Effects of Local Government Behavior on University–Enterprise Knowledge Flow: Evidence from China

Abstract

Government financial investment has been increasingly adopted as a policy support to stimulate university–enterprise cooperation, however, empirical research from the perspective of knowledge flow remains limited. We reason that school–enterprise knowledge flow can be divided into dual stages, namely, knowledge creation and knowledge transfer, and this dual efficiency can be measured with the super-efficiency DEA model. The results show that the average value of knowledge creation efficiency (KCE) is higher than knowledge transfer efficiency (KTE). We adopt the Dynamic Generalized Spatial Model method to study the effect of government support on knowledge flow dual efficiency, and the regression results show that government support has a nonlinear effect on KCE while having a positive impact on KTE. We find that intergovernmental competition has a moderating influence on the relationship between government support and dual efficiency. Fiscal transparency can enhance the moderating effect of intergovernmental competition.

1. Introduction

Innovation is the first driving force, and leads development [1]. The main subjects responsible for technological innovation are universities, public research organizations, and enterprises [2]. In addition, there are cooperative relationships between these organizations to achieve technological progress through innovative interaction. The most typical partnership is university–enterprise collaborative innovation [3], in which university–enterprise knowledge flow plays an important role as a driving force [1]. Specifically, the role of universities in the national innovation system is to create new knowledge, and enterprises obtain economic benefits through the application of this new knowledge. To support universities to carry out innovative activities, enterprises provide universities with funds and other innovative resources for scientific research [1,4]. In university–enterprise knowledge flow, local government behavior has become the best candidate to regulate and strengthen this partnership, and its importance has become increasingly prominent [5]. The alliance between universities, enterprises, and local governments has been confirmed and supported by the triple helix theory [6].

However, unlike the traditional triple helix framework, which exhibits the features of blurry boundaries in the university–enterprise–government relationship, this study attempts to take the government as an external control parameter and examine its influence on university–enterprise knowledge flow. Therefore, we need to find a suitable research context to satisfy this research framework. The literature indicates that China’s institutional system is both unique and universal compared to other countries. Specifically, in contrast to a centralized political system, China has implemented a market economy [7]. The political and economic environment provides the Chinese government with strong institutional power and flexible market power in innovation policy implementation. On the one hand, universities and enterprises in China choose their innovation partners mainly because they match their mutual research goals and conditions [8]. On the other hand, the Chinese government provides a relatively well-developed innovation policy environment and grants financial subsidies for university–enterprise knowledge flow [9]. Therefore, the Chinese institutional context and university–enterprise–government relations align with the research framework of this study. Using China as a case study, the findings can provide policy implications for other developing countries and emerging economies.

In China, the importance of university–enterprise knowledge flow in constructing the national innovation system has always been emphasized, becoming a key support project of the Chinese government [10,11]. On the basis of the perspective of knowledge flow, universities are undoubtedly evolving into knowledge exporters and creators, while enterprises become knowledge importers and users. According to the duality of the innovation value chain theory, the process of knowledge flow involves knowledge creation and knowledge transfer [1,12,13]. The local government can formulate policies to encourage universities and enterprises to organize innovative cooperation activities and expand university–enterprise knowledge flow through financial investment [9].

Following reform of China’s tax-sharing system, local governments have gradually gained fiscal revenues and control power [14]. Corresponding to China’s fiscal decentralization reform is administrative centralization [15]. Such a political and economic environment has fundamentally changed how local governments participate in innovation activities. In order to fully implement the national strategic policy of innovation-driven development, the central government has put forward rigid technological innovation requirements for local governments and incorporated indicators such as technological innovation performance into the local official assessment system [16]. In this context, local governments must improve their competitiveness in order to obtain more innovation-related resources and attract innovation factors. University–enterprise knowledge flow plays a primary role in improving regional innovation capacity [17]. Therefore, promoting university–enterprise knowledge flow becomes one of the goals of intergovernmental competition. Along with intensification of the degree of intergovernmental competition, the impact of government support on university–enterprise knowledge flow may be influenced by different moderating sources.

Local governments centrally control most local innovation resources. The concentration of scientific management power of local governments provides rent-seeking space. In the absence of effective supervision, the rent-seeking space is even greater [18]. Fiscal transparency can curb budget violations by helping to ensure the public’s right to know, participate, and supervise [19]. Fiscal transparency impacts scientific technology investment, including for university–enterprise knowledge flow, reflecting the public’s restriction of local governments. Fiscal transparency can constrain the direction and extent of local government competition [18]. Specifically, fiscal transparency may affect the degree of influence of intergovernmental competition on the relationship between government support and university–enterprise knowledge flow to a certain extent [17].

In this study, we first measured knowledge creation efficiency (KCE) and knowledge transfer efficiency (KTE) and found that the average value of KCE is higher than KTE. Furthermore, the regression results indicate that government support has a nonlinear effect on KCE and a positive impact on KTE. We discovered that intergovernmental competition has a positive moderating effect in the relationship between government support and the dual efficiency of university–enterprise knowledge flow, as well as that fiscal transparency can enhance the moderating effect of intergovernmental competition.

The contributions of our study are as follows. First, in previous studies of university–enterprise cooperation most scholars have focused on university–enterprise cooperative behavior or single-dimensional university–enterprise cooperative innovation performance [20]. We divided university–enterprise knowledge flow into dual stages, namely, knowledge creation and knowledge transfer, which further promote the research progress of university–enterprise cooperation in the field of knowledge management.

Second, while many studies have shown that government support has a significant impact on university innovation performance or enterprise innovation ability [16,21], there is no unified answer as to whether local government behavior can play a role in university–enterprise cooperation. We found heterogeneous effects of government support on university–enterprise knowledge creation and knowledge transfer.

