Electronic Commerce, R&D, Externalities, and Productivity An Empirical Study of Taiwanese Manufacturing Firms
Jong-Rong Chen*
Ting-Kun Liu
Graduate Institute of Industrial Economics
National Central University, Taiwan
and
Cliff J. Huang
Department of Economics
Vanderbilt University, U.S.A.
Abstract
Using a newly constructed panel data on Electronic Commerce (e-commerce)
and R&D investments by Taiwanese manufacturing firms during 1999-2000, this
paper investigates the relationship between knowledge capital and productivity. The
knowledge spillover consists of R&D spillover and e-commerce network externalities.
Furthermore, this paper considers the impact of knowledge capital on capital
productivity and labor productivity through substitution by applying a non-neutral
production function. The empirical results show that: (1) both e-commerce and R&D
capital have positive impact on productivity; (2) e-commerce stock and R&D capital
tend to have a complementary relationship; (3) intra-industry e-commerce network
externalities and inter-industry R&D spillover demonstrate a more significant
contribution to productivity than the other two spillovers.
Draft Submitted for Presentation at Asia-Pacific Productivity Conference
(APPC), July 14-16, 2004, University of Queensland, Brisbane, Australia
Keywords:
Productivity, Taiwanese manufacturing
e-commerce network externalities
firms,
R&D
spillover,
JEL classification: D24; L60
*
Corresponding author, Graduate Institute of Industrial Economics, National Central University,
Chungli 320, Taiwan. Tel: +88634227791. Fax: +88634226134. E-mail: jrchen@cc.ncu.edu.tw
1
1. Introduction
In the dynamic environment of a knowledge-based economy, firms must try to
maintain their advantageous position under extreme competition. According to the
neoclassic school theory, preserving the continued growth of long-term technological
progression is the essential key to the success of firms. For example, the adoption of a
new production technology, new inputs, and other knowledge or inventive factors
changes the equilibrium of the market structure (Geroski and Pomroy, 1990; Chen, et
al., 2000). In the 1990s Taiwanese manufacturing firms faced a dilemma of lacking
the basic level of labor and technological human resources, and they had difficulties
in acquiring land as a result of the rise in environmental protection consciousness. In
order to hold their competitive advantage, some industries began to move their
businesses from Taiwan to Southeast Asia and/or Mainland China, while other
industries still staying in Taiwan sought to upgrade their technology and industry.
As the global economic situation changes with each passing day, with the
progression of information technology (IT), and under the application and diffusion of
Internet and e-commerce, the global operational environment has varied greatly. For
example, Dewan and Kraemer (2000) quoted a statistical survey by the International
Data Corporation (IDC) in 1997, which showed that global IT expenditure increased
from $162 billion in 1985 to $630 billion in 1996 and has a continuing rising trend.
Starting from basically zero in 1995, total electronic commerce is estimated at some
$26 billion for 1997 and predicted to reach $1 trillion in 2003-05 (OECD, 1999).
E-commerce growth to date has been quite impressive. For the sake of digesting
information to create and accumulate knowledge as soon as possible, and for the
purpose of consolidating their competence in competition and globalization,
knowledge capital or technological capital has become the focal point and core factors
of all industries. To reform products, exploit new products, or improve the process to
2
decrease production costs, firms in Taiwan invest in reasonably large amounts of
R&D expenditure. In Table 1 we observe that both R&D expenditure and the number
of U.S. patents granted to Taiwanese firms have an increasing trend, and the rank of
the number of U.S. patents granted represents the achievement of innovation by these
companies.
In the era of a knowledge-based economy, firms who want to raise their market
share should respond to market demand rapidly. As such, they apply electronic
technology to confirm that their commodities are handed over to their clients from the
process of ordering, production, and transportation. Therefore, the application of
Internet and electronic technologies becomes the firms’ new generation investments
and challenges. Table 2 shows the amount of e-commerce and the ratio of e-commerce
transaction of Taiwanese manufacturing firms, showing that both magnitudes also
have an increasing trend.
In view of this, Taiwan’s government encourages firms to engage in R&D and
innovation by financing private firms. The island’s government also helps companies
to establish an e-commerce environment of integrated supply and demand chain
enthusiastically so as to strengthen the integration of up- and down-stream industries,
promote industrial productivity and national competition, and reach the future goal of
e-business development.
Firms face continual and enormous investments in e-commerce and R&D, and they
encounter higher risks. Firms also may have difficulties in repairing and operating
these technical facilities, including excessive costs, lower cooperative willingness of
up- and down-stream industries, and a lack of capability and related technological
human resources. Moreover, the economic performance assessment is an important
indicator of firms’ investment strategies and is a symbol of success of the authorities’
tax reduction or subsidy policies. Consequently, no matter firms promote e-commerce
3
and R&D investments with help by the government or firms accelerate their own
e-commerce and R&D by themselves, it is worth exploring in depth the impact of
e-commerce and R&D on productivity.
There are a large number of studies that have assessed the contribution of R&D to
productivity at the firm level using panel data (e.g. Griliches and Mairesse, 1983;
Cuneo and Mairesse, 1984; Griliches, 1986; Goto and Suzuki, 1989; Mairesse and
Sassenou, 1991; Lichtenberg and Siegel, 1991; Hall and Mairesse, 1995; Mairesse
and Hall, 1996; Branstetter and Chen, 1999; Yang et al., 2001; Yang and Chen, 2002).
