce crime rates, such as institutional factors like taxation (Lenhart, 2021), judiciary (Guerra & Nilssen, 2023), and property rights (Lawson, 2023), as well as natural and social factors(Allen, 2021; Lee et al., 2022; Tealde, 2022; Drugowick & Pereda, 2023). While previous literature has also examined the relationship between crime rates and poverty (Mehlum et al., 2005; Ross, 2020), there are two key limitations. First, it has not adequately addressed the issue of endogeneity between the two variables. Second, it has not specifically discussed the impact of poverty reduction on crime rates within the context of China, the largest developing country, and we will fill these gaps in this study.

Hypotheses and Refinement of Questions

Our core question is whether alleviating poverty will reduce the crime rate. However, there is a "chicken-egg" problem between poverty and crime. Without any external policy shocks or administrative control, the poor are more likely to commit crimes. Conversely, a region with a high crime rate will inevitably hinder the normal functioning of the economy and the improvement of people’s living standards, which form a negative cycle. However, if the government takes various policies to improve the living standards and increase the income of people in a region, this cycle can move in a positive direction. When people’s income levels rise, it reduces criminal behavior, and the reduction in criminal behavior further promotes economic development. Therefore, in the context of this study, we tend to assume that government poverty alleviation projects, while increasing income, will further reduce the crime rate.

In addition, we also want to answer another question: Is the decrease in crime rate symmetrical across different types of crimes? In other words, does an increase in income lead to a decrease in various types of crimes? Different types of crimes correspond to different criminal motives (Toby, 1962). For example, Merton’s strain theory suggests that societal resources are limited, while human desires are unlimited. When individuals in society all strive for the same goals (such as high income or employment), and the resources provided by society are insufficient, those who are unable to acquire resources through legitimate way will feel frustrated, leading to the use of illegal means (crime) to acquire wealth. Additionally, criminologists also have proposed the anarchic crime theory, which suggests that crime occurs in situations of social disorder and low law enforcement efficiency. During economic downturns, individuals in socially disadvantaged areas may be driven by extreme helplessness and choose crime as a means to escape reality. Another perspective is the cultural factor emphasized by the neoclassical school of criminology, where the relative deprivation theory emphasizes emotional and internal reward-based criminal motives (Webber, 2007). Based on these theories, we want to investigate which types of crimes will decrease after the efforts to alleviate poverty.

Details of the Project

China’s poverty alleviation efforts have a long history, but the earlier poverty alleviation projects were different from the current one. In the past, poverty alleviation projects were implemented nationwide without regional divisions (Qu, 2017), which posed challenges in studying the economic and social impacts of poverty reduction. However, unlike the past, the current poverty alleviation project is implemented in targeted counties.

The reason of why some parts of China are classed as non-impoverished and others impoverished is that substantial financial resources are allocated annually for transfer payments to impoverished areas. However, China has approximately 1,300 county-level administrative regions, and there is significant disparity in wealth among different counties (Zhang, 2021). For example, the living standards in the eastern coastal regions are significantly better than those in the western regions. Hence, the government needs to establish certain criteria to identify low-income areas with insufficient local government fiscal revenue. Financial subsidies are then provided to these counties to enhance their living standards.

After two years of design and planning, the corresponding poverty alleviation work began in the relating counties at the end of 2015. The time-line of this poverty alleviation project is presented in Table 1.

Table 1 The time-line of the poverty alleviation project
Full size table

The scale and intensity of this poverty alleviation work are significant. According to official statistics, as of 2020, a total of 255,000 village-based work teams were dispatched nationwide, and a cumulative total of 2.9 million government and state-owned enterprise personnel were deployed to work in the selected counties. According to the requirements of official documents, these personnel were precisely matched with the corresponding impoverished counties, villages, and households, and carried out poverty alleviation work based on local and household characteristics.

Previous poverty reduction projects were mainly aimed at providing transfer payments to county-level governments which have limited fiscal revenue and unable to afford basic public infrastructure. However, the focus of this poverty reduction project is to transform the county’s industrial structure, enhance the people’s ability to create wealth, and eradicate poverty mindset. In other words, it emphasizes “teaching people to fish rather than just giving them fish”. We consulted the government’s official website and summarized the tasks of this poverty reduction project in Table 2.

