In search of quasi-subordinate workers in China: A typology of gig riders by economic dependency and subordination

Wei TU; Xueyu WANG; Wei Tu; Xueyu Wang

doi:10.16995/ilr.23784

The employment status of platform workers has generated debate both in China and internationally. Drawing on legal thresholds from selected developed countries and statistical indicators developed by Eurofound and Eurostat, our study constructs context-specific indicators for identifying quasi-subordinate workers in the Chinese labour market. Between December 2021 and January 2022, we distributed online questionnaires in five Chinese cities to measure the subordination levels of 7,680 platform gig riders. Workers were classified into subtypes based on two dimensions: economic dependency and personal subordination. Our results indicate that only around 19 per cent of gig riders can be classified into the existing “employee versus independent worker” binary framework, while the rest should be grouped into a new “quasi-subordinate” worker category. Statistical analysis reveals significant differences in working conditions across subgroups. Our results suggest that high economic dependency is a predictor for longer working hours, greater work intensity and increased perceived pressure, while high personal subordination is related to low job satisfaction. Social insurance coverage is found to be particularly inadequate among subgroups with higher levels of subordination.

Keywords: platform economy, delivery platforms, gig workers, employment status, economic dependency, subordination, China

How to Cite: TU, W. & WANG, X. (2025) “In search of quasi-subordinate workers in China: A typology of gig riders by economic dependency and subordination”, International Labour Review. 164(4). doi: https://doi.org/10.16995/ilr.23784

Responsibility for opinions expressed in signed articles rests solely with their authors, and publication does not constitute an endorsement by the ILO.

This article is also available in French, in Revue internationale du Travail 164 (4), and Spanish, in Revista Internacional del Trabajo 144 (4).

1. Introduction

The classification of the employment status of digital platform gig workers has long been a bone of contention for labour market governance worldwide (Harris and Krueger 2015; Cherry and Aloisi 2017; Stewart and McCrystal 2019), prompting many countries to explore adjustments in their regulations. Some adhere to the binary “employees versus independent workers” model, while experimenting with more worker-friendly and stringent legal criteria to define independent worker status, as in the case of California Assembly Bill No. 5 in the United States (Palagashvili et al. 2024). Other countries are defining a new category of “employee-like” workers to extend labour protections to digital platform gig workers. This is the approach taken in Australia with the Fair Work Legislation Amendment (Closing Loopholes No. 2) Act 2024. Similarly, in 2021, the European Commission submitted a “Proposal for a Directive of the European Parliament and of the Council on Improving Working Conditions in Platform Work”, designed to facilitate accurate classification of platform workers’ employment status.¹

In China, the emergence of new forms of employment (NFE)² has become the most prominent and transformative trend in the domestic labour market in recent years. According to the ninth national workforce survey conducted by the All-China Federation of Trade Unions (2023), the number of NFE workers in China reached 84 million in 2023. However, owing to the unclear employment status of NFE workers in China’s labour law framework, this large group remains insufficiently protected (Wang 2016; Wang and Wang 2018). In order to ensure appropriate protection for NFE, while fulfilling the broader objectives of labour law – upholding workplace justice and advancing fundamental and universal human rights (Arthurs 2011) – the Ministry of Human Resources and Social Security (MOHRSS), together with eight other ministries and commissions, issued the “Guiding Opinions on Safeguarding the Labour Rights and Interests of Workers in New Forms of Employment” (MOHRSS 2021).³ This guidance introduces a new category of workers who “do not fully meet the current criteria for the identification of labour relations but are subject to enterprise-level management during the labour process”,⁴ thereby moving beyond the dichotomous “employee versus independent worker” framework in Chinese labour law and ushering in a trichotomy in worker identities. By introducing the concept of “quasi-subordinate” workers, the policy aims to prevent misclassification of NFE workers as self-employed workers and provide them with a basic level of protection. This encompasses fair employment practices, improved minimum wage and payment guarantees, reinforced occupational health and safety responsibilities, expanded basic pension and medical insurance coverage and stronger protections against occupational injury.⁵

Although this solution seeks to address social concerns over NFE, it also raises new questions. Currently, the Guiding Opinions (MOHRSS 2021) provide general principles for identifying quasi-subordinate workers but lack detailed guidance on implementation. Local governments in several provinces have expedited efforts to pilot policies and refine specific criteria. For example, in order to operationalize one of the core elements mentioned in the Guiding Opinions – “management over workers” – Jiangsu Province defines it as situations where “workers have strong autonomy in work arrangements, while working online under the constraints of platform rules or algorithms and receiving labour compensation”.⁶ Shanghai, on the other hand, stipulates that “the labour process of workers must comply with the algorithms and rules established by the platform companies.”⁷ Although these localized attempts make some progress towards a unified national framework, the identification of standardized criteria for “quasi-subordinate” workers remains an unresolved issue. Similarly, previous academic discussion of a trichotomous employment status structure has primarily focused on legal analysis and theoretical delineation (Wang 2016; Lu and Chen 2020; Xiao 2021; Ke 2022). While scholars have proposed assessing the employment status of NFE workers using a combination of personal subordination and economic dependency criteria (Ban 2017; Lu and Chen 2020; Wang and Wang 2018; Tian 2019; Xiao 2021; Yang 2022; Shen 2022), empirical validation for such an approach to regulatory reform remains limited and practical criteria have yet to be developed.

In order to address the remaining ambiguity in identifying, classifying and measuring “quasi-subordinate” workers, and to empirically examine their prevalence, working conditions and latent heterogeneity, our article proceeds as follows. First, in section 2, we examine the limitations of existing employment status identification methods in China and review international practice and policy innovations. On this basis, we construct indicators tailored to the Chinese context, drawing on the statistical measurements developed by international statistical institutions Eurofound and Eurostat. Although the diverse NFE workforce could be classified on the basis of case-by-case assessments, we believe that establishing a clear framework of criteria based on quantitative indicators would significantly improve the efficiency and accuracy of worker classification and group status assessments and facilitate large-scale statistical surveys. In section 3, we describe the application of the aforementioned indicators in a survey targeting online food-delivery gig riders,⁸ who provide a representative sample of NFE workers in China. By employing the typology established through latent class analysis (LCA), we identify different gig rider subtypes, estimate the proportion of quasi-subordinate workers within the Chinese gig rider group and characterize the subordination features and working conditions specific to these subtypes. Our discussion and conclusion, in section 4, highlight the policy implications of our findings and identify the relevance of our study for further research.

2. Construction of indicators for the identification of quasi-subordinate workers in China

2.1. NFE and the challenges to current approaches to identifying employment relationships in China

Currently, two approaches are used to determine the existence of employment relationships within China’s labour law system – both operating within a dichotomous framework. The first approach focuses on whether an employment contract has been signed between the employer and the worker. If this is the case and the worker has started to provide substantive work, an employment relationship is deemed to exist. In the absence of a signed contract, the second approach – the principle of the primacy of facts – is applied. This approach broadly aligns with the guidance provided in the ILO Employment Relationship Recommendation, 2006 (No. 198).⁹ It assesses whether the employment situation meets five indicators: (1) both the employer and the worker qualify as subjects of labour relations; (2) the work-related rules and regulations established by the employer are enacted in accordance with the law and are applicable to the worker; (3) the worker is subject to the management of the employer; (4) the worker is engaged in a remunerated labour arranged by the employing unit; and (5) the labour provided by the worker is an integral part of the employer’s business (MOHRSS 2005). Only when all five indicators are jointly satisfied can workers be regarded as direct employees; otherwise, they are considered to be independent contractors or self-employed workers. In the majority of industries in China, these two approaches generally yield accurate decisions in most labour dispute cases and have proven effective in judicial practice (Xie 2018; Wang and Wang 2018).

However, these established methods and specific indicators for determining labour relations come up against significant challenges in the case of NFE. In practice, the proportion of NFE workers who have signed employment contracts with either the platform or a third-party agency¹⁰ remains remarkably low. For instance, data from the Gig Economy Research Centre of the South China University of Technology reveal that only 15.12 per cent of online delivery riders have signed labour contracts with platforms or third-party human resources service agencies, and a mere 29.82 per cent have signed service contracts (Liu, Liu and Liu 2022). Similarly, data from the All-China Federation of Trade Unions indicate that the contract signing rate among NFE workers – such as delivery riders, couriers and e-hailing drivers – is only at 43 per cent (Yuan 2021).