Most of them show that knowledge capital and R&D expenditure do actually have a
significant positive impact on productivity.
Despite the daily attention to the knowledge-based economy, many economists
have claimed that e-commerce is a manifestation of the Internet and related technical
progress, and e-commerce will dramatically reduce transaction costs and lead to a
growth in productivity and economy through shifting the production frontier and/or
increasing returns (Romer, 1986; Grossman and Helpman, 1991; Ajit, 1995; Aghion
and Howitt, 1998 Lucking-Reiley and Spulber, 2001). However, there is surprisingly
little empirical evidence on the impact of e-commerce on firms’ productivity.
With respect to research on e-commerce and productivity, Oliner and Sichel (2000)
adopted “back of the envelope” calculations to explore the impact of e-commerce on
productivity in the U.S. during 1996-99, finding that e-commerce has no significant
impact on MFP growth. Litan and Rivlin (2001) used “judgmental estimates” and
showed the Internet’s contribution to productivity growth to be 0.2-0.4% per year over
the last half of the 1990s. Goss (2001) examined the impact of actual Internet usage
by industry for 1997-1999 by using pooled time-series and cross-section data, using
job-related Internet usage as a proxy variable. Goss suggested that job-related Internet
usage has a positive and statistically significant impact on productivity growth of
4
roughly 0.25% per year. Konings and Roodhooft (2002) made use of a large
representative data set of Belgium firms to study empirically the impact of e-business
on productivity and cost efficiency of firms. They concluded that e-business had no
effect on productivity in small firms, but selling online and e-procurement both had
positive effect on productivity in large firms.
In these studies mentioned above, there is little empirical evidence on the impact of
e-commerce on productivity, because e-commerce is a recent and rapidly evolving
phenomenon and the measurement of e-commerce is difficult (Fraumeni, 2001). The
above empirical studies implemented input-side Internet usage (Goss, 2001), indirect
variable (Oliner and Sichel, 2000), or a dummy as a proxy of e-business or
e-commerce (Konings and Roodhooft, 2002). However, Internet usage prompts
problems of measurement and quantification, and the application of electronic
technology is comprehensive and not limited to Internet.1 In addition, Internet usage
and dummy variables obviously cannot be ample enough to reflect the empirical
dynamic situations of firms that engage in e-commerce investment. Therefore, this
paper adopts an output-side indicator of e-commerce transaction as a proxy of
e-commerce, which would be a proper and alternative way to measure the impact of
e-commerce.
Romer (1986) considered that knowledge enables a firm or an industry to
successfully adopt advanced technology that will naturally spill over to other firms or
industries. Carlsson (1997) suggested that the most important features of
technological systems are the characteristics of knowledge and spillover mechanisms,
which determine the potential spillovers. 2 There are two types of network
1
For details, please see Table 2: Procurement and sale via e-mail and Internet in Taiwan, Table 3:
Electronic technologies adoption in Taiwan, and Table 4: The ways of communication between
manufacturers and up and down-stream partners.
2
Carlsson (1997) defined the technology systems as factory automation, electronics and computers,
biotechnology, and power technology.
5
externalities, direct and indirect effect. The direct network effect comes from an
increase in users, while the indirect effect comes from the development of
applications (Katz and Shapiro, 1985; Mun and Nadiri, 2002). Therefore, another
feature of e-commerce capital that distinguishes it from other traditional inputs is that
it may generate considerable economic externalities. When the number of firms
engaged in e-commerce and the items of electronic technologies adoption from firm
i’s related industries are increasing, this could be helpful for firm i to enhance
productivity. For instance, Taiwan Semiconductor Manufacturing Company (TSMC)
has proposed the concept of Virtual Fab, whereby TSMC can move its wafer factory
into IC clients’ backyards via the Internet. More than 500 clients of TSMC through
the TSMC-Online platform can trace the percentage of scheduled progress and an
analysis of defects. This mechanism also helps them to save production costs and
shorten the product’s time to enter the market. Consequently, we place more emphasis
on the network externalities of e-commerce in our empirical model.
Most analyses of the impact of e-commerce and/or R&D on productivity focus on
manufacturing firms in developed economies, such as the U.S., Europe, and Japan.
Similar studies on this issue in newly industrialized economies (NIEs) have been
scarce, and they neither estimate the impact of e-commerce and R&D on productivity
simultaneously nor apply firm-level panel data to detect this issue. Hence, the result
from Taiwan’s experience could be a consultation for other developing countries.
In order to understand the impact on productivity of the rapid increase in
e-commerce and R&D spending in recent years, this paper use a newly constructed
panel data of Taiwanese manufacturing firms during 1999-2000. We investigate the
relationship between knowledge capital (including both e-commerce and R&D
investments) and productivity with the consideration of externalities. We also employ
an output-side indicator of e-commerce transaction to construct e-commerce stock.
6
Furthermore, this paper considers the impact of knowledge capital on capital
productivity and labor productivity through substitution by applying a non-neutral
production function.The Generalized Method of Moments (GMM) approach is robust
in the presence of heterocedasticity across firms and the correlation of disturbances
within firms over time, and it is adopted to acquire more efficient estimators in this
paper.
