The Digital Economy and Flexible Employment Quality: Empirical Evidence from China

Abstract

The digital economy has reshaped the structure and operation of the labor market through profound technological changes, exerting systematic impacts on the quality of flexible employment. Based on five consecutive periods of data from the China Family Panel Survey (CFPS) between 2014 and 2022, a multidimensional, flexible employment quality indicator system is constructed to empirically examine the effects, heterogeneity, and mechanisms of the digital economy on flexible employment quality. This study shows that the digital economy has significantly improved the overall quality of flexible employment. Specifically, male, low-skilled, young, and flexible workers with a low dependency ratio on the northwest side of the Hu-Huan-Yong Line benefit more significantly. Mechanism testing results indicate that industrial structure optimization, human capital accumulation, and improved matching efficiency are important intermediary pathways through which the digital economy enhances the quality of flexible employment. The conclusion indicates that amplifying the positive empowerment effect of the digital economy on the quality of flexible employment, implementing targeted policies, and activating three indirect transmission pathways—industrial structure, human capital, and supply–demand matching efficiency—are key measures to enhance the quality of flexible employment.

1. Introduction

In the global agenda for promoting inclusive, resilient, and sustainable development [1], high-quality employment is the key driver of economic growth and social equity. As an increasingly important component of the labor market, flexible employment, with its high spatiotemporal flexibility and diverse entry barriers, has become a crucial channel for absorbing labor and alleviating structural unemployment. By the end of 2024, the number of flexible workers in China had reached approximately 240 million, accounting for nearly one-third of the total employed population [2]. However, this large group generally faces the reality of “quantity growth with weak quality”, significant fluctuations in labor compensation, insufficient social security coverage, and unclear career development paths. Without intervention, flexible employment may evolve into an unsustainable form, running counter to the goal of “promoting decent work” in the United Nations 2030 Agenda for Sustainable Development.

At the same time, the rapid development of the digital economy has created new opportunities to address the sustainability challenges associated with flexible employment. In 2023, the scale of China’s digital economy exceeded CNY 53.9 trillion, accounting for 42.8% of GDP [3]. Digital technologies, including artificial intelligence, big data, and 5G, are profoundly reshaping labor processes, employment relationships, and value distribution mechanisms [4,5]. However, the digital economy has dual sustainability effects: on one hand, it empowers marginalized groups to access the labor market and strengthens their economic participation by creating new flexible jobs [6]; on the other hand, the implicit control of algorithmic management, skill-biased technological changes, and regional digital divides may trap vulnerable flexible workers in informal employment [7], further widening the gap in employment quality between groups and undermining social inclusivity.

Therefore, how the digital economy influences and through which mechanisms it affects flexible employment quality, as well as the logical relationships and temporal sequence among these mechanisms, are key issues that urgently need to be addressed. To address the above questions, this paper systematically analyzes the mechanism by which the digital economy affects the quality of flexible employment. Based on continuous five-period panel data from the China Family Panel Survey (CFPS) between 2014 and 2022, this study conducts an empirical test of the impact of the digital economy on the quality of flexible employment. The subsequent sections are organized as follows: Section 2 is a literature review, including an analysis of existing achievements and limitations. Section 3 develops an analytical framework encompassing the supply and demand sides and their matching and proposes research hypotheses. Section 4 details the research design. Section 5 reports baseline regression results, addresses endogeneity issues, and conducts heterogeneity tests. Section 6 highlights mechanism testing to deepen the understanding of the causal pathways. Section 7 outlines the research conclusions and policy implications.

2. Literature Review

2.1. Studies on the Digital Economy and Flexible Employment

Flexible employment quality has long been a central concern in academia [8]. Early studies often used formal employment as a benchmark, measuring the quality of informal employment using objective indicators such as stable employment contracts, reasonable compensation, safe working conditions, social security coverage, and career development pathways [9]. With the rapid global expansion of flexible employment forms, academics have gradually recognized that psychological dimensions such as job satisfaction, perceived autonomy, sense of time control, and sense of meaning are also important components of employment quality [10]. The concept of “decent work,” proposed by the International Labour Organization (ILO, 2017) [11], further advances this integration by defining it as a multidimensional framework encompassing fair income, workplace safety, social security, social dialog, gender equality, skill development, and personal dignity. This framework provides a more inclusive analytical benchmark for assessing the quality of flexible employment.

With the advancement of digital technologies, research on the impact of the digital economy on the quality of flexible employment has become increasingly comprehensive, primarily reflecting two aspects: enabling and constraining effects. Regarding enabling effects, digital platforms significantly lower market entry barriers through algorithmic matching, real-time scheduling, and disintermediation, thereby creating unprecedented economic participation opportunities and time autonomy for women, people with disabilities, residents in remote areas, and part-time workers [12,13]. The digital economy also enhances labor participation willingness through work flexibility and timely income [14], while more precise matching can improve labor productivity and optimize income levels [15]. Regarding constraining effects, extensive empirical studies reveal that the digital economy may reduce the quality of flexible employment. Platforms exercise substantial control over labor processes through algorithmic management while evading employer responsibilities as “independent contractors,” thereby exposing workers to income instability, lack of social security, absence of collective bargaining rights, and decision-making risks arising from algorithmic opacity [16,17]. More critically, platforms often shift operational risks entirely to individuals, creating a mechanism of risk individualization that leaves flexible workers to bear market uncertainties without institutional buffers. Additionally, Aghion et al. (2019) [18] highlighted that the digital economy may reduce the share of labor income in the overall income distribution, further amplifying income fluctuations and unemployment risks for flexible employment groups [19], posing potential threats to labor market sustainability.

2.2. Existing Research Gaps and Marginal Contributions of This Paper

While existing research provides valuable references for this study, it is evident that current studies still have significant limitations in incorporating flexible employment into analyses of work quality. First, while existing literature has preliminarily revealed the dual impact of the digital economy on flexible employment quality, related research remains fragmented at the mechanism analysis level, failing to systematically explain how the digital economy influences flexible employment quality through multiple interactive pathways. Second, the unique characteristics of flexible employment are often overlooked in current employment quality indicator systems. Through de-employment labor relations, enterprises transfer operational risks—including job instability, lack of social security, occupational injury protection gaps, and algorithmic decision-making uncertainties—to individual workers. However, these risks, as a critical component of flexible employment, are not adequately considered. Third, most empirical studies focus on platform-based gig work, whereas service-oriented and localized gig work receive insufficient attention, and even fewer studies differentiate this group of flexible workers.

Compared with the existing literature, this study makes the following potential contributions: First, it establishes a theoretical analytical framework integrating supply-side empowerment, demand-side drivers, and supply–demand matching efficiency enhancement, providing an in-depth analysis of how the digital economy improves the quality of flexible employment. Second, to address the high-risk-bearing characteristic of flexible employment, it develops a five-dimensional evaluation system encompassing work environment, welfare protection, career development, subjective experience, and risk exposure, using the entropy method to assess the quality of flexible employment. Third, drawing on multidimensional perspectives of individual, family, and regional resource endowments, it employs a high-dimensional fixed-effects model to reveal the differentiated impacts of the digital economy on the quality of flexible employment across groups.

3. Theoretical Framework

Flexible employment quality constitutes a comprehensive assessment of the subjective and objective well-being of workers in this sector, integrating both the common features of traditional employment and the unique aspects of flexible work. The digital economy, characterized by its pervasive reach, robust connectivity, and advanced intelligence, exerts a systematic influence on the quality of flexible employment through three key dimensions: demand-side optimization of industrial structures, supply-side enhancement of human capital, and improved matching efficiency (Figure 1). Specifically, demand-side factors determine job opportunities and objective working conditions, while supply-side elements shape workers’ competencies and employment experiences. The matching mechanism further impacts job benefits and career development through the alignment between individuals and positions. These three dimensions interact synergistically to redefine the quality standards of flexible employment.

