Does the high-tech enterprise certification policy promote innovation in China?

This study investigates the impacts of the high-tech enterprise certification policy on enterprise innovation in China by analyzing the unique dataset of listed companies and their affiliates from 2006 to 2015. We exclude firms certified after year 2009 from the sample, because they may have engaged in R&D manipulation. The results show that the high-tech enterprise certification can promote enterprise innovation in China, especially the innovation captured by invention patents. The results of a rich set of robustness tests all support this conclusion. The study also concludes that the high-tech enterprise certification can influence enterprise innovation through “direct” and “indirect” channels. Heterogeneity analysis shows that private enterprises, enterprises in more competitive industries, and equity-inspired enterprises benefit most from high-tech enterprise certification. This paper helps comprehensively understand the validity of the innovation policy and the driving forces of enterprise innovation in China. JEL O31 O32 O38


Introduction
In order to promote the transformation of economic structure, China has gradually implemented an innovation-driven development strategy. Various innovation policies have emerged one after another, with significant resources being invested in the field of scientific and technological innovation. Meanwhile, the number of patent applications has continued to grow, making China the world's largest patent applicant in 2011. As the main power of innovation, Chinese high-tech enterprises (HTEs) are increasingly gaining attention. Aimed at supporting HTEs' development, the Chinese Ministry of Science and Technology, the Chinese Ministry of Finance, and the Chinese National Tax Bureau jointly promulgated 'National High-Tech Enterprise Certification Management Measures' (hereinafter referred to as 'Certification Measures') and 'High-Tech Enterprise Certification Management Working Guidelines' (hereinafter referred to as 'Working Guidelines') on April 14, 2008, which for the first time made the work of HTE certification fully carried out nationwide. On January 29, 2016, the 'Certification Measures' were revised to enlarge the scope of certification, particularly, relevant standards were tilted toward small businesses for the purpose of encouraging them to innovate more.
Promotion of self-driving innovation capability at national and corporate levels along with transformation of government functions are among the key concerns for scholars and policy makers. In this study, we mainly attempt to find out how HTE certification policy drives Chinses enterprises' participation in technological innovation. All in all, whether the government's 'visible hand' really promotes enterprise innovation is an important question that cannot be answered unambiguously yet. Existing theories have put forward two opposite views on this. One view suggests that innovation activities are highly risky, while government support can make up for market failures and pump large amounts of resources into companies, thereby stimulating corporate innovation behavior (Arrow 1972;Romer 1990;David 2000). Another view argues that government cannot effectively allocate resources. Instead, its direct intervention, distortion of competition, and selective support can even inhibit corporate innovation (Michael and Pearce 2009;. To our knowledge, the current literature mostly focuses on a single industrial policy or policy instrument when exploring this subject. So far, very few analyses have been conducted on HTE certification, a more synthetical policy in the Chinese context. Given this, we take the 'Certification Measures' of 2008 as the background and use the specific data of Chinese listed companies' and their affiliates' patent applications to conduct an in-depth research. The possible contributions of our paper are as follows. First, there has been no consistency in arguments about the influence of government support on enterprise innovation, and the fact that Chinese government is employing HTE certification policy serves as an interesting case for reexamining this inconclusive issue. Second, the scarce literature related to Chinese 'Certification Measures' either concentrates on stock market reaction and earnings management (Xu and Zheng 2016;Yang et al. 2017), or has limited dataset and neglects R&D manipulation problem (Xu 2017). On this basis, we focus on 'real' HTEs' innovation behavior after excluding companies engaged in R&D manipulation, and investigate more comprehensive samples to provide micro-level empirical evidence for the disputes over the influence of government support on corporate innovation. Third, according to , Chinese listed companies may transfer their innovation activities to affiliated companies. Therefore, it is not enough to consider only the number of listed companies' patents when measuring innovation outputs. We thus take the sample companies' subsidiaries, associates and joint ventures into account when collect our data. Fourth, in order to effectively address possible identification concerns, we also employ Heckman's two-step method, PSM-DID method, and other robustness tests. Plus, the policy's internal influence channels and corporate heterogeneity effects are analyzed, which deepens the understanding of mechanisms' transmission from macroeconomic policies to micro-market entities.
The remainder of this paper proceeds as follows. Section 2 presents the theoretical analyses and research hypotheses based on the policy background and relevant literature. Section 3 provides the information on sample selection, variable definitions, and summary statistics. Section 4 reports the main empirical results. Section 5 concludes and gives the policy recommendations.

