Government-Subsidized RD and Firm Innovation - RP
Government-Subsidized RD and Firm Innovation - RP
Government-Subsidized RD and Firm Innovation - RP
a
University of Hong Kong, Faculty of Business and Economics, Pokfulam Road, Hong Kong
b
Peking University, School of Economics, 5 Yiheyuan Rd, Haidian, Beijing, P.R. China
c
University of Roehampton, Business School, Roehampton Ln, London SW15 5PU, United Kingdom
Abstract
*Corresponding author: Di Guo, Faculty of Business and Economics, The University of Hong Kong
Pokfulam Rd., Hong Kong Island, Hong Kong (e-mail: guodi2007@gmail.com).
1
1. Introduction
major practice in most countries. The major rationale for such government initiative is
that firms may underinvest in R&D under a free market because of the externalities
generated by these activities (Nelson, 1959; Arrow, 1962), as well as the information
issues associated with these projects (Greenwald et al., 1984; Hall and Lerner, 2009).
failures (Romer, 1986; Aghion and Howitt, 1990). Underinvestment in R&D has been
extent to which government intervention could stimulate firms to invest more in R&D
inconclusive. Griliches and Regev (1998) and Branstetter and Sakakibara (1998) find
Israel and Japan, respectively. Moreover, such firms grow faster (Lerner, 2000),
access other external finances more successfully (Lerner, 2000; Aschhoff, 2009),
invest more in R&D activities (Audretsch et al., 2002; Lach, 2002; Görg and Strobl,
2007; Aerts and Schmidt, 2008; Czarnitzki and Lopes-Bento, 2013), and generate
higher social returns than their counterparts do (Griliches and Regev, 1998; Irwin and
public R&D programs have not stimulated firm performance (Klette and Møen, 1999;
Brander et al., 2008) or have limited positive effects on corporate R&D spending,
except for small firms (Lööf and Hesmati, 2005) or research-oriented projects
2
(Clausen, 2009). Several studies even find that government R&D subsidies crowd out
private R&D inputs (crowding out effect), thereby consequently reducing social
welfare and growth (David, et al., 2000; Wallsten, 2000; Acemoglu et al., 2013).
The mixed findings on the effects of government R&D programs have several
implications. First, institutions may influence the effects of such R&D programs.
activities by profit-driven businesses. Institutions affect the degree of the role of the
market in allocating resources and the efficiency of the government (Acemoglu et al.,
2005). As a result, the institutions under which the market interacts with government
initiatives are ultimately important to determine the success of the government R&D
initiatives. Indeed, empirical studies find that the effects of public R&D subsidies
2000; Cincera et al., 2009). Moreover, a few works based on U.S. data demonstrate a
crowding-out effect of public R&D programs (e.g., Wallsten, 2000; Acemoglu et al.,
2013), whereas most studies based on data from non-U.S. countries find universally
positive effects of such programs despite the variation in the degree of complementary
influence (e.g., Lach, 2002; Cincera et al., 2009; Czarnitzki and Lopes-Bento, 2013).
Second, even under similar institutions, the governance of these public R&D
who allocate the resources. Government agencies play an essential role in allocating
resources through public R&D programs. Thus, the governance of these programs
will expectedly affect their effectiveness. However, to our knowledge, focus on the
endogeneity issues in empirical examinations resulting from data constraints are also
3
a major challenge in existing studies, which may have also contributed to the
This study attempts to fill some of the abovementioned gaps. We examine the
effects of Innovation Fund for Small and Medium Technology-based Firms (Innofund)
on the innovation outputs of firms. Innofund is the largest government program that
innovation (measured by sales from new products and exports) and technological
Innofund governance brought about by the exogenous policy shock in 2005 influence
find that Innofund-backed firms generate significantly higher innovation outputs (both
by the count of newly granted patents) after 2005 when Innofund governance shifted
from a centralized screening system to a more decentralized one. Our results imply
that decentralized governance is more effective than the centralized one in public
R&D investments.
identification issues that result from selection biases and omitted variables. We
4
attempt to address these identification concerns by using two approaches, i.e.,
variables (IVs). We use the PSM approach to match Innofund-backed firms with non-
Innofund-backed firms on the basis of various criteria that may predict the probability
of a firm being selected by Innofund and the future innovation potentials of the firm.
endogeneity issues. The first IV refers to the total number of firms located in high-
tech zones in a given city for a given year. The second IV refers to the ratio of annual
investments in fixed assets made by local governments over GDP at the county level
for a given year. Both IVs reflect how ambitious the local governments are. We
suggest that the more ambitious the local governments are, the more likely they
support local firms to participate in Innofund program competition and also exert
more effort to lobby the upper-level governments for local firms to win Innofund
grants. Statistically, the two-stage estimations confirm the relevance and the
exogeneity of the IVs, thereby indicating that the two IVs are qualified. Our major
Our study differs from and complements the existing literature in three aspects.
First, our study is the first one that links the governance of the government R&D
programs and the effects of such programs. The Innofund program, which was
governance in 2005 when the central government decided to shift from a centralized
project screening system into a relatively decentralized one. This exogenous policy
change provides us with the opportunity to scrutinize the use of the quasi-experiment
approach and determine how governance of public R&D programs influences the
5
effects of such programs. Our study proves that the governance of government R&D
Second, this study is among the first examinations on the effects of government R&D
programs in China. The Chinese government has been deeply involved in businesses,
particularly in resource allocation. The inefficiencies that result from the involvement
of the Chinese government in business are well documented (Brandt et al., 2013; Guo
rampant and the market remains immature. Third, we employ two approaches to
address the identification concerns in this study. Most existing studies on government
R&D programs mainly employ PSM approach to mitigate selection biases. In our
study, we not only employ PSM but also apply two-stage estimations to control the
potential concerns with missing variables. Hence, we attempt to shed some light on
the existing discussions with regard to why empirical findings are inconclusive in
institutional background of the Innofund program and the policy changes it underwent
in 2005. Section 3 describes the sample and data. Section 4 presents the empirical
findings on whether Innofund affects innovation outputs of firms and examines the
robustness of the results. Section 5 reports the findings on the effects of the policy
bring along and attract outside financing for corporate R&D investment of SMTEs.”