3.1. Digital Economy, Industrial Structure, and Flexible Employment Quality

The digital economy drives industrial upgrading through technological integration and business model innovation, fundamentally reshaping labor demand. First, the service-oriented transformation of manufacturing and digitalization of service sectors drive labor migration toward high-skilled manufacturing and modern service industries. Second, industrial upgrading has intensified the demand for highly skilled and digitally literate workers, leading to labor allocation concentration in technology-intensive sectors [20]. Third, the burgeoning gig economy on internet platforms has given rise to numerous hybrid business models [21].

Industrial restructuring fundamentally shapes the quality of flexible employment across two key dimensions: work environment and risk exposure. Technologically advanced industries and modern service sectors have not only dramatically expanded flexible employment opportunities but also boosted labor productivity and corporate profitability. This empowers employers to offer competitive compensation packages, strengthen protections for basic rights, and establish standardized employment protocols, thereby driving systemic improvements in the labor market. Meanwhile, the proliferation of digital platform-based new industries and professions has created a paradox: employer-centric employment models shift operational uncertainties to workers, heightening risk exposure for flexible workers and ultimately constraining the quality of flexible employment.

3.2. Digital Economy, Human Capital, and Flexible Employment Quality

The digital economy has profoundly reshaped the mechanisms of human capital formation and accumulation. First, the development of the digital economy has reduced the direct, opportunity, and psychological costs of human capital investment and optimized the cost–benefit structure of such investment [22]. Second, flexible workers continuously enhance their digital literacy and professional skills through order acceptance, interaction, and feedback, achieving dynamic human capital upgrading through “learning by doing.” Third, the digital economy expands demand for high-skilled positions, reinforcing the returns to human capital and further motivating workers to actively participate in skills training and lifelong learning.

The accumulation of human capital primarily influences the quality of flexible employment in two ways: by enhancing workers’ individual job competencies and by shaping their subjective experiences. Regarding job competencies, workers with higher digital literacy and professional skills possess stronger job capabilities and bargaining power in platform markets. They are capable of undertaking high-value, high-complexity tasks and achieving superior returns on income. Picatoste et al. (2023) [23] indicate that proficiency with digital tools and platform rule adaptation has become a critical factor in improving employment quality. Subjectively, improved skill levels enhance job autonomy and a sense of control, reducing anxiety and frustration caused by inadequate capabilities, thereby boosting subjective job satisfaction. Meanwhile, as skills accumulate and a professional reputation is established, flexible workers gradually transition from temporary gig workers to independent professionals, with their sense of professional experience, social recognition, and confidence in long-term development significantly strengthened.

3.3. Digital Economy, Matching Efficiency, and Flexible Employment Quality

The digital economy has fundamentally restructured labor market mechanisms for job search and matching, significantly enhancing the efficiency of supply–demand coordination. By leveraging big data and algorithmic technologies, digital platforms effectively connect previously fragmented and asymmetrically informed job seekers with employers. This innovation enables workers to access employment information at reduced costs, thereby improving the accessibility and matching efficiency of flexible employment opportunities [24]. Second, digital platforms aggregate large volumes of job listings in real time, delivering personalized recommendations based on users’ historical behavior and skill profiles. This enables workers to quickly identify high-quality opportunities that match their retention wages, thereby reducing structural mismatches caused by information asymmetry. Thirdly, the digital economy has significantly expanded market boundaries and increased job supply. Leveraging algorithmic technologies, it enables real-time, precise matching between job seekers and employers, substantially improving matching efficiency [25].

Efficient matching of supply and demand has significantly improved the welfare benefits and career development opportunities for flexible workers. Regarding job benefits, positions with high matching rates typically offer higher labor compensation, thereby enhancing their bargaining power [26]. Regarding career development, the quality of matches directly affects job satisfaction and career identity. When actual positions align with or exceed search expectations, individuals perceive higher returns and stronger psychological contracts. Platforms also provide highly skilled workers with priority access to orders, credit incentives, or exclusive protections, thereby further reinforcing their job security and career prospects.

Therefore, the following research hypothesis is proposed:

H1: 

The digital economy contributes to improving the quality of flexible employment.

H2: 

The digital economy enhances the quality of flexible employment through industrial structure optimization, human capital accumulation, and improved matching efficiency.

4. Method and Data

4.1. Empirical Method

To examine the impact of the digital economy on flexible employment quality, the baseline model is set as follows:

$${qfe}_{ijt}={\alpha }_0+{\alpha }_1{deindex}_{jt}+{\alpha }_2{Z}_{ijt}+{\mu }_i+{year}_t+{area}_j+{\epsilon }_{ijt}$$

(1)

In Equation (1), qfe_ijt indicates the quality of flexible employment for individuals i in region j and time t; deindex_jt represents the level of digital economic development during a certain period; Z_ijt represents a series of control variables; µ_i, year_t, and area_j represent the individual fixed effect, the year fixed effect, and the province fixed effect, respectively; ε_ijt is the random error term; and α₀, α₁, α₂ are the model’s parameters to be estimated.

4.2. Variable Selection and Description

4.2.1. Dependent Variable: Flexible Employment Quality (qfe)

Using CFPS data and following the methodologies outlined by Myhill et al. (2021) [27] and Berg et al. (2023) [28], we constructed a flexible employment quality index system comprising 20 indicators across five dimensions: work environment, welfare benefits, development potential, subjective experience, and risk exposure (Figure 2; see Appendix A 

Table A1. The flexible employment quality measurement index system.

Primary Indicator	Secondary Indicator	Calculation Method	Direction
Work environment	Career opportunities	How many other jobs have you held in total (0–30)?	Forward direction
	Income from work	Logarithm of total annual income (RMB).	Forward direction
	Working strength	Weekly working hours (hours).	Negative direction
		One-way commute time (minutes) (0–240 min).	Negative direction
	Job security	Where is the main workplace for this job (flexible = 1; inflexible = 0)?	Forward direction
	Job stability	Is a labor contract signed? Yes = 1; No = 0.	Forward direction
Welfare benefits	Security status	What types of insurance are covered by this job? Pension, medical, unemployment, work-related injury, and maternity insurance each count for 1 point, with none being 0.	Forward direction
	Cash benefit	What cash benefits does this job offer? Transportation, meal, housing, and holiday benefits each count as 1 point, with none counted as 0.	Forward direction
	Benefit in kind	This job offers benefits in kind: free breakfast, lunch, and dinner; free accommodation; company-provided car or shuttle; and shopping cards or vouchers. Each item counts as 1 point, with no 0 points awarded.	Forward direction
Development potential	Career advancement	Which of the following promotions is awarded for this job? Administrative promotion or technical title promotion is assigned a value of 1, and neither is assigned a value of 0.	Forward direction
	Accumulation of experience capital	Work experience and income interaction term.	Forward direction
Subjective experience	Job satisfaction	How satisfied are you with this job? 1. Very dissatisfied; 2. Not very satisfied 3. Average; 4. Relatively satisfied; 5. Very satisfied.	Forward direction
	Job status	Is your income in the local area? 1 indicates very low, 5 indicates very high, and values range between 1 and 5.	Forward direction
		How would you rate your social standing locally? 1 indicates very low, 5 indicates very high, and values range between 1 and 5.	Forward direction
	Evaluation	Life satisfaction: 1 means very dissatisfied, 5 means very satisfied, and scores range from 1 to 5.	Forward direction
		Confidence in one’s future: 1 indicates no confidence, 5 indicates very confident, and values range between 1 and 5.	Forward direction
Risk exposure	Job uncertainty	Expected probability of unemployment.	Negative direction
	Uncertainty in healthcare spending	Sudden health expenditure.	Negative direction
	Commercial insurance uncertainty	Commercial paper face value.	Negative direction

for details). The entropy method was employed to measure the current quality of flexible employment.

