RETRACTED: Can Digital Financial Inclusion Help Reduce Urban Crime? Evidence from Chinese Criminal Judgment on Theft Cases

Abstract

The rapid development of digital finance has changed all aspects of human life and has also had a deep impact on the social governance system. This paper constructs an unbalanced panel of data of the theft crime rates for 289 cities in China during 2014–2019 based on the theft criminal judgments published on China’s Judicial Documents website and explores the impact of digital financial inclusion on urban theft crime. It shows that there is a significantly negative correlation between digital financial inclusion and the urban theft crime rate, indicating that the development of digital financial inclusion can effectively reduce urban theft crime, which is also confirmed by instrumental variable analysis based on the spherical distance between cities and Hangzhou, and that digital financial inclusion mainly reduces theft crime committed by more serious and highly educated individuals. In addition, mechanism analysis shows that digital financial inclusion can reduce the expected benefits of theft by enhancing payment convenience and raise the opportunity cost by promoting employment. Therefore, in the Internet era, it is essential for China to continuously improve social governance tools that adapt to the development of new technologies to achieve high-quality urban development.

1. Introduction

Since the reform and opening up in 1978, China’s economy has achieved a remarkable growth miracle, which mainly benefits from institutional transformation and a series of institutional innovations [1]. Meanwhile, social conflicts have also become increasingly prominent, which has led to a persistently high crime rate in China [2,3]. According to the Statistical Bulletin of the National Bureau of Statistics in 2021, China’s crime rate has rapidly jumped to a higher level since 2000, forming the “fifth crime peak” in China. The number of cases prosecuted by the national prosecution authorities rose from 3.788 cases per 10,000 people in 2000 to 7.734 cases per 10,000 people in 2020, and the number of prosecutors rose from 5.593 per 10,000 people in 2000 to 11.139 per 10,000 people in 2020 (as shown in Figure 1). The continuous rise in China’s crime rate has attracted widespread concern in society [4,5]. In this context, the 20th National Congress of the Communist Party of China (CPC) proposed that the governments should improve the public security system, strengthen the overall prevention and control of social security, and severely punish all kinds of illegal and criminal activities that extremely disgust the public according to the law. Social stability is an important prerequisite for healthy economic development. Criminal activities not only threaten the daily personal and property safety of residents, which causes social instability, but also waste a large amount of scarce public resources, which results in the misallocation of resources and a loss of social welfare [6]. What we said above has caused serious negative effects on people’s welfare and regional economic development. Therefore, how to effectively curb the occurrence of criminal activities has become a focus for academics and policymakers [7,8]. What are the factors that are causing the rising crime rate in China? Moreover, what kind of measures does the Chinese government need to adopt to effectively reduce criminal activities? Addressing these issues is very meaningful for building a socialist and harmonious society and enhancing people’s sense of happiness and gain!

What caused the rapid rise in China’s crime rate? Many scholars have conducted theoretical and empirical studies from the perspectives of population mobility, income disparity, urbanization, and social security [9,10,11,12]. It is undeniable that the rising crime rate in China is accompanied by a rapid growth in digital financial inclusion [13]. In recent years, relying on information technologies, such as artificial intelligence, blockchain, cloud computing, and big data, digital financial inclusion, through methods such as Alipay and WeChat Pay, has been in full swing in China. As an emerging industry, digital financial inclusion is a product of the deep integration of digital technology and financial services [14]. Moreover, it has attracted widespread concerns from the Chinese governments and the public since it was proposed in 2016, becoming one of the most active fields at present [15]. Compared with traditional finance, digital financial inclusion can get rid of the shackles of physical outlets because it has a strong geographical penetration and low-cost advantages, and it can also lower the standard for people to use financial products because it can provide efficient payment methods and convenient financial services for customers. In other words, digital financial inclusion has deeply penetrated and integrated into all aspects of people’s daily lives and social operations in the forms of sharing, security, accuracy, efficiency, and convenience, thus becoming a notable economic and social governance mechanism [16]. In 2022, the 20th National Congress of the CPC proposed that the governments should accelerate the development of the digital economy and promote the integration of the digital economy and the real economy. As a representative of the digital economy, digital financial inclusion has a huge development space and application prospects. Although the existing literature has examined the impact of digital financial inclusion on economic behaviors, such as consumption, entrepreneurship, and financing [17,18,19], few scholars have studied the impact of digital financial inclusion on social public safety, especially the urban crime rate [20]. In this context, can digital financial inclusion with its characteristics of a low entry barrier and low transaction cost reduce the crime rate? Does its impact significantly differ among different groups? What are the internal mechanisms? This paper intends to answer these questions through theoretical and empirical analyses systematically, so as to provide references for the Chinese government to improve the governance capacity of social public safety and promote the modernization of the social governance system.

This paper focuses on the mechanisms and effects of digital financial inclusion on the urban crime rate from a theft crime governance perspective. Specifically, first, an altruism-based crime time allocation model is constructed to theoretically explain the impact mechanism of digital financial inclusion on criminal activities. Second, all the first-instance criminal adjudication documents of theft crimes for 289 cities in China during 2014–2019 are collected and the data on urban theft crimes are systematically organized to analyze the effects and mechanisms of digital financial inclusion on urban theft crimes in China using a panel fixed effects model and instrumental variable regression. Compared with the existing studies, the main contributions of this paper can be summarized as follows: Firstly, unlike the existing studies on the various economic effects of digital financial inclusion and digital economy development, this paper explores the social impacts of digital financial inclusion from the perspective of crime governance, providing new ideas and evidence to understand the relationship between the digital economy and high-quality development. Secondly, based on the text data of official criminal adjudication documents of theft in Chinese cities, which are extracted by regular expressions, this paper systematically examines the demographic characteristics of theft crimes and judgments of theft activities, helping to comprehend the current criminal characteristics of theft activities in China. Thirdly, this paper demonstrates the impacts and mechanisms of digital financial inclusion on urban theft crime based on the cost–benefit analytical framework of crime, which makes up for the shortcomings of existing research on criminal economics and criminology, and provides new insights for understanding the social governance effects of Internet development and constructing a governance mechanism to effectively prevent theft crime activities. Fourthly, this paper uses judicial big data based on text recognition to quantitatively study theft crimes, which expands the research data of criminology and lays a solid foundation for subsequent relevant studies.

The rest of this paper is organized as follows: Part 2 reviews the literature relating to the nexus between digital financial inclusion and crime. Part 3 adopts an altruism-based crime time allocation model to explain the impact mechanism of digital financial inclusion on criminal activities. Part 4 presents the data, econometric model, and variables used for empirical analysis. Part 5 discusses the empirical results. Part 6 concludes the study.

