Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
The continuing debate on
firm performance: A multilevel approach to the IT sectors
of Taiwan and South Korea
☆
Yi-Min Chen
⁎
Department of Asia-Pacific Industrial and Business Management, National University of Kaohsiung, 700 Kaohsiung University Road, Nanzih, Kaohsiung 811, Taiwan
a b s t r a c t
a r t i c l e i n f o
Article history: Received 1 April 2008
Received in revised form 1 March 2009 Accepted 1 April 2009 Keywords: Performance Industry organization Strategic management Technological diversification Hierarchical linear modeling
In the economics and strategyfields, researchers seek to understand the antecedents of firm profitability.
How and why do certain private enterprisefirms develop competitive advantages in environments of rapid
technological change while otherfirms do not? This study extends recent variance decomposition research in
three ways. First, this work compares IT sectors in Taiwan and South Korea by using the Standard & Poor's
Compustat® Global Vantage database. Second, this investigation tests industry andfirm effects using both
the multilevel approach of hierarchical linear modeling (HLM) and the conventional variance components
approach (VCA). Third, this study explores the question of why there are significant profitability differences
among technologicalfirms even with similar industrial structural characteristics and leveraged resources and
capabilities in the same IT industry. This study uses data from the U.S. Patent Office to estimate technological
diversification at the level of firm resources for knowledge-based relatedness for the IT firms of Taiwan and
South Korea. The empirical resultsfind that firm effects have great impact on performance of the IT sectors of
Taiwan and South Korea when estimated by either HLM or VCA. However, industry effects dominatefirm
effects on South Korea's IT sectors when the variance is estimated by HLM. From the perspective of conducting patents innovation, both of the specialized and diversified corporate strategies are matter to the
development of these IT sectors, and South Korea's ITfirms are more technologically diversified than those
firms in Taiwan.
© 2009 Elsevier Inc. All rights reserved.
1. Introduction
What matters most to profitability? Industrial organization
economics (IO) and the resource-based view (RBV) of thefirm have
disagreed over this question for more than 60 years. The originator of IO in the late 1930s, Ed Mason, argues that market structure
contributes greatly to firm profitability (Mason, 1939). Meanwhile,
from the RBV perspective,Nourse and Drury (1938)suggest that
firm-specific influences, such as management skills, basically determine
firm advantages and performance. Until recent variance decomposi-tion studies started trying to determine the relative importance of
industry, corporate, and business segment on profitability, little
research had been done to resolve their debate. The current empirical debate about the importance of industry, corporate, and
business-segment effects began withSchmalensee (1985), followed by studies
by Wernerfelt and Montgomery (1988); Kessides (1990); Rumelt (1991);Roquebert, Phillips and Westfall (1996);McGahan and Porter
(1997, 2002); Brush and Bromiley (1997); Brush, Bromiley, and Hendrickx (1999);Bowman and Helfat (2001);Hawawini, Subrama-nian, and Verdin (2003);Ruefli and Wiggins (2003);Hough (2006);
Misangyi, Elms, Greckhamer, and Lepine (2006); andShort, Ketchen, Palmer, and Hult (2007). All of these studies decompose the variance
of business orfirm returns (or business market share, in one study)
into components associated with industry, corporate, and business-segment effects, some including year effects and interaction terms as
well. Later studies weigh only the influence of industry effects
(Powell, 1996), the influence of industry and firm effects together (Cubbin and Geroski, 1987; Mauri and Michaels, 1998), the influence
of industry and organizational effects together (Hansen and
Werner-felt, 1989), or the influence of industry, strategic group and firm on
performance (Short et al., 2007). Unfortunately, the ability to
generalize the results of such studies, particularly variance
decom-position, is limited by their breadth of data and statistical difficulties.
While previous studies along the lines ofRumelt's (1991)work use
the Compustat® database that included mostly U.S.firms, they have
neglectedfirms outside the United States in explaining how much
industry, corporate and business-level factors affectfirm performance
except theKhanna and Rivkin's (2001)work. Additionally, in order to
obtain a deeper understanding of the processes that generate
favorable and unfavorable effects at the industry andfirm-specific
levels,McGahan and Porter (2002)have suggested that they need
☆ An earlier version of this paper was presented at the Annual Conference of CSMOT in 2007. The author thanks two anonymous reviewers, JBR Special Issue Editor Kun-Huang Huarng, and the CSMOT conference reviewers for their thoughtful comments on previous drafts of this paper. The work described in this paper was partially supported by a grant from the National Science Council of Taiwan (NSC 96-2416-H-390-008-MY2).
⁎ Tel.: +886 7 591 9241; fax: +886 7 591 9430. E-mail address:ymchen@nuk.edu.tw.
0148-2963/$– see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.jbusres.2009.04.004
Contents lists available atScienceDirect
detailed studies at the sectoral level. In this study, the relative
importance of industry and firm factors on profitability differences
among ITfirms by employing business databases of Taiwan and South
Korea is assessed.
Taiwan's IT industry has been outstanding during the last two
decades (Chang and Yu, 2001). For example, Taiwan's IT sectors
ranked third in the world, with total output of US$34 billion in 1998, ahead of larger nations like South Korea, and behind only the United
States and Japan (Saxenian and Hsu, 2001). As both consumers and
producers of IT hardware products, Taiwan and South Korea have strengthened their capacities to develop new technologies and successfully supply IT hardware products to the global marketplace (National Science Board, 2006). For example, the IT sectors of Taiwan and South Korea increasingly rely on creating and earning more U.S. patents, a leading indicator of future competitive advantage, and move ahead of France and the United Kingdom to rank third and fourth as the residences of foreign inventors who obtained patents in the United
States. Therefore, scientific, engineering and technological
innova-tions that emerge from research and development (R&D) activities enable high-wage economies like Taiwan and South Korea to engage in today's highly competitive global marketplace. However, techno-logical competencies and specialization patterns vary widely between the IT industries of Taiwan and South Korea. While Taiwan's IT manufacturers have secured the leading positions on the
semicon-ductor, office and computer machinery production in the world, South
Korea's IT sectors have mostly focused on the production of telecommunications equipment. In other words, the foregoing over-arching considerations regarding independent IT development of Taiwan and South Korea indicate that different technological competencies and specialization capacities exist in these IT sectors allowing them to compete with each other in global markets.
