A research on the effectiveness of innovation policy for regional innovation under Chinese long-range plan

Abstract

By adopting a set of panel data from thirty provinces in China from 2007 to 2017, this paper examines how regional innovation efficiency is influenced by different types of innovation policies from a specific long-range plan (‘National Medium- and Long-Term Science and Technology Development Plan (2006–20)’). We incorporate public direct research and development (R&D) grants, R&D tax credits, and intellectual property rights (IPR) protection into our research. We also explore the heterogeneous influences of these factors at different stages in regional innovation development. The results indicate that public direct R&D grants have a significantly negative influence, especially in innovation-catching-up regions. Conversely, the R&D tax credit has a considerably positive impact on innovation-catching-up regions. IPR protection plays various roles according to different innovation development stages. Specifically, IPR protection has a positive impact on innovation-leading regions, while it negatively influences regional innovation efficiency in innovation-catching-up regions.

1. Introduction

Science and technology (S&T) policies play an essential role in improving regional innovation (Boeing and Peters 2021; Gao et al. 2021a). Central and local governments have issued guidelines for regional S&T development and innovation capability building. The European Union (EU) proposes smart specialization as a policy framework to enhance regional and industrial innovation capabilities.¹ The World Trade Organization (WTO) promulgated the Agreement on Subsidies and Countervailing Measures, in which R&D subsidies are classified as non-actionable subsidies (Coppens 2014). The National Medium- and Long-Term Science and Technology Development Plan (2006–20) in China (hereinafter referred to as the MLP) emphasizes constructing a regional innovation system with its characteristics and advantages.²

The existing literature studies have explored the relationship between S&T policy and innovation capability. For example, by employing an exclusive set of panel data from manufacturing firms in Jiangsu province, Gao et al. (2021a) suggested that R&D subsidies, in general, do promote firms’ exploratory innovation. At the same time, some scholars believe that S&T policies also have negative effects, which are not conducive to promoting corporate R&D investment (Acemoglu et al. 2018), and subsidies are misappropriated (Boeing and Peters 2021). Using a panel of Chinese provinces during the 2000–10 time period, Boeing et al. (2022) found that an increase of R&D subsidies by one standard deviation decreases private R&D expenditures in large- and medium-sized enterprises by 6.5 per cent. However, few studies analyse the effectiveness of innovation policies from the perspective of regional innovation efficiency (RIE) (Min et al. 2020). RIE is the ratio of innovation input to output for a given region and is a crucial indicator of regional performance. RIE primarily reflects the capability of regions to transform innovation investment into knowledge and market value (Guan and Chen 2012). In most cases, the goal of S&T policy is to promote the R&D input and output of regions. However, an upward/downward change in R&D expenditures may not translate into an upward/downward change in innovation efficiency. Therefore, it is necessary to explore the relationship between S&T policy and RIE (Broekel 2015).

The implementation of a continuous S&T policy and China’s progress in innovation provide good opportunities for the study of regional innovation policy and innovation efficiency (Boeing et al. 2022). Since the Chinese economic reform and opening-up in 1978,³ China has achieved remarkable economic development. In 2005, China became the world’s fourth-largest economy. China surpassed France and became one of the world’s top five countries in scale in terms of gross domestic product (GDP), measured at current prices. However, China’s economic growth was heavily dependent on energy and resource consumption, and the national indigenous innovation capability was weak.⁴ It was imperative to improve the overall innovation efficiency. To optimize the allocation of resources, transform the investment-driven model, and improve China’s scientific and technological capability and innovation efficiency, China proposed the MLP (2006–20) in 2006. The promulgation of these supportive policies tends to promote the optimization of resource allocation in various regions and enhance the scientific and technological innovation capabilities and RIE.

Support from the MLP contains a series of policy combinations, including public direct R&D grants, R&D tax credits, and institutional measures, such as the powerful protection and promotion of intellectual property rights (IPRs) (Liu et al. 2011, 2017). Extant studies also show that in the process of implementing MLP, the Chinese government’s S&T innovation policy has promoted the R&D input and output of enterprises to some extent (Guo et al. 2016; Liu et al. 2016; Zhou et al. 2020). However, these policy instruments still have shortcomings. For example, the government selecting R&D grants may provoke corporate alienation and rent-seeking behaviours (Acemoglu et al. 2018), which may reduce RIE. Thus, the Chinese government adopts different mechanisms to support S&T innovation. However, the impact of such mechanisms on innovation efficiency may vary. Research on the effects of innovation efficiency is still scarce. The MLP ended in 2020, and it is time to review and discuss the influence of MLP implementation on RIE.

Furthermore, regional innovation development stages vary in China (Liu et al. 2018). Due to imbalances in resource endowments, institutional construction, and economic development, the effects of S&T support policies on RIE may differ (Fan et al. 2019; Fu 2008). However, the empirical evidence related to this regional contingency factor is still scarce. This study attempts to fill the aforementioned research gaps and answer two main research questions: (1) how do S&T policies in an MLP (e.g. public direct grants, tax credits, and IPR protection) influence RIE? (2) What are the heterogeneous effects of these S&T policies on RIE at different innovation development stages?

To answer these questions, we use the data envelopment analysis (DEA) method to calculate the RIE in China (Chen and Guan 2012; Guan and Chen 2012) and analyse the effects of public direct R&D grants, R&D tax credits, and IPRs on RIE during the MLP period. We will also examine the heterogeneous impacts of innovation policies at different development stages. Due to the heterogeneity of regional geographical location and resource endowment (Fu 2008; Sun 2000), regional innovation capability also varies. That is, regions may be at different development stages, and hence, we classify regions into innovation-leading regions and innovation-catching-up regions according to the Chinese regional innovation index (Liu et al. 2018). The results demonstrate that these policies have contrasting effects on RIE. For example, public direct grants have a negative and significant effect on RIE, particularly in innovation-catching-up regions, whereas R&D credit has a positive and significant impact on RIE. The impact of IPR protection depends on the development stages. The IPRs substantially positively impact innovation-leading regions while negatively impacting innovation-catching-up regions.

The rest of this paper is organized as follows: we introduce the institutional background of China’s MLP in Section 2 and review the theoretical background and develop hypotheses in Section 3. In Section 4, we describe the data resources, variables, and econometric methods. Following the empirical results in Section 5, we then provide the conclusions and limitations of our study in Section 6.

