Optimizing Sustainable Entrepreneurial Ecosystems: The Role of Government-Certified Incubators in Early-Stage Financing

Abstract

In the sustainable evolution of the entrepreneurial ecosystem, the efficiency of early-stage capital allocation directly affects the intergenerational transmission capacity of innovation resources. The financing barriers caused by information asymmetry urgently require institutional solutions. This study, based on tracking data from 19,463 startups in China’s information technology sector (2016–2019), analyzes how government-certified incubators (GCIs) optimize the sustainability of the entrepreneurial ecosystem through signaling mechanisms. The empirical results show that collaboration with a GCI can significantly increase the likelihood of IT startups securing venture capital by approximately 25%. This effect is not only due to the strict screening and resource support provided by GCIs, but also due to their role in amplifying internal signals from startups, such as the experience of founders and intellectual property. Notably, in the IT sector, the impact of GCIs is more significant for startups traditionally disadvantaged, particularly those led by female founders. Our research demonstrates that GCIs drive the sustainable development of the entrepreneurial ecosystem through three signaling mechanisms: (1) institutional certification screening, which optimizes the intergenerational allocation efficiency of ecosystem resources; (2) the signaling validation–amplification mechanism, which enhances the value of intellectual property and founder experience, alleviating investors’ challenges in quantifying startup potential; (3) inclusive signal rebalancing, where GCI certification significantly improves the funding success rate of female founders, breaking traditional market biases in screening disadvantaged groups and supporting the inclusive and sustainable development of the entrepreneurial ecosystem. These findings provide a new pathway for emerging economies to optimize the resilience of their entrepreneurial ecosystems through policy tools: for governments, GCIs achieve sustainable development goals at low institutional cost; for investors, the signal integration mechanism reduces investment information friction; and for entrepreneurs, certification endorsements accelerate market validation of sustainable business models.

1. Introduction

Venture capital not only alleviates financial constraints but also provides critical managerial expertise and network resources, making it a pivotal objective for early-stage startups [1]. However, securing venture capital remains exceptionally challenging, with only 0.05% of startups succeeding globally [2]. This scarcity is amplified in entrepreneurial ecosystems characterized by fragmented financial support networks (FSNs), where information asymmetry, power imbalances, and institutional voids hinder resource flows between startups and investors [3,4,5]. In emerging economies like China, these challenges are further exacerbated by underdeveloped legal frameworks, limited capital exit channels, and resource allocation inefficiencies [6].

To bridge these gaps, startups increasingly rely on signaling mechanisms to convey value to investors. While prior studies emphasize internal signals (e.g., founder experience, intellectual property) and third-party endorsements (e.g., venture capitalist affiliations) [7,8,9], their applicability to early-stage ventures in institutionally fragile contexts remains underexplored. Crucially, early-stage startups face a “noisy” signaling environment where weak internal signals risk being overlooked [10]. Recent work suggests that third-party certifications—particularly those involving governmental legitimacy—may amplify startup signals by reducing investor uncertainty [11,12].

In institutionally fragile contexts like China, where governmental legitimacy profoundly shapes market trust, state-endorsed intermediaries emerge as critical signal amplifiers. Since 1987, China has strategically employed incubators to bridge entrepreneurial ecosystem gaps, with over 13,000 incubators operating by 2022 [13]. To elevate quality standards, the Torch Center of the Ministry of Science and Technology annually certifies top performers as “Government-certified incubators” (GCIs) based on stringent criteria including infrastructure, service capacity, and tenant outcomes. Unlike generic incubators, GCIs are hypothesized to address FSN weaknesses through dual institutionalized mechanisms: (1) ex-ante quality filtering via rigorous startup screening and (2) ex-post resource empowerment through mentorship and network brokerage [14,15]. Yet, whether GCI affiliation effectively signals startup quality—and how this signal interacts with internal venture attributes—remains empirically unresolved in early-stage financing.

Using data from China’s IT sector, this study examines three underexplored dimensions: (1) Signaling validity: whether GCI certification serves as a credible third-party signal to mitigate information asymmetries in early-stage financing. (2) Signal interaction: how GCI affiliation interacts with internal signals (founder experience, intellectual property) to enhance venture capital acquisition. (3) Gender paradox: whether GCI certification alleviates or exacerbates gender-based financing disparities predicted by gender role congruity theory. Our findings reveal that GCI affiliation increases startups’ venture capital likelihood by approximately 25%, primarily through signal verification and amplification effects. Contrary to the Matthew effect, GCIs disproportionately benefit female-founded ventures by counteracting gender stereotypes—a redistribution mechanism absent in prior studies. Our research setting and findings contribute to the current literature on signal transmission and interaction in the early financing stages of startups in emerging economies. First, we expand the sources of third-party signals available for early-stage startup financing. Our research demonstrates that government certification of incubators provides behavioral added value, sending effective signals to the outside world and increasing the likelihood of startups securing venture capital. The combination of government and incubators offers a dual guarantee for startups accessing venture capital. Second, our study explores how internal signals (e.g., founders’ prior experience) and external signals (e.g., GCI certification) of startups work together in the context of early-stage financing, enhancing startups’ access to venture capital by reducing investor uncertainty. Finally, our research reveals new insights into the interaction between signaling theory and gender role congruity theory. While gender stereotypes make it harder for female entrepreneurs to secure funding [15,16], the rigorous vetting mechanisms of GCIs make female founders admitted to these incubators more likely to gain investor attention and recognition, thereby increasing their chances of obtaining venture capital and reducing the “Matthew effect” to some extent [3,17].

From another perspective, our research enriches the understanding of how incubators influence startups’ access to financing. Although the mechanisms by which incubators affect startup performance remain underexplored [18], existing studies suggest that incubators facilitate external funding through direct support and network agency [13]. Our research, based on a large sample from China’s IT sector, not only confirms these mechanisms but also shows that reputable incubator locations serve as a complementary legitimacy strategy, enhancing startups’ access to venture capital through signaling.

Our findings also have practical implications. For governments, certifying incubators alleviates information asymmetry in corporate finance at a “lower cost” compared to subsidies, tax incentives, or direct funding [19]. For startups, gaining admission to a GCI is a strategic choice that empowers them with stronger resources and increases their chances of securing venture capital by amplifying weak signals or reversing disadvantageous positions. Similarly, for investors, partnering with reliable GCIs reduces information costs and broadens investment opportunities [20].

This paper is structured as follows. Section 2 presents the theoretical analysis and research hypotheses. Section 3 introduces the data background, explains the sample selection, defines key variables, and describes the empirical model. Section 4 reports the empirical results, including descriptive statistics, hypothesis tests, and robustness checks. Section 5 summarizes the main findings, discusses theoretical and practical contributions, and highlights the limitations and future research directions.

2. Theoretical Analysis and Research Hypotheses

2.1. Weak Networks in Entrepreneurial Ecosystems

The entrepreneurial ecosystem (EE) in which startups are located consists of a knowledge subsystem and a business subsystem. Firms in the knowledge subsystem hold innovations that need to be commercialized into products and services within the business subsystem to transfer core technology and generate profits [4]. A key link between these two systems is the financial support network (FSN) [21,22]. However, in Europe, where the financial system and institutional environment are relatively well-developed, there is a significant gap between the players in the FSN, especially startups and venture capitalists [23]. Venture capital for startups is scarce, with reports indicating that as little as 0.05% of startups receive such funding (Where Startup Funding Really Comes From (Infographic)|Entrepreneur, https://www.entrepreneur.com/money-finance/where-startup-funding-really-comes-from-infographic/230011 (accessed on 30 January 2025)). This gap is even wilder in emerging economies, where market mechanisms for resource allocation are dysfunctional, the environment for property rights protection is relatively weak, and capital exit channels such as IPOs and M&A are blocked [11].