Most of the variance decomposition studies use two statistical
techniques to estimate the influences of industry, corporate, and
business-segment on profitability: nested analysis of variance
(ANOVA) and variance components analysis (VCA). Although they are widely applied approaches in this type of studies, they have recently been criticized because of their methodological shortcomings
and incomplete interpretation of results (Brush and Bromiley, 1997;
Brush et al., 1999; Bowman and Helfat, 2001; Ruefli and Wiggins, 2003; Hough, 2006; Misangyi et al., 2006). Brush and Bromiley (1997), for example, question the merit of using VCA, which may produce unreliable variance estimates or estimates that
under-represent the true importance of the effect. Rumelt (1991) and
Bowman and Helfat (2001)point out the difficulties of using nested ANOVA that the corporate effects must be entered into the model before business-segment effects; otherwise the results will inaccu-rately estimate the relative importance of these two factors. Therefore, recent researchers propose new statistical methods (e.g., multilevel
methodology of hierarchical linear modeling (HLM) byHough 2006
and Misangyi et al., 2006) to explore the relative importance of
industry, corporate, and business-segment on profitability. However,
the possible shortcomings of the VCA technique do not cause us to disregard its important contributions that give us ordinary-least squares estimates under non-restrictive conditions and random effects. For comparison purposes, the VCA results are compared to the results from HLM using the same dataset in this study.
2. Identifying the IT industrialization of Taiwan and South Korea While very large business groups (the chaebol) dominate South Korea's IT industrialization, Taiwan's IT development is mainly led by small and medium-sized enterprises (SMEs), often involved in direct contract relations with U.S. and European manufacturers. For example, Taiwan Semiconductor Manufacturing Corporation (TSMC), founded as a joint venture with the Dutch multinational Philips in 1986, uses its superior manufacturing capabilities and production processes to build
the integrated circuits (IC) foundry model. TSMC breaks away from the traditional vertically-integrated modes of production in the semiconductor industry, previously monopolized by the U.S., Japan and South Korea, to establish Taiwan as one of the world's leading semiconductor producers. How can a thriving group of Taiwanese IT SMEs survive and gain the world's leading position in a fast-moving environment of incessant innovation, especially when facing intense
competition from large and existingfirms in South Korea and Japan?
Different approaches explain how different development attributes exist on the IT industrializations of Taiwan and South Korea.
First, the theory of industrial ecosystem clarifies this situation. The
economic growth of Taiwan's IT industry is simply the product of a
unique“industrial ecosystem” that brings governmental intervention,
knowledge, technical and financial resources together (Mathews,
1997). From the perspective of industrial organization economics, the
Taiwan authorities recognize that as a latecomer industrializer they need some extra governmental interference (e.g., creating the Hsinchu Science-based Industrial Park (HSIP), providing a range of
taxation benefits and allowances, including low-interest loans, R&D
matching funds, and investment allowances) to offer IT firms
attractive terms to emulate the IT achievements in Silicon Valley. In addition, the publicly built HSIP is a facilitating platform to leverage the advanced technologies from around the world and accelerate the uptake and master of these technologies for Taiwanese IT
private-sector development (Mathews, 1997). Bringing together IT firms,
venture capital organizations, university–private sector partnerships
and interactions with other components of the industrial ecosystem has raised Taiwan as the world's leading IT producers since the 1990s. Therefore, while many Western observers have asserted that Taiwan's IT sectors are internationally competitive because of cooperation among Taiwanese rivals and active government policies in competi-tion and industrial upgrading orchestrated by the Industrial Technol-ogy Research Institute (ITRI) and other governmental agencies (Kraemer et al., 1996; Mathews, 1997; Thorbecke et al., 2002),
Hamilton and Biggart (1988)find that the South Korea state actively sponsors the extraordinary growth of selected chaebol in selected industries by careful planning and implementing strategic policies.
Second, while the governmental intervention policies underpinning the development of innovative sectors, Taiwan's IT industry is also a free market-driven and competition-oriented sector. For example, Taiwan's government has never imposed any protective tariffs on its IT hardware products. Even publicly-funded but privately-operated demonstration
ventures such as TSMC were quickly privatized and floated on the
Taiwan Stock Exchange as they achieved commercial success in international marketplace. The emergent role of free entrepreneurship
and uniquefirm-level capabilities to successful Taiwan's IT
industriali-zation leads to the importance of resource-based view of thefirm.
Taiwan's IT entrepreneurs participate in a quite advanced form of “organizational learning,” allowing firms to absorb technology and
resources from a variety of sources and“recombine” them through the
exercise of“combinative capabilities” (Kogut and Zander, 1992). For
South Korea,Choung, Hwang, and Yang (2006)indicate that cumulative
learning in technology, organization and public institution is a critical factor to upgrade technological capabilities in the innovation systems of IT industry. The process of recombining technological resources into technological capabilities represents a fusion of technology strategies at
thefirm level and international technology transfers at the level of the IT
sectors as a whole (Cusumano and Elenkov, 1994). Overall, the IT
firm-level analysis of Taiwan and South Korea can provide a dramatic
demonstration of the proposition that afirm's unique resources and
capabilities are the basis of core competencies and competitive
advantages (Brush et al., 2001).