Theoretical Analyses and Research Hypotheses
Innovation investment is a significant part of long-term strategy, directly affecting companies' future profitability and even the whole country's competitive advantages (Manso 2011). However, unlike ordinary investment, investment in R&D is characterized by positive externalities, high risks, high uncertainties, yield lags (Arrow 1972;Holmstrom 1989;Dosi et al. 2006), and high adjustment costs (Hall 2004), which leads to firms' lack of innovation. Under these circumstances, government departments began to engage in private enterprises' innovation activities by using financial measures as well as administrative control (Kang and Park 2012;Rao 2016). The HTE certification policy is an attempt of the Chinese government to address the challenge of boosting companies' technological innovation through the following channels.

Tangible Channel
At this stage, the Chinese market mechanism is not mature enough, and the government still controls the allocation of important resources required for enterprises' survival and development. HTE certification can ensure that enterprises have sufficient tangible support, which smooths R&D expenditure path and decentralizes the risk of corporate innovation activities.
Generally speaking, certified HTEs will obtain direct or indirect economic benefits on the three levels mentioned below. The first level involves national unified polices. It is stated in the 'Certification Measures' that in accordance with the 'Enterprise Income Tax Law' and its 'Implementation Regulations', the 'Law of the People's Republic of China on the Administration of Tax Collection' and its 'Implementation Regulations', enterprises that have obtained HTE certificates are entitled to such preferential taxation policies as 15% preferential income tax rates, R&D expenses deductions, and deductible taxes for energy-saving and environmental-friendly equipment. The reduction of tax burden cuts down enterprises' cash outflow to a certain extent, which improves their intrinsic capacity of financing innovation activities (Duchin et al. 2010).
The second level involves local policies. Certified HTEs can also enjoy various rewarding policies provided by local governments, such as government subsidies, easy financing approval, land lease preferences. As a compensation for pioneer companies trying to innovate, government subsidies directly 'transfer' some economic benefits to micro market players, which reduces corporate innovation costs and increases the funds available for research and development (Tether 2002;Chen et al. 2014).
Finally, HTE certificates are jointly awarded by the three government departments, which makes it possible to authoritatively prove enterprises' scientific and technological stance. So HTEs can use these 'brands' to quickly enhance their reputation, which is conducive to accessing innovative resources from other sources, thereby enhancing innovative capacities. H1: HTE certification can promote enterprise innovation through the tangible channel.

Intangible Channel
Enterprise innovation as an investment decision essentially depends on companies' power structure. In many cases, the failure of a corporate entity to innovate is not due to the absence of favorable conditions, but the lack of motivation (Frenkel 2000). For instance, the separation of listed companies' ownership and control power will lead to the 'principal-agent problem'. Corporate executives tend to concentrate on their own interests and prefer avoiding risks due to their individual wealth relying on a single company (Bertrand and Mullainathan 2003), which is inconsistent with corporate long-term development. HTE certification also provides intangible benefits, spurring enterprises to engage in more innovative activities.
First, the acceptance of innovation in corporate culture can greatly influence innovation behavior (Deshpande et al. 1993), and the stronger the emphasis on innovation, the more resources will be allocated for innovative activities (Hurley and Hult 1998). HTE certification policy helps corporate executives better understand the true value of innovation activities, thus reducing their short-sightedness and stimulating their innovation enthusiasm. Besides, the mandatory provisions under the 'Certification Measures' also have a deterrent effect on HTEs, making the senior executives implement innovation projects for achieving policy goals.
Second, HTE certification requires enterprises to enhance disclosure of relevant information through jointly established information disclosure mechanisms in various departments, which fosters creative behavior by making it easier for companies to attract the attention of angel investors, institutional investors, and even news media.
Finally, HTE certification improves corporate human capital. HTEs are generally more capable of retaining existing scientific research talents as well as attracting more management personnel with higher professional skills, who are more likely to support independent innovation activities (Holmstrom 1989).
H2: HTE certification can promote corporate innovation through the intangible channel.