2. Background of China’s MLP

After more than 40 years of exploration and reform in S&T innovation policies, China has established a relatively comprehensive S&T innovation policy system, including public direct R&D grants, R&D tax credits, and IPR protection. Since joining the WTO, China has attracted foreign capital and technology to promote and upgrade its domestic industries, thus increasing China’s economic strength. However, this change exacerbated China’s economic dependence on alien technology, especially in some sectors. In response to this issue and to improve its indigenous innovation capabilities, China launched the MLP in 2006 with the following goal: ‘By 2020, the proportion of the national research and development investment in GDP will increase to more than 2.5 per cent. The degree of foreign technology dependence will be reduced to less than 30 per cent. The annual number of patents granted by Chinese people and the number of citations of international scientific papers will both rank among the top five in the world.’

To ensure the realization of MLP goals, the State Council of China has issued sixty-six supporting policies to encourage independent innovation from ten aspects, including tax policy, financial policy, government procurement, and scientific and technological talent team building.⁵ The policy principles include indigenous innovation, leap-frogging in fields of strategic priority, enabling sustainable development, and leading the future. To achieve MLP goals and improve RIE, the Chinese government also launched additive policy instruments. Different ministries published sixty-eight detailed innovation policy documents affiliated with the MLP.⁶ There are three types of supporting policies in the MLP: public direct R&D grants, R&D tax credits, and IPR protection.

First, the MLP emphasizes the role of government public direct R&D grants. Governments at all levels should significantly increase investment and ensure their budgets and steady growth in S&T investment. The government established a diversified and multichannel technology investment system. Financial S&T investment focuses on supporting basic research, social welfare research, and cutting-edge technology research. The government needs to give full play, at all levels, to the guiding role of monetary funds in encouraging independent innovation of enterprises, innovating the investment mechanism, integrating government funds, increasing support, and enabling enterprises to carry out technological innovation and the digestion, absorption, and re-innovation of imported advanced technologies. The state has formulated a series of projects to ensure investment in S&T. For example, for supporting essential tasks to carry out research work, the National High-tech Research and Development Program (863 Program) mainly supports scientific research institutes, colleges, universities, domestic-funded institutes or domestic-funded holding enterprises with independent legal personality in mainland China. The National Key Basic Research and Development Program (973 Program) mainly supports scientific research institutions, colleges,universities, domestic-funded institutes, or domestic-funded holding enterprises, with legal person status in mainland China to carry out basic research for primary national strategic needs and to undertake related major scientific research programmes.

Second, governments at all levels increase the pre-income deduction for investment in enterprises’ independent innovation. Companies can deduct the current year’s taxable income at 150 per cent of the actual costs in technology development incurred in the current year. According to tax law, insufficient deduction of the basic technology development expenses in the current year can be carried forward for up to 5 years. If the employee education fund drawn by the enterprise is within 2.5 per cent of the total taxable salary, it can be deducted before the enterprise income tax. The government will improve tax policies to promote the development of high-tech enterprises and will improve the pre-tax deduction policy for taxable salaries of high-tech enterprises. The pre-tax deduction policy for corporate income tax wage expenditures adjusts the pre-tax deduction limit for corporate wage expenditures to 1,600 yuan per capita per month.

Third, the national S&T department, together with the intellectual property management department, establishes an intellectual property information service platform, supports the processing of intellectual property information and strategic analysis, and provides intellectual property information services for the creation and market development of independent IPRs. China’s government actively participates in the formulation of international standards and promotes domestic technical standards. The government also establishes and improves the IPR system, increasing legal enforcement and protection of IPRs. The State Intellectual Property Office reformed the patent examination method for inventions, enhancing the efficiency of the substantive examination of patents. The National Development and Reform Commission, the Ministry of Science and Technology, the Ministry of Finance, and other departments have issued a series of regulations to create a favourable intellectual property environment. For example, the ‘Catalogue of Key Technologies and Important Products with Independent IPRs in Chinese Information Industry’ supports companies that develop technologies and products in the cataloguing of patent applications, standard formulation, international trade, and cooperation.

Since launching the MLP, China has improved its indigenous innovation capabilities, promoted economic and social development with S&T support, and optimized the environment for innovation policies. According to Ministry of Science and Technology of the People's Republic of China at the National S&T Conference in 2019, the proportion of R&D expenditure to GDP is 2.15 per cent and it has been improving over the years. The number of international scientific papers ranks second, and the number of applications and granted invention patents ranks first in the world.⁷

3. Hypothesis development

3.1 The effect of public direct R&D grants on RIE

Public direct R&D grants are the most commonly-used policy instrument, which can directly compensate for the resource shortage and increase the return of R&D activities by providing low-cost or even free funds to enterprises (David et al. 2000; Dimos and Pugh 2016). As innovation efficiency is a measure of transformation of R&D inputs into outputs, the relationship between public direct R&D grants and efficiency is more complicated. We analyse the relationship between them from different angles.

Public direct R&D grants via governmental R&D projects can provide enterprises with extra resources for innovation (David et al. 2000; Guan and Yam 2015). Direct grants also enhance the marginal rate of return on R&D investment to compensate for enterprises’ loss due to the attributes of public goods of R&D activities (David et al. 2000; Zuniga-Vicente et al. 2014). At the same time, direct financial support from governments will increase firm-level risk tolerance for high-risk innovation (Chapman and Hewitt-Dundas 2018; Gao et al. 2021c). Furthermore, direct R&D grants create a resource munificent environment to protect recipient enterprises from unfavourable uncertainties and risks (Jourdan and Kivleniece 2017). Thus, by protecting enterprises from potential adverse selection, direct R&D grants may encourage enterprises to invest more in their R&D activities, which is regarded as input additionality (Dimos and Pugh 2016). By helping enterprises overcome resource constraints and encouraging more resource deployment on R&D, direct grants can facilitate enterprises’ technological capabilities, which are essential for innovation (Hottenrott and Richstein 2020; Lazzarini 2015). For example, R&D grants can promote enterprises’ investment in the recruitment of high-quality R&D personnel, as tacit knowledge carriers, to enhance enterprises’ technological capabilities (Gao et al. 2021a). Higher technological capabilities can, in turn, support firms in allocating resources more reasonably to facilitate innovation efficiency (Yam et al. 2004). Therefore, the following hypothesis is proposed:  

However, direct R&D grants, as an ex ante administrative subsidy through a non-market exchange mechanism, may lead to crowding-out effects on private R&D investment, poor efficiency of public funds, and subsequently productive insufficiency of R&D activities due to the misalignments of interest and information asymmetries (Gao et al. 2021b; Guan and Yam 2015; Hall and Van Reenen 2000; Perez-Sebastian 2015). More specifically, enterprises in nature are profit-seekers, while governments are more concerned about social welfare (Jourdan and Kivleniece 2017; Li et al. 2018). Compared with investment in R&D activities, which may produce high-quality technological achievements and further social welfare but at higher-level uncertainty, enterprises are more likely to allocate resources to projects that are high yielding and low risk but less innovative (Gao et al. 2021a; Tang et al. 2019). Although enterprises are sponsored by R&D grants, the grant recipient enterprises may only use these ‘cheap’ grants for innovation without providing additional resources or even simply substitute their own R&D investments with direct grants, as enterprises tend to avoid the loss of their own resources (Lach 2002; Zhou et al. 2020). This issue is exacerbated by the lack of a complete selection and supervision mechanism and the lack of effective evaluations; hence, the effectiveness of direct grants is highly dependent on the information available to the public agencies that manage the allocation of R&D grants (Castellacci and Lie 2015; Gao et al. 2021a).