Finally, recent analysts have moved beyond the simple state–
market debate to examine other antecedents of economic perfor-mance such as networks and geography of production. While a community of U.S.-educated entrepreneurs and engineers has
coordinated a decentralized process of reciprocal industrial upgrading by transferring capital, sharing managerial and technological skills, and facilitating collaborations among specialist producers in the U.S.
and Taiwan regions (Saxenian and Hsu, 2001), Taiwanese IT firms
benefited from a variety of strong linkages with large business groups,
foreign R&D, marketing and manufacturing affiliates, and an early
participation in international production networks that have facili-tated their learning and capability formation. After comparing the organizational patterns of businesses of Taiwan and South Korea,
Hamilton and Biggart (1988)demonstrate differences in the particular
network configurations of firms are important to their economic
success in international competitiveness, and propose two different models of state/business networks for South Korea and Taiwan. They
are South Korea's“strong state model” in which the state intervenes
heavily in the economy and Taiwan's“strong society model” in which
the state refrains from such action. In fact, social networks common to
Chinese society thatflourished in the absence of state intervention
explains widespread entrepreneurship in Taiwan's IT sectors. While the successful HSIP cluster in Taiwan has led noted industry and innovation analyst John Mathews to aptly describe this success as a
“Silicon Valley of the East” (Mathews,1997), evidence on the importance
of industrial linkages to agglomeration processes in South Korea's IT
sectors exists as well. For example,Park (1991)argues that polarized
regional economies with a diversity of industries generate innovative
firms and spin-offs, which in turn encourage further growth. Park
(1994)further indicates that vertical disintegration and the clustering of small plants increase the productivity of labor-intensive industries.
Overall, the different experiences on IT industrialization of Taiwan and South Korea are quite valuable for those emerging countries interested in creating a national policy and innovation strategy for transforming themselves from producers of low-value goods to producers of high-technology products. However, a study dedicated to performance differences among Taiwanese and South Korean IT firms has not yet been proffered. The majority of previous studies on IT competitiveness has been carried out from an industrial cluster perspective with little attention on the question of the principal
source of profits (industry or organizational behavior) among firms. In
fact, while clusters reveal that the immediate business environment outside companies is important, what happens inside companies plays a vital role as well.
3. The model and its operationalization 3.1. The model
Recognizing that ITfirms in Taiwan are not large corporations, this
study uses the term“firm” to designate an autonomous competitive
unit within an industry in contrast to the term “company” or
“corporation”—a legal entity considered in previous research which
owns and operates one or more business-segments (Rumelt, 1991;
McGahan and Porter, 1997, 2002; Mauri and Michaels, 1998; Brush et al., 1999; Ruefli and Wiggins, 2003; Hough, 2006). Thus the term “firm effects” indicate both intra-industry dispersion among
theore-ticalfirms and difference among firms that are not explained by their
patterns of industry activities. In this regard, this study uses the term “firm effects” to include “corporate effects” and “business-unit
effects.”
The analysis of the effects of industry,firm, and year factors relies
on the following descriptive model, which is similar to the models of
Schmalensee (1985), andRumelt (1991):
rijt=μ:::+αi+βj+γt+eijt ð1Þ
where i = 1,…; m refers industries, j=1,…; nirefersfirms (niis the
number offirms within industry i); t years and rijt the accounting
profit in year t for firm j in industry i. The first right-hand-side term is
µ…, which is the average profit over the entire period for all firms (the three dots indicate averages over indices i, j, and t). The next three
terms αi,βj, andγt represent the random industry,firm, and year
effects, respectively, and thefinal term εijtis the random error term.
The termαiis the increment to profit associated with participation in
industry i and reflects the influence of industrial structural
character-istics onfirm performance. Industry effects derive from differences in
the average returns to individualfirm within each different industry.
The termβjis the increment to profit associated with the specific
situation offirm j, including both corporate and business unit effects,
and represents the influence of firm-specific factors such as
hetero-geneity in tangible and intangible assets and differences in reputation, operational effectiveness, organizational processes and managerial
skills amongfirms (Hawawini et al., 2003). Firm effects derive from
differences in the average annual returns to each differentfirm. The
termγtrepresents the difference betweenμ… and the average profit
allfirms in year t and captures the impact of broad economic trends.
Year effects derive from differences in the average returns to
individualfirm in each observation year.
The assumption of this descriptive model is that random
disturbances εijt occur independently from a distribution with a
mean of zero and unknown variance σε2
. This descriptive model
makes the additional assumption that all of the other main effects (αi,
βj, andγt), like the error term, are realizations of random processes
with a mean of zero and constant variances,σα2,σβ2, andσγ2. These
three sources of variation in firm profitability represent random
industry,firm effects, yearly macroeconomic fluctuations, and random
disturbances. The advantage of this random-effects assumption is, as
Rumelt's (1991, 172) states, “that the differences among effects,
whatever their source, are‘natural’, not having been controlled or
contrived by the research design, and are independent of other
effects”. In other words, random effects occur when observations are
drawn from an underlying and unobservable probability distribution. In summary, this variance decomposition model, or descriptive model,
makes its estimations using dummy variables for industries,firms and
year effects. The strength of the descriptive model using three relevant effects is that this model requires no causal or structural explanation
for profitability differences across years, industries, or firms, and
simply posits the existence of differences in profitability associated
with these categories. 3.2. The operationalization
Previous studies on the relative importance of industry, corporate and business-segment effects rely on accounting returns such as ROA to
measure profitability that reflects historical profits relative to book value
of assets. ROA is calculated by dividing the operating income (i.e., earnings before interest and taxes) by total assets. Although many analysts have debated that ROA does not necessarily capture the net
present value of all returns on investment (e.g.,Fisher and McGowan,
1983), this work as previous variance decomposition studies does not
have a priori hypothesis about the nature and direction of bias caused by calculations of systemic risk and accounting conventions, thus proceeds with using ROA as the performance measure.
4. Data sample and statistical methodology 4.1. Data
Schmalensee (1985)and Rumelt (1991) use U.S. Federal Trade Commission's (FTC) Line of Business (LOB) data to measure industry,
corporate and business-segment effects on profitability, while further
empirical studies along the lines ofRumelt's (1991)work have used
the Compustat® database. Unfortunately, previous variance
decom-position research have mostly focused on the U.S.firms and neglected
differ radically in non-U.S. settings (Khanna and Rivkin, 2001; McGahan and Porter, 2002; Chen and Lin, 2006a,b). With regard to
studies of subpopulations,McGahan and Porter (1999), for example,
find that the influences of industry and corporate effects are substantially different for high and low performers, and further suggest that detailed studies at the sectoral level are needed (McGahan and Porter, 2002). The study assesses the relative importance of benign industrialization policy (industry factors) and
unique organizational processes (firm factors) on profitability
differences the IT firms of Taiwan and South Korea by employing
S&P Compustat® Global Vantage database, which covers market
indicators and financial reports for 25,000 companies listed and
traded on their stock exchanges of 80 countries in the world. For each
country, although the number of firms collected in the S&P
Compustat® Global Vantage database is smaller than the firms
actually issued on their stock exchange, it would be useful to employ this database to do research for cross-country comparisons.