2. Literature Review

When it comes to the nexus between digital financial inclusion and the crime rate, previous research has studied the factors influencing criminal activities and the economic impacts of digital financial inclusion. Therefore, in this section, we can review the previous literature about the two aspects.

2.1. Factors Influencing Criminal Activities

In fields including sociology, law, and psychology, crime has long been a popular research topic since it is a widespread social occurrence [21,22]. Becker [23] pioneered the study of crime in economics and created a theoretical model of crime economics with cost–benefit analysis at its foundation, arguing that criminal activity is the result of rational choices made by individuals who weigh the costs and benefits of criminal activities. Subsequently, in light of this reasoning, a large number of studies have conducted extensive research on the factors influencing criminal activities and obtained relatively insightful results [24,25,26]. First, many scholars have proposed unequal income distribution may be responsible for the ongoing rise in crime rates, arguing that the growing income gap implies a decline in the opportunity cost of crime and an increase in the potential benefits, which favors more criminal activities [27,28]. For example, based on the unbalanced panel data for 39 countries from 1965–1995, Fajnzylber et al. [29] studied the effect of income distribution on crimes, using generalized moments analysis. The results showed that income inequality was a crucial factor for the high crime rate in regions. By using the state-level panel data for Brazil from 1981–1995, Sachsida et al. [30] found that the widening income gap was the Granger cause of criminal activities. Since then, a large number of empirical studies have supported the idea that income inequality resulted in an increase in crime rates [31,32]. Second, a few studies investigated how demographic factors, including regional demographics, labor market development, and the unemployment rate, affects crime rate. Theoretically, the higher the population quality and education in a region, the higher the opportunity cost of crime, and the greater the social abhorrence and rejection of criminal activities, thus leading to fewer criminal activities [33,34]. For example, based on micro-survey data over the period of 1960–1980 in the U.S., Lochner and Moretti [35] analyzed the effects of education on criminal activities. They confirmed that schooling significantly reduced crime rates. However, rising unemployment reduces the opportunity cost of crime and increases the propensity for crime among those with low levels of education [6]. Sachsida et al. [30] exhibited that the unemployment rate had a significant positive effect on the crime rate, but improvements in labor market conditions significantly reduced property crime. In addition, the proportion of young people, especially the proportion of young males, has a significant positive effect on crime rates [9]. Third, a few studies analyzed how urbanization affects crime rates, arguing that as cities grow, crime opportunities increase, raising the expected returns on crime and encouraging more people to commit crimes in metropolitan areas [11,30]. For example, based on the panel data from 1993–1999 for 46 provinces in Spain, Buonanno and Montolio [36] analyzed the economic and demographic factors of criminal activities. They detected that urbanization has a significantly positive effect on crime rates. Using the panel data for 73 Russian regions during the period 1995–2007, Ivaschenko et al. [37] showed a significantly positive nexus between urbanization and criminal activity. Moreover, several other studies also supported that urbanization enhanced crime rates [38,39]. Fourth, other scholars have explored the impact of social welfare spending on crime rates and argued that social relief can help alleviate the sense of social isolation among low-income groups and enhances their willingness to comply with moral norms, thus helping to reduce crime activities [40,41]. For example, based on the panel data for 81 cities in the U.S. from 1930–1940, Fishback et al. [42] analyzed the impact of social welfare expenditures on crime during the Great Depression. They confirmed that social welfare expenditures significantly inhibited urban crime activities. Meloni [43] empirically examined the nexus between social welfare expenditures and criminal activities for 23 Argentine provinces over the period 2002–2005, using panel data analysis. The results revealed that social welfare expenditures reduced crime rates and had the most significant inhibitory effect on property crime. The above findings have also been confirmed in many subsequent studies [44,45].

2.2. The Economic Impacts of Digital Financial Inclusion

Using modern technologies, such as electronic payments, information technology, big data, and cloud computing, digital financial inclusion improves users’ service, promotes financial inclusion, and significantly lowers transaction costs and entry barriers through a series of functions, such as financing, investment, and payment [20,46]. In terms of the economic impact of digital financial inclusion, scholars currently focus on the intension of digital financial inclusion [47], the poverty reduction effect of digital financial inclusion [48], the contribution of digital financial inclusion to the real economy [49], and the risks and regulation of digital financial inclusion [50]. Firstly, at the macro level, some studies have demonstrated that digital financial inclusion enhances bank operations [51], reduces poverty and income inequality [48], narrows the urban–rural income gap [52], and boosts social welfare [53,54]. This is due to the fact that digital financial inclusion develops together with e-commerce, lowering the threshold for financial services and resulting in scale and long-tail effects. As a result, owing to the decline in transaction costs, more people gain access to financial services, thus improving financial asset allocation efficiency [55,56]. Moreover, except for efficiency, the growth in digital financial inclusion can also contribute to enhancing social justice [57]. For example, Das and Chatterjee [58] showed that, due to the severe and persistent lack of financial services in economically disadvantaged regions, the growth in digital financial inclusion makes convenient financial services available to the poor and low-income groups who are not covered by formal finance. Therefore, in regions with uneven economic development, digital financial inclusion reduces income disparity and promotes inclusive economic growth. Secondly, at the micro level, the existing literature performs empirical analysis, using matched data from household or enterprise surveys and the regional level of financial inclusion development. The results reveal that the growth in digital financial inclusion has a significantly positive impact on residential consumption [17,59], household financial demands [60], private lending [19,61], entrepreneurial opportunities [18,62], and corporate technological innovation [63,64]. For example, Lai et al. [17] discovered that digital financial inclusion promoted consumption by alleviating residents’ liquidity constraints and enhancing payment convenience, and the effects were greater and more significant for households in rural, low- and middle-income, or less-developed areas. However, Luo and Li [59] contended that the main factor why digital financial inclusion boosted residential consumption was the relaxation of payment convenience, rather than liquidity constraints. Using the cross-sectional data from the Chinese Household Finance Survey, Wang et al. [60] demonstrated that digital financial inclusion met household financial diversification demands by stimulating consumption and alleviating liquidity constraints and effectively avoided household over-indebtedness. This conclusion was also supported by Li et al. [19] and Yue et al. [61]. Based on the digital financial inclusion index and Chinese household sample survey data, Yang et al. [62] showed that the growth in digital financial inclusion met female demand for financial inclusion and reduced the financial constraints on female entrepreneurship, which in turn stimulates female entrepreneurial activity and promotes more entrepreneurial opportunities. Using the data from Chinese listed companies covering the period 2011–2020, Yao and Yang [64] exhibited that digital financial inclusion effectively absorbed financial market resources and transformed them into effective supply and deeply optimized traditional financial institutions and businesses to improve quality and effectiveness, thus creating favorable conditions for the technological innovation in enterprises.

