R&D EFFICIENCY EVALUATION OF COMPUTER INDUSTRY IN CHINA BASED ON AN IMPROVED DEA MODEL
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Journal of Theoretical and Applied Information Technology
31st December 2012. Vol. 46 No.2
© 2005 - 2012 JATIT & LLS. All rights reserved.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
R&D EFFICIENCY EVALUATION OF COMPUTER
INDUSTRY IN CHINA BASED ON AN IMPROVED DEA
MODEL
CHENGDONG WANG, LIANGQUN QI, YUANYUAN CAI
School of Management, Harbin University of Science and Technology, Harbin150080, China
ABSTRACT
The computer industry plays an important role in promoting the development of national economy and
national comprehensive power of China. As an important evaluation indicator, the R&D efficiency of
computer industry becomes one of the most important research fields currently. According to the shortages
of traditional C2R model of DEA, this paper constructs an improved DEA model for measuring the R&D
efficiency of computer industry and does an empirical research on R&D efficiency of computer industry in
different regions. This research shows that: the improved DEA model overcomes the shortages of
traditional C2R model and it is able to reflect the difference of R&D efficiency between different regions
more effectively; Compared with efficiency frontier, the R&D efficiency of Chinese computer industry has
potentials to improve; The R&D efficiency of Chinese computer industry is influenced by various factors,
such as government support capacity, market structure and enterprise scale; Improve the environment of
R&D activities is more effectively than increase the input of R&D resources in improving the R&D
efficiency of Chinese computer industry.
Keywords: Computer Industry, Evaluation of R&D Efficiency, Improved DEA Model, Influence Factor
1. INTRODUCTION a lot of useful reference information for studying
the R&D efficiency of computer industry. From the
As a typical high-tech industry, the computer previous research on efficiency evaluation we can
industry not only has an important influence on the see that, the DEA model is obviously different form
other industries and national economy, but also
the nonparametric evaluation methods, because it
promotes the development of them. Therefore,
could analyze the DMU which with multiple input
developing the computer industry has become a
high-effectively way to promote China's economic and output indicators. And this characteristic makes
development. As one of the most important the DEA model exactly suitable for evaluating the
measure indicator of development situation, the efficiency of high-tech industry. Beside that, the
R&D efficiency of computer industry becomes one DEA model does not need to set the specific
of the most important research fields currently. function form, thus the DEA model avoids the
The outline of the paper is as follows. In section errors and problems caused by wrong setting of
2 we review previous research on previous frontier function. For these advantages, DEA model
improvement of traditional DEA model. In section is widely used by a lot of scholars, such as
3, the improved DEA model is proposed. Suzuki(1989) [1] and Encaoua(2000) [2]. However,
Evaluation of R&D efficiency by the improved there are still some shortages of the traditional DEA
DEA model of computer industry is discussed in model. For example, the traditional DEA model can
section 4. Section 5 gives the conclusions and
not distinguish DMU form each other effectively in
political suggestions.
some cases, as well as the weights of its input and
output indexes are set up inconsistent with the
2. PREVIOUS RESEARCH REVIEW
actual. In view of the weaknesses of the traditional
DEA model, the scholars have put forward a series
Previous researches on R&D efficiency of
of improved DEA model. The first kind of
computer industry are seldom both in China and
improved DEA model is given some prior
abroad. However, the research achievements on
information, such as weight restrictions
high-tech industry are more plenty and they provide
706Journal of Theoretical and Applied Information Technology
31st December 2012. Vol. 46 No.2
© 2005 - 2012 JATIT & LLS. All rights reserved.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
information, preference structure information and 3. THE CONSTRUCTION OF IMPROVED
value efficiency analysis information. Specifically, DEA MODEL
the improved DEA model with prior information
includes the following: direct weight restrictions 3.1 Introduction of Traditional DEA Model- C2R
Model
model, which is built up by Dyson,
C2R is built up by Charnes, Cooper and Rhodes
Thanassoulis(1988) [3] and Roll(1990)[4]; cone
at 1978, its basic assumption is unchangeable scale
ratio model, which is built up by Charnes（1989,
to return. The essence of C2R is evaluating the
1990）[5] [6]; assurance region model established
relative efficiency of organization by mathematical
by Thompson （ 1986 ） [7] and Beasley’s virtual
programming techniques [17]. At present, most of
inputs and outputs restrictions model （1990）[8].