4.2. Sample
The sample sets of IT sectors in Taiwan and South Korea cover the five-year period from 1998 to 2002 to represent the growth stage of IT sectors. The sample is screened in various ways. This study excludes firms that do not contain a primary standard industry classification
(four-digit SIC code),firms that report results with missing values, and
firms that are inactive in the same industry classification over the
five-year period. The final sample of Taiwan's IT sectors contains 195
observations for 42firms across seven industry classifications, such as
21 firms for computer and office equipment, three for electrical
components and equipment, one for telecommunication equipment, 13 for semiconductors and electronic computer accessories, one for telephone communications, two for computer programming and data
processing services, and one for high-tech business services. Thefinal
sample of South Korea's IT sectors contains 145 observations for 37 firms across seven industry classifications, such as four firms for
computer and office equipment, seven for electrical components and
equipment, six for telecommunication equipment, seven for
semi-conductors and electronic computer accessories, five for telephone
communications, six for computer programming and data processing services, and two for high-tech business services. These IT industry
classifications of Taiwan and South Korea and the number of sample
firms are also shown inTable 1.
The sample sets affirm technological competencies and
specializa-tion patterns differ between the IT industries of Taiwan and South Korea. While most of Taiwan's IT manufacturers focus on
semicon-ductor, office and computer machinery production, South Korea's IT
sectors mostly produce telecommunications equipment. 4.3. Statistical methodology
In general, most of variance decomposition studies have used nested
ANOVA and VCA methods to decompose the variances offirm
profit-ability. Nested ANOVA estimates an ordinary linear regression model using dummy variables. The importance of an effect is measured by the
percentage variance arrived at by comparing the influences of dummy
(independent) variables on profitability (dependent variable). Thus, the
order of entry of the independent variables can have a large impact on which variable explains the most variance. To avoid the methodological problem of sequential ordering of entry of variables in nested ANOVA, the VCA approach, sometimes termed random-effects ANOVA, is a popular method of estimating the relative effect of each factor on
performance. For example,Schmalensee (1985)is thefirst to use VCA to
decompose the variance of business profitability into components of
variance. This method treats each effect as though each effect is generated by an independent, random sample drawn from an
under-lying population of the class of effects (Neter et al., 1996). Using the
descriptive model of Eq. (1), the equation for estimating variance
components is developed and decomposing the total variance of performance into its components as follows:
σ2 r =σ 2 α +σ 2 β+σ 2 γ+σ 2 e ð2Þ
The dependent variable rijtin the descriptive model has constant
variance and is normally distributed because the model is a linear combination of independent normal random variables. To estimate the different variance components by using VCA approach, this work uses the VARCOMP procedure in SAS® packages to control for biases that arise from the order of entry of independent effects by rotating entries and adjusting estimates of the variables. The VCA estimation is particularly suited to this study because the VCA does not require a data set covering the whole population while at the same time allowing the results to be generalized. This method is useful since constructing a
data set that covers allfirms in each industry is impossible.
4.3.1. Applying multilevel analysis
Despite VCA method is theflagship approach for this type of
variance decomposition studies, researchers have identified various
issues such as the potentially unreliable and negative estimates (Brush
and Bromiley, 1997), and the assumption of independence between effects, an assumption thought not to be met by the inherently nested
data structures (Brush and Bromiley, 1997; Bowman and Helfat,
2001). Therefore, recent researchers propose new statistical methods,
for example, the HLM method byHough (2006)andMisangyi et al.
(2006), to explore the relative importance of various effects on
profitability. HLM represents an approach to analyzing data that
attempts to address the non-independence between effects, and uses iterative estimation to simultaneously estimate all variance compo-nents and produce feasible (i.e., non-negative) estimates. In other words, HLM offers a means for moving beyond simple models of variance decomposition toward complex models that examine explanatory variables at multiple levels of a data hierarchy.
Given the above methodological context, employing the HLM
technique presented by Raudenbush and Bryk (2002)empirically
examines the relative importance of various effects. The HLM analysis allows for regression-like modeling of relationships at the lowest level of analysis, while regression-like models describe how the within-unit
relationships vary (Hough, 2006; Misangyi et al., 2006). This model
partitions the variation infirm performance as the model is allocated
across time,firms, and industries. Thus, this study posits the following
two-level unconditional model as Level 1:
rijt=βij:+βij1ðYearÞijt+ eijt ð3Þ
Level 2:
βij:=αi::+δij ð4Þ
Table 1
Taiwan's and South Korea's IT industry classifications and the number of sample firms. IT SIC Class description No. offirms
in Taiwan
No. offirms in South Korea 3570 Computer and office equipments 21 4 3600 Electrical components and equipments 3 7 3660 Telecommunications equipments 1 6 3670 Semiconductors and electronic computer
accessories
13 7 4810 Telephone communications 1 5 7370 Computer programming and data processing
services
2 6
where the meaning of notation i, j, t, and rijtis the same as in Eq.(1).
At thefirst level of analysis, the first right-hand-side term βij.is the
mean ROA offirm j in industry i, and the next term βij1represents year
effects (i.e., the impact of macroeconomic fluctuations on firm
activity). At the second level of analysis, αi..is the grand mean of
firm ROA. These two levels model their own random within-firm and
within-industry residuals, eijtandδij, respectively. The HLM analysis
assumes that eijtandδijare distributed normally, with a mean of zero
and variances of σβ2
and σα2
, respectively. Thus, this modeling
partitions the total variance in profitability into two components:
within-firm, σβ2, and within-industry,σα2, and a chi-squared test is
used to determine whether significant variance across firms and
across industries exists. The amount of total variance attributable to
each level is calculated as follows:σβ2/ (σβ2+σα2) is the proportion
of variance within eachfirm; and σα2/ (σβ2+σα2) is the proportion of
variance betweenfirms. Year effects may be estimated through their
incorporation at the time level of analysis as shown in Eq.(3).