In conclusion, there are many findings from the enormous studies that the academic community has conducted regarding the factors influencing crime rates and the economic effects of digital financial inclusion. However, the following issues remain with the current research: Firstly, crime is a widespread social issue. Despite the fact that the academic community has performed an extensive number of studies on the issue, most of them focused on developed countries, and there is still a lack of literature on developing countries and transition economies. Secondly, in relation to the causes of the ongoing rise in the crime rate, despite the academic community’s useful discussions and empirical research from the perspectives of urbanization level, income disparity, population migration, and social welfare spending, it has not taken into account the impacts and mechanisms of digital financial inclusion on crime. Thirdly, although the existing literature has empirically studied criminal activities based on macro-statistical data at the national or regional level, there is very little literature on the usage of micro-survey data to explore the features of crime in a country or region, especially studies on the use of judicial big data to reveal the general patterns of local crime. Therefore, the theft crime statistics in this study for 289 Chinese cities during 2014–2019, which match pertinent variables at the prefecture level, are extracted from the judicial big data published on China’s Judicial Documents website. Then, using a panel two-way fixed effects model and instrumental variable regression, this paper analyzes the mechanisms and effects of digital financial inclusion on urban crime in China from the perspective of theft crime governance.

3. Theoretical Analysis and Research Hypothesis

Referring to Becker [23] and Freeman [65], this paper analyzed the nexus between digital financial inclusion and theft crime by constructing a crime time allocation model embedded with digital financial inclusion and family altruism. Specifically, consider an economy where the individual’s utility function is concave, i.e., U′ > 0 and U″ < 0, and the total time of an individual is unitized to one. Assume that t_i is the time that individual i engages in theft and p is the probability that an individual is arrested for theft. If the criminal individual i commits theft but is not apprehended, he can obtain revenue I₁ with the probability 1 − p. Therefore, the revenue I₁ can be expressed as follows:

$$I_1=A+W(fin)+G(t_i)$$

(1)

where A represents the initial wealth of criminal individual i, and W is the family’s return rate, which is the function of digital financial inclusion fin. For simplicity, we assume that digital financial inclusion is continuous, so the higher the level of digital financial inclusion fin, the higher the family’s return rate W correspondingly. G(t_i) is the net revenue of criminal individual i from theft, assuming that G′ > 0, G″ < 0.

Individual i will be punished and sent to jail if he is apprehended for theft. For simplicity, assume that he can no longer enjoy family revenue after being arrested, indicating that W is equal to zero. Thus, his payoff can be formulated as follows:

$$I_2=A+G(t_i)-F(t_i)-\alpha (fin)f(t_i)$$

(2)

where F(t_i) represents the loss of an individual by his arrest, which depends on the crime time. Assume that F′ > 0, F″ > 0, meaning that the severity of the penalty may increase with crime time and the seriousness of the offense. α(fin) denotes the altruism of the offender to his family members. Although the economic theory assumes that people are self-interested, existing studies show that people are exceedingly altruistic within their family, which can promote the long-run benefits of family members. The degree of altruism α is an increasing function of digital financial inclusion fin, indicating that the higher the level of digital financial inclusion is, the more altruistic the criminal is in the family and the more he cares about the losses caused by his criminal behavior to family members. f(t_i) represents the cost shared by family members for theft crime, such as reputational loss, mental anguish, and various economic losses caused by reputation damage. Similarly, suppose that f′ > 0, f″ > 0. α(fin)f(t_i) denotes the extra cost that the offender should pay, meaning that he must consider the losses caused by his theft crime activities to family members. Therefore, this potentially raises the cost of theft and indicates that the family joint and several liability systems are constructed.

In summary, the expected payoff of individual i who committed theft is as follows:

$$EU=(1-p)U(I_1)+pU(I_2)$$

(3)

Substituting I₁ and I₂ into Equation (3), we can obtain the equation below:

$$EU=(1-p)U\left[A+W(fin)+G(t_i)\right]+pU\left[A+G(t_i)-F(t_i)-\alpha (fin)f(t_i)\right]$$

(4)

By using the first-order condition, the derivative of Equation (4) with respect to crime time t_i is obtained as follows:

$$D_1=\frac{\partial EU}{\partial t_i}=(1-p)U^{\prime }(I_1)G^{\prime }(t_i)+p{U}^{\prime }(I_2)\left[G^{\prime }(t_i)-F^{\prime }(t_i)-\alpha (fin)f^{\prime }(t_i)\right]=0$$

(5)

Following Equation (5), the optimal condition of crime time allocation is as follows:

$$\frac{(1-p)U^{\prime }(I_1)}{p{U}^{\prime }(I_2)}=-\frac{\left[G^{\prime }(t_i)-F^{\prime }(t_i)-\alpha (fin)f^{\prime }(t_i)\right]}{G^{\prime }(t_i)}$$

(6)

Based on Equation (5), we further obtain the second-order condition as follows:

$$\begin{array}{l}{D}_2=\frac{{\partial }^{2}EU}{\partial t_i^2}=(1-p)U^{″}(I_1){\left[G^{\prime }(t_i)\right]}^2+(1-p)U^{\prime }(I_1)G^{″}(t_i)\\ +p{U}^{″}(I_2)\left[G^{\prime }(t_i)-F^{\prime }(t_i)-\alpha (fin)f^{\prime }(t_i)\right]+p{U}^{\prime }(I_2)\left[G^{″}(t_i)-F^{″}(t_i)-\alpha (fin)f^{″}(t_i)\right]\end{array}$$

(7)

In Equation (7), it is clear that D₂ < 0, indicating that the problem of maximizing individual expected returns holds. Moreover, Equation (5) implies that t_i is the implicit function of fin, so we can obtain the derivative of t_i with respect to fin:

$$\frac{\partial t_i}{\partial fin}=-\frac{\left[(1-p)U^{″}(I_1)K_1+p{U}^{″}(I_2)K_2\right]W^{\prime }(fin)-p{U}^{\prime }(I_2)K_3}{D_2}<0$$

(8)

where K₁ = G′(t_i), K₂ = G′(t_i) − F′(t_i) − α(fin)f′(t_i), and K₃ = α′(fin)f′(t_i).

Equation (8) illustrates that the improvement of digital financial inclusion will lead to a decrease in the distribution of an individual’s crime time. Accordingly, this paper proposes the following research hypothesis to be verified:

Hypothesis 1 (H1).

The development of digital financial inclusion can reduce the theft crime rate.

The main reason behind hypothesis H1 is that the financing, investment, and payment functions of digital financial inclusion can generate a series of economic and social governance effects through such mechanisms as reducing financial transaction costs, lowering financial entry barriers, narrowing geographical exclusion, and reducing financing constraints. Particularly, digital financial inclusion has a positive impact on theft crime through the following two mechanisms.