4.2.2. Core Explanatory Variables: Digital Economy (Deindex)

Based on the measurement approaches to the digital economy proposed by Xiang et al. (2025) [29], we constructed a Digital Economy Development Index by selecting 19 indicators across four dimensions: digital infrastructure, digital industrialization, industrial digitalization, and digital transaction development (see Appendix B 

Table A2. Digital economy measurement index system.

Primary Indicator	Secondary Indicator	Tertiary Indicator	Unit	Attribute	Data Sources
Digital infrastructure	Network-based coverage	Optical cable line density	km/square km	Forward direction	State Statistical Bureau
		Mobile phone base station density	Per square kilometer	Forward direction
		Proportion of administrative villages with Internet broadband services	%	Forward direction
	User access capability	Mobile penetration rate	Department/100 people	Forward direction
		Internet users as a percentage of the Permanent population	%	Forward direction
		Per capita telecom traffic	CNY/person	Forward direction
Digital industrialization	Size of the core digital industry	Software revenue	100 million	Forward direction
		Information transmission, software, and information technology service industry Employees in urban units	Thousands of people	Forward direction	EPS database
	Digital innovation capability	Full-time equivalent of r&d staff in high-tech enterprises	Per person/year	Forward direction
		Number of granted patent applications	Piece	Forward direction
Development of digital transaction	Digital transaction basis	Number of domain names	Thousands	Forward direction
		Number of websites owned by the Enterprise	Individual	Forward direction
		Ipv4 address count	Thousands	Forward direction	EPS database
		Number of web pages	Thousands	Forward direction	State Statistical Bureau
	Digital transaction effect	Number of enterprises engaged in e-commerce transactions	Individual	Forward direction	EPS database
		E-commerce sales	100 million	Forward direction
		Per capita express delivery volume	Items/person	Forward direction
Industrial digitization	Industrial convergence empowerment	Total index of integration of informatization and industrialization	/	Forward direction	Blue Book on the Evaluation of the Integration and Development Level of Informatization and Industrialization in China
	Comprehensive development level	Digital inclusive finance composite index	/	Forward direction	Peking University Digital Inclusive Finance Index

). The comprehensive index was measured using the entropy method.

4.2.3. Mediating Variables

Industrial Structure

On the one hand, following the methodology of Zhang et al. (2021) [30], the ratio of employment in technology-intensive industries to total employment is used to measure the trend of industrial structure upgrading, thereby forming an industrial structure index. Based on the National Economic Industry Classification of the National Bureau of Statistics of China (GB/T4754−2011) [31], industries are categorized into labor-intensive, capital-intensive, and technology-intensive sectors (industry1) (see Appendix C 

Table A3. Classification of national economic industries.

Industry Type	Industry Category
Labor-intensive	Agriculture, forestry, animal husbandry and fishery; construction; wholesale and retail trade; accommodation and food services; labor-intensive manufacturing industries (including food processing, beverage manufacturing, tobacco products, textiles, textile apparel, footwear and hat manufacturing, leather, fur, feather and down products, wood processing, bamboo, rattan, palm and straw products, furniture manufacturing, paper and paper products, printing and recording media reproduction, cultural, educational and sports equipment manufacturing, rubber products, plastic products, non-metallic mineral products, waste resource and recycled material recycling, handicrafts, and other manufacturing sectors)
Capital-intensive	Warehousing and postal services, real estate, capital-intensive manufacturing (petroleum processing, coking and nuclear fuel processing, chemical raw materials and chemical products manufacturing, chemical fiber manufacturing, ferrous metal smelting and rolling, non-ferrous metal smelting and rolling, metal products manufacturing, general equipment manufacturing, specialized equipment manufacturing, transportation equipment manufacturing, electrical machinery, and equipment manufacturing)
Skill-intensive	Modern service industries (including information transmission, software and IT services, finance, leasing and business services, scientific research and technical services, education, healthcare and social work, culture, sports and entertainment, water conservancy, environmental and public facility management, public administration, social security and social organizations, international organizations, and other service sectors), as well as technology-intensive manufacturing industries (such as pharmaceutical manufacturing; production of communication equipment, computers, and other electronic devices; and manufacturing of instruments, meters, and cultural and office machinery).

). On the other hand, following the approach provided by Chen and Yang (2021) [32], the industrial structure level coefficient is used to illustrate the changes in the industrial structure of each province, with the calculated rule as follows:

$$industry2=\sum _{i=1}^3{q}_i\times i=q_1\times 1+q_2\times 2+q_3\times 3$$

(2)

In Equation (2), $q_i$ represents the proportion of the output value of industry i.

Human Capital

First, drawing on the approach highlighted by Sun (2026) [33], we use two dimensions—years of education and digital literacy level—to measure the accumulation of human capital among flexible workers and employ an equal-weight method to synthesize years of education with digital literacy-related indicators into a comprehensive index (human1 see Appendix D 

Table A4. Measurement index system of digital literacy.

Primary Indicator	Secondary Indicator	Tertiary Indicator
Tool Literacy	Use the device to access the internet	Do you use a mobile device to access the internet?
		Do you use a computer to access the internet?
Technology Application Literacy	Do you have digital transaction skills?	Do you shop online?
	Do you have digital entertainment skills?	Online entertainment
	Do you have digital learning skills?	Is it online learning?
Information Literacy	Cognition of network value	The importance of the network in daily life

). Second, the per capita human capital data released by the Center for Human Capital and Labor Economics (CHLR) at the Central University of Finance and Economics are used as an additional variable to characterize the level of human capital (human2).

Matching Efficiency (Match)

Drawing from the methodology outlined by Wang et al. (2025) [34], the matching efficiency index is constructed from three dimensions: reducing search cost, reducing mismatch degree, and improving search efficiency, and the comprehensive matching efficiency index is synthesized (match1). To validate the robustness of this mechanism, adopting the methodology of Bai and Lin (2025) [35], we use years of education to represent skill levels. The deviation between workers’ skill levels and the average of their occupations and annual groups is calculated; an absolute negative value indicates skill-matching efficiency (match2).

4.2.4. Controlled Variables

Combining theoretical analysis with existing research findings, the control variables are structured across three dimensions: first, at the individual level—age (age), gender (gen), household registration (hukou), health (health), marriage (marriage), and communist (communist); second, at the family level—household size (fml), child dependency ratio (childr), elderly support ratio (elderr), household savings (sav), informal financial obligations (unfinance), and bank loans (bankloan) (The specific explanations for the control variables are as follows: age (18–65 years); gender (female is coded as 0; male is coded as 1); household registration (non-agricultural is coded as 0; agricultural is coded as 1); health status (1–5; higher values indicate better health); marital status (unmarried is coded as 0, married is coded as 1); political affiliation (non-party member is coded as 0; party member is coded as 1). Family size refers to the total household population, whereas the child dependency ratio measures the proportion of children under 15 in the total household population. The elderly support ratio is the proportion of seniors aged 65+ in the total family population. Household savings represent the total cash and deposit balance (in CNY). Informal financial sources include outstanding loans from relatives or informal lenders, and bank loans include any outstanding bank loans); third, at the regional level—the provincial unemployment rate (unemployment) and the provincial economic density (ecodensity).

