Labor market regulation intensity and manufacturing entrepreneurship: evidence from Chinese cities

In China, labor market regulation refers to government oversight and enforcement of labor laws, imposing penalties on companies that violate these laws. Labor market regulation is a key variable in entrepreneurial research (Angulo-Guerrero et al., 2023; da Fonseca, 2022; Dilli, 2021; Henrekson, 2020; Kacperczyk and Marx, 2016; Wennekers et al., 2002). As a labor-intensive industry, manufacturing entrepreneurship is particularly sensitive to labor market regulations. Recently, the Chinese government has intensified labor market regulation (Li et al., 2020), with administrative penalties for violations rising from 5,412 to 8,982 in five years, a 65.96% increase. How will this intensified regulation affect manufacturing entrepreneurship in China? According to existing literature (Agostino et al., 2020; Chowdhury et al., 2019; Kim et al., 2022), manufacturing entrepreneurship is fundamental to the development of China’s manufacturing sector. As a manufacturing powerhouse, China has held the world’s largest manufacturing scale for 13 consecutive years. The impact of labor market regulation on manufacturing entrepreneurship in China is crucial for global welfare. Thus, investigating this relationship holds significant theoretical and practical value.

Labor is a crucial input for entrepreneurial ventures (Boudreaux and Nikolaev, 2019; De Clercq et al., 2013; Ghani et al., 2014). Most scholars agree that labor market regulation significantly impacts entrepreneurship (Angulo-Guerrero et al., 2023; Desai et al., 2021; Dilli, 2021; Fu et al., 2018; Stephan et al., 2023; Van Stel et al., 2007). One perspective argues that strict labor market regulation suppresses entrepreneurship (Desai et al., 2021; Van Stel et al., 2007). While another contends it can promote it (Angulo-Guerrero et al., 2023; Dilli, 2021). There are three gaps in existing research on the relationship between labor market regulation and manufacturing entrepreneurship in China. First, most studies focus on entrepreneurship across all sectors, while manufacturing, being labor-intensive, is more sensitive to labor market regulation and deserves specific attention. Second, existing studies present conflicting conclusions, indicating the need for targeted research on the relationship between labor market regulation and manufacturing entrepreneurship. Third, since 2008, the digital transformation of China’s manufacturing sector has altered labor demands, impacting the complex relationship between labor market regulation and manufacturing entrepreneurship – a factor not yet addressed in existing studies.

To address these gaps, we adopt the approach of Bracht et al. (2024), using signaling theory as a framework, incorporating AI innovation and digitalization in manufacturing. We will hypothesize and empirically test the direct, indirect and heterogeneous effects of labor market regulation on manufacturing entrepreneurship using panel data from all Chinese cities from 2008 to 2020, using a two-way fixed effects model.

Our research contributes to signaling theory in the field of entrepreneurship. Following Spence’s (1973) establishment of signaling theory in the labor market, this theory has been widely applied in entrepreneurship research, leading to a growing body of literature (Bafera and Kleinert, 2023; Connelly et al., 2011). This literature examines signals emitted by entrepreneurial firms, including founders’ qualifications (Bublitz et al., 2018), strategic information (Lanchimba et al., 2021), legal forms (Bracht et al., 2024), technological research and development (Wang et al., 2019) and media usage (Courtney et al., 2017). It also explores third-party signals from relationships with microfinance institutions (Anglin et al., 2020) and prestigious universities (Johnston et al., 2023) such as affiliations with prestigious universities (Johnston et al., 2023). These signals are generated by market participants, not by government actions. Our study focuses on signals emitted by government actions in labor market regulation and their impact on manufacturing entrepreneurship. To our knowledge, our research is the first to explore the effects of government-generated signals on entrepreneurship, thereby expanding the application of signaling theory in this field. In addition, our study identifies four main types of signal recipients:

1. entrepreneurs and potential entrepreneurs in manufacturing;

2. laborers providing human capital;

3. banks offering financial resources to manufacturing entrepreneurs; and

4. existing manufacturing and nonmanufacturing firms.

This contrasts with Spence’s (1973) focus on a single signal recipient, thus broadening the scope of signaling theory.

Our research narrows the focus of entrepreneurship studies to the manufacturing sector. Labor is one of the core resources necessary for entrepreneurship (Autio and Acs, 2010; Boudreaux and Nikolaev, 2019; Bhagavatula et al., 2010; De Clercq et al., 2013). Labor market regulation influences factors such as wage rates and labor adjustment costs (Wennekers et al., 2002) and is a significant determinant of entrepreneurship levels (da Fonseca, 2022). Scholars have extensively studied how labor market regulation affects entrepreneurship. These studies focus on entrepreneurship across all industries without distinguishing specific sectors. Manufacturing is a labor-intensive sector, making it particularly sensitive to labor market regulation. Therefore, following existing literature (e.g. Abd Rashid et al., 2023; Aparisi-Torrijo et al., 2023; Estrin et al., 2013; Stephan and Pathak, 2016), we define new manufacturing firms as those registered within 42 months or between 3 and 42 months. This allows us to derive a dependent variable reflecting the level of manufacturing entrepreneurship in cities, enabling empirical analysis of the impact of labor market regulation on this level. Thus, our research refines the study of entrepreneurship specifically to the manufacturing sector.

Our research expands the paradigm of studies on factors influencing entrepreneurship. Existing literature typically examines the direct relationship between labor market regulation and entrepreneurship (e.g. Angulo-Guerrero et al., 2023; Dilli, 2021; Desai et al., 2021; Van Stel et al., 2007). However, labor market regulation may also indirectly influence the level of manufacturing entrepreneurship through other variables. Notably, since 2008, China’s manufacturing sector has undergone a digital transformation, enhancing its digitalization and accelerating AI innovation alongside the promotion of the digital economy. Based on signaling theory, we argue that labor market regulation enhances AI innovation and digitalization in manufacturing, both of which benefit entrepreneurship. Thus, labor market regulation indirectly affects manufacturing entrepreneurship through AI innovation and digitalization. Empirical tests confirm this indirect effect. This research on indirect effects broadens the study of entrepreneurship factors.

The remainder of the paper is organized as follows: Section 2 develops the research hypotheses, Section 3 describes the data and variables, Section 4 presents empirical tests and Section 5 discusses the findings.

“Different people know different things”; information asymmetry arises when decision-makers possess varying quantities and qualities of information (Connelly et al., 2011; Stiglitz, 2002). Information asymmetry is widespread in economic life (Chen et al., 2022). Spence (1973) established signaling theory to address information asymmetry regarding job candidates’ qualities between employers and job seekers. This theory posits that information asymmetry can be mitigated through the transmission of various “signals” (Saxton et al., 2019). Signal transmission involves three elements: the signal sender, the signal receiver and the signal itself (Bergh et al., 2014; Ernst et al., 2022; Mirowska and Mesnet, 2022; Spence, 1973; Stiglitz, 2002). The signal sender is an insider with relevant information not accessible to outsiders (Kirmani and Rao, 2000; Ross, 1977; Spence, 1973). The signal receiver is an outsider lacking relevant information but seeking to obtain it (Connelly et al., 2011). A signal is an activity or attribute of an individual or organization that can alter the beliefs of others in the market (Di Pietro et al., 2023). Local government labor market regulation can transmit signals to market participants, thereby influencing manufacturing entrepreneurship (Figure 1).