The first mechanism is the expected revenue effect. According to Article 264 of the Criminal Law of the People’s Republic of China (PRC), theft is divided into four types, namely multiple thefts, burglary, armed theft, and pickpocketing. Among them, burglary and pickpocketing are the main forms of theft crimes, of which the primary targets are property with monetary or economic value, such as cash, mobile phones, and gold. Harbaugh [66] detected that more potential thieves are likely to commit theft activities as the value of stolen items increases. However, digital financial inclusion has improved the convenience of payment. Most of the traditional consumption that require cash payment, including shopping, dining, and transportation, can be performed through third-party payment platforms, thus reducing the frequency of exposing cash in public. In addition, with the development of digital financial inclusion, residents are also keeping less and less cash on their hands, which greatly hinders traditional cash theft. Based on a general equilibrium analytical model, Becket al. [67] proved that mobile payment can effectively reduce the incidence of theft crime related to funds. Therefore, the following research hypothesis can be proposed:

Hypothesis 2 (H2).

Digital financial inclusion changes the mode of payment, which decreases the expected payoff of theft, ultimately reducing the theft crime rate.

The second mechanism is the opportunity cost effect. It should be noted that the payoff of legal work also affects criminal activities. Usually, the theft offenders are relatively young, have relatively low levels of education, belong to low-skilled workers, and have relatively fewer employment options, which makes them more susceptible to exclusion in the labor market. Therefore, the labor market conditions play a vital role in the benefits of their legal work. Existing studies show that increases in the unemployment rate, which damages the welfare of local residents and intensifies social tensions, will raise the theft crime rate. A high incidence of theft activities, in turn, makes low-skilled workers more susceptible to exclusion and reduces the opportunity cost of theft, which further leads to an increase in theft crime [45]. On the contrary, a rise in the income from legal work leads to an increase in the opportunity cost of theft. If the expected payoff of labor is higher, the decision makers are more willing to choose employment rather than theft. In addition, the development of digital financial inclusion is conducive to promoting entrepreneurship. As a financial infrastructure, digital financial inclusion has released market potential, reducing financing and transaction costs, which provides favorable conditions for the emergence of new economic forms and job opportunities, such as express delivery, takeaway, live streaming, and the gig economy. Therefore, digital financial inclusion brings more entrepreneurship and employment opportunities in an active job market, which raises the opportunity cost of theft crime. The higher the employment rate and labor income in a region, the greater the gains from legal work. Accordingly, it means that the higher the opportunity cost of theft in the region, the lower the expected net revenue from theft. Based on the above analysis, the third hypothesis can be put forward as follows:

Hypothesis 3 (H3).

The inhibitory effect of digital financial inclusion on the theft crime rate is influenced by labor market conditions. The higher the labor income and employment rate in a region, the greater the effect.

4. Methodology, Variables, and Data Sources

4.1. The Specification of a Panel Econometric Model

Based on the previous theoretical analysis, this section will explore the mechanisms and effects of digital financial inclusion on urban theft crime from an empirical analysis perspective. Specifically, because the empirical sample is unbalanced panel data, referring to Ren et al. [68], this paper uses a two-way fixed effects (FEs) panel regression model to conduct the empirical analysis. The model can be expressed as follows:

Crime_it = α + βDfi_it + γ X_it + μ_i + δ_t + ε_it

(9)

where the subscript i denotes the city and the subscript t denotes the year. The explained variable Crime_it denotes the crime rate of the city i in year t; the core explanatory variable Dfi_it denotes the development level of digital financial inclusion in the city; and the variable X_it is a series of control variables, including per capita GDP, per capita fiscal expenditure, the rate of total deposits and loans to GDP, population density, the number of urban internet users, the number of higher education students, the registered unemployment rate, the ratio of secondary industry employment, the number of other criminal cases excluding theft, per capita judicial monitoring expenditure, etc. Parameters α, β, and γ are the regression coefficients of each variable in the model, where parameter β measures the effect of digital financial inclusion on urban theft crime. When β > 0, it indicates that the development of digital financial inclusion worsens urban theft crime, and conversely, when β < 0, it indicates that digital financial inclusion can reduce urban theft crime. In addition, μ_i denotes city-fixed effects that do not vary over time to control for the bias in the estimation results from factors such as factor endowments and cultural traditions in each city, δ_t denotes year-fixed effects that do not vary with cities, and ε_it is a random error term.

4.2. Variables Selection

(1) Explained variable: urban theft crime rate (crime). Most of the existing literature measure the level of crime rate at a macro level. For example, Wu et al. [5] used the number of arrests approved by procuratorial organizations per 10,000 people to measure the regional crime rate. Unlike the existing studies, this paper constructs an indicator of the urban theft crime rate based on the text data of criminal adjudication documents publicly available in China’s Judicial Documents website. Specifically, we first retrieved 1,043,100 first-instance criminal judgment documents from China’s Judicial Documents website with the keyword “theft”, then we identified the case information and extracted the data of 1,306,581 defendants by regular expressions, and finally, we matched the data with the location of the trial court, the trial time of the case, and city-level variables, so that the indicator of urban theft crime rate could be formed. It should be noted that the theft crime rate is measured by the number of defendants in the judgment documents, which inevitably encounters the problem of measurement error. First, the complete measurement of the actual crimes is an international challenge. If the value of a theft case is small, it seems that the case will not be reported by the victim and will not be filed by the public security authorities. Second, the measurement of crime rates must adhere to legal facts based on procedures and evidence rather than “objective truth”, so cases of theft that are solved but not prosecuted or adjudicated are excluded in this study. Moreover, there are certain standards for theft crimes in terms of case value. The above factors have led to a significant gap between the number of cases filed by public security authorities and the number of cases judged by courts, indicating that the defendants of theft crimes published by judgement documents are the perpetrators of more serious theft crimes. Despite the existence of the dark figure of crime (undetected, unreported, unfiled, etc.) in criminal cases, using official data to measure crime rates is still not unreliable since unreported cases are generally minor. Therefore, the explained variable in this paper is the theft crime rate by region, which is measured by the number of defendants with first-instance criminal judgment documents for theft per 100,000 people.

(2) Core explanatory variable: the level of digital financial inclusion (Dfi). The Digital Financial Inclusion Index, which is jointly compiled by the Digital Finance Research Center of Peking University and Ant Group Research Institute, is used in the study to measure the development level of digital financial inclusion in each city. The composition of this index is detailed in China’s Digital Inclusive Finance Indicator System and Index Compilation published by the Digital Finance Research Center of Peking University (2021). The total index of digital financial inclusion can also be divided into three sub-indicators, the coverage breadth, the usage depth, and the digitization level, which are used to measure the coverage level of digital financial inclusion, the usage frequency of digital financial inclusion, and the convenience and cost of digital financial inclusion, respectively.