Scholars have explained this phenomenon in the entrepreneurial ecosystem from different perspectives. One is the theory of power advantage. Most of the resources in the entrepreneurial ecosystem are held by powerful firms or political parties, and startups can access external resources by reducing their actual dependence on these potential resource providers and maximizing their ability to establish common goals with them [24]. However, this requires significant effort and cost, which is challenging for startups at a power disadvantage. Second is market information asymmetry. Investors screen targets using size, growth potential, etc. as key evaluation criteria [25]. However, the technological advantages, market prospects, expected returns, and potential risks of startups are often held by their insiders, who believe that releasing such relevant information will enable competitors to obtain intelligence, which will deprive them of a competitive advantage, and there is a large cost to disclose the information. The information of many startups is opaque, which makes it difficult for investors to obtain effective signals to make investment decisions [26,27]. Third is return uncertainty. This is particularly prominent in the high-tech industry, which provides innovative knowledge essential for economic development, but the assessment of new knowledge is more complex, there is a high degree of uncertainty about the success of its R&D projects, and high-tech firms lack sufficient collateral to hedge against this uncertainty, leading to the high risk of investment [8].

2.2. Signaling Mechanisms Mitigate Weak Networks in Entrepreneurial Ecosystems

The signaling mechanism, by reducing information asymmetry, serves as a critical tool in alleviating weak FSN within entrepreneurial ecosystems. Startups actively release strategic signals to convey implicit value to external stakeholders, bridging the trust gap caused by insufficient network connections. The core mechanisms are manifested in three aspects: (1) Third-Party Signals: startups signal technological reliability and market potential through endorsements, such as accreditation from reputable incubators, patent grants, or industry awards [14]. (2) Social Capital Signals: founding teams build a “bundle of capability signals” by showcasing credentials, such as prior experience at top-tier companies, degrees from prestigious universities, or associations with renowned technical advisors [9]. (3) Strategic Investment Signals: the involvement of cornerstone investors with significant industry status (e.g., strategic investments by BAT) triggers a “herding effect” [5].

Startups face a noisy signaling environment in the early stages of financing, where internal signals—such as firm characteristics and behaviors—are often overlooked by investors overwhelmed with information [28,29]. Therefore, early-stage startups seeking venture capital face the challenge of not only capturing investor attention but also effectively conveying their potential in such an environment. Research by Plummer et al. (2015) indicated that affiliation with third parties becomes a new signal that stands out in the signaling environment and can unlock the value of other weaker signals [14], enhancing the ability of startups to access external funding at early stages. Startups that need to switch identities to access venture capital during their early years can strategically use third-party signals to bridge the gap with potential investors and highlight their potential value [3].

2.3. Government-Certified Incubator as an Effective Signal in the Network

2.3.1. Why Do Incubators Need to Be Certified by the Government?

Incubators are considered one of the most important tools for supporting early-stage ventures and the core of entrepreneurial ecosystems [30]. They can partially alleviate the information asymmetry issues between investors and startups to help startups access funding resources more easily [29]. Over several decades, incubators have proliferated globally, with the United States already having about 2700 incubators and accelerators (Data resource: 2.7K+ Accelerators & Incubators in United States—Tracxn, https://tracxn.com/d/investor-lists/accelerators-incubators-in-united-states/__ipYdkLo_FgbzV1gMFal2rWv-EyhedJX0mEbP-IXdRgs (accessed on 30 January 2025)), and a report on European incubators shows that there are more than 1200 incubators in the United Kingdom, France, Germany, Italy, and Spain, achieving a multiplier growth (Data resource: First report on the impact of European incubators and accelerators (phys.org), https://phys.org/news/2020-10-impact-european-incubators.html (accessed on 30 January 2025)). The number of incubators in China has increased dramatically over the past 30 years, and by 2022, official figures show that there will be more than 13,000 incubators in the country [31].

A large number of incubators have emerged in the market, but they are mixed and not all of them are trustworthy; startups, investors, etc. often need to spend more and more effort to find a target incubator [32]. The increase in numbers has also exposed various issues. Firstly, due to disparities in the development levels of incubators and the limited capabilities of internal staff, assessment criteria and industry preferences lead to selection biases, resulting in a portion of potential enterprises being overlooked [33]. Secondly, the diversity of incubators complicates entrepreneurial support networks, making it challenging for investors to identify reliable incubators to focus on [34]. Given these issues, it is essential to recognize the limitations of incubators in fulfilling their function within the entrepreneurial ecosystem. Dvouletý et al. (2018) emphasize the necessity and importance of rigorous government evaluation of incubators [33]. In a market environment where the number of incubators is huge and difficult to identify, the implementation of hierarchical classification management by the government can guide the market to proceed in an orderly manner.

2.3.2. Validity of Government Certification

Research on signaling theory indicates two necessary conditions for a signal to be effective: (1) observability and variability, meaning the signal must be perceptible to the public, and the signal sender can adjust the signal; (2) the signal must cover costs negatively correlated with the quality of the actor’s behavior, implying that low-quality participants must exert more effort to obtain the signal and find it difficult to imitate the signal transmitted by high-quality participants [35,36]. These conditions ensure the reliability of the signal. Governments typically have broader social and economic objectives and establish professional evaluation departments to invite industry experts to rigorously oversee the evaluation criteria and process [37]. Simultaneously, the entire evaluation process is based on the relevant operational data provided by the incubator and its developmental history, making it difficult for the incubator to engage in fraudulent activities [38]. Therefore, government evaluation results are generally more widely recognized by the public. Furthermore, uncertified incubators need to exert significant effort to enhance their operational and profitability capabilities to gain government recognition [39,40]. In conclusion, we can argue that the government’s certification for incubators meets the two necessary conditions for signaling. The relationship with GCIs is an effective third-party signal for early-stage enterprises. Affiliation with GCIs sends signals to investors in two ways: by acting as a filter for the quality of the startups and by conveying that GCIs will provide effective substantial assistance to early-stage enterprises.

Filter: Upon accepting applications from entrepreneurs, GCIs evaluate the feasibility, innovativeness, and team capabilities of the projects, commonly having more stringent screening criteria. Gaining entry into GCIs serves as effective recognition of the quality of the enterprise [32]. As GCIs need to maintain their reputation consistently, they prudently avoid collaborating with startups that are less likely to succeed. To investors, GCIs act as a high-quality filter, allowing them to focus their primary efforts on enterprises within GCIs, as this implies a higher likelihood of discovering hidden gems within, thus reducing search costs and investment risks [14].

