Does COVID-19 Exacerbate Regional Income Inequality? Evidence from 20 Provinces of China

Abstract

The COVID-19 pandemic has had a profound impact on almost all aspects of society and the world’s economy. This study aimed to examine the impact of COVID-19 on regional income inequality in China. Based on provincial data from 20 provinces (covering 218 prefecture-level cities) for the period from 2013 to 2022, the study revealed the significant impact of COVID-19 on widening regional income inequality, using a continuous difference-in-differences (DID) model. The results were robust when applying a parallel trend test, lagging control variables, and alternative measures of regional income inequality to test the results. Additional analysis suggested that the adverse impact of COVID-19 on regional income inequality was increased by regions’ foreign trade dependence and their share in the service industry but may have been weakened by the development of the digital economy. The findings highlight the adverse effects of COVID-19 on regional income inequality.

1. Introduction

The COVID-19 pandemic has had a huge impact on China, as well as on the rest of the world, affecting almost every aspect of society and the world’s economy. The pandemic and its consequences have been a global concern for the past three years. As the virus spread widely across the world, social distancing policies and lockdowns, along with other corresponding policies, were implemented by governments in different countries and regions to control the pandemic. Consequently, the economies of these countries and regions were deeply impacted.

China’s COVID-19 control measures, known as the zero-COVID policy, were more stringent than those of other countries. Specifically, grid governance measures, including city lockdowns, household surveys, and quarantines of residents, were implemented at the grassroots level of society [1]. The zero-COVID policy had two main impacts. In the short term, it restricted human mobility between regions, enabling quick control of the epidemic and reducing the spread of the virus. This facilitated a faster return to normal production, which positively affected the economy. However, in the long run, the continuing strict control measures disrupted the normal order of production and life, thereby adversely affecting the economy [2,3].

Beginning in December 2022, the zero-COVID policy was no longer in effect. On 23 February 2023, the National Health Commission of China announced that the spread of the COVID-19 pandemic had ended in China (China Daily, 24 February 2023: https://www.chinadaily.com.cn/a/202302/24/WS63f7f724a31057c47ebb08c8.html, accessed on 21 March 2023). While the impact of COVID-19 has diminished in China, its effects continue. Over the past three years, significant economic losses have been observed, with China’s average annual GDP growth rate falling from 6.9% before the pandemic period (2012–2019) to 4.5% during the pandemic period (2019–2022). Moreover, the COVID-19 pandemic has had far-reaching consequences on various aspects of people’s lives, including employment, income, education, health, and social welfare [4,5,6].

Among the various impacts of the COVID-19 pandemic, one worth noting is resulting inequalities. Inequalities include income inequalities, education inequalities, gender inequalities, and digital inequalities [7,8,9,10]. Existing studies showed that the COVID-19 pandemic exposed and exacerbated pre-existing inequalities and created new disparities. It affected different groups of people in various ways [8,11,12,13]. In fact, in 2021, the poorest 20% of the world’s population experienced a steeper drop in income than that of high-income people [14]. Income inequality is closely linked to the incidence of poverty. Suryahadi et al. [15] estimated the impact of the COVID-19 outbreak on poverty in Indonesia and found that COIVID-19 significantly reduced the economic growth rate and increased the poverty rate.

Therefore, understanding the ways in which COVID-19 has affected different dimensions of inequality and developing policy interventions to address such inequalities is crucial.

Despite a wealth of literature contributing significantly to the research on the impact of COVID-19 on socioeconomic inequality, most studies were conducted with a focus on enterprises, industries, gender, age, etc., with relatively few studies on the impact of COVID-19 on regional income inequality [6,9,10,16,17,18,19].

In China, income inequality exists mainly in the form of regional inequality, urban–rural inequality, and inequality among different groups of residents. Of these three aspects of inequality, income inequalities between urban and rural residents and among different groups of residents have received the most research attention. Research by Shen et al. [20] showed that the COVID-19 pandemic significantly impacted rural incomes, leading to a further widening in the urban–rural income inequality in China. Qian and Fan [11] suggested that longstanding status markers in China, such as education, family economic status, Communist Party membership, state-sector employment, and urban hukou registration, mitigated the adverse effects of the COVID-19 outbreak on individuals’ income losses.

In contrast, few studies have examined the impact of COVID-19 on income inequality at the regional level. Nevertheless, the issue of unbalanced regional development in China has persisted, without significant resolution. In addition, China’s unique and stringent pandemic prevention and control measures had both short-term positive and long-term negative effects on local economies. These measures complicated the understanding of the impact of the pandemic on regional income disparities. Therefore, this study aimed to examine the impact of COVID-19 on regional income inequality in China.

This study makes two contributions. First, it provides new empirical evidence about the impact of COVID-19 on regional income inequality. The findings show that COVID-19 had a significant impact on widening regional income inequality in China, which aligns with existing studies [11]. However, in contrast to previous studies, our research results also suggest that while the cumulative impact of COVID-19 widened the regional income inequality of both urban residents and rural residents within the provinces, it narrowed the income gap between urban and rural residents within the provinces. Second, we analyzed factors that may have moderated the impact of COVID-19. The results suggest that the impact of COVID-19 in widening regional income inequality was affected by regions’ foreign trade dependence and their shares in the service industry, but that impact may have been offset by the development of the digital economy. These findings may help us better understand the impact of the COVID-19 pandemic on income inequality in different provinces and, accordingly, help policymakers in formulating policies.

2. Literature Review

2.1. Income Inequality Studies

Over the years, income inequality has consistently captured significant attention due to its strong correlation with economic growth and other related concerns. Given the paramount significance of this issue, extensive research has been conducted to investigate various facets of income inequality, such as its causes and its intricate association with economic growth and poverty [21,22,23,24,25,26]. Cumulative evidence unequivocally demonstrates that elevated levels of inequality not only impede prospects for economic growth but also pose formidable obstacles to poverty alleviation endeavors [21].

