The Effect of Income Polarization on Crime: Evidence From Court Judicial Documents in China

1 Introduction

Crime generates substantial social costs (Becker 1968). Existing studies demonstrate that criminal activities not only significantly reduce the residents' consumption (Mejía and Restrepo 2016) and human capital accumulation (Macmillan 2001), but also diminish corporate competitiveness (Loayza et al. 2000). To prevent crime, international organizations have set a number of strategic goals. For example, the United Nations integrates the violence prevention and crime reduction into its Sustainable Development Goals, while the World Bank implements crime mitigation through Urban and Social Development Programs. As an important participant in global crime governance, China has put forward strategic goals such as Safe China and Law-based China. Effective crime reduction requires not only enhanced law enforcement—a costly governance approach (Kearney et al. 2014; Fone et al. 2023)—but also a deeper analysis of criminal behavior determinants to develop more targeted social governance policies.

Income distribution is an important socioeconomic factor that affects crime. Extensive empirical evidence confirms that income inequality leads to higher crime rates, with some studies revealing a positive correlation between criminal activities and the Gini coefficient (Kelly 2000; Enamorado et al. 2016). Beyond income inequality, income polarization is also an important indicator for measuring income distribution. Economists define income polarization as the phenomenon of the concentration of income distribution towards the two ends or a particular point. More generally, income polarization can be manifested as the concentration of income levels at multiple poles. Currently, income polarization pressures are intensifying. On the one hand, the rise of new-generation information technologies such as digital economy and artificial intelligence brings about significant income and job polarization effects (Autor and Dorn 2013; Michaels et al. 2014; Wang and Dong 2023). On the other hand, China's middle class faces the risk of downward mobility due to the decline in the working-age population, the slowdown in economic growth, and the increased uncertainty of the international environment (Xu and Liu 2019).

Compared with income inequality, income polarization is a more important factor in determining social stability (Bhattacharya et al. 2015). Rising income polarization amplifies social tensions by intensifying social conflicts (Esteban and Ray 1994, 2011; Østby 2008). However, there are scarce studies examining the impact of income polarization on crime (Lee and Shin 2011; Li et al. 2019), with existing studies not adequately addressing the endogeneity issue as well as the related mechanism analyses.

To address this gap, this paper exploits the variation in income polarization and crime rates across cities and over time in China to analyse the impact of income polarization on criminal activities and to investigate the underlying mechanisms. To accurately measure the crime indicators at the city level, we performed textual analysis of judgment documents at all levels of courts in China, thereby compiling data on the number of criminal cases and criminals in 2014, 2016, and 2018. Moreover, we also compiled data on the number of cases and criminals for different types of crimes to facilitate heterogeneity analyses.

We find that an increase in income polarization can significantly raise crime rates. The baseline results indicate that a one-standard-deviation increase in the income polarization index corresponds to an approximately 4% increase in both criminals per 10,000 people and criminal cases per 10,000 people. After employing the rice-to-wheat planting area ratio in 2000 as an instrumental variable (IV) to address the endogeneity issue and conducting a variety of robustness checks, the above conclusion remains robust. The heterogeneity analyses find that income polarization significantly increases violent crime, robbery and stealing, drug crime and financial fraud. Regarding city characteristics, the impact of income polarization on crime is stronger in cities with a higher proportion of young and migrant populations. To shed some light on the mechanisms through which income polarization affects crime, we decompose the income polarization index into two parts: identification and alienation. Our findings show that alienation, rather than identification, drives the increase in crime rates. In addition, we also provide empirical evidence that income polarization, especially alienation, induces criminal activities by reducing the job-seeking willingness of the unemployed and the subjective happiness of residents.

Our study contributes to several strands of literature. First, our study supplements the literature on the impact of income distribution on crime. Existing studies have focused on the impact of income inequality and poverty on crime (Kelly 2000; Enamorado et al. 2016; Hu et al. 2024), and have reached contradictory conclusions (Pazzona 2024). Building on the work of Li et al. (2019), and using millions of judicial documents, this paper refines the impact of income polarization on crime to the city level in China, providing empirical evidence that income polarization significantly leads to an increase in crime rates. Second, based on the “rice theory” proposed by Talhelm et al. (2014), this paper uses the rice-to-wheat planting area ratio in 2000 to measure the degree of collectivism in different cities, and constructs an IV for income polarization, providing a rigorous identification framework for studying the socioeconomic impact of income polarization.