Labor market regulation actions can convey signals to market participants. Laws are formulated at the national level but are implemented by local governments (Chan et al., 2010; Meyer and Nguyen, 2005). To prioritize short-term economic growth, local governments in China may relax labor law enforcement to varying degrees, resulting in differing levels of labor market regulation across regions (Lee, 2016). The intensity of labor market regulation within the same region also varies over time. In other words, to balance economic development and workers’ rights, local governments adjust their regulatory stance, where frequent regulation signals a preference for worker protection, and less regulation signals employer protection. Local governments’ labor market regulation actions send a dynamic signal of their “protection stance,” influencing decisions of manufacturing entrepreneurs and potential entrepreneurs, workers, banks and incumbent firms in both manufacturing and other sectors. According to signaling theory, the signal sender is the local government, while the receivers include manufacturing entrepreneurs, incumbent manufacturing and nonmanufacturing firms, workers and banks. The signal is the local government’s preference for worker protection or adherence to the rule of law.

The signals conveyed by labor market regulation actions have the greatest impact. To address the information asymmetry regarding protection preferences, local governments can use various methods, such as holding press conferences, media briefings or issuing calls for labor law compliance via Weibo and WeChat. However, the signals conveyed by labor market regulation actions are the strongest. The reason is that implementing labor market regulation requires local governments to strictly follow legal procedures, such as on-site investigations, confirmation of violations and delivery of penalty notices, which are time-consuming, manpower-intensive and costly. Compared to other methods, labor market regulation actions incur higher costs and the signals they convey are correspondingly costlier. According to signaling theory, higher signal costs reduce information asymmetry more effectively (Bergh et al., 2014), thereby increasing the signal’s impact. On the other hand, signaling theory posits that stronger signals capture the target audience’s attention more effectively (Connelly et al., 2011; Mirowska and Mesnet, 2022), thus amplifying their impact. The more frequent the labor market regulation actions, the stronger the signals they send to market participants, the more they attract the attention of manufacturing entrepreneurs and other signal recipients, and the greater their impact.

The signals conveyed by labor market regulation actions carry different meanings for manufacturing entrepreneurs and other market participants. According to signaling theory, signal recipients interpret the meaning of signals based on their preferences and positions (Lu and Chen, 2021). First, the interpretation by manufacturing entrepreneurs and potential entrepreneurs. Labor is one of the three essential capitals for entrepreneurship (Autio and Acs, 2010; Boudreaux and Nikolaev, 2019; Bhagavatula et al., 2010; De Clercq et al., 2013). Manufacturing is a labor-intensive industry, making its entrepreneurs and potential entrepreneurs more sensitive to labor market regulation signals, which are more likely to capture their attention. They typically interpret the signal as indicating that more frequent labor market regulation reflects a stronger inclination of local governments to protect workers, and vice versa. Second, the interpretation by incumbent manufacturing and nonmanufacturing firms. Labor market regulation, as a form of administrative regulation, has a deterrent effect (Hu et al., 2023; Syed, 2024; Wang et al., 2023), which intimidates law-abiding incumbent manufacturing and nonmanufacturing firms. Based on this deterrence, they also tend to interpret the signal as indicating that more frequent labor market regulation reflects a greater inclination of local governments to protect workers. Based on this deterrence, they also tend to interpret the signal as indicating that more frequent labor market regulation reflects a greater inclination of local governments to protect workers. Third, the interpretation by workers who provide human capital for manufacturing entrepreneurship. With the continuous implementation of China’s legal awareness programs, workers’ legal consciousness has been increasingly strengthened (Whiting, 2017), leading them to focus more on protecting their rights through legal means. Therefore, these workers often interpret local government labor market regulation as an indication of stronger protection for them, believing they can more readily seek redress through legal avenues when their rights are violated (labor arbitration and litigation following disputes between workers and employers represent passive protection by local governments, whereas labor market regulation actions are proactive measures by labor administrative departments. The more frequent these actions, the stronger the signal of active protection for workers). Fourth, the interpretation by banks. As mentioned earlier, in China, laws are formulated at the national level but implemented by local governments (Chan et al., 2010; Meyer and Nguyen, 2005). Local governments vary in their intensity of law enforcement. More frequent labor market regulation by local governments indicates a greater capacity and willingness to act in accordance with the law. Therefore, unlike other market participants, banks usually interpret this signal as reflecting local governments’ adherence to the rule of law.

The signals conveyed by labor market regulation actions influence manufacturing entrepreneurship; the more frequent the regulation, the more favorable it is for manufacturing entrepreneurship. The reason lies, first, in the early elimination of weaker manufacturing entrepreneurs, choosing to forgo entrepreneurship, leading to a purified entrepreneurial environment in the manufacturing sector. China has a vibrant entrepreneurial environment, with the second-largest number of start-ups globally after the USA (Dushnitsky and Yu, 2022). However, the potential manufacturing entrepreneurs vary in strength. When potential entrepreneurs perceive that local governments are inclined to protect workers more, weaker individuals may assess their resources and deem it difficult to meet regulatory requirements, thus abandoning entrepreneurship. This preemptively prevents these entrepreneurs from violating workers’ rights due to insufficient strength, preserving the region’s labor market reputation and enabling stronger entrepreneurs to attract high-quality workers for manufacturing entrepreneurship. Meanwhile, the early exit of weaker entrepreneurs creates more market space for stronger manufacturing entrepreneurs and potential entrepreneurs, increasing their chances of success. Second, labor resources are more secure. Existing manufacturing and nonmanufacturing firms interpret local government labor market regulation as a sign of worker protection. Based on the governance effects of administrative regulation (Zhang and Liu, 2023), they enhance internal management, improve internal controls and strengthen worker rights protection, thereby improving the region’s labor market reputation. Simultaneously, the policy’s coercive power and punitive mechanism serve as a strong deterrent, encouraging enterprises to prioritize the protection of workers’ rights and interests in their operations. When enterprises violate relevant regulations, they face significant penalties such as fines, business suspensions or even revocation of their business licenses. These direct financial consequences motivate enterprises to proactively enhance internal management, improve labor protection systems and regulate employment practices to avoid penalties due to violations. Unlike strategic responses based on interpreting signals in signaling theory, this behavior adjustment of enterprises emphasizes the direct influence of the policy’s mandatory constraints on corporate conduct. Laborers contributing to manufacturing entrepreneurship, interpreting government labor regulation as protective of workers, are more inclined to seek employment in such areas. China has built the world’s largest high-speed rail network (Yang et al., 2023; Zhou et al., 2023). Workers can easily and affordably relocate to regions with superior labor market reputations and stronger guarantees of worker rights. This indicates that the behavior of existing firms and laborers, driven by their interpretation of signals, supports manufacturing entrepreneurs in securing high-quality labor, thereby ensuring robust labor resources. Third, financial resources are more secure. Unlike Western countries, China has a bank-dominated financial system (Chen et al., 2022), with banks controlling most of the financial resource allocation. Large banks in China play a decisive role in financial resource allocation, distributing resources across regions through a head-office and branch system. These banks interpret local government labor market regulation as a signal that these governments have greater capacity and willingness to act in accordance with the law. For banks, this interpretation implies that if loans in these regions default, recovering funds through legal proceedings is more secure, reducing potential loan losses. Consequently, they are more inclined to allocate financial resources to these regions. All else being equal, manufacturing entrepreneurs are more likely to secure financial resources from banks to facilitate their entrepreneurial activities.

