Energy Policy 89 (2016) 106–113

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

The impact of government subsidies and enterprises’ R&D investment:

A panel data study from renewable energy in China
Feifei Yu a, Yue Guo a,n, Khuong Le-Nguyen b, Stuart J. Barnes c, Weiting Zhang a
a
Business School of Hohai University, #8 West Focheng Road, Jiangning District, Nanjing, China
b
Cleveland State University, Cleveland, OH, United States
c
King’s College London, UK

H I G H L I G H T S

Government subsidies have a significant crowding-out effect on enterprises’ R&D.

The moderating role of the attributes of company ownership is examined.

A panel threshold regression model is used to explore the influence of subsidy.

First examining the effect of subsidy in the renewable industry in China.

art ic l e i nf o a b s t r a c t

Article history: In this research, we aim to understand the influence of government subsidies on enterprises’ research
Received 3 May 2015 and development (R&D) investment behavior, particularly in China’s renewable energy sector. We are
Received in revised form also interested in examining how the attributes of enterprise ownership act as a moderating variable for
9 September 2015
the relationship between government subsidies and R&D investment behavior. Three classical panel data
Accepted 5 November 2015
analysis models including the pooled ordinary least squares (OLS) model, the fixed effect model and the
Available online 28 November 2015
random effect model are employed. We find that government subsidies have a significant crowding out
Keywords: influence on enterprises’ R&D investment behavior and that the influence is further moderated by the
Renewable energy attributes of enterprise ownership. Moreover, a panel threshold regression model is used to demonstrate
Enterprise
how the influence of government subsidies on enterprises’ R&D investment behavior will change when
Government subsidy
government subsidies increase. Two thresholds, 0.6% and 10.1%, are identified. We recommend that re-
Ownership
R&D investment levant government departments should motivate enterprise R&D investment behavioral intention by
Threshold effect increasing subsidies within a certain range. Different attributes of enterprise ownership should also be
considered as part of policy reform and re-structuring relating to government subsidies.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction increase in China’s domestic energy demand have drawn the at-
tention of various Chinese government departments including the
The G7 countries have formulated a series of policies to support State Council of China, the Bureau of Energy and The Ministry of
firms in the renewable energy industry. For example, U.S. Pre- Environmental Protection of China. The Chinese central govern-
sident Obama released an ambitious 2016 Federal Government ment has recently urged these departments to collaborate to-
Budget Proposal to invest $7.4 billion in clean energy technology gether to consider various policies and specific measures for pro-
programs across all agencies (Laporte, 2015). Japan and the Eur- moting the development of renewable energy.
Since the Renewable Energy Law of the People’s Republic of
opean Commission also published relevant policies and national
China was enacted in 2005, the Chinese government announced a
programs to help renewable energy firms achieve sustainable
series of subsidy policies to boost the renewable energy industry
development (see EU renewable energy policy (2015)). Conse-
(Shen and Luo, 2015). Generally speaking, government subsidies
quently, fierce international energy competition and a quick may be offered through direct or indirect supporting approaches
in terms of funding allocation, such as VAT returns, fiscal subsidies,
n
Correspondence to: Department of Marketing, Business School of Hohai Uni-
tax incentives for innovation, price control, demand assurance and
versity, #8 West Focheng Road, Jiangning District, Nanjing, China. compulsory allocation (Zhang et al., 2014; Shen and Luo, 2015).
E-mail address: [email protected] (Y. Guo). Moreover, government can set up specified subsidies for different

http://dx.doi.org/10.1016/j.enpol.2015.11.009
0301-4215/& 2015 Elsevier Ltd. All rights reserved.
 F. Yu et al. / Energy Policy 89 (2016) 106–113 107