The Impacts of Corporate Heterogeneity on the Effectiveness of HTE Certification
HTE certification under the 'Certification Measures' of 2008 is a policy for all enterprises across China. However, due to the existence of corporate heterogeneity, its effectiveness may exhibit discrepancy among different companies. We mainly discuss the influence of ownership, corporate governance structure, and market competition.
First, we analyze firms' ownership. Private enterprises face more resource constraints than state-owned enterprises. For example, state-owned commercial banks that monopolized credit market have a natural 'financial discrimination' against private enterprises. Also, owing to the imperfection of China's related systems, there is a more serious information asymmetry problem between private enterprises and external investors. These problems faced by private enterprises can be effectively mitigated with the help of HTE certificates, which will drive the development of private enterprises' innovation activities. In contrast, state-owned enterprises have more abundant original resources (Greve 2003), but due to the 'resource curse' 1 their technological innovation can be insufficient. Besides, state-owned enterprises often have more social responsibilities, which distorts their business objectives. Hence, compared with state-owned enterprises that are not in urgent need to obtain resources and are less efficient in transforming resources into innovative results (Carman and Dominguez 2001), private enterprises can enjoy greater positive effects of HTE certification on their innovation performance.
Second, we study corporate governance structure. Compared with traditional performance-based compensation incentive plans, the implementation of equity incentive plans can effectively align the interests of management with the interests of shareholders (Wu and Tu 2007), thus reducing the adverse effects of principal-agent problems, helping to prevent corporate managers from 'enjoying a calm life' (Bertrand and Mullainathan 2003), and increasing corporate risk-taking level (John et al. 2008;Atanassov 2013). Therefore, as an innovation-oriented corporate governance structure (Jensen and Murphy 1990), the long-term incentives can provide management with opportunities to share corporate profits and motivate them to focus on corporate technological innovation (Wu and Tu 2007;Armstrong et al. 2013).
Finally, we discuss market competition. Arrow (1972) believes that corporate innovation incentives under competitive conditions are significantly higher than those under monopolistic conditions. The 'natural laws' of survival of the fittest in market competition will drive out inefficient companies. Therefore, if companies want to survive and maintain a large market share, they have to continuously carry out innovation activities and accelerate commercialization of innovation results. Some empirical studies have already proven that moderate market competition encourages enterprises to extensively upgrade products, services, and technologies (Jaffe 1988;Zucker and Darby 2007). We believe that enterprises in the environment with relatively high market competition have stronger innovation driving force, and for them HTE certification will have a better innovation promotion effect.
H3: Private enterprises, equity-inspired enterprises, and enterprises in industries with higher competition can enjoy greater positive effects of HTE certification on their innovation performance compared with state-owned enterprises, non-equity-inspired enterprises, and enterprises in industries with lower competition.

The Influence of R&D Manipulation on the Effectiveness of HTE Certification
It should not be overlooked that HTE certification is based on the ex-ante information delivered by enterprises, so there exists the same possibility of rent-seeking as for many other government-supported initiatives. That is, in order to meet the policy requirements, companies are prone to some 'support-oriented' adverse selection behavior. For example, companies may engage in 'pseudo' research and development to reach certification standards, or carry out 'strategic' innovation 2 to gain relevant policy preferences (Li and Zheng 2016;Yang et al. 2017;Chen et al. 2018). If these 'pseudo' HTEs can be effectively identified and excluded from our sample, the remaining 'real' HTEs should have better innovation performance.
H4: After excluding the companies engaged in R&D manipulation, we can observe that HTE certification has a greater positive impact on corporate innovation.

Sample Selection
The data used in this research mainly comes from Chinese CSMAR database, Wind database, CNINFO (http://www.cninfo.com.cn/cninfo-new/index) and the network of HTE certification management (http://www.innocom.gov.cn/). Some data on HTE certification and R&D is manually collected from the HTE certification announcements and listed companies' annual reports. Our sample includes all the listed companies traded on Shanghai and Shenzhen Stock Exchanges from 2006 to 2015. The reason why the sample period starts from 2006 is that a new Chinese accounting standard was adopted in 2006. Since then, R&D input data of Chinese listed companies has been disclosed more systematically. Besides, we take the 'Certification Measures' of 2008 as the background, so the sample period is up to 2016 when the 'Certification Measures' was amended.
We treated the original data in the following ways. First, we excluded all financial companies and companies that have suffered losses for two or more consecutive fiscal years. Second, we excluded clearly unreasonable sample observations and made up for some missing data. Third, we performed 1% winsorize processing at the beginning and at the end of all continuous variables to eliminate the influence of extreme values. Eventually we obtained 15,825 sample observations for 2,409 companies. Among them, 1,188 companies issued the project information on HTE certification. They were defined as HTEs in this paper.