4.3. Data Sources and Descriptive Statistics

The data on the digital economy, industrial structure, and provincial unemployment rates in this paper are sourced from the “China Statistical Yearbook”, “Statistical Bulletin of National Economic and Social Development”, EPS database, the National Bureau of Statistics, “China Informatization and Industrialization Integration Development Level Assessment Blue Book”, and Peking University Digital Inclusive Finance Index. Micro-level data are derived from five consecutive surveys conducted by Peking University’s CFPS between 2014 and 2022. Specifically, indicators related to employment quality, age, gender, education level, marital status, and health conditions are obtained from personal questionnaires. The child dependency ratio, elderly support ratio, and household population size are sourced from family relationship questionnaires, while household deposit balance, informal lending, and bank loans are derived from household economic questionnaires. Human capital data are obtained from the “China Human Capital Report (2024)” by the Center for Human Capital and Labor Economics (CHLR) at Central University of Finance and Economics.

Table 1. Descriptive statistics of primary variables.

Type of Variable	Variable Name	Symbol	No.	Mean	Standard Deviation	Lowest Value	Crest Value
Explained variable	Quality of Flexible Employment	qfe	16,281	0.138	0.103	0.009	0.748
Explanatory variable	Digital Economy	deindex	16,281	0.207	0.133	0.042	0.613
Individual-level control variables	Age	age	16,281	38.19	11.884	18	65
	Sex	gen	16,281	0.552	0.497	0	1
	Hukou	hukou	16,281	0.795	0.404	0	1
	Political Status	communist	16,281	0.03	0.17	0	1
	Health	health	16,281	3.304	1.096	1	5
	Marriage	marriage	16,281	0.759	0.428	0	1
Household-level control variables	Size of Family Population	fml	16,281	4.108	2.018	1	15
	Child Eependency Ratio	childr	16,281	0.307	0.424	0	6
	Age Dependency Ratio	elderr	16,281	0.127	0.307	0	3
	Savings	sav	16,281	65,929.133	197,303.42	0	7,000,000
	Informal Finance	unfinance	16,281	0.137	0.344	0	1
	Bank Advance	bankloan	16,281	0.109	0.311	0	1
Regional characteristic control variable	Jobless Rate	unemploy	16,281	3.153	0.642	1.35	5.15
	Urban Economic Density	codensity	16,281	5244.769	11,808.369	60.311	76,987.586
Metavariable	Industrial Structure	Industrial1 Industrial2	16,281 16,281	0.109 242.735	0.068 9.544	0.024 222.5	0.264 283.8
	Human Capital	Human1 Human1	5976 16,281	0.065 1.924	0.466 0.834	1.723 0.993	1.124 5.302
	Atching Efficiency	Match1 Match1	14,963 14,963	0.014 0.771	1.493 0.620	8.714 4.994	5.988 0

presents the descriptive statistical results of the main variables. The mean value of flexible employment quality was 0.138, indicating that the overall quality of flexible employment remains at a relatively low level. The standard deviation of the digital economy index was 0.133, indicating substantial disparities across provinces. The sample primarily consisted of middle-aged, married individuals with agricultural household registration. In accordance with the research requirements and referencing the approach of Yang et al. (2023) [36], the study scope was set to non-agricultural samples. Additionally, formal employment samples were excluded based on “whether pension insurance was paid” and “whether a labor contract was signed”, retaining only samples aged 18–65 with no missing indicators. Ultimately, 16,281 observations were selected.

5. Empirical Results

5.1. Baseline Results

Table 2. Benchmark regression results.

Variables	(1)	(2)	(3)	(4)
	OLS	FE	RE	MFE
deindex	0.033 ***	0.060 **	0.044 ***	0.163 **
	(4.34)	(2.27)	(5.70)	(2.44)
age	−0.001 ***	0.004 ***	−0.001 ***	−0.001
	(−14.23)	(6.28)	(−13.03)	(−0.22)
gen	0.008 ***	−0.004	0.008 ***	−0.003
	(4.97)	(−0.11)	(4.42)	(−0.10)
hukou	−0.021 ***	−0.002	−0.020 ***	−0.003
	(−9.86)	(−0.30)	(−9.20)	(−0.59)
communist	0.027 ***	0.039 **	0.028 ***	0.038 **
	(5.38)	(2.28)	(5.89)	(2.06)
health	0.002 ***	0.003 *	0.003 ***	0.003 *
	(3.41)	(1.84)	(3.50)	(1.85)
marriage	0.009 ***	0.011	0.010 ***	0.011
	(3.90)	(1.62)	(4.15)	(1.49)
fml	−0.000	0.002 *	−0.000	0.001
	(−0.33)	(1.68)	(−0.83)	(1.14)
childr	0.000	−0.001	0.001	−0.002
	(0.09)	(−0.25)	(0.39)	(−0.51)
elderr	−0.005 *	−0.008	−0.004	−0.008
	(−1.95)	(−1.49)	(−1.42)	(−1.46)
sav	0.000 ***	0.000 **	0.000 ***	0.000 **
	(5.87)	(1.99)	(9.56)	(2.06)
unfinance	−0.007 ***	−0.000	−0.006 **	−0.000
	(−3.00)	(−0.04)	(−2.41)	(−0.00)
bankloan	0.025 ***	0.006	0.023 ***	0.005
	(9.22)	(1.44)	(8.77)	(1.08)
unemploy	−0.002	0.001	−0.001	−0.000
	(−1.29)	(0.31)	(−0.83)	(−0.07)
ecodensity	0.000 ***	−0.000	0.000 ***	−0.000
	(5.96)	(−0.51)	(6.14)	(−0.49)
_cons	0.169 ***	−0.059 *	0.165 ***	0.127
	(24.29)	(−1.81)	(23.06)	(0.61)
individual fixation		YES		YES
Fixed time		NO		YES
Province fixed		NO		YES
Obs.	16,281	16,281	16,281	9154

Notes: The values in parentheses represent robust standard errors, with *, **, and *** indicating p-values less than 0.10, 0.05, and 0.01, respectively.

regression results demonstrate that the digital economy significantly enhances flexible employment quality across four models: mixed OLS, individual FE, random RE, and high-dimensional FE. All models incorporate control variables at individual, household, and regional levels, confirming Hypothesis 1: digital economic development substantially improves flexible employment quality. The digital economy provides flexible workers with expanded job opportunities, more flexible work arrangements, and fairer income distribution mechanisms. The positive regression coefficient for Communist Party membership aligns with the research of Chen et al. (2025) [37], as party members typically possess stronger organizational coordination capabilities, broader social networks, and policy access advantages—factors that enable them to seize opportunities in digital economy-driven flexible employment markets. The positive coefficient on health status indicates that health is a core labor-supply factor influencing the quality of flexible employment. Given the inherent challenges of flexible employment—including fluctuating work intensity and inadequate social security coverage—healthy workers demonstrate greater adaptability to flexible work rhythms and stable income streams, thereby improving employment quality. This aligns with the core proposition of labor market theory that “health capital constitutes a vital component of human capital.” Furthermore, the model’s goodness of fit improved from 0.046 in column (1) to 0.585 in column (4), demonstrating that the multi-dimensional fixed-effects model effectively captures unobservable heterogeneity. This further underscores the significant positive impact of the digital economy on the quality of flexible employment.