Two key differences between HLM and VCA in conducting an assessment of the relative importance of effects exist. First, while VCA assumes the random processes that generate the effects are independent from each other, HLM allows modeling of the idiosyn-cratic relationships between levels of effects and even permits complex error structures at each level of the hierarchy to address dependence between levels of analysis. Thus, the use of HLM to examine such hierarchically ordered system avoids statistical
con-cerns such as lack of independence (Hofmann, Griffin and Gavin,
2000; Hough, 2006). Second, the fully unconditional model in HLM is
equivalent to VCA with one importance caveat—the method of
estimation (Hough, 2006). Previous scholars such asRumelt (1991)
and McGahan and Porter (1997) use sum of squares estimation techniques in their VCA, which can potentially produce negative variances. In contrast, HLM uses iterative estimation procedures such as maximum likelihood that do not produce negative variances (Hough, 2006).
5. Empirical results
Using the S&P Compustat® Global Vantage database, the
magni-tude offirm and industry effects that are sensitive to profitability is
tested.Table 2shows the variance component estimates of the VCA
(left panel) and those of the HLM (right panel) for the independent variables that sum up to the variance in the dependent variable for IT firms from Taiwan and South Korea. Note that these two estimation
techniques, VCA and HLM, do not affect the ordering of entry of the
two factors.Table 2also reports the percentages of the total variance
of the dependent variable explained by the independent effects.
The results reveal thatfirm effects dominate long-term
perfor-mance of Taiwan's IT sectors irrespective of whether variance is decomposed by VCA or HLM. Firm effects estimated by VCA and HLM are 55.8% and 94.7%, respectively. In comparison, the corresponding figures for industry effects are 0% (i.e., originally negative estimate)
and 5.3%, respectively. For South Korea's IT sectors, although firm
effects are greater than industry effects when variance is estimated by
VCA, industry effects dominatefirm effects when estimated by HLM.
Firm effects and industry effects estimated by VCA are 40.7% and 6.9%, and those effects estimated by HLM are 34.2% and 65.8%, respectively.
The results provide strong support for the idea that thefirm-related
resources and capabilities of ITfirms in Taiwan and South Korea have
important influences on profitability. However, industry factors, such
as industry membership and structural characteristics of IT sectors,
have different impacts on profitability between Taiwan and South
Korea. For Taiwan's IT sectors, industry effects have little or no
influences on profitability while the variance is estimated by the HLM
or VCA. On the other side, for South Korea's IT sectors, industry effects
dominate firm effects when variance is estimated by HLM, but
industry effects only have small influences on profitability when
variance is estimated by VCA. Finally, year effects have some impact on
Table 2
Absolute values and percentages of the variance contributed by predictor variables for Taiwan's and South Korea's IT industries.
Taiwan's IT industries Variance
component
Variance estimate Percentage Variance estimate Percentage
VCA HLM Year effect 18.7 13.2 4.8 3.7 Industry effect 0 0 6.9 5.3 Firm effect 79 55.8 122.7 94.7 Model 97.7 69 Error 43.9 31 South Korea's IT industries
Variance component
Variance estimate Percentage Variance estimate Percentage
VCA HLM Year effect 2.2 0.6 2.3 0.4 Industry effect 26.8 6.9 351.6 65.8 Firm effect 157.9 40.7 183.1 34.2 Model 186.9 48.2 Error 201.3 51.8 Table 3
Comparison of percentages of total variance of the component effects. Variance component Schmalensee (1985) Rumelt (1991)a Roquebert et al. (1996)b McGahan and Porter (1997) McGahan and Porter (2002)
Data source FTC FTC Compustat® Compustat® Compustat® Method VCA VCA VCA VCA Nested
ANOVA Year effect N/A N/A 0.4 2.4 0.4 Industry 19.6 4.0 10.2 18.7 10.3 Effect Firm effectc 0.6 45.8 55.0 36.0 47.6 Error 80.4 44.8 32.0 48.4 41.7 Variance component Hawawini et al. (2003)d Hough (2006)e Short et al. (2007)f
Data source Compustat® Compustat® Compustat® Method VCA VCA HLM VCA HLM Year effect 1.0 N/A N/A N/A N/A Industry
effect
8.1 5.8 5.3 19.2 19.3 Firm effectc
35.8 44.7 60.3 65.8 65.8 Error 52.0 46.6 34.5 N/A N/A Variance component Chen and Lin (2006a)g
Current study
Data source TEJ Compustat® Global Vantage Sectoral coverage All Taiwan Taiwan's IT
industries
South Korea's IT industries Method VCA VCA HLM VCA HLM Year effect 2.5 13.2 3.7 0.6 0.4 Industry effect 0.6 0 5.3 6.9 65.8 Firm effectc
41.6 55.8 94.7 40.7 34.2 Error 55.3 31 N/A 51.8 N/A
a
Only the results of sample B inRumelt (1991)are reproduced here. b
Only averages across samples inRoquebert et al. (1996)are reproduced here. c
Firm effects include both corporate and business-level effects. d
Only the results of performance measure ROA inHawawini et al. (2003)are reproduced here.
eOnly the results of statistical methods VCA using SAS® and HLM inHough (2006) are reproduced here.
f
Only the results of performance measure ROA and statistical methods VCA and HLM inShort et al. (2007)are reproduced here.
g
TEJ is the Taiwan Economic Journal data set which coversfinancial reports more than 600 companies listed and traded on the Taiwan Stock Exchange. Only the results of performance measure ROA inChen and Lin (2006a)are reproduced here.
Taiwan's IT sectors, and account for 3.7% and 13.2% when estimated by HLM and VCA, respectively. For South Korea's IT sectors, no matter the variance is estimated by either HLM or VCA, year effects have very
small influences on performance, and account for 0.4% and 0.6%,
respectively. By definition, year effects are macroeconomic
fluctua-tions that affect allfirms to the same degree in a particular year.