In summary, according to signaling theory, the interpretation of signals from local governments’ labor market regulation actions by market participants, and the resulting decisions, ultimately benefit manufacturing entrepreneurship. The more frequently local governments implement labor market regulation, the more likely weaker potential manufacturing entrepreneurs are eliminated early, and the region gains greater favor from workers and banks. Manufacturing entrepreneurs and potential entrepreneurs in the region are better able to secure labor and financial resources, both essential capitals for entrepreneurial activities such as manufacturing entrepreneurship (Autio and Acs, 2010; Boudreaux and Nikolaev, 2019; Bhagavatula et al., 2010; De Clercq et al., 2013). Therefore, labor market regulation is beneficial to manufacturing entrepreneurship. To this end, we propose the following hypotheses:

Based on the interpretation of signals by market participants, labor market regulation is conducive to AI innovation. The rationale is: first, incumbent firms (both in manufacturing and nonmanufacturing sectors) adopt AI to replace labor, facilitating AI innovation. Companies frequently resort to underreporting, misreporting or underpaying social security contributions to ease tax burdens (Li et al., 2021). Incumbent firms, perceiving local governments’ inclination to protect workers, adhere more rigorously to social security obligations mandated by law. This leads to higher costs. AI’s capability to automate tasks helps firms substitute labor with AI to manage increased costs (Acemoglu and Restrepo, 2019). This boosts the demand for AI innovation, thereby driving AI innovation from the demand side. Second, the protection of workers’ rights allows them to focus on AI innovation without concerns. AI innovation entails uncertainty, making it difficult to predict outcomes beforehand. This tendency often results in incumbent firms underpaying workers who contributed to successful AI innovation. Incumbent firms, perceiving local governments’ inclination to protect workers, adhere more strictly to labor laws and contracts, thus minimizing the chances of underpayment. This motivates workers to invest greater effort into AI innovation (Acharya et al., 2013, 2014). Furthermore, incumbent firms, interpreting a proworker stance from local governments, follow labor contract laws more rigorously, thereby lowering the risk of arbitrary layoffs. Such security encourages workers to invest in specialized human capital for AI innovation (Filippetti and Guy, 2020), enhancing their knowledge and skills, which, in turn, fosters AI innovation Third, easing financial constraints facilitates AI innovation. AI innovation, among other high-risk innovative activities, (Chen and Zhang, 2023; Chen et al., 2023) demands continuous financial backing. Banks, interpreting that local governments have greater capacity and willingness to act lawfully, allocate more financial resources to regions with more frequent labor market regulation. This leads to increased financial resources for incumbent firms engaging in AI innovation, thereby enhancing the progress of AI innovation.

AI innovation and other digital technology innovations create new entrepreneurial opportunities (Nambisan, 2017). This trend promotes manufacturing entrepreneurship focused on products like drones, self-driving cars, smart home devices, smart speakers and similar intelligent products. First, employees of incumbent firms might resign to start their own manufacturing businesses. In China, returnees typically gain experience and resources at parent companies before leaving to start their own ventures. For employees at manufacturing giants like Huawei, entrepreneurship is a prominent career path. AI-driven entrepreneurial opportunities may prompt incumbent manufacturing workers to resign and pursue their own ventures. These workers often choose to establish new ventures within the same industry as their previous employers (Agarwal et al., 2004; Bruneel et al., 2013; Garvin, 1983). To seize AI-driven opportunities, incumbent manufacturing workers are likely to pursue entrepreneurship in the manufacturing sector after leaving. Second, incumbent firms actively seek manufacturing entrepreneurship opportunities. With the emergence of new entrepreneurial opportunities, incumbent manufacturing firms might leverage their existing technologies and knowledge to launch new ventures for growth or strategic advantage (Covin and Miles, 1999; Ireland et al., 2009; Li et al., 2020; Wallin and Dahlstrand, 2006). This trend further encourages manufacturing entrepreneurship. Moreover, nonmanufacturing employees and firms, motivated by diversification strategies, might venture into manufacturing entrepreneurship leveraging AI innovation.

In conclusion, market participants’ interpretation of labor market regulation signals promotes active AI innovation and the provision of necessary labor and financial resources, thus elevating AI innovation levels. In turn, this creates new entrepreneurial opportunities, fostering manufacturing entrepreneurship. Accordingly, we propose the following hypotheses:

Based on the interpretation of signals by market participants, labor market regulation can promote manufacturing digitization. The underlying logic is: first, opting for digitization as a proactive response to increased costs. As discussed earlier, incumbent firms, perceiving a proworker stance from local governments, adhere more strictly to social security obligations, leading to higher costs. Since 2008, manufacturing digitization has become a significant trend (Chen et al., 2022; Horvat et al., 2019). Digitization boosts production efficiency in manufacturing (Horvat et al., 2019; Luo et al., 2012) while saving resources, minimizing waste and enhancing resource utilization efficiency (Brüggemann et al., 2020). This creates opportunities for incumbent manufacturers to cope with higher costs, thus advancing manufacturing digitization. Second, securing workers’ rights encourages active implementation of digitization. Digitization represents a fundamental transformation of elements such as corporate technology, organizational structure and identity, value propositions and business strategies (Drechsler et al., 2020; Vial, 2021; Wessel et al., 2021). It also entails the restructuring of collaborative relationships between firms and their upstream suppliers and downstream distributors (Butollo, 2021). In other words, digitization is a synthesis of multiple digital innovations (Hinings et al., 2018), requiring sustained and proactive participation from workers across all stages and processes over extended periods. As discussed earlier, incumbent firms, perceiving local governments’ inclination to protect workers, adhere more strictly to labor laws and contracts. Corporate workers’ rights are more securely protected. This strengthens the stability of corporate workers, securing their long-term active involvement in the digitization of manufacturing firms, thus enhancing manufacturing digitization levels. Furthermore, banks, perceiving that local governments are more capable and willing to enforce laws, tend to channel more financial resources to regions with stringent labor market regulation, supporting digitization in incumbent manufacturing firms and advancing the digitization level.

Digitization in manufacturing can drive manufacturing entrepreneurship. First, it intensifies industry competition, prompting incumbent manufacturing firms to pursue entrepreneurship. Through online platforms and other channels, digitization expands the geographical reach of products and shortens their distribution time (Lee and Falahat, 2019). It minimizes spatial and temporal gaps between companies and customers, thereby reducing entry barriers for newcomers (Knudsen et al., 2021). Consequently, this escalates competition within the manufacturing sector. In addition, digitization unlocks new business opportunities (Bouncken and Kraus, 2022), generating fresh entrepreneurial prospects for manufacturing firms and inspiring innovative entrepreneurial activities (Dabbous et al., 2023). To counteract growing competition, incumbent manufacturing firms might proactively engage in manufacturing entrepreneurship. Second, it facilitates the development of an entrepreneurial ecosystem, which supports manufacturing entrepreneurship. Manufacturing digitization leads to the digitization and networking of the production and business activities of incumbent manufacturing firms. Connections between incumbent manufacturing firms become closer, making resource and knowledge sharing more convenient and efficient. Therefore, manufacturing digitization enables coupling and technological complementarity among incumbent manufacturing firms (Bouncken and Kraus, 2022). This notably enhances the economic environment for manufacturing entrepreneurship and other entrepreneurial activities (Acs et al., 2016; Roundy et al., 2017), supporting manufacturing entrepreneurship. Moreover, digitization in manufacturing and other industries helps reduce information asymmetry between banks and firms, encouraging banks to allocate more financial resources for manufacturing entrepreneurship by incumbent firms. This further drive manufacturing entrepreneurship.