At the same time, under an incomplete supervision mechanism, sponsored enterprises may lack internal discipline due to information asymmetry (Jourdan and Kivleniece 2017). This may result in the moral hazard behaviour of some grant recipient enterprises, which may actually be unwilling to invest in R&D and hence be non-compliant with funding rules (Boeing and Peters 2021), thereby leading to a low-effective usage of public resources for R&D. Furthermore, it is difficult for governments to set effective evaluations to measure the quality of R&D and the real growth in technological capabilities of sponsored enterprises (Jia et al. 2019). Under low-effectiveness evaluations, some enterprises will simply pursue the number of low-quality patents to quickly and more prudently satisfy government requirements or to allocate resources to non-productive rent-seeking activities to obtain continuous government support (Antonelli and Crespi 2013; Gao et al. 2021a; Wang et al. 2020). This may result in resource distortion related to enterprises’ R&D activities. Consequently, public R&D grants fail to achieve their purpose and they reduce the efficiency of regional innovation due to resource misallocation and market distortion. Therefore, the competing hypothesis is proposed as follows:  

3.2 The effect of R&D tax credits on RIE

To remedy the shortages of direct grants, another type of R&D-subsidized instrument, R&D tax credits, is designed (Hall and Van Reenen 2000). Enterprises enjoy tax deductions according to their R&D investment through the support of R&D tax credits (Busom et al. 2014; Chen and Yang 2019). The basic mechanism of tax credits is the direct reduction of the marginal cost of R&D, and firms are required to invest in R&D in response to linked tax credits (David et al. 2000). Unlike direct grants, R&D tax credits minimize the ‘picking-the-winner’ issue during the selection process of sponsored enterprises, which is an unbiased approach to the industry and the nature of the enterprises (Castellacci and Lie 2015; Czarnitzki et al. 2011). This means that, regardless of projects or industrial sectors, tax credits are available to all enterprises undertaking R&D activities. Enterprises can gain more tax credits as their R&D investment increases. Thus, enterprises are encouraged to allocate more additional resources to R&D according to their own decisions based on tax credit rules (Castellacci and Lie 2015). R&D tax credits are expected to avoid the crowding-out effect on firm-level R&D expenditure and are expected to enable subsidized enterprises to gain the highest rate of private return with less governmental control on the usage of R&D subsidies (Busom et al. 2014; Castellacci and Lie 2015). Under less governmental control, enterprises can enhance the efficiency of resource usage. Thus, R&D tax credits may alleviate the government’s resource misallocation and help improve innovation efficiency. Therefore, the following hypothesis is proposed:  

However, tax credits on R&D activities may also be less effective. First, with tax incentives, enterprises have higher autonomy to decide the composition of R&D and are more likely to concentrate financial resources on projects that will generate fast returns (Czarnitzki et al. 2011). Instead, higher innovation with more social welfare but long-term projects may be less favoured under the sponsorship of tax incentives. Furthermore, once the tax credits for R&D expenses incurred in the past year are available, enterprises can use the funds generated by tax credits at their discretion without necessarily using them for further R&D (Xie et al. 2021). Second, an enterprise does not enjoy the benefits of the tax credit before R&D expenses are incurred. This implies that the enterprise must raise funds for its R&D and needs to mitigate the risks involved in self-financing, market competition, and business operations on its own. In other words, unlike direct grants, R&D tax credits do not reduce the risks inherent in R&D activities (Xie et al. 2021). Third, the risk of adverse selection also exists in the case of using R&D tax credits. Enterprises may relabel their expenditures to inflate actual R&D costs to gain more tax benefits (Hall and Van Reenen 2000; Xie et al. 2021). Chen et al. (2021) show that enterprises in China inflate their R&D expenses by 25 per cent on average to ‘sneak’ into tax credits. Thus, enterprises’ risk aversion to innovative projects and inflation of R&D expenses will reduce innovation efficiency through tax credits. The competing hypothesis is proposed as follows:  

3.3 The effect of IPRs on RIE

IPRs promote enterprises’ innovation by granting intellectual property owners a certain degree of monopoly power to protect their legal innovation income (Xiao et al. 2013). Specifically, patents, as a formal and well-defined appropriation mechanism, can clarify the ownership of the intellectual output (Arundel 2001). IPR protection incentivizes enterprises to innovate and provides them with a technology monopoly for a certain period (Landes and Posner 2009). Enterprises can generate excess revenue to compensate for the loss due to technology externalities and can continuously invest profits from patents in R&D. Thus, stronger IPR protection increases enterprises’ returns to innovation (Allred and Park 2007). With a better IPR protection environment, enterprises are more willing to invest their resources in R&D, rationally allocate innovation resources, and improve innovation efficiency. Apart from the appropriation mechanism, IPR protection can enhance enterprises’ patent applications to assess and reward R&D personnel, promote licence income, and improve the potential for technological knowledge exchange and trading (Blind et al. 2006). IPR protection can shape the institutional environment (INS) to facilitate RIE. Thus, the following hypothesis is proposed:  

3.4 The heterogeneous effects of regional development stages

The effects of governmental support for RIE may vary under different innovation development stages at the regional level (Oughton et al. 2002).

In innovation-leading regions, knowledge flow is faster with higher efficiency, and legal systems are more complete and well established. More importantly, enterprises’ technological capabilities are higher in terms of undertaking more innovative R&D. Thus, the potential negative effect of direct R&D grants on RIE may be weakened in innovation-leading regions.