The principal criteria for applying to Innofund are as follows: The project
should comply with national industrial technology policies, exhibit relatively high
potential for economic and social benefits, and competitive in the market. The
applicant should be a business corporation with generally not more than 500
employees, not less than 30% of which should have received higher education. The
annual R&D investment of the firm should be at least more than 3% of the total sales,
and the number of R&D employees should be more than 10% of the total number of
employees. Firms with leading products in the market with an economy of scale
production must also exhibit good economic performance. The following projects are
rights and high value added; projects established by researchers or overseas returnees
firms, universities, and research institutions; and projects that utilize new and
advanced technologies to revive the stock assets of traditional industries and drive job
creation.
free bank loans, and equity investment. Appropriation is provided as start-up capital
1
Source: http://www.innofund.gov.cn/
7
for small firms founded by a researcher with scientific achievements. Partial subsidies
are also provided to SMEs for the development of new products and pilot production.
The total amount of subsidies for an individual project is generally between 1–2
the funded projects. Interest-free loans are provided mainly to SMEs that require
projects. Generally, equity investment is reserved for projects that use advanced
technology, have high innovation capacity, and have market potential in emerging
industries. On average, Innofund support should not exceed 20% of the registered
From 1999 to 2011, Innofund provided more than 19.17 billion RMB to
30,537 projects, 27,498 (86%) of which were supported through appropriation, 2,880
through interest-free loans, and 1,159 through other forms, including bank loan
insurance, equity investment, and other forms of subsidies. The size of direct
for government R&D in China. However, according to official reports, Innofund has
induced 1:11 external financing from local governments, banks, and venture
high-tech firms, such as Zhongxingwei and Huawei. Since 1999, the program has
created approximately 450,000 new jobs and generated 209.2 billion RMB in sales,
22.5 billion RMB in tax income, and 3.4 billion RMB in exports. By the end of 2008,
82 out of 273 publicly listed companies in China’s SME Stock Exchange were once
supported by Innofund.2
2
http://www.innofund.gov.cn/.
8
2.2 Innofund Governance Before and After 2005
Innofund. At the central level, the Innofund Administration Center (IAC) under the
including the issuance of the application guide, preparation of proposals for the
preferred fields and industries to support for each year,3 screening and evaluation of
project assessments. The Ministry of Finance (MOF) plays a regulatory role and
approves the application guidelines and yearly budget, transfers funds to the IAC
twice a year, and assesses IAC operation. The MOST and MOF report yearly to the
State Council on the operation and performance of Innofund. The IAC must adhere to
the principles of honest application, fair processing, strict selection, and transparent
administration. According to IAC reports, fraudulent cases for each year constituted
less than 0.5% of the total projects for the past 10 years.
At the local level, each province has an Innofund office under the Provincial
Science and Technology Committee, which reports to the IAC. The role of the local
administration was reformed. The policy changes simplified the application processes,
decentralized the screening and evaluation of projects, and delegated more power to
Local Innofund offices principally acted as bridges between IAC, and the local firms
3
A consulting committee composed of technology and management specialists, economists, and
entrepreneurs help identify preferred areas to support and provide advice on Innofund guidelines.
9
had no considerable involvement in project selection. The local Innofund offices had
three major responsibilities. These offices delivered and promoted IAC guidelines and
policies to local firms or agencies to guide them in preparing the required application
documents. The local offices also collected the application documents and certified
the qualifications of applicants. Finally, the local offices recommended and forwarded
experts at the IAC evaluated the submitted applications and promulgated the final
funding decisions. Local Innofund offices were only recommendatory bodies that did
allocated by the local governments to the recommended projects until the IAC
announced its final decision. After the IAC reached a decision, the provincial Bureau
of Finance was normally required to match 50% of the total support from the central
In 2005, the operations and governance of Innofund were reformed, and a new
application and screening system was introduced. The system considerably increased
selection. The role of local Innofund offices was substantially shifted. Local
governments at the provincial level were now required to set up their own Innofund
programs and take responsibility for the initial project selection. In particular, project
assessments by local Innofund offices constitute 30% of the final decision of the IAC.
commit at least 50% of the proposed support (25% for provinces in Western China) to
selected local projects before even recommending the projects to the IAC. The list of
projects that local offices plan to recommend must be published in their websites for
10
two weeks before the applications are submitted to IAC. Accordingly, these offices
screening of the Innofund Program. The major feature of this policy change is the
delegation of power on project selection to local Innofund offices. Compared with the
reduce inefficiencies that result from the hierarchical decision making process by
solving information issues and considering that local officers have more knowledge
on local firms. Thus, information issues can be addressed immediately. The delegation
of decision making power and the newly introduced co-investment mechanism also
aligned with the interests of local and central governments, and provided more
Indeed, anecdotes reveal that the reforms introduced in 2005 ushered in creative
such as city or county governments. In Chongqing and Hunan provinces, the local
Innofund offices cooperated with other government and consulting agencies, such as
the local industrial and commercial bureau, tax bureau, law firms, and accounting and
auditing firms, to collect information on candidate firms for project selection. These
efforts are also reflected in the total amount of funds granted by the local Innofund
offices. According to the official report of the Innofund program in 2005, local
committed by local government was more than 1.2 billion RMB or approximately six
times that used by the local government to provide matching funds before 2005. We
11
expect this systematic change in the governance of Innofund to influence the effects
of the program.
commercialized and technological innovation outputs of the firms after they receive
sales from new products and exports of a firm, whereas technological innovation
outputs are measured by the number of newly granted patents of a firm for each year.
We also control several firm-specific variables including age, size, leverage ratio, and
Table A-1).
