The Digital Economy and Common Prosperity: Empirical Evidence from Multidimensional Relative Poverty in China

Abstract

The swift advancement of the digital economy presents new pathways toward achieving common prosperity in China. Based on microdata derived from the China Family Panel Studies (2010–2022), this study employs the “Broadband China” pilot policy as a quasi-natural experiment to explore how digital economy development influences multidimensional relative poverty. We develop a multidimensional relative poverty index encompassing economic, health, education, and living condition aspects utilizing the Alkire–Foster dual cutoff method and employ a staggered Difference-in-Differences design for empirical analysis. Results show that the policy leads to an average decrease of 1.8 percentage points in the probability of multidimensional relative poverty across households. The effect is more pronounced in central and western regions, rural households, and those with a high proportion of non-labor force, particularly in the dimensions of economic, health, and living conditions dimensions. Mechanism analysis via interaction term regression indicates that increased population mobility and improved informal employment are key channels. These findings suggest that enhancing digital infrastructure and tailoring mobility and employment policies to fit regional and urban–rural contexts can effectively alleviate multidimensional relative poverty. This study contributes empirical evidence connecting the advancement of the digital economy to poverty alleviation and aligns with the United Nations Sustainable Development Goal 1 (No Poverty).

1. Introduction

Alleviating relative poverty is essential for advancing common prosperity [1]. Poverty remains a persistent challenge for developing countries [2]. Despite the eradication of absolute poverty in China, the world’s largest developing country, relative poverty is expected to persist [3,4,5]. This indicates that a portion of the population continues to face disadvantages in accessing economic and social resources. Consequently, China’s poverty governance policy has transitioned from an emphasis on absolute poverty to tackling relative poverty, and efforts are underway to create a comprehensive framework for this objective [6].

Relative poverty is inherently multidimensional [7]. Traditional measures often rely on a single indicator such as income or consumption. Under these measures, individuals earning below a specific proportion of the per capita or median income—commonly 40%, 50% or 60%—are classified as relatively poor [8,9,10]. Nevertheless, concentrating solely on income neglects deprivations in other vital dimensions of life [11]. As Townsend emphasized, relative poverty also encompasses inadequate access to diet, social participation, living conditions, education, and health [12]. Empirical evidence further shows that the overlap between multidimensional poverty and income poverty is often low, sometimes as little as 31% [13], highlighting the necessity to assess relative poverty through multiple dimensions.

Multidimensional relative poverty persists as a critical challenge in urban and rural areas nationwide [14]. Nevertheless, existing literature has predominantly focused on rural relative poverty [15,16,17], frequently neglecting urban or national aspects. This limits a comprehensive understanding of China’s multidimensional relative poverty challenges under the common prosperity agenda.

The digital economy has emerged as the core engine powering modern advancement, profoundly influencing economic growth, productivity, resource allocation, and common prosperity. Its rapid expansion and integration across various sectors have attracted extensive academic attention. Prior research has explored its effects on economic growth [18,19], low-carbon development [20,21,22], energy transition [23,24], technology transfer [25], total factor productivity [26,27,28], employment [29,30], consumer behavior [31,32], residents’ subjective well-being [33,34], rural revitalization [35,36], and poverty alleviation [37,38].

Existing studies on the digital economy’s impact on poverty outcomes can generally be divided into two primary strands. The first investigates its effects on single-dimensional poverty, with most focusing on income poverty [39,40]. A smaller body of work examines energy poverty, analyzing the role of digital economy development [38,41], and information technology facilities [42]. The second strand, most relevant to this study, investigates how the digital economy affects multidimensional poverty. For instance, studies have shown that mobile internet, an important channel for digital economic growth, significantly reduces rural multidimensional poverty [43]. Other studies highlight digital information technology’s effect on mitigating multidimensional poverty by enhancing social capital [44]. Moreover, digital financial inclusion not only mitigates household multidimensional poverty but also may influence multiple aspects of poverty reduction [14].

In the digital era, the widespread adoption of digital technologies exerts substantial effects on relative poverty. However, empirical research in this domain remains limited. Existing studies have predominantly concentrated on the Internet or digital finance, overlooking the inherent connection between the digital economy’s evolution and infrastructure development—a key policy instrument for poverty reduction [14].

In August 2013, the State Council launched the “Broadband China” (hereafter BC) Strategy and Implementation Plan to develop a new generation of digital infrastructure that meets the demands of socio-economic advancement. The initiative aimed to increase broadband prevalence, expand network coverage, and enhance service quality. To advance this strategy, the Ministry of Industry and Information Technology, together with the National Development and Reform Commission, designated 117 BC pilot cities across three phases in 2014, 2015, and 2016. By the end of 2023, the number of fixed broadband users had reached 636 million households, an increase of 46.66 million from the previous year (The data comes from “2023 Statistical Bulletin of the Communications Industry”. https://www.gov.cn/lianbo/bumen/202401/content_6928019.htm (accessed on 8 June 2025)). The BC pilot policy thus provides a valuable case for examining the influence of digital infrastructure on multidimensional relative poverty, yet few studies have addressed this question directly.

While most studies evaluate the digital economy’s impact on poverty primarily through income or energy dimensions, poverty has progressed from being viewed in absolute terms to relative terms and from a single-dimensional to a multidimensional concept. Despite this evolution, research on the effects of the BC pilot policy on multidimensional relative poverty remains limited, leaving a significant empirical gap.

This study addresses these gaps employing the China Family Panel Studies (CFPS) micro-level panel data, focusing on the BC pilot policy. We apply a staggered difference-in-differences (DID) strategy to explore the effects of the digital infrastructure on common prosperity from the perspective of multidimensional relative poverty. This study also investigates the mechanisms—namely, population mobility and informal employment—via which the BC policy affects multidimensional relative poverty, including population mobility and informal employment. The findings are expected to provide theoretical insights as well as applied guidance in interpreting the social impacts of the digital economy, designing effective relative poverty governance policies, and advancing the common prosperity agenda.

The present research offers three primary advancements beyond prior studies. Firstly, while most prior research has examined the digital economy’s effects on income poverty or energy poverty, and only a few have explored multidimensional poverty via channels such as Internet use or digital finance, this study concentrates on digital infrastructure—specifically, the BC pilot policy—and the influence it exerts on multidimensional relative poverty, a perspective that has been largely overlooked. Secondly, despite extensive research on the BC initiative’s impact on economic growth, industrial transformation, and employment, its potential to alleviate relative poverty remains unexplored. Using nationally representative micro-survey data (CFPS), this study provides new empirical evidence on how broadband expansion affects poverty outcomes through mechanisms such as population mobility and informal employment. Thirdly, this study uncovers the heterogeneous effects of digital infrastructure on multidimensional relative poverty across different poverty dimensions (economic, health, and living conditions), household structures, and regional as well as urban–rural divides, offering more nuanced insight.

The following sections of this study are arranged as follows. Section 2 presents the theoretical analysis and formulates four hypotheses. Section 3 describes the staggered DID approach and the data employed. Section 4 reports the analytical findings, covering benchmark regression, robustness checks, analysis of underlying mechanisms, and heterogeneity. Finally, Section 5 concludes the study and discusses the policy insights derived.

2. Theoretical Analysis and Research Hypothesis

The digital economy, as the linchpin of contemporary economic endeavors, has witnessed swift growth, largely attributable to the advancement of information infrastructure [45]. The BC pilot policy is designed to promote the advancement of broadband networks in targeted regions by establishing exemplary cities, while also underscoring the role of broadband networks as essential strategic public infrastructure. The execution of this program has greatly boosted the rollout and diffusion of network facilities across various regions, thereby enhancing the reach and velocity of broadband networks.