In conclusion, market participants’ interpretation of signals from labor market regulation drives active manufacturing digitization or financial support for it, thus advancing manufacturing digitization levels. In turn, this foster manufacturing entrepreneurship. Therefore, we propose the following hypotheses:

China is vast with uneven development (Chen et al., 2022), and the endowments of resources and firm density vary significantly across regions. Our data shows that in 2020, the lowest firm density was 0.1361 operating firms per square kilometer, whereas the highest was 81.0634. High firm density is beneficial for manufacturing entrepreneurship. This is because entrepreneurship is rooted in innovation, and manufacturing entrepreneurship demands ongoing innovation to surpass existing products (Bradley et al., 2012). It necessitates the integration of diverse expertise to produce creative ideas (Kollmann et al., 2020; Sciascia et al., 2013). High firm density signals both market presence and opportunities for firms to interact and exchange expertise (Walzer et al., 2007). Higher firm density makes it easier for incumbent manufacturers to access diverse expertise needed for innovation, thereby enhancing their chances for manufacturing entrepreneurship. Moreover, higher firm density leads to more frequent exchanges between workers of incumbent firms and those of other companies, facilitating the exchange of creative ideas and motivating workers to promote manufacturing entrepreneurship internally or pursue it independently. Higher firm density likely means more upstream and downstream partners available for manufacturing entrepreneurship, thereby supporting it more effectively. Therefore, firm density has a positive impact on manufacturing entrepreneurship and other entrepreneurial activities (Lowrey, 2004). In addition, labor and financial resources are essential capitals required for manufacturing entrepreneurship and other entrepreneurial activities (Autio and Acs, 2010; Boudreaux and Nikolaev, 2019; Bhagavatula et al., 2010; De Clercq et al., 2013). In a given period, the manufacturing labor force and financial resources allocated by banks to manufacturing entrepreneurship in a region are limited, thereby capping the level of manufacturing entrepreneurship. As a result, once firm density has elevated manufacturing entrepreneurship, the potential for labor market regulation to further boost it diminishes, and vice versa. Consequently, the higher the firm density, the less impact labor market regulation has on further enhancing manufacturing entrepreneurship levels. Therefore, we propose the following hypotheses:

Entrepreneurial activity in China is extremely vibrant, with the number of start-ups second only to the USA (Dushnitsky and Yu, 2022). Overall, entrepreneurs in sectors other than manufacturing are also actively driving entrepreneurship. Similarly, resource endowments vary from region to region, and the levels of nonmanufacturing entrepreneurship vary from region to region. Our data indicates that in 2020, nonmanufacturing entrepreneurship levels ranged from a minimum of 26.6290 to a maximum of 297.0066 firms per 10,000 people. Nonmanufacturing entrepreneurship is beneficial for manufacturing entrepreneurship. This is because start-ups frequently encounter resource shortages and have a high tendency to make mistakes. As a result, start-ups have a high failure rate (Azoulay et al., 2020) and represent a high-risk activity (Astebro et al., 2014). Fear of failure is a major deterrent to entrepreneurship (Arenius and Minniti, 2005; Caliendo et al., 2009).Regions with higher nonmanufacturing entrepreneurship levels often have more risk-taking individuals, leading to higher manufacturing entrepreneurship levels. More importantly, manufacturing entrepreneurs typically consult specific groups (like friends, colleagues and family) before initiating their ventures. Positive feedback from these groups significantly influences the formation of entrepreneurial intentions, which can subsequently translate into manufacturing entrepreneurship actions (Ajzen, 1991; Hagger et al., 2022). The higher the level of nonmanufacturing entrepreneurship, the more willing people in the region are to take risks and the more likely specific groups are to give positive evaluations of manufacturing entrepreneurs’ actions, thereby facilitating manufacturing entrepreneurship. Similarly, the level of manufacturing entrepreneurship in a region is capped within a certain period. After nonmanufacturing entrepreneurship drives manufacturing entrepreneurship, the room for labor market regulation to further enhance manufacturing entrepreneurship levels becomes relatively smaller, and vice versa. In other word, with rising levels of nonmanufacturing entrepreneurship, the impact of labor market regulation on boosting manufacturing entrepreneurship declines progressively. Therefore, we propose the following hypotheses:

In summary, we propose five hypotheses. The relationships among them are illustrated in Figure 2.

Regarding the factors influencing entrepreneurship, some scholars conduct empirical tests based on microlevel data, while others use macrolevel data. Given data availability and the macro nature of the research question, we use macrolevel data for empirical testing. Administratively, China is divided into 31 provincial-level units (excluding Hong Kong, Macau and Taiwan), 293 prefecture-level cities and 1,301 counties. Due to China’s vastness and uneven development, provincial-level data struggles to reflect this diversity adequately. County-level data lacks sufficient detail. Therefore, scholars typically use prefecture-level city data for empirical tests. Furthermore, Beijing, Shanghai, Tianjin and Chongqing are considered provincial-level administrative units in China. Scholars usually include them in the 293 prefecture-level cities collectively referred to as Chinese cities for empirical testing. Following this approach, we excluded cities with severe data deficiencies and obtained a sample of 280 cities.

China’s new Labor Contract Law was implemented in 2008. Data from the “China City Statistical Yearbook,” issued by the National Bureau of Statistics, extends up to 2020. Therefore, we constructed a city-level panel data set based on samples from 280 Chinese cities from 2008 to 2020 for empirical testing. Data processing steps included:

• removing missing samples;

• excluding cities with a single observation;

• applying linear interpolation for 2010 city green area data; and

• using natural logarithms to minimize outliers’ effects and 1% Winsorization for other continuous variables.

As a result, we obtained 3,367 year-city observations.

Data on start-ups across all industries, manufacturing industry start-ups, the number of operating firms, labor administrative penalties and various administrative penalties were sourced from Shanghai DZH Data Technology Co., Ltd., using Web scraping from different administrative agencies. AI patent application data was collected by searching city by city with the keyword “artificial intelligence” from Suzhou Industrial Park PatSnap Information Technology Co., Ltd., (www.zhihuiya.com), with sources from the China National Intellectual Property Administration. Similarly, data on various patent applications was obtained city by city from Suzhou Industrial Park PatSnap Information Technology Co., Ltd., allowing for searches of all types of patent applications without keywords. Data on the number of entrepreneurial policies in each city was collected using the keyword “entrepreneurship” from Bailu Thinktank (www.bailuzhiku.com), with sources from local government data. Other data is sourced from the “China City Statistical Yearbook” published by the National Bureau of Statistics and the People’s Bank of China (www.pbc.gov.cn).

Based on previous studies, Table 1 presents the independent, dependent, mediating and control variables for this research. To enhance the visual presentation of the estimation results, we adjusted the decimal points of certain variables. While this adjustment alters the scale of the coefficients, it does not affect the overall conclusions.

The level of manufacturing entrepreneurship is the dependent variable in this study. Abd Rashid et al. (2023) and Stephan and Pathak (2016) consider firms established for more than 3 months but less than 42 months as start-ups. The listing criteria for China’s Growth Enterprise Market require enterprises to have been operating for at least three years, or 36 months. We consider 42 months to be an appropriate threshold to differentiate startup enterprises from other types of enterprises. Following their approach, we retrieved data from Shanghai DZH Caifu Data Technology Co., Ltd., which sources its data from the State Administration for Market Regulation but offers more convenient access. We extracted the number of manufacturing firms registered in cities for more than 3 months but less than 42 months, defined as Cnew342, and divided it by the total population of the city to obtain MENP, which serves as a proxy for the level of manufacturing entrepreneurship. As mentioned earlier, entrepreneurs who exit the market early do not need to be identified by regulatory authorities, as they have not established a company. In addition, Aparisi-Torrijo et al. (2023) and Estrin et al. (2013) regard firms established for less than 42 months as start-ups. Following their approach, we retrieved data from Shanghai DZH Caifu Data Technology Co., Ltd., extracting the number of manufacturing firms registered for less than 42 months, defined as Cnew42, and divided it by the city’s total population to obtain rMENP, which serves as a proxy for the level of manufacturing entrepreneurship in robustness tests – this ensures that entrepreneurial firms with an operational lifespan of less than 3 months (yet demonstrating entrepreneurial spirit) are not omitted.