Our data are acquired mainly from three sources. Basic information on
publicly announced on the website each year since 1999. The website provides the
names and addresses of the firms, the nature of the projects, the date the firm was
granted Innofund support, the type of support from Innofund, and the results of
performance evaluation of the project (i.e., terminated during the process or finished
on time and achieved the proposed goal). Firm-level data on financial information,
sales from new products, exports, and other firm-specific characteristics are obtained
from the Above-scale Industrial Firms Panel 1998-2007 (ASIFP). ASIFP is composed
of all state- and non-state-owned industrial firms with annual sales of at least
12
5 million RMB (US$750,000) between 1998 and 2007. This database provides
industry, age, and ownership structure. Patent data are obtained from the State
Intellectual Property Office (SIPO) patent database. The SIPO database provides
complete information on all patents granted in China, including the application and
publication number of the patent, application and grant year, IPC classification
The first challenge in this study is data matching because the names of the
firms listed in the three databases may not be fully consistent. First, we need to match
the list of Innofund-backed firms in the Innofund website with the list in the ASIFP
database to identify which firms in the ASIFP database have won Innofund support
and obtain detailed financial information for these firms. We employ both
computerized matching and manual matching approaches to match the two databases.
As mentioned, both the Innofund website and ASIFP provide information for the
names, locations (at city level), and industries of the firms by year. ASIFP also
provides information for the legal person code of all the firms in the database. We use
which is similar to that used by the NBER Patent Data Project4, to ensure accuracy of
the matching. In the first step, we standardize the firm names in the two databases to
prepare for the matching. Under the Company Law of China, a company name must
4
https://sites.google.com/site/patentdataproject/Home.
13
contain four elements, namely, a trade name, industry sector, legal entity identifier
(e.g., Limited Liability Company or Joint Stock Limited Liability Company), and the
administrative region. We first create a “standard name” for a firm by removing the
\, etc.)5 and standardizing the legal entity identifiers (e.g., we converted Limited into
Ltd.). This step is carried out to prevent the matching quality from being affected by
inconsistencies in the formats of firm names listed in the two databases. Moreover, we
created a “stem name” for each firm by removing the administrative region and legal
entity identifiers in the firm name (e.g., a firm called “Beijing Tian Fa Logistics Ltd”
is changed to “Tian Fa Logistics”). This step is carried out to prevent the matching
quality from being affected by the mistake driven by input errors with legal entity
conducting matching with “standard names”, “stem names”, and other information in
the two databases by Innofund awarding year. We first accurately match the two
databases using the “standard names”, locations (at city level), and standard industry
codes (SICs) of firms for the year when Innofund-backed firms won the grant. If a
firm was awarded an Innofund grant in 2000, then we use the aforementioned
matching information of the firm listed in the Innofund website of the said year to
match with that of the firms listed in ASIFP of the same year. We generate a matched
file called “full marching_2000” for the matching results of 2000. Year and location
company has an exclusive right to its name on a regional basis. A company name
5
These characters may be input into the names of the databases by mistakes.
14
must be unique and identical within its region. Thus, if a firm has exact the same
Chinese name and location in the two databases for the same year, then it should be
the same firm. We repeat this procedure for each Innofund awarding year and the
counterpart year of ASIFP, and generate the matching files by year accordingly.
the “standard names” with the “stem names” and generate a matched file called
“partial matching” for each Innofund awarding year. We use ”stem names” to conduct
using ”standard names” (we may not have exhausted all the expressions of the legal
entity identifiers and converted them into standard identifiers when we created
“standard names”). Finally, we combine the matching results of the ”full matching”
and “partial matching” by year and delete duplicates using the legal person codes of
dataset for each year between 1998 and 2007 in which Innofund-backed firms are
identified in ASIFP for the year when they obtained the grant.
check all Innofund-backed manufacturing firms that are not matched by computerized
matching using online search engines such as Google and Baidu. We mainly focus on
checking the names, business nature, legal person codes, and Innofund granting
records of the firms to ensure that we do not miss some observations because of slight
variations of the trade names of firms listed in the two databases. Similarly, after the
combine the yearly matching results from the computerized and manual matching by
year and create a pooled cross-sectional dataset entitled “final matching” for each year
in which all Innofund-backed firms are identified from ASIFP for the year they won
the grant. We thereby obtain the legal person codes of all identified Innofund-backed
firms and distinguish the time when the firm was awarded an Innofund grant. Finally,
2,638 firms that won backing from Innofund at least once between 1999 and 2007 are
identified for the estimations.6 We build the panel data for the identified firms by
merging the firms listed in “final matching” into ASIFP by year and adding two
dummy variables into ASIFP to distinguish whether the firm won and when it won
Innofund (Brandt et al., 2012). The final sample consists of 18,224 firm-year
With this matching strategy, we ensure that the variations or the changes of
firm names over the years will not affect the quality of our matching. First of all, by
controlling the “standard names”, locations (at city level), industries of firms, and the
year in computerized matching, we may ensure that type I error does not occur in
matching. According to the Company Law of China, a company name must contain
four elements, namely, a trade name, industry sector, legal entity identifier (e.g.,
Limited Liability Company or Joint Stock Limited Liability Company), and the
6 The number of Innofund-backed firms for the estimations dropped substantially from 11,977 (the number of
project backed between 1999 and 2007) to 2,638 for estimations during the examination period because of several
reasons. The ASIFP database covers manufacturing firms only; therefore, we cannot include non-manufacturing
firms backed by Innofund, thus reducing the number of Innofund-backed firms in the sample. Non-state-owned
firms with sales of less than RMB 5 million are also not included in the ASIFP. Hence, we may have missed
several micro-sized firms backed by Innofund. One of the aims of the study is to estimate the ex-post effects of
Innofund. An Innofund-backed firm that lacks information on the year when it received funding is also excluded.
Theoretically, we included all state-owned manufacturing firms supported by Innofund and non-state-owned
manufacturing firms with more than 5 million RMB in sales (in the year of application) backed by Innofund for the
estimations.
16
administrative region. Moreover, a company has an exclusive right to its name on a
regional basis. Thus, if the Chinese name, location, and industry of a company are
shown exactly the same in both databases in the same year, then it must be identified
as the same company. Year is an important factor to secure the accuracy of the
matching. Firms are considered identical only when the firm names can be matched in
the same year shown in both databases. Moreover, to prevent type II errors in
“stem names” and manual matching. Such procedures prevent the matching quality
from being affected by the variations in firm names shown in the two databases.
changes of firm names over time. First, the final panel database for Innofund-backed
firms is not built up by firm names. Rather, we establish the panel by legal person
codes of firms based on the cross-sectional data matched by firm names and other
information by year. According to the China’s Company Registration Rules, the legal
person code of a company is unique nationwide and will not change after the
registration of a legal entity even if its name or business nature is changed. By using
the legal person codes, we identified firms by year to match and build up the panel
database. Thus, the changes of firm name over time cannot affect the matching quality.
firms belong to high-tech industries. The allocation of Innofund is consistent with the
received their first round of Innofund grant. We show the distribution of awarding
year for the sampled Innofund-backed firms in this study and the full sample of
Innofund-backed projects. Results show that from 1999 to 2007, the sampled
Innofund-backed firms have similar year distributions -like those in the full sample,
This study also needs to match the firms in the ASIFP database with those in
the SIPO patent database to identify patent information for all firms in the estimations.