industries and at business development stages, such as subsides Our paper proceeds as follows. The next section reviews the
for the development of renewable energy technologies. In this theoretical foundation for the study. The third section describes
research, the term “government subsidies” mainly refers to public the research methodology employed in the investigation. Section 4
R&D research funding granted by Chinese central and local gov- provides the results of our data analysis and this is followed by a
ernment departments for enterprises to develop new technologies summary of our research findings in Section 5. The final section
of renewable energy and relevant innovative activities. This is our rounds off with a discussion of the contributions and policy im-
research focus because we believe that the development of plications of our research.
emerging and cutting-edge renewable energy technologies plays a
pivotal role in ensuring green, sustainable, renewable energy de-
velopment. Recent studies show that the average research and 2. Literature review
development (R&D) intensity of international renewable energy
enterprises is higher than that of Chinese listed renewable en- In this section we consider government subsidies for renewable
terprises. The average research and development (R&D) intensity technology development, their impact on private enterprises, and
of Chinese listed renewable enterprises is only 0.76% and so the the influence of firm ownership.
main barrier for R&D and innovation in the renewable energy
industry in China is cost (Costa-Campi et al., 2014). 2.1. Government subsidies for renewable energy technology
Despite the growing significance of government subsidies in development
solving the shortage of private R&D investment of renewable en-
ergy enterprises, there is limited understanding of the impact of The effect of government subsidies for technology develop-
government subsidies on renewable energy enterprises’ beha- ment has been extensively examined in prior studies from differ-
vioral intention to strengthen R&D at the micro-level. A systematic ent aspects, such as increasing the rate of green innovation
literature review of prior studies on renewable energy develop- (Johnstone et al., 2010; Aalbers et al., 2013), improving the value of
ment and government subsidies was undertaken that resulted in renewable power technologies (Davis and Owens, 2003), pro-
the identification of a research gap. Existing research on renewable moting the market diffusion of a niche renewable energy tech-
energy development has mainly focused on exploring various nology (Bointner, 2014) and increasing international trades and
energy policy instruments and the effect of portfolio standards on domestic R&D (Kim and Kim, 2015).
climate and energy industry development in developed countries
and regions, such as the United States (Menyah and Wolde-Rufael, 2.2. Differential effect of government subsidies on enterprises’ pri-
2010; Roe et al., 2001; Bang, 2010) and the European Union (Ja- vate R&D investment
cobsson et al., 2009; Haas et al., 2004; Reiche and Bechberger,
2004). We contend that such findings may not be the same for Prior studies show that the actual effect of government sub-
China. A few other researchers have further examined the poten- sidies on improving enterprise private R&D investment con-
tial macroeconomic influence of renewable energy policies on the siderably varies across industries. For instance, Arias and van Beers
energy industry in developing countries (e.g., Zhang et al., 2009; (2013) found a negative relationship between government sub-
Wang et al., 2010; Cherni and Kentish, 2007; Zhao et al., 2011). As sidies and some renewable energy firms’ R&D in terms of the
noted earlier, due to the lack of understanding of the exact effect of amount of renewable energy technology inventions and patents,
government subsidies on enterprises’ intention to strengthen R&D particularly those related to the solar and wind renewable energy
investment at the micro-level, it is still unclear how the R&D level sector. Lach’s (2002) empirical data from Israeli manufacturing
of firms with government subsidies is to be compared with that of enterprises in the 1990s indicated that firms were more like to
firms without government subsidies. Specifically we want to know increase their private R&D investment over the long-term pro-
if the relationship between government subsidies and enterprises’ vided that they could concurrently obtain funding from the Israeli
intention to strengthen R&D investment is linearly correlated and Ministry of Industry and Trade. In other words, government sub-
if government subsides complement or substitute private R&D sidies should be viewed as an alternative funding source instead of
investment. These two questions have important implications for replacing enterprises’ private R&D investment (Koga, 2005). In-
the implementation of appropriate policies of government sub- terestingly, some studies found inconsistent empirical findings
sidies in developing countries. There is a dearth of literature in- where government subsidies have a negative (i.e. crowding-out)
vestigating the above-mentioned issues. (Goolsbee, 1998; Kelette et al., 2000) or limited effect (Wallsten,
In this study, we address the gap in prior research by in- 2000) on enterprises’ private R&D investment. Moreover, some
vestigating the influence of government technology development elements of context (e.g., the background of industries and coun-
subsidies on enterprises’ R&D investment behavior, focusing on try characteristics) have a moderating effect on the relationship
China’s renewable energy sector. Moreover, we are also interested government subsidies and enterprises’ private R&D investment
in examining how the attributes of enterprise ownership moder- (Czarnitzki and Toole, 2007; Görg and Strobl, 2007; Czarnitzki
ate the relationship between government subsidies and en- et al., 2007).
terprises’ R&D investment behavior. Unlike renewable energy en-
terprises in developed countries, Chinese renewable energy en- 2.3. Ownership influence on enterprises’ R&D activities and sub-
terprises inherently have different types of ownership due to the sidies received
unique institutional environment and political background. The
amount of government subsidies acquired by renewable energy During the economic transition process over the past two
manufacturers with different attributes of ownership may vary decades, China has formed a special institution in which company
greatly. Specifically, the more a company has a close relationship ownership characteristics directly affect their R&D performance
with the Chinese central government, the more financial support and subsequent innovation activities. Compared with other kinds
this company can obtain. In addition, we further examine how of enterprises (e.g., private enterprises), state-owned enterprises
these factors – including government subsidies, enterprise own- even stress how to innovate effectively and efficiently because
ership attributes, and political connections of enterprise owners – they have huge political resources that are rather useful for ob-
work together and demonstrate an interactive effect on enterprise taining government subsidies (Yu et al., 2010; Wu and Liu Cheng,
R&D investment. 2011 and Wu et al., 2012).
108 F. Yu et al. / Energy Policy 89 (2016) 106–113