The rest of this paper is organized as follows. Sections 2 and 3 present the
empirical models and data sources. In Section 4 we analyze the econometric results.
Conclusions are in Section 5.
2. Empirical framework and measurement of spillover stocks
The framework in which we measure the contribution of e-commerce and R&D to
productivity follows the standard approach to analyze the contribution of R&D to
productivity. This paper assumes that the production function can be approximated by
a Cobb-Douglas function:
Yit  Aet CAPit LABit KNOit e it ,
(1)
where Y is value added, CAP, LAB, and KNO are the physical capital, labor input, and
knowledge capital, respectively, and knowledge capital consists of firms’ own
e-commerce stock (ECS) and R&D stock (RDS). The subscripts i and, t refer to the
firm i and the current year t, while λ is the rate of disembodied exogenous technical
change andεis the error term reflecting the effect of unknown factors and other
disturbances.3
If the investment decisions on e-commerce and R&D are independent, then
3
The time trend t will be replaced with time dummies in estimation.
7
equation (1) can be written as:
Yit  Aet CAPit LABit ECSit RDS it e it .
(2)
We can take logarithms to equation (2) and obtain the linear regression to implement
the estimation of the Cobb-Douglas function as shown below:
yit  a  t  capit  labit  ecsit  rdsit   it ,
(3)
where the lower-case letter denotes the logarithms of variables, and  ,  , and
especially  and  (the elasticity of value added with respect to e-commerce and
R&D) are the parameters of interest.
According to Kambil (1995), firms use the Internet to improve innovation,
production, sales, and service processes. Kambil noticed that as universities and firms
increasingly go online with working papers, technical reports, and journal articles,
individuals have instant access to relevant materials to support research and
innovation, and information about innovations or innovations themselves can be
distributed in a matter of minutes. Madden and Coble-Neal (2002) indicated the
emergence of e-commerce driven by this net adoption of Internet services and
continual technological innovation. This implies that the behavior for firms to invest
in e-commerce and R&D activities is not only for industrial transition and government
policies, but also for the reason of relative technological application connections.
In accordance with previous studies and surveys, the decisions to invest in
e-commerce and R&D investments by firms can be influenced interactively. Therefore,
this paper releases the restriction of independence between e-commerce and R&D
investments, and we allow this relationship to be a substitute or a complement. Thus,
the knowledge capital KNOit should be a function including the stock of
accumulated past e-commerce and R&D investments.
Because e-commerce and R&D investments may generate considerable economic
8
externalities, such as knowledge spillovers and network externalities, the exogenous
variations of spillovers can influence the decision of firms’ and have an impact on
productivity. When the effect of knowledge spillovers on firm i from external sources
is taken into account, it is partly determined by firms and serves as an endogenous
variable (Adams, 2000). Therefore, we take account of spillovers of e-commerce and
R&D into our model, and we can express KNOit as follows:
KNOit  f ( ECSit , RDS it , KSit ) ,
(4)
where KSit is the stock of spillover.
To account for the different abilities of firms to internalize other firms’ knowledge,
equation (4) is extended by adding weights, wij, which stands for firm i’s ability to
internalize pieces of firm j’s knowledge stock. The larger these weights are, the more
firm i can gain from firm j’s knowledge stock. Bernstein (1988) provided the
following indicator:
N
KSi   wij K j ,
j i
(5)
where wij represents firm i’s ability to absorb firm j’s knowledge stock, and K j is
the knowledge stock for firm j..
This paper follows the work of Chen and Lu (2002) to construct two types of R&D
spillover stocks - Rstra and Rster. Term Rstra represents the intra-industry R&D
spillover, while Rster represents the inter-industry R&D spillover by using the
following definitions, respectively:
N
Rstrai   wi1RDS j
(6)
j i
N
Rsteri   wi 2 RDS j ,
(7)
j i
where wi1 is the ratio of firm i’s R&D expenditure to firm i’s 4-digit industry’s total
R&D expenditure and wi 2 is the R&D expenditure of firm i’s 4-digit industry’s total
9
R&D expenditure to firm i’s 2-digit industry’s total R&D expenditure.4
In order to take into consideration the effect of e-commerce network externalities,
we modify the of model Chen and Lu (2002) to construct two types of e-commerce
network externalities stocks. Term Estra represents the intra-industry e-commerce
network externalities, while Ester represents the inter-industry e-commerce network
externalities by using the following definitions, respectively:
N
Estrai   ni1ECS j , ni1  dei1  iei1
(8)
i j
N
Esteri   ni 2 ECS j , ni 2  dei 2  iei 2 ,
(9)
i j
where ni1 is the weight of intra-industry network externalities, and is composed of
two effect:
1) direct effect dei1 is the ratio of the number of firms engaging in
e-commerce of firm i’s 4-digit industry to the total number of firms in firm i’s 4-digit
industry.
2) indirect effect iei1 is the ratio of firm i’s electronic technology
application index to firm i’s 4-digit industry’s total electronic technology application
index. 5 Term ni 2 is the weight of inter-industry network externalities and is
composed of two effect:
1) direct effect dei 2 is the ratio of the number of firms
engaging in e-commerce in firm i’s 4-digit industry to the number of firms engaging
in e-commerce in firm i’s 2-digit industry. 2) indirect effect iei 2 is the ratio of firm
i’s 4-digit industry’s total electronic technology application index to firm i’s 2-digit
industry’s total electronic technology application index.