At first, the higher level of regional innovation development implies a higher quality of local government with higher capabilities of public governance (Jia et al. 2019; Lau and Lo 2015). Governments in innovation-leading regions may have closer and more frequent interactions with local enterprises to reduce information asymmetry (Lau and Lo 2015). The public officials in government branches in innovation-leading regions may be more diligent in additional information collection to scrutinize subsidy applicants and select more appropriate recipients. This can, to some extent, reduce the bias and resource distortion generated by the ‘picking-the-winner’ strategy and strengthen the enhancing effect of direct grants. Second, innovation-leading regions usually have more complete legal systems (Fan et al. 2019; Peng 2002) to prevent enterprises’ cheating and rent-seeking behaviours. Complete legal systems have stricter constraints on enterprises and they can penalize cheating and rent-seeking behaviours on innovation. Consequently, the opportunities to avoid cheating and rent-seeking actions that harm innovation efficiency will be reduced under the complete legal systems. In addition, the technological capabilities of enterprises in innovation-leading regions are higher, which can provide more benefits for knowledge sharing, especially valuable tacit knowledge (Gao et al. 2021a). This can also promote the effectiveness of direct R&D grants in terms of improving recipients’ technological capabilities and thereby strengthening innovation efficiency. Thus, the following hypothesis is proposed:  

Regional heterogeneity also exists in the effect of R&D tax credits (Freitas et al. 2017). As mentioned previously, enterprises are required to undertake R&D activities and invest related resources before they receive the tax credit. In doing so, R&D tax credits provide enterprises more autonomy to allocate their R&D investment. Thus, unlike direct grants, enterprises supported by tax incentives do not have to invest heavy resources in frontier R&D projects with long-term and high-risk R&D activities. In particular, in innovation-catching-up regions, enterprises’ technological capabilities are lower, and regional innovation developments rely more on the secondary innovation model to undertake incremental innovation at a fast pace (Wu et al. 2009). Enterprises in innovation-catching-up regions may invest more funds in low-risk but fast-returning R&D activities and hence reduce inefficient R&D due to the conduct of their R&D projects that are sponsored by direct grants but are beyond their technological capabilities. At the same time, tax incentive recipients attempt to maximize the success rate of R&D projects, as tax credits are ex post subsidies (Zuniga-Vicente et al. 2014). Thus, recipients often have in-depth control over their R&D activities, and these efforts may lead to a more rigorous and structured operation of R&D activities to enhance innovation efficiency. Therefore, R&D tax credits may be more beneficial for enterprises in innovation-catching-up regions, which rely heavily on imitation and secondary innovation.

At the same time, the R&D tax credits modify, to some extent, the resource misallocation caused by the government’s picking-the-winner strategy. Tax credits directly provide tax relief based on enterprises’ R&D investment, avoiding the efficiency loss caused by government choice. This is particularly important to the innovation-catching-up regions where the public governance capabilities are comparatively low to precisely evaluate innovative enterprises. Furthermore, tax credits, as a type of inclusive policy with less discrimination, can benefit more enterprises and effectively ease the financial constraints on R&D activities in innovation-catching-uping-up regions. Thus, innovation-catching-up regions can benefit more from R&D tax credits. The following hypothesis is proposed:  

The effects of IPR protection also vary across innovation-leading and innovation-catching-up regions (Allred and Park 2007). In innovation-leading regions, due to developed INSs with IPRs, enterprises with higher technological capabilities can more easily increase the appropriability of their R&D achievements and can make more profits (Landes and Posner 2009). These profits can be used for continuous R&D investment and for supporting new R&D programmes. In addition, technologically leading enterprises are generally more affected by knowledge externalities (Arora et al. 2016). With higher IPRs, enterprises with higher technological capabilities are more willing to exchange their R&D achievements through IPRs to improve technology externalities (Allred and Park 2007). Knowledge exchange among enterprises can increase technology externalities and facilitate the technological capabilities of other enterprises within the same region.

The innovation-catching-up regions that are more distant from the knowledge frontier may benefit more from weaker IPR protection. More specifically, in innovation-catching-up regions, due to technological disadvantages and external dependencies, technological development is primarily manifested in incremental imitative innovation (Kim and Lee 1987; Xiao et al. 2013). Thus, knowledge spillover may be an important source of inputs for innovation in catching-up regions. However, higher IPR protection, which leads to technological monopolies, leads innovation resources based on knowledge spillover to become less accessible. Thus, the following hypothesis is proposed:  

The framework of our study is shown in Fig. 1. We put the three major S&T policies in the MLP into a unified framework to analyse the impact of policies on RIE. Besides, we analyse the regional heterogeneity of different development stages.

4. Research design

4.1 Sample selection and data sources

The data sources are mainly from five databases. The provincial-level economic data are from the ‘China Economic Statistics Database’ of the National Information Centre and the ‘China Labour Statistics Yearbook’ from the National Bureau of Statistics. The R&D subsidy and innovation performance data are from the ‘Statistical Yearbook of Scientific and Technological Activities of Industrial Enterprises’ and the China Intellectual Property Index Report (CIPR) published by the CIPR Research Group. In addition, some other data such as foreign direct investment (FDI) are from the Wind database. We also use data from the National Bureau of Statistics and provincial statistical yearbooks to supplement missing data and verify the data’s accuracy and reliability. After matching and merging these databases, this study obtains panel data for thirty provinces from 2007 to 2017.⁸

4.2 Measurement of RIE

4.2.1 Measurement methodology

Innovation is a complex and non-linear process (Cruz-Cázares et al. 2013; Guan and Chen 2012) involving multiple processes, such as idea generation, innovation input and output, and commercialization (Dziallas and Blind 2019). Thus, calculating innovation efficiency should combine the effect of multiple input–output variables. At present, the commonly-used methods for measuring innovation efficiency in academia include the DEA method and stochastic frontier analysis (SFA) (Guan and Chen 2012). Compared with SFA, DEA has many advantages in calculating the RIE. First, DEA is a non-parametric estimation method. It builds an optimal model based on the regional innovation’s input–output variables and performance optimization analysis, avoiding the pitfalls of the subjective weighting method (Charnes et al. 1978). Second, DEA does not need a specific production function. Due to significant variances in resource endowments and development stages among Chinese provinces (Fan et al. 2019; Fu 2008; Tian et al. 2016), it is not easy to find a universal production function suitable for RIE in China. Third, DEA is suitable for calculating multiple input–output variables and comprehensively measuring RIE (Cruz-Cázares et al. 2013). Fourth, the unit does not restrict the choice of input–output indicators. It retains the original data in the statistical yearbook to the greatest extent and reduces the impact of data processing on the results (Zemtsov and Kotsemir 2019). Therefore, we choose DEA to measure RIE in this paper.