3. Research method (Wind Info1), located in the Lujiazui Financial Center in Shanghai.
Wind Info. is a leading integrated service provider of financial
3.1. Research design data, information, and software. Data with missing key values
were removed and 172 completed cases were finally collected.
We chose to conduct longitudinal research to examine the Moreover, we only consider enterprises that have received sub-
delayed effect of government subsidies on enterprises’ private R&D sidies for at least twelve consecutive half-years to examine the
investment behavior. Such a longitudinal analysis design will allow differential effect of government subsidies on enterprises’ private
us to identify and compare the influence of government subsidies R&D investment. As a result, the financial data for 147 renewable
on enterprises private R&D investment behavioral intention at energy enterprises was finally used in our study.
different stages (Doh and Kim, 2014; Lee and Cin, 2010; David Background information on firms, ownership characteristics
et al., 2000). In this research, we decided to introduce a lag vari- and types of actual firm owners were manually collected from
able demonstrating the effect of government subsidies within 0–1 company annual reports. A further triangulation was carried out
year. Specifically, private R&D investment in year t is influenced by based on information found on the official websites of enterprises.
government subsidies in previous years: t, t  1 and t  2. We hy- All estimations were calculated using Stata 11.
pothesize that government subsidies should have a direct effect on
enterprises’ behavioral intention to increase R&D investment 3.3. Panel regression model
during the first and second half-years after they obtain funding. As
noted earlier, prior studies employing different research methods Our panel regression model is given in Eq. (1). In this section
found that the effect of government subsidies on enterprises R&D we will explain the model.
investment varies. For example, linear regression methods have
been extensively employed to examine the effect of government R&Dputi, t = α1GSi, t + α2GSi, t − 1 + α3GSi, t − 2 + β1incomei, t
subsidies on motivating enterprises’ private R&D investment + β2topsharei, t + β3roai, t + β4 scalenumi, t + χ1CT
(Wolff and Reinthaler, 2008). We argue that such linear regression
methods ignore the moderating role of subsidy intensity on the + χ2 PR + ϕ1GSi, t − 1*CT + ϕ2GSi, t − 2*CT + ϕ3GSi, t*PR
relationship between government subsidies and enterprises’ pri- + ui + εi, t (1)
vate R&D investment behavioral intentions. Therefore, we adopt
the panel threshold regression (PTR) method proposed by Hansen
(1999) to examine the threshold effect of subsidy intensity on
moderating the relationship between government subsidies and 3.3.1. Dependent variable
enterprises’ private R&D investment. R&Dputit represents the private R&D investment of enterprise i
Traditional methods employed by prior studies more or less in the period t. Private R&D investment refers to R&D staff’s in-
assume that government subsidies are randomly allocated to all novative ideas, information and other intangible elements, which
eligible enterprises. However, this is not always the case, because are difficult to accurately measure. However, enterprises have to
policymakers often determine the allocation model of government spend considerable funding to ensure these subjective elements.
funding resources. The amount of subsidies for a firm will be Thus, private R&D investment may be viewed as an indicator of the
significantly influenced by the firm’s ownership attributes and relative strength of enterprises R&D spending. Prior studies on
political background. In order to explore the interactive role of measuring private R&D investment usually use the amount of R&D
government subsidies, we employ a semiannual panel dataset over expenditure of enterprises during a certain period as a dependent
the period from June 2009 to December 2014. A panel data re- variable.
gression is carried out to examine how government subsides in-
fluence an enterprise’s intention to increase private R&D invest- 3.3.2. Explanatory variables
ment toward a positive (i.e. complementary) or negative (i.e. GSi, t represents the amount of government subsidies of en-
crowding-out) effect. In addition, government subsidies usually terprise i in the period t; GSi, t − 1 and GSi, t − 2 represent the amount of
demonstrate a lagged effect on a firm’s innovative activities and government subsidies of enterprise i in the t  1 and t  2 lagged
R&D investment. For that, we add a lag variable for government periods, respectively. Data on government subsides for new tech-
subsidies in the model in this study (see Section 3.3 for the full nology development and transfer were empirically obtained from
model). Moreover, in order to ascertain how Chinese market listed enterprises’ semi-annual reports. incomei, t refers to the
characteristics and the institutional environment influence the amount of business income of enterprise i in the period t;
relationships among firm ownership attributes, firm types and topsharei, t refers to shares of the largest shareholder of enterprise i
government subsidies, a threshold regression model is developed in the period t; roai, t represents the return on assets of enterprise i
for non-dynamic panels with individual-specific fixed-effects. In in the period t, which includes return on total assets, return on
this research, we use this model to examine the differential effects total core assets, return on net assets and return on total net
of government subsidies on enterprises’ private R&D investment worth; scalenumi, t represents the total number of staff of en-
under different subsidy intensities. A time frame is chosen due to terprise i in the period t.
limited high quality data resources. This time frame demonstrates The criteria for identifying these variables are as follows:
key phases of the development of domestic renewable industries
supported by the Chinese government. (1) Considering the effect of government subsidies on the private
R&D investment is usually delayed, we thus introduced GSi, t − 1
3.2. Data and GSi, t − 2 into the panel model, to capture a lagged effect as
they have been used in previous studies (Antonelli and Crespi,
Data was collected from renewable energy enterprises
listed on the Shanghai and Shenzhen stock exchanges. Through
1
semiannual panel data collection over the period from June 2009 Wind Info. has built up a substantial, highly-accurate, first-class financial
database, which includes stocks, funds, bonds, FX, insurance, futures, derivatives,
to December 2014, relevant information on private R&D invest- commodities, and macroeconomic and financial news. Moreover, institutional in-
ment and the amount of government subsidies for renewable vestors can acquire the latest information via regular updates provided by Wind
energy enterprises were provided by Wind Information Co. Ltd Info.
 F. Yu et al. / Energy Policy 89 (2016) 106–113 109