On one hand, the BC pilot policy has played a pivotal role in augmenting both the demand and supply for informal employment sectors. From the perspective of demand, the policy’s implementation establishes a strong foundation for diversifying employment options, thereby expanding the range of job opportunities available to the household workforce. With the pervasiveness of broadband networks, novel employment paradigms such as online part-time positions, remote employment, digital services, and e-commerce have burgeoned. These developments not only afford the household workforce the luxury of flexible working schedules and locations, presenting a plethora of employment alternatives, but also engender a wealth of opportunities within the informal employment sector.

Conversely, from the supply-side perspective of informal employment, the BC pilot policy has elevated the region’s level of informatization, effectively dismantling the geographical barriers that previously impeded information exchange between disparate regions [46]. By reducing the communication gap between urban and rural areas and enhancing interregional connectivity, the policy has broadened the public’s access to employment-related information. This enhancement significantly diminishes the search costs associated with job hunting and mitigates the likelihood of mismatches between job positions and candidates [47]. Consequently, the household workforce is now better positioned to align with employment opportunities that are commensurate with their skills and preferences.

The enhanced accessibility of information has provided the household workforce with a broader understanding of market demands, enabling them to swiftly seize employment opportunities and participate more actively in economic activities. Moreover, the rapid expansion of broadband networks has created abundant online training and educational opportunities for the household workforce. These resources are crucial for equipping them with the skills necessary to enhance their employability and, in turn, improve the quality and productivity of their informal employment engagements.

The growth of informal household employment exerts a profound impact on mitigating various dimensions of relative poverty within familial. First, informal employment serves as an additional source of income, increasing household financial resources and directly improving their economic status, thereby reducing financial strain. Second, the flexibility inherent in informal employment allows family members to choose jobs that align with household needs and individual abilities. This flexibility helps them achieve a more balanced integration of work and domestic responsibilities, ultimately improving their overall quality of life.

In sum, as the digital economy continues to develop, the scale of informal employment has expanded. This expansion plays an essential role in alleviating the complexities of relative poverty, highlighting the crucial contribution of informal employment to the economic and social well-being of families.

From the perspective of the push-pull theory, the rapid growth of the digital economy has also created a wide range of employment opportunities, which exert a geographic “pull” on population flows—particularly toward urban centers with advanced digital economic infrastructures. In the employment sector, the development of information infrastructure has been vital in driving economic growth and industrial transformation in pilot regions [48]. A variety of emerging industries based on information technology have flourished in these regions, creating numerous job opportunities for the local workforce.

The development of the information infrastructure has significantly reduced the costs of acquiring and exchanging information. This reduction has facilitated the workforce’s access to employment opportunities across different regions, thereby encouraging labor migration to areas with more abundant job markets. From the perspective of public services, the BC pilot policy has also played a crucial role in expanding access to services such as online education and telemedicine. These advances have improved the overall standard of public service provision, enabling residents to access more convenient and higher-quality services, thereby enhancing the quality of life in these pilot areas and attracting further population inflows.

With the resulting population flows and migration, family members are generally able to secure higher incomes, thereby improving their economic status. At the same time, migration allows family members to access better educational and medical resources, thereby fostering the accumulation of human capital within the family. Overall, the development of the digital economy promotes workforce mobility, which in turn helps alleviate multidimensional relative poverty within families.

Regional developmental disparity remains a prominent issue in China. On one hand, the eastern coastal regions have benefited from early implementation of reform and opening-up policies, combined with geographical advantages, resulting in stronger economic development and a more diversified industrial structure. In contrast, the central and western inland regions remain comparatively underdeveloped due to factors such as harsh natural conditions, inadequate infrastructure, and an unfavorable investment climate, which have led to a more homogeneous industrial structure. This regional imbalance has resulted in an uneven distribution of resources and employment opportunities, thereby affecting income levels and quality of life for the local population.

On the other hand, urban areas, driven by the agglomeration effects of the economy, provide more abundant employment opportunities, better public services, and higher income levels for their residents. In contrast, rural areas face challenges such as labor outflows, inadequate infrastructure, and a shortage of educational resources, which collectively constrain the living standards and developmental potential of the rural population. Based on these observations, this study proposes the following four hypotheses:

Hypothesis 1.

The BC pilot policy is instrumental in mitigating household multidimensional relative poverty.

Hypothesis 2.

The policy ameliorates household multidimensional relative poverty through the promotion of population mobility.

Hypothesis 3.

The policy mitigates multidimensional relative poverty by augmenting informal employment opportunities for families.

Hypothesis 4.

The policy’s efficacy in addressing multidimensional relative poverty exhibits significant regional and urban–rural disparities.

3. Materials and Methods

3.1. Model Specification

In order to precisely delineate the causal nexus between the digital economy and multidimensional poverty, we treat China’s BC pilot policy as an exogenous policy shock and employs the DID methodology to assess its effects on multidimensional poverty. Initiated in 2014, the State Council of China officially launched the BC pilot program, delineating a roster of inaugural demonstration cities and progressively incorporating additional cities in 2015 and 2016. This program is designed to expedite the advancement of networking infrastructure within the designated cities, tackling pivotal issues such as the acceleration of broadband access, expansion of coverage, and proliferation of applications. It is hypothesized, in alignment with these policy objectives, that the digital economic development in China’s BC pilot cities would exhibit enhanced progression. In recognition of the phased implementation of the BC pilot policy, this study constructs a staggered DID model to rigorously evaluate the policy’s efficacy, and all empirical analyses are conducted using Stata 18.

$$mr{p}_{ict}={\alpha }_1+{\theta }_{1}polic{y}_{ict}+{\gamma }_1{X}_{ict}+{\mu }_i+{\upsilon }_t+{\phi }_c+{\epsilon }_{ict}$$

(1)

Here, $mr{p}_{ict}$ denotes the multidimensional poverty status of household $i$ in city $c$ at time $t$. the key explanatory variable, $polic{y}_{ict}$ signifies whether the city $c$ where household $i$ is located is a BC pilot city at time $t$, operationalized as a dummy variable. A vector of control variables, $X_{ict}$, is included to account for factors at time $t$ that may affect the multidimensional relative poverty status of household $i$. ${\mu }_i$, ${\phi }_c$, and ${\upsilon }_t$ represent fixed effects specific to households, cities, and years, respectively; ${\epsilon }_{ict}$ represents the error term.

3.2. Choice of Variables

3.2.1. Dependent Variable

$mrp$, the dependent variable, represents whether the household falls into multidimensional relative poverty. The measurement framework utilized in this study is based on the work of Alkire and Foster [49] and adheres to the indicator selection principles established by the United Nations Development Programme (UNDP) for the Multidimensional Poverty Index (MPI). It is formulated encompassing four dimensions: economy, health, education, and living conditions. The household’s multidimensional poverty status is assessed through the dual-threshold method, which calculates the relevant indicators.

Consistent with the methodologies adopted by the UNDP, as well as the studies conducted by [50,51,52,53], the threshold for identifying a household as being a state of multidimensional poverty in this study is set when the household’s poverty level $k>1/3$. In such cases, the household is categorized as experiencing multidimensional poverty, and the variable $mrp$ is assigned a value of 1. In all other instances, the value of 0.