(3) Control variables. In addition to digital financial inclusion, there are other factors that affect the urban theft crime rate. To avoid regression bias from omitted variables, this paper also introduces a series of control variables related to economic crime in the model, including per capita GDP (a measure of economic development), per capita fiscal expenditure (a proxy variable of public safety spending), population density (controlling the convenience of criminal activities), the proportion of total deposits and loans to GDP (a measure of financial development), the number of Internet users (a measure of Internet penetration), the number of students enrolled in higher education (a measure of educational attainment), the registered unemployment rate, and the ratio of employment in the secondary industry. In addition, to reduce the influence of other crimes on the number of theft crimes and to measure the regional differences in the efforts to combat crime and the proportion of adjudication documents uploaded in each region, this paper also considers the number of first-instance judgments of other criminal cases excluding theft crime per 100,000 people. Additionally, to exclude the influence of video surveillance on theft crimes, this paper extracts the annual amount of purchasing video surveillance equipment in each city from China’s Government Procurement Network. For simplicity, the depreciation rate of video surveillance is not considered, and then the per capita judicial monitoring expenditure in each city is used as a proxy variable for the distribution of video surveillance.

4.3. Data Sources and Description

Based on the data availability and the actual needs of the study, this paper selects the unbalanced panel data of 289 cities in China during 2014–2019 as the empirical study sample (excluding Tibet, Hong Kong, Macau and Taiwan) to explore the nexus between digital financial inclusion and urban theft crimes. The data are obtained from the China City Statistical Yearbook in previous years, the Digital Finance Research Center of Peking University, and China’s Judicial Documents website of the Supreme People’s Procuratorate of the PRC. Specifically, the data of urban theft crimes are constructed by retrieving 1,043,100 first-instance criminal judgment documents from China’s Judicial Documents website with the keyword “theft”, then the regular expressions are adopted to identify the case information and extract the data of 1,306,581 defendants, which are matched with city attributes finally. The data on digital financial inclusion are obtained from the Digital Inclusive Finance Index jointly compiled by the Digital Finance Research Center of Peking University and the Ant Group Research Institute, which can be divided into three sub-indicators, the coverage breadth of digital financial inclusion, the usage depth of digital financial inclusion and the digitization level of digital financial inclusion. The data of the control variables in the model are obtained from the China City Statistical Yearbook in previous years. For a small portion of missing data in some prefecture-level cities, this paper uses the statistical yearbooks of the corresponding provinces to supplement as much as possible. In order to eliminate the impact of price fluctuations, all nominal value variables in the model are deflated to real variables using the consumer price index of each city in 2014. In addition, all absolute value variables in the model are logarithmized to eliminate the heteroskedasticity as much as possible. The descriptive statistics of all variables are shown in

Table 1. The results of description statistics of all variables.

Symbol	Definition	Obs	Mean	Std. Dev.	Min	Max
Crime	the rate of urban theft crime	1733	14.2	9.2	1.3	73.3
Dfi	the index of digital financial inclusion	1733	202.4	40.3	105.6	321.6
PGDP	per capita GDP	1731	10.8	0.5	9.2	12.3
Fiscal	per capita fiscal expenditure	1732	9.1	0.4	8.0	11.7
Finance	the rate of total deposits and loans to GDP	1732	2.6	1.4	0.7	21.3
Pop_density	population density	1733	470.6	560.1	5.7	6626.3
Internet_per	the number of urban Internet users	1718	0.2	0.1	0.0	1.3
Education	the number of higher education students	1687	180.9	207.0	2.4	1148.4
Unemp	registered unemployment rate	1705	0.1	0.0	0.0	0.3
Second_emp	the ratio of secondary industry employment	1731	43.7	14.5	7.4	83.4
Other_crime	the number of other criminal cases excluding theft	1733	52.5	24.2	2.1	318.4
Moni_jud	per capita judicial monitoring expenditure	1733	1.9	1.7	0.0	7.0

.

5. Empirical Analysis

5.1. Baseline Regression Analysis

To explore the impact of digital financial inclusion on the urban theft crime rate, this paper uses a two-way fixed effects model for empirical analysis according to Model (9). The regression results for the full sample are shown in

Table 2. The regression results of impact of digital financial inclusion on urban theft crime for full samples.

Variable	(1)	(2)	(3)	(4)
Dfi	−0.505 ***	−0.613 ***	−0.621 ***	−0.577 ***
	(−3.31)	(−3.99)	(−3.90)	(−3.98)
PGDP		0.417 ***	0.362 ***	0.343 ***
		(3.68)	(2.79)	(2.78)
Fiscal		−0.069	−0.187	−0.157
		(−0.49)	(−1.12)	(−1.08)
Finance		0.013	0.006	0.003
		(0.66)	(0.25)	(0.14)
Pop_density			−0.000	−0.000
			(−1.11)	(−1.10)
Internet_per			−0.029	−0.026
			(−0.15)	(−0.15)
Education			0.002 ***	0.001 **
			(3.14)	(2.19)
Unemp			−0.908	−0.863
			(−1.23)	(−1.14)
Second_emp			0.002	0.001
			(0.48)	(0.31)
Other_crime				0.009 ***
				(3.81)
Moni_jud				0.063 **
				(2.49)
_Cons	−0.841 ***	−4.854 ***	−3.419 *	−3.727 **
	(−3.95)	(−2.88)	(−1.80)	(−2.21)
Year FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Obs	1732	1730	1649	1649
R2	0.190	0.198	0.210	0.298
F Statistic	45.96	33.05	20.60	19.37
Number of Cites	289	289	287	287

Note: t value in parentheses. ***, **, and * mean significance at the 1%, 5%, and 10% levels, respectively.

. Specifically, Column (1) shows the regression results when controlling only for the two-way fixed effects. Column (2) shows the regression results when controlling for the macroeconomic variables at the prefectural level, including per capita GDP, per capita fiscal expenditure, financial development, and the two-way fixed effects. Column (3) shows the regression results when controlling not only for the prefecture-level macroeconomic variables and the two-way fixed effects but also the demographic factors, which include population density, the number of urban Internet users, the number of higher education students, the urban registered unemployment rate, and the ratio of employees in the secondary industry. Column (4) shows the regression results when further controlling for the number of other criminal cases excluding theft and the per capita judicial monitoring expenditure on the basis of Column (3).