Since each of these elements interacts with one another, these interactive effect and
the specific functional form of KNOit are unknown. Here we take the Tornquist input
4
Chen and Lu (2002) used a 2-digit industry instead of a 3-digit one to avoid the estimation problems
due to similarity among industries.
5
Firm i’s electronic technology application index is measured by the ratio of the number of items of
electronic technology application that firm i has adopted to the total number of items of electronic
technology application in our survey data. Please see Table 3 for details on items of electronic
technology application.
10
value index to construct the KNOit index for our empirical model as follows:
KNOit  [( X 1it )1 ( X 2it ) 2 ( X 3it )3 ( X 4it ) 4 ( X 5it )5 ( X 6it )6 ( X 7 it )7 ]
(10)
where X1 ~ X 7 denotes ECS, RDS, ECS  RDS , Estra, Ester, Rstra, and Rster,
respectively. i is the share of each variable of knowledge capital in the total value
of the knowledge capital bundle.
Besides, intensive use of knowledge capital is likely to raise the capital (CAP)
productivity and labor (LAB) productivity through substitution (Lucking-Reiley and
Spulber, 2001). Therefore, the end result to production is a non-neutral shift of
observed output; not only will productivity of inputs change, but also will the
marginal rate of technical substitution (Huang and Liu, 1994). In this paper we apply
a non-neutral production function different from the conventional model of equation
(1). Our reformation of non-neutral production function is the following:
Yit  Ae1t  2 knoit CAPit1  2 knoit LABit1   2 knoit e it
(11)
Taking logarithms to equation (11) we obtain equation (12), which can be used to
implement the relationship between knowledge capital and productivity.
yit  a  1t  2 knoit  1capit   2 knocapit  1labit   2 knoitlabit   it
(12)
Equation (12) will be explored in this paper, and the general specification of equation
(3) will also be discussed in our study for comparison.
In order to investigate the impact of related variables of knowledge capital on
productivity, we use the estimated coefficients of equation (12) to calculate the partial
average productivity (PAP) in the following way:
PAPi 
yi( 2 )  yi(1)
X i (t )  X i (t 1)
(13)
where yi(1) and yi( 2 ) are the output contribution of variables i at time t-1 and t
respectively while holding other variables constant, X i ’s are defined in equation
11
(10).
On investigating the relationship between knowledge capital and productivity, as is
common with panel data, we allow for the existence of individual effect which are
potentially correlated with the right-hand side regressors, such that:
 it  vit  ui ,
(14)
where, ui is a firm effect that corresponds the permanent, unobserved heterogeneity
across firms, but not within a firm over time, and vit is a “white noise” error term,
representing a period-specific stock for firm i that is assumed to be independent across
firms and over time. Using a “within firm” panel estimator is a standard method to
eliminate the individual effect (Yang et al., 2001). Mairesse and Hall (1996) and Yang
et al. (2001) studied the relationship between knowledge capital and productivity.
They proved that the Generalized Method of Moments (GMM) is better in robustness
in the presence of heteroskedasticity across firms and a correlation of disturbances
within time. It can also be efficient even under a weak assumption on the disturbance.
Therefore, this paper adopts the GMM developed by Arellano and Bond (1991) and
Ahn and Schmidt (1992).
3. Data sources and variable definitions
Owing to a lack of integrated data about e-commerce, R&D, and related firms’
operational information in Taiwan, this paper merges the automation and e-commerce
survey and the plant survey of Taiwanese manufacturing firms during 1999-2000
provided by MOEA to obtain a more complete database. At the same time, taking into
account the representation of industrial distribution and the length of time period, we
construct a balanced panel data of 3,698 Taiwanese manufacturing firms for the
survey period of 1999-2000.
In equation (2) of this paper, due to materials not included in the model, we use
12
value-added (VAL) as the proxy variable of dependent variable Y. According to
MOEA’s definition, we measure VAL by the following specification:
VAL is equal to
total output sales subtracting intermediate inputs. Total output sales have been
deflated using a wholesale price index and intermediate inputs have been deflated
using a price index of intermediate inputs. With respect to the explained variables of
the right-hand side equation, physical capital (CAP) is total fixed capital and we use a
capital price index to adjust for inflation, where labor (LAB) is the number of workers.
Internet usage incurs problems of measurement and quantification, and the
application of electronic technology is comprehensive and not limited to the Internet.6
In addition, Internet usage and dummy variables obviously are not ample enough to
reflect the empirical dynamic situations of firms engaging in e-commerce investment.
Therefore, this paper adopts an output-side indicator of e-commerce transaction as a
proxy of e-commerce. Furthermore, e-commerce stock (ECS), R&D stock (RDS), and
spillover stock (KS) are used to construct knowledge capital.