Table 2 The details of the projects
Full size table

Regarding the selection of impoverished counties, the official documents explicitly state that it was based on the per capita gross domestic product (GDP), per capita general budgetary revenue, and per capita net income of rural residents. Finally, a total of 592 impoverished counties were selected as targets based on these indicators (Appendix 1 provide more details about the classification of the county).

Data

Crime Rate Data

In this article, we have used two measures to assess crime rates: the number of criminal cases and the number of offenders. To examine whether there are differences in the changes of crime rates across different types of crimes, we further divided criminal cases into four categories: property crimes, robbery and theft crimes, violent crimes, and intentional injury crimes. The motivation behind this classification is to explore the impact of poverty reduction programs on different crime motivations. Property crimes include robbery and theft crimes, but considering that robbery and theft may be more common in economically depressed areas with low income, we have separated them as a distinct category for regression analysis.

Since the poverty alleviation project implemented at the county level, official county-level crime rate data in China is not publicly available at present. Therefore, in this article, county-level crime rate data is obtained by aggregating national court verdicts published by the Supreme People’s Court. The method of obtaining the crime rate data through court verdicts is reliable for the subsequent analysis in this article due to the following reasons.

Firstly, the Supreme People’s Court of China has explicitly stipulated that, except for cases involving state secrets, personal privacy, crimes committed by minors, case end with mediation, and other specific types, all other cases should be made public and published on their official websites. Moreover, the verdicts published on these websites follow a standardized format, which facilitates the automated retrieval of a large amount of data using programs. Specifically, for each case, we obtained data such as location, time, and type of crime using automated programs. Previous studies have also utilized this measure of crime rate and have confirmed its reliability (Khan & Li, 2020).

One important point to note is that the public disclosure of court verdicts began in 2014. Therefore, a challenge for this article is the limited availability of pre-implementation data for the poverty alleviation project, which prevents us from conducting parallel trend analysis at the annual level. However, we can still analyze the crime rate data on a quarterly basis, which will be used for testing in subsequent sections.

County-level Data

We also need to include relevant data at the county level as control variables, including fiscal revenue, fiscal expenditure, population, and per capita GDP. The reason for using these variables as control variables is their association with the classification of poverty and non-poverty counties. If we do not control for these variables in the regression, the experimental group cannot be compared with the control group, and the regression results may be biased.

We determined the experimental group and control group based on the official list of impoverished counties. For other county-level data such as per capita gross domestic product (GDP), per capita general budgetary revenue, and per capita income of rural residents, we obtained them from multiple sources such as the China County Statistical Yearbook, EPS database(a commercial database that contains various micro and macro data), CEIC database(a commercial database contains economic data about most counties). These data have been widely used in research on China-related issues (Clay et al., 2021; Groh et al., 2022; Li et al., 2016).

It is important to note that county-level statistical data may have some missing values, so it was not possible to obtain complete data for all 592 impoverished counties. After excluding samples with missing data, the final analysis included 592 impoverished counties as the experimental group and 1,355 non-impoverished counties as the control group. In addition, to discuss the possible mechanisms in this article, we also utilized data on changes in people’s income levels from the China Family Panel Studies (CFPS) survey, which provides information after the implementation of poverty alleviation policies (Buchmueller et al., 2021; Crossley et al., 2023; Cutler et al., 2022).

Other Relevant Data

In addition to the aforementioned variables, we also included other variables at the county level in the covariate balance test of the difference-in-differences model, such as industrial added value, employment, fixed asset investment, and student population. Furthermore, in the robustness test, we utilized indicators such as population mobility and income. Table 3 presents the names, definitions, and descriptions of these variables. In the subsequent sections of the research design, we will provide detailed explanations for the use of each variable.