5.2. Endogeneity Test

Although the benchmark regression model has made every effort to control for key variables affecting flexible employment quality—including demographic characteristics, household characteristics, and regional contexts—it may still exhibit endogeneity. To enhance the credibility of the results, we first adopt the research methodology of He and Song (2020) [38] by selecting the interaction term between the spherical distance from provincial capitals to Hangzhou and the national internet user count from the previous year as the first instrumental variable (IV1). Hangzhou is a hub city for China’s digital economy, hosting leading digital platform enterprises such as Alibaba and Ant Group, serving as a core source of digital technology R&D, business model innovation, and digital resource diffusion. The closer the spherical distance between provincial capitals and Hangzhou, the easier it is to accommodate digital technology spillovers and attract digital industry transfers. The completeness of digital infrastructure construction and the level of digital economy development are also relatively higher, meeting the correlation requirements of instrumental variables. Meanwhile, the spherical distance between provincial capitals and Hangzhou is an inherent geographic variable that does not directly affect the quality of flexible employment but only indirectly influences it by promoting local digital economy development, indicating strong exogenous characteristics. Simultaneously, drawing on the approaches of Nunn and Qian (2014) [39], we employ the interaction terms between the number of post offices per million population in each province in 1984 and the national internet port count from the previous year, as well as the interaction terms between the number of post offices per million population in 1984 and the national internet user count from the previous year as the second (IV2) and third (IV3) instrumental variables, respectively. The application and development of digital technologies rely on improved postal and telecommunications infrastructure. Communication methods in workers’ regions in the past may influence local digital technology development from various perspectives, thereby meeting the correlation requirements of instrumental variables. Meanwhile, the number of post offices in 1984 was determined by historical administrative planning and does not directly impact the quality of flexible employment at present. It can only indirectly affect the quality of flexible employment through its influence on the construction of information infrastructure, indicating high exogeneity. Additionally, to align with panel data with provincial and temporal dimensions, the instrumental variables described above were constructed by multiplying the national internet port count and the internet user count from the previous year. These two indicators are exogenous at the macro level and exhibit annual changes primarily driven by national factors such as digital infrastructure construction plans and technological iterations. They do not depend on the socio-economic characteristics of individual provinces and do not directly affect the quality of flexible employment.

Table 3. The results of the instrument variable test.

Variables	IV1		IV2		IV3
	deindex	qfe	deindex	qfe	deindex	qfe
Instrumental variable	0.011 *** (9.44)		0.001 *** (15.63)		0.001 *** (9.86)
deindex		0.185 *** (8.32)		0.641 *** (8.41)		0.878 *** (6.48)
_cons	0.423 *** (28.74)	0.105 *** (9.31)	0.372 *** (46.33)	−0.087 *** (−2.63)	0.385 *** (46.82)	−0.186 *** (−3.24)
Controlled variable	YES	YES	YES	YES	YES	YES
Individual fixation	YES	YES	YES	YES	YES	YES
Fixed time	YES	YES	YES	YES	YES	YES
Province fixed	YES	YES	YES	YES	YES	YES
Anderson canon. corr.LM statistic	1784.687		211.614		85.910
Cragg–Donald Wald F-statistic	667.396		214.190		86.281
Obs.	16,281	16,281	16,281	16,281	16,281	16,281

Notes: The values in parentheses represent robust standard errors, with *** indicating p-values less than 0.01.

presents the instrumental variable test results. All three instrumental variables demonstrated statistically significant positive coefficients at the 1% level in the first-stage regression, with F-values exceeding the weak instrumental variable threshold of 10, confirming the validity of the instrumental variable selection. The second-stage regression showed that the coefficients for the digital economy remained significantly positive at the 1% level, consistent with the baseline regression results. This indicates that levels of digital development have a significant positive impact on the quality of flexible employment, a conclusion that holds across different instrumental variable strategies, demonstrating both credibility and robustness.

5.3. Robustness Test

First, the estimation model was updated. Given that the dependent variable ranges from 0 to 1, we re-estimated using the Fractional Logit Quantile Response Model, which better accommodates the distribution characteristics of the constrained dependent variable and avoids prediction values exceeding the range that may occur with linear models. The regression results showed a significantly positive coefficient for the digital economy, as presented in column (1) of

Table 4. Robustness test results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Fractional Logit Model	Tobit Model	PCA Method for Measuring Digital Economy Index	Remove Outliers	Lagged Explanatory Variable	Relaxing the Age Boundary
dy/dx Deindex	0.283 *** (4.39)	0.041 *** (5.43)	0.035 *** (2.76)	0.164 ** (2.45)	0.178 *** (2.69)	0.174 *** (2.64)
Pseudo r2/r2	0.0049		0.585	0.585	0.585	0.586
sigma_u		0.051 *** (35.18)
sigma_e		0.087 *** (105.51)
Controlled variable			YES	YES	YES	YES
Individual fixation			YES	YES	YES	YES
Fixed time			YES	YES	YES	YES
Province fixed			YES	YES	YES	YES
Obs.	16,281	16,281	9154	9154	9154	9282

Notes: The values in parentheses represent robust standard errors, with **, and *** indicating p-values less than 0.05, and 0.01, respectively.

. Additionally, a two-truncated Tobit model was employed for robustness testing, thereby mitigating estimation bias arising from the constrained dependent variable. The test results confirmed the positive coefficient for the digital economy, as shown in column (2) of Table 4.

Second, the empowerment approach of the digital economy was changed. Using Principal Component Analysis (PCA) to recalculate the comprehensive index of the digital economy, we re-examined the effect of the digital economy on the quality of flexible employment. The test results are significantly positive, as shown in column (3) of Table 4.

Third, outliers were removed. Using the winsorization method, the 1st and 99th percentiles of continuous variables in the flexible employment quality indicator system were subjected to bilateral truncation. This approach retains all samples, avoids potential sample selection bias and information loss caused by outliers, and mitigates the potential interference of extreme values on estimation results. Column (4) of Table 4 shows that the test result is significantly positive.

Fourth, the explanatory variable was lagged by one period. Following the relevant methodology [40,41,42,43], we employed data with the core explanatory variable lagged by one period to test robustness and demonstrate the time-accumulation effect of digital economy development on flexible employment quality. The results are statistically significant, as shown in column (5) of Table 4.

Fifth, the age range was expanded. Given the informal and inclusive nature of flexible employment, some participants under 18 or over 65—such as dropouts in e-commerce or live streaming, and seniors in community services or street vending—were included. Thus, the sample age range was extended from the baseline 18–65 to 16–75, with statistically significant results, as shown in Table 4, column (6). These findings align with the baseline regression, confirming the digital economy’s positive impact on the quality of flexible employment, which is not influenced by model specifications, indicator construction, or sample selection methods, thereby validating the robustness of the baseline results.

5.4. Heterogeneity Analysis

The digital economy’s impact on the quality of flexible employment is heterogeneous, varying significantly across individuals based on their personal traits, family backgrounds, and regional contexts. To explore this, we conduct group regression analyses across these three dimensions to comprehensively examine the heterogeneous effects of the digital economy on the quality of flexible employment.

5.4.1. Individual Level

Gender heterogeneity: When the sample is divided into female and male groups, regression analysis reveals a significantly positive coefficient for the digital economy in the male group, while the coefficient in the female group is insignificant (see

Table 5. Heterogeneity test.

Variables	(1)		(2)		(3)		(4)		(5)
	Sex Heterogeneity		Skill Heterogeneity		Age Heterogeneity		Heterogeneity of the Family Dependency Ratio		Regional Heterogeneity
	Female	Male	Low	High	36–65	18–35	Low	High	Northwest Side	Southeastern Side
Digital economy	0.011 (0.12)	0.301 *** (3.19)	0.186 *** (2.63)	0.066 (0.27)	0.104 (1.20)	0.274 ** (2.33)	0.187 *** (2.62)	−0.399 (−0.72)	9.456 * (1.85)	0.161 ** (2.38)
r2	0.579	0.594	0.569	0.617	0.579	0.594	0.586	0.704	0.593	0.586
Controlled variable	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Individual fixation	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Fixed time	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Province fixed	YES	YES	YES	YES	YES	YES	YES	YES	YES	YES
Obs.	4069	5070	7815	1145	4802	3840	8829	108	1022	7998

Notes: The values in parentheses represent robust standard errors, with *, **, and *** indicating p-values less than 0.10, 0.05, and 0.01, respectively.