The remainder of this paper is organized as follows. Section 2 presents the literature review. Section 3 describes the data. Section 4 presents the empirical strategy used in this study. Section 5 reports the empirical results, including the baseline results, IV regression results, robustness checks, and heterogeneity analyses. Section 6 explores the potential mechanisms underlying the relationship between income polarization and criminal activities. Section 7 concludes.

2 Literature Review

3 Data

3.1 Data on Crime Rates

The crime data used in this paper are obtained from the official judicial document platform of the Supreme People's Court of China, namely the "China Judgments Online" (CJO) Database. On January 1, 2014, the Supreme People's Court issued the “Regulations on the Publication of Judgments by People's Courts on the Internet”. Adhering to the principle that “openness is the norm and nondisclosure is the exception”, the CJO ensures the comprehensiveness and accuracy of the judicial documents data to the greatest extent (Yi et al. 2023). The format of criminal litigation documents is uniformly regulated. The main body of each document is divided into five parts: the heading, the facts and evidence, the rationale, the judgment outcome, and the concluding section. In addition, the documents also include the year in which the documents were filed.

This fixed format allows us to use natural language processing to identify key information in the documents. For example, the number of criminals and their locations can be identified in the heading, the type of crime can be identified in the judgment outcome, and the court name and judgment date can be extracted from the concluding section. We focus on criminal cases that violate the Criminal Law of the People's Republic of China, as the social costs resulting from criminal activities are much higher (Loayza et al. 2000).

The specific processing procedure of judicial documents is as follows: by searching for documents on the CJO, selecting “criminal cases” as the cause of action and “first-instance judgment” as the document type, we can obtain all documents with complete content from 2014 to 2020. After counting the number of documents and identifying the number of criminals in each case, we derive the annual totals of criminal cases and criminals for each city.

There are two major challenges involved in this process, namely, determining the time and location of the crime. Many criminal cases are characterized by continuous criminal activity, resulting in more than one identifiable crime time on the judicial documents. To address this, following Ma et al. (2023), we use the judgment date as a proxy for the crime time.

Regarding the location of the crime, Article 24 of China's Criminal Procedure Law stipulates that “Criminal cases shall be under the jurisdiction of the people's court where the crime was committed. If it is more appropriate for the people's court where the defendant resides to conduct the trial, the court where the defendant resides may have jurisdiction.” Yi et al. (2023) pointed out that criminal cases are mainly under the jurisdiction of the people's court where the crime was committed. Therefore, this paper identifies the location of the crime by recognizing the name of the court, that is, the county where the court is located.

Then, the number of crime cases and the number of criminals from courts at all levels within a city are aggregated to obtain the crime data for each city per year. Figure 1 reports the number of crime cases and criminals calculated using the CJO from 2014 to 2018. It can be observed that criminal activities show an upward trend throughout the period under study. Additionally, Figure 1 demonstrates that the crime data derived from the CJO are consistent in trend with the official crime data published in the annual work reports of China's Supreme People's Court (SPC), which validates the reliability and representativeness of the crime data used in this study. In addition, we also classify cases based on judicial documents and aggregate crime data of different types of crime at the city-level.

3.2 Data on Income Polarization

Income polarization refers to the phenomenon where the income distribution converges towards the high and low ends of the distribution from the middle class. More generally, it can be the convergence towards multiple poles. In terms of measuring income polarization, previous work has been done by Wolfson (1994) and Esteban and Ray (1994). The latter, through subjective determination, divided the total population into several groups. For example, dividing income groups based on the median as a dividing line. However, in regions with large (small) income gaps, certain individuals whose income is higher (lower) than the local median income may actually belong to the low (high) income level. To overcome this shortcoming, Duclos et al. (2004) introduced the income density function to divide income groups, and thereby constructed the DER index to measure the degree of multi-pole income polarization:

$$$\mathrm{DER}={P}_{a}(f)=\iint {f(x)}^{1+\alpha }f(y)|x-y|{dxdy}$$$()

where, α is a sensitive parameter of multi-polarization, and α ∈ [0.25, 1]. The larger the value of α, the higher the weight assigned to within-group identification, and the greater the difference between the calculated DER index and the Gini coefficient. f(x) and f(y) represent the density functions of income levels x and y, respectively.