It is difficult to confirm the existence of employment relationships between NFE workers and digital platforms through contract signing alone. Consequently, in the event of labour disputes, the dominant approach for assessing workers’ legal status relies on the principle of the primacy of facts. However, it is difficult for NFE workers to meet all five indicators simultaneously: digital platforms operate under a business model that differs from traditional hierarchical enterprises, such that they often deny that they are the employers responsible for work arrangements. Instead, they claim to be mere intermediaries, integrating and matching information and resources between the labour supply and demand sides (De Stefano 2016; Stewart and Stanford 2017). As a result, indicators (3), (4) and (5) are consistently the subject of controversy and debate in both academic and judicial circles.

Yet, categorizing NFE workers as self-employed is also inappropriate. Many scholars, drawing on labour process theory, argue that platforms exert substantive management and control over workers across various dimensions, including income, emotional dynamics and working time and intensity. This control is facilitated by invisible algorithmic technology and bureaucratic organizational systems (Chen 2022; Li and Jiang 2020; Wu and Li 2018; Zhao and Han 2021; Wu et al. 2019). Moreover, digital platforms routinely aggregate various types of workers simultaneously, including full-time and part-time (“crowdsourced”) workers (Zhao and Luo 2024; Sun, Chen and Rani 2023; Wang and Meng 2024), and even those “hobbyists” who work purely for fun and fresh life experience (Rosenblat 2018, 50–52), creating internal diversity and complexity that hinder consistent and unambiguous assessment. Therefore, both the rigorous “primacy of facts” principle and dichotomous classification prove inadequate when addressing the complexities of new employment patterns. This is reflected in judicial practice, where the number of labour dispute cases concerning the determination of labour relations for NFE workers has rapidly increased. For instance, the Shanghai Second Intermediate People’s Court published a White Paper that summarizes NFE-related labour dispute trial cases from 2017 to the first half of 2022.¹¹ The data show that while the number of cases has significantly risen since 2020, the determination of employment status remains the focal point of dispute, accounting for 45.78 per cent of all cases. Ambiguous legal identification makes it even more challenging to ensure the legal rights of workers and hinders their ability to obtain labour protection (Zhang 2019). Thus, regulation innovation and adjustment are imperative.

2.2. Identifying quasi-subordinate workers: International practice and theoretical foundation

The complexity of determining the employment status of NFE workers and the demand for new identification criteria are not new challenges. Before the emergence of the platform economy, several countries had already experimented with an intermediate “third category” of workers in their labour law frameworks. For example, in 1975, Canadian legislation introduced the concept of “dependent contractor” to describe small business owners who are economically dependent on a larger company, thereby extending collective bargaining rights to them (Arthurs 1965; Langille and Davidov 1999).¹² In 2007, the category of “economic dependent self-employed worker” (trabajador autónomo económicamente dependiente) was established in Spanish legislation,¹³ granting enhanced labour protections in response to the accelerated trend of companies hiring self-employed subcontractors (Sánchez Torres 2010; Cherry and Aloisi 2017). The German Equal Treatment Act 2006, recognizes “employee-like workers” (Arbeitnehmerähnliche Personen) as individuals who are not fully integrated into the organizations they work for.¹⁴ Although these workers are rarely subject to the direct control of the organization, the work provides them with their primary source of income, warranting protections similar to those afforded to standard employees (Däubler 1999; Sorge 2010; Wang 2017; Ban et al. 2023). The United Kingdom introduced the legal concept of “worker” in section 230 of the Employment Rights Act 1996, which includes not only all standard employees but also casual, freelance and self-employed individuals, granting them a subset of employee rights (Grimshaw et al. 2017).

In the context of NFE, the introduction of a trichotomous framework for worker status has sparked considerable discussion worldwide. For instance, to address the challenges faced by platform gig workers – who often fall into a legal grey area that bars them from comprehensive legal protections and sufficient individual bargaining power – US scholars have proposed the creation of a new “independent worker” category to provide specific welfare protections (Harris and Krueger 2015; Harris 2018). In addition, responding to evolving dynamics in the world of work, in 2018 the ILO revised its traditional dichotomous classification approach. The International Classification of Status in Employment (ICSE-93) standard had previously drawn a clear distinction between employees and self-employed workers (ILO 1993). The new standard introduced the new category of “dependent contractors” to describe workers “employed for profit, who are dependent on another entity that exercises control over their productive activities and directly benefits from the work performed by them” (ILO 2018). This dependency may be operational and/or economic.

In the light of these comparative legislative practices and policy proposals concerning the recognition of “third-category” workers, this article focuses on understanding how to identify and assess the legal status of such workers in practice and establish appropriate statistical criteria. By reviewing the classification criteria developed by international statistical institutions Eurofound and Eurostat and legal thresholds in selected developed countries, we find that – despite variations in specific indicators and quantitative boundaries – a general global consensus has emerged regarding the identification of “third-category” workers. This consensus centres on a combination of economic dependency and personal subordination (Davidov, Freedland and Kountouris 2015; Schoukens and Barrio 2017; ILO 2018; Bozzon and Murgia 2022).

Economic dependency refers to the degree to which workers rely on a single employer for financial support and the resulting vulnerability they face in their economic circumstances. It is considered to be high when the income earned from providing labour to an employer constitutes the worker’s sole or primary source of income (Schoukens and Barrio 2017), highlighting structural inequalities in the labour market (Xiao 2021; Davidov 2017). Although the legal frameworks for defining economically dependent workers vary across countries and are often fragmented or based on intuitive criteria, certain common elements can be identified (Rosioru 2014). At the empirical level, two main indicators are commonly used to measure economic dependency. The first is the “primary client” indicator. This concept was first introduced by French scholar Alain Supiot (2001), who defined economically dependent self-employed individuals as those working exclusively for a single dominant client. Similar approaches were subsequently adopted by the British Labour Force Survey, Eurofound and Eurostat (Bozzon and Murgia 2022). According to the European Industrial Relations Dictionary, “economically dependent workers” typically hold a sort of “service contract” with the employer and depend on a single employer for all or much of their income (Eurofound 2018). As a result of the “primary client” situation, the risk of business cooperation termination by the main client is significantly higher for economically dependent workers than for independent self-employed individuals (Bagari 2020). The second indicator – the revenue proportion indicator – derives mainly from legislative and judicial practice. Countries such as Germany and Spain have established clear quantitative thresholds in their labour legislation to determine economic dependency among self-employed individuals. In Germany, section 12a of the Collective Agreements Act (Tarifvertragsgesetz – TVG) stipulates that workers who primarily work for a single principal and derive more than 50 per cent of their total remuneration from that relationship meet the threshold for economic dependency. Notably, even if a worker provides services for several principals, they may still be classified as economically dependent if income from one legal relationship constitutes “the decisive means of existence” (Sorge 2010, 249). In Spain, the 2007 Self-Employed Workers’ Statute¹⁵ defines “dependent self-employed workers” as individuals who engage in economic or professional activities for remuneration, personally and primarily providing services to one person or organization and economically relying on that client for 75 per cent of their income (Sánchez Torres 2010).

The concept of personal subordination underscores the employer’s authority to direct, instruct, monitor and discipline workers in the labour process (Schoukens and Barrio 2017; Countouris 2019). It reflects the employer’s power to determine the nature of the work, the methods employed and the time and place of its execution (Zou 2017). Under this arrangement, the worker is expected to personally perform assigned tasks, without the use of agents or subcontractors, and is not allowed to provide services to other entities concurrently. Furthermore, the worker is required to work in the pursuit of the employer’s objectives, rather than their own business interests (Xiao 2021). Given that the labour process is subject to the employer’s supervision and control, workers occupy a relatively disadvantaged position compared with employers (Wang and Wang 2018; Xiao 2021; Fang 2022). Australia’s Fair Work Legislation Amendment (Closing Loopholes No. 2) Act 2024 is an example of applying the concept of personal subordination in determining employment status in NFE. The Act lists “a low degree of authority over the performance of the work” (15P(1)(e)(iii)) as one of the criteria for identifying “employee-like” workers in the digital platform work.