In general, three types of patents exist in China, namely, invention, utility, and design
other major patent offices in the world. This type of patent is given 20 years of
protection and may be granted to the methods and products. Both utility and design
patents are given 10 years of protection. Utility patents are generally granted to
normally granted to shapes and patterns with patentable aesthetic appeal. Firms have
patents are the most technologically innovative and thus require more R&D efforts
than the two other types. In this study, we measure patent outputs using two values:
the number of invention patents and the number of patents of all types granted to a
firm in a given year. Given that creating patentable works and applying for a patent
take time, we use filing time of newly granted patents as a basis in panel estimations.
We also use the one-year lag of filing time for all estimations to check the robustness
of the results.
18
The matching strategy we apply to match ASIFP and SIPO is significantly
similar to that which is used to match the name lists of Innofund-backed firms and the
ASIFP. However, the major difference is that SIPO does not provide information for
the industry of a firm that we do not control in the matching. However, this issue will
not affect the quality of our matching. As we discussed earlier, firm names, location
and year are the details that are needed to secure the accuracy of the matching.
for the subsidiaries of firms. According to the Patent Law of China, organizational
patent applicants must provide the registration license while applying to file a patent,
thereby suggesting that a firm that applies for patents must be an independent legal
entity. Patents applied by subsidiaries that are not registered as independent legal
entities will be filed to the parent firm. Similarly, only an independent legal entity will
based on both the names and locations of firm (for cross-sectional data matching by
year) and legal person codes (for panel construction), should not be affected
group with several steps to ensure that our results are not driven by a specific
matching method and control for selection biases. We first identify firms that are
eligible to apply for Innofund but did not apply or did not win the grant from the
ASIFP Database. The Innofund selection criteria are officially announced each year.
19
A firm is eligible for Innofund application if its SIC 7 is similar to the SICs of the
awarded group, has fewer than 500 employees, and has a leverage ratio lower than
70%. The Innofund program also requires that R&D investments of a firm should be
more than 3% of the total sales, and the number of R&D employees should be more
Rosenbaum and Rubin (1985) to construct the control group on the basis of the
identified pool of eligible firms. In the context of our study, the propensity score
potential to generate high economic and social benefits, firms with leading products in
the market, firms with projects that utilize new and advanced technologies or with
independent intellectual property rights and high value added, and projects that utilize
new and advanced technologies will be prioritized. That is, innovation potentials are
innovation performance is our major focus in designing the PSM algorithm. When
the matched non-Innofund-backed firms are selected based on their two-digit SIC
industry code, location, size, leverage ratio, sales from new products, exports, and
stock of patents. Following the suggestion of Démurger et al. (2002), we control the
which may affect the results. We also match the size and leverage ratio of Innofund-
7The National Bureau of Statistics in China updated the SIC system in 2003. Thus, we amend the two-digit SIC
before 2003 to maintain consistency with the latest code system.
8ASIFP does not provide information on human capital and presents only data on R&D investment from 2005 to
2007. Thus, we cannot utilize the R&D investment and human capital information as criteria for the control group
sample construction.
20
backed firms and their counterparts. These criteria ensure that Innofund-backed and
non-Innofund-backed firms are similar in many aspects, which may affect the
probability of being supported by Innofund and their innovation outputs in the future.
can be found by following Rosenbaum and Rubin (1985). Table 2 presents the results
of the balance tests of both the randomly drawn sample 10 and the PSM matched
variables are balanced between the Innofund-backed firms and the PSM matched
sample. 11
The major shortcoming of the ASIFP database that affects our PSM is that
ASIFP provides data on R&D expenditure only from 2005 to 2007 and does not
investment and information on human capital as criteria to construct the control group
sample. However, given that R&D expenditure is one of the most important factors
that may affect innovation outputs, we utilize R&D investment information as criteria
to match a subsample of Innofund firms that obtained funding in 2005 and 2007 to
backed firms in the PSM control group, including the number of observations, mean,
9 Our results are robust after we remove the common support restrictions.
10 We construct a randomly draw sample of the control group and present the difference between the sample
matched by PSM approach and the randomly draw sample to further justify why we have to employ the PSM
sample in order to reduce the selection biases.
11 The balance tests for other variables are available upon request.
21
maximum, and standard deviations across the entire examination period. On average,
newly granted patents and sales from new products. Similarly, these firms are younger
and larger in size as measured by total sales and total assets. These firms also have
We test whether the Innofund Program helps firms generate more innovation
outputs by implementing fixed effect panel data regression through the following
where i indices a firm, indices time, and yit are dependent variables used to measure
the innovation output of firm at time . The innovation outputs include sales from
new products, exports, and newly granted patents. InnoAftit is a dummy variable
equal to 1 if the firm gained Innofund support at time t and equals zero if otherwise. A
22
vector of control variables is indicated by . controls time-invariant firm-specific
unobserved variables, and controls yearly fixed effects. The effect of Innofund on
estimate (1.1) when the dependent variables are exports, sales from new products, and
patent counts in log-link formulation. We utilize a logit model for panel data to
estimate (1.2) when the dependent variables are dummy variables of sales from new
products and exports. The standard errors for correlation are adjusted within the
firms. Models (1) to (4) show that is significantly and positively associated
with sales from new products and exports of firms, whether these values are measured
firms generate significantly higher sales from new products and exports after gaining
government support compared with non-Innofund-backed firms and the same firms
before funds were infused. Meanwhile, the probability that Innofund-backed firms
generate sales from new products and exports is significantly higher than that of non-
Innofund-backed firms and the same firms before the funds were infused. For
example, Model (2) shows that, given the other things being equal, the probability that
a firm generates sales from new products will increase by 7.88% after the firm wins
Innofund support. Similarly, Model (4) shows that winning Innofund support can help
Models (5) and (6) present the estimations of how Innofund affects newly
granted patents and show that Innofund significantly and positively motivates firms to
23
generate more patents. We examine both the number of patents of all types and the
Models (5) and (6) are significantly positive, thereby indicating that Innofund-backed
firms generate more new patents of all types and more invention patents after winning
Innofund support compared with non-Innofund-backed firms and the same firms
before the grant. For instance, Model (5) shows that the growth rate of newly granted
patents of all types for Innofund-backed firms after the grant is 13.2% higher than that
of non-Innofund-backed firms and the same firms before winning the grant. Model (6)
shows that the growth rate of newly granted invention patents for Innofund-backed
firms after the grant of funds is 8.6% higher than that of non-Innofund-backed firms
and the same firms before the grant. In summary, Table 4 shows that Innofund
effectively influences the innovation outputs of awarded firms. Our results remain the
same when we use the one-year lag for patent filing time.