Labor market regulation intensity is the independent variable in this study. Unlike existing research, we construct a city-level measure of labor market regulation intensity. In China, the “Regulations on Labor Security Supervision” stipulate that labor administrative departments may impose administrative penalties and measures based on investigation and inspection results for violations of labor laws, regulations or rules. Among these, administrative penalties are the most severe and best reflect the intensity of labor market regulation. Therefore, we collected data from Shanghai DZH Caifu Data Technology Co., Ltd., which uses Web scraping technology to gather information from various levels of government on the number of labor administrative penalties received by firms registered in cities each year and the number of operating firms. Then, we calculated LPUN, a proxy for the intensity of labor market regulation, by dividing the number of labor administrative penalties by the number of operating firms. The natural logarithm of the number of labor administrative penalties received by firms registered in cities each year, plus one, was calculated to obtain rLPUN, another proxy for labor market regulation intensity, for robustness testing.

AI innovation level (AI): Following Chen et al. (2023), we use the natural logarithm of the number of AI patent applications filed by firms within a city, plus one, as a proxy for AI innovation level.

Manufacturing digitization level (MDIG): There are no publicly available data on the level of MDIG in cities, and China’s NBS divides the manufacturing industry into 29 segments, including processing of food from agricultural products, manufacture of tobacco, textile, furniture, medicine, vehicle, and so on. The digitization requirements of each segment differ. We constructed the keywords listed in Table 2 based on the specific needs of each segment. Based on these keywords, we collect the number of enterprises whose business scope contains these keywords and their registered capital from the State Administration for Market Supervision and Administration and obtain 58 ( = 29 * 2) data items for each city from 2008 to 2022. For the few data items that are zero, we refer to Li et al. (2020) and fill them with random numbers between 0 and 0.0001. We then use factor analysis to construct an urban manufacturing digitization index.

The Kaiser–Meyer–Olkin (KMO) and Bartlett’s spherical tests are prerequisites for the factor analysis method, with KMO = 0.886, which is greater than 0.6. The chi-squared statistic of Bartlett’s test is 3.52e + 05, with a p -value < 0.000158. A total of 66 data items were subjected to factor analysis, resulting in 16 factors with a cumulative contribution rate of 0.7630. Thus, the data meet the prerequisites of factor analysis. Based on the factor analysis method, we measured the raw digital index of each city and then calculated Mdigindex using (digital-Min)/(MAX-Min)*10 to obtain Mdigindex as the MDIG. MIDG indicates the extent of digitalization within the manufacturing industry; a higher value signifies a more advanced level of digital integration.

All control variables are listed in Table 1, and their actual selection varies depending on the dependent variables, as detailed later.

Table 3 presents the descriptive statistics for the dependent, independent, mediating and control variables. As shown in the table, first, the mean of manufacturing entrepreneurship level (MENP) is 5.2749, meaning that on average, there are 5.2749 manufacturing start-ups per 10,000 people, with a minimum of 0.3747 and a maximum of 47.1100. The gap between the minimum and maximum is significant, aligning with the fundamental national condition of uneven development. Second, the mean of manufacturing entrepreneurship level (rMENP) is 5.7231, higher than that of MENP, as the latter excludes manufacturing firms established for less than 3 months. Third, the mean of labor market regulation intensity (LREG) is 0.4567, indicating that, on average, every 10,000 firms receive 0.4567 administrative penalties from labor authorities, with a minimum of 0.0000 and a maximum of 7.7441. The significant gap also aligns with the fundamental national condition of uneven development.

The main variables in this study are manufacturing entrepreneurship level (MENP, rMENP), labor market regulation intensity (LREG, rLREG), AI innovation level (AI) and manufacturing digitization level (MDIG). The analysis in Section 2 suggests that there should be positive linear relationships among these variables. Our correlation tests show that there are indeed significant positive correlations among these variables, but the strengths of the correlations vary (see the left panel of Figure 3).

In the left panel, both the horizontal and vertical axes represent key variables. The numbers inside the circles at the intersections indicate the correlation coefficients between the variables; ***denotes significance at the 1% level. The right panel shows the regression coefficients and 95% confidence intervals of labor market regulation intensity from Columns (1) to (4) in Table 4. Purple corresponds to Column (1), red to Column (2), green to Column (3) and black to Column (4). The regression coefficients of labor market regulation intensity are represented by error bars. An error bar that crosses the red vertical line indicates that the coefficient is not significantly different from zero; error bars entirely on either side of the red line indicate robust negative or positive correlations, respectively. The same applies throughout.

China is a geographically vast and unevenly developed country, as noted by Chen et al. (2022, 2023). Each city in China has unique characteristics that impact its climate-friendly technology innovation, and these characteristics remain stable over time. To account for this variable, we include city fixed effects in our study. In addition, China follows a government-led economic structure, wherein the central government annually executes initiatives to promote various types of entrepreneurship including manufacturing entrepreneurship, throughout all cities. These policies have a pervasive effect on every city and see annual modifications. To account for this factor, we include year fixed effects in our study. To inform our study, we consult the relevant scholarly works on China, such as the publications by Chen et al. in 2022 and 2023. Accordingly, we constructed the following econometric model to assess the impact of labor market regulation intensity on manufacturing entrepreneurship levels:

where i and t, respectively, denote the city and year indices; captures the city-specific fixed effects; captures the year-specific fixed effects; and represents the random error term is the dependent variable, signifying the level of manufacturing entrepreneurship in city i in year t. is the independent variable, indicating the intensity of labor market regulation in city i in year t, with as its coefficient. If is significantly positive, then labor market regulation is inferred to enhance the city’s level of manufacturing entrepreneurship denotes control variables, and following existing literature (e.g. Agostino et al., 2020; Ali et al., 2020; Darnihamedani et al., 2018; Levie and Autio, 2011; Yoon et al., 2018), we control for Government Technology Expenditure, Innovation Level, Population Density, Human Capital Level, Unemployment Rate, Economic Development Level, Economic Growth Rate, Industrial Structure Level, Financial Development Level. The analysis of Section 2 indicates that enterprise density and the level of nonmanufacturing entrepreneurship also affect the level of manufacturing entrepreneurship, so we control for these two variables as well. In addition, administrative regulations other than labor market regulation could impact manufacturing entrepreneurship; thus, we control for the nonlabor Regulation Intensity. Given that cities may develop entrepreneurship policies that are tailored to their realities in response to central-level entrepreneurship policies, we also control for entrepreneurship policy support.

We use MENP as the dependent variable and gradually add control variables. A two-way fixed-effects (FE) panel model with year and city effects is used to estimate the impact of labor market regulation intensity on manufacturing entrepreneurship levels. The results are presented in Table 1, Columns (1)–(4).As shown in Table 1, the regression coefficients for labor market regulation intensity are significantly positive at the 1% significance level (p < 0.0001). According to Column (4), for each unit increase in labor market regulation intensity – where one unit corresponds to one additional penalty per 10,000 firms – the level of manufacturing entrepreneurship rises by 0.2137 units, where one unit represents one additional manufacturing start-up per 10,000 people. In other words, if labor administrative authorities impose one labor penalty per 10,000 firms on average, the number of manufacturing start-ups per 10,000 people increases by 0.2137.

The right panel of Figure 3 presents the regression coefficients of labor market regulation intensity and their 95% confidence intervals from Columns (1)–(4) in Table 4. In the right panel of Figure 3, the error bars for labor market regulation intensity are all positioned to the right of the red vertical line, indicating a consistently significant positive correlation between labor market regulation intensity and manufacturing entrepreneurship levels. Figure 4 illustrates the marginal effect of labor market regulation intensity on urban manufacturing entrepreneurship levels. Figure 4 shows that after progressively controlling for other factors, manufacturing entrepreneurship levels in cities exhibit an upward trend as labor market regulation intensity increases.