Substantial assistance: When a business is affiliated with GCIs, it sends a signal to investors that the startup will derive substantial benefits from being associated with prestigious GCIs, thereby facilitating its success [14]. GCIs empower enterprises through capacity building and resource linkages [40]. Firstly, once a startup is selected to join a GCI, the incubator organizes various forms of entrepreneurial guidance and establishes mentoring relationships with entrepreneurs to accelerate the learning curve of the enterprise [41]. Secondly, the network constructed by the incubator can help link early-stage startups with other entrepreneurial teams, financial institutions, research institutions, government entities, and other external resources, which is particularly crucial for early-stage startups [13]. Based on the foregoing analysis, we propose the following foundational hypothesis:

H1. 

Affiliation with a GCI serves as a valid third-party signal that enhances the likelihood of startups securing venture capital.

2.4. Signal Interaction Effects

Several signals can originate from within the business itself, such as excellent founders, intellectual property, sales performance, etc. However, the challenge for early-stage startups facing investors is that these signals are often hard to discern due to their uncertainty, ambiguity, and the fact that they can be mixed with other signals in noisy environments [14]. However, by exposing themselves to third-party signals, these signals become easily identifiable. We explore whether affiliation with GCIs can resolve signal ambiguity and release the value of internal signals from startups by testing signal interactions.

Previous experience of the founder. A wealth of experience can help organizations identify the best opportunities in the market, reduce errors, and save time and resources [42]. Therefore, for startups, the founder’s previous experience may serve as a valuable quality signal [43,44]. However, the interpretation of this signal is ambiguous, making it difficult for investors to ascertain whether such experience will translate into sound entrepreneurial decisions, as the effectiveness of experience is contingent upon its relevance, practicality, and the individual capabilities of the manager. GCIs can validate the credibility of entrepreneurs’ experience through their operational mechanisms and can also unleash the potential of entrepreneurs’ historical experience through empowerment. Thus, the location of GCIs enables potential investors to understand and believe that the founder’s previous experience is effective and elevates this weak signal above the noise in the early-stage financing environment [14]. Therefore, we assume the following:

H2. 

Third-party signals affiliated with GCI can amplify the signal value of the founder’s previous experience, and the interaction of these two signals increases the likelihood of early-stage ventures obtaining venture capital.

Intellectual property of startups. Patents symbolize the innovativeness of a firm and serve as a measure of its commercial potential. In the biotechnology sectors in the UK and Germany, high-quality patent applications expedited the process of firms securing venture capital [45]. The effectiveness of patents as a signal in enhancing a firm’s financing capabilities also applies in the software sector [46]. However, in the “noisy” signal environment of early-stage financing, startups relying solely on “patents” struggle to stand out and capture the attention of discerning investors. Additionally, patent signals entail significant uncertainty, as there is still a long journey from patenting to the commercial profitability of products in the market [47].

Affiliation with GCI first draws attention to the startups, subsequently bringing the patent signal into focus. Furthermore, due to GCI’s long-term embeddedness in relevant industries and markets, and its deep familiarity with client companies, investors have greater confidence in GCI’s ability to reliably handle patent signals [14]. Moreover, the mentoring capabilities and network resources of GCI can significantly ensure that patents translate into tangible market returns for firms [41]. Under the mechanisms of signal amplification and signal combination, the endorsement of multiple patents and GCI generates a linear effect between these two signals, resulting in a synergistic effect where 1 + 1 > 2 [8]. Therefore, we assume the following:

H3. 

Third-party signals affiliated with GCIs can amplify the signaling value of startups’ patents, i.e., the signaling benefits of being affiliated with GCIs will be more effective for startups with more patents, increasing the likelihood of their access to venture capital.

Gender of the founder. Gender role congruity theory suggests that the prevalent societal perception is that men are responsible for “making money”, while women are more often associated with caregiving roles. This gender stereotype results in natural bias and overtly discriminatory behaviors by investors towards female entrepreneurs, leading to greater difficulties for female entrepreneurs in accessing funding [15]. Furthermore, female entrepreneurs have simpler social networks and lack avenues for conveying legitimacy signals. Consequently, female entrepreneurs face greater information asymmetry and obstacles in signal transmission in their interactions with investors [48,49]. In fact, there is substantial evidence indicating that female entrepreneurs are at a disadvantage within the entrepreneurial ecosystem, particularly in the technology sector, where gender disparities are more pronounced [50,51]. This gender imbalance is no exception in the incubation field, as evidenced by (1) significantly more male than female tenants in incubators and (2) the marginalization of female entrepreneurs in the incubation process [52,53]. Female entrepreneurs struggle to benefit from a male-dominated incubation network. Taken together, the signal benefits of affiliation with GICs may still be more beneficial to male entrepreneurs in accessing venture capital due to gender bias, initially low legitimacy, and the influence of male-dominated incubation networks. Therefore, we assume the following:

H4a. 

The fact that the main founder of a startup is a female has a negative impact on its access to venture capital. And the marginal value of third-party signals affiliated with GCI is greater for startups with male founders.

However, the concept of “signal congruence” suggests that when a signal aligns with expectations, it is likely to be easily overlooked or rejected. Conversely, when a signal exhibits a significant deviation from expectations and conveys positive information, it can unexpectedly generate substantial value [37]. When a female entrepreneur qualifies for GCI affiliation, it is a rare event [54], and “rarity enhances value”—investors are more likely to take notice of this phenomenon. Furthermore, due to the existence of biases, investors may believe that female entrepreneurs selected by GCI must possess unique advantages, as GCI signifies more rigorous standards and scrutiny [52]. Therefore, we infer that the marginal value of third-party signals affiliated with GCI is greater for startups with female founders. Formally:

H4b. 

The fact that the main founder of a startup is a female has a negative impact on its access to venture capital. But the marginal value of third-party signals affiliated with GCI is greater for startups with female founders.

3. Data and Methodology

3.1. Data Background

3.1.1. Empirical Setting: China Torch Programme—Incubator Survey

Our research data are derived from the China Incubator Survey Database, which is currently the most comprehensive survey on both incubators and startups in China. The data are collected nationwide by a specialized statistical survey team organized and trained by the Torch Center of the Ministry of Science and Technology of China. The post-collection data undergo professional verification and validation, and are reported to the National Bureau of Statistics, ensuring the accuracy of the data through a rigorous process. The database consists of two sub-databases: one for incubators, which includes information on their basic profile, investment composition, revenue sources, space allocation, management personnel, entrepreneurial guidance, and operational management; and the other for incubated startups, which covers their basic profile, economic overview, employment situation, and intellectual property status.

The sample spans from 2016 to 2019, and the regional and technological distribution of the startup samples is presented in

Table 1. Regional and fields distribution of startups in China.