Table 1 subsidies? and (2) Is there a crowding-out or incentive effect on the

Definition of variables. relationship between government subsidies and enterprises’ R&D
investment as the amount of government subsidies change? We as-
R&Dputit The amount of renewable energy enterprise R&D Expenditure
The amount of government R&D subsidies for enterprise i in the
sume that there is a threshold level of r, and the effect of gov-
GSi, t
period t ernment subsidies on improving enterprises’ R&D activities will
GSi, t − 1 The amount of government R&D subsidies for enterprise i in the change significantly when the amount of government subsidies
first lagged period of period t. (Supportit) is more than r. We introduce a dummy variable called
GSi, t − 2 The amount of government R&D subsidies for enterprise i in the
second lagged periods of period t.
Dit . When Supportit ≤ r , Dit ¼1; otherwise, Dit ¼0. We define
incomei, t The amount of business income of enterprise i in the period t It (r ) = {Supportit ≤ r}, of which, I (*) is an indicative function. If
topsharei, t Shares of the largest shareholder of enterprise i in the period t Supportit ≤ r , then I (*) takes value 1, otherwise it is 0.
r o a i, t Return on assets of enterprise i in the period t
In the threshold regression model, variable i represents an
Scalenum the number of employees enterprise i and variable t represents the time period. Support is a
CT Dummy variable that takes 1 representing the weakest political threshold variable and r is a specific threshold value. Variable x is a
connections and 7 representing the strongest political connections set of controlling variables including the number of total staff,
PR Dummy variable that takes 1 if enterprises are state-owned, business income, return on assets, shares of the largest share-
otherwise 0
holder, enterprise ownership attributes, and owners’ political
connections; α1 and α2 respectively refer to the influence ratio of
2013; Montmartin and Herrera, 2015; Colombo et al., 2013); explanatory variables by explained variables of threshold variables
(2) The amount of firms’ private R&D investment is usually in- Supportit when Supportit ≤ r and Supportit > r . ξit refers to sto-
fluenced by firm scale. Researchers often find that firms with a chastic disturbance. It is noted that although our threshold model
large scale have a risk-taking attitude, and are willing to add to only includes one threshold value, two or more threshold values
R&D investment. Thus, we introduced the variable scalenumi, t possibly exist in practice. In this research, we set a double
into the panel model; threshold regression model as:
(3) The amount of R&D private investment is usually constrained R&Dinputit = α0 + α1GSi, t − 1*I (Supportit ≤ r1) + α2GSi, t − 1
by annual income and Return on Assets (ROA), both of which
we assume have a positive effect on R&D private investment. *I (r1 < Supporti, t ≤ r2) + α3GSi, t − 1*I (Supportit > r2)
Thus, we introduced these two variables, incomei, t and roai, t , + θχit + ξit (3)
into the panel model; and
(4) The share of the largest shareholder reflects capital cen- In function r1 o r2 we argue that multiple threshold models can
tralization and the efficiency of resource allocation. We esti- be further developed based on the above single and double
mate that over-centralization of the capital will negatively threshold models.
influence private R&D investment. Thus, we introduced a
variable called topsharei, t into the panel model. The definitions
of the variables are shown in Table 1. 4. Results

3.3.3. Interaction terms 4.1. Descriptive statistical analysis

In this model, we identify three interactions among govern-
ment subsidies, lagged government subsidies, enterprise owner- Descriptive statistics and correlation coefficients are shown in
ship attributes, and owners’ political connections. We believe that Tables 2 and 3. Means, standard deviations, maximum and mini-
firm ownership characteristics and owners’ political connections mum values are summarized for all variables in Table 3. Pairwise
affect the amount of subsidies they can obtain from the Chinese correlation coefficients with significance levels are also provided.
government. In addition, the willingness of renewable energy Moreover, variance inflation factors (VIFs) are used to examine
enterprises to increase private R&D investment in China depends bias estimates derived from multicollinearity, where two or more
on their political resources to some degree. We use a dummy variables in a multiple regression model are highly correlated.
variable, CT, to indicate the degree of owners’ political connections Under this situation, the coefficient estimates of multiple regres-
based on their background scored on a 7-point Likert scale with a sions may change in response to a small change. We computed the
score of 7 indicating ‘the strongest political connections’ and a VIFs and found most to be around 2 and less than the conservative
score of 1 indicating ‘the weakest political connections’. Another threshold of 5. The computed VIFs suggest that multicollinearity
dummy variable, PR, refers to a set of legal rights of a firm and its was not a major issue in our study.
possession of assets (Claessens et al., 2008; Marcelin and Mathur,
2015). This dummy variable takes the value of 1 if a firm is state- 4.2. Regression analysis
owned and 0 otherwise. ui refers to an “enterprise individual ef-
fect” including entrepreneurship, innovation background and According to the basic model (1), we implemented a regression
motivation. εi, t indicates error terms.
Table 2
3.4. Threshold regression model A descriptive statistics.

Our threshold regression model is shown in Eq. (2). Variables Mean Std. dev. Min Max

R&Dinputit = α0 + α1GSi, t − 1*I (Supportit ≤ r ) + α2GSi, t − 1 R&Dinput (ten thousand ) 4103.802 8128.015 0 128959.900
GS (ten thousand) 2289.681 5772.743 0 84270.030
*I (Supporti, t ≥ r ) + θχit + ξit (2) Income (ten thousand) 156790.600 236411.200 0 2,599,373
Topshare (%) 32.876 15.000 0 86.830
We hypothesize that the relationship between government ROA (%) 5.609 5.740  68.064 48.979
subsidies and enterprises’ intentions to add private R&D invest- Scalenum (individuals) 2326.542 2511.418 42 19,682
CT 3.742 1.520 1 7
ment is unlikely to be a linear relationship. Specifically, we ask: (1) PR 0.350 0.477 0 1
Is there an optimal interval for the supporting degree of government
110 F. Yu et al. / Energy Policy 89 (2016) 106–113

Table 3
Pairwise correlations coefficients.

Variables R&Dinput GS GS_lag GS_lag2 Income Scalenum Topshare ROA CT PR CTGSlag2 CTGSlag GSPR

R&Dinput 1.0000
GS 0.400 1.000
0.000
GS_lag 0.214 0.435 1.000
0.000 0.000
GS_lag2 0.385 0.521 0.550 1.000
0.000 0.000 0.000
Income 0.735 0.483 0.273 0.465 1.000
0.000 0.000 0.000 0.000
Scalenum 0.506 0.323 0.328 0.323 0.541 1.000
0.000 0.000 0.000 0.000 0.000
Topshare  0.020 0.161 0.163 0.161 0.129  0.061 1.000
0.421 0.000 0.000 0.000 0.000 0.009
ROA 0.099 0.110  0.027 0.116 0.087  0.053 0.093 1.000
0.000 0.000 0.258 0.000 0.000 0.017 0.000
CT 0.160  0.032  0.034  0.036 0.016 0.114  0.042  0.044 1.000
0.000 0.154 0.149 0.138 0.485 0.000 0.068 0.048
PR 0.143 0.169 0.170 0.171 0.307 0.205 0.231  0.150  0.074 1.000
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001
CTGSlag2 0.544 0.763 0.582 0.627 0.505 0.397 0.113 0.103 0.125 0.148 1.000
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
CTGSlag 0.261 0.590 0.637 0.601 0.338 0.316 0.204  0.031  0.004 0.282 0.539 1.000
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.188 0.876 0.000 0.000
GSPR 0.387 0.744 0.486 0.677 0.560 0.234 0.251 0.056  0.055 0.378 0.585 0.588 1.000
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.013 0.013 0.000 0.000 0.000

Table 4 Table 5
Regression results of property nature, government subsidy and innovation invest- Government subsidy effect on R&D investment of enterprises with different
ment of new energy enterprises. ownerships.