At the operational level, both Eurofound and Eurostat have conducted labour market surveys that include indicators for personal subordination. Both agencies developed distinct measurement indicators and they also adopt differing perspectives on the relationship between, and importance of, various subordination attributes (Eurofound 2017; Eurostat 2018; Williams and Horodnic 2018 and 2019). Eurofound primarily uses two indicators: (1) the authority to hire others, and (2) decision-making autonomy over specific tasks. In contrast, Eurostat relies on a single indicator: the right to control one’s working hours. In terms of the relationship between economic dependency and personal subordination, Eurofound adopts a more lenient approach. It does not assign different weights to the two types of indicators and considers individuals to fall within the “third category” of dependent workers if at least two of the following three indicators are met: (1) they work solely or primarily for one client; (2) they lack the authority to hire others; and (3) they lack autonomy in decision-making and discretion over specific work activities. By contrast, Eurostat applies a stricter definition, classifying individuals as dependent self-employed workers only if they meet both the economic dependency and personal subordination criteria (Bozzon and Murgia 2022).

2.3. Constructing context-specific indicators for identifying quasi-subordinate workers in China

Drawing on the aforementioned practices in other countries, this article constructs context-specific indicators for identifying quasi-subordinate workers in China under NFE (see table 1 for details). To measure economic dependency, we adopt the widely used revenue proportion indicator – applied in Germany and Spain – using a five-option question. Economic dependency is considered to exist when more than 50 per cent of a worker’s total earnings is derived from platform-based work.¹⁶ In order to assess personal subordination, we focus on control over working time and tasks, respectively. To measure riders’ autonomy over job tasks more comprehensively, we further decompose this indicator into three subdimensions: (1) task-content subordination; (2) decision-making subordination; and (3) autonomy in hiring versus personal performance of tasks. Each subdimension is presented as a categorical question, allowing respondents to choose between “Agree” or “Disagree”. An accumulative scoring method is used to describe the level of personal subordination, with each affirmative response assigned a score of 0 and each negative response a score of 1. The total score across the four indicators provides the overall degree of personal subordination.

Table 1

Indicators for identifying quasi-subordinated workers under new forms of employment

Indicator	Subindicator	Items
Economic dependence	–	The share of income obtained through the platform enterprise in the overall remuneration of individuals (less than 25%, 25–50%, 51–75%, 76–99%, 100%).
Personal subordination	Control of time	1. The ability to make autonomous decisions on daily online and offline time (time subordination).
Personal subordination	Control of tasks	2. The ability to choose independently which orders to accept (task-content subordination). 3. The ability to handle unexpected situations and resolve issues on their own during the labour process (decision-making subordination). 4. The ability to hire someone else to perform the task on their behalf (personal performance).
Note: In the questionnaire, the items are asked in plain language and adjusted to suit the daily work scenario of food-delivery gig riders (e.g. “Do you agree with the statement ‘I can have someone else deliver the orders for me’?”). Source: Our own compilation.

When interpreting our results, we classify NFE workers exhibiting strong indicators of both economic dependency and personal subordination as employees, whereas individuals who fulfil only a subset of the criteria for either personal subordination or economic dependency are designated as quasi-subordinate workers. At the same time, we maintain an open and exploratory attitude toward the potential internal heterogeneity among quasi-subordinate workers.

3. Quasi-subordinate workers: Typology, characteristics of subordination and working conditions

3.1. Sample, data collection and measurement

We selected food-delivery gig riders as a representative NFE group, given its size and rapid growth. For instance, in 2023, over 7.45 million riders earned income through Meituan, a leading digital food-delivery platform in China. This was a 19.4 increase from 6.24 million riders in 2022, and a 41.4 per cent increase from 5.27 million riders in 2021 (Meituan 2023 and 2024). Examining the subordinate traits of gig riders offers insights into the broader landscape of workers operating within these new employment arrangements and contributes to a general understanding of their situation.

Data were collected through the distribution of online questionnaires in five Chinese cities: Beijing, Chengdu, Hangzhou, Harbin and Shenzhen.¹⁷ In cooperation with the platform company, we disseminated the survey to the entire rider base through the app’s backend system and finalized it after a designated period. All active gig riders in these five cities received the survey while working through the delivery platform, but participation was voluntary. Between December 2021 and January 2022, a total of 7,680 valid responses were collected. To ensure clarity and facilitate data cleaning, the questionnaire primarily comprised multiple-choice questions, resulting in most variables being categorical. In addition to indicators of economic dependency and personal subordination, the survey captured socio-demographic characteristics, objective work conditions (including income, work intensity, working hours and social security coverage) and subjective perceptions, such as work pressure and job satisfaction. The descriptive socio-demographic characteristics of the sample are shown in table 2. Strikingly, only 374 respondents (4.9 per cent) are women. This gender imbalance is broadly consistent with the 7.4 per cent female share of riders reported by the Meituan Research Institute in its 2020 nationwide survey (Sun, Zhao and Zhang 2021). It reflects a broader characteristic of location-based platform work: in the app-based taxi and delivery sectors, male dominance is common worldwide, with women accounting for fewer than 10 per cent of workers (ILO 2021).

Table 2

Basic socio-demographic characteristics of the sample

Item	Classification	N	%	Mean	Standard deviation
Gender (0,1)
	Male (0)	7 306	95.1	1.05	0.215
	Female (1)	374	4.9	1.05	0.215
Age
	19–25 years old	983	12.8	34.38	8.198
	26–35 years old	3 779	49.2
	36–45 years old	2 095	27.3
	46–55 years old	688	8.9
	56–61 years old	135	1.8
Marital status (1–3)
	Unmarried (1)	3 109	40.5	1.66	0.592
	Married (2)	4 083	53.2
	Other (3)	488	6.4
Child support status (0–3)
	No children	3 597	46.8	0.85	0.929
	One child	2 015	26.2
	Two children	1 689	22
	Three or more children	379	4.9
Education (1–5)
	Lower secondary education and below (1)	2 899	37.7	1.88	0.883
	Upper secondary education (2)	3 336	43.4
	Vocational short-cycle tertiary education (3)	1 023	13.3
	Undergraduate (4)	325	4.2
	Graduate and above (5)	97	1.3
Type of household registration (1–4)
	Local non-farm (1)	648	8.4	3.40	0.987
	Local rural (2)	875	11.4
	Non-local non-farm (3)	952	12.4
	Non-local rural (4)	5 205	67.8
Note: Except for age, the means and standard deviations of the remaining ordinal categorical variables were calculated by sequentially assigning values to each item. Source: Our own calculations based on survey data.

3.2. Analytical methodology

To construct a typology of gig riders, we used the LCA in Mplus to identify clusters and subtypes based on varying patterns of economic dependency and personal subordination. LCA is a statistical method that detects unobserved heterogeneity within a population and classifies individuals into distinct subtypes based on their responses to categorical variables (Hagenaars and McCutcheon 2002; Geiser 2013; Oberski 2016). Compared with traditional clustering methods, LCA is considered to be more statistically robust, offering advantages, such as reduced arbitrariness, in determining the appropriate number of clusters and the ability to assess model fit based on an underlying statistical model (Karnowski 2017; Sinha, Calfee and Delucchi 2021).

We then employ the Kruskal–Wallis H test and binary logistic regression, conducted using SPSS, to examine the potential influence of various subordinate attribute patterns on the gig riders’ objective working conditions and subjective work-related perceptions. The Kruskal–Wallis H test is a rank-based, non-parametric alternative to analysis of variance (ANOVA), allowing for group comparisons without requiring data to meet the assumptions of normality. It can be used to determine whether there are statistically significant differences between two or more groups of an independent variable on a continuous or ordinal dependent variable.¹⁸ Binomial logistic regression is a statistical method used to determine the relationship between predictor variables and a binary categorical outcome variable.¹⁹ With the exception of social security coverage, which is a binary categorical variable, all other working conditions can be regarded as ordinal variables. Accordingly, the Kruskal–Wallis H test is appropriate for examining intergroup differences across these variables, while binomial logistic regression is used to test whether latent classification influences social security participation.