Table 3contains the comparablefigures fromSchmalensee (1985),
Rumelt (1991),Roquebert et al. (1996),McGahan and Porter (1997, 2002),Hawawini et al. (2003),Hough (2006), andShort et al. (2007)
on the various effects. This study diverges from previous variance decomposition studies in that the investigation focuses on the sources of performance differences on IT industry rather than on all or
manufacturing industries.Table 3shows thatfirm effects dominate
long-term profitability in these studies, which range from 35.8% in
Hawawini et al. (2003) to 94.7% in current Taiwan's IT industries estimated by HLM. One exception for this tendency is that industry
factors explain more variations offirm performance than firm factors
do in current South Korea's IT industries estimated by HLM. 6. Discussion
In the economics and strategyfields, researchers seek to provide
understanding of the antecedents offirm profitability to address the
question of how and why certain private enterprisefirms develop a
competitive advantage in an environment of rapid technological
change (Teece, Pisano and Shuen, 1997). Schumpeter (1947) and
March and Simon (1958)argue that inter-organizational variation in technological knowledge can occur even within the same industry as competitive dynamics encourage managers to pursue lines of research distinct from those of competitors and explore new combinations of knowledge. Although corporate management by itself cannot build core competencies, which generally reside within an organization and
consist in part of organizational routines (Nelson and Winter, 1982),
Nelson (1991)argues that to some extent profitability differentials are the result of different strategies that are used to guide
decision-making at various levels infirms, and top management may have an
important role in targeting and helping to develop and sustain
organizational capabilities (Castanias and Helfat, 1991, 1992). Thus, as
Bowman and Helfat (2001) suggest, corporate management and corporate strategy in theory have some impact on, but do not have
complete control of corporate-level factors that influence profitability.
From the perspective of corporate management, two potential
strategies affecting technological firms' performance exist: one is
adopting specialized corporate strategy and the other one is adopting
diversified corporate strategy. Concerning tensions between
econo-mies of scale and specialization econoecono-mies (e.g., Rosenberg, 1963),
Arora and Gambardella (1997) argue that more efficient and
specialized firms in larger markets will gain particularly great
advantages owing to the increased sales opportunities for their goods and services in their base markets. Much of the management
literature on diversification follows the resource-based view of the
firm (Barney, 1986; Peteraf, 1993; Rumelt, 1982; Wernerfelt, 1984) in
explaining the dominant rationale for the multiproduct firm as
economies of scope (Panzar and Willig, 1981; Teece, 1982) in assets
subject to transaction costs (Williamson, 1975). Following the
resource-based and transaction cost logic (e.g.,Teece, 1980),firms
should be expected to gain from diversification when the assets
involved in multiple businesses are primarily knowledge-based.
Considering how a firm's knowledge base interacts with its
product market activity, Miller (2006) creates a new measure,
technological diversity, to indicate afirm's opportunity for corporate
diversification based on economies of scope in valuable knowledge
assets for both single- and multi-businessfirms. Similar to
technolo-gical diversity proposed byMiller (2006), this study estimates the
average difference in technological diversity between specialized and
diversified firms by using data from the U.S. Patent Office to determine
technological diversification at the level of firm resources for
knowl-edge-based relatedness.
The technological diversity the ITfirms of Taiwan and South Korea
is measured by different classes of patents that afirm had during the
year 2000–2006. Each patent is identified by International Patent
Class (IPC) and given a count of“patent equivalents” using the first
four-digit patent class codes. Thefirm's core industry is defined by
finding its segment with the most patents in the same category. Any patent equivalent within the same four-digit patent class code as the core industry is assigned a zero, the counts sharing the same three-digit patent class code are assigned a one, the same two-three-digit code a two, the same one-digit code a three, and in different one-digit codes the patent equivalents are assigned a four. To summarize, the index, Di,j,
is a measure of the dispersion of patents i and j within thefirm. Di,jis
lower if thefirm is focused on closely related patents, and higher if the
firm adopts a technological diversified strategy and has many different classes of patents.
The result inTable 4separates the sample of 5155 patents among
158 Taiwan's ITfirms into mean and quartiles. In the dataset, the
minimum Di,jis 0, the maximum Di,jis 3.5, and the mean Di,jis 1.2. The
first quartile Di,jis 0, the second quartile Di,jis 1, and the third quartile
Di,jis 2.1. Since many of Taiwan's ITfirms in the dataset are engaged in
developing their own patents in one single category, that is, the
specialized corporate strategy, the value of Di,jis 0 for a large portion,
44.3%, of observations. From the perspective of technological diversity
by comparing patents dispersion of eachfirm less than and greater
than mean Di,j(1.2), the number of Taiwan's ITfirms with specialized
and diversified-oriented corporate strategies is 82 (51.9%) to 76
(48.1%). Taking an extreme example of specialized corporate strategy
(Di,j= 0), even 70 (44.3%) ITfirms engage in developing their own
patents in one single category. These almost-equal portions of
observations imply that both the specialized and diversified corporate
strategies are matter to the development of Taiwan's ITfirms, which
can partly explain the profitability differentials between these firms.
The result inTable 5separates the sample of 1368 patents among
40 South Korea's ITfirms into mean and quartiles. In the dataset, the
minimum Di,jis 0, the maximum Di,jis 3.6, and the mean Di,jis 1.4. The
Table 4
Technological diversity of Taiwan's ITfirms.
n = 158 Mean = 1.2 Minimum = 0 Maximum = 3.5 Mode = 0 Mean = 1.2 Di,j≤1.2 Di,j≥1.2
Number offirms 82 76 (Percentage) (51.9%) (48.1%) n = 158 1st quartile = 0 2nd quartile = 1 3rd quartile = 2.1
Quartiles Di,j= 0 0bDi,j≤1 1bDi,j≤2.1 2.1bDi,j Number of
firms
70 10 38 40
(Percentage) (44.3%) (6.3%) (24.1%) (25.3%)
Table 5
Technological diversity of South Korea's ITfirms.