Variable Heteroscedasticity robust model Variable (R&D State-owned group Non state-owned group
investment) (PR ¼1) (PR ¼0)
GS  0.038 (0.652)
GSi, t − 1  0.660*** GS 0.533***  0.023(0.320)
GSi, t − 2  1.098 *** GS_lag1  1.396***  0.355***
GS_lag2  1.041***  0.832**
incomei, t 0.022***
Income 0.018*** 0.028***
topsharei, t  48.441* Topshare  3.502(0.967)  37.700*
roait 31.233(0.212) roa  104.530(0.302) 2.137(0.889)
scalenumi, t 1.105*** Scalenum 1.349*** 0.683***
CT  700.565*** CTGSlag 0.252*** 0.071**
PR  2135.144*** CT 1563.484*  1030.986***
GSi, t − 2*CT 0.209*** CTGSlag2 0.201*** 0.169**
_cons  14891.8*** 6592.272***
GSi, t − 1*CT 0.170***
R-squared 0.6932 0.6498
GS*PR 0.341*
Adjusted R-squared 0.6861 0.6462
_cons 3528.682*
N 444 1011
R-squared 0.664
F 61.45 109.02
Adjusted R-squared 0.661
N 1764 *
Represents po 0.1.
F 53.88 **
p o 0.05.
***
p o0.01.
Note: The estimation method of is heteroscedastisity robust type model.
**p o 0.05.
*
p o 0.1.
crowding-out effect was found at the 1% level of significance. An
***
p o 0.01. increase in government subsidies by 1% will lead to 0.66% and
1.098% reductions in private R&D investment during periods t  1
analysis based on panel data for new energy enterprises from 2009 and t 2 respectively. Such a significant lagged crowing-out effect
to 2014 as shown in Table 4. A Hausman test was carried out means that Chinese renewable energy firms are not willing to take
(p ¼0.0243) and indicated that a fixed effect model should be risks in initial R&D stages because of high costs. Compared to
adopted. developed countries, the Chinese government acts as a key driver
A White’s (1980) test for heteroscedasticity showed positive for the renewable energy sector and offers the majority of R&D
results (p o0.001). Thus, compared with a fixed effects model, a investment for enterprises. Thus, the Chinese government needs
heteroscedasticity robust model could provide a more accurate to consider appropriate policies to encourage enterprises’ inten-
estimation of the relationship between government subsidies and tions to increase the funding of R&D.
the private R&D investment of energy enterprises ( Table 5). Our research findings showed that the relationship between
As shown in Table 4, the coefficient for government subsidies enterprises’ property rights and government subsidies is sig-
and enterprises’ private investment in the current period is nificant at the 1% level. The relationship between owners’ political
0.038. This result means that if the government increases sub- networks and government subsidies is also significant in the t 1
sidies by 1%, renewable energy enterprises are likely to reduce and t  2 lagged periods (p o0.01). This means that owners’ per-
their private R&D investment by 0.038%. Moreover, a lagged sonal political connections play a role in obtaining government
 F. Yu et al. / Energy Policy 89 (2016) 106–113 111