Third, different competition objectives may cause variance in the influence of government financial support on university–enterprise knowledge flow. Meanwhile, fiscal transparency, as one of the public’s supervisory modes for government management, can affect local government behavior, including intergovernmental competition [15]. Therefore, we studied the joint moderating effect of intergovernmental competition and fiscal transparency within the impact mechanism of government support and university–enterprise knowledge flow.

Finally, considering that innovation subjects have knowledge spillovers in their spatial location [2], the spatial correlation and temporal dependence of panel data may be ignored when using the multiple linear regression method. To reduce the estimation error in empirical research, we adopt the Dynamic Generalized Spatial Model (DGSM) method.

2. Theoretical Framework and Hypothetical Development

2.1. The Dual Efficiency of University–Enterprise Knowledge Flow

In previous studies related to knowledge flow, the innovation value chain theory has been applied the most widely [3,22,23,24]. The innovation value chain theory was first proposed in [22]. The authors considered innovation as a step-by-step process, dividing it into three stages: idea generation, conversion, and diffusion. Innovation value chain theory combines the core ideas of innovation theory and value chain theory, that is, knowledge and value. The basic idea of innovation value chain theory is that the innovation value chain has the characteristics of a chain structure, and the chain structure is a physical form made up of connected independent links consisting of basic units and linking units. Thus, the innovation process can be broken down into different links, each link corresponds to different units, and each unit is interrelated with the knowledge flow.

With scholars’ in-depth study of the innovation process, the three-stage structure of the innovation value chain theory has gradually been simplified into an ambidextrous perspective consisting of exploratory and exploitative innovation [25,26,27]. In addition, as a process, there is the possibility of dividing knowledge flow into multiple stages, for example, knowledge acquisition, knowledge integration, and knowledge application [2,3,13]. Therefore, based on the ambidextrous perspective of innovation value chain theory, research has divided knowledge flow into the dual stages of knowledge creation and knowledge transfer [1,9]. According to the above studies, we believe that the dual-stage structure can be applied to university–enterprise knowledge flow as well. Moreover, the dual stages of university–enterprise knowledge flow involve specific inputs and outputs. By measuring the input–output ratio, we can better monitor the efficiency of resource allocation in university–enterprise knowledge flow [1]. We applied the super-efficiency DEA model to measure the dual efficiency.

2.2. Government Support and the Dual Efficiency of University–Enterprise Knowledge Flow

In the traditional triple helix framework, the boundary features of university–enterprise–government relations are blurry. In order to examine the influence of government support on university–enterprise knowledge flow, this study attempts to construct a new research framework in which government is considered as an external control parameter. In the process of university–enterprise knowledge flow, universities, as the knowledge exporter, are the main body that creates knowledge value in the knowledge creation stage [1]. In the knowledge creation stage, the government’s financial expenditure on science and technology can provide innovation support for R&D cooperation activities between universities and enterprises and accelerate the university–enterprise knowledge creation process, thus contributing to the improvement of KCE [28,29]. However, if government innovation support is excessive, this may force universities and enterprises to participate in negotiations and communication with local governments involving the knowledge creation efforts of both partners. In addition, in the event that universities and enterprises rely mainly on government fiscal expenditures to carry out innovation cooperation activities, this will undoubtedly have a negative impact on their knowledge creation capacity. In the long-run, government support inevitably has a negative impact on KCE.

In the knowledge transfer stage, enterprises need to transform knowledge achievements created by universities into tools that they can effectively use [9]. In this process, the government’s financial expenditure on science and technology can build a well-established technical environment for the knowledge transfer activities of both universities and enterprises. At the same time, it can provide a strong driving force to promote the university–enterprise knowledge transfer process, thus promoting KTE [6,30]. Nevertheless, as government financial expenditures on science and technology continue to increase, such fiscal investment may crowd out the input of innovation resources by innovative entities, leading to a loss of efficiency in the knowledge transfer process over the long term [31,32]. In other words, when the government’s fiscal expenditures on science and technology continue to increase and exceed a certain threshold, government support may negatively impact the KTE.

Based on this theoretical analysis, we propose our first set of hypotheses:

H1a. 

Government support has an inverted U-shaped influence on KCE.

H1b. 

Government support has an inverted U-shaped influence on KTE.

2.3. The Moderating Effect of Intergovernmental Competition

Intergovernmental competition is a series of tax incentives and financial subsidies adopted by local governments competing to attract business investment, such as foreign direct investment (FDI), to promote regional innovation capacity and achieve local economic growth under China’s current performance appraisal system [14,33]. In the context of implementing China’s policy of innovation-driven development, the construction of a regional innovation system has gradually become an important target in the performance appraisal of local governments. It has been confirmed that business investment, especially foreign direct investment, has a significant positive impact on attracting the concentration of innovation resources and promoting the flow of regional innovation factors [34,35]. Therefore, local governments engage in competitive activities with other governments to achieve the relevant performance appraisal target, that is, “competition for innovation”.

According to previous research, university–enterprise knowledge flow is often closely related to national innovation systems [3,36]. Improving the dual efficiency of university–enterprise knowledge flow is beneficial to local governments in achieving innovation assessment performance. Fiscal expenditures are the basic means by which the government promotes university–enterprise knowledge flow, and intergovernmental competition may stimulate the impact of government financial support on university–enterprise knowledge flow [5,37]. Specifically, when government support has a positive effect on the efficiency of university–enterprise knowledge flow, a higher degree of intergovernmental competition increases this positive effect. When government support has a negative effect on the efficiency of university-enterprise knowledge flow, however, a higher degree of intergovernmental competition increases the negative effect.

Accordingly, we propose our second set of hypotheses:

H2a. 

Intergovernmental competition plays a positive moderating effect on the relationship between government support and KCE.

H2b. 

Intergovernmental competition plays a positive moderating effect on the relationship between government support and KTE.