3.3. Results and findings

3.3.1. A typology of gig riders based on economic dependency and personal subordination

Before conducting clustering analysis to explore the internal heterogeneity among gig riders and building typologies based on economic dependency and personal subordination, the descriptive statistics on the overall distribution of subordinative characteristics presented in figures 1, 2, 3 provide some key insights.

Figure 1

Gig rider distribution on each personal subordination dimension (percentages)

Source: Our own calculations based on survey data.

First, gig riders exhibit certain characteristics associated with self-employment in specific indicators of personal subordination, and the expression of this subordination varies across subdimensions. Given that all gig riders in the sample work for the same nationwide digital delivery platform, these variations may stem from differences in local operational modes or the involvement of third-party intermediaries between the platform and the workers.²⁰ Overall, riders report the highest level of autonomy in managing their working hours, with 84.14 per cent indicating that they could freely choose when to log on and off. However, autonomy is more limited in decision-making and subcontracting within the labour process. Specifically, 66.54 per cent of gig riders were required to perform tasks in person, without being able to hire others (figure 1).

Second, economic dependency appears more pronounced than personal subordination among gig riders. Approximately 44.80 per cent of riders rely on the platform as their sole source of income (figure 3), whereas only 9.13 per cent of riders satisfy the criteria of all four indicators of personal subordination (figure 2). This suggests a degree of independence between the two subordination attributes.

Figure 2

Distribution by personal subordination level (percentages)

Notes: The personal subordination level indicates the number of subordinative dimensions the respondent reports to have experienced during the labour process. Level 0 (25.39%) represents respondents who reported no experience of any subdimensions of personal subordination.

Source: Our own calculations based on survey data.

Figure 3

Distribution by economic dependency level (percentages)

Note: The economic dependency level indicates the share of workers’ total remuneration earned through the platform.

Source: Our own calculations based on survey data.

Next, we conducted an LCA to classify gig riders based on subordination attributes, including economic dependency²¹ and the four subdimensions of personal subordination, each represented by binary variables. We first estimated a single-class model, then we increased the number of classes until we identified the latent class model with the best fit. Table 3 presents the statistical criteria that we used to evaluate model fit. Among these, the Akaike information criterion (AIC) and Bayesian information criterion (BIC) are indexes derived from the maximum likelihood estimates of a fitted model, with lower values indicating better model fit (Weller, Bowen and Faubert 2020; Sinha, Calfee and Delucchi 2021). BIC is the most reported indicator of model fit and it is considered to be the most reliable (Nylund, Asparouhov and Muthén 2007; Killian et al. 2019). The Vuong–Lo–Mendell–Rubin adjusted likelihood ratio test (VLMR-LRT), Lo–Mendel–Rubin likelihood ratio test (LMR-LRT) and the bootstrapped likelihood ratio test are used to determine whether a model with k classes is better than one with k – 1 classes (Nylund, Asparouhov and Muthén 2007). A p-value of less than 0.05 indicates that the comparative advantage of model fit is significant. Entropy is a diagnostic measure of a model’s classification accuracy, with values closer to 1 indicating better class definition (Weller, Bowen and Faubert 2020). In addition, the size of each class should be evaluated in terms of both proportionality and theoretical plausibility. Beyond statistical indicators, the theoretical expectations and interpretability of the results are also important considerations when evaluating model fit.

Table 3

Comparing latent class models: Statistical model-fit criteria

Number of classes	1	2	3	4	5
Log likelihood	–22 240.275	–20 084.665	–19 930.264	–19 916.066	–19 907.294
Akaike information criterion (AIC)	44 490.550	40 191.330	39 894.528	39 878.133	39 872.589
Bayesian information criterion (BIC)	44 525.282	40 267.740	40 012.616	40 037.900	40 074.033
VLMR likelihood ratio test	–	<0.001	<0.001	0.0016	0.0096
LMR likelihood ratio test	–	<0.001	<0.001	0.0018	0.0104
Bootstrap likelihood ratio test	–	<0.001	<0.001	<0.001	0.0128
Entropy	–	0.840	0.586	0.786	0.750
The proportion of samples in each classification (%)	100	24/76	21/20/60	50/2/19/29	15/49/3/3/29
Source: Our own calculations based on survey data.

Based on the above principles and statistical criteria, we ultimately selected the four-class model as the optimal latent classification. Figure 4 presents the predicted conditional probabilities for each subdimensions of subordination attributes, indicating the likelihood of each type of subordination among workers across the different subtypes. Figure 5 indicates the proportion of workers in our sample per subtype.

Figure 4

Predicted conditional probabilities of different subordinate attributes

Source: Our own calculations based on survey data.

Figure 5

Estimated shares of workers by subtypes

Source: Our own compilation based on survey data.

According to the analytical results, gig riders can be classified into four distinct subtypes. Subtype 1 (19.14 per cent) includes riders who exhibit both high economic dependency and high personal subordination across all subdimensions. Only these workers can be classified into the traditional “employee versus independent worker” dichotomy and should be considered to have an employment relationship with the platform or third-party intermediaries and classified into the “employee” category regardless of the actual contractual status and part-time/full-time status.²² This is the “employee-like” group. Riders in subtypes 2, 3 and 4,²³ who partially meet certain subordination attributes, are classified as “quasi-subordinate workers”. The clustering results indicate that those who “do not fully meet the current criteria for the identification of an employment relationship but are subject to enterprise-level management during the labour process” are not a completely homogeneous group. Subtype 2 (49.93 per cent) comprises workers who exhibit certain personal subordination traits alongside a high degree of economic dependency. Their subordination is primarily reflected in the requirement to perform tasks personally – without the option to delegate – and limited autonomy in decision-making during the labour process. This group represents nearly half of all gig riders, making it the predominant category and a typical subtype within the “quasi-subordinate” classification. We label gig riders in this subtype “partially subordinate dependants”. Subtype 3 (1.95 per cent) comprises workers who do not exhibit economic dependency on the platform but display high levels of personal subordination across all subdimensions. We label this subtype “Subordinate independents”. By contrast, workers belonging to subtype 4 (28.97 per cent) show great economic dependency but low levels of personal subordination. We label gig riders in this subtype “economic dependants”. Overall, “quasi-subordinate workers” – those who do not fully meet the criteria for subordination attributes – account for 80.86 per cent of the gig rider population.

3.3.2. A comparison of working conditions across gig rider subtypes

A Kruskal–Wallis H test was run to determine the existence of differences in workers’ average monthly income, working hours and work intensity,²⁴ perceived work pressure and job satisfaction across the four subgroups of participants with different subordinate attribute patterns. Visual inspection of boxplots indicates that there is little similarity in the distributions of these variables across groups (see supplementary online Appendix 1). Pairwise comparisons were conducted using Dunn’s (1964) procedure, with a Bonferroni correction for multiple comparisons, and adjusted p-values. Statistical significance was determined at p < .0083. Unless otherwise indicated, values are reported as mean ranks (see detailed parametric description in supplementary online Appendix 2). Binomial logistic regression was then used to examine the effects of latent subgroup classification on the likelihood of gig riders participating in the “social insurance for urban and rural residents”, “social insurance for flexible employment” and “commercial insurance”.²⁵ Socio-demographic characteristics – such as gender, age, marital status, child support status, education level and household registration type – were included in the model as control variables.²⁶ Detailed parametric results for each of the three social insurance types are presented in supplementary online Appendix 3, tables SA3.1, SA3.2 and SA3.3, respectively.