Our measurements of e-commerce and R&D stocks follow that of Mairesse and
Hall (1996) and Yang and Chen (2002). They define the equation of knowledge
capital stock as follows:
K t  I t  (1   ) I t 1  (1   ) 2 I t  2   ,
(16)
where K represents the e-commerce and R&D stocks, I is e-commerce transaction or
R&D expenditure, and  is the depreciation rate. Because of the limitation of the
survey period, this paper adopts only one lagged year to construct knowledge capital
stock. We assume that the above knowledge capital stocks have a constant
depreciation rate of 15%, according to the general setting of previous papers.7
6
For details, please see Table 3: Procurement and sale via e-mail and Internet in Taiwan, Table 4:
Electronic technologies adoption in Taiwan, and Table 5: The ways of communication between
manufacturers and up- and down-stream partners.
7
Please see Hall and Mairess (1995), Hall and Mairess (1996), Yang et al. (2001), Yang and Chen
13
As for the spillover effect, this paper considers the spillover effect of intra-industry
and inter-industry. The total (e-commerce or R&D) spillover effect is the sum of
intra-industry spillover effect (Estra or Rstra) and inter-industry spillover effect (Ester
or Rster). We apply equations (6)-(9) to estimate intra-industry and inter-industry
spillover effect. Table 5 gives sample statistics for our key variables.
4. Empirical results
In this section we present the results of the impact of e-commerce and R&D activities
on productivity, which were achieved by applying GMM. The previous equation (2) is
considered as the starting point of the analysis.8 We separate equation (2) into two
estimative modes: mode (i) is a regression under the assumption that e-commerce
and R&D activities are independent (i.e. equation (3) of this paper); and mode (ii) is a
regression that releases the restriction of e-commerce and R&D activities being
independent, and it subsumes the e-commerce and R&D spillover effect of
intra-industry and inter-industry under consideration, and we use those variables to
construct the knowledge capital (i.e. equation (12) of this paper).
This paper tests the over-identifying restrictions by the Hansen J statistic, which is
consistent in the heteroskedasticity presence. The test results of all modes do not
reject the validity of instruments, as indicated by accepting the null hypothesis with a
p-value above 0.05. The empirical results are presented in Table 6.
The second column of Table 6 shows the empirical results of mode (i), where the
labor coefficient is higher than the capital coefficient, and they both have significant
impact on productivity level. The first kind of knowledge capital, e-commerce stock
(2002), and Chen and Lu (2002). This paper also tries to use other depreciation rates (20% and 30%) to
construct knowledge capital stocks, and we find that there is no significant difference between
empirical results.
8
In this estimation we use all independent variables and all lagged capital as instruments.
14
logECS and its parameter  is positive significance at the 1% statistical level (the
coefficient of e-commerce stock elasticity is 0.032). This provides evidence that
Taiwanese firms devoting more e-commerce investment efforts have a better
productivity performance, and this result is consistent with Konings and Roodhooft
(2002), Litan and Rivlin (2001), and Goss (2001) in their findings. Goss suggested
that job-related Internet usage had a positive and statistically significant impact on
productivity growth of roughly 0.25% per year. This implies that there is considerable
room for productivity promotion for Taiwanese firms to invest in electronic activities
in order to construct a better e-business and/or e-commerce environment.
We also explore another important knowledge capital, R&D stock, and detect the
impact of R&D stock on productivity. The result shows that the coefficient of R&D
stock logRDS is 0.203, and the estimated R&D stock contribution to productivity is
on the high end of Mairesse and Sassenou (1991) and Mairesse and Hall (1996). Their
estimated R&D stock elasticity range of 0.008-0.210 in developed countries. This
magnitude is also higher than the average R&D elasticity of 0.036 for Taiwanese
manufacturing firms in 1990-1997 as conducted by Yang and Chen (2002). This
reveals that the contribution to productivity by investing in R&D activities by
Taiwanese manufacturing firms is not inferior to those advanced countries. The above
results have positive and constructive meanings to inspire Taiwan’s government to
encourage and assist firms in investing and popularizing e-commerce and R&D
activities.
The third column of Table 6 presents the estimates of the non-neutral production
function of mode (ii). In order to investigate whether knowledge capital has influence
on productivity, we use the F-test to test the joint null hypothesis about the parameters
in this non-neutral production function model. The joint null hypothesis is H0:
2  2  2  0 , and the alternative hypothesis is H1: 2  0 , 2  0 , or 2  0 , or
15
all are nonzero. Since F=16.55 and P-value <0.01, we reject H0 and conclude that at
least one of them is not zero, and thus knowledge capital has an effect upon
productivity.
In the third column of Table 6, we find that all coefficients of parameters are
significance at the 5% statistical level. In order to investigate the impact of knowledge
capital on productivity of mode (ii) further, this paper estimates the partial average
productivity of each variable of knowledge capital. The results of the partial average
productivity estimates are shown in Table 7.
With respect to the empirical results of the relationship between e-commerce stock
and productivity, we find that partial average productivity of e-commerce stock is
small than the contribution of R&D stock toward output. The magnitude of partial
average productivity of e-commerce stock is approximately 0.045%, while the partial
average productivity of R&D stock is 0.272%. The gap between R&D contribution to
productivity and e-commerce contribution could be the reason why Taiwanese firms
did not widely adopted electronic technologies and the popularization of e-commerce
is not so common among all industries.9 In addition, many of the external benefits
from e-commerce, including automation of transaction, cost saving, added
convenience and reorganization of firms, that will gain productivity and are not
properly showed in the productivity statistics (Litan and Rivlin, 2001; Lucking-Reiley
and Spulber, 2001). Given the low usage rate of e-commerce activities, and the huge
amount of e-commerce transaction, there exists significant room for enhancing
productivity through a substantial expansion in e-commerce activities. Therefore, just
as Fraumeni (2001) mentions that although at this point e-commerce may represent a
relatively small percentage of both B2B and B2C, its future size and impact on
9
For example, the ratio of the number of firms that invest at least in one survey year in e-commerce to
the total number of observations is approximately 14.5% in our data set.