2.4. The Joint Moderating Effect of Intergovernmental Competition and Fiscal Transparency

Local governments often compete to attract business investment in the form of tax incentives and financial subsidies, suggesting that intergovernmental competition is closely related to regional fiscal management [14]. Thus, the adequacy of government financial management mechanisms can directly affect the target achievement of intergovernmental competition. Additionally, government financial transparency is an important indicator of the financial management mechanism [19]. Especially in the Chinese context, where local governments have fiscal autonomy, fiscal transparency implies the governments’ efforts to combat corruption and rent-seeking behavior [18,38]. In other words, regional fiscal transparency is a direct indicator of fiscal management capability and governmental integrity, and can indirectly reflect the business environment. Therefore, fiscal transparency is crucial to the competitiveness of local governments in attracting foreign investment, which affects intergovernmental competition activities. Thus, the higher the regional fiscal transparency, the more conducive it is to guaranteeing intergovernmental competition in the university–enterprise knowledge flow. Fiscal transparency may enhance the impact of intergovernmental competition on the relationship between government support and the dual efficiency of university–enterprise knowledge flow.

Therefore, we propose the following hypothesis:

H3. 

Fiscal transparency positively influences the moderating effect of intergovernmental competition.

3. Materials and Methods

3.1. Model

According to the first law of geography, provincial university–enterprise cooperation depends on economic development and human capital factors. It is affected by university–enterprise cooperation in the surrounding provinces [39]. Therefore, spatial effects cannot be ignored in research on the efficiency of university–enterprise knowledge flow. In addition, the efficiency of university–enterprise knowledge flow in a certain province is usually related to the previous period. More precisely, the efficiency of university–enterprise knowledge flow has a spatial spillover effect and shows a temporal dynamic effect. As a powerful tool for studying spatial economics, the DGSM method can meet the needs of this study. The DGSM can explain the spatial spillover effect of the efficiency of university–enterprise knowledge flow in surrounding provinces in the local area and resolve the temporal dynamic effect of the efficiency [40]. The equation expression of DGSM is:

$${\mathrm{Y}}_{\mathrm{it}}=\mathsf{\alpha }+{\mathsf{\delta }\mathrm{Y}}_{\mathrm{it}-1}+\mathsf{\rho }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}{\mathrm{Y}}_{\mathrm{jt}}+{\mathsf{\beta }\mathrm{X}}_{\mathrm{it}}+{\mathsf{\eta }}_{\mathrm{i}}+{\mathsf{\nu }}_{\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{it}} {\mathsf{\mu }}_{\mathrm{it}}=\mathsf{\lambda }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}{\mathsf{\mu }}_{\mathrm{jt}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

Here, Y_i,jt is the dependent variable, i,j = 1, 2, …, N (N = 30) reflect 30 provinces in China, t = 1, 2, …, T (T = 14) reflect 14 years of data, Y_it−1 is the dependent variable that lags one year, α is the constant term, δ refers to the coefficient of the efficiency that lags one year, X_it denotes the independent variables, ρ is the regressive spatial coefficient, λ is the spatial autocorrelation coefficient, β is the coefficient of the independent variables, W_ij denotes the spatial weight matrix, η_i refers to the province effect, ν_t denotes the time effect, µ_i,jt reflects the spatial fixed effects, and ε_it is the error term.

On this basis, the equation expression to test the inverted U-shaped influence of government support on the efficiency of university–enterprise knowledge flow is as follows:

$$\begin{array}{cc}\hfill \mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{it}}=& \mathsf{\alpha }+\mathsf{\delta }\mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{it}-1}+\mathsf{\rho }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}\mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{jt}}+{\mathsf{\beta }}_1{\mathrm{lnGST}}_{\mathrm{it}}+\hfill \\ & {\mathsf{\beta }}_2\mathrm{lnGST}{2}_{\mathrm{it}}+{\mathsf{\beta }}_{\mathrm{n}}{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\eta }}_{\mathrm{i}}+{\mathsf{\nu }}_{\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{it}} {\mathsf{\mu }}_{\mathrm{it}}=\mathsf{\lambda }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}{\mathsf{\mu }}_{\mathrm{jt}}+{\mathsf{\epsilon }}_{\mathrm{it}}\hfill \end{array}$$

(2)

In Formula (2), lnKCE and lnKTE_i,jt are the dependent variables in this study, representing the efficiency of university–enterprise knowledge flow, while lnKCE and lnKTE_it−1 are the efficiencies of university–enterprise knowledge flow that lags one year; lnGST_it is the core independent variable in this study, representing government support, lnGST2_it is the quadratic term of government support, X_it denotes the control variables selected in this study, and β₁ and β₂ are the regression coefficients that are the main focus of this paper. Among them, when β₁ is significantly positive and β₂ is significantly negative, it indicates that there is an inverted U-shaped relationship between government support and the efficiency of university–enterprise knowledge flow.

The equation expression to test the moderating effect of intergovernmental competition on the relationship between government support and the efficiency of university–enterprise knowledge flow is as follows:

$$\begin{array}{cc}\hfill \mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{it}}& =\mathsf{\alpha }+\mathsf{\delta }\mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{it}-1}+\mathsf{\rho }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}\mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{jt}}+{\mathsf{\beta }}_1{\mathrm{lnGST}}_{\mathrm{it}}+\hfill \\ & {\mathsf{\beta }}_2\mathrm{lnGST}{2}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{lnIGC}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{lnGST}}_{\mathrm{it}}\times {\mathrm{lnIGC}}_{\mathrm{it}}+{\mathsf{\beta }}_5\mathrm{lnGST}{2}_{\mathrm{it}}\times {\mathrm{lnIGC}}_{\mathrm{it}}+{\mathsf{\beta }}_{\mathrm{n}}{\mathrm{X}}_{\mathrm{it}}+\hfill \\ & {\mathsf{\eta }}_{\mathrm{i}}+{\mathsf{\nu }}_{\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{it}} {\mathsf{\mu }}_{\mathrm{it}}=\mathsf{\lambda }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}{\mathsf{\mu }}_{\mathrm{jt}}+{\mathsf{\epsilon }}_{\mathrm{it}}\hfill \end{array}$$