In conclusion, H1 is supported.

This figure is based on the results from Columns (1)–(4) in Table 4 and illustrates the marginal effect of labor market regulation intensity on manufacturing entrepreneurship, holding other factors constant. The upper-left panel corresponds to Column (1), and the upper-right panel corresponds to Column (2) of Table 4. The lower-left panel corresponds to Column (3), and the lower-right panel corresponds to Column (4) of Table 4.

Theoretical analysis and empirical results presented earlier confirm that labor market regulation intensity has a significant impact on manufacturing entrepreneurship levels. Data on labor market regulation intensity and manufacturing entrepreneurship levels were extracted via web scraping by Shanghai DZH Data Technology Co., Ltd., from labor administrative authorities and the State Administration for Market Regulation, which may introduce measurement errors. Such measurement errors can result in endogeneity issues in labor market regulation intensity (Anderson et al., 2019). Furthermore, endogeneity may arise due to omitted variables and unobservable factors. We use a two-way FE panel model with city and year fixed effects to minimize the impact of unobservable factors and omitted variables. The two-stage least squares (2SLS) method, based on instrumental variables, can effectively address endogeneity (Anderson et al., 2019). Therefore, we address endogeneity using the 2SLS approach.

For the instrumental variable, we construct it by multiplying the ratio of secondary industry value added to agricultural value added in 2006 by the growth rate of labor market regulation intensity. This method follows the approaches of Liu et al. (2024), Chen and Qin (2022) and Chen et al. (2024), and is grounded in the Bartik instrumental variable technique. It leverages historical industrial structure indicators alongside the growth rate of labor regulation intensity to generate an estimated measure of labor market regulation intensity. Specifically, the 2006 ratio of secondary to agricultural value added captures the city’s initial industrial structure, while the growth rate of labor market regulation intensity reflects evolving regulatory efforts by labor authorities. Multiplying these two components produces an estimate closely linked to labor market regulation intensity, thereby meeting the relevance criterion.

This instrumental variable is derived from historical industrial structure data from 2006, representing well-established historical facts that do not directly impact the current level of manufacturing entrepreneurship. In addition, the labor market regulation intensity for each city is calculated using the same data collection approach, with its growth rate shaped by a common institutional environment and measurement system rather than city-specific characteristics. By controlling for city and year fixed effects, this instrumental variable remains uncorrelated with the residuals of other factors influencing manufacturing entrepreneurship, thereby meeting the exogeneity criterion.

We use ivLREG as instrumental variables and estimate equation (1) using 2SLS. The Cragg–Donald F-statistic for weak instrument testing is 51.13, surpassing the critical value of 8.96 at a 15% bias level, thereby rejecting the weak instrument hypothesis. The Anderson canonical correlation LR statistic is 51.15, with a p-value < 0.0001, leading to the rejection of the under-identification hypothesis. Thus, ivLREG are confirmed as valid instrumental variables. The first-stage estimation results indicate that ivLREG is positively correlated with labor market regulation intensity at the 1% significance level, while the per capita labor administrative penalty rate of other cities in the same year is negatively correlated with labor market regulation intensity at the 1% significance level. Using ivLREG as instrumental variables, we reestimate equation (1) using 2SLS, with the results presented in Table 5, Panel A, Column (1). As observed in Column (1), the regression coefficient of labor market regulation intensity remains significantly positive at the 1% significance level, confirming that H1 still holds after addressing endogeneity.

We replace the independent variable with rLREG and reestimate equation (1) using a FE model, with the results presented in Table 5, Panel A, Column (2).The dependent variable is replaced with rMENP, and equation (1) is reestimated using the FE model, with the results shown in Table 5, Panel A, Column (3). The analysis in Section 2 indicates that AI innovation level and manufacturing digitization level also affect manufacturing entrepreneurship. We treat them as mediating variables and include them as control variables. Consequently, controlling for these two variables, we reestimate equation (1) using the FE model, with results presented in Table 5, Panel A, Column (4).

The impact of labor market regulation on manufacturing entrepreneurship may involve not only concurrent effects but also lagged effects or cumulative effects. From a theoretical perspective and based on previous research, numerous studies support the inclusion of lagged terms of regulatory intensity for testing. Studies on labor market policies highlight that there is often a significant time lag between the introduction and enforcement of regulations and their eventual effects on factors (such as employment and wage levels) in the labor market. When analyzing the impact of labor market regulation on manufacturing entrepreneurship, focusing solely on the current period’s regulatory intensity may therefore overlook important long-term consequences. For instance, while stringent labor market regulations might initially raise costs and dampen entrepreneurial motivation of enterprises, over time they can encourage enterprises to invest more in innovation, enhance production efficiency and ultimately foster entrepreneurial activity (Acemoglu and Restrepo, 2019). Consequently, to thoroughly and accurately evaluate the impact of labor market regulation on manufacturing entrepreneurship, it is both necessary and appropriate to incorporate lagged regulatory intensity terms in the analysis, as this approach helps to uncover the dynamic, long-term mechanisms at play.

We control for province fixed effects and reestimate equation (1) using a FE model, with the results presented in Table 5, Panel B, Column (1). We reestimate equation (1) using a random-effects model, with the results shown in Table 5, Panel B, Column (2). Our research focuses on analyzing the concurrent effects of labor market regulations on manufacturing entrepreneurship. Considering elements such as enterprise decision-making processes, the pace of policy implementation, and environmental inertia, regulatory measures may produce delayed or cumulative behavioral effects. To account for this, we introduced a one-period lag for the other variables and reestimated equation (1), with the results displayed in Table 5, Panel B, Column (3). We reestimate equation (1) using the maximum likelihood estimation method, with the results presented in Table 5, Panel B, Column (4).

From Table 5, Panels A and B, the regression coefficients of labor market regulation intensity remain significantly positive at the 1% significance level. Therefore, H1 is robustly supported.

Following existing literature on indirect effects (e.g. Bochkay et al., 2022; Chen et al., 2022; Chen et al., 2023), we design the following model:

In equation (2), $AIit$ represents the level of AI innovation in city i in year t. $Z$ includes control variables, following existing literature (e.g. Chen et al., 2023), such as human capital level, financial efficiency, population growth rate, population density, economic development level, industrial structure level, government intervention, nonlabor regulation intensity. Other variables are the same as in equation (1). Equation (3) adds the AI innovation level to the base of equation (1).

We estimate equations (2) and (3) using a FE model, with the results presented in Table 6, Panel A, Column (1), and Panel B, Column (1). In Panel A, Column (1), the regression coefficient of labor market regulation intensity (LREG) is significantly positive at the 1% level, suggesting that labor market regulation promotes AI innovation. The AI innovation level is measured as the natural logarithm of the number of AI patent applications filed by firms within a city, plus one. Thus, the regression coefficient of LREG in Panel A, Column (1), approximately indicates that a one-unit increase in labor market regulation intensity raises AI innovation levels by 5.46%. In Panel B, Column (1), the regression coefficient of AI innovation level is 0.4764, approximately indicating that for each unit increase in AI innovation level, manufacturing entrepreneurship increases by 0.4764 units, equivalent to an additional 0.4764 manufacturing start-ups per 10,000 people.