Definition of Variables
Innovation level of enterprise (LnPat & LnInv). In contrast with R&D investment, innovation output can directly reflect a company's innovation ability (Aghion et al. 2005;Hagedoorn and Wang 2012). We take the number of the sample companies' and their affiliates' total patent applications as a measure of corporate innovation. According to the 'Patent Law of the People's Republic of China', patents include inventions, utility models, and designs. Among them, inventions have the highest originality, thus they can better represent the level of enterprise innovation and are considered 'substantial' corporate innovations (Tan et al. 2015). We use the number of corporate invention patent applications as an auxiliary measure of corporate innovation. We take natural logarithm of (one plus) the raw patent data to construct the measure of main innovation level according to the usual practice.
HTE certification (Tec). HTE is defined in the 2008 'Certification Method' as follows: 'An enterprise implementing innovative activities in 'High-Tech Fields Supported by the State' in order to form its core independent intellectual property rights and use the rights as a basis to carry out business activities. The enterprise should be a resident company registered in China (excluding Hong Kong, Macao and Taiwan) for more than one year.' We here introduce a dummy variable of HTE, which, if a company obtains the certificate in a given year, has a value of 1, otherwise, is equal to 0.
Control variables. Referring to current empirical studies on enterprise innovation, such as He and Tian (2013) The detailed definition of these key variables is shown in Table 1.  Table 2 shows the summary statistics for major variables. We report on companies that have obtained HTE certificates, and those have not obtained HTE certificate respectively. As can be seen from the table, the mean of the number of total patent applications (invention patent applications) for sample companies is greater than its median, regardless of whether those companies are certified as HTEs, that is, the number of patent applications shows a clear right-aligned nature. Besides, the standard deviation is very large, which demonstrates that Chinese companies' innovative capabilities are uneven. On average, companies with HTE certificates in the sample are more innovation-intensive activities and have higher number of 'substantive' innovations: certified companies applied for 51.77 patents per year, of which 24.15 were inventions, but non-certified companies applied for 43.26 patents each year, of which 19.30 were inventions. The characteristics of other control variables are similar to those of previous studies and are not repeated here. Figure 1 shows certified companies' number of total patent applications 3 adjusted by industry and time before and after certification. We generated this figure by subtracting the average number of patent applications for all companies in a certain industry from the number of patent applications filed by certified companies each year, and then averaging the differences. 0 point on the horizontal axis represents the year when a company obtained an HTE certificate for the first time, 1 (-1) point is the first year after (before) the certification year, other abscissas are analogous. Before certification, the adjusted level of corporate patent applications was low and did not exhibit any obvious time trend. However, after HTE certification, the volume of companies' patent applications has maintained a significant upward trend at a relatively high level.