Table A1. Multidimensional relative poverty index.

Dimensions	Indicator	Deprived Conditions
Economy	Per capita household income	Under half of society’s median income.
	Per capita household consumption	Below 2300 yuan.
	Employment	There is unemployment among the working-age population.
Health	Self-rated health	All family members self-rate their health as being in an unhealthy state.
	Chronic diseases	There are family members suffering from chronic diseases.
	Medical insurance	All family members lack medical insurance.
Education	Years of schooling	The average educational attainment of household members older than 16 does not exceed 6 years.
	Child school dropout	There are children aged 6–15 who have dropped out of school.
Living Conditions	Clean water source	The household mainly relies on unclean water sources such as river, lake, and rainwater, as well as water from storage pits.
	Clean fuel	The main fuel used by the household is biomass fuels such as firewood and straw or fossil fuels such as coal.
	Owned housing	The household does not have housing provided by the government or the employment unit, nor do they own housing with legal ownership rights.

Notes: The deprived condition of per capita household income is made according to the OECD standard. The deprived condition of per capita household consumption is made according to China’s poverty alleviation standard.

lists the specific indicators that constitute the multidimensional relative poverty index for each dimension.

3.2.2. Key Independent Variable

We employ the BC pilot policy as an externally imposed policy shock to gauge the digital economy’s advancement. Specifically, the policy indicator, denoted as $polic{y}_{ict}$, is operationalized such that it takes on a value of 1 when the city $c$ inhabited by household $i$ is designated as a pilot city in year $t$. In all other instances, where the city is not part of the pilot initiative in a given year, $polic{y}_{ict}$ is assigned a value of 0. This binary coding allows for a clear and direct assessment of the policy’s presence and its potential impact on multidimensional relative poverty.

3.2.3. Control Variables

We meticulously select a comprehensive group of control variables across three distinct levels: the household head, the family unit, and the city context. At the household head level, the control variables encompass demographic and relational factors, such as the gender, age, and marital status of the household head. Moving to the family level, the control variables are composed of the proportion of elderly and children, family size, net family assets, and a binary indicator for urban versus rural residence. At the city level, the control variables are designed to reflect the broader economic and social environment, encompassing measures such as the economic development level, the industrial structure, the government expenditure level, and the social consumption level.

Table A2. Variable meanings.

Variable Name	Variable Meanings
mrp	Multidimensional Relative Poverty Status. This variable takes a value of 1 if a household is in multidimensional poverty, and 0 if not.
policy	BC Pilot Policy Shock. A city included in the pilot list is marketed with a value of 1; in the absence of such inclusion, it is assigned a value of 0.
gender	The Household Head’s Gender. The value takes a value of 1 if the household head is male, and 0 otherwise. This variable takes a value of 1 if the household head is male, and 0 otherwise.
age	The Household Head’s Age. The value is obtained by calculating subtracting the birth year from the survey year.
mar	The Household Head’s Marital Status. The assigned value is 1 for a household head who is married; for all other marital statuses, the value is 0.
raise	Family Elderly and Child Proportion. It is calculated by dividing the total number of elderly individuals and children under 18 within the family by the total family’s total population.
fsize	Family Size. It is denoted by the aggregate count of all individuals comprising the family unit.
asset	Family’s Net Assets. It is represented by the total value of the family’s assets.
urban	Urban–Rural Variable. This variable is binary, taking a value of 1 for urban households and 0 for rural households
ggdp	Economic Development Level. It is represented by the logarithm of per capita GDP.
stru	Industrial Structure. It is represented by the proportion of the value added by the tertiary sector relative to the secondary sector.
expen	Government Expenditure Level. It is indicated by the ratio of general public budget expenditure to GDP.
consu	Social Consumption Level. It is represented by the proportion of total retail sales of consumer goods to GDP.

Notes: Since the family’s net assets include negative numbers, logarithmic transformation is not applicable; therefore, the absolute scale is used for representation.

outlines detailed descriptions of all variables.

Table 1. Descriptive statistics.

Variable	Treatment Group		Control Group
	Mean	Std. Dev.	Mean	Std. Dev.
mrp	0.135	0.341	0.182	0.386
policy	0.496	0.500	0	0
gender	0.565	0.496	0.599	0.490
age	52.126	13.456	52.015	12.794
mar	0.863	0.344	0.882	0.322
raise	0.261	0.313	0.256	0.300
fsize	3.661	1.708	3.918	1.820
asset	82.442	187.711	29.110	67.648
urban	0.597	0.490	0.408	0.491
ggdp	6.700	3.705	3.688	2.083
stru	1.248	0.676	1.022	0.565
expen	0.177	0.086	0.226	0.400
consu	0.396	0.103	0.451	2.049

Notes: (1) Families in the BC pilot cities are included in the treated group, otherwise they are included in the control group. (2) Treatment Group (n = 21,343) Control Group (n = 24,233). (3) mrp = multidimensional poverty status; policy = BC pilot policy shock; mar = household head’s marital status; raise = family elderly and child proportion; fsize = family size; asset = family’s net assets; urban = urban–rural variable; ggdp = economic development level; stur = industrial structure; expen = government expenditure level; consu = social consumption level.

presents the descriptive statistics of the key variables for both the treatment and control groups. The treatment group consists of households located in BC pilot cities, while the control group includes households outside these pilot areas. Overall, the two groups display broadly similar demographic characteristics: the average age of household heads is around 52 years in both groups (52.126 vs. 52.015), the proportion of married household heads is 0.863 in the treatment group and 0.882 in the control group, and the share of male household heads is 56.5% and 59.9%, respectively. Some economic and locational characteristics appear to differ. Households in the treatment group tend to have smaller family sizes (3.661 vs. 3.918). In addition, 59.7% of treatment households are in urban areas, compared with 40.8% of the controls. These statistics provide a descriptive overview of the sample characteristics across groups, rather than formal evidence of significant differences.

To assess potential multicollinearity, we conducted variance inflation factor (VIF) tests for all explanatory variables (including the key explanatory variable and control variables) reported in Table 1.

Table A3. Variance inflation factor tests.

Variables	policy	gender	age	mar	raise	fsize	asset	urban	ggdp	stru	expen	consu	Mean
VIF	1.423	1.01	1.013	1.077	1.077	1.121	1.235	1.078	2.172	2.27	8.013	7.716	2.433
1/VIF	0.703	0.99	0.987	0.929	0.928	0.892	0.81	0.928	0.46	0.44	0.125	0.13	0.693

shows that the VIFs for most variables are below 5; only controls $expen$ and $expen$ exceed 5 (8.013 and 7.716, respectively). Notably, the VIFs of all explanatory variables remain below the widely accepted critical threshold of 10. The average VIF is 2.434, suggesting a relatively low overall level of multicollinearity among these explanatory variables.

3.3. Data Source

The data for the dependent variable, namely the multidimensional poverty index, as well as the control variables at both the household head and family levels, are obtained from the China Family Panel Studies (CFPS). This comprehensive database includes samples from 25 provinces, municipalities, and autonomous regions, offering an in-depth look at every family member and various dimensions, such as family income, consumption, education, healthcare, and living standards. The CFPS data is nationally representative and has been extensively utilized in poverty research studies, as evidenced by the works of [54,55]. The CFPS has conducted biennial surveys on the same subjects since its inception in 2010 to track changes over time. For this study, we use data from seven consecutive waves (2010–2022) to ensure sample consistency. Furthermore, the catalog of BC pilot cities is obtained from the Ministry of Industry and Information Technology of China, while the city-level control variables are compiled from the respective statistical yearbooks of each city.