(3)

In Formula (3), lnIGC_it is the moderating variable in this study, representing intergovernmental competition, while β₄ and β₅ are the regression coefficients that are our main focus. Among them, if β₁ is significantly positive and β₂ is significantly negative in Formula (2), and if β₅ is significantly positive here, this indicates that intergovernmental competition plays a positive moderating role on the inverted U-shaped relationship between government support and the efficiency of university–enterprise knowledge flow. Moreover, if β₁ passes the significant test, β₂ is not significant in Formula (2), and β₄ is significantly positive here, this indicates that intergovernmental competition plays a positive moderating effect on the positive relations between the government support and the efficiency of university–enterprise knowledge flow.

The equation expression to test the moderating effect of fiscal transparency on the moderating role of intergovernmental competition is as follows:

$$\begin{array}{cc}\hfill \mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{it}}=& \mathsf{\alpha }+\mathsf{\delta }\mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{it}-1}+\mathsf{\rho }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}\mathrm{lnKCE}\backslash {\mathrm{lnKTE}}_{\mathrm{jt}}+{\mathsf{\beta }}_1{\mathrm{lnGST}}_{\mathrm{it}}+\hfill \\ & {\mathsf{\beta }}_2\mathrm{lnGST}{2}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{lnIGC}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{lnGST}}_{\mathrm{it}}\times {\mathrm{lnIGC}}_{\mathrm{it}}+{\mathsf{\beta }}_5\mathrm{lnGST}{2}_{\mathrm{it}}\times {\mathrm{lnIGC}}_{\mathrm{it}}+\hfill \\ & {\mathsf{\beta }}_6{\mathrm{lnFTY}}_{\mathrm{it}}+{\mathsf{\beta }}_7{\mathrm{lnGST}}_{\mathrm{it}}\times {\mathrm{lnFTY}}_{\mathrm{it}}+{\mathsf{\beta }}_8\mathrm{lnGST}{2}_{\mathrm{it}}\times {\mathrm{lnFTY}}_{\mathrm{it}}+\hfill \\ & {\mathsf{\beta }}_9{\mathrm{lnIGC}}_{\mathrm{it}}\times {\mathrm{lnFTY}}_{\mathrm{it}}+{\mathsf{\beta }}_{10}{\mathrm{lnGST}}_{\mathrm{it}}\times {\mathrm{lnIGC}}_{\mathrm{it}}\times {\mathrm{lnFTY}}_{\mathrm{it}}+ {\mathsf{\beta }}_{11}\mathrm{lnGST}{2}_{\mathrm{it}}\times {\mathrm{lnIGC}}_{\mathrm{it}}\times \hfill \\ & {\mathrm{lnFTY}}_{\mathrm{it}}+{\mathsf{\beta }}_{\mathrm{n}}{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\eta }}_{\mathrm{i}}+{\mathsf{\nu }}_{\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{it}} {\mathsf{\mu }}_{\mathrm{it}}=\mathsf{\lambda }{\sum }_{\mathrm{j}=1}^{\mathrm{N}}{\mathrm{W}}_{\mathrm{ij}}{\mathsf{\mu }}_{\mathrm{jt}}+{\mathsf{\epsilon }}_{\mathrm{it}}\hfill \end{array}$$

(4)

In Formula (4), lnFTY_it is the moderating variable in this study, representing the fiscal transparency, while β₁₀ and β₁₁ are the regression coefficients that are our main focus. Among them, if β₁ is significantly positive and β₂ is significantly negative in Formula (2), β₅ is significantly positive in Formula (3), and β₁₁ is significantly positive here, this indicates that fiscal transparency has a positive moderating effect on the moderating role of intergovernmental competition. Moreover, if β₁ passes the significant test, β₂ is not significant in the Formula (2), β₄ is significantly positive in Formula (3), and β₁₀ is significantly positive here, this indicates that fiscal transparency has a positive moderating effect on the moderating role of intergovernmental competition as well.

3.2. Variables

3.2.1. Dependent Variables

The dependent variables are the efficiency of university–enterprise knowledge flow, namely, KCE and KTE as measured with the super-efficiency DEA model. Referring to the index selection of past studies [1,20,41,42,43,44], the evaluation index system of KCE and KTE in this study is shown in

Table 1. Evaluation index system of the efficiency of university-enterprise knowledge flow.

Variables	Primary Indicator	Secondary Index
KCE	Input 1	Number of R&D personnel at that time in the universities
	Input 2	Amount of cooperative funding entrusted by enterprises
	Output 1	Number of academic papers
	Output 2	Number of patents authorized
KTE	Input 1	Number of academic papers
	Input 2	Number of patents authorized
	Output 1	Number of R&D technology transfer contracts
	Output 2	Total funds for R&D technology transfer

. It is worth mentioning that our reason for choosing the number of R&D technology transfer contracts and total funds to measure KTE is that technology transfer is one of the main pathways by which universities can transfer their research outputs to the industrial sector [42,43,44,45,46,47].

3.2.2. Core Independent Variables

The core independent variables are government support (GST) and its quadratic term (GST2). As mentioned above, fiscal expenditure is the most basic way for local governments to support regional innovative activities [21]. It is reasonable to use government fiscal expenditure to reflect government support for university–enterprise knowledge flow [33]. Therefore, we adopted the proportion of scientific and technological expenditure in local government fiscal expenditure to express the degree of GST.

3.2.3. Moderating Variables

Based on theoretical analysis, we believe that the degree of intergovernmental competition and fiscal transparency play a joint moderating role in the influence mechanism of GST on KCE and KTE. Therefore, we chose intergovernmental competition (IGC) and fiscal transparency (FTY) as the moderating variables in this study.