DEA is a systematic analysis of the relative efficiency of input and output factors proposed by Charnes et al. (1978). The DEA model mainly uses mathematical programming and statistical data to determine a relatively effective production frontier while keeping the input or output of the decision-making unit (DMU) unchanged. Then, each DMU is projected onto the DEA production frontier surface, and its relative effectiveness is evaluated by comparing the degree of deviation of the DMU from the DEA frontier surface.

In the DEA model, the most widely used model is the |${C^2}B$| model, which uses a fixed-scale assumption and linear programming to estimate the production boundary to measure the relative efficiency of each DMU. There are |$n$| DMUs of the same type, and each DMU has m types of input and p types of output. The input and output vectors are |${X_k} = {\left( {{x_{1k}},{x_{2k}}, \cdots ,{x_{mk}}} \right)^T}$| and |${Y_k} = {\left( {{y_{1k}},y, \cdots ,{y_{pk}}} \right)^T}$|⁠, and the following model is built:

In the formula, |$\theta $| represents the effective utilization of input relative to output, which is the efficiency of the DMU. If |$\theta = 1$|⁠, it means that the DMU is on the frontier surface, which is indicative of a technically effective DMU. If |$\theta \lt 1$|⁠, it indicates that there is a technical loss. |${\theta _k}{x_{ik}}$| represents the product of |${\theta _k}$| and |${x_{ik}}$|⁠. |${s^ + }$| and |${s^ - }$|are slack variables representing the input redundancy and output shortage of |${\rm{DM}}{{\rm{U}}_k}$| relative to the frontier, respectively. |$\lambda $| represents the weight of the jth DMU.

4.2.2 Dependent variable: input–output efficiency variables

RIE is the ratio of input to output obtained by regional innovation. It reflects the ratio of innovation agents’ actual output of an innovation factor in a specific innovation environment to the highest possible innovation output possible in correspondence with the factor input. Therefore, the choice of input–output variables is important for evaluating RIE (Cruz-Cázares et al. 2013; Guan and Chen 2012). In the knowledge production function, capital, labour, and technology are the key elements that affect knowledge outcomes , which are also input variables commonly used in DEA models (Bai 2013; Cruz-Cázares et al. 2013; Guan and Chen 2012; Zemtsov and Kotsemir 2019). This study uses these three variables as input variables. We adopt the efficiency value to measure the regional input and output and avoid differences in the absolute amount of input caused by each region’s area and population. The full-time equivalents of regional R&D personnel per capita reflect the human capital input level of the RIE. The proportion of regional R&D expenditures in the regional GDP is used as the capital input level of the RIE. In addition, because of the widespread ‘secondary innovation’ model of regional innovation in China (Guan et al. 2009; Wu et al. 2009), we also use per capita foreign technology contract transaction volume as the cost input for core technology.

Innovation production includes two stages: the knowledge production process and the knowledge commercialization process (Chen and Guan 2012; Guan and Chen 2012). At the stage of knowledge production, patents are often regarded as important outputs (Bai 2013; Cruz-Cázares et al. 2013; Guan and Chen 2012). Compared to the number of patents granted, the number of patent applications can be timelier and can comprehensively reflect regional innovation information (Li 2015). Therefore, we used the number of patent applications per capita in the regional statistics from the National Bureau of Statistics as the innovation output variable.

At the stage of knowledge commercialization, we take technology products and technology transactions into account. We used the per capita income of high-tech industries as a variable of technology products of innovation. High-tech industries are industries whose main products are technology and play a leading role in economic development.⁹ High-tech industries are innovation-intensive industries with a large amount of innovative output. In the Organization for Economic Co-operation and Development (OECD) area, high-technology industries account for more than 53 per cent of total manufacturing R&D.¹⁰ Drawing on the classification method of the OECD, the National Bureau of Statistics of China classifies manufacturing industries with relatively high R&D input intensity as high-tech industries, including pharmaceutical manufacturing, aviation, spacecraft and equipment manufacturing, electronic and communication equipment manufacturing, computer and office equipment manufacturing, medical equipment and instrument manufacturing, information chemicals manufacturing, and six other categories.¹¹ Hence, we consider the per capita income of high-tech industries as a variable of regional innovation output. It is measured by the ratio of the regional high-tech industry’s main business income to the total population. In addition, the commercialization process includes the economic value that is transformed through technology market transactions. This article thus takes the per capita technology market transaction contract value as the output variable of RIE.

The input–output variables of the DEA model, which are used to measure innovation efficiency, are shown in Table 1.

4.3 Independent and control variables

4.3.1 Independent variables

4.3.1.1 Public direct R&D grants (GRANT).

Existing research generally uses direct government subsidies to enterprises and dummy variables of government subsidies as measurements (Dimos and Pugh 2016). In China, governments at all levels provide companies with different types of R&D grants. This makes public direct R&D grants a very inclusive variable. We measured public direct R&D grants by the amount of government funding in a region’s internal R&D expenditures. In the China Statistical Yearbook, there are four sources of internal expenditure of R&D funds: government funds, enterprise funds, foreign funds, and other funds. Among them, government funds refer to all kinds of funds from the government finances at all levels in the internal expenditure of R&D funds, including funds for scientific and technological expenditures and other financial functions for the actual expenditures of R&D activities.¹² Due to regional differences in economic aggregates, population aggregates, and geographic areas, there is a large gap in public direct R&D grants. Consequently, we used a percentage of public direct R&D grants over government revenue to measure the intensity of public direct R&D grants.

4.3.1.2 R&D tax credit (TAX).

In some enterprise-level research, scholars use the actual amount of tax credit and dummy variables to measure R&D tax credit (Busom et al. 2014; Czarnitzki et al. 2011). However, these measures are not suitable for regional-level research. The Statistical Yearbook of Industrial Enterprise Science and Technology Activities provides the total amount of tax credit for each province. Although the rate of tax credit for R&D is uniform across the country, the actual implementation varies from region to region.¹³ To this end, we measure tax credit using total tax credits as a percentage of R&D investment.

4.3.1.3 IPRs.

They are generally measured using the survey method, legislative strength, and comprehensive evaluation method. A survey method is usually used to evaluate scholars, business managers or lawyers, and other relevant personnel. Sampling and the cognition of respondents can affect the survey results. Legislative strength is commonly used internationally to evaluate IPRs. The most representative index isproposed by Ginette and Park (abbreviated as GP- index) (Cho et al. 2015; Ginarte and Park 1997). However, China implements the same set of methods in different regions. The legislative strength makes it difficult to distinguish the differences in the enforcement of IPRs in different regions. The CIPR is published by the CIPR Task Group, evaluating IPRs in various regions by four dimensions: output of intellectual property, the level of intellectual property market movement, comprehensive performance of intellectual property, and possibilities of creations of intellectual property.¹⁴ Considering the comprehensive situation of China’s regional IPRs, it is more suitable to study the impact of IPRs on RIE. This article adopts the IPR index in the CIPR to measure the intensity of regional IPRs.