16
e-business process and economic growth may be large.
Moreover, the sign of partial average productivity for the interaction effect between
e-commerce and R&D stock is positive, implying that e-commerce and R&D
activities tend to have a complementary relationship in the production of knowledge
during the period of our study. We can infer that R&D and e-commerce activities
could integrate with each other to improve productivity, and this finding also
intensifies the credibility of our model’s setting.
As for the spillover effect, the intra-industry e-commerce network externalities
have played a more important role than the inter-industry e-commerce externalities in
terms of their impact on productivity. Its partial average productivity is 2.657% as
shown in Table 7. On the other hand, the inter-industry R&D spillover has a greater
contribution to productivity than the intra-industry R&D spillover. We also find that
e-commerce network externalities and R&D spillover have more significant impact on
productivity than e-commerce and R&D stocks. This connotes that some studies may
overestimate the impact of the knowledge capital of productivity when they neglect
the consideration of putting e-commerce and R&D spillover effect into their model.
Our empirical finding verifies the claim of Carlsson (1997) that the most important
features of technological systems are the characteristics of knowledge and spillover
mechanisms.
5. Conclusion
Over the past decade, quite a few studies have adopted R&D as the proxy variable of
knowledge capital in order to explore the relationship between knowledge capital and
productivity. Nevertheless, knowledge capital is not restricted to R&D activities only;
e-commerce activities also show a link to knowledge capital. For this reason, this
study uses a newly constructed panel data on e-commerce and R&D investments of
17
Taiwanese manufacturing firms during 1999-2000 to investigate the impact of R&D
and e-commerce on productivity. The knowledge spillover consists of R&D spillover
and e-commerce network externalities. Furthermore, this paper considers the impact
of knowledge capital on capital productivity and labor productivity through
substitution by applying a non-neutral production function.
The empirical results of this study indicate that: (1) both e-commerce and R&D
capital have positive impact on productivity; (2) e-commerce stock and R&D capital
tend to have a complementary relationship; (3) intra-industry e-commerce network
externalities and inter-industry R&D spillover demonstrate a more significant
contribution to productivity than the other two spillovers. This means that a firm’s
productivity is not only affected by self-owned inputs, but also affected by the
technological knowledge diffusion from up- and downstream firms or the same
industry.
Finally, our empirical results would be a good reference for other developing
countries. For further studies, the decomposition of e-commerce transaction into sales
and procurement (i.e. e-sales or e-procurement), and the relationship between
knowledge capital and spillover variables could be discussed in a detailed and
accurate way in the future if a more detailed data were available.
References
Adams, James D. (2000) “Endogenous R&D Spillovers and Industrial Research
Productivity,” National Bureau of Economic Research Working Paper, No.
7484.
Aghion, Philippe and Peter Howitt (1998) “Endogenous Growth Theory,” MIT Press,
Cambridge, MA.
Arellano, Manuel and Stephen Bond (1991) “Some Tests of Specification for Panel
18
Data: Monte Carlo Evidence and an Application to Employment Equations,”
Review of Economic Studies, April 1991, 58 (2), 277-97.
Ahn, Seung C. and Peter Schmidt (1992) “Efficient Estimation of Panel Data with
Exogenous and Lagged Dependent Regressor,” Journal of Econometrics, 68,
5-27.
Basant, Rakesh and Brian Fikkert (1996) “The Effect of R&D, Foreign Technology
Purchase, and Domestic and International Spillovers on Productivity in Indian
Firms,” Review of Economics and Statistics, 78(2), 187-99.
Bernstein, Jeffrey I. (1988) “Costs of Production, Intra- and Inter-industry R&D
Spillovers: Canadian Evidence,” Canadian Journal of Economics, 21 (2),
324-47.
Branstetter, Lee G. and Jong-Rong Chen (1999) “The impact of technology transfer
and R&D on productivity growth in the Taiwanese electronics industry: Microeconometric analysis using plant-level data” Paper presented at the international
conference on the role of technology transfer in East Asian economic growth,
University of California at Davis, U.S.A.
Carlsson, Bo (1997) “On and Off the Beaten Path: The Evolution of Four
Technological Systems in Sweden,” International Journal of Industrial
Organization, 15 (6), 775-99.
Chamberlin, G. (1984) “Panel Data, in Z. Griliches and M. Intriligator eds.,” The
Handbook of Econometrics, Volume II, Amsterdam, North Holland.
Chen, Jong-Rong, Chih-Hai Yang and Wen-Nueng Cheng (2000) “The diffusion of
Multiple Technologies-Evidence from CNC and CAD/CAM,” Journal of Social
Science and Philosophy, 12(3), 433-457.