, column 1). This disparity directly reflects the gender bias in the empowerment effects of the digital economy. The first possible reason is the inertia of the gender division of labor and insufficient policy support. Women often bear greater responsibility for family care, and in China’s existing childcare system, the care of children aged 0–3 is highly dependent on families, forcing many women with medium skills to engage in low-level, flexible jobs that require high flexibility and low continuity of time. The second reason is the implicit gender bias in platform algorithms. Algorithms often match positions and push business recommendations based on historical data, making it easier for experienced men with long tenure to obtain high-value-added jobs, while female flexible workers are at a competitive disadvantage due to data feedback, thereby falling into low-level flexible employment positions [44]. One more reason is that traditional occupational segregation persists. Although job postings do not specify gender requirements, extensive data reveals significant gender disparities in flexible employment sectors like ride-hailing services, food delivery platforms, and courier work, resulting in women’s voluntary exit from these fields.

Skill level heterogeneity: Using years of education as a proxy for workers’ skill levels, this study divides participants into high-skilled (≥15 years of education, i.e., bachelor’s degree or higher) and low-skilled (<15 years of education) groups. While “skill level” typically refers to work experience or professional expertise, micro-level survey data indicate that educational attainment remains the most stable and comparable indicator of human capital quality, closely linked to digital skill acquisition. Therefore, this research uses years of education as the skill stratification criterion to examine the differential impacts of the digital economy across human capital endowments. Regression results show that the digital economy coefficient is significantly positive in the low-skilled group, while remaining statistically insignificant in the high-skilled group (see Table 5, column 2). This outcome reflects a ceiling effect, likely because highly skilled workers already enjoy higher wages, greater job autonomy, and greater career flexibility, resulting in limited marginal gains from digital economy development. Traditional employment scenarios for low-skilled, flexible workers often present challenges, including unstable income and scarce job opportunities. The digital economy has significantly expanded the supply of low-skilled positions through platform-based models, while lowering barriers to employment and reducing information-search costs. This makes the digital economy a crucial pathway for low-skilled groups to improve their employment quality.

Age heterogeneity: The youth group (18–35 years) and middle-aged group (36–65 years) were defined accordingly. Regression analysis revealed a statistically significant positive coefficient for the youth group in the digital economy, whereas the middle-aged group’s coefficient did not meet the significance thresholds (see Table 5, column 3). This discrepancy may stem from differences in digital adaptability and employment compatibility. The youth, having grown up in the digital era, demonstrate higher digital literacy, more flexible employment perspectives, and quicker adaptation to emerging flexible employment models in the digital economy. Their strong risk tolerance and ability to benefit from empowerment also contribute to this advantage. In contrast, the middle-aged group faces challenges such as insufficient digital skills, rigid human capital structures, limited acceptance of new flexible employment forms, and high transition costs due to family responsibilities, which collectively prevent the positive empowerment effects of the digital economy from becoming statistically significant.

5.4.2. Family Level

Heterogeneity in family dependency ratios: Families were grouped by the number of children under 16 years: those with no children or only one minor child were classified as low childcare-burden groups, whereas families with two or more minor children were categorized as high-childcare-burden groups. This classification aims to capture the structural constraints on time allocation and work continuity in multi-child households. Regression analysis revealed that the digital economy coefficient was significantly positive in low-dependency-ratio groups, whereas the coefficient in high-dependency-ratio groups failed to pass significance testing (see Table 5, column 4). This discrepancy may stem from the constraining effect of family caregiving pressure on digital employment participation. Flexible workers in low-dependency-ratio households face lighter childcare burdens and more flexible time allocation, enabling them to fully engage in digital flexible jobs such as online customer service, remote data processing, and part-time design. These positions typically require continuous time availability and timely responsiveness. Reduced caregiving pressure enables workers to consistently secure employment opportunities in the digital economy, thereby improving job quality. In contrast, flexible workers in high-dependency-ratio households must balance work and childcare responsibilities, resulting in fragmented time schedules that struggle to meet the demands of digital jobs that require sustained focus and continuity. This forces them to shift to traditional forms of flexible employment with lower levels of digitalization and unstable income, such as street gig work and offline temporary services, thereby hindering the effective implementation of digital economy empowerment.

5.4.3. Regional Level

Dividing the sample along the Hu-Huan-Yong Line into Southeast and Northwest groups, regression analysis reveals regional disparities in the digital economy’s impact intensity, with significantly higher effects in the Northwest than in the Southeast (see Table 5, column 5). This may be attributed to the Northwest’s late-developed flexible employment market with low standardization, where policy interventions are in a “low base, high elasticity” phase, offering substantial potential for marginal improvement. In contrast, the Southeast’s market has been integrated into a mature system, resulting in diminishing marginal effects from existing regulations.

6. Further Discussion: Spatial and Temporal Heterogeneity

The benchmark regression and robustness tests demonstrate that the digital economy enhances the quality of flexible employment. Furthermore, this study examines the mechanisms and channels through which digital economic development affects the quality of flexible employment. A literature review and mechanism analysis indicate that the digital economy primarily influences the quality of flexible employment through three mechanisms: industrial structure, human capital, and matching efficiency. Building on this, this study adopts the two-step approach outlined by Jiang (2022) [45] and incorporates panel data characteristics to construct an intermediary testing model that controls for individual, time, and provincial fixed effects. The specific model specification is as follows:

$$\begin{array}{c}{industry}_{jt}={\beta }_0+{\beta }_1{deindex}_{jt}+{\beta }_2{Z}_{ijt}+{\mu }_i+{year}_t+{area}_j+{\epsilon }_{ijt}\\ {human}_{ijt}={\gamma }_0+{\gamma }_1{deindex}_{jt}+{\gamma }_2{Z}_{ijt}+{\mu }_i+{year}_t+{area}_j+{\epsilon }_{ijt}\\ {match}_{ijt}={\delta }_0+{\delta }_1{deindex}_{jt}+{\delta }_2{Z}_{ijt}+{\mu }_i+{year}_t+{area}_j+{\epsilon }_{ijt}\end{array}$$

(3)

In Equation (3), ${industry}_{jt}$ refers to industrial structure, ${human}_{ijt}$ refers to human capital, and ${match}_{ijt}$ refers to matching efficiency. ${\beta }_0$, ${\beta }_1$, ${\beta }_2$, ${\gamma }_0$, ${\gamma }_1$, ${\gamma }_2$, ${\delta }_0$, ${\delta }_1$ and ${\delta }_2$ are the coefficients of variables, and the meanings of the other variables are the same as those in Equation (1).

6.1. Industrial Structure Mechanism Test

The regression results in column (1) of

Table 6. The results of the intermediary effect and robustness test of the industrial structure.