From the above regression results, it can be seen that the development of digital financial inclusion has a significant negative impact on the urban theft crime rate, which tentatively validates the hypothesis H1 mentioned in this study. Specifically, according to the results in Column (4), there is a significantly negative correlation between digital financial inclusion and the urban theft crime rate, namely, each standard deviation in the digital financial inclusion index is associated with a 0. 577 standard deviations decrease in the number of defendants for theft per 100,000 people, indicating that digital financial inclusion can effectively reduce urban theft crime. This conclusion is highly consistent with the study of Harbaugh et al. [66].

5.2. Robustness Analysis

The results in

Table 3. The robustness test results of impact of digital financial inclusion on urban theft crime.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Dfi	−0.609 ***	−0.575 ***	−3.214 **	−0.603 ***	−0.400 ***	−0.486 ***	−0.402 ***
	(−4.26)	(−3.96)	(−2.32)	(−4.31)	(−2.91)	(−3.51)	(−3.11)
_Cons	−4.088 **	−3.705 **	−60.154 **	−4.083 **	−3.860 ***	−2.426	−4.936 **
	(−2.44)	(−2.19)	(−2.27)	(−2.45)	(−2.64)	(−1.59)	(−2.27)
Control Variable	YES	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES	YES	YES
Obs	1626	1649	192	1649	1649	1649	1632
R2	0.307	0.298	0.395	0.304	0.301	0.276	0.429
F Statistic	19.51	19.20	6.04	19.93	20.29	14.53	11.67
Number of Cites	283	287	33	287	287	287	282

Note: t value in parentheses. *** and ** mean significance at the 1% and 5% levels, respectively.

further confirm the robustness of the baseline regression. Considering that the political attention, economic development level, population migration, population agglomeration, and public security level in municipalities may be quite different from other cities, Column (1) of Table 3 shows the regression results when excluding four municipalities, namely Beijing, Shanghai, Tianjin, and Chongqing. The results reveal that the inhibitory effect of digital financial inclusion on urban theft crime is still robust, and the coefficient is very close to Column (2) of Table 2.

It is worth noting that theft often occurs in crowded areas, especially railway stations and bus stations. On the one hand, theft crimes in railway stations and carriages may be more related to current assets, such as cash and mobile phones, which are more significantly affected by digital financial inclusion. On the other hand, theft crimes in railway stations and carriages are often committed by gangs of professional offenders whose cost–benefit analysis will be more complicated. In addition, criminal activities in railway stations and carriages are handled by the railway public security departments, and cases are adjudged by the railway court. However, the crimes handled by a single railway court include not only the cases that occurred in the city where the court is located but also the cases that occurred in other surrounding cities along the railway. Therefore, after removing the cases decided by the railway court, the regression results are reported in Column (2) of Table 3, indicating that the conclusion is still consistent with the baseline regression. Furthermore, Column (3) shows additional regression results, in which such theft cases heard by the railway court are retained, involving 6517 theft defendants for 33 prefecture-level cities during 2014–2019. Considering the characteristics of theft crimes committed on the railway, this paper uses the number of theft defendants as the explained variable and the development level of digital financial inclusion in the city where the railway court is located as the core explanatory variable. The results show that the development of digital financial inclusion significantly reduces theft crimes against train passengers.

There are 38,027 unpublished documents in the first-instance theft judgment documents of 2014–2019. If these unpublished cases are mainly caused by the non-standard online operation of court documents, it may lead to estimation errors. In the baseline regression, these judgment documents have been removed. Considering that most thefts (82.44% of the cases in public) are committed alone, we suppose that each unpublished case has one defendant, and the regression results are reported in Column (4) of Table 3, indicating that the results still support the above conclusions in baseline regression.

Nearly 17% of the theft cases are committed by gangs (cases with 2 or more defendants in 177,165 documents). In view of the fact that the motivations of the gang crime are more complex, which may be affected by social interaction factors, all cases with more than one defendant have been removed in the study, and the regression results reported in Column (5) of Table 3 are still significantly negative. The distribution of fines in theft shows that there are some cases with huge fines, and the circumstances and motivations of these cases are often more complicated, such as the massive theft of valuables such as gasoline. In Column (6) of Table 3, the theft cases with fines exceeding the 99% quantile (The numerical value of the 99% quantile is CNY 60,000) are fully removed, and the estimation results still support the main conclusions of this study. In Column (7), in order to avoid errors due to control variables that may be affected by the development level of digital financial inclusion, the interaction term between the initial control variables (the variables in 2014) and the year fixed effect is incorporated into the econometric model as the control variables. Although the estimation result is smaller than the baseline results in Table 2, the impact of digital financial inclusion on urban theft crime is still significantly negative at the 1% significance level.

5.3. Instrumental Variable Regression Analysis

The results of baseline regression and robustness tests have shown that there is a significant negative correlation between digital financial inclusion and the theft crime rate. Then, is there a causal relationship between them? Theoretically, due to the high expected risk of suffering from theft in higher larceny areas, residents are more willing to choose electronic payment, which will lead to the issues of omitted variables. Although the two-way fixed effect model can to some extent reduce some errors due to omitted variables, it cannot reduce errors caused by variables that change over time and cities. Moreover, because the theft crime rate may also be affected by such unobservable factors as urban supervision, case filing standards, and law enforcement, it is difficult to control for all factors in the regression analysis.

Referring to the current academic studies on digital financial inclusion [52,63], the interaction term between the spherical distance from each city to Hangzhou and the national digital financial inclusion index for each year is used as the instrumental variable of the regional digital financial inclusion index in this study. The main reason is that the development of digital financial inclusion is affected by geospatial factors. On the one hand, the farther it is to Hangzhou, the more difficult it is to promote the development of digital financial inclusion, which meets the relevant requirements. On the other hand, the spherical distance to Hangzhou is unlikely to correlate with the local theft risks or unobservable institutional factors, such as law enforcement and the supervision level, which meets the exclusive requirements. However, the spherical distance from each city to Hangzhou is just cross-sectional data. In order to match the sample panel data, we multiply it by the annual digital financial inclusion index at the national level to form an instrumental variable of the digital financial inclusion index that changes over time. It should be noted that the development of national digital financial inclusion will affect the development of regional digital financial inclusion, which meets the relevant requirements. Meanwhile, the national digital financial inclusion is unlikely to affect the level of law enforcement and other unobservable institutional factors in cities, which meets the exclusive requirements. The regression results are shown in Column (1) and Column (2) of

Table 4. The IV regression results of impact of digital financial inclusion on urban theft crime.

Variable	Full Sample		Sample Excluding Hangzhou
	(1)	(2)	(3)	(4)
	Stage One	Stage Two	Stage One	Stage Two
	Dfi	Crime	Dfi	Crime
Dfi		−1.720 ***		−0.644 ***
		(−5.90)		(−2.66)
Distance_HZ	−0.000 ***		−0.000 ***
	(−11.35)		(−10.48)
Control Variable	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Obs	1648	1648	1588	1588
R2		0.210		0.275
F Statistic		23.58		27.76
Number of Cites	286	286	276	276

Note: t value in parentheses. *** means significance at the 1% level.