Chen, Jong-Rong and Wen-Cheng Lu(2003) “Effect of R&D Spillover on Firm
Growth- The Case of the Electronics and Electrical Machinery Industry in
19
Taiwan,” paper presented at the WEAI meeting in Denver, USA.
Cuneo, Philippe and Jacques Mairesse (1984) “Productivity and R&D at the Firm
Level in French Manufacturing,” R&D, Patents, and Productivity, NBER
Conference Report. Chicago and London: University of Chicago Press 375-92.
Dewan, and K. L. Kraemer (2000) “Information Technology and Productivity:
Evidence from Country-Level Data,” Management Science, 46:4, 548-562.
Fraumeni, Barbara M. (2001) “E-Commerce Measurement and Measurement Issues,”
American Economic Review, 91 (2), 318-22.
Geroski, Paul A. and Richard Pomroy (1990) “Innovation and the Evolution of
Market Structure,” The Journal of Industrial Economics, 38: 3, 299-314.
Goss, Ernest (2001) “The Internet’s contribution to U.S. Productivity Growth,”
Business Economics, 32-42.
Goto, Akira and Kazuyuki Suzuki (1989) “R&D Capital, Rate of Return on R&D
Investment and Spillover of R&D in Japanese Manufacturing Industries,”
Review of Economics and Statistics, 71 (4), 555-64.
Grossman, Gene M. and Elhanan Helpman (1991) “Innovation and Growth in the
Global Economy,” MIT Press, Cambridge, MA.
Griliches, Zvi (1986) “Productivity, R&D, and the Basic Research at the Firm Level
in the 1970's,” American Economic Review, 76 (1), 141-54.
Griliches, Zvi and Jacques Mairesse (1983) “Comparing Productivity Growth: An
Exploration of French and U.S. Industrial and Firm Data,” European Economic
Review, 21(1-2), 89-119.
Hall, Bronwyn H. and Jacques Mairesse (1995) “Exploring the relationship between
R&D and Productivity in France Manufacturing Firms,” Journal of
Econometrics, 65, 263-293.
Huang, J. Huang and Jin-Tan Liu (1994) “Estimation of a Non-Neutral Stochastic
20
Frontier Production Function,” The Journal of Productivity Analysis, 5,
171-180.
Kambil, Ajit (1995) “Electronic Commerce: Implications of the Internet for Business
Practice and Strategy,” Business Economics, 30 (4), 27-33.
Katz, Michael L. and Carl Shapiro (1985) “Network Externalities, Competition, and
Compatibility,” American Economic Review, 75(3), 424-40.
Konings, Jozef and Filip Roodhooft (2002) “The Effect of E-Business on Corporate
Performance: Firm Level Evidence,” De Economist, 150 (5), 569-81.
Lichtenberg, Frank R. and Donald Siegel (1991) “The Impact of R&D Investment on
Productivity-New Evidence Using Linked R&D-LRD Data,” Economic Inquiry,
29 (2), 203-29.
Litan, Robert E. and Alice M. Rivlin (2001) “Projecting the Economic Impact of the
Internet,” American Economic Review, 91(2), 313-317.
Lucking-Reiley, David and Daniel F. Spulber (2001) “Business-to-Business Electronic
Commerce,” Journal of Economic Perspectives, 15(1), 55-68.
Madden, Gary and Grant Coble-Neal (2002) “Internet Economics and Policy: An
Australian Perspective,” The Economic Record, 78 (242), 343-357.
Mairesse, Jacques and Bronwyn Hall H. (1996) “Estimating the Productivity of
Research and Development in French and United States Manufacturing Firms:
An Exploration of Simultaneity Issues with GMM Methods,” International
productivity
differences:
Measurement
and
explanations,
Amsterdam,
North-Holland, 233, 285-315.
Mariesse Jacques and Mohamed Sassenou (1991) “R&D and Productivity: A Survey
of Econometric Studies at the Firm Level,” Science-Technology Industry Review,
8, 317-348.
Mun, Sung-Bae and M. Ishaq Nadiri (2002) “Information Technology Externalities:
21
Empirical Evidence from 42 U.S. Industries,” NBER Working Paper, No. w9272,
1-43.
OECD (1999) “Growth of Electronic Commerce: Present and Potential,” The
Economic and Social Impact of Electronic Commerce, Preliminary Findings and
Research Agenda, 27-54.
Oliner, Stephen D. and Daniel E. Sichel (2000) “The Resurgence of Growth in the
Late 1990’s: Is Information Technology the Story?” Journal of Economic
Perspectives, 14(4), 3-22.
Romer, Paul M., (1986) “Increasing Returns and Long-Run Growth,” Journal of
Political Economy, 94(5), 1002-37.
Wong, Poh-Kam and Phyllisis M. Ngin (1997) “Automation and Organizational
Performance: The Case of Electronics Manufacturing Firms in Singapore,”
International journal of Production Economics, 52(3), 257-268.
Yang,
Chih-Hai
and
Productivity-Evidence
Jong-Rong
from
Chen
Taiwanese
(2002)
“R&D,
Manufacturing
Patents
Firms,”
and
Taiwan
Economic Review, 30 (1), 27-48.