In the case of discrete data, the probability density functions f(x) and f(y) can be obtained through non-parametric estimation methods. Taking f(y) as an example, the probability density function can be estimated using a kernel density estimation $$f(y)=\frac{1}{{nh}}{\sum }_{i=1}^{n}K(\frac{y-{y}_{i}}{h})$$, where $$h$$ is the bandwidth, and $$K({\rm{\cdot }})$$ is the kernel function. If the distribution is concentrated around a few income points (i.e., multi-polar income polarization), it is reflected by the higher probability density at these income poles. Therefore, the distinction and identification of income groups by the DER Index are based on the inherent clustering characteristics of the income distribution itself, rather than being subjectively determined.

Unlike the Gini coefficient, which reflects the average dispersion of income distribution relative to the mean, the polarization index captures the income clustering within the income distribution (Gornick and Jäntti 2013). The numerical example illustrated in Table 1 below, using simulated data, may further explain the DER index. It is assumed that the first and second columns represent two income sequences with the same total income and population size but different distributions. Specifically, the first column follows a uniform distribution, while the second column exhibits a multi-peaked distribution.

Table 1. A numerical example of DER index.

	Uniform distribution		Multi-peaked distribution
	# of people	Income	# of people	Income
	3	0	1	0
	3	25	7	25
	3	50	3	50
	3	75	1	75
	3	100	3	100
	3	125	1	125
	3	150	3	150
	3	175	7	175
	3	200	1	200
Total	27	2700	27	2700
Gini coefficient		0.3703		0.3663
DER index ($$\alpha =0.25$$)		0.2908		0.2966
DER index ($$\alpha =0.5$$)		0.2290		0.2405
DER index ($$\alpha =0.75$$)		0.1807		0.1953
DER index ($$\alpha =1$$)		0.1430		0.1589

• Note: Table 1 presents a numerical example of DER index.

It can be observed that, although the Gini coefficient of the uniform distribution is greater than that of the multi-peaked distribution, the DER index is exactly the opposite in magnitude. This is because the DER index is more sensitive to the income clustering or identification. Furthermore, it can be observed that as the value of α increases (giving more weight to within-group identification), the relative difference between the two types of income distributions becomes more pronounced.

We use per capita household annual income to calculate city-level income polarization based on CHFS. Conducted by the Center for Household Finance and Research at Southwestern University of Finance and Economics, the CHFS is a nationwide survey that has been carried out five waves between 2011 and 2019, covering 29 provinces, 181 cities, and 1481 communities in China, and is highly representative.1 To ensure the representativeness of the sample at the city level and the accuracy of measurement of income polarization, we make two adjustments. First, we adjust for the level of income polarization using the sampling weights. Second, we exclude cities with fewer than 60 households in the sample. Figure 2 shows that income polarization in China shows a gradual downward trend during the period from 2010 to 2018.

After excluding time-varying confounders at the provincial level, Figure 3 illustrates the relationship between income polarization and the number of criminal cases and criminals. It indicates that there is a positive correlation between income polarization and crime. In the subsequent empirical analysis, we will further examine the causal relationship between the two.

4 Empirical Strategy

$${X}_{{ct}}$$ denotes a set of city-level control variables. Following Enamorado et al. (2016) and Fone et al. (2023), we use GDP per capita, urbanization, unemployment, human capital, public security expenditure, fiscal expenditure, and financial development as the control variables.2 $${u}_{c}$$ and $${v}_{{pt}}$$ represent city fixed effects and province-year fixed effects, respectively. City fixed effects control for factors that do not vary over time at the city level. Province-year fixed effects, on the other hand, account for the impact of time-varying confounders at the provincial level. $${\varepsilon }_{{ct}}$$ denotes the standard errors clustered at the city level. Table 2 presents the summary statistics of the main variables.

Our empirical results may be subject to potential endogeneity issues. First, although we have considered the representativeness of the sample by using sampling weights and excluding cities with a small number of observations, there may still be measurement errors in the estimation of income polarization. Second, there may be reverse causality between criminal activities and income polarization, as individuals from higher-income groups may migrate away from areas with severe crime. Third, it is clearly challenging to control for all potential confounders.