Income inequality is influenced by a myriad of factors, with the significance of each factor varying. To shed light on this issue, Li et al. [22] conducted a comprehensive analysis based on data from 625 county-level administrative units in China in 2017. Their findings showed that commerce, population footprint, industrialization, and investment are the main factors influencing region’s income. Additionally, globalization and foreign direct investment (FDI) were recognized as crucial determinants of regional income inequality. Prior research has indicated that globalization and FDI tend to exacerbate regional income disparities. However, recent trends suggest that domestic capital has emerged as the primary driver of regional inequality [23,25].

2.2. The Economic Impact of COVID-19

Existing studies examined various aspects of the economic impact of COVID-19, including the impact on production, employment, income, poverty, exports and imports, FDI, and overall economic growth [5,15,27,28,29,30,31]. Among the macroeconomic consequences of COVID-19, the most prominent and widely studied aspect pertains to the impact on economic growth. Numerous studies showed that COVID-19 had a significant negative impact on economic growth [5,15,32,33]. Using data for the last quarter of 2017 to the third quarter of 2020, Martinho [32] found that the COVID-19 pandemic eliminated the convergence of GDP per capita in OECD countries. Additionally, Ghecham [33] suggested that the decline in GDP growth during the first year of the pandemic was larger in countries with a higher reliance on services and in countries with more restrictive measures (lockdowns). In addition to the study of the impact of COVID-19 on economic growth, Suryahadi et al. [15] suggested that the pandemic would lead to a significant increase in the poverty rate.

In addition to the aforementioned studies, several scholars have conducted research on the influence of COVID-19 on international trade. For instance, Gruszczynski [34] investigated the short-term and long-term effects of the pandemic on international trade, suggesting that while the short-term consequences were severe, the adverse impact may diminish over time. Correspondingly, Hayakawa and Mukunoki [35] observed substantial negative effects of COVID-19 on international trade for both exporting and importing countries. In general, existing research indicated a substantial short-term negative impact of the pandemic on international trade.

2.3. The Impact of COVID-19 on Income Inequality

Since the outbreak of the COVID-19 pandemic, numerous studies have explored its impact on income inequality across different countries and regions. For instance, Deaton [36] investigated the effect of the pandemic on global income inequality among countries. The results indicated a decrease in international income inequality when considering the higher number of deaths in rich countries per capita. However, it is important to note that this finding may be less convincing given the limitations of the research timeframe. Conversely, when countries were weighted by population, international income inequality increased [36]. Delaporte et al. [37], on the other hand, examined the impact of major epidemics over the past two decades on income distribution. They suggested that the scale of COVID-19, compared to other pandemics, would lead to an increase in the Gini coefficient and raise the income share of higher-income deciles. Similarly, Prakash et al. [38] provided a prognosis based on an analysis of the effects of five previous major epidemics in the century. Their findings revealed significant and persistent reductions in disposable income, along with increases in unemployment, income inequality, and public debt-to-GDP ratios. In contrast to the aforementioned studies, Wildman [39] investigated the adverse effect of income inequality on the number of COVID-19 deaths in OECD countries. The results demonstrated that countries with high levels of income inequality tended to perform worse in terms of cases and deaths in their response to COVID-19.

In addition to cross-country studies, numerous research endeavors have focused on examining the impact of the COVID-19 pandemic on income inequality at the country and regional levels. For instance, Blundell et al. [19] investigated the potential effects of the pandemic on inequalities in the UK. Their results revealed a significant rise in income disparities between wealthier and poorer households as a consequence of the pandemic. Turning to China, Shen et al. [20] adopted the distribution dynamics approach to study the impact of COVID-19 and social distancing policies on regional income inequality. Their analysis was based on a sample of 295 prefecture-level cities in 31 provinces. Their findings showed that regional income inequality intensified in cities with prolonged durations of stringent social distancing policies during the pandemic. Conversely, in cities with shorter policy durations, the impact on income inequality diminished. Moreover, the study highlighted that rural residents experienced a more persistent impact on their incomes compared to urban residents. Similarly, Zhang et al. [40] confirmed that the pandemic exacerbated income inequality in China, particularly impacting the per capita income of rural households.

Apart from the above studies, researchers also examined the impact of the pandemic on income inequality from a micro perspective. For instance, Qian and Fan [11] conducted an investigation into the individual-level economic consequences of the COVID-19 outbreak in Mainland China. Drawing on data collected from the region, their study showed that education, family economic status, Communist Party membership, state-sector employment, and urban hukou registration—all long-standing status markers in China—mitigated the adverse effects of the pandemic on individuals’ income losses. Furthermore, individuals residing in households or areas disproportionately affected by the pandemic were at a higher risk of experiencing income reductions. Zhao [41] examined the impact of the pandemic on the income gap between capital-intensive and labor-intensive firms in China, finding that the negative impact of the pandemic on the income of employed residents in labor-intensive enterprises was greater than that in capital-intensive enterprises; thus, widening the income gap among employed residents in heterogenous enterprises. Based on high-frequency labor surveys conducted in Asian countries, Jurzik et al. [42] showed that the COVID-19 pandemic had increased the inequality, as job losses had disproportionately affected low-income workers, women, and youth. Moreover, their study further showed that the pandemic not only exacerbated pre-existing social inequalities but also created new forms of disparities.

In addition to studies on the direct impact of the pandemic on income inequality, there is a large body of research focusing on gender inequality as the impact of the pandemic on gender disparities varies in term of income and employment outcomes [10,17]. Moreover, income inequality caused by working from home (WFH) during the pandemic has also been a topic of interest, with several studies indicating that WFH will increase income inequality between different groups [43,44].