Part A: Region and Year Distribution of Startups and Incubators
Province	2016		2017		2018		2019
	Startups	Incubators	Startups	Incubators	Startups	Incubators	Startups	Incubators
Jiangsu	28,312	548	38,579	610	39,638	695	45,241	832
Guangdong	19,271	576	28,109	754	37,284	962	39,993	1013
Shandong	13,387	216	17,486	303	20,697	378	20,788	358
Zhejiang	10,846	160	15,464	235	19,787	321	21,390	363
Henan	8066	162	10,479	148	10,928	169	10,929	167
Hubei	5917	67	11,463	176	12,619	192	13,939	216
Shanghai	7485	156	8911	176	9781	180	9611	175
Beijing	6341	101	8099	105	11,391	152	11,377	130
Sichuan	6580	108	8423	143	9192	147	10,077	168
Tianjin	6474	108	5531	71	5532	72	5582	81
Others	47,158	1053	65,706	1342	73,079	1581	79,574	1703
Sum	159,837	3255	21,8250	4063	249,928	4849	268,501	5206
Part B: Technical Area and Year Distribution of Startups
Field		2016		2017		2018		2019
Information Technology		82,373		107,826		122,916		130,256
Advanced Manufacture		25,825		34,983		36,937		40,375
Biomedicine and Devices		13,719		17,452		20,260		22,471
New Materials		9634		11,568		13,161		14,015
New Energy		7717		9679		10,922		11,613
Cultural Creativity		7222		17,559		24,313		27,234
Modern Agriculture		5288		8072		8805		9141
Environmental Protection		5090		6713		7682		8451
Modern Transport		1428		2322		2791		2907
Aerospace		828		1061		1192		1159
Earth, Space and Oceans		624		727		732		717
Nuclear Applied Technology		89		288		217		162
Sum		159,837		218,250		249,928		268,501

. Part A ranks the top ten provinces in terms of the number of startups and incubators, with Jiangsu, Guangdong, and Zhejiang being the most developed regions in China’s private economy. Beijing, Shanghai, and Tianjin possess relatively well-established scientific and technological resources and financial markets, while provinces such as Shandong, Henan, Hubei, and Sichuan have made significant breakthroughs in entrepreneurial ecosystem development in recent years [55]. Both the number of startups and incubators show an increasing trend. Part B presents the distribution of startups across technological fields, with the information technology sector having the highest number of startups, accounting for approximately half of the entire sample. This is followed by advanced manufacturing (approximately 16%) and biomedicine and devices (approximately 9%).

3.1.2. Sample Selection

We selected the sample of startups in the information technology sector from this database for research, as they are highly motivated to secure external capital. Since the rise of information technology, its development has been closely intertwined with the venture capital industry. On one hand, these startups require substantial investment in technological innovation and software development, and obtaining external funding can help them bear these costs [12]. On the other hand, the IT sector experiences rapid market development and continuous product iteration and upgrades. To maintain sufficient competitiveness, startups are often eager to obtain external capital to enhance their flexibility in responding to market changes [56]. The proportion of startups in the IT sector is the highest in the database we utilized, providing ample sample support for subsequent research.

First, we matched the sub-databases of incubators and startups based on the names of the incubators to eliminate unmatched samples and startups that have already graduated from the incubator. Next, we excluded samples with missing values for key variables, such as registered capital, venture capital, operating income, research and development expenses, and the number of research personnel. Additionally, as we are focusing on the early stages of companies, we excluded all firms established more than 10 years ago as of 2019 [3]. Finally, we organized the data into balanced panel data spanning from 2016 to 2019, resulting in a total of 19,463 startups and 77,852 samples.

3.2. Variable Design and Measurement

Dependent variable. The dependent variable we selected is whether startups obtained venture capital in the current year from angel investors or venture capitalists, represented as a dummy variable where obtaining venture capital is coded as 1, and otherwise as 0.

Independent variable. The independent variable considers the affiliation of startups with GCIs, with a value of 1 if affiliated with a GCI, and 0 otherwise.

Control variables. Taking into account other factors that may influence a startup’s ability to obtain venture capital, we controlled for other variables such as startup size, startup age, startup revenue, level of research and development investment, and human capital level, as described in

Table 2. Description of variables.

Variable Type	Variable Name	Symbol	Description
Dependent variable	Whether receive venture capital	$VC$	Dummy variable with a value of 1 for venture capital received in the current year and 0 otherwise
Independent variable	Whether affiliated with GCI	$incu\_certifi$	Dummy variable, affiliation to GCI in the current year, takes the value of 1, otherwise 0
Control variables	Startup size	$com\_size$	The registered capital of the startup, taken in logarithms (Unit: 1000¥)
	Startup Age	$com\_age$	Take the logarithm of the difference between the statistical year and the year of establishment of the startup
	Startup Revenue	$com\_rev$	Take the logarithm of the annual operating income for that year (Unit: 1000¥)
	R&D Level	$com\_RD$	Take the logarithm of the startup’s R&D expenditure (Unit: 1000¥)
	Human Capital	$com\_edu$	The proportion of employees with a college degree or above in the startup
Moderating variables	Founder’s previous experience	$experience$	The dummy variable takes a value of 1 for serial entrepreneurs and 0 for first-time entrepreneurs
	The intellectual property status of the startup	$patent$	Take the logarithm of the number of invention patents obtained by the startup in the current year
	The gender of the main founders	$gender$	The dummy variable takes a value of 1 when the founder’s gender is female, and 0 when it is male

Notes: for a value of 0, add 1 for logarithmic processing.

. Additionally, we included year, industry, and regional dummy variables to control for fixed effects. Industry dummy variables were defined according to the “National Standard of the People’s Republic of China for National Economic Industry Classification”, and regional dummy variables were set based on the province where the startup is located. To mitigate the impact of outliers, the control variables were subjected to Winsorization at the 1st and 99th percentiles.

Moderating variables. As discussed in the theoretical hypothesis, to verify whether certain internal signals from startups are activated due to affiliation with GCIs or whether affiliation with GCI provides greater support for specific types of startups, we included founder previous experience, the number of intellectual property rights held by the startup, and the gender of the startup’s founder as moderating variables, as defined in Table 2.

Alternative explanatory variables. In the robustness tests in Section 4.3, we introduced additional control variables to eliminate other potential alternative explanations. As analyzed in Section 2.3.2, GCIs typically possess larger facilities, mentor resources, and incubation networks, and can facilitate startups’ access to venture capital through capacity building and resource linkage [13]. Considering that establishing coaching relationships is the primary step and main channel for incubators to provide capacity building to startups, we constructed a dummy variable “$if\_coached$” based on the survey question “whether a coaching relationship was established with a startup mentor”, with a value of 1 if a coaching relationship was established, and 0 otherwise. Furthermore, based on the survey question “whether the incubator is located in a high-tech zone”, we set a dummy variable “$incu\_if\_htz$”, with a value of 1 if located in a high-tech zone, and 0 otherwise. High-tech zones are special areas stimulated by the Chinese government for innovation and entrepreneurship, with abundant entrepreneurial resources, and the presence of an incubator in a high-tech zone implies access to rich network resources as a way of controlling for the alternative explanation of resource linkage [11]. From the perspective of startups, we also controlled for other strong signals related to government quality certification for startups that investors may receive, using a dummy variable “$if\_hightech$” based on the survey question “whether the startup is a high-tech enterprise”, with a value of 1 if it is a high-tech enterprise, and 0 otherwise.