the researchers are familiar with the traditional C2R
there second kind of improved DEA model is
model:
without prior information, such as super efficiency
Set n comparable DMU, and denoted by DMU1-n,
model(Andersen, Petersen （ 1993 ） [9], cross-
each DMU has m types input and s types output.
evaluation model(Sexton （ 1986 ） ) [10] and
So the input vector can be written as
multiple objective approach(Li, X. B, （1999）)
X j = ( x1 j , x 2 j ,L , x mj ) T , and the output vector can
[11]. In addition, scholars improved the traditional
DEA model by changing the input and output be written as:
variable type both abroad and in China. For Y j = ( y1 j , y 2 j ,L, y sj ) T , ( j = 1,L , n) , and they
example, Cooper （ 1999 ） analyzed the DEA are constant. v = (v1 , v 2 , L , v m ) T are m types of
model with categorical variables[12], Cook, Kress
and Seiford （ 1993 ） given an improved DEA input weight vectors, and u = (u1 , u 2 , L , u s ) T are
model with ordinal numbers as its input and output s types of output weight vectors, they are
variables[13], Herbert and Thomas (2004) studied unknown variable.
the efficiency network DEA model for evaluating So, the efficiency of Kth DMU can be rewritten
efficiency of complex structured system[14]. The as equation (1):
improved DEA models mentioned above still have uTY
Vk = T k (1)
some shortages: On one hand, the models with v Xk
priority information cannot avoid the error or bias Set Vk , the efficiency value of DMUk , as
in judgment of value and priority information. And
objective function, and constraint to the efficiency
these shortages may cause the evaluation results are
value of all DMU, we get fractional programmed
inconsistent with the facts. On the other hand, the
C2R model as equation (2):
improved DEA model without prior information
always calculates the weight vector of input and u T Yk
max T
output indicators from the most advantageous angle v Xk
for their own DMU, which leads to low u Y j
T (2)
T ≤ 1, j = 1,L , n
comparability of efficiency between different
v X j
DMU. In order to solve the shortages of the DEA
model mentioned above, this paper builds up an u ≥ 0, v ≥ 0
improved DEA model by drawing lessons from the
previous academic achievements of Liu Transform equation (2) into equation (3) :
Yingping(2006) [15] and Sun Kai(2008) [16], and 1
t= , α = tv, β = tu (3)
does an quantitative research on evaluation of R&D vT X k
efficiency of computer industry in China based on
Where α is m -dimension column vector, β is
the improved DEA model. According to the
evaluation results, this paper analyzes the current
s -dimension column vector, so,
situation and developing rules of computer α = (tv1 , tv 2 ,L, tv m ) T , β = (tu1 , tu 2 ,L , tu s ) T .
industry, and provides policy reference for Accordingly, the model changes into equation (4):
promoting the development of computer industry.
707Journal of Theoretical and Applied Information Technology
31st December 2012. Vol. 46 No.2
© 2005 - 2012 JATIT & LLS. All rights reserved.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
max β T Yk max β T Yn +1
T T
α X j − β Y j ≥ 0, j = 1,L , n α X j − β Y j ≥ 0, j = 1,L , n + 2
T
(4) T
T T
α X k = 1 α X n +1 = 1
α ≥ 0 , β ≥ 0 α ≥ 0 , β ≥ 0
(5)
3.2 The Improved DEA Model: By Introducing So, DMUn+1 is obviously DEA efficiency, and
Virtual DMU
V * n +1 ≡ 1 . But the optimum solutions α * and β *
The first step: construct virtual DMU. To
are infinite. So, choosing of the infinite weight
construct two virtual DMU named DMUn+1 and
vector is needed in order to identify "reasonable"
DMUn+2. DMUn+1 is the optimal decision making public weight vector for all the n +2 DMU.