Table 3 Description of variables
Full size table

Research Design

In the context of this article, the division between the experimental group and the control group is relatively clear; therefore, we will employ a difference-in-differences model for regression analysis. The difference-in-differences model has been widely used to analyze the impact of government policies. Specifically, we employ the following regression equation to study the causal relationship between poverty and crime rates.

$$Crim{e}_{it}=\alpha +\beta Trea{t}_{i}\times Pos{t}_{t}+{X}_{it}+{\sigma }_{i}+{\theta }_{t}+{\varepsilon }_{it}$$
(1)

In Eq. (1), \(Crim{e}_{it}\) represents the growth rate of crime of county \(i\) in year \(t\), \(Pos{t}_{t}\) is a dummy variable indicating whether the time is after 2015, \(Trea{t}_{i}\) takes 1 if the county belongs to experimental group. \({X}_{it}\) represents the control variables that we need, \({\sigma }_{i}\) and \({\theta }_{t}\) represent county and year fixed effects, respectively, while \({\varepsilon }_{it}\) represents the disturbance term.

In an ideal scenario, if the experimental group and the control group were randomly selected, we could directly compare the results of the two groups to determine the causal relationship between them. However, in the context of this article, the selection of the experimental group and the control group is not entirely random; the experimental group is inherently more impoverished. Therefore, in order to accurately assess the impact of poverty reduction policies, it is necessary to follow the common practice in the literature. Prior to regression analysis, a covariate balance test is conducted on the experimental and control groups, and relevant variables are included in the regression equation as control variables. We use the following equation to conduct the covariate balance test between the experimental and control groups and set the final regression equation based on the test results.

$$Covariat{e}_{it}=\alpha +\beta Trea{t}_{i}+{X}_{it}+{\upsilon }_{it}$$
(2)

We are interested in the coefficient \(\beta\) which indicates whether the control group and experimental group have significant difference on a battery of pre-determined variables. Then we will set the final regression equation based on the results of balanced test. Table 4 presents the results of these tests. In Panel A, we first display four variables based on official grouping criteria. It can be observed that there are significant differences between the experimental and control groups in all four variables. Furthermore, in Panel B, we select additional variables for testing and examine the conditional differences between the two groups after controlling for the aforementioned four variables. We find that the experimental and control groups still exhibit noticeable differences in other variables, but achieve conditional balance after controlling for the selection criteria. Moreover, in Panel C, we examine the differences in the main dependent variables of this study between the two groups and find that they are also conditionally balanced.

Table 4 Balance test
Full size table

Based on the results of the balance tests mentioned above, it is evident that there are differences between the experimental group and the control group. However, when we control for the government’s classification criteria in the regression analysis, the experimental group and the control group no longer exhibit significant differences in other variables. Therefore, it is necessary to control for these classification criteria variables in the regression equation. Finally, the regression equation in this article is formulated as follows:

$$Crim{e}_{it}=\alpha +\beta Trea{t}_{i}\times Pos{t}_{t}+fisca{l}_{it}+populatio{n}_{it}+expenditur{e}_{it}+GD{P}_{it}+{\sigma }_{i}+{\theta }_{t}+{\varepsilon }_{it}$$
(3)

Analysis and Findings

We present the mean, standard deviation, and sample size of the main variables mentioned in the article in Table 5, and compare the experimental group and the control group. According to this table, there are two points that need to be noted. Firstly, in terms of the total number of criminals and cases, the control group has higher absolute values, which could be attributed to the higher population density in the control group. However, when considering the growth rates of crime rates, the experimental group shows a significant increase compared to the control group. Therefore, we will use both absolute values and growth rates of these variables as explanatory variables in subsequent analyses. Secondly, based on the per capita GDP, per capita fiscal revenue, and expenditure data, it is evident that the control group has better economic conditions compared to the experimental group. This indicates that the selection criteria for the experimental and control groups are indeed related to the level of poverty.

Table 5 Descriptive statistics of main variables
Full size table

According to Table 5, there are two points that need to be noted. Firstly, in terms of the total number of criminals and cases, the control group has higher absolute values, which could be attributed to the higher population density in the control group. However, when considering the growth rates of crime rates, the experimental group shows a significant increase compared to the control group. Therefore, we will use both absolute values and growth rates of these variables as explanatory variables in subsequent analyses.

Secondly, based on the per capita GDP, per capita fiscal revenue, and expenditure data, it is evident that the control group has better economic conditions compared to the experimental group. This indicates that the selection criteria for the experimental and control groups are indeed related to the level of poverty.