In summary, existing research suggested that the pandemic generally led to an increase in income inequality at various levels, including countries, regions, industries, and demographic groups. However, most studies primarily focused on the early stages of the outbreak, neglecting the long-term effects. Additionally, previous research has not provided a comprehensive explanation for the underlying mechanisms that moderate the relationship between the pandemic and regional income inequality. To address these gaps, our study utilized a provincial panel dataset covering the period from 2013 to 2022 to empirically verify the impact of COVID-19 on regional income inequality. Further analysis demonstrated that this impact was moderated by the degree of dependence on foreign trade, the share of services, and the level of digital economy development.

3. Variables and Data

3.1. Sample and Data Source

The sample of this study contains data from 20 provincial regions from 2013 to 2022 in Mainland China (there are 337 prefecture-level (and above) cities in 31 provinces in China (excluding Hong Kong, Macau, and Taiwan). In this study, the sample covers 218 prefecture-level cities from 20 provinces. Some cities are not included in the sample as the data are not available). To be specific, the data used in this article come from three sources. Macroeconomic data at the provincial level come from the National Bureau of Statistics of China, including per capita GDP, population, value added of the secondary industry and value added of the tertiary industry (National Bureau of Statistics of China: https://data.stats.gov.cn/easyquery.htm?cn=E0103, accessed on 3 March 2023). The prefecture-level data used to calculate income inequality within the province come from the CEIC database (CEIC database: https://info.ceicdata.com/zh-hans/ceic-database-baidu-ads_china_premium, accessed on 3 March 2023). The raw prefecture-level city data of each province include per capita disposable income, per capita disposable income of urban residents, per capita disposable income of rural residents, aggregate population, urban population and rural population. Additionally, the data on confirmed COVID-19 cases by province are sourced from the Social Science Data Integrated Service Platform (Social Science Data Integrated Service Platform: http://www.ppmandata.cn, accessed on 18 February 2023).

3.2. Variables Specification

3.2.1. Measure of Regional Income Inequality

There are several approaches to measure income inequality, including the Gini coefficient, the Theil index, the coefficient of variation, the mean-log deviation (MLD), and others [21,36]. In this study, we adopted the Gini coefficient as the main way to measure regional income inequality. To be specific, the Gini coefficient of each province was calculated from the per capita disposable income and the corresponding population of each prefecture-level city within the province, including the overall Gini coefficient (AGini), the income Gini coefficient of urban residents (UGini), and the income Gini coefficient of rural residents (RGini). Furthermore, we employed additional measurement indicators to test the robustness of our results, providing supplementary evidence.

3.2.2. Explanatory Variable

The primary objective of this study is to examine the impact of the COVID-19 pandemic on regional income inequality. However, it is essential to acknowledge that the pandemic has affected all provinces to varying degrees over the course of several years. Therefore, it is not appropriate to analyze the impact of COVID-19 solely by creating a dummy variable and assigning a value of 1 to provinces where the outbreak occurred. In recognition of the fact that the impact of COVID-19 differs in intensity across regions, this study utilizes the number of confirmed COVID-19 cases in different regions as a proxy variable to measure the strength of the pandemic’s impact. This approach acknowledges that the impact of COVID-19 is not simply a binary variable (presence or absence), but rather varies in magnitude across different regions. This approach offers two distinct advantages. Firstly, it enables us to determine whether a province has experienced an outbreak of COVID-19. Secondly, it allows us to differentiate the severity of the outbreak across different regions, thereby providing a more accurate assessment of the impact on each province. By considering both the occurrence and intensity of the outbreak, we can better capture the effect of the pandemic on regional income inequality. This is because the response of different local governments to control the outbreak are roughly related to the number of confirmed cases in their area (according to Shen et al. [20], the duration and extent to which local governments implemented strict social distancing policies following the Emergency Response Situation (ERS) varied with the number of confirmed cases, and the intensity of the policy was roughly proportional to the number of confirmed cases). Moreover, because the pandemic has raged for years, its impact is likely to be cumulative and persistent. For this reason, we used two variables: one is the annual number of confirmed cases (ConfirmYear) and the other is the total number of confirmed cases (ConfirmTot) in each province.

3.2.3. Control Variables

This study includes several control variables that are likely to influence regional income inequality. The study controls for GDP per capita (RGDP) as it measures the level of regional economic development, which is routinely employed as a determinant of income inequality [45,46,47]. The population (POP) is also included as a control variable as it measures the size of a region, which may also affect income inequality [48]. Additionally, according to previous studies, the industrialization process may also affect income inequality [45,49]. Therefore, we take the share of the secondary industry outputs in provincial GDP (INDR) as a control variable to control for its potential effect. GDP per capita (RGDP) and population (POP) are treated into log forms.

Table 1. Variables Selection.

	Variables	Definition	Operation
Dependent Variables	AGini	Overall Gini coefficient
	UGini	Income Gini coefficient of urban residents
	RGini	Income Gini coefficient of rural residents
Explanatory Variables	ConfirmYear	Annual number of confirmed cases
	ConfirmTot	Total number of confirmed cases
Control Variables	RGDP	GDP per capita	Log
	POP	Population	Log
	INDR	The share of the secondary industry outputs in GDP

briefly presents the selection of dependent variables, explanatory variables, and control variables.

3.3. Descriptive Statistics of Variables

Table 2. Descriptive Statistics.