The results indicate statistically significant differences across subgroups in average monthly income (χ²(3) = 434.160, p < .001), average weekly working days (χ²(3) = 211.491, p < .001), average daily working hours (χ²(3) = 343.206, p < .001), average daily number of orders (χ²(3) = 380.904, p < .001), perceived work pressure (χ²(3) = 17.571, p < .001) and job satisfaction (χ²(3) = 129.137, p < .001). Specifically, gig riders with higher economic dependency (employee-like and economic dependants subtypes) tend to have higher overall income, longer daily working hours, greater work intensity and report higher perceived work pressure. With the exception of the employee-like subtype, the remaining three subtypes – classified as “quasi-subordinate workers” – do not exhibit significant variance in weekly working days. Notably, workers with high levels of personal subordination (subordinate independents and employee-like subtypes) report lower levels of job satisfaction. Surprisingly, employee-like gig riders have the lowest participation rates across all three types of social insurance, with the subordinate independents presenting the second-lowest rates for both “social insurance for urban and rural residents” and “social insurance for flexible employment”, and partially subordinate dependants presenting the second-lowest rates in the case of “commercial insurance”. The employee-like subtype – with both high personal subordination and high economic dependency – is arguably most in need of social insurance protection to mitigate the risks associated with this unstable occupation. In summary, the observed differences in working conditions across subgroups highlight the latent heterogeneity within the gig worker population.

4. Discussion and conclusion

By proposing practical indicators for identifying quasi-subordinate worker groups, this study responds to ongoing policy ambiguity and contributes to validating the relevance of two key subordination attributes – economic dependency and personal subordination – in distinguishing different subtypes of gig workers. Under conventional Chinese criteria for determining employment status, economic dependency has been relatively overlooked. However, formally integrating this factor alongside the existing criterion of personal subordination may enhance future judicial and legislative practices.

Our results suggest that indicators such as “primary employer” and “revenue proportion” can effectively identify quasi-subordinate workers who retain autonomy over work schedules and task selection yet remain economically reliant on the platform and subject to its operational rules. Economic dependency may also serve as a key indicator in assessing the need for worker protection, as it is associated with excessive daily working hours, increased work intensity and heightened perceived pressure. Moreover, since our typology is grounded in a labour law perspective and based on quantified legal concepts, it offers the means of capturing the intrinsic nature of the relationship without being constrained by complex multi-layered business structures and ambiguous contractual arrangements.²⁷

This approach could overcome limitations in previous literature that categorizes Chinese gig workers based on platform business models or operational structures, which often focus on the structural, contractual or registrational relationships between workers, intermediaries and digital platforms (Zhao and Luo 2024; Sun, Chen and Rani 2023; Qiu, Sun and Chen 2021; Zhen et al. 2021; Wang and Cooke 2021; Wang and Meng 2024). It also offers advantages for large-scale statistical survey design. Gig workers may not accurately report their contractual status owing to limited legal knowledge (Beijing Zhicheng Migrant Workers Legal Aid and Research Centre 2021). However, they can reliably indicate whether all or most of their income is derived from the platform and whether they are subject to platform or intermediary instructions regarding working time, location, task content, supervision, performance evaluation and disciplinary measures.

Through the collection of large-scale survey data for clustering, estimating proportional distributions and comparing working conditions across latent subgroups, our study offers quantitative empirical evidence that illustrates the heterogeneity within China’s gig worker population. The clustering results support the policy rationale underpinning the trichotomous employment status structure outlined in MOHRSS (2021), while the observed differences in working conditions reinforce the validity of the proposed categorization.

Our findings identify four main subtypes of gig workers, with only approximately 19 per cent conforming with the current binary employment status framework. The remaining 81 per cent should be grouped into a new “quasi-subordinate” worker category. Formal labour relations should be established and contracts signed for the 19 per cent “employee-like” workers, while “quasi-subordinate workers” should be afforded the labour rights proposed by relevant policy frameworks.²⁸ Moreover, nearly half of gig riders exhibit partial personal subordination and high economic dependency, indicating complex platform control and a disadvantaged position within the platform ecosystem. Workers with different subordination attributes exhibit significant variance in both objective working conditions and subjective work-related perceptions. High economic dependency is a strong predictor of longer working hours, greater work intensity and heightened perceived pressure, whereas high personal subordination is linked to low job satisfaction. Social insurance coverage is particularly inadequate for subtypes with higher levels of subordination.

This Chinese-context analysis contributes to global understanding of the heterogeneity within the gig worker population. Researchers in other countries increasingly acknowledge the need to pay more attention to this heterogeneity when analysing platform economy ecosystems (Rosenblat 2018, 49–72; Cansoy et al. 2020; Schor et al. 2020; Congregado et al. 2022), arguing that the reliance of gig workers on platform income to cover basic expenses – as opposed to earning supplemental income – accounts for the variation in their experiences (Schor et al. 2020; Schor 2021). However, few studies have performed quantitative clustering analyses to estimate the proportions of domestic subtypes or compare their working conditions. Policymakers and researchers in other countries exploring the extension of the legal employment status to the trichotomous framework may find our results useful in case study comparisons and as reference material. They may even consider conducting similar empirical analyses to better understand the characteristics of their own platform labour ecosystems.

The key limitation of this study is the gender imbalance within our sample. Gendered dynamics in the platform economy are a critical topic that warrants further exploration (Sun, Zhao and Zhang 2021; Zheng, Qiu and Yang 2024). Although such gender imbalance is typical of platform-based delivery work, future research should aim to include more data on female gig workers and other location-based gig work sectors, such as platform-based care and cleaning work. Such an approach would allow for comparisons of subordination characteristics across gender groups. Although the descriptive comparisons that we have presented here demonstrate and highlight the heterogeneity among gig workers, they fall short of fully explaining the underlying mechanisms driving the observed differences in working conditions – particularly the notably low rates of social insurance participation among subtypes with higher levels of subordination. Further field-based research is needed to uncover the root causes of these disparities.

Notes

See https://www.europarl.europa.eu/legislative-train/theme-a-europe-fit-for-the-digital-age/file-improving-working-conditions-of-platform-workers (accessed 7 August 2025). ⮭

In the Chinese labour market context, NFE include workers who rely on digital online platforms to access job opportunities, including delivery riders, taxi-hailing/truck-hailing drivers, housecleaning workers and livestreamers. NFE is a common concept in Chinese policy documents, and its connotations are similar to those of “platform work”. ⮭

It should be noted that the Guiding Opinions exist only in the form of regulations and are not part of the labour law framework (Ke 2022). They can, nevertheless, be used as a reference in court decisions. ⮭

The original text is “不完全符合确立劳动关系情形但企业对劳动者进行劳动管理”. ⮭

In addition, the “Guidelines on Safeguarding the Rights to Rest and Remuneration for Workers in New Forms of Employment”, the “Guidelines on Publicizing Labour Rules for Workers in New Forms of Employment” and the “Service Guide for Protecting the Rights and Interests of Workers in New Forms of Employment”, issued in November 2023, provide detailed provisions on the labour rights and protections to which workers in new forms of employment are entitled (MOHRSS 2023). ⮭

See https://jshrss.jiangsu.gov.cn/art/2021/12/28/art_77279_10234493.html (accessed 19 August 2025). ⮭

See https://rsj.sh.gov.cn/tldgx_17731/20220210/t0035_1405575.html (accessed 19 August 2025). ⮭

In this article, “gig riders” refer to workers in NFE who register on digital platforms in the food-delivery industry, accept customer orders through these platforms and deliver meal orders to specified locations within required time frames according to order requirements. ⮭

According to Recommendation No. 198, countries should consider the specific indicators of the existence of an employment relationship in their laws and regulations. Suggested indicators include: “…(a) the fact that the work: is carried out according to the instructions and under the control of another party; involves the integration of the worker in the organization of the enterprise; is performed solely or mainly for the benefit of another person; must be carried out personally by the worker; is carried out within specific working hours or at a workplace specified or agreed by the party requesting the work; is of a particular duration and has a certain continuity; requires the worker’s availability; or involves the provision of tools, materials and machinery by the party requesting the work; (b) periodic payment of remuneration to the worker; the fact that such remuneration constitutes the worker’s sole or principal source of income; provision of payment in kind, such as food, lodging or transport; recognition of entitlements such as weekly rest and annual holidays; payment by the party requesting the work for travel undertaken by the worker in order to carry out the work; or absence of financial risk for the worker.” ⮭

In the gig work marketplace in China, there are many third-party intermediaries between digital platforms and gig workers, such as subcontractors and temporary employment and human resources service agencies (Beijing Zhicheng Migrant Workers Legal Aid and Research Centre 2021). ⮭

Available at https://www.shezfy.com/book/bps/2022/p05.html (accessed 19 August 2025). ⮭