Yang, Chih-Hai, Jong-Rong Chen and Lee G. Branstetter (2001) “Technology
Sourcing, Spillover and Productivity: Evidence from Taiwanese Manufacturing
Firms,” Paper presented at the 28th EARIE Conference, Dublin, Ireland.
22
Table 1: R&D and Patents of Taiwan (1998-2002) (Million US$)
R&D
Total Government Budget R&D /GDP Number of U.S. Patents
Expenditures Appropriations for R&D
(%)
Granted and the Rank
1998
8,600.3
NA
1.97
3,100 (7)
1999
9,616.8
NA
2.05
3,693 (5)
2000
10,326.0
3,336.4
2.05
4,667 (4)
2001
11,014.2
3,519.4
2.16
5,371 (4)
2002
12,246.6
4,313.7
2.30
5,431 (4)
Source: Main Science and Technology Indicators, 2003, OECD, U.S. Patent and
Trademark Office, and Intellectual Property Office, MOEA.
Note: Number of U.S. patents granted and the rank exclude new designs.
Table 2:
Procurement and Sales via E-mail and Internet in Taiwan (2001-2002)
Procurement
Sales via
via e-mail
e-mail
(billion NT$) (billion NT$)
Procurement via e-mail Sale via e-mail /
/ total procurement
total sales revenue
expenditure（%）
（%）
2001
1,699
2,539
6.51
5.97
2002
2,140
3,177
7.39
6.81
Procurement
Sales via
via Internet
Internet
(billion NT$) (billion NT$)
Procurement via Internet Sales via Internet /
/ total procurement
expenditure（%）
total sales revenue
（%）
2001
903
2,095
3.46
4.93
2002
1,741
2,747
6.01
5.89
Source: The survey of automation and e-commerce of manufacturing firms during
1995-2002, The Department of Statistics, a staff unit of the Ministry of Economic
Affairs (MOEA).
23
Table 3:
Electronic Technologies Adoption (2002) (Multiple Choice, %)
Metal and Information and
Chemical
Total Engineering Electronics
Industries
Industries
Industries
Network construction 63.98
ERP
59.02
Civil
Industries
57.39
70.52
58.54
64.90
59.85
67.70
54.74
43.71
SFCS
37.79
31.28
35.11
42.28
47.02
PDM
32.48
35.71
36.59
25.47
27.48
EDI
27.91
30.54
32.44
22.22
21.19
ASP
21.00
20.44
21.04
18.70
24.50
CRM
16.27
15.27
17.48
15.45
15.89
SCM
15.58
19.21
17.33
10.84
12.58
E-marketplace
8.62
9.85
8.74
6.78
8.94
Other
0.29
0.49
0.30
0.27
0.00
Note: Network contains Internet, Intranet, and Extranet construction and management.
ERP means enterprise resource planning, SFCS means shop floor control system,
PDM means product data management, EDI means electronic data interchange, ASP
means application service provider, CRM means customer relations management, and
SCM is supply chain management.
Source: The same as Table 2.
Table 4: The Methods of Communication Between Manufacturers and Up- and
Down-stream Partners (Multiple choice, %)
Total
Metal and
Engineering
Industries
Information and
Electronics
Industries
Chemical
Industries
Civil
Industries
2000 E-mail
2002
92.47
91.76
90.77
87.72
95.77
96.39
92.73
89.16
89.14
89.40
2000 EDI
2002
2000 Internet
2002
23.52
24.73
39.39
57.80
25.68
25.93
37.39
57.31
27.88
25.41
43.85
61.07
19.39
25.30
37.58
53.49
17.57
20.63
36.74
55.59
Source: The same as Table 2.
24
Table 5: Statistics on Variables (After Deflation), 1999-2000 (Million NT$)
Variable
Name
Mean (S.E.)
Value added
Capital stock
Number of employees
E-commerce stock
R&D stock
E-commerce network externalities stock
of intra-industry
E-commerce network externalities stock
of inter-industry
VAL
CAP
LAB
ECS
RDS
Estra
418.026 (1864.660)
1,032.440 (6863.457)
177.186 (394.483)
132.274 (2450.291)
29.398 (199.031)
713.670 (66.289)
Ester
2,689.660 (8280.890)
R&D spillover stock of intra-industry
R&D spillover stock of inter-industry
Rstra
Rster
20.858 (123.067)
288.356 (1022.265)
Note: The numbers in parentheses are standard errors.
Table 6: Estimate of the Empirical Model (1999-2000)
(i)
(ii)
log ECS
log RDS
0.0326 a
0.2030 a
log CAP
log LAB
logKNO
logKNO log CAP
logKNO log LAB
constant
t
R2
0.1271 a (0.0277)
0.7090 a (0.0573)
(0.0125)
(0.0264)
0.882
0.2881a (0.0125)
0.8365a (0.0200)
0.0380a (0.0084)
-0.0010b (0.0023)
-0.0058 b (0.0036)
-3.3563a (0.8860)
0.0318a (0.0100)
0.855
Note: 1. The numbers in parentheses are standard errors. 2. a, b, and c represent the
1%, 5%, and 10% significant levels, respectively.
25
Table 7: Partial Average Productivity of knowledge capital
Partial Average Productivity (%)
ECS
RDS
0.045
0.272
0.181
2.657
1.362
0.174
1.379
ECS  RDS
Estra
Ester
Rstra
Rster
26