In Figure 5, the purple error bars in the left panel display the regression coefficient of labor market regulation intensity (LREG) from Panel A, Column (1), along with its 95% confidence interval. The first plot on the upper left of Figure 6 illustrates the change in AI innovation level resulting from variations in labor market regulation intensity, holding other factors constant, along with its 95% confidence interval. In the right panel of Figure 5, the purple error bars represent the regression coefficients of AI innovation level (AI) and labor market regulation intensity (LREG) from Panel B, Column (1), along with their 95% confidence intervals. The first plot on the lower left of Figure 6 illustrates the change in manufacturing entrepreneurship level resulting from variations in AI innovation level, holding other factors constant, along with its 95% confidence interval. These figures consistently indicate that labor market regulation promotes AI innovation, which, in turn, facilitates an increase in manufacturing entrepreneurship.

Collectively, these figures confirm that H2 holds.

For the same reasons outlined in Section 4.1.3, labor market regulation intensity may be endogenous when estimating equation (2). Here, we use the same instrumental variables as in Section 4.1.3 and reestimate equation (2) using the 2SLS method, with the results presented in Table 6, Panel A, Column (2). In the left panel of Figure 5, the red error bars display the regression coefficient of labor market regulation intensity (LREG) along with its 95% confidence interval. The second plot from the left on the upper side of Figure 6 illustrates the impact of changes in labor market regulation intensity on AI innovation level, holding other factors constant, along with its 95% confidence interval.

Referring to the method of Chen et al. (2022), we compute the average number of digital technology patent applications (encompassing artificial intelligence, blockchain, big data, cloud computing and the Internet of Things) from other cities in the same year as ODTCH and use it as an instrumental variable. Along with GREE and OLRG, we reestimate equation (3) using 2SLS, with results displayed in Table 6, Panel B, Column (2). In the right panel of Figure 5, the red error bars represent the regression coefficients of AI innovation (AI) and labor market regulation intensity (LREG) along with their 95% confidence intervals. The second graph from the left at the bottom of Figure 6 shows how changes in AI innovation levels affect manufacturing entrepreneurship while keeping other conditions unchanged, with the 95% confidence interval displayed.

The visual analysis confirms that, after addressing endogeneity, labor market regulation fosters AI innovation, which subsequently enhances manufacturing entrepreneurship. Therefore, after controlling for endogeneity, H2 is supported.

Furthermore, given that manufacturing digitization can create demand for AI innovation, thereby fostering AI advancements. The independent variable is substituted with rLREG, and equation (2) is reestimated using the FE model, with results shown in Table 6, Panel A, Column (3). The regression coefficient of rLREG and its 95% confidence interval are represented by the green error bars in the left panel of Figure 5. The third graph from the left at the top of Figure 6 displays how changes in labor market regulation intensity influence AI innovation levels under unchanged conditions, with the 95% confidence interval. Manufacturing digitization is added as a control, and equation (2) is reestimated using the FE model, with findings shown in Table 6, Panel A, Column (4). The regression coefficient of LREG and its 95% confidence interval are represented by the black error bars in the left panel of Figure 5. The fourth graph from the left at the top of Figure 6 presents how changes in labor market regulation intensity affect AI innovation levels under constant conditions, with the 95% confidence interval. These visualizations further confirm that labor market regulation fosters AI innovation.

The independent variable is substituted with rLREG, and equation (3) is reestimated using the FE model, with results displayed in Table 6, Panel B, Column (3). In Figure 5 (right panel), the green error bars illustrate the regression coefficient of rLREG along with its 95% confidence interval. The third graph from the left at the bottom of Figure 6 demonstrates how changes in AI innovation levels affect manufacturing entrepreneurship while keeping other factors unchanged, with the 95% confidence interval. The dependent variable is replaced with rMENP, and equation (3) is reestimated using the FE model, with findings shown in Table 6, Panel B, Column (4). The fourth graph from the left at the bottom of Figure 6 shows how changes in AI innovation levels impact manufacturing entrepreneurship while keeping other conditions unchanged, with the 95% confidence interval displayed. These visuals confirm that AI innovation facilitates manufacturing entrepreneurship.

In summary, the validity of H2 is confirmed with robustness.

Similarly, we design the following model with reference to existing literature on indirect effects (e.g. Bochkay et al., 2022; Chen et al., 2022; Chen et al., 2023):

In equation (4), $MDIGit$ represents the manufacturing digitization level of city i in year t. Rdenotes the control variables. Following existing studies (e.g. Chen et al., 2022), we control for economic development level, economic growth rate, industrial structure level, government intervention, human capital level, financial development, financial efficiency, population growth rate, urbanization rate, population density and nonlabor regulatory intensity. Other variables are the same as in equation (1). Equation (5) extends equation (1) by including the manufacturing digitization level.

Following the procedure in Section 4.2, we estimate equations (4) and (5), with the results presented in Table 7. The regression coefficient of labor market regulation intensity and its 95% confidence interval from Table 7 Panel A are shown in the left panel of Figure 7. The regression coefficients and 95% confidence intervals for labor market regulation intensity and MDIG in Panel B are shown in the right panel of Figure 7. Purple corresponds to Column (1), red to Column (2), green to Column (3) and black to Column (4). Based on the estimation results from Panel A, the four upper plots of Figure 8 display the changes in manufacturing digitization level resulting from variations in labor market regulation intensity, holding other factors constant, along with their 95% confidence intervals. Based on the estimation results from Panel B, the four lower plots of Figure 8 show the changes in manufacturing entrepreneurship level caused by variations in manufacturing digitization, under constant conditions, with their 95% confidence intervals. From left to right, these correspond to Columns (1) through (4), respectively.

Taken together, these figures support the validity of H3; the conclusion remains robust after accounting for endogeneity and conducting additional robustness checks.

To examine whether the effect of labor market regulation on manufacturing entrepreneurship varies with enterprise density, we construct the following model:

In equation (6), $EDENit$ epresents the enterprise density of city i in year t. $EDENit*LREGit$ is the interaction term between labor market regulation intensity and enterprise density is included, with $ℑ$ as its coefficient. A significantly negative $ℑ$ suggests that the effect of labor market regulation on manufacturing entrepreneurship weakens as enterprise density increases.

Using MENP and rMENP as dependent variables and LREG as the independent variable, we estimate equation (6) using the FE model. The results are presented in Columns (1) and (2) of Table 8. Replacing the independent variable with rLREG, we reestimate equation (6) using the FE model. The results are presented in Columns (3) and (4) of Table 8. As shown in Table 8, the coefficients of labor market regulation intensity and enterprise density are significantly positive at the 1% level, consistent with our analysis in Section 2. However, the interaction terms between them are significantly negative at the 1% level. This suggests that, holding other factors constant, the effect of labor market regulation on promoting manufacturing entrepreneurship gradually weakens as enterprise density increases. Based on the estimation results from Columns (1) to (4) in Table 8, Figure 9 illustrates the marginal effects of labor market regulation on manufacturing entrepreneurship along with their 95% confidence intervals, as enterprise density increases while holding other variables constant. Both Table 8 and Figure 9 support H4.

In the figure, the first subplot on the top corresponds to Column (1) of Table 8, and the second subplot on the top corresponds to Column (2); the first subplot on the bottom corresponds to Column (3) and the second subplot on the bottom corresponds to Column (4). The horizontal axis represents enterprise density, and the vertical axis represents the marginal effect of labor market regulation intensity on manufacturing entrepreneurship.

To examine whether the impact of labor market regulation on manufacturing entrepreneurship changes with the level of nonmanufacturing entrepreneurship, we construct the following model:

In equation (7), $OENPit$ represents the level of nonmanufacturing entrepreneurship in city i in year t. $OENP*LREGit$ is the interaction term between labor market regulation intensity and the level of nonmanufacturing entrepreneurship. $ℑ$ is its coefficient. If it is significantly negative, it indicates that the effect of labor market regulation on manufacturing entrepreneurship gradually weakens as the level of nonmanufacturing entrepreneurship increases.