Therefore, to address the potential endogeneity issues, this paper employs 2SLS regression using IV. Building on the “rice theory” proposed by Talhelm et al. (2014), which posits that agricultural production methods can influence social organization and culture in different regions, we use the ratio of rice-to-wheat planting areas in each city in the year 2000 as an IV for income polarization. Specifically, rice cultivation is more labor-intensive than wheat cultivation and requires greater cooperation among residents and families to undertake complex projects such as repairing irrigation systems, thereby emphasizing collaboration. In contrast, wheat farming involves less labor and fewer tasks that require collective effort.

Regarding the validity of IV, we provide explanations from two aspects: relevance and exogeneity. Firstly, agricultural practices may have a profound impact on the culture and ideological concepts of a region. Talhelm et al. (2014) found that the cooperation and coordination required for rice cultivation give rise to a “collectivist” cultural mindset among local residents, while individualism is characterized by self-centered values; and that the former emphasizes group affiliation and social norms of mutual assistance among individuals. It is manifested in cultural traits such as focusing on collective goals, desiring consistency with others, and aiming to maintain harmonious relationships (Hofstede 2001; Triandis 1995; Twenge et al. 2010).

Existing research indicates that the cooperative spirit derived from rice cultivation not only enhances social trust but also promotes mutual assistance behavior among individuals, such as financial lending, caring for each other's family members, and support in job searching and housing construction (Butler and Fehr 2024). Butler and Fehr (2024) further discovered that a 10-percentage-point increase in the proportion of land used for rice cultivation is associated with a 2.1-percentage-point reduction in income loss during natural disasters. Additionally, rice cultivation can also facilitate the provision of public goods (Zhou et al. 2023; Butler and Fehr 2024). It shows that the collectivist mindset derived from rice cultivation not only prevents individuals from excessively pursuing personal profit maximization but also improves local income distribution through promoting mutual assistance and increasing public goods. Therefore, there is a close theoretical link between the relative planting area of rice and wheat and income polarization.

Secondly, regarding the exogeneity, on the one hand, the ratio of rice-to-wheat cultivation is directly determined by the region's natural endowments which are relatively exogenous. On the other hand, historical data are unlikely to be directly correlated with current crime occurrences. Furthermore, to avoid potential confounding effects of climatic factors reflected by the relative planting ratio of rice and wheat on the IV regression, we have additionally controlled for each city's annual average temperature, annual cumulative precipitation, and annual cumulative sunshine duration.

The variable of the rice and wheat planting areas in 2000 is derived from the 1 × 1 km gridded rice and wheat planting area data provided by Luo et al. (2020) for China. By clipping and aggregating the grid data, we summarize the planting areas of rice and wheat to the city level, thereby calculating the ratio of rice-to-wheat planting areas at the city level. Compared with existing literature, this data set offers higher accuracy in crop classification. To construct a time-varying IV, this paper adopts a matching approach. Specifically, for each city in a given year, we calculate the average income polarization value of other cities whose GDP falls within 75% to 125% of the focal city's GDP in that year (excluding the focal city itself). Our IV is the product of this average value of income polarization and the rice-to-wheat planting area ratio.

5 Empirical Results

This section presents the empirical results of the effect of changes in income polarization on criminal activities. We first report the baseline results, followed by 2SLS regression using the rice-to-wheat planting area ratio in 2000 as IV and a series of robustness checks. Finally, we conduct heterogeneity analyses based on crime types and city characteristics.

5.3 Robustness Checks

5.3.1 DER Index With Alternative Values of α

Duclos et al. (2004) pointed out that the size of the DER index varies with different values of the parameter α. The larger the value of α, the greater the contribution of identification. Thus, the relative degree of income polarization in each city may also differ. Therefore, we conduct sensitivity tests by setting the value of α to 0, 0.5, 0.75, and 1, respectively. Table 5 shows that, regardless of the value of α, income polarization has a significantly positive impact on the crime rates. Consistent with the results of Li et al. (2019), as the value of α increases, the coefficient of the income polarization index also increases. In columns (1) and (5), the effect of the Gini coefficient (i.e., DER (α = 0)) on crime is significant, but the coefficients are smaller than those in the baseline results (Table 3) and other columns. This indicates that income polarization plays a more important role in explaining criminal activities compared to the Gini coefficient (Li et al. 2019).

Table 5. Sensitivity test-alternative values of $$\alpha $$.