Variables	(1)	(2)	(3)	(4)	(5)
	Industrial Structure	Heterogeneous Coefficient of Industrial Structure	Lagging First-Generation Digital Economy	The Measurement of the Digital Economy via PCA	Digital Economy Excluding the Outliers
deindex	0.375 ***	14.406 ***			0.375 ***
	(20.09)	(8.89)			(20.09)
deindex_lag			0.356 ***
			(19.46)
Deindex_pca				0.067 ***
				(18.12)
_cons	0.220 ***	224.925 ***	0.239 ***	0.223 ***	0.220 ***
	(4.81)	(35.63)	(4.94)	(4.83)	(4.81)
r2	0.951	0.953	0.951	0.951	0.951
Controlled variable	YES	YES	YES	YES	YES
Individual fixation	YES	YES	YES	YES	YES
Fixed time	YES	YES	YES	YES	YES
Province fixed	YES	YES	YES	YES	YES
Obs.	9154	9154	9154	9154	9154

Notes: The values in parentheses represent robust standard errors, with *** indicating p-values less than 0.01.

demonstrate that the coefficient of industrial structure is significantly positive at the 1% significance level, indicating that the digital economy exerts a significant influence on industrial structure development. To validate the robustness of the industrial structure mechanism, regression analysis was conducted using hierarchical coefficients of industrial structure, lagged core explanatory variables, change measurement methods, and outlier removal. As shown in columns (2)–(5) of Table 6, the regression results remain significantly positive at the 1% level, confirming the robustness of the conclusion that digital economic development enhances flexible employment quality through industrial structure adjustment.

The development of the digital economy has significantly driven the optimization of regional industrial structures. On the one hand, the industrial structure has shifted from industry-dominated to service-oriented [46]; on the other hand, technological progress is typically not factor-neutral but biased toward specific factors. While digital technologies have substantially replaced labor-intensive positions, they have simultaneously increased demand for technology and capital-intensive roles [47]. As Adam Smith observed, there is a close, intrinsic connection between the evolution of industrial structure and the deepening of the labor division, which can create more job opportunities for society. The Petty–Clark Theorem also reveals the general pattern of labor transfer with industrial restructuring: labor gradually shifts from the primary sector to higher-productivity secondary and tertiary industries. During this transition, industrial structures evolve from low-value-added to high-value-added and from labor-intensive to technology or knowledge-intensive sectors. The average employment quality of flexible workers varies significantly across industries, as shown in Figure 3. In knowledge-intensive service sectors such as scientific research, technical services, information technology, education, healthcare, and social work, the quality of flexible employment is generally higher. This is attributed to the high demands for professional skills, innovation capabilities, and digital literacy in these industries, which enhance practitioners’ bargaining power and career development prospects. Conversely, in traditional labor-intensive industries like manufacturing, construction, transportation, and hospitality, despite the large number of flexible workers, their average employment quality remains relatively low, reflecting the profound impact of industrial structure on the quality of flexible employment. The upgrading of the industrial structure will guide the redistribution of the labor force across industries, and the labor force will gradually shift from sectors with low productivity and low wages to those with high productivity and high wages, thereby improving the quality of employment.

6.2. Human Capital Mechanism Test

The regression results in

Table 7. Intermediary effect of human capital and robustness test.

	Human Capital	Labor Average Human Capital	Expand the Age Range	The Measurement of the Digital Economy via PCA	Digital Economy Excluding the Outliers
deindex	1.511 **	0.392 ***	1.494 **		1.527 **
	(2.40)	(5.56)	(2.37)		(2.43)
deindex2				0.311 **
				(2.62)
_cons	8.512	−0.733 **	8.508	8.077	8.518
	(0.95)	(−2.31)	(0.94)	(0.88)	(0.95)
r2	0.776	0.999	0.777	0.776	0.776
Controlled variable	YES	YES	YES	YES	YES
Individual fixation	YES	YES	YES	YES	YES
Fixed time	YES	YES	YES	YES	YES
Province fixed	YES	YES	YES	YES	YES
Obs.	1892	1892	1896	1892	1892

Notes: The values in parentheses represent robust standard errors, with **, and *** indicating p-values less than 0.05, and 0.01, respectively.

demonstrate that both human capital and labor per capita human capital coefficients are statistically significant at the 5% level. Robustness tests conducted by expanding the sample scope, modifying the estimation method, and excluding outliers consistently show positive significance at the 5% level, confirming the digital economy’s role in promoting human capital accumulation. It should be noted that the measurement of digital literacy within human capital variables was only included in the 2020 and 2022 questionnaires. Therefore, this section exclusively utilizes the 2020 and 2022 CFPS data. After applying a high-dimensional fixed effects model regression, the sample size was adjusted to 5976 observations.

Human capital theory posits that human capital—formed through education, skill development, and knowledge accumulation—is a key driver of productivity growth, employment opportunities, and income enhancement [48]. On the one hand, this theory suggests that quality education not only enriches workers’ knowledge and skills but also boosts their productivity and innovation potential, thereby securing higher wages in the labor market. Typically, highly educated individuals earn significantly more than their less educated counterparts [49]. Higher levels of education directly determine the job fit and income-generating potential of flexible workers [50], thereby improving their employment quality. On the other hand, the digital economy enhances digital literacy among flexible workers by reducing barriers to skill acquisition and accelerating technological adoption. This not only enhances work flexibility but also drives skill development, optimizes the employment environment, and improves job quality. Figure 4 illustrates the dynamic evolution of “internet usage” and employment quality among China’s flexible workforce from 2014 to 2022. The data show that the proportion of internet-connected flexible workers rose steadily from 68.09% in 2014 to 84.67% in 2022, reflecting the deep integration of digital technologies into the labor market. The quality of flexible employment among non-internet users has shown consistent annual improvement, yet remains significantly lower than that of internet users, with the gap maintained between 0.03 and 0.07. While the quality of flexible employment among internet users declined slightly from 2016 to 2018, it stabilized and rebounded after 2020, reaching 0.105 in 2022. This trend likely stems from the early stages of digital economy development, when internet access primarily served groups with higher human capital and technical adaptability, who predominantly engaged in technology-intensive or high-value-added emerging sectors, resulting in relatively higher-quality employment. However, with the expansion of digital infrastructure and the penetration of platform economies into lower-tier markets, internet access has rapidly expanded to the “long-tail population”—individuals with lower education levels and weaker skill reserves. Although this inclusive expansion has significantly broadened the reach of digital dividends and enhanced social equity, the initially low employment quality of these new groups temporarily reduced the overall average internet penetration rate. As platform regulations improve, digital skills training becomes more widespread, and workers’ adaptability strengthens, the quality of flexible employment among internet users has steadily improved. These findings demonstrate that digital literacy serves as a critical capability that links individuals to the digital economy and flexible employment [51].

6.3. Matching Efficiency Mechanism Test

Regression analysis in

Table 8. The mediating effect of matching efficiency and robustness testing.

Variables	Matching Efficiency 1	Matching Efficiency 2	Lagging First-Generation Digital Economy	Digital Economy Excluding the Outliers	Expand the Age Range
deindex	5.883 *** (7.92)	0.530 * (1.84)		5.883 *** (7.92)	5.978 *** (8.11)
deindex_lag			6.971 *** (8.67)
_cons	−5.881 ** (−2.00)	−0.288 (−0.27)	−6.142 ** (−2.09)	−5.881 ** (−2.00)	−5.620 * (−1.92)
r2	0.575	0.806	0.577	0.575	0.577
Controlled variable	YES	YES	YES	YES	YES
Individual fixation	YES	YES	YES	YES	YES
Fixed time	YES	YES	YES	YES	YES
Province fixed	YES	YES	YES	YES	YES
Obs.	8070	8070	8070	8070	8180

Notes: The values in parentheses represent robust standard errors, with *, **, and *** indicating p-values less than 0.10, 0.05, and 0.01, respectively.

reveals that both matching efficiency 1 and matching efficiency 2 demonstrate statistically significant positive coefficients at the 5% level. Robustness tests conducted by applying lagged core explanatory variables, removing outliers, and expanding the sample scope consistently show positive results at the 1% level. These findings indicate that the development of the digital economy has significantly improved human–job matching efficiency, with the results demonstrating robustness.