The monetary effect of the funding is also examined. The estimation focuses
on the total amount of Innofund support given to firms. Thus, we may obtain more
insightful ideas on the extent to which government R&D funding addresses the
The results are presented in Table 5. Models (1) to (4) show that InnoAmtit is
significantly and positively correlated with the sales from new products and exports of
firms. These findings imply that firms that win a larger Innofund grant may generate
significantly higher sales from new products and exports. Meanwhile, the probability
of generating sales from new products and exports increases as the size of Innofund
support increases. For example, Model (2) shows that if a firm wins a funding of
24
1 million RMB, then the probability that the firm generates sales from new products
increases by 11.73%. Similarly, Model (4) shows that if a firm wins a funding of
1 million RMB, the probability that the firm generates export increases by 2.63%.
Model (5) shows that after winning a funding of 1 million RMB, the growth of
newly granted patents of all types generated by Innofund-backed firms is 20% higher
than that of non-Innofund-backed firms and the same firms before the grant. Model (6)
shows that a funding of 1 million RMB results in the 10% higher growth of newly
Innofund-backed firms and the same firms before the grant. The results shown in
Table 5 imply that the amount of Innofund support affects the innovation outputs of
firms significantly.
We further examine the effects of the relative weights of the funds. The ratio
of Innofund support over total free cash of the firm is used to replace InnoAmtit, and
the regressions conducted in Table 5 are repeated. However, the results show that the
abovementioned relative measure does not affect the innovation outputs of firms
R&D input is an important factor that may affect the grant of funds. However,
the ASIFP database does not provide complete information on the R&D expenses of
firms and only contains information for 2005, 2006, and 2007. Hence, a subsample
analysis is conducted to test whether our results are biased because of the missing
information. We focus on firms that obtained their first round of Innofund funding in
2006 and 2007. Innofund-backed firms are matched with non-Innofund-backed firms
in the year prior to the awarding of the Innofund grant on the basis of the two-digit
SIC industry code, location, total sales, exports, sales from new products, number of
25
patents, and R&D expense. Estimations in Table 4 are duplicated based on the newly
matched subsample. The results are presented in Table A-2, which shows that our
main conclusion remains the same after we control R&D input in the PSM process.
4.2 Identification
firms in terms of sales from new products, exports, and newly awarded patents after
potential ex-ante selection effects are controlled using the PSM approach. One
significant limitation of the PSM methodology is its inability to determine the effects
improved innovation outputs. For instance, we could not measure the R&D ability of
firms or observe the management capability of executives on the basis of existing data.
However, both factors may contribute to the innovation outputs of the firms.
with unobserved variables that may affect dependent variables. The first IV used is
the total number of firms in high-tech zones of the city where the firm is located in
each given year (Frm_HTZ). High-tech zone is a distinctive type of special economic
zone (SEZ) in China where central and local governments seek to attract foreign
direct investment and consequently stimulate the development of the local economy.
Information on the number of firms in high-tech zones at city level is obtained from
the China Statistical Yearbook on Science and Technology (1999–2007). The second
IV is the ratio of total investment in fixed assets made by local governments over the
total GDP at the county level each year (Fxd_Asst). Information on local government
26
investment across 1998 to 2007 is obtained from city yearbooks. Both IVs reflect the
effort level of the local governments in the developing local economy. We suggest
that these two IVs can help to identify the probability of a firm winning an Innofund
grant. However, the IVs should not be directly correlated with the error terms that
The choice of the two IVs is mainly based on the understanding in institutions
in China. Under the regionally decentralized authoritarian regime in China, the central
government governs the state through personnel control, whereas local governments
manage economic activities and allocate resources (Xu, 2011). During the economic
reform era, local governors compete with each other in terms of economic growth, the
search for resources, and support from the central government to obtain promotion
governments are more likely to support local firms in competing for the Innofund
program and to exert more effort in lobbying the upper-level governments for local
firms to win Innofund grants. The level of effort of local governments in attracting
foreign investment and investing in fixed assets is a good indicator to measure how
ambitious the local governments are. We consequently expect the two IVs to be
However, the two IVs should not be directly correlated with the error terms of
estimations on innovation outputs of individual firms. The two IVs used are measured
either at the city level or county level, whereas innovation outputs are measured at the
firm level. That is, we should not expect a direct relationship between the
limited, and results are mixed. Several studies find that the establishment and
physical capital, exports, or outputs of foreign firms at the city or province level
(Cheng and Kwan, 2000; Wang, 2013; Alder et al., 2013). However, Hu (2007) did
on the basis of firm-level data, Schminke and Van Biesebroeck (2013) reveal that
firms within SEZs do not generate higher total factor productivity. The existing
literature suggests that the relationship between the IVs we use and the innovation
outputs of individual firms may be unclear. Thus, we statistically test the exogeneity
6 presents the results from the first stage of estimation. The results demonstrate that
the number of firms in local high-tech zones and the investment in fixed assets made
wins Innofund backing at a given year. These results suggest that a firm has a greater
local governments are more ambitious and provide more support to local firms. The
The results of the second stage of estimation are presented in Panel B of Table
6. Sargan tests are performed to test the exogeneity of the two IVs. The results of the
28
Sargan tests indicate that the null hypothesis, which states that the two instrumental
variables are uncorrelated to the residuals, cannot be rejected for all estimations.