Intergovernmental competition mainly refers to the economic behavior of local governments in attracting liquidity elements and serving their jurisdictions through taxation, environmental policies, and welfare [48]. In fact, due to the restrictions of the registration system and the financial system, the liquidity elements of labor and capital are often insufficient [27]. Therefore, as a liquidity element with significant economic benefits and spillover effects, scholars have often favored FDI to measure the degree of local government competition in past studies [14,19]. Thus, we used each province’s ratio of FDI to GDP to define IGC.

Fiscal transparency reflects the extent to which taxpayers can participate in fiscal activities and the degree of fiscal democracy [49]. The “Report on China’s Fiscal Transparency” issued by the Public Policy Research Center of Shanghai University of Finance and Economics is an authoritative earlier assessment of fiscal transparency in China. The report takes 31 provinces in China as its survey objects. It uses various methods, such as information disclosure applications, online searches, and document collection (ranging from public budgets, fund budgets, financial account management capital budgets, state-owned capital operating budgets, government asset debts, departmental budgets, social insurance fund budgets, state-owned enterprises, and the attitude and sense of responsibility of those surveyed) to examine the degree of openness of provincial fiscal information and calculate the provincial fiscal transparency index. The value range of the index is 0–100, with larger values indicating higher fiscal transparency in government. We used this index to measure the FTY of each province.

3.2.4. Control Variables

In order to eliminate the error in the regression results, we selected control variables according to previous studies [1,16], including the level of human capital (HCL), industrial structure (ILS), infrastructure (INF), and marketization (MAR). Among these, we used the number of college students per 10,000 people to measure the HCL variable and the ratio of the added value of tertiary industry to GDP to measure the ILS variable. The INF variable was measured by the per capita road area, while the MAR variable was characterized by the proportion of non-state investment in total regional investment.

3.3. Data Sources

Based on the principle of data availability, the sample in this study contained 30 provinces, autonomous regions, and municipalities in China, excepting Tibet, Taiwan, Hong Kong, and Macau. The data were derived from the Compilation of Science and Technology Statistics of Institutions of Higher Learning, the China Statistical Yearbook, and the Easy Professional Superior (EPS) data platform. The sample period was from 2007 to 2020. We used the linear interpolation method to adjust and supplement the data and adopted the logarithmic form of all the variables in this paper (In order to avoid negative values after taking the logarithm of values less than 0, we follow the method of adding 1 to the value of all variables and then taking the logarithm.). The symbols and descriptive analysis results of the variables are shown in

Table 2. The results of summary statistics of variables.

Variables	Symbol	Observation	Mean	Standard Deviation	Min	Max
Knowledge creation efficiency	lnKCE	420	0.671	0.205	0.017	2.026
Knowledge transfer efficiency	lnKTE	420	0.257	0.213	0.008	1.705
Government support	lnGST	420	0.058	0.337	0.024	0.084
Intergovernmental competition	lnIGC	420	0.006	0.022	0.000	0.016
Fiscal transparency	lnFTY	420	3.681	0.472	2.582	4.570
Human capital	lnHCL	420	5.206	0.308	4.173	6.242
Industrial structure	lnILS	420	0.251	0.133	0.136	0.558
Infrastructure	lnINF	420	3.109	1.165	0.641	5.315
Marketization	lnMAR	420	0.121	0.062	0.036	0.277

.

4. Results

4.1. Spatial Autocorrelation Test Results

In order to test the spatial correlation of lnKCE and lnKTE, we measure the global Moran’s I based on two spatial-weight matrices (It should be noted that the two spatial-weight matrices used in this paper are specified with W₁ being the geographic adjacency matrix and W₂ the geographic distance matrix.), that is, W₁ and W₂, respectively. In

Table 3. The results of the Global Moran’s I of lnKCE and lnKTE.

Year	W1		W2
	lnKCE	lnKTE	lnKCE	lnKTE
2007	0.403 ***	0.239 ***	0.171 ***	0.134 ***
2008	0.348 ***	0.309 ***	0.153 ***	0.181 ***
2009	0.357 ***	0.271 ***	0.190 ***	0.155 ***
2010	0.322 ***	0.285 ***	0.224 ***	0.192 ***
2011	0.319 ***	0.227 **	0.207 ***	0.136 ***
2012	0.264 **	0.187 **	0.176 ***	0.098 **
2013	0.218 **	0.244 ***	0.125 **	0.175 ***
2014	0.178 *	0.268 ***	0.159 ***	0.188 ***
2015	0.227 **	0.215 **	0.131 **	0.129 ***
2016	0.179 *	0.426 ***	0.168 ***	0.112 ***
2017	0.213 **	0.356 ***	0.106 **	0.083 **
2018	0.246 **	0.309 ***	0.145 **	0.131 ***
2019	0.359 ***	0.267 ***	0.117 **	0.172 ***
2020	0.368 ***	0.253 ***	0.179 ***	0.154 ***

Note: *, ** and *** are significant at the statistical levels of 10%, 5%, and 1%, respectively.

, we find that lnKCE and lnKTE have a positive spatial correlation between 2007 and 2020. The results show a positive spatial spillover effect of lnKCE and lnKTE in China, meaning that empirical analysis should be performed using spatial econometric models.

4.2. Main Results

In the main results, the dependent variables of Models 1 and 2 are lnKCE while those of Models 3 and 4 are lnKTE. Models 1 and 3 are regression results based on W₁, while Models 2 and 4 are based on W₂. Moreover, the Hausman test was performed on all models, with the results showing that the fixed-effect model should be used for Models 1–4. The regression results for all models are shown in

Table 4. The main results of lnGST on lnKCE and lnKTE.