3.4. Changes in Multidimensional Relative Poverty Incidence of Pilot Cities

Figure 1 presents the direct changes in multidimensional relative poverty incidence in cities treated by the BC pilot policy. As shown in Figure 1, after the implementation of the BC policy, the multidimensional relative poverty incidence in most pilot cities has significantly decreased. This demonstrates that the BC pilot policy may be an important reason for alleviating multidimensional relative poverty in China. Given this, this study will further empirically identify the specific effects and mechanisms of the BC pilot policy on the reduction in multidimensional relative poverty in China through a staggered DID model.

4. Results and Discussion

4.1. Benchmark Regression Results

The estimation results of model (1) using the staggered DID method are reported in Table 2. Columns (1)–(4) gradually incorporate control variables at the household head, household, and city levels. We controlled for year, household, and city fixed effects in the regression. Further, Column (5) uses a continuous multidimensional poverty score—where a higher score indicates a greater likelihood of being in multidimensional relative poverty—instead of the discrete multidimensional poverty status used in other regressions.

(1)

The effects of the BC pilot policy

As per the data in Table 2, the implementation of the BC pilot policy has led to a marked decrease in household multidimensional relative poverty, irrespective of the inclusion of household-, year-, and city-fixed effects, along with control variables at the level of household head, household, and city. This finding substantiates Hypothesis 1, indicating that the digital economy helps alleviate relative poverty [4] and promote common prosperity. Echoing these results are scholars such as [43], who explored the nexus between mobile internet usage and multidimensional poverty, concluding that mobile internet utilization can effectively mitigate multidimensional poverty.

When controlling for pertinent variables, the BC pilot policy has been found to decrease the likelihood of a household falling into multidimensional relative poverty by an average of 1.8 percentage points, a statistically significant result at the 1% confidence level. Although the magnitude of this reduction may appear modest, its practical significance should not be underestimated: given the large population base in rural China, even a 1.8- percentage-point reduction translates into millions of individuals experiencing improvements in living conditions. It is also important to note that the BC initiative was not originally designed as a poverty alleviation program but rather as a national strategy to expand digital infrastructure and information accessibility. Prior studies have predominantly focused on targeted poverty alleviation programs—such as minimum living security and public pension programs [56,57]—while the broader welfare effects of macro-level digital infrastructure initiatives remain underexplored. Consequently, the policy’s effect on multidimensional poverty operates indirectly. This helps explain why the estimated effect size is relatively modest, while still highlighting the broader policy relevance of digital infrastructure for poverty reduction.

The result in column (5) reveals that the BC policy reduced the score of multidimensional relative poverty by 0.014, an effect that is statistically significant at the 1% level.

It is worth noting the substantial difference in R² between column 4 and column 5. Both models include the full set of control variables; however, column 4 uses a binary variable indicating whether a household is in multidimensional relative poverty, while column 5 uses the continuous multidimensional relative poverty score. The higher R² in column 5 largely reflects the statistical property that continuous variables contain more information and allow more variance to be explained, whereas binary outcomes inherently have limited variance. Nevertheless, using a binary dependent variable in model 4 is theoretically and empirically justified: it directly captures whether a household is in multidimensional relative poverty, which is the key policy-relevant outcome. In contrast, the continuous score only reflects the likelihood of being in poverty—the higher the score, the greater the probability—but it does not directly indicate poverty status. Therefore, despite its lower R², column 4 provides a more policy-relevant measure of poverty incidence, while column 5 complements this analysis by offering insight into the factors associated with higher multidimensional poverty scores.

Table 2. Benchmark regression results.

Variables	(1)	(2)	(3)	(4)	(5)
	mrp	mrp	mrp	mrp	mrps
policy	−0.017 ***	−0.017 ***	−0.019 ***	−0.018 ***	−0.014 ***
	(0.006)	(0.006)	(0.006)	(0.006)	(0.002)
	[−0.030, −0.005]	[−0.029, −0.005]	[−0.031, −0.006]	[−0.031, −0.006]	[−0.017, −0.010]
gender		−0.017 ***	−0.018 ***	−0.018 ***	−0.010 ***
		(0.004)	(0.004)	(0.004)	(0.001)
		[−0.026, −0.009]	[−0.027, −0.009]	[−0.027, −0.009]	[−0.013, −0.008]
age		0.002 ***	0.002 ***	0.002 ***	0.001 ***
		(0.000)	(0.000)	(0.000)	(0.000)
		[0.001, 0.002]	[0.001, 0.002]	[0.001, 0.002]	[0.001, 0.001]
mar		−0.003	−0.003	−0.003	−0.000
		(0.009)	(0.009)	(0.009)	(0.003)
		[−0.203, 0.014]	[−0.020, 0.015]	[−0.021, 0.014]	[−0.005, 0.005]
raise			0.077 ***	0.078 ***	0.035 ***
			(0.011)	(0.011)	(0.003)
			[0.055, 0.099]	[0.056, 0.099]	[0.029, 0.040]
fsize			0.004 *	0.004 *	0.002 ***
			(0.002)	(0.002)	(0.001)
			[−0.000, 0.008]	[−0.001, 0.008]	[0.000, 0.003]
asset			0.000 **	0.000 **	−0.000 ***
			(0.000)	(0.000)	(0.000)
			[0.000, 0.000]	[0.000, 0.000]	[−0.000, −0.000]
urban			−0.011	−0.011	0.000
			(0.011)	(0.011)	(0.003)
			[−0.033, 0.010]	[−0.032, 0.010]	[−0.006, 0.006]
ggdp				0.001	−0.001 **
				(0.002)	(0.001)
				[−0.002, 0.004]	[−0.002, −0.000]
stru				−0.036 ***	−0.015 ***
				(0.008)	(0.002)
				[−0.053, −0.020]	[−0.019, −0.010]
expen				−0.009	−0.027 *
				(0.056)	(0.014)
				[−0.118, 0.100]	[−0.054, −0.000]
consu				−0.003	−0.005 **
				(0.010)	(0.003)
				[−0.023, 0.017]	[−0.010, −0.000]
Year FE	YES	YES	YES	YES	YES
Household FE	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES
R-squared	0.491	0.492	0.493	0.494	0.702
Observations	43,505	43,505	43,505	43,505	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, ** p < 0.05, * p < 0.1. (3) Standard errors in parentheses are clustered at the household level. (4) Values in square brackets indicate 95% confidence intervals. (5) mrp = multidimensional poverty status; mrps = multidimensional poverty score; policy = BC pilot policy shock; mar = household head’s marital status; raise = family elderly and child proportion; fsize = family size; asset = family’s net assets; urban = urban–rural variable; ggdp = economic development level; stur = industrial structure; expen = government expenditure level; consu = social consumption level.

Table 2. Benchmark regression results.