“c 76 The Labour Relations Amendment Act, 1975”, Ontario: Annual Statutes, Vol. 1975, Article 78. ⮭

“Ley 20/2007, de 11 de julio, del Estatuto del trabajo autónomo”, Boletín Oficial del Estado No. 166, 12 July 2007. ⮭

“Allgemeines Gleichbehandlungsgesetz” (AGG), BGBl. I S. 1897, enacted on 14 August 2006. ⮭

See note 13. ⮭

In the Chinese NFE context, 50 per cent is a good threshold, since there are obvious differences in the employment situation of workers above and below it. In the pre-analysis of the sample data, where platform income share is over 50 per cent, gig riders’ full-time rate exceeds the part-time rate. ⮭

These five cities were selected for their representativeness. Beijing and Shenzhen are the cities with the most developed platform economy in China, and they represent the Beijing-Tianjin-Hebei economic region (in the north of China) and the Pearl River Delta region (in the south), respectively. The other three cities are second-tier cities. Chengdu is the largest city in south-west China, Hangzhou is the city with the most developed platform economy in the Yangtze River Delta region, and Harbin is a city in north-east China that is broadly representative of the industrial structure and demographic challenges of the region’s resource-dependent urban centres. ⮭

Laerd Statistics, “Kruskal-Wallis H Test Using SPSS Statistics”. https://statistics.laerd.com/spss-tutorials/kruskal-wallis-h-test-using-spss-statistics.php (accessed 30 July 2025). ⮭

Laerd Statistics, “Binomial Logistic Regression Using SPSS Statistics”. https://statistics.laerd.com/spss-tutorials/binomial-logistic-regression-using-spss-statistics.php (accessed 30 July 2025). ⮭

Chinese delivery platforms use different modes of operation (Zhao and Luo 2024; Sun, Chen and Rani 2023; Wang and Meng 2024). Crowdsourced workers may be subject to higher personal subordination. Outsourced workers, by contrast, are tied to different local third-party intermediaries, such as subcontractors and temporary employment and human resources service agencies; they are, therefore, subject to varying degrees of control and management, which could lead to different expressions of personal subordination. ⮭

Here, the economic dependency is transformed into a binary variable using a threshold of 50 per cent, as mentioned in section 2.3. ⮭

Even part-time workers should be regarded as falling under “part-time employment”. For further information on this concept, see MOHRSS (2003). ⮭

Notably, the latent class analysis did not recognize the “true self-employed” category with both low economic dependency and personal subordination among the gig rider samples. This is likely because the employment model in the Chinese food-delivery industry has undergone multiple complex and rapid iterations and evolutions, such that a “pure form” of crowdsourcing gig work no longer exists. Instead, crowdsourced riders are transitioning from flexible employment to more stable and “sticky” employment, which is primarily reflected in an increase in the proportion of full-time workers, extended years of service and a significant growth in working hours, consequently heightening economic dependency. Furthermore, food-delivery platforms employ constant and pervasive management and supervision over riders through digital devices and methods, thereby increasing the degree of personal subordination. (Beijing Zhicheng Migrant Workers Legal Aid and Research Centre 2021; Sun, Chen and Rani 2023). ⮭

In this study, working hours are derived from the average weekly working days and average daily working hours, while work intensity is measured using the average daily delivery order amount. ⮭

These three types of social insurance represent the primary options available to gig workers within the current Chinese social security system. ⮭

Among these control variables and our key independent variable “latent classification”, only age is a continuous variable. Accordingly, we do not need to test for multicollinearity. ⮭

The marketplace is chaotic, with many third-party intermediaries – such as subcontractors and temporary employment and human resources service agencies – operating between digital platforms and gig workers. Some workers are subjected to the management of these intermediary firms but do not have formal (and correct) employment contracts with them. Furthermore, the digital platform still implicitly controls workers’ labour process through intelligent algorithm systems and intermediaries. Reported relationships can thus be misleading, with bogus self-employed/independent contractors being common (Beijing Zhicheng Migrant Workers Legal Aid and Research Centre 2021). In addition, multi-layered business networks make structural relationships complex and ambiguous, such that contractual or registration status under certain platform business models may not accurately reflect the nature of the employment relationship. ⮭

See note 5. ⮭

References

All-China Federation of Trade Unions. 2023. 第九次中国职工状况调查: 报告卷 [Ninth China Workers’ Status Survey Report]. Beijing: China Workers Publishing House.

Arthurs, Harry W. 1965. “The Dependent Contractor: A Study of the Legal Problems of Countervailing Power”. University of Toronto Law Journal 16 (1): 89–117. http://doi.org/10.2307/825096.

Arthurs, Harry W. 2011. “Labor Law after Labor”. In The Idea of Labour Law, edited by Guy Davidov and Brian Langille, 13–29. Oxford: Oxford University Press.

Bagari, Sara. 2020. “The European Pillar of Social Rights: An EU-Level Response to the Social Protection of the (Economically) Dependent Self-Employed?” Zbornik znanstvenih razprav/Ljubljana Law Review 80 (2): 101–118. http://doi.org/10.51940/2020.2.101-118.

Ban, Xiaohui. 2017. “论‘分享经济’下我国劳动法保护对象的扩张——以互联网专车为视角” [A Research on the Expansion of the Protection Scope of Labour Law in the Sharing Economy from the Perspective of Hiring Cars by Internet Platform]. 四川大学学报(哲学社会科学版)/Journal of Sichuan University (Social Science Edition) 209 (2): 154–161.

Ban, Xiaohui, Tobias Beck, René Bormann, Wolfgang Däubler, Li Kungang, Wang Fayang, Wang Qian, and Yang Yang. 2023. Platform Economy in China and Germany: Labour Law Policy Recommendations for Decent Work. Shanghai: Friedrich-Ebert-Stiftung Shanghai.

Beijing Zhicheng Migrant Workers Legal Aid and Research Centre. 2021. 外卖平台用工模式法律研究报告 [Legal Research Report on Employment Models of Food Delivery Platforms]. Beijing.

Bozzon, Rossella, and Annalisa Murgia. 2022. “Independent or Dependent? European Labour Statistics and Their (In)ability to Identify Forms of Dependency in Self-Employment”. Social Indicators Research 160 (1): 199–226. http://doi.org/10.1007/s11205-021-02798-1.

Cansoy, Mehmet, Samantha Eddy, Isak Ladegaard, and Juliet. B. Schor. 2020. “Homines Diversi: Heterogeneous Earner Behaviors in the Platform Economy”. Sociologica 14 (3): 143–165. http://doi.org/10.6092/issn.1971-8853/11508.

Chen, Long. 2022. “Labor Order under Digital Control: Research on Labor Control of Take-Out Platform Riders”. Journal of Chinese Sociology 9: Article No. 17. http://doi.org/10.1186/s40711-022-00171-4.

Cherry, Miriam A., and Antonio Aloisi. 2017. “‘Dependent Contractors’ in the Gig Economy: A Comparative Approach”. American University Law Review 66 (3): 635–689. http://doi.org/10.2139/ssrn.2847869.

Congregado, Emilio, María Isabel de Andrés, Eimear Nolan, and Concepción Román. 2022. “Heterogeneity among Self-Employed Digital Platform Workers: Evidence from Europe”. International Review of Entrepreneurship 20 (1): 45–68.

Countouris, Nicola. 2019. Defining and Regulating Work Relations for the Future of Work. Geneva: ILO.

Däubler, Wolfgang. 1999. “Working People in Germany”. Comparative Labor Law and Policy Journal 21 (1): 77–98.

Davidov, Guy. 2017. “Subordination vs Domination: Exploring the Differences”. International Journal of Comparative Labour Law and Industrial Relations 33 (3): 365–389. http://doi.org/10.54648/ijcl2017016.

Davidov, Guy, Mark Freedland, and Nicola Kountouris. 2015. “The Subjects of Labor Law: ‘Employees’ and Other Workers”. In Comparative Labor Law, edited by Matthew W. Finkin and Guy Mundlak, 115–131. Cheltenham: Edward Elgar.