n = 40 Mean = 1.4 Minimum = 0 Maximum = 3.6 Mode = 0 Mean = 1.4 Di,j≤1.4 Di,j≥1.4
Number offirms 19 21 (Percentage) (47.5%) (52.5%) n = 40 1st quartile = 0 2nd quartile = 1.6 3rd quartile = 2.4
Quartiles Di,j= 0 0bDi,j≤1.6 1.6bDi,j≤2.4 2.4bDi,j Number of
firms
13 7 10 10
first quartile Di,j is 0, the second quartile Di,j is 1.6, and the third
quartile Di,jis 2.4. From the perspective of technological diversity by
comparing patents dispersion of eachfirm less than and greater than
mean Di,j(1.4), the number of South Korea's ITfirms with specialized
and diversified-oriented corporate strategies is 19 (47.5%) to 21
(52.5%). Taking an extreme example of specialized corporate strategy
(Di,j= 0), a statistic of 13 (32.5%) ITfirms engaged in developing their
own patents in one single category. Like Taiwan's ITfirms, both of the
specialized and diversified corporate strategies are matter to the
development of South Korea's ITfirms. However, from the perspective
of conducting patents innovation, South Korea's ITfirms are more
technologically diversified than Taiwan's IT firms.
7. Conclusion
This study revisits questions about the relative importance of
industry- and firm-level effects on profitability differentials among
firms and extends recent research in three ways. First, this work compares IT sectors in Taiwan and South Korea by using the Standard & Poor's Compustat® Global Vantage database, because they have become very competitive and demonstrated outstanding performance
since the 1990s. Second, this investigation tests industry and firm
effects using both the multilevel approach of HLM and the
conven-tional VCA. Third, the question of why there are significant
profit-ability differences among technological firms even with similar
industrial structural characteristics and leveraged resources and capabilities in the same IT industry will be explored. This study uses
patents information from the United States Patent Office to estimate
technological diversification at the level of firm resources for
knowl-edge-based relatedness for the ITfirms of Taiwan and South Korea.
When seeking to explain the sources of performance differentials
among IT firms, firm effects dominate industry effects in both
economies except when the variance is estimated by HLM for South
Korea's IT sectors, which firm effects explain less variation on
profitability. One possible explanation for this finding is that although
the governments of NIEs, including Taiwan and South Korea, take a
fairly aggressive role in directing IT industry development (Mathews,
1997; Hernandez, 2005), industry effects matter little to Taiwan's IT firms' profitability, which implies, as Michael Porter argues, very powerful micro, diamond-type factors, rather than government intervention, that ultimately play the dominant role in driving their
success exist (Snowdon and Stonehouse, 2006). For South Korea's IT
sectors, although the results indicate that both industry- and
firm-level factors are clearly important in shaping strategy and
perfor-mance, industry effects have greater influences than firm effects.
These findings provide strong support of the industrial policy
perspective of South Korea's IT success, where the South Korean government since the late 1980s have employed economic strategies that are regarded as having played a key role in forging its IT sectors' economic miracle. The analysis yields more evidence to support the
idea that Taiwan's ITfirms gain competitive advantages by deploying
their organizational capabilities, or sound microeconomic
fundamen-tals rather than traditional top–down approach, and that South
Korea's IT firms deploy their sound microeconomic fundamentals
along with very aggressive government intervention in directing economic activities, across industry boundaries.
In fact, the results indicate that firm effects are much more
important to profitability especially for Taiwan's IT sectors, even if an
investment and innovation-friendly environment is created that is conducive to sustainable growth. These facts raise an interesting
question whetherfirm effects persist longer than industry effects. Are
the competitive advantages offirms within their industries sustained
longer than industry influences? To investigate this issue, future
research needs to decompose the persistence of incremental
industry-andfirm- specific effects on profitability among IT firms.
References
Arora A, Gambardella A. Domestic markets and international competitiveness: generic and product-specific competencies in the emerging sector. Strateg Manage J 1997;18(6):53–74.
Barney JB. Strategic factor markets: expectations, luck, and business strategy. Manage Sci 1986;32(10):1231–41.
Bowman EH, Helfat CE. Does corporate strategy matter? Strateg Manage J 2001;22(1):1-23. Brush CG, Greene PG, Hart MM. From initial idea to unique advantage: the entrepreneurial
challenge of constructing a resource base. Acad Manage Exec 2001;15(1):64–78. Brush TH, Bromiley P. What does a small corporate effect mean? A variance components
simulation of corporate and business effects. Strateg Manage J 1997;18(10):825–35. Brush TH, Bromiley P, Hendrickx M. The relative influence of industry and corporation on business segment performance: an alternative estimate. Strateg Manage J 1999;20(6):519–47.
Castanias RP, Helfat CE. Managerial resources and rents. J Manage 1991;17:155–71. Castanias RP, Helfat CE. Managerial and windfall rents in the market for corporate
control. J Econ Behav Organ 1992;18:153–84.
Chang CY, Yu PL. Made by Taiwan: booming in information technology era. Singapore: World Scientific Publishing; 2001.
Chen YM, Lin FJ. Sources of superior performance: industry versusfirm effects among firms in Taiwan. Eur Plan Stud 2006a;14(6):733–51.
Chen YM, Lin FJ. Regional development and sources of superior performance across textile and IT sectors in Taiwan. Entrep Reg Dev 2006b;18(3):227–48.
Choung JY, Hwang HR, Yang H. The co-evolution of technology and institution in the Korean information and communications industry. Int J Technol Manag 2006;36 (1–3):249–66.
Cubbin J, Geroski P. The convergence of profits in the long run: firm and inter-industry comparisons. J Ind Econ 1987;35(4):427–42.
Cusumano M, Elenkov D. Linking international technology transfer with strategy and management: a literature commentary. Res Policy 1994;23:195–215.
Fisher F, McGowan J. On the misuse of accounting rates of return to infer monopoly profits. Am Econ Rev 1983;73(1):82–97.
Hamilton G, Biggart N. Market, culture, and authority: a comparative analysis of management and organization in the Far East. Am J Sociol 1988;94:S52–94. Hansen GS, Wernerfelt B. Determinants offirm performance: the relative importance of
economic and organizational factors. Strateg Manage J 1989;10(5):399–411. Hawawini G, Subramanian V, Verdin P. Is performance driven by industry- orfirm-specific
factors? A new look at the evidence. Strateg Manage J 2003;24(1):1-16.
