Advances in Economics, Business and Management Research, volume 71

4th International Conference on Economics, Management, Law and Education (EMLE 2018)

Government Subsidy, Research and Development

Expenditure and Independent Innovation of
Enterprise*

Xiaoxu Kong
College of Business Administration
Capital University of Economics and Business
Beijing, China 100070

He Zhu** Yang Yu
Peking University HSBC Business School Research Institute of Maritime Silk-Road (RIMS)
Beijing, China 100872 Peking University HSBC Business School
**Corresponding Author Beijing, China

Abstract—In recent years, the government's emphasis on the effect of government R&D investment from a macro
corporate innovation and policy support has been increasing. perspective are relatively uniform. Most studies believe that
It is necessary to analyze the effectiveness of government's the government R&D investment has an inductive effect on
innovation support policies. Based on the sample of 108 A- corporate R&D expenditure. However, at the micro level,
share listed companies from 2010 to 2014 and the GMM due to certain differences in sample selection and research
method of dynamic panel data, this paper examines the role of methods, the unified research paradigm has not yet formed,
government subsidies for enterprise research and development leading to inconsistencies in the conclusions of the study.
expenditures and independent innovation results. According to The existing academic achievements indicate that the
this study, we can know that after controlling the enterprise
government's innovation support policy will generate
scale and the number of scientific research personnel,
government subsidies can significantly stimulate the research
innovation incentives for enterprises. For example, Xu Zhi
and development expenditure of the enterprise. The impact of and Shi Ping (2005) found that the government's technology
government subsidies on the number of inventions and the investment has significantly promoted corporate expenditure
number of designs is significantly positive. However, the during the sample period. However, some scholars have
impact on the number of utility models is not significant. This analyzed that government research and innovation subsidies
paper also analyzes the efficiency of government's innovation have an "extrusion" effect on R&D investment (Lv Jiuqin,
support with the DEA method. For the sample enterprises, the Yu Dandan, 2011).
government innovation support policy has certain inefficiency.
There are two main reasons why the above research
Keywords—government subsidy; enterprise research and conclusions are so different. At the methodological level, due
development expenditure; enterprise independent innovation; to the different choices of measurement techniques, the
GMM method of dynamic panel research conclusions are not stable. At present, the academic
community mainly uses three types of methods to study the
effectiveness of government innovation support policies. The
I. INTRODUCTION first category is the least squares method, including ordinary
In recent years, under the guidance of the macro-strategy least squares method, two-stage least squares method, etc.,
of the national innovation system, the Chinese government such as the study of Yang Dewei and Tang Xiangxi (2011).
has invested a large amount of funds each year to encourage The second category is the instrumental variable method.
innovation. The main investment is to support innovative The representatives are the studies of Xing Fei and Zhang
activities and technology-based science and technology Jianhua (2009) and Zhang Jie et al. (2015). The third
service activities. As the government attaches more category is the application of differential GMM or system
importance to corporate innovation and policy support, it is GMM for panel data analysis, such as the studies of Xie
necessary to analyze the effectiveness of government Weimin et al (2009) and Yang Xiangyang (2014). Second,
innovation support policies. the conclusions caused by the different sample data selection
are not stable. Scholars analyze the sample of enterprises in
The conclusions of scholars at home and abroad studying
different regions and industries, and the differences in the
*Fund Project: The research on the short-term effects of environmental use of government innovation support policies by enterprises
protection inspectors on the Chinese economy commissioned by the
Ministry of Ecology and Environment of the People's Republic of China.

Copyright © 2018, the Authors. Published by Atlantis Press.

This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). 253
 Advances in Economics, Business and Management Research, volume 71