Figure 1. Patent applications adjusted by industry and time before and after certification
Simple graphic analysis cannot fully portray the relationship between HTE certification and corporate innovation, so in order to study it, we established a benchmark model.
The explained variable LnPat (LnInv ) is a natural logarithm of one plus the number of total patent applications (invention patent applications) for enterprise i in year t, reflecting the level of innovation. The main explanatory variable is . If enterprise i has obtained an HTE certificate, equals 1, otherwise it is 0. is a vector of control variables. We also control for year fixed effects ∑Year, industry fixed effects ∑Ind and province fixed effects ∑Prov, is the residual error. Since the number of patents cannot be negative, we adopted Tobit estimation method 4 in the study.
To test Hypothesis 4, we first obtained a sub-sample by excluding companies certified as HTEs after 2009. Then we successively regressed the full sample and sub-sample. Results are shown in Table 3. 4 Unless specified otherwise, simple regressions using the patent data as an explained variable all adopted Tobit estimation method in this paper.
Industry-Year Adjusted Patent Notes: The first and second columns provide results for the full sample. The third and fourth columns provide results for the sub-sample which has excluded HTEs certified after 2009. ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.
As can be seen from table 3, first, the coefficients of Tec are all significant and positive，so we can initially assume that HTE certification has improved the average level of corporate innovation. Second, the coefficients' values of the sub-sample are greater than those of the full sample, which can be explained as follows. According to Yang (2017), companies under tax incentives may manipulate R&D investment to reach to the HTE certification yardstick, and companies involved in R&D manipulation usually have poor innovative performance since they are just seeking policy benefits. We noticed a series of quantitative conditions for HTE certification that were aimed at the overall performance of enterprises in the past three fiscal years. Therefore, there are reasons to believe that for companies with R&D manipulation it was impossible to be certified before 2009 5 . When we used the full sample, the existence of these profit-driven companies hindered the appearance of HTE certification policy's positive effects. In the above subsample we excluded companies certified after 2009, therefore excluding companies possibly involved in R&D manipulation. The Tec coefficients are now larger, and Hypothesis 4 is verified. In order to ensure the validity of the subsequent analyses, they will be based on the subsample.
The coefficients of control variables indicate that companies with larger size, younger age, stronger product market competition, higher leverage ratios, higher institutional investors' shareholding ratios and lower operating income growth rates will have more patent applications. In addition, the corporate structure with integration of chair and general manager positions is also conducive to corporate innovation.

Analysis Based on Heckman Two-step Method
Considering the possible sample selection bias, we used Heckman two-step method to do further research. The specific process can be divided into two stages. The first stage was constructing a selection model. We used Probit model to estimate the probability of a company obtaining an HTE certificate and then constructed the Inverse Mills Ratio parameter Lambda. In the second stage, the Lambda parameter was added as an additional explanatory variable to our previous benchmark influence model. For effective identification, Heckman two-stage model requires the selection model containing at least one exclusive variable, so we need to find a variable that will determine whether a company obtains an HTE certificate but will not directly affect company's innovation level. The findings of Guo et al. (2016) suggest that the total number of firms in high-tech zones of the cities where HTEs are located in each given year can be used as an instrument variable to identify the probability of an 5 According to Yang Guochao et al. (2017), it is not difficult to track down the traces of companies' R&D manipulation. For instance, Nanling Civil Explosion company announced that it obtained the HTE certificate on May 26, 2011. Then it was found that in 2009, 2010, and 2011, the company's R&D investment accounted for 3% of the sales revenue for the year, which is exactly the regulatory threshold of 'Certification Measures' promulgated in 2008. enterprise acquiring government R&D subsidies. Similarly, for our model, we selected the total number of companies that have obtained HTE certificates in the given company's home province to be used in the selection model. This variable choice is mainly based on our understanding of the 'Certification Measures'. Accordingly, the science and administrative departments of provinces, autonomous regions, directly administered municipalities, and municipalities with independent planning status together with the financial and taxation departments at the same level constitute the administrative organs that carry out the certification work within their administrative areas. Therefore, the more companies are certified as HTEs in a certain province, the better the certification work in the region is carried out. Due to path dependence, there can be a certain basis for local enterprises' applications for HTE certificates, which can influence whether companies in the area can be certified. However, the number of companies that have been certified in a province is a macro-variable compared with enterprises' own attributes, and it is not an important factor affecting a company's innovation activities. Table 4 reports the results of this method's second step. Notes: Control are all the control variables, for simplicity, we will not report their estimation results anymore. ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.
From the table, we can see that coefficients of Lambda and Tec are both significant and positive. It proves that there exists an issue of sample selection. In addition, after mitigating this bias, we find that obtaining HTE certificates still significantly increases companies' number of patent applications.

Analysis Based on PSM-DID Method
Apart from the selection bias, there may also be endogenous problems caused by the mixed bias and reverse causality. We subsequently used PSM-DID method to better control these issues. The basic idea is that if a company which will obtain an HTE certificate before its certification is completely similar to a counterpart that will not be certified, then, the only factor leading to their difference in innovation level is whether the company is certified as an HTE.
The first step is Propensity Score Matching (PSM), which provides a feasible strategy for match by turning high-dimensional corporate characteristics into propensity scores. We matched the sample treatment group and control group in the same year and in the same industry according to important corporate characteristic variables, including the number of accumulated patent applications (LnPats) and the growth rate of the total patent applications (PatGrowth 6 ). Figure 2 shows the comparison of score density before and after matching. Overall, our matching quality is satisfying, as the distributions of score density in the two groups are very close after matching.