3.3. Empirical Model

Given that the dependent variable is a dummy variable, we employ the Logit model and progressively control for time fixed effects, regional fixed effects, and industry fixed effects in the regression. The model testing Hypothesis 1 is presented as follows:

$$ln\left({\displaystyle \frac{P\left({VC}_{i,t}=1\right)}{1-P\left({VC}_{i,t}=1\right)}}\right)={\beta }_0+{\beta }_1{inc{u}_{certifi}}_{i,t}+{\beta }_c{control}_{i,t}+{\delta }_{year}+{\gamma }_{industry}+{\mu }_{province}+{\epsilon }_{i,t}$$

(1)

where $P({VC}_{i,t}=1)$ represents the probability of enterprise $i$obtaining venture capital at time $t$, and ${\displaystyle \frac{P({VC}_{i,t}=1)}{1-P({VC}_{i,t}=1)}}$ is referred to as the odds ratio. ${control}_{i,t}$ denotes the control variables, ${\delta }_{year}$ represents the year dummy variables, ${\gamma }_{industry}$ signifies the industry dummy variables, ${\mu }_{province}$ indicates the regional dummy variables, and ${\epsilon }_{i,t}$denotes the random disturbance term. Once an investor invests in a specific company within a certain incubator, they are likely to invest in other startups within the same incubator. This selection dependence may lead to bias in the regression results. Therefore, we employ robust standard errors clustered at the incubator level in the regression to address the departure from the independence assumption. Additionally, the explanatory variables and control variables are lagged by one period to avoid endogeneity issues arising from reverse causality [9]. If Hypothesis 1 holds true, the coefficient ${\beta }_1$ of ${incu\_certifi}_{i,t}$ should be positive.

The model testing Hypothesis 2 is presented as follows:

$$\begin{array}{c}ln\left({\displaystyle \frac{P\left({VC}_{i,t}=1\right)}{1-P\left({VC}_{i,t}=1\right)}}\right)\hfill \\ & ={\beta }_0+{\beta }_1{inc{u}_{certifi}}_{i,t}+{\beta }_2{experience}_{i,t}+{\beta }_3{inc{u}_{certifi}}_{i,t}\times {experience}_{i,t}+{\beta }_c{control}_{i,t}\hfill \\ & +{\delta }_{year}+{\gamma }_{industry}+{\mu }_{province}+{\epsilon }_{i,t} \hfill \end{array}$$

(2)

If Hypothesis 2 holds true, the coefficient ${\beta }_2$ of ${experience}_{i,t}$ should be positive, and the coefficient ${\beta }_3$ of the interaction term ${incu\_certifi}_{i,t}\times {experience}_{i,t}$ should also be positive.

The model testing Hypothesis 3 is presented as follows:

$$\begin{array}{c}ln\left({\displaystyle \frac{P\left({VC}_{i,t}=1\right)}{1-P\left({VC}_{i,t}=1\right)}}\right)\hfill \\ & ={\beta }_0+{\beta }_1{incu\_certifi}_{i,t}+{\beta }_2{patent}_{i,t}+{\beta }_3{incu\_certifi}_{i,t}\times {patent}_{i,t}+{\beta }_c{control}_{i,t}\hfill \\ & +{\delta }_{year}+{\gamma }_{industry}+{\mu }_{province}+{\epsilon }_{i,t}\hfill \end{array}$$

(3)

If Hypothesis 3 holds true, the coefficient ${\beta }_2$ of ${patent}_{i,t}$ should be positive, and the coefficient ${\beta }_3$ of the interaction term ${incu\_certifi}_{i,t}\times {patent}_{i,t}$ should also be positive.

The model testing Hypothesis 4 is presented as follows:

$$\begin{array}{c}ln\left({\displaystyle \frac{P\left({VC}_{i,t}=1\right)}{1-P\left({VC}_{i,t}=1\right)}}\right)\\ & ={\beta }_0+{\beta }_1{incu\_certifi}_{i,t}+{\beta }_2{gender}_{i,t}+{\beta }_3{incu\_certifi}_{i,t}\times {gender}_{i,t}+{\beta }_c{control}_{i,t}\hfill \\ & +{\delta }_{year}+{\gamma }_{industry}+{\mu }_{province}+{\epsilon }_{i,t}\hfill \end{array}$$

(4)

If Hypothesis 4a holds true, the coefficient ${\beta }_2$ of ${gender}_{i,t}$ should be negative, and the coefficient ${\beta }_3$ of the interaction term ${incu\_certifi}_{i,t}\times {gender}_{i,t}$should also be negative. If Hypothesis 4b holds true, the coefficient ${\beta }_2$ of ${gender}_{i,t}$ should be negative, and the coefficient ${\beta }_3$ of the interaction term ${incu\_certifi}_{i,t}\times {gender}_{i,t}$should be positive.

In various regressions, due to the lack of variability in the dependent variable within firms, the maximum likelihood estimation utilized in the Logit regression fails to converge, resulting in fluctuations in the number of observations across different regressions [9]. Nonetheless, our ample sample size warrants the accommodation of such minor fluctuations.

4. Results

4.1. Descriptive Statistics

Table 3. Descriptive statistics of variables.

Variable	N	Mean	Sd	Min	Max
$VC$	77,852	0.074	0.262	0.000	1.000
$incu\_certifi$	77,852	0.547	0.435	0.000	1.000
$com\_size$	77,852	6.746	2.429	0.000	17.241
$com\_age$	77,852	1.226	0.594	0.000	2.303
$com\_rev$	77,852	6.106	3.051	0.000	15.563
$com\_RD$	77,852	2.804	2.958	0.000	15.277
$com\_edu$	77,852	0.862	0.195	0.000	1.800
$experience$	77,852	0.312	0.463	0.000	1.000
$patent$	77,852	0.062	0.280	0.000	5.935
$gender$	77,852	0.148	0.355	0.000	1.000
$if\_coach$ed	77,852	0.617	0.486	0.000	1.000
$incu\_if\_htz$	77,852	0.424	0.494	0.000	1.000
$if\_hightech$	77,852	0.094	0.292	0.000	1.000

presents the descriptive statistics of the main variables. On average, only 7.4% of the sample’s startups have obtained venture capital, making it a rare occurrence for startups. Among all the samples, 54.7% are affiliated with GCIs, 31.2% of the entrepreneurs are serial entrepreneurs, 14.8% of the entrepreneurs are female, 61.7% of the startups have received mentorship, 42.4% of the startups are located in national high-tech zones, and 7.7% of the startups are recognized as high-tech enterprises.

Table 4. Correlation coefficient matrix.

	$\mathit{V}\mathit{C}$	$\mathit{i}\mathit{n}\mathit{c}\mathit{u}\_\mathit{c}\mathit{e}\mathit{r}\mathit{t}\mathit{i}\mathit{f}\mathit{i}$	$\mathit{c}\mathit{o}\mathit{m}\_\mathit{s}\mathit{i}\mathit{z}\mathit{e}$	$\mathit{c}\mathit{o}\mathit{m}\_\mathit{a}\mathit{g}\mathit{e}$	$\mathit{c}\mathit{o}\mathit{m}\_\mathit{r}\mathit{e}\mathit{v}$	$\mathit{c}\mathit{o}\mathit{m}\_\mathit{R}\mathit{D}$	$\mathit{c}\mathit{o}\mathit{m}\_\mathit{e}\mathit{d}\mathit{u}$
$VC$	1.000
$incu\_certifi$	0.030 ***	1.000
$com\_size$	0.062 ***	0.100 ***	1.000
$com\_age$	0.017 ***	0.111 ***	−0.016 ***	1.000
$com\_rev$	0.105 ***	0.155 ***	0.187 ***	0.327 ***	1.000
$com\_RD$	0.171 ***	0.207 ***	0.203 ***	0.163 ***	0.452 ***	1.000
$com\_edu$	0.037 ***	0.053 ***	0.079 ***	−0.047 ***	−0.010 ***	0.053 ***	1.000

Notes: *** p < 0.01.

presents the Pearson correlation coefficients for the key variables. It is evident that the independent variable $incu\_certifi$ is significantly and positively correlated with the variable “whether a startup receives venture capital investment” ($VC$), providing preliminary support for Hypothesis 1. This suggests that affiliation with a GCI can effectively enhance a startup’s likelihood of securing venture capital financing. Furthermore, the correlations among most variables are not strong, with the highest being 0.452, indicating that our regression model does not include highly interrelated variables that could lead to severe multicollinearity, thus ensuring efficient and unbiased estimates and preserving the accuracy of our model regression [9,57]

4.2. Hypothesis Tests

Table 5. The impact of signal effects from GCI on startups’ access to venture capital.