within the region and industry will inevitably affect the and the analysis of the role of government subsidies on
research conclusions. independent innovation of enterprises. The fifth is the
analysis of government innovation support efficiency based
This paper believes that the GMM method of application on DEA method. The sixth part summarizes the research
dynamic panel data has incomparable advantages over other conclusions of the full text.
methods in terms of the impact of government innovation
support policies on enterprise innovation motivation and
innovation results. First, the regression analysis using the II. THE LOGIC MECHANISM AND RESEARCH HYPOTHESIS
ordinary least squares (OLS) method can reflect the OF GOVERNMENT SUBSIDY TO STIMULATE ENTERPRISE
influence of government support on enterprise innovation. INNOVATION
However, it does not consider the estimation bias caused by First of all, from the historical evolution of enterprise
the endogeneity of variables. Second, the instrumental technology innovation activities, the knowledge/technology
variable method is scientifically effective. However, it is in the innovation process has certain externalities. Especially,
difficult to achieve at the actual operational level. And the for some basic research, its social benefits far outweigh the
selection of instrumental variables will have an impact on the private benefits, and the role of the government in it is
research results to a large extent. Third, the GMM method of particularly important. The government's incentive subsidy
dynamic panel data can solve the relevant problems of for enterprises reduces the innovation cost of enterprises,
unobservable variables and interpretation variables, or the reduces the gap between private income and social benefits
bias caused by the omission of some important explanatory of innovation activities, and increases the return on
variables to some extent. It is especially important for investment of enterprises engaged in innovative activities,
empirical testing. Based on the above reasons, this paper stimulating enterprises to increase R&D expenditure.
chooses to apply the GMM method of dynamic panel data to Especially for high-tech enterprises listed on the GEM, the
study the impact of government innovation support policies scale and other restrictions make the R&D projects of
on enterprise innovation. enterprises have high investment and high risks, and the
In addition, previous academic research focused on the R&D funds are short. If they can obtain government
impact of government support on corporate R&D subsidies, they will generally expand the R&D scale and
expenditures (ie, the motivation for corporate innovation). increase R&D investment. Therefore, the hypothesis 1 of this
However, these studies did not pay much attention to the paper is proposed. There is significant positive correlation
research and the results of independent innovation. Therefore, between government subsidies and R&D expenditures of
this paper adopts the GMM method of dynamic panel data to enterprises. That is to say, government subsidies will
examine the role of government subsidies for corporate R&D stimulate the increase of R&D expenditures of enterprises.
expenditures and independent innovation results. According Secondly, if Hypothesis 1 is set up, after receiving
to this study, we can know that after controlling the government subsidies, the company will increase the R&D
enterprise scale and the number of scientific research investment for innovation activities. In the long run, it will
personnel, government subsidies can significantly stimulate bring certain innovations, which are reflected in the increase
the research and development expenditure of the enterprise. of invention number, utility models and designs. Therefore,
The impact of government subsidies on the number of the research hypothesis 2 of this paper is proposed. The
inventions and the number of designs is significantly positive. government subsidies will have a significant positive impact
However, the impact on the number of utility models is not on the independent innovation results of enterprises.
significant. This paper also analyzes the efficiency of Combined with the statistics in the patent research database
government's innovation support with the DEA method. For of CAMAR China listed companies, the enterprise
the sample enterprises, the government innovation support innovation results can be represented by three indicators:
policy has certain inefficiency. invention number, utility model number and design number.
The innovations are reflected in the following two This paper has designed three second-level hypotheses as the
aspects. First, it should pay attention to the effect of following. Hypothesis 2a: government subsidy will increase
government innovation support policies on R&D the number of inventions significantly. Hypothesis 2b:
expenditures of enterprises, and also focus on the impact of government subsidies will significantly increase the number
government subsidies on independent innovation of utility models. Hypothesis 2c: government subsidies will
achievements (ie patents, utility models and designs). Second, significantly increase the number of corporate designs.
the author also uses the DEA efficiency analysis method to Finally, the government's support for the technology
study the inefficiency of the government's innovation support innovation activities is beneficial. For some industries and
policy, which helps to guide the rational allocation of enterprises, the government's innovation support is
government innovation support at the micro level. increasing. However, it has not brought about a significant
The rest are as the following. The second is the logical increase in the enterprise innovation and independent
mechanism and research hypothesis of government subsidies innovation results. It is mainly influenced by the different
to stimulate enterprise innovation. The third is research preferences of the government and enterprises on technology.
design. And the fourth is empirical analysis based on This difference in preferences ultimately led to the
dynamic panel data, including the analysis of the role of government's support not meeting expectations, and the
government subsidies on R&D expenditure of enterprises, policy effect was not significant. From the perspective of
input and output, it can be considered that the government's

254
 Advances in Economics, Business and Management Research, volume 71

innovation support policy is less efficient. Therefore, the introduce lag items to explain these economic relations. The
research hypothesis 3 of this paper is proposed. The dynamic panel data model, that is, the lagging term of the
government innovation incentive policy has certain explained variable is included in the interpretation term of
inefficiency. the panel data model to reflect the dynamic lag effect. This
paper designs a dynamic panel data model. The basic form
III. RESEARCH DESIGN of the dynamic panel data model is as follows:
A. Model Setting and Sample Data:
1) Model setting: In the real world, many economic
relations are dynamic. Sometimes it is necessary to
n
Yit = Yit-1    xit
i=1
i
 it