unit, its input vector can be written as
X n+1 = ( x1,n +1 , x2, n+1 , L , xi ,n +1 ,L , xm, n+1 )T The third step: choose the public weight
vector. From the equation
and its output vector can be written as
β * Yn+1
T T
u * Yn +1
Yn +1 = ( y1,n +1 , y2, n+1 ,L , yr , n+1 ,L , ys , n+1 ) . T V *
n +1 = = ≡ 1 we can infer that,
α * X n +1
T T
v * X n +1
DMUn+2 is the worst decision making unit, its optimal decision making unit, DMU n+1
input vector can be written as
meet β *T Yn +1 − α *T X n +1 = 0 . In order to make
X n + 2 = ( x1,n + 2 , x 2, n + 2 , L , xi ,n + 2 , L , x m ,n + 2 ) T
sure there are “reasonable” public weight vector,
And its output vector can be written as we build model as:
Yn + 2 = ( y1,n + 2 , y 2,n + 2 ,L, y r ,n + 2 ,L, y s ,n + 2 ) T . Set the min β T Yn + 2
T
α X j − β Y j ≥ 0, j ≠ n + 1
T
input index of optimal decision making unit equal
to the minimum value of X , as well as the output T
α X n + 2 = 1
index of optimal decision making unit equal to the T
β Yn +1 − α X n +1 = 0
T
maximum value of X . So,
α ≥ 0 , β ≥ 0
xi,n +1 = min( xi1 , xi 2 , L, xin ) ,
(6)
y r ,n +1 = max( y r1, y r 2, L , y rn ) . The model (6) aims at the minimum efficiency
Correspondingly, the input and output indexes of value of the worst decision making unit DMUn+2,
the worst decision making unit are maximum value and adds constraint β T Yn +1 − α T X n +1 = 0 into the
and minimum value of X , that is model in order to get maximum value of the
xi,n + 2 = max ( xi1 , xi 2 , L , xin ) , optimal decision making unit DMUn+1. Change a
kind of statement, form infinite group weight vector
Obviously, DMUn+1 and DMUn+2 may not exist
of optimal decision making unit, model (6) can
inside the production possibility set, and their
choose a group of weight vector which makes the
introduction is prepared for the following steps as a
efficiency value of worst decision making unit to be
reference.
minimum one.
The second step: establish the efficiency The fourth step: to calculate efficiency value
evaluation model of optimal DMUn+1. Take the for DMU. By using the following equation (7) and
efficiency evaluation of n +2 DMU as constraint; optimal solutions α * * and β * * of equation (6),
we build up the new DEA model by setting calculate efficiency value of every decision making
efficiency evaluation V * n +1 of the optimal DMU unit.
β ** Yj
T
as objective function. The new DEA model as
equation (5) shows: V j* * = ( j = 1, L, n)
α ** X j
T
(7)
The fifth step: Reset the optimal and worst
DMU by using the evaluation results of the
708Journal of Theoretical and Applied Information Technology
31st December 2012. Vol. 46 No.2
© 2005 - 2012 JATIT & LLS. All rights reserved.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
fourth step. V **
j
in fourth step is calculated by The sixth step: calculate efficiency value by
public weight vector, so it is objective and rational, repeating second step to fourth step. From the
and it can solve the problem of traditional C2R fifth step we can see that, DMUα has the maximum
model. The problem of traditional C2R model is efficiency value in all the DMU, and it is obviously
that, there are more than one DMUs’ efficiency effective compared with other DMU, so it is set as
value is 1, and this caused difficulty in 1. To use DMUα and DMUβ instead of DMUn+1 and
distinguishing and sequencing of all the DMU. DMUn+2 in equation (4) and equation (5), then
However, the efficiency value is far less than 1 repeat second step to fourth step until we get
sometimes because DMUn+1 and DMUn+2 are virtual satisfying efficiency value of all DMU.
ideal decision making unit. And more or less, this
situation reduces the discrimination degree of 4. R&D EFFICIENCY EVALUATION OF
COMPUTER INDUSTRY IN CHINA
different DMU on efficiency evaluation. Therefore,
replace DMUn+1 by decision making unit DMUα 4.1 Selection of Influencing Factors and
with maximum efficiency value, and replace Evaluation Indexes
DMUn+2 by decision making unit DMUβ R&D efficiency evaluation of computer industry
( 1 ≤ α，β ≤ n ) with minimum efficiency value, in China by DEA model depends on two aspect
and introduced into models above. In this way, not endogenous factors: input resources of computer
only the results of the fourth step are used, but also industry in R&D process, and final output results.