	Model 1	Model 2	Model 3	Model 4
Variables	lnKCE	lnKCE	lnKTE	lnKTE
L.lnKCE\L.lnKTE	0.093 ***	0.069 **	0.134 ***	0.214 ***
	(0.028)	(0.030)	(0.043)	(0.043)
lnGST	0.120 **	0.136 **	0.079 **	0.102 ***
	(0.053)	(0.063)	(0.035)	(0.027)
lnGST2	−0.085 *	−0.118 **	0.139	−0.181
	(0.049)	(0.045)	(0.257)	(0.253)
lnIGC	0.466 *	0.391	0.303 *	0.312
	(0.220)	(0.431)	(0.162)	(0.269)
lnFTY	0.289 **	0.405 ***	0.230 ***	0.336 *
	(0.125)	(0.139)	(0.071)	(0.188)
lnHCL	−0.147	0.399 **	−0.286 **	−0.168 *
	(0.579)	(0.158)	(0.142)	(0.101)
lnILS	−0.591	−0.403	−0.632	−0.360
	(17.824)	(13.265)	(9.195)	(6.743)
lnINF	0.197 ***	0.284 ***	0.144	0.232 ***
	(0.052)	(0.045)	(0.107)	(0.067)
lnMAR	−0.093 *	−0.075	0.062	0.125 **
	(0.051)	(0.064)	(0.070)	(0.061)
ρ	0.478 ***	0.535 **	0.411 ***	0.240 **
	(0.162)	(0.242)	(0.137)	(0.115)
λ	−0.303 **	−0.269 **	−0.639 ***	−0.382 **
	(0.148)	(0.124)	(0.213)	(0.189)
Time effect	Fixed	Fixed	Fixed	Fixed
Province effect	Fixed	Fixed	Fixed	Fixed
Observations	420	420	420	420
Adj-R2	0.542	0.369	0.638	0.317

Note: *, ** and *** are significant at the statistical levels of 10%, 5%, and 1%, respectively. L.lnKCE and L.lnKTE represent the efficiency of university–enterprise knowledge flow that lags one year.

. By comparing statistical indicators such as regressive spatial coefficient (ρ), spatial autocorrelation coefficient (λ), and goodness of fit (R²), we chose the results of DGSM based on W₁ for analysis, that is, Models 1 and 3. It is worth mentioning that we adopt the spatial autocorrelation (SAC) method here to conduct the empirical research.

First, we note that the impact of lnGST on lnKCE is positive (β = 0.120, p < 0.05) and that lnGST2 negatively impacts lnKCE (β = −0.085, p < 0.10) in Model 1. This indicates an inverted U-shaped relationship between government support and knowledge creation efficiency, which supports hypothesis H1a. Second, we observe that while the spatial regressive coefficient of Model 1 is positive (β = 0.478, p < 0.01), the spatial autocorrelation coefficient is negative (β = −0.303, p < 0.05). Therefore, the results show that lnKCE has a positive spatial spillover effect in adjacent regions. These results indicate that the higher the university–enterprise knowledge creation efficiency is in a region, the more it can help its neighboring regions improve their efficiency, which is line with previous research [1]. However, lnGST has a negative role on lnKCE in adjacent regions, which may be due to the mobility of innovation elements and the attractiveness of government support to innovative elements in neighboring provinces. Finally, the lagging phase of lnKCE has a positive impact on itself (β = 0.093, p < 0.01), which indicates that increased knowledge creation efficiency has time inertia, that is to say, it has the effect of innovation accumulation.

In Model 3, although the influence of lnGST on lnKTE is positive (β = 0.079, p < 0.05), there is no significant correlation between lnGST2 and lnKTE (β = −0.139, p > 0.10), which shows that government support has a positive effect on knowledge transfer efficiency. Thus, hypothesis H1b is not supported. This may be because the financial support provided by the Chinese government for the transformation of knowledge achievements in university–enterprise cooperation has not reached the required value at this stage. Specifically, the impact of government support on the efficiency of university–enterprise knowledge transfer has not passed the inverted U-shaped inflection point, and acts as a promotion effect. This result is very similar to the study of Zhang and Wang [9]. Moreover, while the spatial regressive coefficient is positive (β = 0.411, p < 0.01), the spatial autocorrelation coefficient is negative (β = −0.639, p < 0.01) in Model 3. These results reveal that lnKTE has a positive spatial spillover effect in adjacent regions. Simultaneously, lnGST has a negative role on lnKTE in adjacent regions, the same as the results of the spatial coefficients in Model 3. Furthermore, the lagging phase of lnKTE has a positive impact (β = 0.134, p < 0.01) as well, which indicates that the increase of knowledge transfer efficiency has time inertia which is expressed as innovation accumulation.

4.3. Results of the Influence Mechanism

In order to study the influence mechanism of lnGST on lnKCE and lnKTE, we examined the joint moderating effect of lnIGC and lnFTY by following the empirical model construction from previous studies [50,51,52]. Among them, the dependent variable of Model 5 is lnKCE, while the dependent variable of Model 6 is lnKTE. The Hausman test was again performed on the models, and the results showed that the fixed-effect model should be adopted in Models 5 and 6; we the SAC method was again used to carry out the empirical research. The regression results are shown in

Table 5. The results of the influence mechanism of lnGST on lnKCE and lnKTE.