Using MENP and rMENP as dependent variables and LREG as the independent variable, we estimate equation (7) with a FE model. The results are presented in Columns (1) and (2) of Table 9. Replacing the independent variable with rLREG, we reestimate equation (7) using the FE model, and the results are reported in Columns (3) and (4) of Table 9. As shown in Table 9, the coefficients of labor market regulation intensity and nonmanufacturing entrepreneurship are significantly positive at the 1% level, consistent with our analysis in Section 2. However, their interaction terms are significantly negative at the 1% level. This suggests that, holding other factors constant, the positive effect of labor market regulation on manufacturing entrepreneurship weakens as nonmanufacturing entrepreneurship increases. Based on the estimation results from Columns (1) to (4) of Table 9, Figure 10 illustrates the marginal effect of labor market regulation on manufacturing entrepreneurship and its 95% confidence interval as nonmanufacturing entrepreneurship increases, holding other conditions constant. Both Table 9 and Figure 10 provide support for H5.

In the figure, the first subplot on the top corresponds to Column (1) of Table 9, the second subplot on the top to Column (2); the first subplot on the bottom to Column (3) and the second on the bottom to Column (4). The horizontal axis shows nonmanufacturing entrepreneurship levels, while the vertical axis depicts the marginal impact of labor market regulation intensity on manufacturing entrepreneurship.

Our study examines how labor market regulation implemented by local governments in China affects local manufacturing entrepreneurship, including its direct, indirect and heterogeneous impacts. Using signaling theory, we conceptualize local governments as signallers and various stakeholders – entrepreneurs and potential entrepreneurs in manufacturing, workers supplying human capital, banks providing finance and incumbent firms – as receivers. We examine the signals conveyed through labor regulation and how these signals are interpreted and acted upon by receivers. Building on this, we propose that labor regulation eliminates weaker potential entrepreneurs early, incentivizes stronger candidates to launch ventures and encourages ongoing entrepreneurship among existing manufacturers. Meanwhile, such regulation also attracts support from laborers who contribute human capital and banks that finance manufacturing entrepreneurs. Ultimately, labor market regulation is conducive to manufacturing entrepreneurship. According to signal interpretation by receivers, labor market regulation can encourage incumbent firms to adopt AI in place of labor, thereby stimulating AI innovation and promoting manufacturing entrepreneurship. Labor regulation may likewise incentivize manufacturing firms to embrace digitalization, thereby improving manufacturing digitization and fostering entrepreneurship. Furthermore, the effect of labor regulation in promoting manufacturing entrepreneurship becomes weaker as enterprise density and the level of nonmanufacturing entrepreneurship increase. Using city-level panel data from 2008 to 2020 in China, we conduct empirical analysis with a two-way fixed effects model and find evidence supporting our theoretical framework.

Signaling theory was introduced by Spence (1973) to explain how information asymmetry between employers and job applicants, particularly concerning applicant quality, can be addressed in the labor market. It primarily analyzes how job applicants convey meaningful signals to employers to indicate their abilities. The original framework assumes a one-to-one relationship between the sender and receiver of the signal. With further academic development, signaling theory has been widely adopted across disciplines including economics, management and marketing (Bafera and Kleinert, 2023; Connelly et al., 2011). Recently, researchers in entrepreneurship have also embraced signaling theory, with growing interest in third-party signals (e.g. Anglin et al., 2020; Johnston et al., 2023). Our research contributes to the literature with several distinct differences: First, there is an asymmetry in status: signal senders (local governments) act as regulators, while some signal receivers (such as incumbent firms) are the regulated, creating a hierarchical regulatory relationship. Second, the receivers of the signal are diverse and numerous. Unlike the single-sender, single-receiver structure in Spence’s (1973) original model, our study considers multiple heterogeneous receivers. In this study, local governments act as signal senders, and the signal receivers encompass four main groups – manufacturing entrepreneurs (both actual and potential), laborers providing human capital, banks offering financing and incumbent manufacturing and nonmanufacturing firms – with a large number of actors in each group. In summary, this research broadens and deepens the application of signaling theory.

Existing international literature (e.g. Desai et al., 2021; Van Stel et al., 2007) generally suggests that stringent labor market regulations can hinder entrepreneurship. This is because such regulations tend to increase labor costs and complicate management for enterprises. Innovation often goes hand in hand with business growth (Braunerhjelm and Thulin, 2023), which exposes enterprises to more rigorous regulatory requirements, potentially dampening entrepreneurial motivation. From a theoretical standpoint, high regulatory costs function similarly to a tax on profits, reducing enterprises’ incentives to invest in innovation and leading to a decline in overall innovation levels. In some countries, while strict employment protection laws effectively safeguard workers’ rights and interests, they may also discourage employers from hiring, thereby limiting entrepreneurial opportunities. In addition, overly stringent bankruptcy liquidation laws and regulations can heighten entrepreneurs’ concerns about the consequences of failure, further inhibiting entrepreneurial activity.

Our research reveals that, unlike in other contexts, strict labor market regulations in China play a supportive role in fostering manufacturing entrepreneurship. In China, local governments actively implement strict labor market regulations, contrasting with the more reactive enforcement typically seen in labor litigation and arbitration. This proactive approach highlights local authorities’ commitment to protecting workers’ rights and interests. As Chinese workers become increasingly aware of their legal protections, this stance helps attract the skilled labor force essential for manufacturing entrepreneurs. At the same time, our analysis also indicates that the implementation of labor market regulations by local governments in China encourages existing enterprises to comply with the Labor Law and the Labor Contract Law, thereby motivating banks to allocate more financial resources. Together, these factors create a favorable environment for manufacturing entrepreneurship. In addition, our study takes into account the fact that “the Chinese government’s strong promotion of the digital economy since 2008, which has driven significant progress in AI innovation and manufacturing digitalization.” By examining how these developments are perceived, we explore the influence of labor market regulation on AI innovation and digitalization, and how these elements subsequently impact manufacturing entrepreneurship. Notably, our work is among the first to investigate the indirect effects of labor market regulation on manufacturing entrepreneurship through these intermediary factors.

Research indicates that, under current conditions, strengthening labor regulations can effectively encourage entrepreneurship within the manufacturing sector. The government may consider appropriately enhancing the rigor of labor regulations. It is important to note, however, that the positive impact of such regulations on manufacturing entrepreneurship differs across regions, with a more pronounced effect observed in regions characterized by low business density and limited nonmanufacturing entrepreneurial activity.

In regions with a less developed industrial base and scattered entrepreneurial activities (such as certain cities in central and western regions), it is advisable to moderately enhance the frequency of labor market supervision and improve the transparency of law enforcement. By reinforcing labor inspections and standardizing employment contracts, a strong message of “worker protection” can be conveyed. This strategy not only helps to identify and filter out less compliant entrepreneurs but also attracts high-quality labor, fostering a more favorable market environment for manufacturing entrepreneurship.

In regions with concentrated manufacturing industries and a high density of enterprises (such as industrial parks in the eastern region), regulatory efforts should prioritize “precision.” Enforcement should be intensified in locations with a high frequency of labor disputes, while ensuring that regulations do not become overly burdensome, so as to avoid increasing compliance costs for high-quality enterprises.

Our study focuses on how local governments’ labor market regulation affects manufacturing entrepreneurship. Administrative regulation by local governments targeting violations of labor laws and labor contracts by manufacturing firms may also impact manufacturing entrepreneurship. Due to the lack of statistical data, our study is unable to isolate the effect of such administrative regulation on manufacturing entrepreneurship, which is a limitation of this paper and a direction for future research.