Yit is the explained variable; xit is the explanatory variable; i is the coefficient;  it stochastic disturbance term.
Under the analytical framework of this paper, variables designs each year. Xit is innovative support for the enterprise.
are given practical meaning. Y1it is the enterprise R&D i represents the enterprise. t represents time. t-1 represents
expenditure. Y2it, Y3it, and Y4it are the independent lagging period. i is the parameter to be estimated. The
innovation achievements of the enterprise, representing the
number of newly applied inventions, utility models and empirical analysis model of this paper is obtained as follows:
 Model 1: 模型1：RDit  RDit 1  1GOVSit  2SCALit  3NUMSit  it

 Model 2a: 模型2a：INVENTION it  RDit 1  1GOVSit  2SCALit  3NUMSit  it

 Model 2b: 模型2b：UTILITYit  RDit 1  1GOVSit  2SCALit  3NUMSit  it

 Model 2c: 模型2c：DESIGN it  RDit 1  1GOVSit  2SCALit  3NUMSit  it

In the above model, RD stands for enterprise R&D to bias in the estimation results. From the actual situation,
investment. INVENTION, UTILITY and DESIGN represent when the government department puts the subsidy to the
the number of newly applied inventions, utility models and enterprise, it usually needs to follow a certain process, and
designs each year. GOVS represents the government's R&D the actual fund allocation often needs certain time. Under
subsidy for enterprise innovation. SCAL represents the scale
this circumstance, although the government subsidies can be
of the company. NUMS represents the number of corporate
R&D personnel. observed in the year, there has different degrees of time lag
in the real disposable funds of corporate. To accurately
B. Data Source and Variable Definition estimate the impact of government subsidies, this paper
refers to the study of Yang Xiangyang and Tong Xinle
The basic data used in this paper mainly comes from
CSMAR database and RESSET financial research database. (2014), and adopts the value of lagging periods of
The sample period spans from 2010 to 2014. After removing government subsidy.
the enterprise samples with incomplete data information, the 4) Control variables: Regarding the control variables,
valid sample is 108 A-share listed company. this paper refers to the practice of similar literature, and
considers the scale of the enterprise and the number of R&D
1) Enterprise R&D expenditure: The enterprise R&D personnel. The scale of the enterprise is represented by the
expenditure is the R&D expenses incurred by the enterprise total assets in the listed company's balance sheet. The
for the scientific and technological innovation activities. specific data comes from the financial statement database of
The specific data comes from the patent research database CSMAR China listed companies. The number of R&D
of CSMAR China listed company. personnel in the company is based on the number of people
2) The independent innovation results of enterprises: engaged in scientific research and technology research and
The independent innovation results of enterprises include development published in the annual report of the listed
the number of inventions, utility models and designs. The company. The data comes from RESSET Financial
specific data comes from the patent research database of Research Database.
CSMAR China listed company.
3) Government subsidy: The specific data comes from C. Descriptive Statistics of Variables
the database of financial statements of CSMAR China listed This paper use Eviews 7.0 software to perform the
companies. It should be pointed out that the direct use of descriptive statistical analysis of the sequence groups. The
current government subsidies for empirical testing may lead basic descriptive statistics of each variable are obtained as

255
 Advances in Economics, Business and Management Research, volume 71

shown in "Table I". From "Table I", we can know that from number of employees engaged in research and development
2010 to 2014, the average R&D expenditure of the sample work is 748.88810. From the bias of the median and the
company is 43,139,649 yuan, and the mean of newly applied average, in addition to the concentrated distribution of utility
inventions, utility models and designs is 52 pieces, 15 pieces models and designs, the distribution of the remaining
and 4 pieces per year, respectively. The mean of government variables is relatively scattered. In summary, it can be seen
subsidies is 53,441,204 yuan. The mean of assets from the descriptive statistics that at the level of each
representing enterprise scale is 4.58E+09 yuan. The average variable, the gap among sample companies is large.

TABLE I. DESCRIPTIVE STATISTICAL ANALYSIS OF VARIABLES

Mean Median Maximum Minimum value Standard deviation

RD 43139649 0.000000 3.48E+09 0.000000 2.58E+08
INVENTION 51.89591 3.000000 5241.000 0.000000 370.6437
UTILITY 14.81413 2.000000 749.0000 0.000000 62.59734
DESIGN 3.973978 0.000000 134.0000 0.000000 15.09360
GOVS 53441204 9348245. 2.37E+09 0.000000 2.07E+08
SCAL 4.58E+09 1.57E+09 1.09E+11 1.30E+08 1.27E+10
NUMS 748.8810 216.0000 30876.00 25.00000 2820.888

In order to further verify the change of the sample uniform trend in the change. Secondly, regarding the
company as a whole in the time dimension, the mean and standard deviation, the enterprise R&D expenditure and
standard deviation were statistically analyzed in each year. design are stable year by year. The standard deviation of
The results are shown in "Table II". In "Table II", in addition other indicators is also larger in different years. This shows
to the utility model and the design, the mean of all the that the gap among sample companies is changing as the
variables vary greatly in the time dimension, and there is no year changes.