Hernandez Z. Industrial policy in East Asia: in search of lessons. World Development Report; 2005. p. 1-33. Background Paper.
Hofmann DA, Griffin MA, Gavin MB. The application of hierarchical linear modeling to organizational research. In: Klein KJ, Kozlowski SWJ, editors. Multilevel theory, research, and methods in organizations: foundations, extensions, and new directions. San Francisco, CA: Jossey-Bass; 2000. p. 467–511.
Hough JR. Business segment performance redux: a multilevel approach. Strateg Manage J 2006;27(1):45–61.
Kessides IN. Internal versus external market conditions andfirm profitability: an exploratory model. Econ J 1990;100:773–92.
Khanna T, Rivkin J. Estimating the performance effects of business groups in emerging markets. Strateg Manage J 2001;22(1):45–74.
Kogut B, Zander U. Knowledge of thefirm, combinative capabilities, and the replication of technology. Organ Sci 1992;3(3):383–97.
Kraemer K, Dedrick J, Hwang CY, Tu TC, Yap CS. Entrepreneurship,flexibility, and policy coordination: Taiwan's computer industry. Inf Soc 1996;12:215–49.
March JG, Simon H. Organizations. New York: Wiley; 1958.
Mason ES. Price and production policies of large scale enterprises. Am Econ Rev 1939;29 (1):61–74.
Mathews JA. A Silicon Valley of the East: creating Taiwan's semiconductor industry. Calif Manage Rev 1997;39(4):26–54.
Mauri AJ, Michaels MP. Firm and industry effects within strategic management: an empirical examination. Strateg Manage J 1998;19(3):211–9.
McGahan AM, Porter ME. How much does industry matter, really? Strateg Manage J 1997;18(1):15–30.
McGahan AM, Porter ME. The persistence of shocks to profitability. Rev Econ Stat 1999;81(1):143–53.
McGahan AM, Porter ME. What do we know about variance in accounting profitability? Manage Sci 2002;48(7):834–51.
Miller DJ. Technological diversity, related diversification, and firm performance. Strateg Manage J 2006;27(7):601–19.
Misangyi VF, Elms H, Greckhamer T, Lepine JA. A new perspective on a fundamental debate: a multilevel approach to industry, corporate, and business unit effects. Strateg Manage J 2006;27(6):571–90.
National Science Board (NSB). Science and engineering indicators 2006, NSB-06-2. Arlington, VA: National Science Foundation; 2006.
Nelson RR. Why dofirms differ and how does it matter? Strateg Manage J 1991;12 (8):61–74.
Nelson RR, Winter SG. An evolutionary theory of economic change. Cambridge, MA: Belknap Press of Harvard University Press; 1982.
Neter J, Kutner MH, Nachtsheim CJ, Wasserman W. Applied linear statistical models. 4th ed. Chicago: Irwin; 1996.
Nourse EG, Drury HB. Industrial price policies and economic progress. Washington, DC: The Brookings Institution; 1938.
Panzar J, Willig R. Economies of scope. Am Econ Rev 1981;71(2):268–72.
Park SO. High-technology industries in Korea: spatial linkages and policy implications. Geoforum 1991;22:421–31.
Park SO. Industrial restructuring in the Seoul metropolitan region: major triggers and consequences. Environ Plan A 1994;26:527–41.
Peteraf MA. The cornerstones of competitive advantage: a resource-based view. Strateg Manage J 1993;14(3):179–91.
Powell TC. How much does industry matter? An alternative empirical test. Strateg Manage J 1996;17(4):323–34.
Raudenbush SW, Bryk AS. Hierarchical linear models: applications and data analysis methods. 2nd edition. Thousand Oaks, CA: Sage; 2002.
Roquebert JA, Phillips RL, Westfall PA. Markets vs. management: what “drives” profitability? Strateg Manage J 1996;17(8):653–64.
Rosenberg N. Technological change in the machine tool industry, 1840–1910. J Econ Hist 1963;23(4):414–43.
Ruefli TW, Wiggins RR. Industry, corporate and business-segment effects and business performance: a non-parametric approach. Strateg Manage J 2003;24(9):861–79. Rumelt RP. Diversification strategy and profitability. Strateg Manage J 1982;3
(4):359–69.
Rumelt RP. How much does industry matter? Strateg Manage J 1991;12(3):167–85. Saxenian AL, Hsu JY. The Silicon Valley—Hsinchu connection: technical communities
and industrial upgrading. Ind Corp Change 2001;10(4):893–920. Schmalensee R. Do markets differ much? Am Econ Rev 1985;75(3):341–51. Schumpeter JA. The creative response in economic history. J Econ Hist 1947;7
(2):149–59.
Short JC, Ketchen DJ, Palmer TB, Hult TM. Firm, strategic group, and industry influences on performance. Strateg Manage J 2007;28(2):147–67.
Snowdon B, Stonehouse G. Competitiveness in a globalised world: Michael Porter on the microeconomic foundations of the competitiveness of nations, regions, and firms. J Int Bus Stud 2006;37(2):163–75.
Teece DJ. Economies of scope and the scope of the enterprise. J Econ Behav Organ 1980;1:223–45.
Teece DJ. Towards an economic theory of the multiproductfirm. J Econ Behav Organ 1982;3:39–63.
Teece DJ, Pisano G, Shuen A. Dynamic capabilities and strategic management. Strateg Manage J 1997;18(7):509–33.
Thorbecke E, Tung AC, Wan H. Industrial targeting: lessons from past errors and successes of Hong Kong and Taiwan. World Econ 2002;25(8):1047–61. Wernerfelt B. A resource-based view of thefirm. Strateg Manage J 1984;5(2):171–80. Wernerfelt B, Montgomery CA. Tobin's q and the importance of focus in firm
performance. Am Econ Rev 1988;78(1):246–50.
Williamson OE. Markets and hierarchies: analysis and antitrust implications. New York: Free Press; 1975.