De Stefano, Valerio. 2016. “The Rise of the ‘Just-in-time Workforce’: On-Demand Work, Crowd Work, and Labor Protection in the ‘Gig-Economy’”. Comparative Labor Law and Policy Journal 37 (3): 471–504. http://doi.org/10.2139/ssrn.2682602.

Dunn, Olive Jean. 1964. “Multiple Comparisons Using Rank Sums”. Technometrics 6 (3): 241–252. http://doi.org/10.1080/00401706.1964.10490181.

Eurofound (European Foundation for the Improvement of Living and Working Conditions). 2017. Exploring Self-Employment in the European Union. Luxembourg: Publications Office of the European Union.

Eurofound (European Foundation for the Improvement of Living and Working Conditions). 2018. “Economically Dependent Worker”. In European Industrial Relations Dictionary. Dublin. Article published 11 June 2007; updated 26 June 2018. https://www.eurofound.europa.eu/en/european-industrial-relations-dictionary/economically-dependent-worker (accessed 11 August 2025).

Eurostat. 2018. Labour Force Survey (LFS) Ad-hoc Module 2017 on the Self-Employed Persons: Assessment Report. Luxembourg: Publications Office of the European Union.

Fang, Changchun. 2022. “‘第三类劳动’及其权益保障：问题与挑战” [The Third Type of Labour and Its Rights and Interests: Problems and Challenges]. 人民论坛：学术前沿/Frontiers, No. 8: 52–62.

Geiser, Christian. 2013. Data Analysis with Mplus. Methodology in the Social Sciences. New York, NY: Guilford Press.

Grimshaw, Damian, Mat Johnson, Arjan Keizer, and Jill Rubery. 2017. “The Governance of Employment Protection in the UK: How the State and Employers are Undermining Decent Standards”. In Myths of Employment Deregulation: How It Neither Creates Jobs Nor Reduces Labour Market Segmentation, edited by Agnieszka Piasna and Martin Myant, 225–246. Brussels: European Trade Union Institute.

Hagenaars, Jacques A., and Allan L. McCutcheon, eds. 2002. Applied Latent Class Analysis. Cambridge: Cambridge University Press.

Harris, Seth D. 2018. “Workers, Protections, and Benefits in the U.S. ‘Gig Economy’”. Global Law Review 40 (4): 7–37.

Harris, Seth D., and Alan B. Krueger. 2015. “A Proposal for Modernizing Labor Laws for Twenty-First-Century Work: The ‘Independent Worker’”. Discussion Paper 2015-10. Washington, DC: Hamilton Project.

ILO. 1993. Resolution concerning the International Classification of Status in Employment (ICSE). 15th International Conference of Labour Statisticians, 1993. Geneva.

ILO. 2018. Resolution concerning statistics on work relationships. 20th International Conference of Labour Statisticians, 2018. Geneva.

ILO. 2021. World Employment and Social Outlook 2021: The Role of Digital Labour Platforms in Transforming the World of Work. Geneva.

Karnowski, Veronika. 2017. “Latent Class Analysis”. In The International Encyclopedia of Communication Research Methods, edited by Jörg Matthes, Christine S. Davis and Robert F. Potter. Hoboken, NJ: John Wiley and Sons.

Ke, Zhenxing. 2022. “A Third Employment Category for Platform Workers in China: A Tough Start”. Chinese Journal of Comparative Law 10 (2): 297–322. http://doi.org/10.1093/cjcl/cxac022.

Killian, Michael O., Andrea N. Cimino, Bridget E. Weller, and Chang Hyun Seo. 2019. “A Systematic Review of Latent Variable Mixture Modeling Research in Social Work Journals”. Journal of Evidence-Based Social Work 16 (2): 192–210. http://doi.org/10.1080/23761407.2019.1577783.

Langille, Brian A., and Guy Davidov. 1999. “Beyond Employees and Independent Contractors: A View from Canada”. Comparative Labor Law and Policy Journal 21 (1): 7–46.

Li, Shenglan, and Lihua Jiang. 2020. “新型劳动时间控制与虚假自由—外卖骑手的劳动过程研究” [A New Mode of Labor Time Control and Fake Experience of Freedom: A Study on the Labour Process of Take-Out Platform Riders]. 社会学研究/Sociological Studies, No. 6: 91–112.

Liu, Shanshi, Shubing Liu, and Xiaolang Liu. 2022. 平台劳动者:分类、权益与治理 [Online Platform Labourers: Classification, Rights and Interests and Governance]. Beijing: China Legal Publishing House.

Lu, Haina, and Yiheng Chen. 2020. “The Protection of the ‘Third Kind of Workers’ in Gig Economy from the Perspective of Social Rights”. Journal of Human Rights ( 19) 1.

Meituan. 2023. 2022美团骑手权益保障社会责任报告 [2022 Meituan Riders’ Rights Protection Social Responsibility Report]. Beijing.

Meituan. 2024. 2023年·美团骑手权益保障社会责任报告 [2023 Meituan Riders’ Rights Protection Social Responsibility Report]. Beijing.

MOHRSS (China, Ministry of Human Resources and Social Security). 2003. “关于非全日制用工若干问题的意见” [Opinions on Several Issues Regarding Part-Time Employment]. MOHRSS Document No. 12. Beijing.

MOHRSS (China, Ministry of Human Resources and Social Security). 2005. “关于确立劳动关系有关事项的通知” [Notice on Matters Concerning the Establishment of Labour Relations]. MOHRSS Document No. 12. Beijing.

MOHRSS (China, Ministry of Human Resources and Social Security). 2021. “关于维护新就业形态劳动者劳动保障权益的指导意见” [Guiding Opinions on Safeguarding the Labour Rights and Interests of Workers in New Forms of Employment]. MOHRSS Document No. 56. Beijing.

MOHRSS (China, Ministry of Human Resources and Social Security). 2023. “人力资源社会保障部办公厅关于印发‘新就业形态劳动者休息和劳动报酬权益保障指引’‘新就业形态劳动者劳动规则公示指引’‘新就业形态劳动者权益维护服务指南’的通知” [MOHRSS Circular on the Publication of the “Guidelines on Safeguarding the Rights to Rest and Remuneration for Workers in New Forms of Employment”, the “Guidelines on Publicizing Labour Rules for Workers in New Forms of Employment”, and the “Service Guide for Protecting the Rights and Interests of Workers in New Forms of Employment”]. MOHRSS Document No. 50. Beijing.

Nylund, Karen L., Tihomir Asparouhov, and Bengt O. Muthén. 2007. “Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study”. Structural Equation Modeling 14 (4): 535–569. http://doi.org/10.1080/10705510701575396.

Oberski, Daniel. 2016. “Mixture Models: Latent Profile and Latent Class Analysis”. In Modern Statistical Methods for HCI, edited by Judy Robertson and Maurits Kaptein, 275–287. Cham: Springer.

Palagashvili, Liya, Paola A. Suarez, Christopher M. Kaiser, and Vitor Melo. 2024. “Assessing the Impact of Worker Reclassification: Employment Outcomes Post-California AB5”. Mercatus Working Paper. Arlington, VA: Mercatus Center, George Mason University.

Qiu, Jack Linchuan, Ping Sun, and Julie Chen. 2021. “Labour Management and Resistance among Platform-Based Food Delivery Couriers in Beijing”. In A Modern Guide to Labour and the Platform Economy, edited by Jan Drahokoupil and Kurt Vandaele, 290–307. Cheltenham: Edward Elgar.

Rosenblat, Alex. 2018. Uberland: How Algorithms Are Rewriting the Rules of Work. Oakland, CA: University of California Press.

Rosioru, Felicia. 2014. “Legal Acknowledgement of the Category of Economically Dependent Workers”. European Labour Law Journal 5 (3–4): 279–305. http://doi.org/10.1177/201395251400500307.

Sánchez Torres, Esther. 2010. “The Spanish Law on Dependent Self-Employed Workers: A New Evolution in Labor Law”. Comparative Labor Law and Policy Journal 31 (2): 231–248.

Schor, Juliet B. 2021. “Dependence and Heterogeneity in the Platform Labor Force”. Policy Brief, Governing Work in the Digital Age. Berlin: Hertie School.