Dep. Var.:	Log Casesct				Log Criminalsct
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
$${{DER}}_{{ct}}(\alpha =0)$$	0.603*				0.574*
	(0.308)				(0.302)
$${{DER}}_{{ct}}(\alpha =0.50)$$		1.764**				1.730**
		(0.792)				(0.788)
$${{DER}}_{{ct}}(\alpha =0.75)$$			1.845*				1.821*
			(0.953)				(0.950)
$${{DER}}_{{ct}}(\alpha =1)$$				2.062*				2.015*
				(1.062)				(1.061)
City FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province × Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Number of observations	497	497	497	497	497	497	497	497
Adjusted $${R}^{2}$$	0.799	0.799	0.799	0.799	0.792	0.793	0.792	0.792

• Note: Table 5 reports the results of the sensitive test. The dependent variable is the natural logarithm of the number of total criminal cases and criminals per 10,000. Standard errors clustered at the city level are reported in parentheses.
• * p < 0.1,
• ** p < 0.05, ***p < 0.01.

5.3.2 Different Measures of Crime Timing

In the baseline regressions, we use judgment dates to determine crime timing. However, since court proceedings often involve lags of several months to years, this approach may introduce measurement error. Therefore, we have recalculated the annual number of crime cases and the number of criminals at the city level, based on the recorded crime occurrence time in the judicial documents. The results are presented in columns (1) and (2) of Table 6, which are consistent with the results of the baseline regressions.

Table 6. Additional robustness check.

Dep. Var.	Different measures of crime dates		Drop first-tier cities		Base year controls × year FE
	(1)	(2)	(3)	(4)	(5)	(6)
	Log Casesct	Log Criminalsct	Log Casesct	Log Criminalsct	Log Casesct	Log Criminalsct
$${{DER}}_{{ct}}(\alpha =0.25)$$	1.173**	1.142**	1.171**	1.144**	1.131**	1.027*
	(0.546)	(0.541)	(0.540)	(0.534)	(0.537)	(0.535)
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Province × Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Number of observations	497	497	485	485	474	474
Adjusted $${R}^{2}$$	0.853	0.844	0.805	0.798	0.830	0.825

• Note: Table 6 reports results of additional robustness check. The dependent variable is the natural logarithm of the number of total criminal cases and criminals per 10,000. Standard errors clustered at the city level are reported in parentheses.
• * p < 0.1,
• ** p < 0.05, ***p < 0.01.

5.3.3 Permutation Test

To further confirm that the estimation results are indeed due to income polarization rather than other city-level characteristics, this paper conducts a permutation test. Specifically, we randomly assign all DER indices to each city based on Equations (2) and (3) to obtain the estimated coefficients of income polarization in each estimation. By repeating this random process 1,000 times, the estimated coefficients obtained from the permutation test are shown in Figure 4 below, which clearly deviate from the estimates in the baseline regressions.6 This indicates that the findings of this paper indeed come from the impact of income polarization, rather than other city-level characteristics.

5.3.4 Exclusion of First-Tier Cities

Beijing, Shanghai, Tianjin and Chongqing are municipalities directly under the central government. Compared with other cities, they have significant differences in political attention, level of economic development, size of migrants, degree of population agglomeration, and public security. To avoid the influence of the particularity of first-tier cities on the regression, Table 6 excludes them from the sample. Results in columns (3) and (4) are in line with the baseline results.

5.3.5 Considering “Bad Control” Problem

The correlation between control variables and income polarization may lead to the “Bad Control” problem. Therefore, we replace the control variables with the interaction terms of the values of each city's control variables in 2014 and year fixed effects. The results remain robust.

6 Mechanisms

The preceding sections have explained the causal relationship between income polarization and crime. This section delves further into the underlying mechanisms of this relationship. Following Duclos et al. (2004), the DER index can be decomposed into the product of identification, alienation, and the correlation between the two. Therefore, we further decompose Equation (1).

Therefore, an increase in either the identification within income groups or the alienation between income groups will lead to a rise in the degree of income polarization.