Thus, the results statistically prove that both the IVs satisfy the conditions of
qualifications as IVs. Models (1) to (2) indicate that firms generate more sales from
new products after they obtain Innofund grants compared with non-Innofund firms
and the same firms before receiving Innofund support. Similar results are observed in
the number of newly granted patents of all types and the number of newly granted
invention patents. The outcomes of two-stage estimations are consistent with the
regression results in Tables 4 and 5. These outcomes empirically confirm that winning
Innofund support positively affects innovation outputs, even after considering the
2005 because of policy shock (Section 2). The major feature of this change is that the
screening to local Innofund offices. R&D projects are associated with a high level of
whether the change in ex-ante screening systems may influence the effects of
the organizational structure are abundant. The rationality of human beings is limited.
Moreover, information gathering, transmission, and processing are costly. Sah and
Stiglitz (1991) argue that centralized organizations may delay decision making and
29
reduce the total number of projects because of cost constraints and the lack of local
approach, Aghion and Tirole (1997) and Hart and Moore (2005) further emphasize
that a decentralized decision making system may strengthen the incentives of local
agents in acquiring information and may reduce the overload problem experienced by
the principal. Stein (2002) predicts that decentralized organizations are more
attractive when the needed information is “softer” (i.e., the information is difficult for
outsiders to verify), whereas centralized organizations are more favorable when the
without cost.
derived from soft budget constraints theory. Dewatripont and Maskin (1995) suggest
that a centralized credit market may affect efficiency because of adverse selection and
the lack of a termination mechanism. Qian and Xu (1998) further posit that
bureaucracy often results in more mistakes through the rejection of promising projects,
thus delaying innovation. Decentralized decision making may not only reduce ex-ante
screening costs but may also terminate bad projects ex-post that both types of errors
may reduce. Thus, decentralized organizations may increase the number of selected
refinance bad projects. This effect should be more obvious in investment when the
uncertainty is higher, and the quality of the projects is more difficult to predict ex-ante.
organizational structure. Rajan and Wulf (2006) show a strong movement towards
flatter corporations in the U.S. Caroli and Van Reenen (2001) report a positive
30
association between decentralization and the development of IT adoption. Acemoglu
et al. (2007) find that, apart from younger firms, more technology-oriented firms are
efficiency of the decision making processes and the organizational forms of for-profit
subsidy programs.
high level of uncertainty and severe information-related issues because the Innofund
program targets young firms with potential advanced technology in some frontier
industries. Thus, the efficiency of the information passage and incentives of local
knowledge holders (i.e., local Innofund offices in our context) are important for
project selection. The major policy change in 2005 was to delegate more decision
making power to local Innofund offices. Local Innofund offices had no input in the
final decision of the awardees before 2005. After 2005, their views have 30% weight
in the final decision of the awardees. Moreover, the ex-ante funding commitment after
2005 further enhanced the alignment of interests between the local and central
Innofund offices. Therefore, this policy change may significantly affect the incentives
of local Innofund offices and the effects of Innofund. Indeed, as introduced in Section
2, local Innofund offices took more initiative to experiment with new approaches in
project selection after 2005. On the basis of existing literature (Dewatripont and
Maskin 1995; Qian and Xu, 1998; Ahgion and Tirole, 1997; Hart and Moore, 2005),
the decentralized screening system after 2005 is expected to help in selecting better-
quality projects and consequently have stronger positive effects on firm innovation
31
outputs compared with the centralized screening process before 2005. Moreover, the
conducted by distinguishing firms backed by Innofund before and after 2005, along
below.
where all the variables remain the same as those in Equations (1.1) and (1.2), and the
Innofund dummy variable is replaced with two dummy variables to specify the
that is equal to 1 if the firm has gained Innofund support at time t, and the support was
dummy variable that is equal to 1 if the firm has gained Innofund support at time t,
and the first Innofund has been granted after 2005; otherwise, the dummy variable is
equal to 0.
Table 7 reports the regression results for the effects of the change in the
screening system. Models (1) to (2) show that Inno_2005Befit and Inno_2005Aftit are
significantly and positively correlated with the sales from new products measured by
32
log-link formulation of absolute number and dummy variable. The results are
consistent with the findings shown in Table 4. To test the significance of the policy
change effects, we conduct Lincom tests and statistically examine the difference of
suggest that the difference between the two coefficients is statistically insignificant
Models (1) and (2). Models (3) to (4) present the estimations for exports. The results
are similar to those that we observed with the sales from new products. Models (1) to
(4) indicate that the effects of Innofund on commercialized innovation outputs do not
seem to significantly change after 2005 when the governance of the government R&D
The findings shown in Models (5) and (6) are different. The models show that
newly granted patents of all types and invention patents. Moreover, the coefficients of
in both regression models. Model (5) indicates that after gaining Innofund support, the
selected before 2005 is 11.4% higher than that of non-Innofund-backed firms and the
same firms before winning the grant. The growth of newly granted patents of all types
by Innofund-backed firms selected after 2005 is 16.2% higher than that of non-
Innofund-backed firms and the same firms before winning the grant. Model (6)
demonstrates that after winning the Innofund grant, the growth of newly granted
higher than that of non-Innofund-backed firms and the same firms before winning the
33
grant. After the firms win Innofund support, the growth of newly granted invention
patents by Innofund-backed firms selected after 2005 is 10.1% higher than that of
non-Innofund-backed firms and the same firms before winning the grant. Moreover,
the Lincom tests statistically confirm that the growth of both newly granted patents of
all types and invention patents is significantly higher for firms that win the Innofund
changed in 2005. However, the policy change in 2005 does not seem to affect the
A few alternative mechanisms may exist for the results of the 2005 effects. For
instance, the property rights protection was improved since 2004, which may be one
of the alternative mechanisms that helped enhance the effects of Innofund. 12 With
better protection of private property rights, firms may have stronger incentives to
invest in R&D activities after 2004 than before in general. Second, the improved
protection for intellectual property rights since 2003 may also contribute to the
enhanced Innofund effects after 2005.13 Given that IPR is an important system that
protects and promotes R&D investment, the improved IPR protection in China since
12 Specifically, in 2004, the state constitution of China was amended, and the protection of private property rights
was constitutionalized for the first time. Although the private sector was legally recognized in the mid-1990s, the
protection of private rights was not recognized by the constitution until 2004.