TABLE II. THE STATISTICS OF MEAN AND STANDARD DEVIATION IN EACH YEAR

Variable RD INVENTION UTILITY DESIGN GOVS SCAL NUMS

Mean 18347546 60.05556 10.48148 3.555556 36741069 3.83E+09 598.1852

2010 Standard
0.000000 2.000000 1.000000 0.000000 7399607. 1.23E+09 154.5000
deviation
Mean 29543300 59.32407 15.31481 4.796296 61375575 4.97E+09 727.3889
2011 Standard
0.000000 3.000000 2.000000 0.000000 8975788. 1.49E+09 190.5000
deviation
Mean 42752566 59.49533 19.56075 3.102804 51769781 4.28E+09 788.3645
2012 Standard
0.000000 4.000000 4.000000 0.000000 10281731 1.51E+09 216.0000
deviation
Mean 54004005 44.96296 15.19444 3.555556 59477606 4.66E+09 790.2407
2013 Standard
0.000000 4.000000 3.000000 0.000000 9038270. 1.70E+09 236.0000
deviation
Mean 71308060 35.56075 13.55140 4.859813 57867494 5.17E+09 841.4486
2014 Standard
0.000000 4.000000 1.000000 0.000000 10687789 1.94E+09 247.0000
deviation

In summary, it can be seen from the above descriptive conducts an in-depth study on the impact of government
statistical analysis that the distribution and development innovation support policies on enterprise innovation.
status of sample companies vary widely at both the cross-
section level and the time level. In this case, if this paper D. Discussion on the Applicability of Dynamic Panel GMM
uses OLS to make the estimation, it is difficult to describe Method
the change in the gap, and the resulting regression coefficient
In the past, when scholars studied the impact of
is likely to be biased. Therefore, this paper chooses to use
government innovation support policies on enterprises, they
panel regression to estimate the sample. On the basis of
usually used the ordinary least squares (OLS) to make the
controlling individual differences and time differences, it
regression analysis. It can reflect the impact of government