Main Results

Table 6 presents the baseline regression results of this study. Column 1 displays the regression results using annual data, showing a significant reduction in the annual growth rate of criminal activities after the implementation of the poverty alleviation project. Column 2 further controls for the linear time trend of the control variables (Goodman-Bacon, 2021), and finds that the coefficient remains significant. Finally, in column 3, regression is conducted using monthly data, and it still reveals a significant reduction in the growth rate of criminal activities due to the poverty alleviation project.

Table 6 Baseline results and dynamic effects
Full size table

In addition, we also examined the dynamic effects of the poverty alleviation project on the growth of crime rates according to the literature (Sun & Abraham, 2021). Column 4 of Table 5 presents these results. It is found that the effect is not significant in the first year. This suggests that poverty reduction is a continuous process and may not have an immediate impact on crime rates initially. However, the effects gradually become evident after 2017. Finally, due to the limitations of annual data, we utilized quarterly data to examine the parallel trends between the experimental and control groups. Specifically, we tested the variation of the difference in crime growth rates between the two groups over time using the following equation:

Figure 1 presents parallel trends based on quarterly data (Based on the time-line in Table 1, we take the last quarter of 2015 as the base period), revealing that prior to the initiation of the poverty alleviation project, the regression coefficient fluctuates around zero, indicating no economic or statistical significant difference. However, after the project begins, the regression coefficient gradually becomes negative and attains statistical significance. This indicates that the difference in crime rates between the experimental group and the control group is indeed a result of the poverty alleviation project.

Fig. 1
figure 1

Parallel trends

Full size image

Changing the Measurement of Crime Rate

In the baseline results, we used the annual growth rate of the number of crimes as the dependent variable. In this section, we will first replace the dependent variable with the annual growth rate of the number of crime cases. Table 7 presents the relevant regression results, and it is found that using the annual growth rate of crime cases as the dependent variable does not significantly change the regression results. Then we take the number of offender and crime cases as dependent variable. Table 8 presents the relevant results, it can be observed that using the number of the offender and case as the dependent variable does not change the conclusions of the baseline regression.

Table 7 Take the growth rate of crime cases as dependent variable
Full size table
Table 8 Take the number of offender and crime cases as dependent variable
Full size table

Changing the Control Group

In the previous analysis, we treated all non-poverty counties as the control group. However, there might be spillover effects of poverty alleviation efforts (Durlauf, 1994; Thome et al., 2013), meaning that the poverty reduction effects could simultaneously affect the crime rate in neighboring counties. If such effects exist, the estimation results may be biased (Berg et al., 2021). Therefore, we attempt to select control counties that are farther away from the treatment group(We use adjacency of two counties as the criterion. If two counties are adjacent, we exclude them from the control group. If they are not adjacent, we take them as part of the control group).

As a comparison, we also present the results using neighboring counties as the control group and non-neighboring counties as the control group simultaneously. In Table 9, columns (1) and (2) represent the regression results using neighboring counties as the control group, and it can be observed that both the statistical significance and economic significance decrease to some extent. However, the regression results using non-neighboring counties as the control group remain significant. This suggests that the poverty alleviation project indeed has a certain degree of spillover effect.

Table 9 Changing the control group
Full size table

Considering Criminal Mobility

Since many criminal activities do not occur locally (Andresen, 2011; Freedman & Owens, 2016), it is possible that individuals from poverty counties migrate to non-poverty counties and engage in criminal behavior (Wilström & Torstensson, 1999). If such migration exists, the estimates presented earlier may be biased and could potentially underestimate the crime rate growth in poverty counties. To address this issue, we ranked cities based on their population mobility and removed the 15 and 30 most active cities in terms of population mobility from the sample respectively (Huang & Deng, 2006).

Table 10 presents the corresponding regression results, and it is found that even after considering the cross-regional transfer of criminal behavior due to population mobility, the causal relationship between poverty and crime still exists.