4.3.2 Innovation development stages

4.3.2.1 Innovation development stage (INNO).

Due to differences in geographical location and resource endowments, there is a duality in the innovation and development between regions in China (Liu and Hu 2002; Sun 2000). We differentiate the development stages of regions according to the regional innovation index of China, which is an annual report that comprehensively evaluates the innovation capabilities of China’s provinces (Liu et al. 2017).¹⁵ The regional innovation index includes five first-level indicators, twenty second-level indicators, forty third-level indicators, and 138 fourth-level indicators.¹⁶ The first-level indicators include knowledge creation, knowledge acquisition, enterprise innovation, innovation environment, and innovation performance. The data come from the public data of government departments, such as the National Bureau of Statistics, and the comprehensive score of regional innovation ability is obtained by using the step-by-step weighted utility value method. We determine a region’s stage of development by comparing its innovation score with the national average. If the score is above the mean plus half a standard deviation, it is an innovation leader. If the score is less than half a standard deviation from the mean, it is an innovation catching-up region.

4.3.3 Control variables

4.3.3.1 FDI.

It is measured by the ratio of the actual utilization of FDI in each province to the regional GDP, reflecting the impact on capital investment and foreign technology spillovers for each province (Driffield and Love 2007; Ning et al. 2016). FDI can bring abundant funds to regional innovation and can provide a vital source of external knowledge and technology (Wang et al. 2016), affecting regional innovation through knowledge spillover effects.

4.3.3.2 The regional Openness (OPEN) level.

It is measured by the ratio of the total volume of imports and exports to regional GDP. There will be more foreign trade activities in open areas, involving transfers and exchanges of various knowledge and technologies and showing technology spillover effects (Kim et al. 2016) on RIE.

4.3.3.3 Information infrastructure construction (INFR).

It is measured by the ratio of the total volume of post- and telecommunications business to regional GDP. INFR is a foundation and essential platform for knowledge exchange and transfers among enterprises, universities, and scientific research institutions. Additionally, INFR is conducive to promoting collaborative innovation, knowledge transfer, and spillovers of various subjects in the region, thereby improving RIE.

4.3.3.4 Fixed-asset investment (FAI).

It is measured by the proportion of the total FAI completed in each province to the regional product’s total value. While investment promotes economic development, it also affects RIE through technology transfers and spillovers.

4.3.3.5 Institutional environment (INS).

It is measured by the marketization process index (Fan et al. 2019). The INS is a critical variable that affects enterprises’ innovation decision-making and behaviours (Peng 2002). In the process of market reform, the INS’s development is insufficient and unbalanced between regions (Wang et al. 2015). It mainly impacts resource allocation and corporate decision-making, and then it impacts RIE (Barasa et al. 2017).

4.3.3.6 Regional industrial diversification (DIV).

It uses the improved Herfindahl index to measure regional industrial specialization agglomeration (Ning et al. 2016). Formed by the agglomeration of different industries, knowledge spillovers and technological externalities contribute to the fusion and collision of complementary knowledge and promote regional innovation (Glaeser et al. 1992).

4.3.3.7 Labour mobility (LM).

It is measured by the rate of change in the urban employment population. In a country with high cross-province LM, such as China, R&D investments in innovation-leading regions spill over and positively affect the RIE of innovation-catching-up regions. Therefore, we use changes in the employed population to control for this spillover effect.

4.3.3.8 Industrial upgrading (IU).

It is measured by the ratio of the output value of the tertiary industry and the secondary industry. With the development of the economy and the improvement of the per capita national income level, the evolution trend of labour first transferred from the primary industry to the secondary industry and then to the tertiary industry. The ratio of the output value of the tertiary industry and the secondary industry reflects the situation of IU.

The names, symbols, and meanings of the variables are summarized in Table 2.

4.4 Empirical specification and estimation methodology

This paper constructs the following baseline model testing the influence of public direct R&D grants, R&D tax credits, and IPRs on RIE:

where CONTROL means the control variables of this article, including regional openness, infrastructure construction, R&D human capital, FAI, regional absorptive capacity, INS, and regional DIV. As the moderate variable is categorical in testing regional development, this paper adopts group testing.

5. Empirical results

5.1 Descriptive statistics and correlation analysis

The descriptive statistics and correlation analysis are shown in Table 3. There is a significant negative correlation between public direct R&D grants and RIE, while the relationship is not significant between the R&D tax credit and RIE, nor is it significant between IPRs and RIE. The variance inflation factor (VIF) of each variable is less than 10, and the average value of VIF is less than 5. There is no severe collinearity problem among variables. To further eliminate collinearity, the variables are centralized in the process of constructing the interaction terms.

5.2 Hypothesis testing

5.2.1 Baseline model regression

Based on Model 1 in the Table 4, we examine the impact of government R&D grants, R&D tax credits, and IPRs on RIE and consider truncated data with RIE values between 0 and 1. The estimation result of the ordinary least square (OLS) method cannot guarantee unbiasedness and consistency. This paper chooses the Tobit model for testing. The model’s likelihood ratio (LR) test result is significant, so we choose the random effect panel Tobit in this study. The results of Model 1 to Model 3 are shown in Table 4. Model 4 is the full model that contains all the independent variables.

In Model 1, direct grants have a significantly negative impact on RIE at the 5 per cent level, supporting Hypothesis 1b. The results from Model 2 support Hypothesis 2a, showing that the R&D tax credit has a significant and positive impact on RIE at the 5 per cent level. The results from Model 3 show that IPRs have a significant and negative impact on RIE at the 5 per cent level. The empirical results are contrary to Hypothesis 3.

5.3 Results in different development stage regions

These data are divided into two groups based on the ‘China Regional Innovation Capability Evaluation Report’ to test the various impacts of public direct R&D grants, R&D tax credits, and IPRs on RIE over the development stage. The results in the two groups are shown in Table 5.