Schor, Juliet B., William Attwood-Charles, Mehmet Cansoy, Isak Ladegaard, and Robert Wengronowitz. 2020. “Dependence and Precarity in the Platform Economy”. Theory and Society 49 (5–6): 833–861. http://doi.org/10.1007/s11186-020-09408-y.

Schoukens, Paul, and Alberto Barrio. 2017. “The Changing Concept of Work: When Does Typical Work Become Atypical?” European Labour Law Journal 8 (4): 306–332. http://doi.org/10.1177/2031952517743871.

Shen, Jianfeng. 2022. “数字时代劳动法的危机与用工关系法律调整的方法革新” [The Crisis of Labour Law in the Digital Age and Innovative Approaches to Legal Adjustments of Employment Relations]. 法制与社会发展 [Law and Social Development] 28 (2): 119–135.

Sinha, Pratik, Carolyn S. Calfee, and Kevin L. Delucchi. 2021. “Practitioner’s Guide to Latent Class Analysis: Methodological Considerations and Common Pitfalls”. Critical Care Medicine 49 (1): e63–e79. http://doi.org/10.1097/CCM.0000000000004710.

Sorge, Stefanie. 2010. “German Law on Dependent Self-Employed Workers: A Comparison to the Current Situation under Spanish Law”. Comparative Labor Law and Policy Journal 31 (2): 249–252.

Stewart, Andrew, and Shae McCrystal. 2019. “Labour Regulation and the Great Divide: Does the Gig Economy Require a New Category of Worker?” Australian Journal of Labour Law 32 (1): 4–22.

Stewart, Andrew, and Jim Stanford. 2017. “Regulating Work in the Gig Economy: What Are the Options?” Economic and Labour Relations Review 28 (3): 420–437. http://doi.org/10.1177/1035304617722461.

Sun, Ping, Julie Yujie Chen, and Uma Rani. 2023. “From Flexible Labour to ‘Sticky Labour’: A Tracking Study of Workers in the Food-Delivery Platform Economy of China”. Work, Employment and Society 37 (2): 412–431. http://doi.org/10.1177/09500170211021570.

Sun, Ping, Yuchao Zhao, and Qianyu Zhang. 2021. “平台、性别与劳动：‘女骑手’ 的性别展演” [Platforms, Gender, and Labor: Female Food Delivery Couriers’ Gender Performativity]. 妇女研究论丛/Journal of Chinese Women’s Studies 168 (6): 5–16.

Supiot, Alain. 2001. Beyond Employment: Changes in Work and the Future of Labour Law in Europe. Oxford: Oxford University Press.

Tian, Silu. 2019. “工业4.0时代的从属劳动论” [The Theory of Subordinate Labour in the Era of Industry 4.0]. 法学评论/Law Review, No. 1: 76–85.

Wang, Jing, and Quan Meng. 2024. “Unregulated Flexibility and the Multiplication of Labour: Work in the Chinese Platform Economy”. Social Inclusion 12: Article No. 7719. http://doi.org/10.17645/si.7719.

Wang, Qian. 2017. “德国法中劳动关系的认定” [Identification of Employment Relationships in German Law]. 暨南学报（哲学社会科学版）/Jinan Journal (Philosophy and Social Sciences) 39 (6): 39–48.

Wang, Quanxing, and Wang Qian. 2018. “我国“网约工”的劳动关系认定及权益保护” [Identification of Employment Relationship and Protection of Rights and Interests of ‘Online Gig Workers’ in China]. 法学 [Law Review] 4: 57–72.

Wang, Tianyu. 2016. “基于互联网平台提供劳务的劳动关系认定——以“e代驾”在京、沪、穗三地法院的判决为切入点” [Identification of Employment Relationships among Workers Providing Labour Services through Internet Platforms: A Case Study of “E-Driving” in the Courts of Beijing, Shanghai, and Guangzhou]. 法学 [Law Review], No. 6: 50–60.

Wang, Tianyu, and Fang Lee Cooke. 2021. “Internet Platform Employment in China: Legal Challenges and Implications for Gig Workers through the Lens of Court Decisions”. Relations Industrielles/Industrial Relations 76 (3): 541–564. http://doi.org/10.7202/1083612ar.

Weller, Bridget E., Natasha K. Bowen, and Sarah J. Faubert. 2020. “Latent Class Analysis: A Guide to Best Practice”. Journal of Black Psychology 46 (4): 287–311. http://doi.org/10.1177/0095798420930932.

Williams, Colin C., and Ioana Alexandra Horodnic. 2018. “Evaluating the Prevalence and Distribution of Dependent Self-Employment: Some Lessons from the European Working Conditions Survey”. Industrial Relations Journal 49 (2): 109–127. http://doi.org/10.1111/irj.12206.

Williams, Colin C., and Ioana Alexandra Horodnic. 2019. Dependent Self-Employment: Theory, Practice and Policy. Cheltenham: Edward Elgar.

Wu, Qingjun, Hao Zhang, Zhen Li, and Kai Liu. 2019. “Labor Control in the Gig Economy: Evidence from Uber in China”. Journal of Industrial Relations 61 (4): 574–596. http://doi.org/10.1177/0022185619854472.

Wu, Qingjun, and Zhen Li. 2018. “分享经济下的劳动控制与工作自主性——关于网约车司机工作的混合研究” [Labour Process Control and Job Autonomy in Sharing Economy: A Case Study of Online Car-Hailing Drivers’ Work]. 社会学研究/Sociological Studies 33 (4): 137–162.

Xiao, Zhu. 2021. “劳动关系从属性认定标准的理论解释与体系构成” [Theoretical Explanation and System Composition of Subordinate Attribute Identification Standards for Employment Relations]. 法学 [Law Review], No. 2: 160–176.

Xie, Zengyi. 2018. “互联网平台用工劳动关系认定” [Identification of Employment Relationships in Internet Platform Employment]. Peking University Law Journal, No. 6: 1546–1569.

Yang, Haonan. 2022. “共享经济背景下我国劳动关系认定标准的路径选择” [Path for Labour Relations Identification Standards in the Context of the Sharing Economy in China]. 法学评论/Law Review 40 (2): 100–112.

Yuan, Zhaohui. 2021. “新就业形态人员社会保险状况研究” [An Exploratory Study on the Social Insurance Status of New Forms of Employment]. 中国劳动关系学院学报/Journal of China Institute of Industrial Relations, No. 1: 75–84.

Zhang, Chenggang. 2019. “问题与对策:我国新就业形态发展中的公共政策研究” [Problems and Strategies: A Study of Public Policy on the Development of New Forms of Employment in China]. 中国人力资源开发 [Human Resources Development of China], No. 2: 74–82. http://doi.org/10.16471/j.cnki.11-2822/c.2019.02.006.

Zhao, Bo, and Siqi Luo. 2024. “The Old Conflict in the New Economy? Courier Resistance on Outsourcing Platforms in China”. China Quarterly 258 (June): 495–512. http://doi.org/10.1017/S0305741023001467.

Zhao, Lei, and Yue Han. 2021. “跨越企业边界的科层控制——网约车平台的劳动力组织与控制研究” [Hierarchical Control across Corporate Boundaries: On Labour Organization and Control for Online Ride-Hailing Platforms: A Case Study of W Ride-Hailing Platform in City T]. 社会学研究/Sociological Studies 36 (5): 70–90.

Zhen, Lu, Yiwei Wu, Shuaian Wang, and Wen Yi. 2021. “Crowdsourcing Mode Evaluation for Parcel Delivery Service Platforms”. International Journal of Production Economics 235 (May): Article No. 108067. http://doi.org/10.1016/j.ijpe.2021.108067.

Zheng, Qi, Zitong Qiu, and Weiguo Yang. 2024. “The Shifting Motherhood Penalty and Fatherhood Premium in China’s Gig Economy: Impact of Parental Status on Income Changes”. International Labor Review 163 (2): 173–197. http://doi.org/10.1111/ilr.12407.

Zou, Mimi. 2017. “The Regulatory Challenges of ‘Uberization’ in China: Classifying Ride-Hailing Drivers”. International Journal of Comparative Labour Law and Industrial Relations 33 (2): 269–294. http://doi.org/10.54648/ijcl2017012.