We further explore the theoretical linkages between alienation, identification, and crime. The relationship between alienation and crime can be explained through the following two mechanisms. Firstly, according to Becker (1968)'s classic economic analysis of crime, criminal behavior is the outcome of a cost-benefit calculation. An increase in alienation, manifested as a widening of income gaps between income groups, enhances the economic benefits for lower-income groups to commit crimes against higher-income groups, which in turn raises the opportunity cost of legitimate employment for individuals, thereby prompting some to shift from lawful work to criminal activities. Moreover, alienation is also highly correlated with social exclusion. A substantial body of theoretical and empirical research indicates that socially excluded individuals become vulnerable to radicalism (Moghaddam 2005; Kowalski et al. 2021; Leary et al. 2003; Dugan and Chenoweth 2020), which may subsequently inspire radical actions. Secondly, within-group identification is highly correlated with social identity. Sen (2006) posited that within-group identification fosters exclusivity among group members towards outsiders, thereby exacerbating intergroup conflicts such as criminal activities.

The DER index can be decomposed into alienation and identification (Duclos et al. 2004). Table 9 shows that between-group alienation significantly promotes the occurrence of criminal activities, while the impact of within-group identification is not significant. This indicates that the social confrontation brought about by the increase in income polarization is the key reason for the increase in crime rates.

We then further elaborate on the above mechanism from the perspective of the willingness to work. Firstly, when an increase in income polarization leads to a higher relative payoff from crime, individuals may shift from choosing legitimate employment to engaging in criminal activities. In this case, we should observe that income polarization has a dampening effect on the employment intentions among the unemployed. Secondly, if people are chronically exposed to social exclusion, they may even experience depression, helplessness, and a sense of unworthiness (Riva et al. 2017; Williams 2009). Although it is challenging to directly observe individuals' decision-making between legitimate labor and illegal activities, we can apply the individual-level data from the CHFS to examine the impact of income polarization on the unemployed individuals' job-seeking willingness and their subjective happiness.

Specifically, we match the city-level income polarization with individual-level data and conduct an empirical analysis using the unemployed individuals' responses to the question “Did you look for a job in the past month?” in the CHFS survey, as well as their self-rated scores of subjective happiness. Table 10 shows that the coefficients of income polarization on residents' job-seeking intentions and subjective happiness are both negative. This indicates that after an increase in income polarization, the unemployed are less willing to look for a job, and the level of residents' subjective happiness also decreases, which may be contributing factors to the rise in the crime rates.

7 Conclusion

Despite the long-standing theoretical view that polarization promotes social conflict, the existing empirical literature has predominantly focused on the impact of income inequality or poverty on crime. Using data from the CHFS and judicial documents from 2014 to 2018, this paper constructs city-level indicators of income polarization and criminal activity to empirically examine their relationship. To address the endogeneity issue, this paper constructs an IV for income polarization using the rice-to-wheat planting area ratio in 2000 in different cities, providing causal evidence of the relationship between the two.

The results show that an average increase of one-standard-deviation in income polarization will lead to an increase of 4.71% and 4.58% in the number of criminal cases and the number of criminals per 10,000 people, respectively. This corresponds to an estimated 0.39 criminal cases and 0.42 criminals per 10,000 people. The results still hold after a series of robustness checks.

Furthermore, this study also examines the impact of income polarization on different types of crime based on the rich textual content of judicial documents. The results show that rising income polarization leads to an increase in violent crime, theft, drug-related crime and financial fraud. In addition, heterogeneity analyses indicate that the exacerbating effect of income polarization on crime is more pronounced in cities with a higher proportion of young and migrant populations.

Finally, the mechanism analyses reveal that alienation significantly promotes criminal activities, while the impact of identification is not significant. Further analysis indicates that increasing alienation significantly diminishes the employment intention of the unemployed and the individual subjective happiness, thereby increasing the likelihood of criminal behavior.

Based on our finding, the policy implications are as follows: First, the government should focus on income distribution issues and avoid a structure where the income distribution is characterized by “large at both ends and small in the middle”. Second, it is important to provide employment opportunities and skill training for low-income groups to enhance the job-seeking intentions among the unemployed. Third, community-based mental health services and well-being programs could be established to provide psychological support for the unemployed. Finally, inclusive social policies should be implemented, including anti-discrimination laws and initiatives to reduce social exclusion.

Author Contributions

Jingqi Liu: data curation, formal analysis, investigation, methodology, software, validation, visualization, resources, writing – original draft, writing – review and editing. Chen Wang: conceptualization, methodology, funding acquisition, investigation, project administration, resources, supervision, validation, writing – original draft, writing – review and editing. Yuzhou Wang: data curation, formal analysis, investigation, visualization, methodology, software, writing – original draft.

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.