Comparing Model 1 with Model 4 in Table 5, it is found that public direct R&D grants have a significantly negative effect on RIE in both innovation-leading regions and innovation-catching-up regions. Furthermore, the Chow test results show that the influence coefficients of public direct R&D grants in innovation-leading regions are substantially different from those in innovation-catching-up regions. The negative effect of direct grants in innovation-leading regions is weaker. Comparing Model 2 with Model 5, it is found that tax credits have a negative and significant effect on RIE in innovation-leading regions, while tax credits have a positive and significant effect on RIE in innovation-catching-up regions. Hypothesis 4b is supported. Comparing Model 3 with Model 6, it is found that IPRs have a significantly positive impact on RIE in the innovation-leading region. In contrast, IPRs show a significantly negative impact on RIE in the innovation-catching-up region. Thus, Hypothesis 4c is supported.

5.4 Endogeneity test

The main endogeneity sources include measurement bias, omitted variable bias, and mutual cause and effect. To solve the possible endogenous problems, this paper uses residual regression and IV-Tobit regression to test and control.

5.4.1 Residual regression

This article may have a few errors in measuring public direct R&D grants, R&D tax credits, and IPRs. To test the possible measurement errors, we use two-stage error regression to test endogeneity (Terza 2017). First, we define public direct R&D grants, R&D tax credits, and IPRs as the dependent variables and then take FDI, local openness, infrastructure construction, R&D human capital, FAI, regional absorptive capacity, and other control variables as independent variables. Regression is performed to control related variables’ interference, extract model residuals, and use residuals to replace three independent variables. Later, we repeat hypothesis tests. The results are consistent with the above hypothesis test results and are shown in Table 6.

5.4.2 IV regression

RIE is a comprehensive measurement of a region’s innovation input and output efficiency and is affected by many factors. This may lead to ‘self-selection’ issues among public direct R&D grants, R&D tax credits, IPRs, and RIE.

First, in regions with higher innovation efficiency, the government has more sufficient funds to subsidize enterprises. This paper uses the lag term of public direct grants as an instrumental variable to examine the endogenous problems of public direct R&D grants.

Second, in regions with high innovation efficiency, companies have greater R&D capabilities and R&D investment, and thus, they obtain more R&D tax credits. This paper selects the number of high-tech enterprises and the total internal expenditure on R&D invested by high-tech enterprises. R&D tax credit policy is mainly aimed at encouraging enterprise innovation, and high-tech enterprises are the main body of innovation in China. The more R&D expenditures there are, the more the R&D tax may be deducted.

Third, a higher RIE promotes the local government’s enforcement of IPRs, cooperation of IPR awareness, and promotion of IPR improvement. To solve the endogenous IPRs problem, we use historical instrumental variables and geographic instrumental variables (Ang et al. 2014). This article sets a dummy variable based on British concessions in the late Qing Dynasty¹⁷ as an instrumental variable for IPR protection. The value is 1 if a city in a province was a British concession and 0 otherwise. There are three key considerations for choosing this instrumental variable: (1) the UK was one of the first countries to pay attention to IPR protection and promulgate-related laws. In the British concession, the British will manage according to their laws and regulations. Moreover, to a certain extent, it enhances the awareness of local people’s IPRs. (2) China’s modern legal process started in the concessions and expanded across the country through a point-to-face model. (3) Provinces with British colonies have no evident impact on their innovation efficiency after dynasties and historical changes. The geographical distance from the province where the British concessions are located will not directly affect innovation efficiency.

The results of the IV-Tobit and two-stage OLS test are shown in Table 7. The results suggested that instrument variables in each model are valid. The results of the two-stage OLS are consistent with the hypothesis test results.

5.5 Robustness test

5.5.1 Test use lagged variables

To test the reliability of the results, we conduct a robustness test. First, considering that there may be a certain lag among public direct R&D grants, R&D tax credits, and IPRs and innovation efficiency, we use lagging regression. That is, we analyse the influence of independent variables in year t on dependent variables in year t + 1, and the results are shown in Table 8. The robustness results are consistent with the previous hypothesis tests, and our analysis achieved robustness to a certain degree.

5.5.2 Test use alternative variables

Alternative variables are also used to test the robustness. We use the proportion of public direct R&D grants to R&D expenditures to measure direct grants, which is consistent with the denominator of the tax credit measurement. In China, the R&D tax credit policy is uniformly regulated at the national level, but local governments implement it differently, so we set a dummy variable for the tax credits to test the robustness. We first calculate the proportion of the R&D tax credit to local R&D expenditures. If the proportion of tax credit in R&D expenditure is higher than the national average, then the tax credit is considered to be better implemented and recorded as 1. Otherwise, it is 0. We set a dummy variable based on China’s Intellectual Property Demonstration City Program as the alternative to measuring IPR protection. To promote the construction of a powerful country with IPRs and give full play to the important role of IPRs in driving urban innovation and development, China National Intellectual Property Administration (CNIPA)¹⁸ has formulated the ‘National Intellectual Property Pilot and Demonstration Cities Evaluation Measures’. The CNIPA evaluates the city’s IPR protection based on multiple dimensions, such as resource input and patent output. Only cities whose IPR protection is higher than others can be selected. Since the first batch of national intellectual property demonstration cities is selected in 2012, the CNIPA has selected four batches of sixty-four national intellectual property demonstration cities, which are distributed in twenty-two provinces. In a given year, if there is more than one intellectual property demonstration city, the dummy variable takes a value of 1; otherwise, it takes a value of 0. If there is more than one intellectual property demonstration city in the province, the dummy variable takes a value of 1; otherwise, it takes a value of 0. We redo the hypothesis test using alternative measures. Since the intellectual property demonstration programme started in 2012, we used the data after 2012 to test the effect of IPR protection. The results are shown in Table 9.

To test the robustness of regional heterogeneity, we use the regional R&D investment intensity (R&D investment/regional GDP) to group regions. When the regional R&D investment intensity is greater than the national average, it is an innovation-leading region. Otherwise, it is an innovation-catching-up region. The results are consistent with the previous hypothesis tests, and the analysis in this paper has certain robustness. The results are listed in Table 10.

5.5.3 Test use subsample

Considering the agglomeration effects of municipalities, we excluded four provincial-status municipalities from the sample for the robustness test. The results are listed in Table 11. The results in Table 11 are consistent with the previous hypothesis tests, and the analysis in this paper has certain robustness.

6. Discussion and conclusion

This study explores the impact of three innovation support policies of the Chinese MLP, including public direct R&D grants, R&D tax credits, and IPR protection, on RIE. The empirical results indicate that, in general, public direct R&D grants have a significant negative effect on RIE, and direct grants in innovation-leading regions do not promote innovation efficiency better than in catching-up regions. R&D tax credits have a significant positive effect, but the positive effect mainly occurs in innovation-catching-up regions. R&D tax credits exert a negative effect on RIE in innovation-leading regions. As an indirect innovation support policy, IPRs protection has an overall significant negative effect on innovation efficiency. However, after distinguishing the different regions, IPRs have this negative effect mainly in innovation-catching-up regions but facilitate innovation efficiency in innovation-leading regions.