subsidies. In practice, enterprises do not have an equal opportunity Table 7

to obtain government funding, particularly in China where the Estimation results of threshold value.
institutional environment is distinctive. Further, we found that a
Threshold value Estimation value 95% confidence interval
firm’s political background and network also demonstrates an
important role in obtaining other types of funding, such as bank First one 0.006 (0.005 0.006)
loans (Li et al., 2008). As a result, more and more Chinese firms in Second one 0.101 (0.101 0.101)
the renewable energy industry are willing to spend considerable nnn
p o 0.01
effort in establishing relationships with different Chinese gov-
ernment departments.
Table 8
In order to further examine the influence of firm ownership
Estimation results of threshold effect of government subsidies on firms’ R&D
attributes, we divided the sample into two groups (state-owned investment.
and non-state-owned). Through descriptive statistics, we found
that the average amount of government subsidies for the state- Explanatory value (R&D investment) Threshold model
owned group and the non-state-owned group are 3637.336 and
Income 0.0263***
1580.175 ten thousand Chinese Yuan, respectively. The average Scalenum 1.445***
intensity of government subsidies is similar in both groups: 0.010 Topshare  9.251
and 0.019 for the state-owned group and for the non state-owned roa 0.602
group, respectively. Thus, our research findings suggest that Govsupport 0.203***
CT  299.267**
ownership attributes have a weak effect on helping firms to obtain
PR  931.558
government subsidies. GS_lag_1  0.315***
We found that the nature of firm ownership directly de- GS_lag _2 0.139***
termines the effect of government subsidies on firms’ R&D in- GS_lag _3  0.259***
vestment behavior. For example, for a state-owned new energy _cons  593.595
F value 284.590
firm, the incentive effect of government subsidies on the firm’s Fixed effects test F¼ 10.30 Prob 4F¼ 0.0000
R&D activities in the current period is 0.533, with a significant
level of 1%. However, this effect was significantly reduced for a
non-state-owned new energy firm. Our empirical findings found intensity exceeded 10.1%. In addition, prior marginal effect reaches
that lagged government subsidies for a firm during one period maximum value. We see that when government subsidy intensity
have a significant crowding-out effect on the firm's R&D invest- exceeded 10.1%, the government subsidies demonstrated a
ment:  1.396 for the state-owned group and  0.355 for the non- crowding-out effect again on the firm’s R&D investment activities
state-owned group (p o0.01). The different effects (i.e. incentive and the contribution degree was 0.259*** (Table 8).
vs. crowding-out) for these two types of firms may be due to or-
ganizational heterogeneity.
5. Discussion
4.3. Results of the threshold estimation
The results of our data analysis show that the effect of gov-
Estimation of two threshold values is shown in Table 6. Single ernment subsidies on increasing enterprises’ intention to invest in
and double threshold effects were both found at the significant R&D is an inverted-U relationship. Moreover, firm owners will play
level of 1%, while the triple threshold effect was not significant. an important role in helping a firm to obtain government subsidies
Threshold values of government subsidy intensity (i.e. 0.6% and
if they have substantial political resources. Thus, in the renewable
10.1%) were achieved by bootstrap estimation. According to the
energy industry, we suggest that enterprises with poor political
two threshold values, government subsidies could be divided into
backgrounds should actively foster R&D activities and improve
three groups including low subsidy (less than 0.6%), medium
their own innovative capabilities, which may help to overcome the
subsidy (greater than 0.6% and less than 10.1%) and high subsidy
discrimination of government in obtaining funding. To increase
(greater than 10.1%).
the effectiveness of government subsidies, we recommend that
According to the three intervals divided by each threshold, the
governments should improve funding allocation mechanisms and
contribution of the degree of government subsidies for firms’ in-
place an emphasis on assessing applicant’s research competency
novation investment was significant at the 5% level (see Table 7).
and innovative capability. Third, the two identified threshold ef-
When the intensity of government subsidy was less than 0.6%,
fects (i.e. 0.6% and 10.1%) confirm the non-linear relationship be-
government funding support behavior showed a significant
tween the intensity of government subsidies and its practical ef-
crowding-out effect (i.e. 0.315, po 0.001) on a firm’s R&D in-
fect on increasing a firm’s intention to do private R&D activities. In
vestment activities associated with slight marginal effects. Our
sum, our research findings fully support our hypotheses. Govern-
empirical data show that if government subsidy intensity reached
0.6%, the crowding-out effect disappeared and a significant in- ment subsidies create a crowding-out effect followed by an in-
centive effect of government subsidies (i.e. 0.139***) on the firm’s centive effect. A crowding-out effect will appear again once the
R&D investment activities was created until government subsidy government subsidy intensity increases. This means that govern-
ment should consider how to develop an appropriate subsidy
Table 6 policy for renewable energy enterprises, adapting to the dynamic
Bootstrap of threshold effects. characteristics of the subsidy effects. The Chinese government
needs to avoid the ‘Matthew effect’ (Merton, 1968; Antonelli and
Threshold model F value P value Critical value
Crespi, 2013) when implementing government subsidies.
1% 5% 10%

Single 86.365*** 0.002 34.624 9.019 4.272 6. Conclusions and policy implications
Double 58.671*** 0.000 9.912 3.501 2.108
Triple 34.922 0.112 10.509 3.299 2.030
This research makes several contributions, from both practical
112 F. Yu et al. / Energy Policy 89 (2016) 106–113