Variables	Observations	Mean	Std. Dev.	Min	Max
AGini	140	0.1274	0.0468	0.0019	0.2232
UGini	163	0.0759	0.0421	0.0025	0.1777
RGini	157	0.0859	0.0354	0.0127	0.1762
RGDP	250	56,108	21,882	22,089	144,390
POP	250	5003	3036	316	12,670
INDR	250	0.4077	0.0662	0.1908	0.5576
SERR	250	0.4880	0.0474	0.3466	0.6163
ConfirmYear	71	3241	9548	0	59,491
ConfirmTot	71	4117	10,612	0	61,804

summarizes the statistical characteristics of the data. The mean values of AGini, UGini, and RGini are, respectively, 0.1274, 0.0759, and 0.0859, indicating that the income inequality between urban and rural residents is more significant than the income inequality within urban residents and rural residents. The lowest number of confirmed COVID-19 cases in 2020 and 2021 was 0, including Qinghai (2020) and Tibet (2020, 2021). The maximum number of confirmed COVID-19 cases in 2020, 2021, and 2022 was 38,193 (Hubei), 1591 (Shaanxi), and 59,491 (Guangdong), respectively. Pearson’s correlation matrix, presented in

Table 3. Correlation Matrix.

	AGini	UGini	RGini	RGDP	POP	INDR	SERR
AGini	1	0.754 ***	0.618 ***	0.242 ***	0.313 ***	0.211 **	−0.146 *
UGini	0.800 ***	1	0.694 ***	0.435 ***	0.604 ***	0.241 ***	−0.030
RGini	0.590 ***	0.688 ***	1	0.060	0.470 ***	0.214 **	−0.031
RGDP	0.113	0.387 ***	0.122	1	0.243 ***	0.105	0.340 ***
POP	0.336 ***	0.642 ***	0.526 ***	0.341 ***	1	0.094	0.112
INDR	0.410 ***	0.344 ***	0.422 ***	0.221 ***	0.321 ***	1	−0.742 ***
SERR	−0.198 **	−0.006	−0.151 *	0.233 ***	−0.077	−0.705 ***	1

Notes: Superscript ***, ** and * represent statistical significance at the 1%, 5%, and 10% levels, respectively.

, shows a positive and statistically significant correlation between AGini, UGini, and RGini, indicating that the overall income inequality, the income inequality of urban residents, and the income inequality of rural residents are positively correlated. The multicollinearity issue seems to be very unlikely in the model, as correlation coefficients between other variables are less than 0.80, with Gujarati [50] stating that only values exceeding 0.80 are likely to indicate multicollinearity.

4. Results and Discussion

4.1. Empirical Model

As a purely exogenous event, the COVID-19 outbreak has resulted in two dimensions of differences in regional income inequality. Firstly, from a temporal perspective, the outbreak of COVID-19 may have a time-dependent impact on income inequality within the same region. Secondly, from a spatial perspective, the impact of COVID-19 varies between different provinces and individuals. This aligns with the conditions required for applying the difference-in-differences (DID) method, considering that the COVID-19 epidemic has affected all provinces in China, albeit to varying degrees. Different from a standard DID model, following Qian [51], we adopt a model in which the disposal variable is treated as a continuous variable. To examine the impact of the COVID-19 pandemic on regional income inequality at the province level, we present the identification model in Equation (1) below.

$$Y_{it}={\alpha }_0+{\alpha }_1{C}_{it}\ast { \mathrm{post} }_{2020}+{\alpha }_3{X}_{it}+{\delta }_i+{\theta }_t+{\epsilon }_{it}$$

(1)

In the above equation, the subscript i denotes the province, and t denotes the year.$Y_{it}$ is the dependent variable representing regional income inequality, which is represented by AGini, UGini, and RGini, respectively. $C_{it}$ represents the number of confirmed COVID-19 cases. The variable ${\mathrm{post} }_{2020}$ is a time dummy variable, indicating whether the year is 2020 or later. If the year is after 2019, ${ \mathrm{post} }_{2020}=1$; otherwise, ${ \mathrm{post} }_{2020}=0$. $X_{it}$ represents a series of control variables, including GDP per capita (RGDP), population (POP), and the share of the secondary industry outputs in GDP(INDR). ${\delta }_i$ and ${\theta }_t$ refer to province fixed effect and time fixed effect, respectively. ${\epsilon }_{it}$ is the random error term. In Equation (1), the coefficient of interest is $C_{it}{\ast \mathrm{post} }_{2020}$, denoted as ${\alpha }_1$. If ${\alpha }_1$ is significantly positive, it indicates that COVID-19 has increased income inequality, and if ${\alpha }_1$ is significantly negative, it indicates that COVID-19 has reduced income inequality.

4.2. Regression Results

Table 4. Regression results of the COVID-19 (Confirm Year) on regional income inequality.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	AGini	AGini	UGini	UGini	RGini	RGini
Shockpost1	0.0194 **	0.0217 ***	0.0008 ***	0.0008 ***	0.0006 ***	0.0006 **
	(2.54)	(2.98)	(6.34)	(4.18)	(5.14)	(2.73)
LRGDP		0.1047 **		−0.0637		−0.0296
		(2.30)		(−0.70)		(−0.28)
LPOP		-0.0183		−0.2719		−0.1972
		(−0.29)		(−1.47)		(−0.89)
INDR		−0.2212 ***		0.2682		0.2093
		(−3.01)		(1.48)		(0.92)
Constant	0.1340 ***	−0.7256	0.0720 ***	2.9318	0.0847 ***	1.9774
	(39.88)	(−0.98)	(12.84)	(1.35)	(10.21)	(0.80)
Province FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Observations	140	140	163	163	157	157
R-squared	0.552	0.626	0.188	0.294	0.074	0.140
Number of pid	20	20	20	20	20	20

Notes: Superscript *** and ** represent statistical significance at the 1% and 5% levels, respectively. Robust t-statistics in parentheses.

reports the regression results for Equation (1). Table 4, Models (1), (3), and (5) present the baseline regression results between the COVID-19 shock (Shockpost1) and income inequality (AGini, UGini, and RGini), excluding control variables. In the baseline models, we do not include control variables as we consider the pandemic outbreak to be exogenous to the economy. Model (1) shows that the COVID-19 pandemic is positively and significantly related to the overall level of regional income inequality (α = 0.0194, p < 0.05). Models (3) and (5) show that the pandemic also has a significantly positive effect on the regional income inequality of both urban residents (α = 0.0008, p < 0.01) and rural residents (α = 0.0006, p < 0.01). Models (2), (4), and (6) present the regression results that include the variable of interest, COVID-19 shock (Shockpost1), as well as control variables. The coefficients of Shockpost1 are positive and statistically significant in Model (2) (α = 0.0217, p < 0.01), Model (4) (α = 0.0008, p < 0.01), and Model (6) (α = 0.0006, p < 0.05), indicating that the COVID-19 shock is positively related to regional income inequality from the perspective of all residents, urban residents, and rural residents. These findings clearly indicate that regions more affected by the pandemic have experienced an increase in regional income inequality.