In addition, exogenous influence factors, such as
makes optimal and worst decision making unit have
market structure, enterprise size, enterprise
actual significance (the optimal and worst decision ownership, financial support of government
making unit are come from production possibility departments and financial institutions and technical
set). support of scientific research institutions have a
certain degree influence on R&D efficiency of
computer industry in China [18]. To establish R&D
efficiency evaluation index system of computer
industry in China as shown in figure 1:
Fig. 1. The R&D Efficiency Evaluation Index System of Computer Industry
4.2 Determination of Evaluation Data Yearbook 2011”, “China Statistics Yearbook on
Considering the timeliness and availability of data high technology industry 2011” and “China
in this paper, we choose “China statistical yearbook Statistical Yearbook on science and technology
2011”, “China Industrial Economy Statistical 2011” to be the data sources. By collecting and
709Journal of Theoretical and Applied Information Technology
31st December 2012. Vol. 46 No.2
© 2005 - 2012 JATIT & LLS. All rights reserved.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
calculating, we got 23 group data of efficiency Table 3 Efficiency Evaluation Results Of Chinese
evaluation index of regional computer industry, and Regional Computer Industry
shown in table 1:
Table I Data Of Regional R&D Evaluation Indexes
Descriptive statistics results of the computer Note: DRS, decreasing returns to scale; -,
industry R & D investment, R & D outputs and constant returns to scale; IRS, increasing returns to
influence variables in this paper as shown in table scale
2:
TABLE 2 Results Of Descriptive Statistics 4.3 Analysis on R&D Efficiency Evaluation
Result of Computer Industry in China
The analysis on evaluation result of computer
industry R&D efficiency in China is based on the
data in table 1 and table 3, and it shows that:
Firstly, from the perspective of industry, the
average efficiency evaluation results of Chinese
computer industry without and with influence
Using the improved DEA model above, this factors are 0.6603 and 0.7787 respectively.
paper evaluates the R&D efficiency value of According previous research result of reference [18],
Chinese regional computer industry with and the R&D efficiency of Chinese computer industry
without influence factors of R&D process. The is in the middle to upper level of high-tech industry.
efficiency evaluation results are shown in table 3: The reason is that, most R&D activities on Chinese
computer industry are set focus on implicational
research field, and the research results have a high
market value and a high conversion rate.
Secondly, from the perspective of regions: the
R&D efficiency with influences factors of
developed regions of computer industry, such as
Guangdong, Jiangsu and Zhejiang provinces, is
lower than 0.9. This situation shows that, the
current R&D efficiency of computer industry is low
in China, and high level output is depends on high
level input rather then high level R&D efficiency.
At present, the output of Chinese computer industry
does not reach the maximum output which the
710Journal of Theoretical and Applied Information Technology
31st December 2012. Vol. 46 No.2
© 2005 - 2012 JATIT & LLS. All rights reserved.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
frontier production function showed. In the other (3) To improve the R&D environment is more
word, redundancy of inputs or produce insufficient effective than increasing R&D input in improving
or both of them are existed, and these problems will the R&D efficiency. So, exogenous influence
grow bigger with the increasing investment of R&D factors, such as market structure, enterprise size,
activities. enterprise ownership, financial support of
Thirdly, from the perspective of returns to scale: government departments and financial institutions
22 in 23 regions are decreasing returns to scale. But and technical support of scientific research
when the exogenous influence factors are institutions should be considered in order to
considered, this figure dropped to 2 and 21 in 23 improve the R&D efficiency of computer industry.
regions are increasing returns to scale. These data In addition, ways of financial support need to be
show that, to improve the R&D efficiency only by changed form “process support” to "result support".
increasing R&D investment is not effectively. The
exogenous influence factors, such as market ACKNOWLEDGEMENTS
structure, enterprise size, enterprise ownership, This work is supported by the following projects:
financial support of government departments and The National Natural Science Foundation of China,
financial institutions and technical support of No.70773032; Philosophy and Social Science
scientific research institutions should be considered Planning Projects of Heilongjiang Province, No.
in order to improve the R&D efficiency of 11B060; Youth Academic Support Projects of
computer industry in China. Heilongjiang Province.