256
 Advances in Economics, Business and Management Research, volume 71

support on enterprise innovation. However, the scholars did certain endogeneity among enterprise R&D investment, the
not consider estimation deviation caused by the endogeneity self-output innovation of the enterprise, the enterprise scale,
of variables. In addition, Xie Weimin (2009) and other government subsidies and other explanatory variables. Direct
scholars have found that the impact of government regression analysis will lead to significant deviation of the
innovation support policies on corporate innovation behavior estimation results. GMM estimation method based on the
is lagging. At the same time, this lag is not only reflected in dynamic panel data model can better eliminate the
the lag of the explanatory variable to the dependent variable, endophytism.
but also in the influence of the previous dependent variable
on the current dependent variable. That is to say, the Based on the above analysis, this paper analyzes the
innovative behavior of the enterprise in the early stage will impact of government subsidies on corporate R&D
have an impact on the current innovation behavior. Therefore, expenditure and enterprise independent innovation with the
the research in this paper needs to add the lag term of the generalized moment estimation method (GMM).
dependent variable to the explanatory variable, but this will
inevitably cause endogeneity problems. In the dynamic panel IV. EMPIRICAL ANALYSIS BASED ON DYNAMIC PANEL
model, the delayed period of the explained variable as an DATA
explanatory variable is likely to cause the explanatory
variable being correlated with the random disturbance term, A. The Role of Government Subsidies in Corporate R&D
which leads to the biased and inconsistent parameter Expenditure: Stimulation or Squeeze
estimation. And it may affect the final test result. 1) Testing result of Variable correlation: Before
Arellano and Bond (1991), Blundell and Bond (1998) estimating the parameter of the model, the variable
proposed GMM estimation. According to the setting of correlation should be first tested. The testing results of the
moment conditions, the endogeneity and hysteresis problems variable correlation involved in the model 1 are shown in
among variables are well solved. GMM (Generalzed method "Table III". On the one hand, from the correlation between
of moments) is a kind of estimation method based on the fact explanatory variables and other explained variables,
that the actual parameters of the model satisfy certain government subsidies, the enterprise scale and the number
moment conditions, which is the generalization of the
of enterprise R&D personnel are highly correlated with
moment estimation method. As long as the model is set
correctly, it is always possible to find several moment R&D expenditures. On the other hand, the correlation
conditions that the actual parameters of the model satisfy. between government subsidies, enterprise scale, and
And then, it will use GMM estimation. enterprise R&D personnel is not low. That is to say, the
model may have a certain degree of multicollinearity.
Compared with traditional econometric estimation However, the enterprise scale and the number of enterprise
methods, GMM has the following advantages. First,
R&D personnel are only used as control variables of the
traditional econometric estimation methods, such as ordinary
least squares method, instrumental variable method and model, this paper will retain these data.
maximum likelihood method have their own limitations.
TABLE III. TESTING RESULTS OF VARIABLE CORRELATION (MODEL 1)
Namely, the parameter estimator must be a reliable estimator
when certain assumptions are met, for example, the random RD GOVS SCAL NUMS
error term of the model obeys normal distribution or known
RD 1 0.657367 0.742116 0.898521
distribution, and the GMM does not need to know the exact
distribution information of the random error term. It allows GOVS 0.657367 1 0.627831 0.504536
random error terms to have heteroscedasticity and sequence SCAL 0.742116 0.627831 1 0.813334
correlation. The resulting parameter estimators are more
efficient than other parameter estimation methods. Second, NUMS 0.898521 0.504536 0.813334 1
the GMM method can solve the correlation between
unobservable variables and explanatory variables to some 2) Dynamic panel GMM estimation results: This paper
extent, or the bias caused by the omission of some important uses Eviews 7.0 software to make the GMM estimation of
explanatory variables. It is difficult to include all possible Model 1. The results are shown in "Table IV".
significant explanatory variables in any model design, which
is particularly important for empirical testing. Third, there is

TABLE IV. GMM-BASED MODEL ESTIMATION RESULTS (MODEL 1)

Parameter estimation
Variable Standard deviation t statistics P ratios
result
RD(-1) 1.098289 0.004199 261.5394 0.0000
GOVS1 0.073484 0.010011 7.340645 0.0000
SCAL 0.001338 0.000335 3.995503 0.0001
NUMS -9329.216 3405.923 -2.739115 0.0065
J-statistics: 4.609098; the sequence of instrumental variable: 6

257
 Advances in Economics, Business and Management Research, volume 71

The GMM estimation of the dynamic panel data does not large scales have stronger willingness to carry out innovation
have the goodness of fit and the F statistic of the classical activities. And they will pay more attention to innovation
regression analysis. However, it takes the Sargan test by activities. This is reflected in the increase of R&D
using the J statistic. In the null hypothesis, the model is over- expenditure.
constrained correctly. If the sargan test is rejected, the model
is set incorrectly. The P value of the J statistic can be B. The Role of Government Subsidies for Independent
calculated by the chidist (x, ir-v) function. d is the value of Innovation of Enterprise: Stimulation or Inhibitions
the J statistic. ir is the rank of the instrument variable. v is the
number of parameters to be estimated. From the estimation 1) Testing result of variable correlation: Before
results of Model 1 in "Table IV", J=4.609098, ir=11, v=3. estimating the parameter of the model, the correlation test of
From the P value of sargan test (0.7984), the selection of the the variables should be first carried out. The testing results
tool variables is reasonable, and the model 1 is set correctly. of the correlation among the variables involved in the model
2a, the model 2b and the model 2c are shown in "Table V".
3) Interpretation of model results: As can be seen from On the one hand, from the correlation between explanatory
"Table IV", the parameter estimation result corresponding to variables and explained variables, the government subsidies,
the government subsidy is 0.073484. The accompanying the enterprise scale and the number of R&D personnel are
probability of the T test is P=0.0000, indicating that the highly correlated with enterprise inventions, utility models
government subsidy is significantly positively correlated and designs. On the other hand, the correlation among each
with the enterprise R&D expenditure. The government explanatory variable is not low. That is to say, the model
subsidizes companies to encourage innovation activities. may have a certain degree of multicollinearity problem.
This will stimulate independent innovation. And corporate However, the enterprise scale and the number of R&D
R&D expenditure will increase significantly. Hypothesis 1 personnel of the enterprise are the control variables of the
is supported. model. We can make the same processing as Model 1, and
The results in "Table IV" also show that the level of
all variables are retained.
enterprises R&D expenditure is also significantly positively
correlated with the scale of enterprises. The enterprises with