	(1)	(2)	(3)	(4)	(5)
	$\mathit{V}\mathit{C}$	$\mathit{V}\mathit{C}$	$\mathit{V}\mathit{C}$	$\mathit{V}\mathit{C}$	$\mathit{V}\mathit{C}$
$incu\_certifi$	0.314 ***	0.301 *	0.281 **	0.230 **	0.222 **
	(0.107)	(0.105)	(0.106)	(0.107)	(0.107)
$com\_size$		0.041 ***	0.040 ***	0.029 ***	0.029 ***
		(0.013)	(0.013)	(0.012)	(0.012)
$com\_age$		−0.302 ***	−0.254 ***	−0.261 ***	−0.270 ***
		(0.041)	(0.050)	(0.052)	(0.052)
$com\_rev$		0.046 ***	0.044 ***	0.048 ***	0.049 ***
		(0.015)	(0.015)	(0.014)	(0.014)
$com\_RD$		0.125 ***	0.126 ***	0.121 ***	0.120 ***
		(0.014)	(0.014)	(0.013)	(0.013)
$com\_edu$		0.592 ***	0.588 ***	0.602 ***	0.611 ***
		(0.187)	(0.186)	(0.183)	(0.177)
$constant$	−2.773 ***	−3.795 ***	−3.812 ***	−4.338 ***	−4.516 ***
	(0.076)	(0.207)	(0.205)	(0.250)	(0.423)
Year Fixed Effects	No	No	Yes	Yes	Yes
Region Fixed Effects	No	No	No	Yes	Yes
Industry Fixed Effects	No	No	No	No	Yes
Wald chi2	8.65	261.23	296.32	468.57	706.86
Prob > chi2	0.003	0.000	0.000	0.000	0.000
Log likelihood	−15,112.21	−14,681.45	−14,662.81	−14,352.25	−14,247.41
Observations	57,624	57,624	57,624	57,549	64,322

Notes: cluster robust standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

presents the regression outcomes using a logit model with robust standard errors clustered at the incubator level, primarily testing whether affiliation with a GCI is conducive to startups’ acquisition of venture capital. Variables were introduced stepwise from column (1) to column (5), with controls for time, regional, and industry fixed effects. Across all regressions, the coefficients are significantly positive, indicating that GCI affiliation conveys a positive signal that enhances the likelihood of internal startups securing venture capital, supporting Hypothesis 1. The analysis is ultimately based on the results after introducing control variables and controlling for three types of fixed effects. The model’s goodness-of-fit is evaluated using the “percentage of correct predictions” in the logit model, which indicates a correct classification rate of 92.59%, signifying an excellent fit. Observing the coefficient for $incu\_certifi$ in column (5), which is 0.222 and significant at the 5% level, indicates that affiliation with a GCI increases the odds ratio ${\displaystyle \frac{P({VC}_{i,t}=1)}{1-P({VC}_{i,t}=1)}}$ by 1.25 ($e^{0.222}$), translating to a 25% increase in the probability of a startup securing venture capital. The majority of control variables align with expectations: startups with greater registered capital, larger revenue scales, and more substantial R&D spending and human capital are more likely to attract venture capital investments. Intriguingly, our results reveal a significant negative relationship between a firm’s operational history ($com\_age$) and its success in obtaining venture capital, diverging from previous findings [58]. As our sample comprises startups in the information technology sector, where businesses typically have shorter life cycles due to rapid technological evolution and the swift obsolescence of older technologies, firms must continuously innovate to distinguish themselves as young and potential rich entities in the marketplace to attract venture capital investors [57]. Venture capitalists tend to prefer younger enterprises with higher creativity, flexibility, and growth potential, whereas traditional financiers, such as banks, are more inclined to collaborate with longer-standing, more experienced companies [56].

As described in Section 3.3, we introduce interaction terms $incu\_certifi\times experience$,$incu\_certifi\times patent$, and $incu\_certifi\times gender$ on the basis of the benchmark regression to test hypotheses 2–4.

From the results in column (1) of

Table 6. Signal interactions with intra-enterprise signals.

	(1)	(2)	(3)
	$\mathit{V}\mathit{C}$	$\mathit{V}\mathit{C}$	$\mathit{V}\mathit{C}$
$incu\_certifi$	0.166 **	0.206 *	0.198 **
	(0.110)	(0.108)	(0.108)
$experience$	0.153 **
	(0.060)
$incu\_certifi\times experience$	0.337 ***
	(0.058)
$patent$		0.294 ***
		(0.112)
$incu\_certifi\times patent$		0.167 **
		(0.120)
$gender$			−0.388 ***
			(0.128)
$incu\_certifi\times gender$			0.248 **
			(0.141)
$com\_size$	0.022 *	0.024 ***	0.028 ***
	(0.012)	(0.012)	(0.012)
$com\_age$	−0.274 ***	−0.287 ***	−0.270 ***
	(0.053)	(0.053)	(0.052)
$com\_rev$	0.049 ***	0.047 ***	0.049 ***
	(0.013)	(0.013)	(0.014)
$com\_RD$	0.118 ***	0.114 ***	0.120 ***
	(0.013)	(0.013)	(0.013)
$com\_edu$	0.613 ***	0.625 ***	0.612 ***
	(0.175)	(0.178)	(0.178)
$constant$	−4.058 ***	−3.958 ***	−3.947 ***
	(0.417)	(0.423)	(0.424)
Year Fixed Effects	Yes	Yes	Yes
Region Fixed Effects	Yes	Yes	Yes
Industry Fixed Effects	Yes	Yes	Yes
Wald chi2		3035.19	2665.41
Prob > chi2		0.000	0.000
Log likelihood		−15,155.41	−15,340.30
Observations	57,165	57,165	57,165

Notes: cluster robust standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

, we can see that the coefficients for $experience$ and interaction term $incu\_certifi\times experience$ are positive at the 5% significance level. This indicates that the founder’s previous experience has a significant positive impact on the probability of a startup obtaining venture capital, and this historical experience interacts with the affiliation to a GCI, producing a positive signal superposition effect, increasing the possibility of a startup obtaining external capital, thus supporting Hypothesis 2.

Similarly, we find evidence supporting Hypothesis 3, which posits that a company’s invention patents and affiliation with a GCI also produce a positive signal interaction effect. As shown in column (2) of Table 6, the coefficient for $patent$ is positive at the 1% significance level, indicating that startups with more patents are more likely to obtain venture capital compared to their industry peers. The coefficient for the interaction term$incu\_certifi\times patent$ is positive at the 5% significance level, suggesting that the number of patents a company holds moderates the signal effect of government certification, with companies holding more patents benefiting more significantly from the signal superposition effect of being affiliated with a GCI.