and theoretical perspectives. First, we examine the effect of gov- and business background of enterprises, the Chinese government
ernment subsidies on micro innovative activities of renewable should apply appropriate government subsidy policies to en-
energy enterprises, and find that government subsidies have a terprises with different business characteristics (e.g., a firm’s
significant crowding-out influence on enterprises’ R&D investment business scale and development stage). For example, compared
behavior. Second, the moderating role of the political and in- with private firms, the effect of government subsidies on in-
stitutional background of enterprises in the relationship between creasing a university-run enterprise’s intention for R&D invest-
government subsidies and enterprise R&D investment is ex- ment should be obvious. This may be attributed to the fact that
amined. Our research shows that the relationship between a firm’s such kinds of firms (i.e. university-run enterprises) have a strong
ownership and obtained government subsidies reaches a statisti- scientific research foundation and resources. Thus, it seems ap-
cally significant level of 10% with a 0.341 value. The relationship propriate for governments to place an emphasis on increasing
between firm owners’ political resources and government sub- funding support for university-run enterprises.
sidies is at a significant level of 1% during the t  1 and t  2 lagged This research also has some implications for policy im-
periods. Third, this research employs the non-standard asymptotic plementations from the perspective of incentive mechanisms.
theory of inference and the bootstrap method to examine how the According to our research findings, government subsidies indeed
influence of government subsidies on new energy enterprises’ help enterprises that experience a serious shortage of specialized
R&D investment varies in accordance with the supporting degree research funding. Our study shows that enterprises with strong
of subsidies. The effect of government subsidies on increasing political background are likely to receive considerable government
enterprises’ intention to invest in R&D demonstrates an inverted-U subsidies. It is necessary to improve current supervision me-
relationship. chanisms in order to reduce the possibility of senior managers’
There are several limitations in this research. First, we only rent-seeking behavior. On the one hand, ensuring that the usage of
employ a semiannual panel dataset over the period from June subsidies is in accordance with relevant provisions and that in-
2009 to December 2014. Our research is based on this short panel formation on government subsidies is fully disclosed to the public
dataset since the Wind database only records R&D data for the is imperative for the public supervision. Relevant government
sample since 2009. Second, we do not further examine the in- departments need to evaluate the efficiency of annual government
centive effect of subsidies on stimulating private R&D investment subsidies, and keep on examining the usage of government sub-
based on different subsidy types. We believe that such a com- sidies. In order to reduce the possibilities of rent-seeking behavior,
parison among different kinds of subsidies is a possible avenue for government departments should promote collective decision-
future research. making instead of individual decision-making.
The Chinese government is playing a pivotal role in promoting The last contribution to policy of our research is how to identify
the development of renewable energy. An appropriate im- suitable subsidy intensity. Our study shows that government
plementation of subsidy policies is essential for the growth and subsidies do not always play an active role in improving the R&D
innovation of strategic emerging industries (SEIs). Some studies capability of a firm, particularly in the later development stage of
suggest that government subsidies have a crowding-out influence the firm. In other words, our research findings suggest that gov-
on enterprises’ R&D investment while others noted that govern- ernment should gradually reduce subsidy for a firm’s R&D activ-
ment subsidies have a positive impact on enterprises’ own R&D ities in the renewable energy industry as the firm develops. Other
investment. In this research, we employ estimations of pooled OLS, subsidy information including supporting duration and expected
fixed effect, and random effect models to examine government outcomes should be made clear as soon as possible.
subsidies’ influence on the R&D investment of renewable energy
enterprises. The moderating effect of ownership attributes on the
above relationship was examined as well. We employed a panel Acknowledgments
threshold model to investigate the threshold effects of subsidy
intensity on the relationship between government subsidies and
This word is supported by Hohai University’s Basic Scientific
enterprises’ R&D investment.
Research Fund for Central Colleges and Universities (Project no.
A panel data consisted of 147 Chinese renewable energy en-
2013B12114), Fundamental Research Funds for the Central Uni-
terprises from 2009 to 2014. Our empirical findings show that
versities (Project no. 2014B18914), the youth project of Humanities
government subsidies have a significant crowding-out effect on
and Social Science Research Fund of the Ministry of Education
new energy enterprises and have no incentive effect. Second, the
(Project no. 12YJC630273) and the youth project of National Social
moderating effect of enterprise owners’ background character-
Science Fund (Project no. 12CGL010).
istics on the relationship between government subsidies and en-
terprises’ R&D investment was significant. Third, for enterprises
with different ownership attributes, government subsidies have
different effects on firms’ private R&D investment. For example, References
government subsidies have a remarkable incentive effect on state-
owned renewable energy enterprises’ R&D investment. However, a Aalbers, R., Shestalova, V., Kocsis, V., 2013. Innovation policy for directing technical
change in the power sector. Energy Policy 63C, 1240–1250.
notable crowding-out effect for government subsidies on state- Antonelli, C., Crespi, F., 2013. The “Matthew effect” in R&D public subsidies: the
owned renewable energy enterprises’ intention to increase private Italian evidence. Technol. Forecast. Soc. Chang. 80 (8), 1523–1534.
R&D investment was found. Finally, estimation results of the panel Arias, A.D., van Beers, C., 2013. Energy subsidies, structure of electricity prices and
technological change of energy use. Energy Econ. 40 (2), 495–502.
threshold model show that government subsidies have a nonlinear Bang, G., 2010. Energy security and climate change concerns: triggers for energy
relationship with enterprise innovation investment. The influence policy change in the United States? Energy Policy 38 (4), 1645–1653.
of government subsidies on enterprise innovation investment will Bointner, R., 2014. Innovation in the energy sector: lessons learnt from R&D ex-
penditures and patents in selected IEA countries. Energy Policy 73, 733–747.
vary in accordance with government subsidy intensity. Generous
Cherni, J.A., Kentish, J., 2007. Renewable energy policy and electricity market re-
government subsidies do not always result in incentive effects on forms in China. Energy Policy 35 (7), 3616–3629.
renewable energy enterprises’ private R&D investment. Claessens, S., Feijen, E., Laeven, L., 2008. Political connections and preferential ac-
cess to finance: the role of campaign contributions. J. Financ. Econ. 88 (3),
Our empirical findings provide some policy implications. Con-
554–580.
sidering a positive or negative effect on motivating renewable Colombo, M.G., Croce, A., Guerini, M., 2013. The effect of public subsidies on firms’
energy enterprises’ R&D investment with the change of ownership investment–cash flow sensitivity: transient or persistent? Res. Policy 42 (9),
 F. Yu et al. / Energy Policy 89 (2016) 106–113 113