Table 10 Control the mobility
Full size table

Placebo test

Under the background of this study, the selected counties are treated as the experimental group, while the remaining counties are the control group. To test the robustness of the baseline results, we bootstrapping 1000 times and estimate them separately. By comparing the estimated results from bootstrapping with the original estimates, we can assess whether the original estimates are affected by omitted variables (Chetty et al., 2009). Following the approach of Chetty et al. (2009) and La Ferrara et al. (2012), this study conducts bootstrapping procedure while controlling for the unchanged proportion of treatment and control observations in the original sample. Figure 2 and 3 presents the results of the placebo test, where we find that the estimated coefficients from the random assignment show a normal distribution centered around zero, while the distribution of the true estimated coefficient differs significantly from bootstrapping results.

Fig. 2
figure 2

Placebo test with growth rate of offenders as the dependent variable

Full size image
Fig. 3
figure 3

Placebo test with growth rate of cases as the dependent variable

Full size image

Income Effect

The goal of poverty alleviation is to increase income. In this section, we examine whether the poverty alleviation policy has significantly increased the income in poverty counties. Furthermore, we investigate whether income plays a decisive role in the relationship between poverty and crime rates. To answer that, we utilize data from the China Family Panel Studies (CFPS), which provides information on household and individual income levels (Flaaen et al., 2019). We integrate this data with the previous dataset (CFPS data does not cover all counties. Therefore, we can only match samples from 34 poverty counties and 51 non-poverty counties). Table 11 presents the regression results.

Table 11 Mechanism
Full size table

In column 1, it is found that the implementation of the poverty alleviation project significantly increases the annual income level of respondents in poverty counties. The coefficient indicates that, compared to the control group, the experimental group experiences an average increase of approximately 10% in per capita annual income. Furthermore, we add per capita annual income as a variable to the regression equation and regress it against the growth of the number of offenders. Column 2 and 3 shows the regression results, it is observed that both the statistical and economic significance of the coefficient decrease. This suggests that the increase in per capita income after the implementation of the poverty alleviation project is indeed an important factor contributing to the reduction in crime rates. Finally, in columns 4 and 5, we use the growth rate of crime cases as the dependent variable and find similar results as mentioned above.

Subdivision of Crime Types

To further verify that the relationship between poverty and crime rates is a result of income level improvement, we further divide all crime cases into different types according to the literature(Wickramasekera et al., 2015;), including property crimes, robbery and theft, violent crimes, intentional injuries. We conduct separate regressions for each crime type. Table 12 presents the regression results. It can be observed that when the crime type is property crimes, the poverty alleviation project significantly reduces the growth rate of crime rates. However, when the crime type is non-property crimes such as violence and personal crimes, the suppression effect of the poverty alleviation project on the growth rate of crime rates is not significant. This once again highlights that the increase in income level in the experimental group is the main mechanism for the reduction in crime rates. Other non-property crimes did not decrease significantly due to the poverty alleviation project.

Table 12 Subdivision of Crime Types
Full size table

Conclusions

According to Merton’s strain theory of deviance which argues that crime occurs when there aren’t enough legitimate opportunities for people to achieve the normal success goals of a society. In such a situation there is a ‘strain’ between the goals and the means to achieve those goals, and some people turn to crime in order to achieve success. Therefore, this article, to some extent, provides empirical evidence for the aforementioned theories using China’s poverty reduction projects as a background (poverty can lead to an increase in crime rates). Specifically, regarding the differences in this theory under the contexts of China and the United States (where the theory was originally applied to by Merton), we attempt to provide an analysis from the perspective of intergenerational social mobility. Relevant research suggests that due to differences in political systems, in the United States, people are more likely to have opportunities to achieve upward social mobility through their own efforts (maybe that’s why we usually call US as “a land of opportunity”) (Chetty et al., 2014; Corak, 2013). Therefore, the motivation for crime based on Merton’s strain theory may be lower compared to China (Please note that Merton’s strain theory defines poverty in terms of relative poverty, which is different from the absolute poverty used in this study, therefore the above discussion is merely a qualitative analysis).

The analysis in this article can be summarized in the following aspects. Firstly, the paper provides a detailed overview of the background of the poverty alleviation project. Due to the selection of the experimental and control groups in this project was not random, which may present significant challenges to the identification in this study. To address this problem, we reviewed relevant official documents of the project and find out the criteria for selecting the experimental group. Specifically, we find that the government use per capita GDP, per capita general budgetary revenue, and per capita net income of rural residents at the county as criteria for selecting the experimental group. By controlling for these variables, we can better estimate the causal relationship.