Finally, the results in column (3) of Table 6 show that the regression coefficient for $gender$ is −0.388 and significant at the 1% level, indicating that when the company leader is female, it has a negative impact on the company’s ability to obtain external funding. However, the coefficient for the interaction term $incu\_certifi\times gender$ is 0.248 and significant at the 5% level, suggesting that location in a GCI conveys a stronger signal value for female entrepreneurs and is more likely to help them obtain venture capital, which supports the expectation of Hypothesis 4a.

4.3. Robustness Checks

In this section, we employ an additional five methods to further examine the robustness of our baseline regression results. We also use clustered robust standard errors at the incubator level in our regression analysis and lag the independent and control variables by one period to address endogeneity concerns.

Dependent Variable as a Continuous Measure. We transform the venture capital obtained by startups into a logarithmic scale, replacing the previously used binary dependent variable. Given the left-censored at zero in our sample data, we adopt the Tobit regression model for analysis [11]. The results, as presented in

Table 7. Results of the robustness checks.

	Model 1	Model 2	Model 3	Model 4
$\mathit{i}ncu\_certifi$	1.364 **	0.234 **	0.214 **	0.256 ***	0.101 **
	(0.631)	(0.108)	(0.102)	(3.31)	(2.02)
$if\_coached$		0.364 ***
		(0.064)
$incu\_if\_htz$		0.236 ***
		(0.110)
$if\_hightech$		0.187 ***
		(0.066)
$com\_size$	0.200 ***	0.027 ***	0.026 **	0.024	0.052 ***
	(0.072)	(0.012)	(0.011)	(1.58)	(5.34)
$com\_age$	−1.532 ***	−0.311 ***	−0.259 ***	−0.350 ***	−0.113 ***
	(0.306)	(0.057)	(0.050)	(−6.37)	(−3.10)
$com\_rev$	0.295 ***	0.045 ***	0.046 **	0.067 ***	0.079 ***
	(0.077)	(0.013)	(0.013)	(4.80)	(9.00)
$com\_RD$	0.748 ***	0.104 **	0.115 **	0.169 ***	0.184 ***
	(0.073)	(0.013)	(0.013)	(14.87)	(24.23)
$com\_edu$	3.909 ***	0.500 ***	0.580 ***	0.535 ***	0.767 ***
	(1.004)	(0.166)	(0.171)	(3.22)	(6.17)
$constant$	−27.402 ***	−4.358 ***	−3.935 ***	−5.525 ***	−4.104 ***
	(2.588)	(0.439)	(0.404)	(−10.10)	(−11.19)
Year Fixed Effects	Yes	Yes	Yes	Yes	Yes
Region Fixed Effects	Yes	Yes	Yes	Yes	Yes
Industry Fixed Effects	Yes	Yes	Yes	Yes	Yes
Wald chi2/F	97.79	732.66	749.99	1080.22	1707.10
Prob >chi2/F	0.000	0.000	0.000	0.000	0.000
Log likelihood	−28,412.32	−15,023.51	−15,243.71	−4765.54	−10,489.48
Observations	57,624	57,165	57,165	20,681	43,323

Notes: cluster robust standard errors in parentheses. ** p < 0.05, *** p < 0.01.

Model 1, show that the coefficient for incu_certifi is 1.364 and significant at the 5% level. So, the baseline assumptions remain valid.

Controlling for Alternative Explanations. As discussed in Section 3.2, to control for alternative explanations, such as the assistance of GCIs in capacity building and resource linking for startups to obtain venture capital, as well as the additional impact of other quality certification signals possessed by startups, we include if_coached, incu_if_htz, and if_hightech as alternative explanatory variables. The results, displayed in Table 7 Model 2, indicate that all three alternative explanations have a positive impact on startups’ ability to secure venture capital. However, even after incorporating these variables, affiliation with a GCI continues to exert a positive signaling effect on obtaining venture capital.

Rare Events Bias. Given that the incidence of startups obtaining venture capital in our sample is quite low, with only 7.4% of the dependent variable being 1, this constitutes a rare event. To mitigate the bias associated with rare events, we utilize the complementary log–log model for our regression [59]. As shown in Table 7 Model 3, we still find evidence supporting our baseline hypothesis.

Subsample Regression Tests. We divide our sample based on geographical characteristics into the central–western and eastern regions of China. The eastern region implies a better business network, and such business environmental differences may also affect the signaling benefits. The first column of Table 7 Model 4 presents the subsample regression results for the central–western region and the second column for the eastern region. The conclusions remain significantly supportive of the baseline regression. Furthermore, interestingly, by comparing the coefficient of the independent variable, we can observe that the signaling benefits of affiliation with a GCI for startups to obtain venture capital are more effective in the central–western region, where the business environment is relatively less developed.

Sample Selection Bias. In practice, the affiliation of startups with GCIs is not randomly determined. It involves a process where startups provide relevant documentation, initiate an application, and are subsequently evaluated by the incubator. Due to the stringent selection criteria within incubators, startups that are admitted into GCIs may inherently possess stronger development potential. This process introduces sample self-selection bias, leading to endogeneity concerns. To address this issue, we employ a propensity score matching (PSM) approach, dividing the sample into treatment and control groups based on a set of observable matching variables [60]. This ensures that the likelihood of entering a GCI is the same for both groups, with only the treatment group actually receiving the opportunity. We then examine whether there is a significant difference in obtaining venture capital between the two groups, thereby mitigating the sample selection bias [3].

Initially, we estimate the propensity scores for each sample using a Logit model, which predicts the probability of a startup entering a GCI based on a set of control variables. Part A of

Table 8. Propensity score matching Logit regression and balance test.

Part A: Logistic Regression		Part B: PSTEST
Variable	$\mathit{i}\mathit{n}\mathit{c}\mathit{u}\_\mathit{c}\mathit{e}\mathit{r}\mathit{t}\mathit{i}\mathit{f}\mathit{i}$	Variable	Unmatched/Matched	%bias
$com\_size$	0.066 *** (15.73)	$com\_size$	U	22.8
			M	−4.4
$com\_age$	−0.009 (−0.47)	$com\_age$	U	13.1
			M	−0.7
$com\_rev$	0.045 *** (12.21)	$com\_rev$	U	32.3
			M	0.5
$com\_RD$	0.151 *** (37.18)	$com\_RD$	U	51.0
			M	−0.0
$com\_edu$	0.588 *** (11.71)	$com\_edu$	U	14.6
			M	−1.6

Notes: cluster robust standard errors in parentheses. *** p < 0.01.

presents the Logit regression results, indicating that the registered capital, revenue, R&D expenditure, and human capital level of startups are positively correlated with the likelihood of being selected into a GCI, all significant at the 1% level. The age of the company is inversely related to GCI selection, although not statistically significant.

Subsequently, we conduct a balance test to verify whether the matching variables’ means are sufficiently close between the treatment and control groups, and whether the matching process has adequately balanced the data. Part B of Table 8 shows the results of the balance test. Following Rosenbaum and Rubin (1985), a post-matching standardized bias (%bias) of less than 20% in absolute value is considered indicative of a good match [60]. In our model, the post-matching standardized bias (%bias) for all variables is reduced compared to pre-matching, and all are less than 20% in absolute value. The matching results pass the balance hypothesis test, indicating reliability. Figure 1 presents the density functions of the propensity scores before and after matching. Compared to the pre-matching scenario, the density curves of the treatment and control groups are closer post-matching, indirectly reflecting a good matching effect and suggesting that the use of propensity score matching can reduce assessment errors.