1605–1623. R&D investment: empirical evidence from Korea. Technol. Forecast. Soc. Chang.
Costa-Campi, M.T., Duch-Brown, N., García-Quevedo, J., 2014. R&D drivers and ob- 77 (6), 881–890.
stacles to innovation in the energy industry. Energy Econ. 46, 20–30. Li, H., Meng, L., Wang, Q., Zhou, L.A., 2008. Political connections, financing and firm
Czarnitzki, D., Toole, A.A., 2007. Business R&D and the interplay of R&D subsidies performance: evidence from Chinese private firms. J. Dev. Econ. 87 (2),
and product market uncertainty. Rev. Ind. Organ. 31 (3), 169–181. 283–299.
Czarnitzki, D., Ebersberger, B., Fier, A., 2007. The relationship between R&D colla- Marcelin, I., Mathur, I., 2015. Privatization, financial development, property rights
boration, subsidies and R&D performance: empirical evidence from Finland and and growth. J. Bank. Financ. 50, 528–546.
Germany. J. Appl. Econom. 22 (7), 1347–1366. Menyah, K., Wolde-Rufael, Y., 2010. CO2 emissions, nuclear energy, renewable en-
David, P.A., Hall, B.H., Toole, A.A., 2000. Is public R&D a complement or substitute ergy and economic growth in the US. Energy Policy 38 (6), 2911–2915.
for private R&D? A review of the econometric evidence. Res. Policy 29 (4), Merton, R.K., 1968. The Matthew Effect in science. Science 159 (3810), 56–63.
497–529. Montmartin, B., Herrera, M., 2015. Internal and external effects of R&D subsidies
Davis, G.A., Owens, B., 2003. Optimizing the level of renewable electric R&D ex- and fiscal incentives: empirical evidence using spatial dynamic panel models.
penditures using real options analysis. Energy Policy 31 (15), 1589–1608. Res. Policy 44 (5), 1065–1079.
Doh, S., Kim, B., 2014. Government Support for SME innovations in the regional Reiche, D., Bechberger, M., 2004. Policy differences in the promotion of renewable
industries: the case of government financial support program in South Korea. energies in the EU member states. Energy Policy 32 (7), 843–849.
Res. Policy 43 (9), 1557–1569. Roe, B., Teisl, M.F., Levy, A., Russell, M., 2001. US consumers’ willingness to pay for
EU renewable energy policy, 2015, April 29. Retrieved from 〈http://www.euractiv. green electricity. Energy Policy 29 (11), 917–925.
com/energy/eu-renewable-energy-policy/article-117536〉. Shen, J., Luo, C., 2015. Overall review of renewable energy subsidy policies in China
Görg, H., Strobl, E., 2007. The effect of R&D subsidies on private R&D. Economica 74 – contradictions of intentions and effects. Renew. Sustain. Energy Rev. 41(C),
(294), 215–234.
1478–1488.
Goolsbee, A., 1998. Does government R&D policy mainly benefit scientists and
Wallsten, S.J., 2000. The effects of government-industry R&D programs on private
engineers? Am. Econ. Rev. 88 (2), 298–302.
R&D: the case of the Small Business Innovation Research program. RAND J.
Haas, R., Eichhammer, W., Huber, C., Langniss, O., Lorenzoni, A., Madlener, R.,
Econ. 31 (1), 82–100.
Verbruggen, A., 2004. How to promote renewable energy systems successfully
Wang, F., Yin, H., Li, S., 2010. China’s renewable energy policy: commitments and
and effectively. Energy Policy 32 (6), 833–839.
challenges. Energy Policy 38 (4), 1872–1878.
Hansen, B.E., 1999. Threshold effects in non-dynamic panels: estimation, testing,
White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a
and inference. J. Econom. 93 (2), 345–368.
direct test for heteroskedasticity. Econometrica: J. Econ. Soc. 48 (4), 817–838.
Jacobsson, S., Bergek, A., Finon, D., Lauber, V., Mitchell, C., Toke, D., Verbruggen, A.,
Wolff, G.B., Reinthaler, V., 2008. The effectiveness of subsidies revisited: accounting
2009. EU renewable energy support policy: faith or facts? Energy Policy 37 (6),
for wage and employment effects in business R&D. Res. Policy 37 (8),
2143–2146.
Johnstone, N., Haščič, I., Popp, D., 2010. Renewable energy policies and technolo- 1403–1412.
gical innovation: evidence based on patent counts. Environ. Resour. Econ. 45 Wu, J., Liu Cheng, M., 2011. The impact of managerial political connections and
(1), 133–155. quality on government subsidies: evidence from Chinese listed firms. Chin.
Kelette, T.J., Møen, J., Griliches, Z., 2000. Do subsidies to commercial R&D reduce Manag. Stud. 5 (2), 207–226.
market failures? Microeconometric evaluation studies. Res. Policy 29 (4), Wu, W., Wu, C., Rui, O.M., 2012. Ownership and the value of political connections:
471–495. evidence from China. Eur. Financ. Manag. 18 (4), 695–729.
Kim, K., Kim, Y., 2015. Role of policy in innovation and international trade of re- Yu, M.G., Hui, Y.F., Pan, H.B., 2010. Political connections, rent seeking, and the fiscal
newable energy technology: empirical study of solar PV and wind power subsidy efficiency of local governments. Econ. Res. J. 3, 65–77.
technology. Renew. Sustain. Energy Rev. 44, 717–727. Zhang, Huiming, Li, Lianshui, Zhou, Dequn, Zhou, Peng, 2014. Political connections,
Koga, T., 2005. R&D subsidy and self-financed R&D: The case of Japanese high- government subsidies and firm financial performance: Evidence from renew-
technology start-ups. Small Bus. Econ. 24 (1), 53–62. able energy manufacturing in China. Renew. Energy 63, 330–336.
Lach, S., 2002. Do R&D subsidies stimulate or displace private R&D? Evidence from Zhang, P.D., Yang., Y.L., Jin., S., Zheng, Y.H., Wang., L.S., Li, X.R., 2009. Opportunities
Israel. J. Ind. Econ. 50 (4), 369–390. and challenges for renewable energy policy in China. Renew. Sustain. Energy
Laporte, A., 2015. Issue brief: Obama FY2016 Budget Proposal: Sustainable Energy, Rev. 13 (2), 439–449.
Buildings, Transportation and Climate (2015, April 29). Retrieved from 〈http:// Zhao, Z.Y., Zuo, J., Fan, L.L., Zillante, G., 2011. Impacts of renewable energy regula-
www.eesi.org/papers/view/issue-brief-2016-budget〉. tions on the structure of power generation in China – a critical analysis. Renew.
Lee, E.Y., Cin, B.C., 2010. The effect of risk-sharing government subsidy on corporate Energy 36 (1), 24–30.