	Model 5	Model 6
Variables	lnKCE	lnKTE
L.lnKCE\L.lnKTE	0.068 **	0.115 ***
	(0.040)	(0.033)
lnGST	0.143 ***	0.062 **
	(0.038)	(0.042)
lnGST2	−0.191 *	−0.128
	(0.112)	(0.143)
lnIGC	0.544	0.806
	(0.936)	(1.132)
lnFTY	0.274 ***	0.482 **
	(0.054)	(0.211)
lnGST × lnIGC	−0.425 *	0.371 *
	(0.257)	(0.193)
lnGST2 × lnIGC	−0.481 **	0.506
	(0.199)	(1.352)
lnGST × lnFTY	−0.221	0.325 **
	(0.431)	(0.147)
lnGST2 × lnFTY	−0.374	−0.403
	(0.528)	(0.916)
lnIGC × lnFTY	0.720 *	0.593
	(0.414)	(0.486)
lnGST × lnIGC × lnFTY	−0.506	0.910 ***
	(0.459)	(0.308)
lnGST2 × lnIGC × lnFTY	−0.827 **	1.215
	(0.410)	(2.161)
ρ	0.382 ***	0.451 ***
	(0.065)	(0.118)
λ	0.257 ***	0.239 *
	(0.087)	(0.130)
Control variables	Fixed	Fixed
Time effect	Fixed	Fixed
Province effect	Fixed	Fixed
Observations	420	420
Adj-R2	0.549	0.440

Note: *, ** and *** are significant at the statistical levels of 10%, 5%, and 1%, respectively. L.lnKCE and L.lnKTE represent the efficiency of university–enterprise knowledge flow that lags one year.

.

From Model 5, it can be seen that the interaction of lnGST2 and lnIGC has a negative influence on lnKCE (β = −0.481, p < 0.05), which indicates that the degree of intergovernmental competition can enhance the inverted U-shaped relationship between government support and knowledge creation efficiency. Therefore, hypothesis H2a is supported. From Model 6, it can be seen that the interaction of lnGST and lnIGC has a positive influence on lnKTE (β = 0.371, p < 0.10), which indicates that intergovernmental competition has a positive moderating effect on the promotional influence of government support on knowledge transfer efficiency. Thus, hypothesis H2b is supported as well. Moreover, these results show that the financial support of the local government for university–enterprise knowledge transfer has not reached the extreme value and verified the governmental objective of “competition for innovation”, which is verified by previous research [15,32].

Furthermore, to identify the differences in the moderating effect of intergovernmental competition under different levels of fiscal transparency, we note the coefficients of the triple interaction terms in Models 5 and 6. On the one hand, the interaction of lnGST2, lnIGC, and lnFTY is negatively correlated with lnKCE in Model 5 (β = −0.827, p < 0.05); on the other hand, the interaction of lnGST, lnIGC, and, lnFTY has a positive effect on lnKTE in Model 6 (β = 0.910, p < 0.01). From these results, we can conclude that hypothesis H3 is verified.

4.4. Results of the Robustness Check

Although the above empirical analysis mainly applies the DGSM method to analyze the impact mechanism of lnGST on lnKCE and lnKTE, there are other manifestations of the commonly used spatial econometric models. Therefore, we performed the regression analysis again using dynamic SAR, SEM, and SDM methods, respectively, to check the robustness of the main regression results. The robustness check results are shown in

Table 6. The regression results of robustness checks.

	Model 7	Model 8	Model 9	Model 10	Model 11	Model 12
Variables	lnKCE	lnKTE	lnKCE	lnKTE	lnKCE	lnKTE
L.lnKCE\L.lnKTE	0.112 **	0.240 ***	0.087 **	0.208 **	0.151 ***	0.163 **
	(0.054)	(0.056)	(0.206)	(0.102)	(0.052)	(0.190)
lnGST	0.107 ***	0.074 **	0.145 **	0.104 *	0.119 ***	0.054 *
	(0.031)	(0.034)	(0.066)	(0.061)	(0.036)	(0.030)
lnGST2	−0.049 **	0.014	−0.093 **	0.136	−0.045 **	0.086 ***
	(0.023)	(0.026)	(0.045)	(0.278)	(0.021)	(0.028)
lnIGC	0.264	0.449 *	1.037	0.706	0.904 ***	0.342 ***
	(0.518)	(0.261)	(3.059)	(2.162)	(0.127)	(0.099)
lnFTY	0.142 *	0.304 *	−0.215	−0.336	0.234 **	0.163 **
	(0.085)	(0.184)	(0.717)	(0.814)	(0.109)	(0.069)
lnGST × lnIGC	−0.515 **	0.281	−0.702 ***	0.481 **	−0.356	0.219 *
	(0.244)	(0.363)	(0.122)	(0.218)	(0.603)	(0.128)
lnGST2 × lnIGC	−0.320 ***	0.245 **	−0.182 **	0.136	−0.438 **	0.179 **
	(0.062)	(0.122)	(0.079)	(0.088)	(0.203)	(0.083)
lnGST × lnFTY	0.202	0.161 ***	−0.220 ***	0.310	0.411	0.245
	(1.025)	(0.041)	(0.077)	(2.813)	(1.529)	(0.794)
lnGST2 × lnFTY	−0.327	−0.112 ***	−0.410 ***	0.183	−0.263	−0.225
	(1.328)	(0.028)	(0.103)	(0.259)	(2.030)	(1.110)
lnIGC × lnFTY	0.518 *	−0.242 **	0.360 **	−0.161	0.721 *	0.295
	(0.290)	(0.120)	(0.152)	(0.399)	(0.432)	(0.816)
lnGST × lnIGC × lnFTY	−0.208	0.370 **	−0.277 **	0.486 ***	−0.144	0.312 **
	(1.133)	(0.162)	(0.118)	(0.132)	(0.729)	(0.106)
lnGST2 × lnIGC × lnFTY	−0.482 ***	0.654 ***	−0.886 **	0.705	−0.560 **	0.480
	(0.045)	(0.102)	(0.373)	(2.452)	(0.261)	(1.417)
ρ	0.164 ***	0.128 ***	0.094 **	0.217 ***	0.362 ***	0.250 **
	(0.034)	(0.026)	(0.046)	(0.049)	(0.108)	(0.113)
Control variables	Fixed	Fixed	Fixed	Fixed	Fixed	Fixed
Time effect	Fixed	Fixed	Fixed	Fixed	Fixed	Fixed
Province effect	Fixed	Fixed	Fixed	Fixed	Fixed	Fixed
Observations	420	420	420	420	420	420
Adj-R2	0.363	0.605	0.774	0.430	0.827	0.725

Note: *, ** and *** are significant at the statistical levels of 10%, 5%, and 1%, respectively. L.lnKCE and L.lnKTE represent the efficiency of university–enterprise knowledge flow that lags one year.