Finally, from the perspective of influence factors:
when the influence factors are considered, the
average R&D efficiency value is 0.7787, and it is REFERENCES:
higher than 0.6603, which is the average R&D
efficiency value without considering the influence [1] Akira Suzuki and Kazuyuki, R & D capital, rate
factors. We can see that, the improvement of R&D of return on R & D investment and spillover of
efficiency of computer industry is depending on the R & D in Japanese manufacturing industry, The
R&D environment. Therefore, to improve the R&D Review of Economics and Statistics, Vol. 71,
environment is more effective than increasing R&D No. 4, pp. 555-564, 1989
input in improving the R&D efficiency of Chinese [2] David Encaoua, Bronwyn H. Hall and Franois
computer industry. Laisney, The Economics and Econometrics of
Innovation (Kluwer Academic Publishers, 2000)
[3] R. G. Dyson and E, Thanassoulis, Reducing
5. CONCLUSIONS AND SUGGESTIONS
weight flexibility in data envelopment analysis,
Journal of the Operational Research Society,
Through the research above, this paper got the Vol. 39, No. 6, pp. 563-576, 1988
following conclusions:
[4] Y. Roll, W. Cook, and B. Golany, Controlling
(1) The improved DEA model designed in this factor weights in data Envelopment analysis,
paper is superior to the traditional C2R model on IIE Transactions Journal, Vol. 23, No. 1, pp. 2-
discrimination and disperse degree in R&D 9, 1990.
efficiency evaluation of Chinese computer industry. [5] A. Charnes, W. W. Cooper, Q. L. Wei and Z. M.
(2) At present, R&D efficiency of computer Huang, Cone ratio data envelopment analysis
industry is in the middle level of all high-tech and multiple objective linear programming,
industry in China, and the output of Chinese International Journal of System Science,
computer industry does not achieve the maximum Vol.20, No. 7, pp. 1099-1118, 1989.
output which the frontier production function [6] A. Charnes, W. W. Cooper, Z. M. Huang and D.
B .Sun, Polyhedral cone-ratio DEA models with
showed. Input redundancy or output deficiency or
an illustrative application to large commercial
both of them are existed, and these problems will banks, Journal of Econometrics, Vol. 46, No. 1,
grow bigger with the increasing investment of R&D pp. 73-91, 1990.
activities.
711Journal of Theoretical and Applied Information Technology
31st December 2012. Vol. 46 No.2
© 2005 - 2012 JATIT & LLS. All rights reserved.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
[7] R. G. Thompson, Jr. Singleton, R. M. Thrall and
B. A. Smith, Comparative site evaluations for
locating a high-energy physics lab in Texas,
interfaces, Vol. 16, No. 6, pp. 35-49, 1986.
[8] J. E. Beasley, Comparing university departments,
International Journal of Management Science,
vol. 18, No. 2, pp. 171-183, 1990.
[9] P.Andersen and N. C.Petersen, A procedure for
ranking efficient units in data envelopment
analysis, Management Science, Vol. 37, No. 8,
pp. 1261-1264, 1993.
[10] T. R. Sexton, Measuring Efficiency: An
Assessment of Data Envelopment Analysis,
(Jossey-Bass, 1986)
[11] X. B. Li and G. R. Reeves, A multiple criteria
approach to data envelopment analysis,
European Journal of Operational Research,
Vol. 115, No. 3, pp. 507-517, 1999.
[12] W. W. Cooper, IDEA and AR-IDEA: Models
for Dealing with Imprecise Data in DEA,
Management Science, pp. 597-607, 1999.
[13] W. D. Cook, M. Kress and L. Seiford. On the
Use of Ordinal in Data Envelopment Analysis.
Journal of the Research Society, Vol. 45, No. 4,
1993, pp. 133-140.
[14] Herbert F. Lewis and Thomas R. Sexton,
Network DEA: Efficiency Analysis of
Organizations with Complex Internal Structure,
Computers & Operations Research, Vol.31, No.
9, pp. 1365-1410, 2004.
[15] Liu Yingping, Lin Zhigui and Shen Zuyi,
Effective Data Envelopment Analysis Method
for Ranking Decision Making Units, Systems
Engineering theory & Practice, Vol. 17, No. 3,
pp.112-116, 2006.
[16] Sun Kai and Ju Xiaofeng, Modified Data
Envelopment Analysis Model for
Distinguishing Efficient Decision Making Units,
Journal of Systems & Management, Vol. 17,
No. 2, pp. 134-137, 2008.
[17] A. Charnes, W. W. Cooper and E. Rhodes,
Measuring the Efficiency of Decision-Making
Units, European Journal of operations
Research, Vol. 2, No. 6, pp. 429-444, 1978.
[18] Chen Xiude and LIANG Tongying. An
empirical research on R&D efficiency and
influence factors of China’s high-tech industry-
based on panel-data SFPF model, Studies in
Science of Science, Vol. 28, N. 8, pp. 1198-
1205, 2010.
712You can also read