Secondly, we estimated the causal relationship between poverty and crime rates and found a significant decrease in crime rates among the experimental group after the implementation of the poverty alleviation project. Additionally, we demonstrated the dynamic effects of this process, revealing that the decline in crime rates gradually manifested approximately one year after the project’s implementation. Lastly, we conducted a series of robustness tests and found that the aforementioned results remained economically and statistically significant when alternative dependent variables, data frequencies, and control groups were used in the robustness checks.

Thirdly, we conducted further analyses on the above results, including mechanism analysis and extended analysis. In the mechanism analysis, we found that the main pathway through which crime rates decreased was the increase in per capita income levels. In the extended analysis, we further categorized different types of crimes and found that the reduction in crime rates was primarily concentrated in property-related crimes, which shared the common characteristic of being motivated by monetary gain. However, we did not observe a significant decrease in violent crimes within the experimental group. By comparing the aforementioned results, it is evident that the national poverty reduction policies, by increasing per capita annual income, have subsequently led to a reduction in crime rates within the experimental group.

According to the analysis in this article, we propose the following policy recommendations. Firstly, based on China’s poverty reduction experience, other developing and developed countries should formulate corresponding policies in education, healthcare, social security, industrial transformation, and public services to improve people’s living standards while effectively reducing crime rates. It is particularly important to focus on increasing per capita income. Secondly, while the government is formulating policies to reduce crime rates, it should also consider different criminal motivations. The results of this article indicate that reducing the poverty rate alone cannot effectively reduce all types of crime rates. For example, the rates of violent crimes and intentional harm do not decrease due to a decrease in the poverty rate. Therefore, the government should match policies with different types of crimes and effectively reduce crime rates across various categories. For violent crimes and intentional harm, the government should establish an effective system for collecting and analyzing crime data, regularly monitor crime trends and patterns, conduct relevant research, and explore the causes and influencing factors of intentional harm and violent crimes. This will provide a scientific basis for policy formulation and implementation.

Thirdly, poverty reduction can have extensive spillover effects. Therefore, when government fiscal expenditures are limited, it is advisable to conduct widespread pilot projects nationwide, as China has done. This approach ensures that poverty reduction efforts are maximized within a certain fiscal expenditure range, yielding the best social effects. At the same time, pilot projects should emphasize flexibility and sustainability, taking into account local conditions and implementing specific policy measures accordingly. It is important to establish a policy evaluation mechanism to assess the effectiveness and outcomes of pilot projects in a timely manner and to summarize experiences and lessons. Based on the evaluation results, policies should be adjusted and improved promptly to ensure the effective implementation and long-term sustainability of poverty reduction policies.

This article also has several limitations. Firstly, this article focuses only on China, a developing country. It is unclear whether the research findings are applicable to other developing and developed countries, and whether the effects of poverty reduction policies are the same in different countries. Secondly, we do not know why there was no reduction in violent crimes and intentional harm, and we are unsure about which policies can effectively reduce these types of crimes. Thirdly, our data is at the county level and lacks individual-level data, which makes it difficult to study income distribution and inequality, as well as to understand the relationship between income, happiness, and crime. Fourthly, regarding the measurement of poverty, we only utilized the government’s list and did not employ other measurement methods (such as infant mortality rate). However, different poverty measurement approaches may impact the empirical results.

Therefore, we look forward to future research primarily focusing on the following aspects. First, if individual-level data can be obtained, further research could be conducted on the relationship between poverty and crime rates, comparing and contrasting the findings with those of this article, while also discussing the relationship between income distribution inequality and crime. Second, other researchers should evaluate the relationship between poverty reduction policies and crime rates in other developing and developed countries, comparing them with China, and discussing the differences while providing reasonable explanations. Third, based on the findings of this article, government policies that increase income primarily reduce property crimes, but have limited impact on violent crimes and intentional harm. Future research should also examine the effects of other government policies on crime rates, discuss whether these policies increase or decrease crime rates, and differentiate between different types of crimes. This will facilitate matching different policies with different types of crimes, enabling governments to implement tailored policies in different regions.