, in which Models 7 and 8 are based on the dynamic SAR method, Models 9 and 10 are based on the dynamic SEM method, and Models 11 and 12 are based on the dynamic SDM method. In addition, by further adopting the LR and Wald tests (p-value < 0.05) we found that the analysis should be conducted for the results of the dynamic SDM model (Due to space limitations, we omit the regression results of the spatial lagged terms of independent variables in the dynamic SDM method; these are available upon request from the authors.), that is, Models 11 and 12.

From Model 11, it can be seen that lnGST2, the interaction of lnGST2 and lnIGC, and the interaction of lnGST2, lnIGC, and lnFTY are all negatively correlated with lnKCE. These results indicate that government support has an inverted U-shaped influence on knowledge creation efficiency which can be positively moderated by intergovernmental competition, while fiscal transparency appears to strengthen the effect of intergovernmental competition. Moreover, the results of Model 12 show that lnGST, the interaction of lnGST and lnIGC, and interaction of lnGST, lnIGC, and, lnFTY are all positively correlated with lnKTE. Therefore, we believe that government support has a promotional impact on knowledge transfer efficiency which can be positively moderated by intergovernmental competition, while fiscal transparency may further enhance the effect of intergovernmental support. From the above analysis, we can conclude that the robustness of the main regression results is verified.

5. Discussion and Conclusions

Based on dual knowledge flow, this study divides university–enterprise knowledge flow into dual stages, namely, knowledge creation and knowledge transfer, and uses the super-efficiency DEA model to measure the dual efficiency. By adopting the DGSM method, we find that government support has an asymmetric effect on the dual efficiency of university–enterprise knowledge flow. Specifically, government support has an inverted U-shaped effect on knowledge creation efficiency while positively impacting knowledge transfer efficiency. In addition, intergovernmental competition plays a moderating role in the relationship between government support and dual efficiency. Moreover, we incorporate fiscal transparency into the theoretical and empirical framework, revealing that fiscal transparency can enhance the moderating effect of intergovernmental competition.

5.1. Theoretical Implications

This study provides theoretical implications for deeper understanding of the mechanism of local government behavior on university–enterprise cooperation. First, from the perspective of knowledge flow, this study reveals the asymmetric mechanism of government support on knowledge creation efficiency and knowledge transfer efficiency. Our findings can help to further understanding of the relationship between government financial investment and university–enterprise knowledge flow, and can enrich the theory of the university–enterprise–government triple helix from the field of knowledge management.

Second, this study verifies the effect of intergovernmental competition on the process of university–enterprise collaborative innovation when using government financial support to improve knowledge flow efficiency. Our findings clarify the moderating role of intergovernmental competition in the relationship between government support and university–enterprise knowledge flow, providing new evidence for the “competition for innovation” view of Chinese local government. These findings help to understand the mechanism of local government behavior and university–enterprise knowledge flow.

Third, this study analyzes the complexity of the joint moderating mechanism of financial transparency and intergovernmental competition as boundary conditions. By revealing the joint moderating effect of fiscal transparency and intergovernmental competition on the impact of government support on knowledge creation efficiency and knowledge transfer efficiency, these findings can help to expand research into the boundary conditions of the influencing mechanism of local government behavior on university–enterprise knowledge flow.

5.2. Policy and Managerial Implications

Based on our research findings, we propose the following practical implications. First, government financial investment in science and technology is a vital booster in improving university–enterprise cooperation. In particular, to promote the improvement of university–enterprise cooperation, local governments should effectively distinguish the different stages of knowledge creation and knowledge transfer and use R&D subsidies rationally in the different stages. Compared to university–enterprise knowledge transfer, local governments should be more cautious in using financial investment in order to avoid potential negative effects on knowledge creation.

Second, local governments should pay special attention to controlling the degree of intergovernmental competition within a reasonable range. While advancing the reform of the fiscal decentralization system, the central government should effectively restrain the self-interested investment preferences of local governments and should adopt a more diversified official evaluation system in order to improve local officials’ implementation of central innovation policy. These measures can guide the innovation preferences of local government competition, thereby strengthening the effect of government support on university–enterprise knowledge flow.

Third, governments should pay attention to the different meanings of financial investment in science and technology for regional university–enterprise knowledge flow under conditions of different fiscal transparency. The specific suggested approach is to raise the level of public participation in supervision of government fiscal investment and increase the transparency of government fiscal expenditures. Meanwhile, governments should actively engage in fiscal transparency by guiding intergovernmental competition, thus improving university–enterprise knowledge flow.

5.3. Limitations and Future Research

This study has several limitations that can offer scope for future research. First, university–enterprise knowledge flow is divided into knowledge creation and knowledge transfer stages in this paper, meaning that we had to choose more than one index in order to measure knowledge creation efficiency and knowledge transfer efficiency. Due to limited data availability, we only chose the secondary index mentioned above. Therefore, it would be possible to choose more indicators in future research in order to re-measure the dual efficiency. Second, because the sample in this study was 30 provinces in China, the research scope is relatively macro-level and insufficiently focused. I would be possible to lower the research scope to the city level and conduct an empirical study using urban panel data to reveal the influence mechanism of local government behavior on university–enterprise knowledge flow in future research. Finally, in the theoretical framework of this paper, fiscal transparency was used as a moderating variable on intergovernmental competition as a way to test its indirect moderating effect. However, fiscal transparency has a direct and significant impact on government fiscal investment. Therefore, in future research, we intend to further investigate its direct moderating effect on government fiscal investment and the efficiency of university–enterprise knowledge flow.