. It is clear that the impact of digital financial inclusion on theft crime is still significantly negative at the 1% level, which further verifies the above hypothesis H1. In Column (3) and Column (4), by removing the observation data of Zhejiang Province, the results show that the effect of digital financial inclusion on urban theft crime is also significantly negative, and the coefficient of digital financial inclusion is very close to the baseline regression results in Table 2. The results illustrate that there is a correlation between the spherical distance to Hangzhou and the theft rate due to the issues of omitted variables, while the impact of the omitted variables is little outside Zhejiang Province.

5.4. Heterogeneity Analysis

5.4.1. Heterogeneity of the Fines

The impact of digital financial inclusion on urban theft crime may be correlated with the value of theft. Generally, in theft cases with fewer fines, the defendants usually commit the crime less frequently. Because the value of the stolen items is low, they are less likely to engage in a rational analysis of cost–benefit. For example, in our study samples, the proportion of theft fines under CNY 10,000 is 79.27% for theft cases of e-bikes and 74.41% for theft cases of mobile phones. However, the proportion of theft fines under CNY 10,000 is 65.19% for cash theft cases. Therefore, this paper uses the fines of theft cases as the proxy variable of the value of theft cases and then divides the sample data into two groups according to the median of it for heterogeneity analysis. The results are shown in Columns (1) and (2) of

Table 5. The heterogeneity results of impact of digital financial inclusion on urban theft crime.

	Fines		Education Level
Variable	(1)	(2)	(3)	(4)	(4)
	More than Mean	Less than Mean	Low	Middle	High
Dfi	−0.523 ***	−0.334 **	−0.470 ***	−0.455 ***	−0.682 **
	(−3.56)	(−2.50)	(−3.60)	(−3.62)	(−2.42)
_Cons	−5.363 **	0.187	−3.078 **	−3.886 ***	−3.820 ***
	(−2.56)	(0.16)	(−2.05)	(−2.59)	(−2.85)
Control Variable	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes
Obs	1649	1649	1649	1649	1529
R2	0.232	0.195	0.282	0.280	0.175
F Statistic	18.28	8.89	17.47	17.44	9.36
Number of Cites	287	287	287	287	265

Note: t value in parentheses. *** and ** mean significance at the 1% and 5% levels, respectively.

, of which Column (1) shows the results where the fine is more than the median and Column (2) where fine is less than the median. It can be seen that the absolute coefficient in Column (1) is greater than that in Column (2), and that the statistical results are more significant, which indicates that the development of digital financial inclusion has a greater impact on theft with a higher value of theft cases. One possible explanation is that most cases with fewer fines are mostly related to the theft of supermarket goods, e-bikes, and mobile phones, and the proportion of recidivists is even smaller. Therefore, the development of digital financial inclusion has less impact on the expected benefits of theft.

5.4.2. Heterogeneity of the Defendant’s Education Level

Some studies have shown that education can effectively curb the occurrence of criminal activities [34]. Moreover, education can increase the benefits of legal jobs, which leads to a higher opportunity cost of crime. In Columns (3) to (5) of Table 5, the education level of the defendants is divided into three groups, namely low, middle, and high groups for heterogeneity analysis. The regression results of the low group are reported in Column (3) when the education level is 0 to 6 years. The results of the middle group and high group are presented in Column (4) and Column (5) when the education level is 7 to 12 years and 13 years or more. It turns out that the development of digital financial inclusion has a greater impact on theft committed by highly educated individuals. A possible explanation is that the group with a higher education level can better adapt to the development of the Internet era and seize the new employment and entrepreneurial opportunities arising from the development of digital financial inclusion. The results are also consistent with the conclusions of Zhang [6].

5.5. Transmission Mechanism Analysis

Although the results mentioned above show that digital financial inclusion can effectively reduce the urban theft crime rate, the mechanisms by which digital financial inclusion affects theft crimes should be explored. The digital financial inclusion index is a multi-dimensional indicator that includes coverage breadth, usage depth, and digitization level. We further analyzed the different impacts of the sub-indicators of the digital financial inclusion index on theft crimes. The coverage breadth reflects the number of Alipay’s users and bank cards. The usage depth reflects the usage of payment, currency, credit, insurance, investment, and credit services. The digitization level reflects the payment convenience and efficiency of regional digital financial inclusion. Referring to Yang et al. [62], as shown in Column (1) and Column (3) of

Table 6. The regression results of impact of different dimension indicators of digital financial inclusion on urban theft crime.

Variable	(1)	(2)	(3)	(4)
Coverage_breadth	0.129			0.061
	(0.96)			(0.52)
Usage_depth		−0.262 **		−0.213 *
		(−1.99)		(−1.79)
Digitization_level			−0.146 ***	−0.140 ***
			(−4.52)	(−4.40)
_Cons	0.469	−1.206	−2.029	−2.583
	(0.24)	(−0.64)	(−1.19)	(−1.55)
Control Variable	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
Obs	1649	1649	1649	1649
R2	0.276	0.280	0.304	0.307
F Statistic	20.21	19.01	19.99	18.52
Number of Cites	287	287	287	287

Note: t value in parentheses. ***, **, and * mean significance at the 1%, 5%, and 10% levels, respectively.

, the coverage breadth, the usage depth, and the digitization level are added into Model (2) for empirical analysis, respectively. Additionally, the three sub-indicators of digital financial inclusion are added together in Column (4). The regression results show that the digitization level has a significant inhibitory effect on theft crime at the 1% significance level. According to the theoretical analysis mentioned above, the convenience of payment can directly affect people’s payment options. The higher proportion of mobile payments indicates that people are more likely to use mobile payments rather than cash payments in the payment process, which reduces the expected benefits of theft crime. The results effectively confirm the hypothesis H2 mentioned in this study.

If the mobile payment function of digital financial inclusion takes effect by reducing the expected benefits of theft, digital financial inclusion will have a greater impact on cash theft. Therefore, by using regular expressions, the cases of cash theft are extracted, and the cases related to mobile phone and e-bike are also extracted for comparison. Column (1) in

Table 7. The regression results of impact of digital financial inclusion on different types of theft.