Regarding control variables in Model (2), the coefficients of LRGDP are positive and statistically significant. This indicates that, in terms of the overall income inequality, provinces with higher level of economic development also have higher levels of income inequality. However, the coefficients of LRGDP are negative for Models (4) and (5), but the results are not statistically significant. The coefficients of LPOP are negative but not statistically significant for Models (2), (4) and (6). As for INDR, the coefficient is negative and statistically significant. This suggests that provinces with a higher level of industrialization tend to experience lower income inequality at the overall level. However, in Models (4) and (6), the coefficients of INDR are not statistically significant, despite having positive values.

The results in Table 4 show that the pandemic has a significant positive impact on regional income inequality. However, the results in Table 4 may still underestimate the impact of the pandemic on regional income inequality, given that the impact of the pandemic can be persistent and lagging. In view of this, we change the core explanatory variable from the annual number of confirmed cases (ConfirmYear) to the cumulative number of confirmed cases (ConfirmTot) to examine the potential cumulative effects of COVID-19 on regional income inequality.

Table 5. Regression results of the COVID-19 (Confirm Tot) on regional income inequality.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	AGini	AGini	UGini	UGini	RGini	RGini
Shockpost2	0.0172 **	0.0195 ***	0.0011 ***	0.0011 ***	0.0014 ***	0.0015 ***
	(2.67)	(3.15)	(10.44)	(7.51)	(13.46)	(9.04)
LRGDP		0.1080 **		−0.0673		−0.0336
		(2.42)		(−0.76)		(−0.33)
LPOP		−0.0220		−0.2225		−0.1166
		(−0.38)		(−1.28)		(−0.60)
INDR		−0.2180 ***		0.3199 *		0.2962
		(−3.14)		(1.85)		(1.44)
Constant	0.1341 ***	−0.7308	0.0720 ***	2.5260	0.0851 ***	1.2946
	(40.78)	(−1.06)	(13.23)	(1.16)	(10.73)	(0.53)
Province FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Observations	140	140	163	163	157	157
R-squared	0.560	0.635	0.231	0.337	0.172	0.245
Number of pid	20	20	20	20	20	20

Notes: Superscript ***, ** and * represent statistical significance at the 1%, 5%, and 10% levels, respectively. Robust t-statistics in parentheses.

presents the regression results of the shock of cumulative COVID-19 confirmed cases (Shockpost2) on regional income inequality. The coefficients of Shockpost2 are all positive and statistically significant in Models (1) (α = 0.0172, p < 0.05), (3) (α = 0.0011, p < 0.01), and (5) (α = 0.0014, p < 0.01). After including control variables, the coefficients of Shockpost2 are still statistically significant, with values of 0.0195, 0.0011, and 0.0015, respectively. The results further confirm the results in Table 4. Besides, the coefficients of Shockpost2 for Models (1) and (2) in Table 3 are smaller than that of Shockpost1 for Models (1) and (2), but the coefficients of Shockpost2 for Models (3), (4), (5), and (6) are larger than that in Table 4. These results may indicate that the cumulative impact of COVID-19 has widened the regional income inequality of both urban residents and rural residents within the province, but narrowed the income gap between urban and rural residents within the same region. The probable explanation is that the ongoing impact of COVID-19 on the incomes of urban residents is greater than that of rural residents.

4.3. Endogeneity Analysis

The outbreak of COVID-19 in various provinces can be viewed as an exogenous shock, so the impact of endogenous issues for our study is relatively weak. However, it is important to acknowledge that geographical and climatic variations across regions may still impact the level of COVID-19 outbreak, such as the number of confirmed cases. This introduces a potential endogeneity issue. Additionally, there may be some degree of inaccuracy in the reported number of confirmed COVID-19 cases in different regions, which could also affect the robustness of our results from a statistical standpoint. To solve the potential endogeneity problem that arises from these problems, we further process the data on confirmed COVID-19 cases in each region. Specifically, we round up, round down, and round off the number of confirmed COVID-19 cases (unit: thousand people) in each region to address the possible impact of geographical, climatic, and statistical errors on the differences in the number of confirmed COVID-19 cases in each province. In addition, the spread of COVID-19 in different regions may also be affected by geographical environment and cultural factors (such as people’s preference for offline social interaction). However, considering that although the number of confirmed COVID-19 cases in each province may not be accurate, the influence of other factors (such as geography and culture) is also limited. They can hardly influence the spread of COVID-19 to a large extent due to strict social distancing policy. In other words, the relative size of the number of confirmed COVID-19 cases in each province is almost unaffected. Therefore, we construct a new variable (Severity) to measure the impact of the COVID-19 pandemic by standardizing the number of confirmed COVID-19 cases.