Direct R&D subsidy policies that include direct grants and tax credits are less effective in innovation-leading regions. This is probably mainly because enterprises in innovation-leading regions choose defensive R&D to maintain their market position (Bloom et al. 2020). After establishing a competitive advantage through innovation and obtaining huge economic returns, they can make strategies and invest in their R&D and innovation activity upon their own needs rather than spurred by external forces. In particular, innovation support policies may lead to a decline in research productivity when additional R&D projects sponsored by direct grants have lower economic returns than privately funded projects (Bloom et al. 2020; Boeing and Hünermund 2020). From the perspective of enterprises, government-funded R&D projects are usually overambitious because policymakers are not omniscient (Haapanen et al. 2014) and are economically inferior to market options. As a result, such explicitly mission-driven innovation policies can be more detrimental to innovation efficiency, especially as the desire to leapfrog and move to entirely new products and technologies can carry significant opportunity costs (Boeing and Hünermund 2020).

Under the support of tax credits, enterprises have higher autonomy to decide the composition of their R&D. For reasons of economic returns, tax credits can undermine enterprises’ R&D investment in exploratory cutting-edge technologies (Hall and Van Reenen 2000), thus inhibiting innovation efficiency in innovation-leading regions. However, tax incentives can be more effective when innovation-catching-up regions are mainly imitating and catching up with mature technologies for incremental innovation. Such innovation models with low risk and fast returns appear to be more efficient. In addition, during the MLP period, the Chinese government adjusted the tax credits policies continuously. For example, tax credit policies have expanded the types of sponsored organizations, gradually increased the proportion of R&D investments enjoying tax credits, and expanded the scope of investments eligible for tax credits. These initiatives are becoming increasingly inclusive and can especially further benefit the participation of enterprises with inherently less technological capabilities in R&D and enhance RIE in innovation-catching-up regions.

IPRs, as an indirect innovation support policy, need to maintain a reasonable balance between the static losses and dynamic gains caused by a monopoly (Allred and Park 2007). For enterprises in innovation-leading regions with higher technological capabilities, better IPR protection helps them capture greater monopoly benefits based on technological advantages. However, for enterprises in catching-up regions, weaker technological capabilities make it difficult for these enterprises to gain benefits from IPRs. At the same time, IPR protection blocks knowledge spillover, raises the cost of knowledge sharing, and undermines the efficiency of imitative incremental secondary innovation that innovation-catching-up regions rely more upon.

This study contributes to extant studies from three aspects. First, this study further explores how different types of innovation policies affect RIE in emerging economies, which are usually technological catching up. Specifically, this study examines the effects of direct R&D grants, tax credits, and IPRs on RIE. Our results indicate that tax credits may be more effective on RIE in the areas at the technological catching-up stage compared with direct grants. These findings further supplement the conclusion of the study by Broekel (2015). Our results also show the regional heterogeneity of IPRs in the Chinese context and extend the research findings of Allred and Park (2007) that the heterogeneous effect of IPRs exists not only at the national level but also at the regional level.

Second, this paper expands the theory of regional innovation systems from the perspective of innovation efficiency. As an essential part of the national innovation system, the regional innovation system emphasizes the relationship and innovation ability among different innovation subjects in the region (Cooke et al. 1997; Liu and White 2001). The existing research ignores the research on the efficiency of regional innovation systems. Innovation efficiency emphasizes not only innovation output but also resource allocation efficiency (Guan and Chen 2012). This paper enriches the research content of regional innovation systems by analysing the influence of S&T policy on RIE. Furthermore, this study enriches the relevant research on regional innovation systems by distinguishing innovation-leading and innovation-catching-up regions. Specifically, innovation-leading and innovation-catching-up regions have different technological capabilities, INSs, and resource endowments.

Third, this study systematically tested the effectiveness of the Chinese government’s innovation policies in MLP at the regional level. Previous literature introduces the background and content of MLP (Chen and Naughton 2016; Gu et al. 2008) and evaluates the impact of MLP (Appelbaum et al. 2016). However, the MLP is a very large and complex long-term programme, and the Chinese government has not summed up the achievements of this programme although the programme has ended. In our study, we evaluated the effectiveness of several specific representative innovation support policies in the MLP through empirical analysis and explored the process mechanism of China’s indigenous innovation.

This study also provides several policy implications. The MLP’s mixed innovation policy is effective in promoting regional innovation. However, several policy designs can be further improved. First, the government should flexibly use the policy subsidy tool according to the level of regional development. Not all regions need R&D subsidies, and in innovation-leading regions, there should be more emphasis on the role of markets and less government intervention. Governments should make more use of universal and non-competitive subsidies, such as R&D tax credits. Second, governments at all levels should create a sound INS. With the continuous enhancement of regional innovation capabilities, it is necessary to further enhance IPR protection and protect knowledge producers’ profitability, encouraging them to invest in research and development. Third, the government should closely interact with enterprises in the region to eliminate information asymmetry. In particular, they need to carry out industry–university–research cooperation and major national projects. Governments, at all levels, should carefully select cooperative enterprises and avoid rent-seeking behaviours of enterprises.

There are some limitations and deficiencies in this research. First, we used invention patent application as the output variable of RIE without considering the patent quality, such as family size and citation measures. In future studies, we will further focus on the analysis of patent quality. Meanwhile, on 27 January 2021, the State Intellectual Property Office issued the ‘Notice on Further Strictly Regulating the Behaviour of Patent Application’,¹⁹ emphasizing the transformation of patent applications from pursuing quantity to improving quality, which also provides an opportunity for patent quality analysis. Second, a concern with RIE is that it obscures the underlying direct mechanism through which a given policy may affect the related outcome, e.g. public R&D grants on R&D expenditures. Therefore, it may be worthwhile to show the policy impact on the first-order target variables of the respective policy. In future research, we will examine the impact of three S&T policies on the first-order target variables and compare them with the results of this study. Another limitation is the data source. Most of the data are derived from official statistical yearbooks in China, which might contain some biases and imprecisions, especially concerning innovation-related indicators. Although in the process of empirical research we have performed some processing of the data, it still might be inflated for some provinces.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No.72104121, 71932009, and 71872170).

Conflict of interest statement.

None declared.

Data Availability

We are currently unable to provide the data freely due to the requirements of the research project. If you are interested in the data sets, please do not hesitate to contact us.

Endnotes

References