Finally, we perform caliper matching on the treatment group and the sample group with a propensity score difference of 3% (less than 2 times their standard deviation) [61].

Table 9. Results of propensity score matching.

Variable	Sample	Treated	Controls	Difference	S.E.	T
$VC$	Unmatched	0.0788	0.0536	0.0252 ***	0.0023	10.97
	Matched	0.0787	0.0685	0.0102 ***	0.0038	2.71

Notes: *** p < 0.01.

presents the results of the propensity score matching analysis. It is evident that, although the disparity in venture capital funding between the treatment and control groups diminished after matching, the difference remained significantly positive at the 1% level, suggesting that the signaling value of GCIs persists. In terms of sample observations, only 30 control group members and 52 treatment group members were off support, indicating that only a minimal number of samples were lost during the matching process, which is unlikely to affect the outcome substantively.

In summary, after addressing endogeneity due to selection bias through the propensity score matching method, the final results remain consistent with the baseline regression findings, supporting the notion that GCIs serve as an effective third-party signal for startups in securing venture capital.

5. Discussion and Conclusions

Incubators typically offer a multifaceted suite of support to startups, including office space, consulting services, and network connections, to help them overcome resource constraints and accelerate growth during their early stages. Government certification of these incubators signifies that they meet certain standards and quality, thereby enabling them to provide a higher level of service and resources to the enterprises they nurture. Location in a GCI may serve as a signal to investors that a startup has undergone vetting and assessment and has received a certain level of endorsement. This signaling effect could influence venture capitalists’ decision making by reducing the uncertainty faced when evaluating a firm’s potential. Venture capitalists might perceive that startups growing within GCIs are more likely to succeed due to their access to professional guidance and support, potentially possessing a stronger business foundation and higher growth potential. Our study, with a pioneering focus on the Chinese IT industry, empirically tests the efficacy of GCIs as a third-party signal in the venture capital process. The findings corroborate that affiliation with GCIs enhances the likelihood of a startup securing venture capital by 25%. This endorsement signal not only mitigates the uncertainty for investors during the due diligence phase but also amplifies the value of intrinsic signals such as the founders’ managerial experience and the status of intellectual property.

Notably, considering the Matthew Effect, female entrepreneurs in gender-imbalanced technology industries often face additional challenges due to stereotypes. However, through the rigorous screening process of GCI, female entrepreneurs can mitigate the negative impact of such biases, enhancing market trust and legitimacy for their enterprises. This helps to compensate for the initial deficiencies in resources and networks, enabling them to secure more opportunities and gain an equitable competitive platform in highly contested environments. Consequently, this facilitates a redistribution of market benefits, which to some extent counteracts the “Matthew Effect”.

5.1. Contribution to Theory and Practice

Theoretically, our study enriches the research stream on startup signaling strategies in emerging markets, particularly during their early stages. We contribute to the entrepreneurial finance literature by underscoring the significance of GCIs as third-party signals within an emerging economy like China. We extend the understanding of the interplay between different signals by demonstrating how GCIs enhance the credibility of a startup’s intrinsic signals, thereby facilitating their access to venture capital. Furthermore, our research adds a new dimension to discussions on gender disparities in entrepreneurship by showing that GCIs may disproportionately benefit female founders, potentially leveling the playing field in a male-dominated venture capital market. Lastly, the signaling effect is also crucial for understanding how incubators influence startups’ funding pathways and success rates. Startups gain not only coaching and network resources from incubators but also an additional signal value by affiliating with high-reputation incubators, which serves as a means to enhance their legitimacy.

From a practical standpoint, there is ample evidence suggesting that government initiatives aimed at stimulating venture capital towards startups, including tax incentives and funding subsidies, have not had a broad impact [62,63]. However, governments possess behavioral additionality, influencing and altering the actions of other stakeholders within the entrepreneurial ecosystem [9]. Our findings indicate that GCIs operate in a more “favorable” manner, as they do not require significant government expenditure while effectively combining governmental authority with the market-oriented attributes of incubators. This reduces information asymmetry between startups and investors while fostering a more equitable entrepreneurial ecosystem. Secondly, in emerging markets, collaborating with reputable incubators is a forward-looking strategy for startups, providing them not only with quality resources but also enhancing their signaling potential to attract venture capital. Lastly, it is a win–win proposition for future investors and high-reputation incubators to collaborate, achieving joint investment in startups that benefits both parties.

Based on the above analysis, to maximize the advantages of GCI, we argue that governments should optimize policy design by establishing stricter selection criteria and oversight mechanisms to ensure the professionalism and efficiency of GCI. Additionally, governments can support female entrepreneurs in accessing GCI through targeted programs, thereby reducing gender disparities while also enhancing the promotion of GCI’s value to attract more high-quality startups to participate. Correspondingly, startups should actively seek affiliation with reputable GCIs to significantly enhance their capacity to attract external investment. Particularly for female entrepreneurs and other disadvantaged groups, recognition from a GCI can help address initial legitimacy deficits. For investors, the signal value of GCI as a third-party certification should be prioritized during the evaluation of startups to mitigate risks at the due diligence stage.

5.2. Limitations and Extensions

While our study provides significant insights, it is not without limitations. The focus on China’s IT industry may restrict the generalizability of our findings, making them less applicable to other sectors and geographic contexts. Furthermore, the retrospective nature of the data may introduce biases related to the survival and success of the sampled startups. Future research could address these constraints by incorporating longitudinal data and extending the analysis to a broader range of industries and countries.

Additionally, this study relies on propensity score matching to address sample selection bias; although effective, this method is not a perfect substitute for experimental design. Future research could employ more robust methodologies to strengthen causal inference. Regarding variable selection, our study uses dummy variables (0/1) to measure the dependent variable, i.e., whether a startup secures venture capital funding. While this approach allows for a clear capture of binary outcomes and simplifies model interpretation, future research could explore the possibility of using ratio scales to measure this variable. For example, by constructing more refined models, researchers could estimate the impact of GCIs on the return on investment for startups. This would not only provide more precise quantitative results but also offer policymakers and investors more detailed data support, thereby advancing research in this field.

We also encourage scholars to explore the long-term effects of GCIs on startups beyond the initial funding stage. Moreover, the informational value carried by government actions may also depend on the nature of the signal receivers, and its differentiated impact on banks, bond firms, and venture capitalists represents a novel research perspective. Lastly, comparative studies across different emerging economies can reveal contextual factors influencing the efficacy of third-party signals, further enriching understanding of signaling dynamics within entrepreneurial ecosystems.

5.3. Conclusions

In summary, our study underscores the pivotal role of GCIs as third-party signals in reducing information asymmetry and facilitating access to venture capital for startups in emerging economies. The certification by GCIs not only validates the intrinsic signals of startups but also acts as a catalyst for resource acquisition, especially for disadvantaged entrepreneurs, such as female business owners. The findings of this research provide practical implications for policymakers and stakeholders within the entrepreneurial ecosystem, emphasizing the need for supportive structures that can enhance the visibility and credibility of emerging enterprises. Moreover, the study reveals the role of government certification in mitigating information asymmetry and promoting a more transparent and equitable entrepreneurial ecosystem.