Figure 2. Propensity Score Matching Effect
The second step is Difference In Difference (DID). We used the matched sample 7 to examine the causal relationship between HTE certification and enterprise innovation, and a general DID model is as follows.
Where × is a dummy variable that captures companies' entering influence. A value of 1 represents the 'enterprise-year' observations in the treatment group after certification, while a value of 0 indicates the 'enterprise-year' observations in the control group or in the treatment group before certification. ∑ represents the control for enterprise individual effects. The definition of other variables is the same as above. Table 5 reports the regression results. Regardless of whether the explained variable is the number of total patent applications or the number of invention patent applications, the coefficients of × are significant and positive, which proves that even after controlling for related endogenous problems, HTE certification's promotion of corporate innovation output remains evident. 6 The cumulative number of patent applications is still natural logarithm of one plus the raw data. The growth rate of the number of total patent applications for a business is the ratio of this year's number to the previous year's reduces one. 7 The matched sample satisfies the common trend condition, which is shown in the appendix.  Notes: ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.

Certification Instances and Corporate Innovation
We also took the amount of times a company is certified (which we further refer to as certification instances) across the sample period to measure HTE certification policy implementation intensity (the more certification instances, the deeper the company is affected by the certification policy) and examined its impact on corporate innovation. Based on the aforementioned theoretical analyses and empirical conclusions, it can be inferred that the impact of innovation will be more obvious with the certification times growing. The specific estimation model is as follows: , −1 is the number of accumulated instances that company i and its subsidiaries, associates, and joint ventures have obtained HTE certificates in year t-1. HTE certificates are valid for three years. After the expiration, the HTE certificates can be renewed multiple times. The definition of other variables is consistent with the previous text. The reason for lagging for a period is to reduce the possibility of reverse causality between certification times and corporate innovation. Even if a business with stronger innovation ability may be able to obtain HTE certificates many times later, its current innovation level cannot affect its certification times in the previous period. It can be seen from Table 6 that the coefficients of , −1 are significant and positive, which once again confirms that HTE certification can promote enterprise innovation. Moreover, with an increase in the number of times that enterprises were certified, that positive correlation also increased. Notes: ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.

Alternative Measure of Innovation
Research based on U.S. data often measures the quality of corporate innovation through patent citations. However, China has not completely disclosed the information on corporate patent citations yet. Referring to the study of Tan et al. (2015), we used the number of granted corporate patents as a measure of patent quality. To be specific, we constructed the variable LnGra (natural logarithm of one plus the number of granted patents in a given year) based on the number of enterprise's patents authorized by the end of 2017, the data update time. In addition, we also used the proportion of invention patent applications (InvRat, invention patent applications/the number of total patent applications) to measure corporate innovation efficiency. Notes: Column 1 and 2 use the sample with alternative measure of innovation. Column 3 and 4 use the sample excluding companies involved in other projects. ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.
From Table 7, we can find that coefficients of Tec are all significant and positive, indicating that HTE certification also enhances the quality and efficiency of corporate innovation. According to the 'Working Guidelines', the HTE evaluation system adopts a scoring system in which the core independent intellectual property rights are the primary target. If a company owns one patent for invention or six intellectual property rights, it can gain the 'A' file (24 to 30 points). That is, HTE certification may make enterprises more inclined to high-level innovations. Our results show that the proportion of invention patents has increased. Again, HTE certification has promoted corporate innovation, especially the innovation of inventions.

Excluding Companies Involved in Other Projects
In addition to HTE certificates, enterprises may also get other types of qualification certificates, such as Technologically Advanced Service Company, Key Enterprises within National Planning and Layout, Enterprise Technology Centers, Innovative Enterprises, the Torch Program, the 863 Program, Comprehensive Utilization of Resources, Technology Business Incubator, Integrated Circuit Design Company, Leading Enterprise, and Emerging Industry Strategic Backbone Enterprise. In order to eliminate this interference, we excluded all the companies that have other certificates except HTE certificates from the sample. Results of the reexamination are also shown in Table 7. The policy effect is still significant and positive and the coefficient sizes are still very close to our previous results, indicating that the main conclusions are not affected.