13 Starting from 2003, China and the United States have held a round-table conference on IPR every year, and they
have reached agreements on many IPR-related issues at two round-table conferences. In 2004, China and Europe
held their first round of talks on IPR in Beijing, and an initial agreement was reached between the two sides on
matters of cooperation related to IPR. With more interactions and cooperation with US and Europe, the
enforcement of IPR protection was significantly improved in China. Statistics have shown a sharp increase in
patent application. Patent applications in China had exceeded two million by March 17, 2004. It took China 15
years for patent applications to reach one million. However it took only four years for the number to double from
2000 to 2004.
34
The abovementioned two institutional changes may be relevant to the
enhanced Innofund effects observed after 2005. However, these institutional elements
time, although the marginal effects may be different for the two types of firms.
terms of innovation outputs while using 2005 as a cut-off. Moreover, as shown in the
data, the rejection rate of Innofund application significantly decreased after 2005, thus
suggesting that the local IAC becomes more careful in project selection when it has
more decision-making power in project screening and needs to commit the matching
more direct factor contributing to the enhanced Innofund effects after 2005. The
results are consistent with the arguments of Dewatripont and Maskin (1995), Ahgion
and Tirole (1997), and Qian and Xu (1998). These researchers propose a more
degree of uncertainty is higher and the information issues are more severe.
6. Conclusion
This paper estimates the effects of Innofund on the innovation outputs of firms.
Innofund is one of the largest Chinese government programs that target corporate
program influences the effects of Innofund aside from its general effects on the
35
Innofund-backed firms generate significantly more innovation outputs
compared with non-Innofund-backed firms and the same firms before Innofund
funding was infused. We use PSM methodology to control the selection issues. The
results remain robust after using two-stage Heckman estimations to further address
the identification problems. These findings are consistent with several existing studies,
which argue that government funding stimulates corporate R&D activities (Irwin and
Klenow 1996; Griliches and Regev, 1998; Audretsch et al., 2002; Lach, 2002; Görg
and Strobl, 2007). Furthermore, Innofund effects differ before and after 2005 when
technological innovation outputs of firms further improved after 2005 when project
screening became more decentralized. These results are consistent with the findings
reported by Dewatripont and Maskin (1995), Ahgion and Tirole (1997), and Qian and
Xu (1998).
This study provides a new perspective for evaluating government R&D policy.
the governance of the government R&D programs and their influence on the effects of
such programs that have been largely neglected by extant literature. Meanwhile, as a
China, this study extends the extant literature by exploring how the market failures
and the government engagements interact under weak institutions in China. Finally,
this study is also related to the literature on general R&D financing mechanisms by
exploring the governance of the financial institutions and the effects of the investment.
This study has important policy implications. The findings of this study
suggest that decentralized governance may ease the information issues and motivate
36
local governments to exert more effort in project selection and ex-post monitoring
activities, thus improving the effects of government R&D programs. Moreover, the
R&D activities. Driven by government policy, China’s R&D expenditure has grown
into the second largest worldwide since 2010 (WSJ, 2010) and is expected to become
the largest worldwide by 2022 (KPMG, 2013). China’s current R&D expenditure over
GDP ratio is higher than that of the European Union (Noorden, 2014), and its total
number of patent applications has surpassed that of the U.S. since 2011 (KPMG,
growth and affects the competitive landscape of the global economy. However, solid
This assessment of Innofund program and its governance should have some important
This study also raises several questions for further research. First, if
government R&D programs indeed contribute to the innovation outputs of the firms,
or profitability of the firms? Second, can other mechanisms (e.g., property right
institutions, IPR protection, financial budget constraints [Qian and Xu, 1998; Huang
and Xu, 1999], product competition, and input markets or trust and relationships
14 Noorden, Richard Van, 2014, “China tops Europe in R&D Intensity,” Nature, 08 Jan. 2014,
(http://www.nature.com/news/china-tops-europe-in-rd-intensity-1.14476); WSJ,2010,“China Surpasses Japan in
R&D as Powers Shift,” Wall Street Journal, 13/12/2010; KPMG, 2013, “Innovated in China: New Frontier for
Global R&D,”
(http://www.kpmg.com/CN/en/IssuesAndInsights/ArticlesPublications/Newsletters/China-360/Documents/China-
360-Issue11-201308-new-frontier-for-global-R-and-D.pdf)
37
[Allen, et al., 2012]) influence the effect of government R&D funding? If so, how do
different mechanisms work together or interact with one another? Third, do the effects
of different forms of government R&D programs vary? If so, what are the
38
Acknowledgements
We are grateful for the helpful comments from three anonymous referees. Special thanks go
to Professor Keun Lee, our editor, for his guidance and insightful comments on different
versions of the paper. Comments from Chenggang Xu, Larry Qiu, Zhigang Tao, Bo Zhao
and seminar participants at the University of Hong Kong, ALEA 2013 annual meeting and
11th International Conference of WEA are appreciated. We thank Jing Qian, Long Hong and
Felix Shiu for their excellent research assistance. All errors remain ours. Di Guo and Kun
Jiang acknowledge financial support from RGC Theme-based Research Scheme (TRS) (T31-
717 112-R) and financial support from GRF (HKU 791113B). Yan Guo acknowledges
financial support from Peking University.
39
REFERENCES
Acemoglu, D., Antràs, P., & Helpman, E. (2007). Contracts and technology adoption. The
American economic review, 97 (3), 916-943.
Acemoglu, D., Akcigit, U., Bloom, N., & Kerr, W. R. (2013). Innovation, reallocation and
growth. NBER working paper (No. w18993).
Acemoglu, D., Johnson, S., & Robinson, J. A. (2005). Institutions as a fundamental cause of
long-run growth. Handbook of economic growth, 1, 385-472.