Variable	(1)	(2)	(3)
	Cash	Phone	E−Bike
Dfi	−0.591 ***	−0.378 **	−0.108
	(−2.69)	(−2.24)	(−0.59)
_Cons	−2.293	−5.734 ***	−5.205 ***
	(−0.75)	(−2.87)	(−3.71)
Control Variable	YES	YES	YES
Year FE	YES	YES	YES
City FE	YES	YES	YES
Obs	1649	1649	1649
R2	0.384	0.461	0.393
F Statistic	39.24	41.89	19.42
Number of Cites	287	287	287

Note: t value in parentheses. *** and ** mean significance at the 1% and 5% levels, respectively.

shows the impact of digital financial inclusion on cash theft. Columns (2) and (3) show the impacts of digital financial inclusion on theft crimes related to mobile phones and e-bikes, respectively. The results indicate that the development of digital financial inclusion has a greater impact on reducing cash theft, which further confirms the hypothesis H2. In Column (3), the results reveal that digital financial inclusion has little impact on e-bike theft, indicating that there is no evidence that the larcener shifts the theft target owing to the development of digital financial inclusion.

In order to verify hypothesis H3, this paper uses the logarithm of per capita GDP and the registered unemployment rate as proxy variables of regional labor market conditions. In Column (1) of

Table 8. The regression results of impact of digital financial inclusion on urban theft crime under different labor market conditions.

			Economic Growth		Unemployment
Variable	(1)	(2)	(3)	(4)	(5)	(6)
			High	Low	High	Low
Dfi	1.125 ***	−0.567 ***	−0.706 ***	−0.064	−0.051	−0.866 ***
	(3.16)	(−3.90)	(−3.63)	(−0.65)	(−0.34)	(−4.53)
PGDP × Dfi	−0.142 ***
	(−4.53)
Unemp × Dfi		1.603 ***
		(3.25)
_Cons	−1.618	−4.114 **	−3.511	−3.446 *	−4.304 **	−3.553
	(−0.98)	(−2.41)	(−1.16)	(−1.88)	(−2.43)	(−0.92)
Control Variable	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES	YES
Obs	1649	1649	840	809	835	814
R2	0.327	0.306	0.328	0.435	0.330	0.347
F Statistic	19.54	18.92	9.17	25.07	15.61	8.88
Number of Cites	287	287	145	142	147	140

Note: t value in parentheses. ***, **, and * mean significance at the 1%, 5%, and 10% levels, respectively.

, the interaction term between digital financial inclusion and the logarithm of per capita GDP is added into the Model (2), and in Columns (2), the interaction term between digital financial inclusion and the unemployment rate is further added. The results indicate that the higher the per capita GDP and the lower the registered unemployment rate, the more significant the impact of digital financial inclusion on the urban theft crime rate, which effectively confirms the hypothesis H3 mentioned in this paper. In addition, in Column (3) and Column (4), the sample cities are divided into developed regions and developing regions by the median of the national per capita GDP in 2014. In Column (5) and Column (6), the samples are further divided into high-unemployment regions and low-unemployment regions by the median of the unemployment rate in 2014. Specifically, the results in Columns (3) and (6) reveal that the impact of digital financial inclusion on theft crime is more significant in developed regions and low-unemployment regions, but the regression coefficient is not significant in developing regions and high-unemployment regions. Moreover, the results also demonstrate that the development of digital financial inclusion has activated the originally developed markets and lowered the standards of market entry, thus providing more opportunities for legal jobs and increasing the opportunity cost of theft, indicating that the hypothesis H3 mentioned above in the study is confirmed effectively.

6. Conclusions

The Internet revolution is another major revolution in the history of human civilization following the invention of the steam engine and the use of electricity, which has engendered a profound impact on the global economy and human life. As an important part of the Internet era, digital financial inclusion also deeply affects social behavior. Meanwhile, as a significant and frequent criminal activity, theft crime can cause great economic losses and mental trauma to the public. Therefore, studies on the prevention and governance of theft activities are of great significance in reducing the loss of life and property and enhancing the sense of people’s security and well-being.

Based on a crime time allocation model and the judicial big data of first-instance theft crime adjudication documents published on China’s Judicial Documents website, this paper constructs unbalanced panel data of theft crime for 289 cities in China covering the period 2014–2019 and explores the mechanisms and effects of digital financial inclusion on the urban theft crime rate. The results show that the development of digital financial inclusion significantly reduces the urban theft crime rate. This effect is mainly generated by the transaction convenience related to digital financial inclusion and has a stronger impact on criminals with more serious cases and higher levels of education. In addition, in terms of impact mechanisms, digital financial inclusion significantly curbs theft crime by reducing the expected benefits for offenders and increasing the opportunity cost of theft.

This paper examines whether criminals choose to commit theft from the perspective of the expected benefits and opportunity costs of theft crime. The implicit assumption is that criminals will choose between theft and legal work. In reality, some criminals may not choose between illegal or legal activities but rather choose the activity with the greatest benefits among different criminal activities to commit crimes, such as economic criminals choosing among fraud, illegal fund-raising, pyramid schemes, etc. If theft offenders make the same decisions, the result that digital financial inclusion reduces theft crime, as verified in this study, may not necessarily lead to an increase in social welfare due to the mutual substitution between different criminal activities. However, this paper argues that the logic may not always applicable for theft crimes. Firstly, existing studies have shown that there are similarities between burglars and violent offenders in multiple dimensions, such as psychology, family background, and education, but there are significant differences compared to economic offenders and drug-related offenders. Secondly, in the sample data, nearly 90% of the defendants in theft cases only received junior high school education or below. This means that the majority of them can only engage in low-threshold, low-skilled criminal activities. If they are also low-skilled workers in the labor market, it is difficult for them to engage in other economic crimes with higher skills. Thus, reducing theft crime is also an important social benefit for the development of digital financial inclusion.

The analysis in this study shows that the impact of the digital economy on China’s high-quality development is not limited to traditional channels, such as consumption, investment, innovation, and entrepreneurship [14,69]. Promoting digital financial inclusion, especially improving the digitization level and enhancing the convenience of payments, not only can help to promote domestic consumption and inclusive growth but also has crime governance functions, which can bring huge potential socio-economic benefits. Therefore, when assessing the role of digital economy development, we should not only consider its economic effects but also focus on its social effects. Regulatory and law enforcement authorities should re-examine the changes in the distribution and trend characteristics of theft crime, actively adapt to the development of China’s digital economy, and pertinently improve crime governance tools and instruments in the Internet era.

In addition, the basic way to deal with theft crimes is to increase the opportunity cost of theft. This requires us to deeply follow the concept of new development, continuously promote inclusive growth, vigorously develop financial inclusion, reduce the exclusion of low-skilled workers in the labor market, lower income disparities, and improve the level of public services, ultimately enabling the people to equally share the gains of China’s reform and opening up. Only by these ways can we effectively reduce the incidence of theft crime and lay a solid foundation for building a safe society in China, achieving high-quality growth, and enhancing the sense of people’s happiness and satisfaction.