Table 6. Regression results of the COVID-19 (treated Confirm Year) on regional income inequality.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Variables	AGini	AGini	AGini	AGini	AGini	AGini	AGini	AGini
CShockpost1	0.0173 *	0.0204 **
	(2.05)	(2.60)
RShockpost1			0.0153 **	0.0167 **
			(2.36)	(2.55)
FShockpost1					0.0188 **	0.0224 **
					(2.11)	(2.75)
Severity							0.0194 **	0.0217 ***
							(2.54)	(2.98)
LRGDP		0.0965 **		0.0889 *		0.0993 **		0.1047 **
		(2.13)		(1.98)		(2.16)		(2.30)
LPOP		−0.0017		−0.0241		−0.0058		−0.0183
		(−0.02)		(−0.32)		(−0.07)		(−0.29)
INDR		−0.2374 **		−0.2175 **		−0.2413 **		−0.2212 **
		(−2.71)		(−2.56)		(−2.80)		(−3.01)
Province FE	YES	YES	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES	YES	YES
Constant	0.1342 ***	−0.7707	0.1339 ***	−0.5105	0.1341 ***	−0.7642	0.1340 ***	−0.7256
	(38.73)	(−0.91)	(39.81)	(−0.62)	(38.72)	(−0.90)	(39.88)	(−0.98)
Observations	140	140	140	140	140	140	140	140
R-squared	0.491	0.563	0.530	0.593	0.496	0.571	0.552	0.626
Number of pid	20	20	20	20	20	20	20	20

Notes: Superscript ***, ** and * represent statistical significance at the 1%, 5%, and 10% levels, respectively. Robust t-statistics in parentheses.

reports the regression results of the number of confirmed COVID-19 cases on regional income inequality after rounding up (CShockpost1), rounding down (FShockpost1), rounding off (RShockpost1), and standardizing (Severity). The coefficients of CShockpost1 are positive and statistically significant, with values of 0.0173 (p < 0.1) in Model (1) and 0.0204 (p < 0.05) in Model (2), respectively. The coefficients of RShockpost1, FShockpost1, and Severity are positive and statistically significant as well. These results are consistent with the findings presented in Table 4 and Table 5, providing further confirmation that the COVID-19 pandemic has led to an increase in regional income inequality.

4.4. Validity and Robustness Checks

The above study has established a significant positive association between COVID-19 and regional income inequality. Next, we try to discuss the validity and robustness of the above results using various methods, including the parallel trend test, lagging control variables and applying alternative measures of region income inequality.

4.4.1. Parallel Trend Test

The assumption of prior parallel trend is the prerequisite for the effective application of the DID method. However, in contrast to the standard difference-in-differences (DID) model, our study takes into account the heterogeneity in the impact of the pandemic on provinces. This implies that the change in the regional (or individual) dimension is not a binary change from 0 to 1, but rather a continuous change. Therefore, there is no strict distinction between experimental and control groups; instead, all provinces are affected by the COVID-19 pandemic. In this case, we draw on the idea of the event approach. We first generate interaction terms between year dummy variables and pandemic impact variables, and then use these interaction terms as explanatory variables for regression. If the coefficient of interaction between year dummy variables and pandemic impact variables before the actual outbreak point is not significantly different from zero, it indicates that there is no time trend of heterogeneity among regions (provinces) before the outbreak of COVID-19.

Figure 1 shows the dynamic effects of changing the year of COVID-19 outbreak. As can be seen from Figure 1, before the outbreak of COVID-19, the coefficients of the interaction term are basically around 0 (the 95% confidence interval included a value of 0), indicating that there is no time trend of heterogeneity in income inequality across provinces before the outbreak of COVID-19, which supports the parallel trend assumption. After the outbreak of COVID-19, the coefficients of the interaction term show an obvious increase, especially in the year of 2022, the coefficient of the interaction term is significantly greater than 0.

4.4.2. Lagging Control Variables

In above study, we do not take into account the possible impact of COVID-19 on the control variables. In fact, the pandemic also has an impact on GDP per capita, regional population, and value added of secondary industry of the year. The impact of COVID-19 on GDP per capita and value added of secondary industry is clear because it affects regional economic growth broadly. At the same time, interregional mobility, population mortality, and infant birth rates are all subject to varying degrees of impact during the pandemic, subsequently affecting the population of a region. Therefore, the values of LRGDP, LPOP, and INDR after 2019 are lagged by one period to mitigate the concurrent effects of COVID-19 on these variables.

Table 7. Regression results of lagging control variables.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	AGini	UGini	RGini	AGini	UGini	RGini
Shockpost1	0.0203 **	0.0008 ***	0.0005 ***
	(2.73)	(4.22)	(3.82)
Shockpost2				0.0183 ***	0.0011 ***	0.0014 ***
				(2.92)	(8.63)	(10.64)
n_LRGDP	0.0996 **	−0.0728	−0.0388	0.1027 **	−0.0709	−0.0408
	(2.31)	(−0.85)	(−0.40)	(2.44)	(−0.84)	(−0.43)
n_LPOP	0.0092	−0.3317	−0.2391	0.0036	−0.2973	−0.1918
	(0.14)	(−1.48)	(−0.93)	(0.06)	(−1.36)	(−0.79)
n_INDR	−0.2223 ***	0.3541	0.3031	−0.2237 ***	0.4001	0.3665
	(−3.67)	(1.46)	(1.00)	(−3.73)	(1.63)	(1.22)
Constant	−0.9016	3.4958	2.3858	−0.8867	3.1633	1.9767
	(−1.31)	(1.45)	(0.88)	(−1.40)	(1.31)	(0.74)
Province FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Observations	140	163	157	140	163	157
R-squared	0.613	0.303	0.147	0.623	0.346	0.249
Number of pid	20	20	20	20	20	20

Notes: Superscript *** and ** represent statistical significance at the 1% and 5% levels, respectively. Robust t-statistics in parentheses.

shows the estimation results of these variables after lagging by one period. The coefficients of Shockpost1 and Shockpost2 are all positive and statistically significant, which are in line with our above results.