Variables	(1)	(2)	(3)	(4)	(5)
	mrp	mrp	mrp	mrp	mrps
policy	−0.017 ***	−0.017 ***	−0.019 ***	−0.018 ***	−0.014 ***
	(0.006)	(0.006)	(0.006)	(0.006)	(0.002)
	[−0.030, −0.005]	[−0.029, −0.005]	[−0.031, −0.006]	[−0.031, −0.006]	[−0.017, −0.010]
gender		−0.017 ***	−0.018 ***	−0.018 ***	−0.010 ***
		(0.004)	(0.004)	(0.004)	(0.001)
		[−0.026, −0.009]	[−0.027, −0.009]	[−0.027, −0.009]	[−0.013, −0.008]
age		0.002 ***	0.002 ***	0.002 ***	0.001 ***
		(0.000)	(0.000)	(0.000)	(0.000)
		[0.001, 0.002]	[0.001, 0.002]	[0.001, 0.002]	[0.001, 0.001]
mar		−0.003	−0.003	−0.003	−0.000
		(0.009)	(0.009)	(0.009)	(0.003)
		[−0.203, 0.014]	[−0.020, 0.015]	[−0.021, 0.014]	[−0.005, 0.005]
raise			0.077 ***	0.078 ***	0.035 ***
			(0.011)	(0.011)	(0.003)
			[0.055, 0.099]	[0.056, 0.099]	[0.029, 0.040]
fsize			0.004 *	0.004 *	0.002 ***
			(0.002)	(0.002)	(0.001)
			[−0.000, 0.008]	[−0.001, 0.008]	[0.000, 0.003]
asset			0.000 **	0.000 **	−0.000 ***
			(0.000)	(0.000)	(0.000)
			[0.000, 0.000]	[0.000, 0.000]	[−0.000, −0.000]
urban			−0.011	−0.011	0.000
			(0.011)	(0.011)	(0.003)
			[−0.033, 0.010]	[−0.032, 0.010]	[−0.006, 0.006]
ggdp				0.001	−0.001 **
				(0.002)	(0.001)
				[−0.002, 0.004]	[−0.002, −0.000]
stru				−0.036 ***	−0.015 ***
				(0.008)	(0.002)
				[−0.053, −0.020]	[−0.019, −0.010]
expen				−0.009	−0.027 *
				(0.056)	(0.014)
				[−0.118, 0.100]	[−0.054, −0.000]
consu				−0.003	−0.005 **
				(0.010)	(0.003)
				[−0.023, 0.017]	[−0.010, −0.000]
Year FE	YES	YES	YES	YES	YES
Household FE	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES
R-squared	0.491	0.492	0.493	0.494	0.702
Observations	43,505	43,505	43,505	43,505	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, ** p < 0.05, * p < 0.1. (3) Standard errors in parentheses are clustered at the household level. (4) Values in square brackets indicate 95% confidence intervals. (5) mrp = multidimensional poverty status; mrps = multidimensional poverty score; policy = BC pilot policy shock; mar = household head’s marital status; raise = family elderly and child proportion; fsize = family size; asset = family’s net assets; urban = urban–rural variable; ggdp = economic development level; stur = industrial structure; expen = government expenditure level; consu = social consumption level.

(2)

The effects of demographic characteristics

Regarding gender, the regression results in Table 2 show that being male is significantly associated with a lower likelihood of multidimensional relative poverty. Specifically, holding other factors constant, being male reduces the probability of experiencing multidimensional relative poverty by about 1.7–1.8 percentage points and the multidimensional relative poverty score by 0.01.

Regarding age, the regression results in Table 2 show that it is positively associated with multidimensional relative poverty. This is consistent with the findings of Peng and Mao [4] and Wang et al. [14]. Specifically, holding other factors constant, each additional year of age raises the probability of experiencing multidimensional relative poverty by approximately 0.2 percentage points and increases the multidimensional relative poverty score by about 0.001. This pattern suggests that older individuals are slightly more prone to higher multidimensional relative poverty, which may reflect reduced labor-market participation after retirement, lower post-retirement income, higher health-related expenditures, and more limited access to new economic opportunities.

(3)

The effects of household characteristics

Regarding the proportion of elderly and children in the household, the regression results in Table 2 show that a one-unit increase in the share of elderly and children in a household raises the probability of experiencing multidimensional relative poverty by approximately 7.7–7.8 percentage points and increases the multidimensional relative poverty score by 0.035. This suggests that households with a higher dependency ratio tend to face greater multidimensional relative poverty, which aligns with the finding of Wang et al. [14].

Regarding family size, the regression results in Table 2 show that larger households tend to face slightly greater multidimensional relative poverty, which is consistent with the finding of Wang et al. [14]. Specifically, an additional household number increases the probability of multidimensional relative poverty by about 0.4 percentage points and the multidimensional relative poverty score by 0.002.

Regarding a household’s assets, although its effect is statistically significant, the estimated coefficients are close to zero, indicating that, in practice, it has a negligible impact on multidimensional relative poverty.

(4)

The effects of macroeconomic characteristics

The coefficients for $ggdp$ are 0.001 (not significant) for multidimensional relative poverty status and −0.001 for multidimensional relative poverty score, indicating that higher economic development has little effect on the probability of a household being in multidimensional relative poverty but significantly reduces multidimensional relative poverty score.

Regarding industry structure, the regression results in Table 2 show that a one-unit increase in the tertiary-to-secondary industry ratio reduces the probability of multidimensional relative poverty by 3.6 percentage points and multidimensional relative poverty score by 0.015. This result suggests that industrial upgrading—characterized by a shift toward a greater share of the tertiary sector—can play a significant role in reducing both the likelihood and score of household multidimensional relative poverty.

The coefficients for $expen$ are −0.009 (not significant) for multidimensional relative poverty status and −0.027 for multidimensional relative poverty score, indicating that a larger government expenditure share has no significant effect on the probability of multidimensional relative poverty but significantly reduces multidimensional relative poverty score.

The coefficients for $consu$ are −0.003 (not significant) for multidimensional relative poverty status and −0.005 for multidimensional relative poverty score. This indicates that a higher level of social consumption has no significant effect on the probability of multidimensional relative poverty but significantly reduces multidimensional relative poverty score.

4.2. Parallel Trends Test

The validity of the staggered DID design rests on the condition that parallel trends hold, which posits that the trajectory of multidimensional poverty within families in both pilot and non-pilot cities should be congruent prior to the policy’s implementation. The event study method offers a clear and visual means to observe and assess the dynamic reactions and disparities in individual behavior before and after the policy’s enactment. Consequently, utilizing this approach to verify the parallel trends preceding the event has become a standard procedure in the current application of the DID methodology [58,59]. In this study, the event study method is employed to analyze parallel trends. The regression model used for this analysis is specified as follows:

$$mr{p}_{ict}={\alpha }_2+{\displaystyle \sum _{t=-2}^{2+}{\theta }_t{D}_{ict}+{\gamma }_2{X}_{ict}+{\mu }_i+{\nu }_t+{\phi }_c+{\epsilon }_{ict}}$$

(2)

In the model, $D_{ict}$ represents a set of dummy variables that indicate whether the city $c$ where household $i$ is located, is designated as a BC pilot city during period $t$. The remaining variables align with Model (1). This paper concentrates on the coefficient ${\theta }_t$, which captures the divergence in multidimensional poverty status between families in pilot cities and those in non-pilot cities during the t^th period subsequent to the rollout of the BC pilot policy.

This paper selects the third period prior to the policy’s enactment as the base period. The outcomes of the parallel trends test are depicted in Figure 2. It is observed that the estimated coefficients for each period leading up to the policy’s implementation are statistically insignificant, whereas the coefficients exhibit a significant departure from zero post-implementation. This suggests that the research sample has successfully passed the parallel trends test.