TABLE V. TESTING RESULTS OF VARIABLE CORRELATION (MODEL 2A, MODEL 2B, MODEL 2C)

INVENTION UTILITY DESIGN GOVS SCAL NUMS

INVENTION 1 0.52784 0.651161 0.546569 0.746646 0.847212
UTILITY 0.52784 1 0.460528 0.522567 0.630874 0.351022
DESIGN 0.651161 0.460528 1 0.473178 0.623708 0.716524
GOVS 0.546569 0.522567 0.473178 1 0.627831 0.504536
SCAL 0.746646 0.630874 0.623708 0.627831 1 0.813334
NUMS 0.847212 0.351022 0.716524 0.504536 0.813334 1

2) Dynamic panel GMM estimation results: We will use model 2b and model 2c. The results are shown in "Table
Eviews 7.0 software to make GMM estimation of model 2a, VI".

TABLE VI. GMM-BASED MODEL ESTIMATION RESULTS (MODEL 2A, MODEL 2B, MODEL 2C)
Model 2a Model 2b Model 2c
coefficient P ratios coefficient P ratios coefficient P ratios
RD(-1) 1.095344 0.0000 -0.033736 0.3529 0.306204 0.0000
GOVS1 1.75E-07 0.0017 1.39E-08 0.7491 3.66E-08 0.0448
SCAL 2.85E-08 0.0437 -3.57E-09 0.0057 -2.91E-09 0.1347
NUMS 0.009317 0.7877 0.020872 0.0060 -0.007121 0.2448
J-statistics: 3.410590 J-statistics: 1.783121 J-statistics: 2.451365
Sargan
The rank of instrumental The rank of instrumental The rank of instrumental
test
variable: 6 variable: 6 variable: 6

The sargan test is performed on model 2a, model 2b and reasonable, and the model is set correctly. In model 2c,
model 2c. The results are as the following. In model 2a, J=2.451365, ir=6, v=3. The P ratio (0.4841) verified by
J=3.410590, ir=6, v=3. The P ratio (0.3325) verified by sargan test shows that the selection of tool variables is
sargan test shows that the selection of instrument variables is reasonable, and the model is set correctly.
reasonable. The model is set correctly. In model 2b,
J=1.783121, ir=6, v=3. The P ratio (0.6186) verified by 3) Interpretation of model results: It can be seen from
sargan test shows that the selection of tool variables is “Table 6” that the parameter estimation result corresponding

258
 Advances in Economics, Business and Management Research, volume 71

to the government subsidy in model 2a is 1.75E-07. The the inventions, utility models and designs, is uncertain.
accompanying probability of the T test is P=0.0017, Therefore, the policy measures adopted by the government to
indicating that the government subsidy is significantly encourage enterprises to carry out innovation activities are
positively correlated with the number of enterprise likely to be inefficient. Combined with the DEA method, we
can analyze the efficiency of government innovation support
inventions. The hypothesis 2a is supported. The parameter
policies. In order to simplify the analysis, this part only
estimation result corresponding to the subsidy is 1.39E-08, selects the input-output data of the sample companies in the
and the accompanying probability of the T test is P=0.7491, "computer, communication and other electronic equipment
indicating that for the sample enterprises, the impact of manufacturing industry" published by China Securities
government subsidies on the number of utility models is not Regulatory Commission (2012 edition) in 2014 for the
significant. The hypothesis 2b is not supported. The analysis on government innovation support efficiency.
parameter estimation result corresponding to the subsidy is
3.66E-08, and the accompanying probability of the T test is A. Selection of Indicators
P=0.0448, indicating that the government subsidy is According to the analysis above, the government subsidy
significantly positively correlated with the number of scale and government R&D expenditure are taken as input
corporate designs, and hypothesis 2c is also supported. variables, and the sum of invention, utility model and design
is taken as the output variable. Based on the DEA method,
V. ANALYSIS OF GOVERNMENT INNOVATION SUPPORT the efficiency of government innovation support to promote
EFFICIENCY BASED ON DEA METHOD independent innovation of enterprises is measured. "Table
According to the empirical analysis based on dynamic VII" shows the input-output data of 13 sample companies in
panel data, the positive impact of government subsidies on the "Computer, Communications and Other Electronic
R&D expenditures of enterprises is significant. However, the Equipment Manufacturing Industry" in 2014.
impact on the independent innovations of enterprise, namely,