Mechanism Test and Heterogeneity Analysis
According to our previous theoretical analysis, we believe that HTE certification can influence companies' innovation activities through tangible and intangible channels. We tested them in turn here.

Tangible Channel
The tangible mechanism mainly works through tax preferences, R&D subsidies and bank credits the businesses can obtain. Referring to Li Weian et al. (2016), we used indicator ℎ = × ( 25% − 1)/ to measure the income tax concessions of certified enterprises, where r is the current income tax rate of a certified enterprise and 25% is the uniform corporate income tax rate in China. The deduction of 1 from the ratio of the two is the proportion of preferential tax rate this enterprise has obtained, which is then multiplied by its current income tax expenses to get the specific amount of tax benefits, and finally adjusted by its EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization). R&D subsidies (Sub) are obtained by taking the natural logarithm of total R&D subsidies a company received plus one. Bank credits (Ldebt) is the natural logarithm of a company's long-term borrowings plus one. Specifically, based on benchmark model (1), we established the following mediation model to test Hypothesis 1: captures the tangible channel, including tax preferences, R&D subsidies, and bank credits company i obtained in year t. Other variables' definitions are consistent with the previous text. Given the results of the benchmark model, if coefficient in equation (4) and 1 coefficient in equation (5) are significant and positive, and the value or significance of coefficient in equation (5) is lower than that in equation (1), then there exists an intermediary effect. Table 8 reports the test results. We found that HTE certification has significantly improved corporate tax preferences and R&D subsidies but has not influenced bank credits 8 . As for tax preferences and R&D subsidies, 1 coefficients in equation (5) are significant and positive, and the value of coefficient in equation (5) is lower than that in equation (1). This indicates that the tangible channel does exist, which mainly takes effects through tax preferences and R&D subsidies. The insignificance of bank credits reflects the fact that China's financial market has not yet been able to serve HTEs in a satisfactory manner. Notes: ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.

Intangible Channel
In order to study whether HTE certification can influence corporate innovation through the abovementioned intangible channel, we selected important innovative elements (R&D capital and R&D personnel) input as the proxy variables to test Hypothesis 2. Similarly, we established the following mediation test model.
where captures the intangible channel, expressed in terms of R&D input intensity (RDC, R&D capital/operating income) and R&D technicians' proportion (RDP, number of R&D technicians/number of total employees). Other variables' definition is consistent with the previous text. According to Table 9, HTE certification has significantly improved R&D investment intensity and the proportion of R&D technicians. 1 coefficient in equation (7) is significant and positive and the value of coefficient in equation (7) is lower than that in equation (1), which indicates that HTE certification can increase patent applications by increasing the input of corporate innovation factors. That is, HTE certification improve companies' motivation to engage in more technological innovation activities, which proves that the intangible channel exists. Notes: ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.

Further Analysis
As for the tangible channel, relatively standard tax preferences and R&D subsidies give companies a stable expectation of funds, which may also increase companies' tendency to overinvest in projects with high returns and short-term cycles, distracting enterprises from technological innovation (Boeing 2016). It is worth mentioning that the effect of taxation and subsidy policy tools under HTE certification in our study is a net effect and the positive effect dominates, which implicates that serious innovative resource constraints still exist in current Chinese context and the promotion effect can be exerted. It may be easier to obtain HTE certificates for enterprises located in high-tech development zones or in more economically, financially, and legally developed regions. Therefore, the number of patents enterprises applied for in these areas is originally much higher. In other words, it is likely that it is not HTE certification that leads to a higher level of technological innovation. Nevertheless, based on the following three points, we can exclude this assumption's interference in our main results. First, we have already controlled for provincial fixed effects. Second, not all companies in these areas have been certified as HTEs, which means that even though the sample includes enterprises in these areas, it contains certified enterprises and non-certified enterprises that can be both affected by this element. In this case, our regression results let us conclude that the promotion of enterprise innovation exists.
Finally, we collected the data on the index of market intermediary organizations and environmental legal system development in various provinces across the country, and then put the index into our model as a control variable (Wang et al 2017). A further test found that even if the regional environment scores were controlled for, HTE certification still significantly promoted corporate innovation. In this paper, two main underlying channels are proposed and tested. Empirically, our results show that even after controlling for the two channels, HTE certification still significantly promotes corporate innovation 9 . As a result, we assume that HTE certification may have a direct promotion effect on enterprise patent applications, or there may exist other undiscovered channels, which requires further study.