Enhancing the alignment between human capital and industrial structure can optimize the external employment environment, thereby laying institutional and market foundations for improving the quality of flexible employment. When regional human capital structures align closely with industrial technological demands, it effectively boosts total factor productivity and drives high-quality economic growth, thereby strengthening government fiscal capacity and social security provision. This enables more precise implementation of policy tools such as employment subsidies, occupational injury protection, and platform-based labor regulations, significantly improving income stability and risk resilience for flexible workers. Simultaneously, effective talent-industry matching promotes industrial upgrading to high-value-added segments, creating numerous new high-quality positions through digital platforms. Even under diversified employment relationships, workers can obtain relatively fair compensation, basic rights protection, and career development pathways, systematically alleviating the structural challenges of traditional flexible employment characterized by “low security and high vulnerability.”

Furthermore, the enhanced matching efficiency boosts individual employability and labor productivity, activating the endogenous growth momentum of flexible workers. On the one hand, the alignment between skills and job requirements significantly improves labor productivity, enabling workers to achieve higher output in shorter working hours, thereby reducing labor intensity and improving work–life balance. On the other hand, in regions with high matching efficiency, enterprises and platforms are more motivated to provide skills training, certification systems, and career development support, driving flexible workers to transition from passive order acceptance to active value creation. Coupled with the refinement of intelligent matching mechanisms, workers’ search costs decrease, cross-platform and cross-industry mobility increases, and professional capital accumulates continuously. This positive cycle of capabilities and opportunities not only elevates overall compensation but also strengthens career identity and subjective well-being, ultimately resulting in a substantial improvement in the quality of flexible employment.

7. Conclusions and Implications

7.1. Research Conclusions

Promoting the quality of flexible employment benefits workers’ employment and income growth, contributes to the common prosperity of society, and supports the sustainable development of flexible employment models. This article systematically analyzes the impact of the digital economy on the quality of flexible employment and the mediating mechanisms of industrial structure, human capital, and matching efficiency in this process. Empirical tests were conducted based on five consecutive periods of data from the China Family Panel Survey (CFPS) between 2014 and 2022. This study found that the digital economy has a positive, empowering effect on the quality of flexible employment. Among these, male, low-skilled, young, low-dependency-ratio, and flexible workers on the northwest side of the Hu-Huan-Yong Line are more likely to benefit from the positive empowering effects of the digital economy. The digital economy can indirectly improve the quality of flexible employment through a chain transmission mechanism: it first optimizes the industrial structure on the demand side, and then accumulates human capital on the supply side, and finally enhances supply–demand matching efficiency.

This study, grounded in China’s national conditions, develops targeted theoretical extensions and policy recommendations. Addressing the core debate about whether the digital economy enhances the quality of flexible employment, this research demonstrates that while digitalization increases risks in flexible employment, its overall empowering effects outweigh its inhibitory impacts. This validates the practical feasibility of improving the quality of flexible employment through digital economic development. To address the fragmented mechanism analysis in existing studies, this paper proposes a chain-based framework that integrates demand-side empowerment, supply-side enhancement, and supply–demand matching efficiency. This unified analytical framework combines three key pathways—industrial structure, human capital, and matching efficiency—addressing gaps in current research on mechanism explanations. Furthermore, by incorporating risk exposure dimensions into the flexible employment quality indicator system, this study not only reveals the dual role of digital development but also makes measurement more practical, thus providing a new theoretical perspective for research on flexible employment quality.

7.2. Policy Implications

First, we must fully leverage the digital economy’s positive impact on the quality of flexible employment, systematically enhancing its dignity and sustainability. By positioning digital economy development as a key strategy for stable employment and quality improvement, we should refine institutional frameworks supporting the healthy growth of platform economies. While encouraging innovation, we must strengthen the protection of workers’ rights and promote employer-independent social insurance systems, ensuring that social security benefits are portable and accumulative across platforms and contexts. It is essential to accelerate the establishment of algorithmic transparency standards, labor regulations, income floor requirements, and occupational safety benchmarks tailored to flexible employment, thereby transforming digital dividends into equitable employment-quality benefits. Through optimized platform governance, expanded social security coverage, and streamlined career development pathways, we will ensure the digital economy not only expands the employment scale but also substantially improves the sense of fulfillment, security, and career prospects for flexible workers.

Second, it is important to implement tiered support strategies that account for beneficiary disparities, empowering key groups while enhancing digital adaptation capabilities for disadvantaged populations. For groups demonstrating strong capacity to acquire digital dividends—such as males, youth, rural migrant workers, low-dependency-ratio households, and residents north of the Hu-Huan-Yong Line—it is crucial to consolidate their advantages through advanced skills training, entrepreneurship support, and investments in digital infrastructure in the northwest region. Simultaneously, policy should address structural barriers faced by non-key groups: develop flexible digital roles for female workers with time flexibility and family care responsibilities, such as remote customer service, content creation, and community e-commerce; create age-friendly digital tools and simplified interfaces for middle-aged and elderly groups, providing “digital companion” support mechanisms to reduce technical barriers; explore digital employment subsidies linked to childcare benefits for high-dependency-ratio households to alleviate concerns about flexible work participation; and in traditionally economically developed yet competitive regions like the southeast of the Hu-Huan-Yong Line, promote the integration of platform economies with local high-end services to guide flexible employment toward high-skilled, high-value-added sectors. Through differentiated and inclusive institutional designs, policy should ensure digital economic dividends reach broader worker groups while preventing the emergence of new “digital marginalized groups.”

Finally, we should strengthen three key channels—industrial structure optimization, human capital development, and supply–demand matching efficiency—to establish a sustainable mechanism for digital economy-driven quality improvement in flexible employment. On the demand side, we must deepen the integration of digital technologies with modern services and advanced manufacturing to create more high-quality flexible jobs. On the supply side, digital literacy and professional skills should be integrated into lifelong learning systems to support continuous human capital development among flexible workers. Meanwhile, we need to accelerate the development of intelligent public employment service platforms to break down information barriers between governments, platform enterprises, and job seekers, enabling efficient and precise matching of job requirements with individual capabilities. By advancing the tripartite reform of industry–talent–market synergy, we can effectively translate the digital economy’s potential advantages into an intrinsic driver of sustained quality improvement in flexible employment.

7.3. Limitations and Future Research Directions

This study develops an employment quality analysis framework suitable for flexible employment groups and assesses the impact of the digital economy on the quality of flexible employment, demonstrating some innovativeness and policy value; however, the analysis results have certain limitations. First, the development of China’s digital economy has distinct characteristics of policy guidance and platform dominance, forming a highly concentrated digital employment ecosystem represented by Alibaba, Meituan, and Tencent, which may limit the general applicability of the research results. Second, China’s labor market exhibits significant institutional segmentation features, including household registration barriers, a high proportion of flexible employment, and insufficient social security coverage. Under these circumstances, flexible employment is a survival strategy for technological empowerment, suggesting that the conclusion that low-skilled groups benefit more significantly, as reported in this paper, may stem from China’s national context rather than the inherent laws of the digital economy itself. Third, in recent years, China has prioritized stable employment in its policies, and the strong supportive role of policies has, to some extent, expanded the marginal utility of the digital economy’s impact on the quality of flexible employment. In summary, the conclusions of this study are primarily applicable to scenarios in which a large-scale, relatively centralized digital employment ecosystem exists, the labor market exhibits institutional segmentation, flexible employment is widespread, and the government actively promotes the standardized development of flexible employment through digital technology and policy support. Future research will incorporate cross-national comparisons to further test the robustness of the digital economy’s impact on the quality of flexible employment under different institutional conditions, thereby developing a more general theoretical framework.