TABLE VII. THE INPUT-OUTPUT DATA OF SAMPLE ENTERPRISE IN "COMPUTER, COMMUNICATIONS AND OTHER ELECTRONIC EQUIPMENT MANUFACTURING"
(2014)
output indicators input indicators
Stock R&D
Corporate Name Invention, utility model government subsidy
Code expenditure
Design (piece) (yuan)
(yuan)
000063 Zhongxing Telecommunication Equipment Corporation 1031 0 3483505000
000100 TCL Corporation 160 1528368000 1048372000
000725 BOE Technology Group Co., Ltd. 2648 830471170 0
000970 Beijing Zhongke Sanhuan Hi-tech Co., Ltd. 11 19073962 0
002308 VTRON Technology Co.,Ltd. 262 24355076 0
002362 Hanvon Technology Co.,Ltd. 27 14893500 1498659
002415 Hangzhou Hikvision Digital Technology Co.,Ltd. 194 844526976 0
Suzhou Shengli Precision Manufacturing Technology Co.,
002426 24 7490398 0
Ltd.
002429 Shenzhen MTC Co., Ltd. 13 59863659 0
Guangzhou Haige Communications Group Intercoporate
002465 41 98870571 6370737
Conpanny
002475 LuXshare Intercorporate Conpanny 11 59945521 0
300065 Beijing Hailander Digital Technology Co., Ltd. 7 15295806 0
300083 Guangdong JANUS Intelligent Group Corporation Limited 90 9786628 1351528

1) DEA analysis results: The DEAP2.1 software was "Table VIII" were obtained, including comprehensive
used to analyze the efficiency of government innovation efficiency, pure technical efficiency and scale efficiency.
support of 13 sample companies, and the results shown in

TABLE VIII. DEA TECHNICAL EFFICIENCY MEASUREMENT RESULTS

comprehensive pure technical scale
Corporate Name returns to scale
efficiency efficiency efficiency
Zhongxing Telecommunication Equipment Corporation 1.000 1.000 1.000 —
TCL Corporation 0.756 0.992 0.762 drs
BOE Technology Group Co., Ltd. 0.940 1.000 0.940 drs
Beijing Zhongke Sanhuan Hi-tech Co., Ltd. 0.911 0.945 0.964 drs
VTRON Technology Co.,Ltd. 1.000 1.000 1.000 —
Hanvon Technology Co.,Ltd. 0.594 0.689 0.862 drs
Hangzhou Hikvision Digital Technology Co.,Ltd. 0.824 0.849 0.970 drs
Suzhou Shengli Precision Manufacturing Technology Co., Ltd. 1.000 1.000 1.000 —
Shenzhen MTC Co., Ltd. 1.000 1.000 1.000 —
Guangzhou Haige Communications Group Intercoporate Conpanny 0.827 0.862 0.959 drs