4.3. Placebo Test

4.3.1. Time Placebo Test

To ensure that the observed disparities in multidimensional relative poverty between families in pilot and non-pilot cities are not driven by temporal variations, we conduct a placebo test by advancing the policy implementation by two periods and one period, as well as delaying it by one period and two periods. The results are reported in

Table 3. Time placebo test.

Variables	(1)	(2)	(3)	(4)
	One Period in Advance	Two Periods in Advance	One Period Lag	Two Periods Lag
policy	−0.010	−0.016	−0.003	−0.002
	(0.007)	(0.011)	(0.006)	(0.007)
	[−0.024, 0.003]	[−0.037, 0.005]	[−0.015, 0.009]	[−0.016, 0.012]
Control	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Household FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
R-squared	0.495	0.494	0.494	0.494
Observations	43,505	43,505	43,505	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Standard errors in parentheses are clustered at the household level. (3) Values in square brackets indicate 95% confidence intervals.

, where 95% confidence intervals for the estimated coefficients are also provided. As shown, none of the coefficients are statistically significant at the 10% level, and all confidence intervals include zero. This indicates that there are no systematic differences in the pre- or post-treatment time trends between the treated and control groups. Therefore, the benchmark regression result—that the BC pilot policy significantly reduces household multidimensional relative poverty—remains robust.

4.3.2. Individual Placebo Test

To mitigate the influence of unobservable omitted variables, this paper conducts an individual placebo test by substituting the treatment group with a simulated counterpart. This method necessitates the random selection of an equivalent number of fictitious treatment group samples, ensuring that the sample size remains consistent across all years for the virtual treatment group. To achieve this, the original data is transformed into a balanced panel dataset. With 583 family samples in the pilot cities presented in the balanced panel each year, an equivalent number of families are randomly selected to establish the simulated-treatment group, while the rest constitute the simulated-control group. This process is repeated over 500 iterations to generate a set of 500 regression coefficients along with their corresponding p-values. The resulting kernel density distributions are illustrated in Figure 3. It is evident that the simulated coefficients are largely concentrated near zero, with the majority of regression outcomes being statistically insignificant. The benchmark regression coefficient, which indicates the BC pilot policy’s significant alleviation of household multidimensional relative poverty, is positioned at the extreme upper tail of the placebo test coefficient distribution, signifying a low-probability occurrence under the individual placebo simulation. The successful passage of this test further substantiates the robustness of the benchmark regression findings.

4.4. Robustness Check

In the baseline regression, we conclude that the BC policy alleviates multidimensional relative poverty. This conclusion is indirectly supported by the study conducted by Ma et al. (2025), which found that digital infrastructure can significantly enhance the quality of life for residents [60]. To ensure the reliability of the findings, we undertook a comprehensive set of robustness checks from various angles.

4.4.1. PSM-DID Analysis

To address potential endogeneity stemming from omitted variables and measurement errors, we use the Propensity Score Matching Difference-in-Differences approach (PSM-DID). This was applied using the nearest neighbor matching technique to reassess the benchmark regression.

4.4.2. Policy Interference Control

Through an examination of Chinese government documents, the study identified the concurrent promotion of the “National Big Data Comprehensive Pilot Zone” and the “National Digital Economy Innovation and Development Pilot Zone” policy during the sample period. To neutralize any confounding effects this policy might have had on household multidimensional poverty status, potentially skewing the regression outcomes, a dummy variable representing this policy was incorporated into the benchmark regression model, and the analysis was re-conducted.

4.4.3. Exclude the Impact of COVID-19

Taking into account the influence of the COVID-19 pandemic on the conclusion, we further excluded the years affected by it and conducted a regression again. The results are shown in the (3) column of Table 4 According to the results, after excluding the influence of the pandemic, the BC policy still significantly reduced the probability of households falling into multi-dimensional poverty.

4.4.4. Sample Consistency

Recognizing that migration behaviors among families could introduce bias in the policy effect estimation, the study excluded samples with changes in their city-level residence and re-estimated the model to maintain sample integrity. Table 4 presents the outcomes of these robustness checks. The findings consistently support the benchmark regression, confirming that the BC pilot policy markedly diminishes household multidimensional relative poverty and that the results are robust.

Table 4. Robustness analysis.

Variables	(1)	(2)	(3)	(4)
	PSM-DID	Eliminate Other Policy Interference	Exclude the Impact of COVID-19	Consider a Change of Residence
policy	−0.023 **	−0.018 ***	−0.021 ***	−0.018 ***
	(0.010)	(0.006)	(0.006)	(0.006)
Control	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Household FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
R-squared	0.567	0.494	0.505	0.494
Observations	38,570	43,505	40,622	43,181

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, ** p < 0.05. (3) Standard errors in parentheses are clustered at the household level.

Table 4. Robustness analysis.

Variables	(1)	(2)	(3)	(4)
	PSM-DID	Eliminate Other Policy Interference	Exclude the Impact of COVID-19	Consider a Change of Residence
policy	−0.023 **	−0.018 ***	−0.021 ***	−0.018 ***
	(0.010)	(0.006)	(0.006)	(0.006)
Control	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Household FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
R-squared	0.567	0.494	0.505	0.494
Observations	38,570	43,505	40,622	43,181

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, ** p < 0.05. (3) Standard errors in parentheses are clustered at the household level.

4.5. Heterogeneity Analysis

4.5.1. Regional and Urban–Rural Heterogeneity

Research on multidimensional relative poverty needs to take into account regional development gaps [61]; therefore, this paper delves into the regional disparities and the urban–rural divide in the efficacy of the BC pilot policy, with the findings detailed in Table 5. On the one hand, column (1) delineates the policy’s varying impacts on multidimensional relative poverty across distinct geographical regions. The coefficient of $policy$ is −0.037, which is significant at the 1% level. This indicates that the BC initiative significantly reduces multidimensional relative poverty in western regions, where economic development and infrastructure are relatively underdeveloped. These regions stand to gain more from the widespread adoption and acceleration of broadband networks, which can markedly improve the living standards, educational opportunities, and healthcare access of local residents. For eastern regions, the interaction term ($policy\times East$) is positive but statistically insignificant, implying the policy’s poverty-reducing effect in the east is not statistically different from that in the west. For central regions, the interaction term ($policy\times Central$) is 0.032 and marginally significant at the 10% level. Summing the main and the interaction effects yields a net impact of −0.005, suggesting that the poverty reduction in central regions is slightly smaller than in western regions. Overall, the BC initiative delivered a clear poverty-reducing benefit in the western and central regions, while its effect in the eastern region is statistically insignificant.

On the other hand, column (2) examines the policy’s differential effects on urban and rural areas concerning multidimensional relative poverty. The estimated impact of the policy is −0.029, which is statistically significant at the 1% level in rural areas, indicating a notable mitigation in multidimensional relative poverty due to the BC pilot. In urban areas, the net effect equals −0.011 (−0.029 + 0.018), which is statistically significant at the 10% level, indicating that the poverty-mitigating effect is comparatively smaller. This could be attributed to the fact that in urban areas, the network infrastructure is relatively complete, which further reduces the substantial direct benefits obtained from broadband enhancement. In rural areas, the expansion of broadband networks serves to augment the information access capabilities of rural dwellers, broaden employment avenues for the agricultural workforce, and facilitate the marketing and sale of agricultural produce, thereby boosting income levels. It also enhances the availability of educational and medical resources in rural districts. These observations underscore the regional and urban–rural heterogeneity in the policy’s impact on multidimensional poverty, thereby substantiating Hypothesis 4.