4.4.3. Alternative Measures of Regional Income Inequality

As mentioned earlier, there are many indexes to measure regional income inequality. Next, we replace the population-weighted Gini coefficient as the explained variable for regression with alternative indicators, including the non-population-weighted Gini coefficient, the mean log deviation (MLD) index, the coefficient of variation, and the Theil index.

Table 8. Regression results of alternative measures of regional income inequality.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Variables	uAGini	uAGini	AMLD	AMLD	ACOV	ACOV	ATheil	ATheil
Panel A: Shockpost1 as the explanatory variable
Shockpost1	0.0059	0.0063	0.0051	0.0055 *	0.0061	0.0068	0.0048	0.0054 *
	(1.23)	(1.16)	(1.69)	(2.07)	(1.30)	(1.49)	(1.56)	(1.98)
LRGDP		0.0218		0.0559 ***		0.1984		0.0595 ***
		(0.64)		(3.50)		(1.45)		(3.54)
LPOP		0.0072		0.0491 *		0.1160		0.0430 *
		(0.08)		(1.81)		(0.56)		(1.76)
INDR		−0.0164		−0.1098 ***		−0.1993		−0.1184 ***
		(−0.29)		(−3.29)		(−1.28)		(−3.50)
Constant	0.1297 ***	−0.1561	0.0291 ***	−0.9282 ***	0.2315 ***	−2.7515	0.0297 ***	−0.9114 ***
	(50.41)	(−0.18)	(8.68)	(−3.48)	(10.28)	(−0.99)	(8.63)	(−3.50)
Province FE	YES	YES	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES	YES	YES
Observations	140	140	144	144	157	157	144	144
R-squared	0.443	0.448	0.353	0.508	0.175	0.223	0.341	0.508
Number of pid	20	20	20	20	20	20	20	20
Panel B: Shockpost2 as the explanatory variable
Shockpost2	0.0063	0.0068	0.0051 *	0.0055 **	0.0069	0.0076 *	0.0048 *	0.0053 **
	(1.43)	(1.35)	(2.06)	(2.36)	(1.67)	(1.93)	(1.83)	(2.20)
LRGDP		0.0249		0.0568 ***		0.1994		0.0604 ***
		(0.74)		(3.58)		(1.46)		(3.59)
LPOP		0.0022		0.0453 *		0.1168		0.0395
		(0.03)		(1.78)		(0.57)		(1.67)
INDR		−0.0170		−0.1077 ***		−0.2104		−0.1163 ***
		(−0.30)		(−3.53)		(−1.34)		(−3.73)
Constant	0.1297 ***	−0.1457	0.0291 ***	−0.9080 ***	0.2316 ***	−2.7631	0.0297 ***	−0.8924 ***
	(51.28)	(−0.18)	(8.73)	(−3.61)	(10.28)	(−1.00)	(8.66)	(−3.53)
Province FE	YES	YES	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES	YES	YES
Observations	140	140	144	144	157	157	144	144
R-squared	0.459	0.466	0.370	0.522	0.184	0.232	0.356	0.520
Number of pid	20	20	20	20	20	20	20	20

Notes: In Panel A, the core explanatory variable is Shockpost1. In Panel B, the core explanatory variable is Shockpost1. Superscript ***, ** and * represent statistical significance at the 1%, 5%, and 10% levels, respectively. Robust t-statistics in parentheses.

reports the regression results of these alternative variables for Shockpost1 and Shockpost2, respectively. Due to space constraints, we only show the effect of these variables on the overall level of regional income inequality.

The coefficients of Shockpost1 (Table 8: Panel A) are all positive, but the levels of statistical significance are different, with only two of them showing as statistically significant, which are 0.0055 (p < 0.1) for Model (4) and 0.0054 (p < 0.1) for Model (8). As for the control variables, despite the difference in significance levels, their signs are consistent with above study. Similarly, the coefficients of Shockpost2 in Panel B are all positive as well. Specifically, the coefficients of Shockpost2 in Models (3), (4), (6), (7) and (8) are 0.0051 (p < 0.1), 0.0055 (p < 0.05), 0.0076 (p < 0.1), 0.0048 (p < 0.1), and 0.0053 (p < 0.05), respectively. All of them are statistically significant. These results further prove the robustness of the findings of this study.

5. Additional Analysis

The above study suggests that the COVID-19 pandemic has exacerbated regional income inequality in China. The previous literature found that COVID-19 had a significant impact on international trade, including imports and exports. These effects are substantial in the short-term and may even have long-term implications [34,35,52]. Moreover, this impact on trade will further affect regional economic growth and residents income. COVID-19 can also affect the way people work, such as moving from traditional offline work to working from home. However, the impact of the pandemic varies due to the uneven development of the digital economy between regions [43]. In addition, during the outbreak of COVID-19, the digital economy played a strong role in driving regional economic development [53]. However, due to the difference in levels of digital economy infrastructure construction in different regions, the impact of COVID-19 may also vary among regions [54]. In addition, there are also significant differences in the impact of the COVID-19 on different industries, among which the service sector has been hit relatively hard due to social distancing control and other strict measures [55,56]. Therefore, the relationship between COVID-19 and regional income inequality is probably influenced by international trade, the digital economy, and industry structure factors. To better understand the influence of these factors, the current study further examines whether the positive COVID-19 effect on regional income inequality is moderated by the foreign trade dependence (TRADE), development of the digital economy (DIG), and the share in the service industry (SERR).