259
 Advances in Economics, Business and Management Research, volume 71

comprehensive pure technical scale

Corporate Name returns to scale
efficiency efficiency efficiency
LuXshare Intercorporate Conpanny 0.692 0.693 0.999 irs
Beijing Hailander Digital Technology Co., Ltd. 1.000 1.000 1.000 —
Guangdong JANUS Intelligent Group Corporation Limited 0.802 1.000 0.802 drs
Mean 0.873 0.925 0.943
a.
Note: irs, drs respectively indicate that the scale returns are constant. The scale returns are increasing, and the scale returns are decreasing.
As can be seen from "Table VIII", there are 5 companies sample companies, the government innovation support
with comprehensive efficiency of 1 among the 13 sample policy has certain inefficiency. Hypothesis 3 is supported.
companies, and the remaining 8 companies are relatively
inefficient. The hypothesis 3 is supported. Overall, the The results of this paper generally support the positive
comprehensive efficiency of the 13 sample companies was role of government subsidies in stimulating corporate
0.873, indicating that 12.7% of government subsidies were innovation. The policy implication of the research is that it is
wasted. The government could use other methods to necessary for the government to support the enterprise to
encourage the enterprise innovation to increase innovation improve the independent innovation capability. However,
output or appropriately reduce government subsidies for there is certain inefficiency in the government subsidy. It is
certain enterprises. Inefficiency includes technical necessary to design a more reasonable government
inefficiency and scale inefficiency. innovation support policy implementation plan to optimize
the effect of government support on corporate innovation.
The comprehensive efficiency type of enterprise refers to
the enterprises with comprehensive efficiency, technical
efficiency and scale efficiency of 1. The five sample REFERENCES
enterprises with comprehensive efficiency of 1 are all
[1] Dong Yaqin. "Government Research and Development subsidy,
efficient. This type of enterprises is in the stage of constant Enterprise R&D Expenditure and Enterprise R&D Performance –
scale returns. Therefore, it is not necessary to reduce input or Empirical Research Based on GEM Enterprises, Accountant No. 10,
increase output. It just needs to maintain the scale of existing 2015.
government subsidies. [2] Jiang Ning, Huang Wan. "The Impact of Government Subsidies on
Corporate R&D Inputs – An Empirical Study Based on China's High-
Pure technological inefficiency means that the scale tech Industries", Science and Technology Management, No. 7, 2010.
efficiency is 1, and the technical efficiency is less than 1. [3] Liang Laixin, Cao Qinrun. "R&D personnel and their changes and
The inefficiency comes from the pure technological corporate R&D expenditures – based on empirical evidence of listed
inefficiency. The comprehensive efficiency needs to be companies in China", Research and Development Management, No. 1,
increased by improving the technical efficiency. There is no 2010.
enterprise with pure technology inefficiency in "Table VIII". [4] Lv Jiuqin, Yu Dandan. "Government Research Innovation subsidy
Scale inefficiency means that the technical efficiency is 1. and Enterprise R&D Investment: squeeze, Substitution or
stimulation?", China Science and Technology Forum, No. 8, 2011.
However, the scale efficiency is less than 1. The inefficiency
[5] Wu Jianhuan, Zhao Junli. Research on the Effectiveness of
comes from the scale inefficiency. There is no enterprise Supporting Enterprises' Independent Innovation Financial Policies,
with scale inefficiency in "Table 8". Theory Monthly, No. 5, 2007.
[6] Xiao Wen, Lin Gaobang. "Government Support, R&D Management
VI. CONCLUSION and Technology Innovation Efficiency – Empirical Analysis Based on
China's Industrial Sector", Management World, No, 4, 2014.
This paper examines the role of government subsidies for [7] Xie Weimin, Tang Qingquan, Lu Shanshan. "Government R&D
corporate R&D expenditures and corporate independent Funding, Corporate R&D Expenditure and Independent Innovation –
innovations, taking 108 A-share listed companies in 2010- Empirical Evidence from Chinese Listed Companies", Financial
2014 as samples. Based on the analysis of relevant literature Research, No. 6, 2009.
and the analysis of the logical mechanism of government [8] Xing Wen, Zhang Jianhua. China's Innovation Support Policy:
Theoretical Analysis and Validity Test, Contemporary Economic
subsidies to stimulate enterprise innovation, it is proposed Science, No. 4, 2009.
that government subsidies can stimulate enterprises to carry [9] Yang Xiangyang et al. "The Impact of Government Support on KIBS
out innovation activities and increase R&D expenditures, Enterprise Innovation Activity", Science and Science and Technology
which will have a positive impact on the independent Management, No. 12, 2015.
innovation results of enterprises. After controlling the [10] Yang Xiangyang, Tong Xinle. "Research on the Impact of
enterprise scale and the number of scientific researchers, Government Support on R&D Investment of KIBS Enterprises",
government subsidies can significantly stimulate R&D Research Management, No. 12, 2014.
expenditures. The hypothesis 1 is supported. The verification [11] Zheng Chunmei, Li Pei. "The Impact of Government Subsidies and
results of Hypothesis 2 indicate that the influence of Tax Preference on Enterprise Innovation Performance — An
Empirical Study Based on GEM High-tech Enterprises", Science and
government subsidy on the number of inventions and design Technology Progress and Countermeasures, No. 16, 2015.
is obviously positive. However, the impact on the number of [12] Arundel A.Innovation survey indicators:what impact on innovation
utility models is not significant. Hypothesis 2a and policy. Science,Technology and Innovation Indicators in a Changing
Hypothesis 2c are supported, and the hypothesis 2b is not World . 2007
supported. On this basis, this paper uses the DEA method to [13] David PA,Hall BH,Toole AA.Is public R&D a complement or
analyze the efficiency of government innovation support. For substitute for private R&D? a review of the econometric evidence.
Research Policy . 2000

260
 Advances in Economics, Business and Management Research, volume 71

[14] Guellec,D,De La Potterie B.V.P.The Impact of Public R&D

Expenditure on Business R&D. Economics ofInnovation and
NewTechnology . 2003
[15] Hall B Van Reenen,J.How effective are fiscal incentives for R&D, a
review of the evidence. Research Policy . 2000
[16] Tassey Gregory.Policy Issues for R&D Investment in a Knowledge-
Based Economy. Journal of Technology Transfer. 2004

261