Table 5. Regional heterogeneity, urban–rural heterogeneity, and heterogeneity in household demographic structure.

Variables	(1)	(2)	(3)
	Regional Heterogeneity	Urban–Rural Heterogeneity	Heterogeneity in Household Demographic Structure
policy	−0.037 ***	−0.029 ***	−0.003
	(0.014)	(0.010)	(0.007)
policy × East	0.019 (0.015)
policy × Central	0.032 * (0.016)
East	1.212 *** (0.015)
Central	1.178 *** (0.028)
policy × Urban		0.018 * (0.010)
policy × High proportion of non-labor force			−0.032 *** (0.008)
Urban		−0.015 (0.011)
Control	YES	YES	YES
Year FE	YES	YES	YES
Household FE	YES	YES	YES
City FE	YES	YES	YES
R-squared	0.497	0.494	0.494
Observations	36,198	43,505	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, * p < 0.1.

Table 5. Regional heterogeneity, urban–rural heterogeneity, and heterogeneity in household demographic structure.

Variables	(1)	(2)	(3)
	Regional Heterogeneity	Urban–Rural Heterogeneity	Heterogeneity in Household Demographic Structure
policy	−0.037 ***	−0.029 ***	−0.003
	(0.014)	(0.010)	(0.007)
policy × East	0.019 (0.015)
policy × Central	0.032 * (0.016)
East	1.212 *** (0.015)
Central	1.178 *** (0.028)
policy × Urban		0.018 * (0.010)
policy × High proportion of non-labor force			−0.032 *** (0.008)
Urban		−0.015 (0.011)
Control	YES	YES	YES
Year FE	YES	YES	YES
Household FE	YES	YES	YES
City FE	YES	YES	YES
R-squared	0.497	0.494	0.494
Observations	36,198	43,505	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, * p < 0.1.

4.5.2. Heterogeneity in Household Demographic Structure

Household demographic structure may shape the poverty-reduction effects of the BC initiative. Generally, families with a higher share of non-employed members often face heavier caregiving burdens and lower income stability. Their poverty is more likely to be alleviated through improved public services and access to informal employment [47]. The expansion of digital infrastructure can reduce information costs and broaden informal employment opportunities, thereby helping to alleviate poverty in such households.

To examine heterogeneous effects across household structures, we incorporate an interaction between the policy variable and the share of non-labor force members—defined as the proportion of elderly and minor children in the household. This specification captures how the policy’s impact varies with household demographic characteristics, with estimates reported in column (3) of Table 5. The results show that the BC policy reduces the probability of multidimensional relative poverty by 0.3 percentage points in families with fewer non-working members, though this effect is not statistically significant. In contrast, in households with a larger share of non-working members—including the elderly and children—the reduction reaches 3.5 percentage points and is significant at the 1% level. These findings demonstrate that the poverty-reducing effect of the policy strengthens as the proportion of dependents within the household increases.

Households with a large share of non-working members often face “dual vulnerability”. Economically, a high proportion of elderly and children increases the caregiving burden, reducing disposable household income. These non-labor force members are also less likely to secure suitable jobs in traditional labor markets. Broadband infrastructure can mitigate these challenges by enabling remote and gig-based work, thereby raising household income. From a non-economic perspective, the elderly and children have greater needs for healthcare and education. Improved broadband access under the BC policy enhances the availability of telemedicine and lowers education costs, thus helping to alleviate poverty in the health and education dimensions.

4.5.3. Poverty Dimension Heterogeneity

This study further investigates the BC policy’s effects on different dimensions of multidimensional relative poverty, namely economic status, health, education, and living conditions. The regression results are presented in Table 6. Based on the 1/3 threshold criterion for multidimensional relative poverty, a family is considered impoverished in a specific dimension if at least one of the three or four indicators in that dimension falls below the poverty line. As shown in Table 6, the BC policy has had a pronounced alleviating influence on economic, health, and living conditions dimensions among families. However, its impact on educational poverty was not significant, possibly due to the long-term nature of educational outcomes and their delayed benefits.

The implementation of the BC policy has improved the availability and quality of broadband services, which in turn has spurred the growth of emerging sectors such as e-commerce and remote work. These developments have created new employment opportunities, particularly for families in remote regions, thereby diversifying their income sources. Furthermore, the expansion of broadband networks has promoted the widespread adoption of e-commerce, broadening and simplifying the families’ choices and thus helping combating economic impoverishment.

In the digital economy, the cost of information acquisition for families has been greatly reduced. They can now easily access online resources for disease prevention and health promotion, which enhance their health awareness and self-care abilities. Additionally, the rise of internet-based healthcare models, such as telemedicine, has diminished regional disparities in medical resources, making high-quality healthcare more accessible and thus effectively mitigating health-related poverty.

The BC initiative has steadily enhanced informatization, fostering greater environmental awareness among families [42]. This has led to a reduction in the use of non-clean resources and an improvement in living conditions, thereby alleviating poverty in this dimension. These findings highlight the regional and urban–rural heterogeneity in the policy’s impact on multidimensional relative poverty, thus supporting Hypothesis 4.

Table 6. Heterogeneity of dimensions.

Variables	(1)	(2)	(3)	(4)
	Poverty in Economy	Poverty in Health	Poverty in Education	Poverty in Living Standards
policy	−0.037 ***	−0.024 **	−0.002	−0.087 ***
	(0.008)	(0.009)	(0.006)	(0.008)
Control	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Household FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
R-squared	0.521	0.413	0.794	0.529
Observations	43,505	43,505	43,505	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, ** p < 0.05.

Table 6. Heterogeneity of dimensions.

Variables	(1)	(2)	(3)	(4)
	Poverty in Economy	Poverty in Health	Poverty in Education	Poverty in Living Standards
policy	−0.037 ***	−0.024 **	−0.002	−0.087 ***
	(0.008)	(0.009)	(0.006)	(0.008)
Control	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Household FE	YES	YES	YES	YES
City FE	YES	YES	YES	YES
R-squared	0.521	0.413	0.794	0.529
Observations	43,505	43,505	43,505	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, ** p < 0.05.

4.6. Mechanism Analysis

To better understand how the BC pilot policy reduces multidimensional poverty, this section examines two key transmission mechanisms: (1) the promotion of population mobility and (2) the expansion of informal employment. These mechanisms reflect heterogeneous behavioral responses across different household types and development contexts. The regression outcomes related to the mechanism analysis are reported in

Table 7. Mechanism analysis.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Total	Urban	Rural	Total	Urban	Rural
Policy × Movement	−0.027 ***	−0.012	−0.033 **
	(0.010)	(0.013)	(0.016)
Policy × Informal				−0.036 ***	−0.031 ***	−0.034
				(0.007)	(0.008)	(0.021)
Movement	−0.021 ***	0.007	−0.036 ***
	(0.006)	(0.009)	(0.009)
Informal				0.021 ***	0.025 ***	0.003
				(0.005)	(0.006)	(0.016)
Policy	−0.005	−0.002	−0.025	0.012 *	0.014 **	0.012 *
	(0.008)	(0.009)	(0.016)	(0.007)	(0.007)	(0.007)
Control	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Household FE	YES	YES	YES	YES	YES	YES
City FE	YES	YES	YES	YES	YES	YES
R-squared	0.520	0.501	0.517	0.511	0.496	0.511
Observations	34,455	16,516	16,984	43,505	20,861	43,505

Notes: (1) Robust standard errors are shown in parentheses. (2) Statistical significance is indicated as follows: *** p < 0.01, ** p < 0.05, * p < 0.1.