Alder, S., Shao, L., & Zilibotti, F. (2013). The effect of economic reform and industrial
policy in a panel of Chinese cities. Mimeo, University of Zurich.
Aerts, K., & Schmidt, T. (2008). Two for the price of one?: Additionality effects of R&D
subsidies: A comparison between Flanders and Germany. Research Policy, 37(5), 806-822.
Aghion, P., & Howitt, P. (1990). A model of growth through creative destruction. NBER
working paper (No.w3223).
Aghion, P., & Tirole, J. (1997). Formal and real authority in organizations.Journal of
political economy, 105 (1), 1-29.
Aghion P., Akcigit, U., Howitt P. (2013). What do we learn from Schumpeterian growth
theory?. Working paper.
Allen, F., Qian, J and Qian, M. (2005). Law, Finance, and Economic Growth in China.
Journal of Financial Economics, 77(1), 57-116.
Allen, F., Qian, J., Zhang C. and Zhao M. (2012). China’s financial system: opportunities and
challenges. NBER working paper (No. w17828).
Arora, A., Belenzon, S., & Rios, L. A. (2011). The organization of R&D in American
corporations: the determinants and consequences of decentralization. NBER working paper
(No. w17013).
Arrow, K. (1962). Economic welfare and the allocation of resources for invention. In The
rate and direction of inventive activity: Economic and social factors (pp. 609-626). UMI
Audretsch D. B., Link, A. N., & Scott. J. T. (2002). Public/Private Technology Partnerships:
Evaluating SBIR-supported Research. Research Policy. 31(1): 145-158.
Baker, G. P., & Hubbard, T. N. (2004). Contractibility and asset ownership: on-board
computers and governance in US trucking. The Quarterly Journal of Economics, 119(4),
1443-1479.
Brander, J. A., Egan, E., & Hellmann, T. F. (2008). Government sponsored versus private
venture capital: Canadian evidence. NBER working paper, (No. w14029).
Brandt, L., Tombe, T., & Zhu, X. (2013). Factor market distortions across time, space and
sectors in China. Review of Economic Dynamics, 16(1), 39-58.
40
Brandt, L., Van Biesebroeck, J., & Zhang, Y. (2012). Creative accounting or creative
destruction? Firm-level productivity growth in Chinese manufacturing. Journal of
Development Economics, 97(2), 339-351.
Branstetter, L., & Sakakibara, M. (1998). Japanese research consortia: a microeconometric
analysis of industrial policy. The Journal of Industrial Economics, 46(2), 207-233.
Caroli, E., & Van Reenen, J. (2001). Skill-biased organizational change? Evidence from a
panel of British and French establishments. The Quarterly Journal of Economics, 116(4),
1449-1492.
Cheng, L. K., & Kwan, Y. K. (2000). What are the determinants of the location of foreign
direct investment? The Chinese experience. Journal of international economics, 51(2), 379-
400.
Clausen, T. H. (2009). Do subsidies have positive impacts on R&D and innovation activities
at the firm level?. Structural Change and Economic Dynamics, 20(4), 239-253.
Czarnitzki, D., & Lopes Bento, C. (2011). Innovation subsidies: Does the funding source
matter for innovation intensity and performance? Empirical evidence from Germany. ZEW-
Centre for European Economic Research Discussion Paper, (11-053).
David, P. A., Hall, B. H., & Toole, A. A. (2000). Is public R&D a complement or substitute
for private R&D? A review of the econometric evidence. Research Policy, 29(4), 497-529.
Démurger, S., Sachs, J. D., Woo, W. T., Bao, S., & Chang, G. (2002). The relative
contributions of location and preferential policies in China's regional development: being in
the right place and having the right incentives. China Economic Review, 13(4), 444-465.
Dessein, W. (2002). Authority and communication in organizations. The Review of Economic
Studies, 69(4), 811-838.
Dewatripont, M., & Maskin, E. (1995). Credit and efficiency in centralized and decentralized
economies. The Review of Economic Studies, 62(4), 541-555.
Fan G.,Wang X. & Others (2009). Marketization Index for China's Provinces.
National Bureau of Economic Research, Beijing.
Görg, H., & Strobl, E. (2007). The effect of R&D subsidies on private R&D.
Economica, 74(294), 215-234.
Greenwald, B. C., Stiglitz, J. E., & Weiss, A. (1984). Informational imperfections in the
capital market and macro-economic fluctuations. NBER working paper.
Griliches, Z., & Regev, H. (1998). An econometric evaluation of high-tech policy in Israel.
In ATP-conference in Washington, DC.
Guo, D., Jiang, K., Kim, BY., & Xu, C. (2014). The political economy of private firms in
China. Journal of Comparative Economics, 42 (2), 286-303.
Hall, B. H., & Lerner, J. (2009). The financing of R&D and innovation. NBER working paper
(No. w15325).
41
Hart, O. & Moore, J. (2005). On the Design of Hierarchies: Coordination versus
Specialization. Journal of Political Economy: 113(4), 675-702.
42
Sargan, J. D. (1958). The estimation of economic relationships using instrumental
variables. Econometrica: Journal of the Econometric Society, 393-415.
Sah, R., & Stiglitz, J. (1991). The quality of managers in centralized versus decentralized
organizations. Quarterly Journal of Economics, 106(1), 289-295.
Schminke, A., & Van Biesebroeck, J. (2013). Using export market performance to evaluate
regional preferential policies in China. Review of World Economics,149(2), 343-367.
Sianesi, B. (2004). An evaluation of the Swedish system of active labor market programs in
the 1990s.Review of Economics and Statistics, 86 (1): 133-155.
Stein, J. C. (2002). Information production and capital allocation: Decentralized versus
hierarchical firms. The Journal of Finance, 57(5), 1891-1921.
Wallsten, S. J. (2000). The effects of government-industry R&D programs on private R&D:
the case of the Small Business Innovation Research program. RAND Journal of
Economics, 31(1), 82-100.
Wang, J. (2013). The economic impact of special economic zones: Evidence from Chinese
municipalities. Journal of Development Economics, 101, 133-147
Xu, C. (2011). The fundamental institutions of China's reforms and development. Journal of
Economic Literature, 49(4), 1076-1151.
43