Table 9. Results of moderation effects.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	AGini	AGini	AGini	AGini	AGini	AGini
Shockpost1	0.0217 ***	0.0123	0.0176	0.0793	0.0217 ***	0.0163 **
	(3.01)	(1.61)	(0.83)	(1.27)	(3.02)	(2.62)
TRADE	0.0048	0.0339
	(0.14)	(0.97)
SH_TRADE		0.0378 **
		(2.18)
DIG			0.0587 **	−0.0449
			(2.54)	(−0.49)
SH_DIG				−0.1254
				(−1.26)
SERR					−0.4253 ***	−0.2603
					(−3.10)	(−1.61)
SH_SERR						0.1849 **
						(2.60)
LRGDP	0.1049 **	0.0942 *	0.0836 *	0.0767 **	0.1054 **	0.1027 **
	(2.29)	(2.01)	(1.94)	(2.33)	(2.41)	(2.44)
LPOP	−0.0113	−0.1024	−0.0700	−0.0479	−0.0260	−0.0964 *
	(−0.17)	(−1.56)	(−1.44)	(−0.85)	(−0.49)	(−1.78)
INDR	−0.2185 ***	−0.1742 **	−0.1314 *	−0.0981 **	−0.6384 ***	−0.5858 ***
	(−3.12)	(−2.10)	(−2.03)	(−2.17)	(−4.36)	(−3.91)
Constant	−0.7888	0.0642	−0.1148	−0.2204	−0.2904	0.2242
	(−0.95)	(0.07)	(−0.24)	(-0.41)	(−0.46)	(0.37)
Province FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Observations	140	140	121	121	140	140
R-squared	0.626	0.666	0.390	0.473	0.644	0.681
Number of pid	20	20	19	19	20	20

Notes: Superscript ***, ** and * represent statistical significance at the 1%, 5%, and 10% levels, respectively. Robust t-statistics are in parentheses.

, Models (1) and (2) present the regression results for the moderating role of foreign trade dependence on the relationship of COVID-19 with regional income inequality. Foreign trade dependence (TRADE) is a continuous variable. The interaction between COVID-19 shock and foreign trade dependence (SH_TRADE) in Models (1) and (2) is applied to test the moderating effect of international trade (in the analysis of the moderation effect, since TRADE and Shockpost1 are both continuous variables, the construction of interaction term is the result of data centralization of both variables. The same is true for tests of DIG and SERR). The coefficient of the interaction term SH_TRADE shows the impact of COVID-19 on regional income inequality with respect to regional foreign trade dependence. The coefficient of SH_TRADE is positively significant (α = 0.0378, p < 0.05), as shown in Model (2). This specifies that, controlling for other factors, the increase in regional income inequality due to COVID-19 becomes more of a problem for provinces with more dependence on international trade. The findings suggest that a higher level of foreign trade dependence amplifies the impact of COVID-19 on regional income inequality.

Table 9, Models (3) and (4) report the regression results for the moderating role of digital economy development on the impact of COVID-19 on regional income inequality. The results show that, although the coefficient of the interaction term (SH_DIG) is negative (α = −0.1254), it is not statistically significant. However, this insignificance may be caused by data imperfection, as there are a lack of DIG data for the year 2021 and 2022. To some extent, the results suggest that the development of the digital economy may be able to blunt the impact of COVID-19 shocks on regional income inequality. The study also examines the moderating role of the share of the service industry on the association between COVID-19 and regional income inequality. The coefficient of the interaction term (SH_SERR) is significantly positive (α = 0.1849, p < 0.05), reflecting that provinces with a higher share of the service industry may experience a greater increase in regional income inequality under the impact of COVID-19.

6. Conclusions

This study examined the impact of COVID-19 on regional income inequality in China. Consistent with our expectations, the study detected a significantly adverse impact of COVID-19 on regional income inequality, using provincial level data from 20 provinces (covering 218 prefecture-level cities) during the period from 2013 to 2022. The findings were robust when applying a parallel trend test, lagging control variables, and alternative measures of regional income inequality to test the results. Further analysis suggested that the adverse impact of COVID-19 on regional income inequality was increased by regions’ foreign trade dependence and their share of the service industry but may have been weakened by the development of the digital economy. In other words, provinces that rely more on international trade and the services industry have experienced greater increases in income inequality under the shock of COVID-19. Moreover, provinces with higher levels of digital economy development are likely to experience a relatively smaller impact of COVID-19 on income inequality.

The study provides new evidence on the impact of COVID-19 on regional income inequality. The results of this study show that the pandemic has an unequal impact on residents’ incomes across provinces, and this impact is reflected not only at the overall level, but also at the level of urban and rural residents. In addition, different from previous studies, our research also suggested that while the cumulative impact of COVID-19 widened the regional income inequality of both urban residents and rural residents within the province, it narrowed the income gap between urban and rural residents within the province. One possible reason is that cities are more affected than rural areas in the same region because of their high population density and tighter controls. In addition, the moderation effect analysis also suggested that international trade and service industry had been relatively hard hit by the pandemic, and the development of the digital economy could circumvent the adverse impact of the COVID-19 pandemic to some extent.

While the pandemic is basically over, its impact on regional income inequality has not been eliminated and may even deepen in the future. Therefore, it remains necessary to implement measures to address the exacerbated income inequality among people in different regions due to COVID-19. Firstly, the government should increase the allocation of transfer payments to areas severely affected by the pandemic to assist them in overcoming difficulties and returning to normalcy as soon as possible. Secondly, local governments should provide tax incentives for foreign trade enterprises and service industries in regions that are more dependent on international trade and service industry development. Thirdly, the central government and local authorities should promote the construction of the digital infrastructure to enhance the development level of the digital economy, which might mitigate the persistent effects of the pandemic on regional income inequality.

There are also some limitations in this study. Firstly, the sample coverage is not extensive enough due to the lack of data at the district, county, and some prefecture-level city levels. As a result, our sample does not include Beijing, Shanghai, Tianjin, Chongqing, and other provinces. This might limit the generalizability of the conclusions in this study to some extent. Secondly, due to the limited time range of the study, the potential long-term effects of the pandemic on regional income inequality were not examined. Future research could address these limitations.