.

4.6.1. Population Mobility

Population mobility can significantly reduce the risk of the labor force falling into relative poverty [62]. The first mechanism investigates whether improved broadband access promotes population mobility and thus alleviates multidimensional relative poverty. To test this, we interact the policy variable with a dummy variable, Movement, which equals one if any household member has engaged in outbound labor migration. As reported in Table 7, column (1), the interaction coefficient is negative and statistically significant, suggesting that the BC policy increases household mobility and thereby reduces multidimensional relative poverty. This finding supports Hypothesis 2. The mechanism operates through broadband’s ability to reduce information asymmetries, lower search costs, and expand access to labor market opportunities in more developed regions. By enabling faster and broader information flows, broadband connectivity makes it easier for individuals—particularly those from rural areas—to identify and pursue income-generating opportunities elsewhere. It also strengthens long-distance social networks and reduces the uncertainty surrounding migration decisions.

The heterogeneous effects are evident across urban and rural households, as reported in columns (2) and (3) of Table 7. The policy’s influence on population mobility is much stronger for rural households. Historically, rural areas have faced substantial geographic and informational barriers, making labor migration a crucial strategy for escaping poverty. Broadband infrastructure mitigates these constraints by providing real-time access to job information, lowering relocation costs, and fostering interregional social ties. By contrast, urban households already operate in relatively open labor markets with lower migration costs and greater access to information. Therefore, the marginal effect of broadband access on their mobility behavior is limited and statistically insignificant.

While broadband infrastructure promotes geographic labor mobility—particularly among rural households—its impact extends beyond physical relocation. In urban areas, where digital ecosystems are more developed, broadband access fosters a distinct type of labor-market flexibility: informal and platform-based employment. The next subsection examines this complementary mechanism.

4.6.2. Informal Employment

The second mechanism examines whether broadband promotes informal employment and thereby alleviates multidimensional relative poverty. We create an interaction term between the policy indicator and a dummy variable, Informal, which equals one if the household head works without employee insurance. As shown in column (4) of Table 7, the coefficient for this interaction is negative and statistically significant, indicating that the BC policy promotes informal employment, which in turn contributes to a reduction in multidimensional relative poverty. This supports Hypothesis 3 and aligns with evidence from Chiplunkar and Goldberg (2022), who found that mobile internet expansion significantly increases informal and self-employment [63]. This mechanism reflects how the digital economy is reshaping labor markets. Broadband connectivity fosters flexible work arrangements, including remote work, gig employment, and online freelancing [64], and lowers entry barriers for self-employment and entrepreneurship [65], enable more people to earn income through digital platforms and online services. These forms of informal employment offer households additional channels to improve their livelihoods, particularly for those excluded from formal labor markets.

Further analysis reveals substantial heterogeneity across household types. Columns (5) and (6) of Table 7 show that this channel is particularly effective among urban households. Urban regions tend to have more advanced digital infrastructure, greater demand for services, and more vibrant digital ecosystems, such as platform-based gig economies and online marketplaces. As a result, urban residents are better equipped to benefit from informal employment opportunities enabled by broadband. Conversely, the effect is not statistically significant among rural households. The predominance of agricultural work, the limited availability of digital job opportunities, and insufficient digital literacy hinder rural residents’ ability to capitalize on broadband-enabled employment. Consequently, while the policy has a notable effect on informal employment overall, its relative poverty-reduction impact through this channel is largely concentrated in urban areas.

5. Conclusions and Policy Implications

5.1. Conclusions

This study offers novel empirical evidence on how the development of the digital economy contributes to common prosperity by focusing on the BC pilot policy and investigating its effects on multidimensional relative poverty. Distinct from previous studies that largely relied on single-dimensional poverty indicators, our approach adopts a multidimensional relative poverty framework and employs a staggered difference-in-differences (DID) strategy using longitudinal CFPS data (2010–2022). This methodological design enables us to identify the causal impact of digital infrastructure on poverty reduction with improved robustness.

The findings can be summarized into three key insights: Firstly, the BC pilot policy significantly alleviates multidimensional relative poverty, reducing both the probability of experiencing multidimensional relative poverty and the multidimensional relative poverty score. Specifically, the policy is associated with a decrease of approximately 1.8 percentage points in the probability of households falling into multidimensional relative poverty and a reduction of 0.014 in their multidimensional relative poverty scores. Secondly, the policy’s effects are heterogenous. The impacts are substantially stronger in central and western regions, rural areas, and households with a high proportion of non-labor force members. Dimension-wise analysis further shows that the BC policy reduces deprivation mainly in the economic, health, and living-condition dimensions, while its effect on education-related deprivation is limited. Finally, mechanism analyses indicate that the BC policy operates primarily by promoting interregional population mobility and expanding informal employment opportunities, thereby enhancing household income generation and resource access.

These findings enrich the literature by providing new evidence that digital infrastructure can act as an equalizing force, narrowing spatial and social disparities in the progress toward common prosperity. They also highlight that strengthening digital infrastructure in disadvantaged areas may yield disproportionately large poverty reduction benefits.

Nevertheless, some limitations remain. Our analysis is constrained by the latest available wave of CFPS data (up to 2022). Future research could extend this work by incorporating newer data to capture long-term effects. Additionally, since BC represents a specific type of digital infrastructure policy, further research should investigate the effects of other digital initiatives—such as the Digital Village strategy or the rollout of 5G networks—to assess the broader applicability of our conclusions.

Overall, this study underscores the potential of digital economic development as a strategic pathway to reduce multidimensional relative poverty and advance common prosperity, especially for vulnerable groups and lagging regions.

5.2. Policy Implications

Based on the results of this study, several policy recommendations can be proposed. First, strengthening the development of digital infrastructure is crucial. Given the significant role of the BC pilot initiative in alleviating multidimensional relative poverty, government support should be further reinforced. Measures should include expanding the policy’s coverage, accelerating nationwide information network construction, and placing particular emphasis on central and western regions as well as rural areas. Improving network coverage and speed, while narrowing the digital divide, will enable residents to access more reliable and efficient internet services, thereby fostering the growth of the digital economy and contributing to relative poverty alleviation in pursuit of common prosperity.

Second, the government should promote population mobility and urban–rural integration. Population mobility is a key mechanism through which the BC initiative contributes to a reduction in multidimensional relative poverty. The government should ease household registration restrictions, refine mobility-related regulations, and encourage rational migration between urban and rural areas. Leveraging digital technology to improve services for the mobile population would help safeguard their rights, promote social integration, and maximize the digital economy’s role in relative poverty alleviation.

Third, support measures for informal employment should be refined. This study finds that the BC policy alleviates household multidimensional relative poverty by promoting informal employment. Given the unique challenges faced by informal workers—such as insufficient social security and limited digital literacy—the government should strengthen vocational training and employment guidance. Providing entrepreneurship subsidies and improving the social security system would help address these challenges, adapt to the transformations of the digital era, and fully leverage the digital economy’s potential for relative poverty reduction.

Finally, it is crucial to address regional and urban–rural disparities in relative poverty reduction. Given the heterogeneous impacts of the BC policy, the government should take these differences into account during policy implementation. For central and western regions and rural areas, targeted measures and increased investment are needed to bridge the digital divide and enhance policy effectiveness. Moreover, policymakers should pay attention to the specific dimensions of multidimensional relative poverty and adopt tailored interventions to address challenges in each domain.