Corporate Heterogeneity's Impact on the Effectiveness of HTE Certification
In order to further examine the influence of firms' heterogeneity, the following empirical analysis was conducted. First, we ran a test for the interaction of a dummy of firm's ownership and a dummy of HTE certification. Second, we ran a test for the interaction of a dummy of equity incentive (EI; when there is equity incentive, the value is 1, otherwise it is 0) and a dummy of HTE certification. Third, we calculated the Herfindahl index (HHI) based on company's operating income (The larger the index, the greater the degree of industry concentration and the lower the level of market competition). Also, a test was run for the interaction of HHI and a dummy of HTE certification. Table 10 presents all the results. Notes: ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.
We can see that, first, for each group, HTE certification is significantly and positively associated with corporate innovation. Second, as our theoretical part predicts, the coefficients of interaction terms for ownership are significant and negative. In other words, compared with state-owned enterprises, private enterprises are more sensitive to the innovation incentive effect of HTE certification. The coefficients of interaction terms for corporate governance are significant and positive. That is, equity-incentive enterprises enjoy a greater positive adjustment effect of HTE certification on their innovation ability compared with non-equity-incentive companies. The coefficients of interaction terms for market competition are significant and negative, which means that the higher the degree of competition of the industry where the company belongs, the greater the promotion effect of HTE certification. In summary, Hypothesis 3 holds.

Conclusion
Enterprise innovation has always been among the priorities of government policy and academic research. In conjunction with 'National HTE Certification Administrative Measures' promulgated in 2008, we study Chinese enterprises' patent application status after excluding enterprises involved in R&D manipulation from the sample. We find that HTE certification has significantly promoted Chinese enterprises' innovation activities, especially boosting their patented inventions. This conclusion remains sound after addressing a series of identification issues, moreover, the more certification instances, the stronger the innovative promotion effect an enterprise can achieve. Further, we analyze underlying mechanisms and find that there exist tangible and intangible channels of the policy's innovation promotion effect. Also, corporate heterogeneity factors such as ownership, equity incentives, and industry competition level can influence the effectiveness of HTE certification.
This study reveals the role of the Chinese government's 'visible hand' in enterprise innovation and has important policy implications. First, HTE certification policy indeed promotes innovation in 'real' HTEs, but there also exist noises from 'pseudo' HTEs. Hence, the certification criteria and audit procedures should comprehensively evaluate companies' innovation capabilities to leave less space for R&D manipulation. In addition, follow-up tracking and supervising governance need to be strengthened, thus identifying 'pseudo' HTEs as early as possible to avoid the misallocation of market resources. Second, it is necessary to prudently employ policy instruments such as tax preferences and R&D subsidies. More importantly, the internal and external environment of companies should be improved. That is, the philosophy of independent innovation among enterprises should be fundamentally strengthened. Third, we believe that only policies based on corporate characteristics can be effective. Therefore, enterprises' heterogeneity should be considered and appropriate dynamic adjustments should be made in the process of policy implementation. Fourth, the Chinese capital market should be developed constantly. It will help investors to correctly understand and support corporate innovation, thus creating a good 'mass innovation' atmosphere.
This paper also leaves some areas for further research. First, patent data cannot fully reflect enterprises' innovation level, thereby a better measurement for enterprise innovation remains to be found. Second, the influence of macroeconomics on micro-subjects is complex. Although we try to reveal the underlying mechanism, a better quantification of tangible and intangible channels needs to be developed. Third, while excluding the companies certified after 2009, we may have also accidentally excluded some innovative ones. Therefore, identifying enterprises with R&D manipulation using more accurate methods is also among the further research goals. Notes: ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets.

Appendix D: Further analysis
Appendix Table 3 Economic, financial and legal institutional environment ( Notes: ***, **, and * indicate statistical significance at the 1 percent, 5 percent, and 10 percent levels, respectively. T statistics are in brackets. Equation (5) and equation (7) were merged to obtain the following equation (10), and the regression was performed. Results are shown in Appendix Table 5.