Success in High-Technology Markets: Is Marketing Capability Critical?

http://books.google.com.pk/books?
id=IhjyJwmJR80C&pg=PA36&dq=changing+perspectives+in+marketing+planning&hl=en&sa=X&ei=74I7T5XfCMrwrQeFlvSGAQ&ved=0CEIQ6AEwAw#v=onepage&q=c hanging%20perspectives%20in%20marketing%20planning&f=false
Success in High-Technology Markets: Is Marketing Capability Critical?

Shantanu Dutta Om Narasimhan Surendra Rajiv
Marshall School of Business, University of Southern California, Los Angeles, California 90089 sdutta@sba2.usc.edu narasim@usc.edu Graduate School of Business, University of Chicago, Chicago, Illinois 60637, fsrajiv@gsbsxr.uchicago.edu
Abstract
We propose a conceptual frameworkwith the resourcebased view (RBV) of the firm as its theoretical underpinningto explain interfirm differences in firms profitability in high-technology markets in terms of differences in their functional capabilities. Specifically,we suggest that marketing,R&D,and operations capabilities,along with interactions among these capabilities,are important determinants of relative financial performance within the industry. This paper contributes to the RBV literature by proposing the input-output perspective to conceptualize the notion of capabilities. Specifically,this approach entails modeling a firms functional activitiesviz.,marketing,R&D and operationsas transformation functions that relate the productive factors/resources to its functional objectives,if the firm were to deploy these resources most efficiently. Any underattainment of the functional objective,then,is attributable to functional inefficiency,or equivalently,to a lower functional capability of the firm. The input-output conceptualization of a firms capabilities is then estimated using the stochastic frontier estimation (SFE) methodology. SFE provides the appropriate econometric technique to empirically estimate the efficient frontier and hence the level of efficiency achieved by the various firms. Our study contributes to a number of literatures,both methodologically and substantively. First,it contributes both conceptually and methodologically to the RBV literature. Conceptually,our study suggests that firm capabilities can be viewed in an inputoutput framework. Methodologically, the study suggests the use of stochastic frontier estimation to operationalize and estimate firm capabilities. This methodology is,to the best of our knowledge,the first to allow the researcher/manager to infe capabilities from archival r data. Substantively,our study contributes to the literature on market orientation by suggesting that a stronger market orientation of a firm should be re ected in a higher marketing capability. It also adds to the literature on design for manufacturability by explicating the complementarity among the various functional capabilities and offering empirical evidence on their relative importance in in uencing a firms performance. Finally,our study builds on prior literature that
07322399/99/1804/0547/$05.00 1526-548X electronic ISSN
has highlighted the importance of marketingR&D coordination as important determinants of new product development and success. We highlight below some of our main findings. A strong base of innovative technologies enhances a firms sales by favorably in uencing consumers expectations about the externality benefits associated with its product. This suggests that a past track record of consistent innovation is a credible signal to current and potential customers of the firms continued excellence in a technologically evolving market. Given the importance of in uencing customers,managers need to tailor their marketing activities around the need to inform customers of the technological excellence of their firm. Thus,customers need to be informed of the innovative technologies that the firm possesses and of the future R&D initiatives undertaken by it. Similarly,any potential applications of innovative technology developed by the firm,and of technologies under development,should be emphasized. Marketing capability has its greatest impact on the (quality-adjusted) innovative output for firms that have a strong technological base. In other words,firms with a strong R&D base are the ones with the most to gain from a strong marketing capability. Marketing capability strongly in uences the width of applicability of innovations,i.e.,a firms marketing capability enhances its ability to generate innovative technologies that have applications across a range of industries. This result carries a strong message for managers: A strong market orientation is one of the most fertile sources of ideas for innovation. Thus,marketing needs to be involved from the beginning of the innovation processnamely,right at the stage when technological ideas are being generated. The most important determinant of a firms performance is the interaction of marketing and R&D capabilities. This supports the assertion that firms in high-technology markets need to excel at two things: the ability to come up with innovations constantly,and the ability to commercialize these innovations into the kinds of products that capture consumer needs and preferences. This finding offers further evidence on the importance of coordination between R&D and marketing,as suggested in the extant marketing literature. Marketing Science 1999 INFORMS Vol. 18,No. 4,1999,pp. 547568
SUCCESS IN HIGH-TECHNOLOGY MARKETS: IS MARKETING CAPABILITY CRITICAL?
Finally,using archival data,our methodology can be used to benchmark a firms capabilities,with other firms in the industry,along various functional dimensions. This would be an important step in making more informed resourceallocation decisions. Thus,the firm can spend more money on those capabilities where it most lags the competition,or
on those capabilities that are shown to have the maximum impact on firm performance. (High-Technology Markets ; Resource-Based View ; FirmCapabilitie ; Stochastic ; R&D and Innovation Specific ; Patent s Frontier ; Cross-Functional Coordination s and Patent ; MarketingCitations Manufacturing ; Market Orientation ) Interface
1. Introduction
Firms in high-technology markets are growing at twice the rate of the economy as a whole and have generated returns for their shareholders in significant recent However,despite the rapid growth and years. profit- for these markets,there exists significant ability vari- in the performance of the firms,often within ation the same industry Business Week 1998). 1 Interestingly ( most of the extant literature has attributed , variation inperformance to external market interfirm factors,a major component in uencing the firms where performance is its ability to curtail competitive rivalry 1980,Montgomery and Wernerfelt 1988). (Porter Other have attempted to explain interindustry studies differ- in R&D investment and innovative ences performance by identifying different appropriability and opportunity conditions across industries (Griliches 1984, Boulding and Staelin 1995). Virtually no role is ascribed to firm-specific factors in these studies. few existing detailed case studies of The individual firms in high-technology markets emphasize the role of R&D and manufacturing in enhancing firms performance (e.g.,Iansiti and West 1997). The role of marketing is,however,rarely acknowledged. We argue picture of high-technology industries is that this seri- incomplete. A firm might have a strong ously R&D but be incapable of converting it into ability commercially viable products because of a poor marketing example,Xerox PARC came up with ability. For revolutionary concepts such as the graphical user inter-and the laser printer but was unable to face exploit
1
them because of its poor marketing ability. Similarly,AMD came up with the fastest chip in although the market (the K6),it was unable to pose a serious chal- to Intel because of the latters superior lenge marketing operations and abilities. The above discussion highlights the importance of considering marketing,R&D,and operations capabilities together to understand interfirm differences in performance. To this end we propose a conceptual frameworkwith the resource-based view (RBV) of firm as its theoretical underpinningto the assess firm-specific determinants of firms financial perfor- in high-technology markets. The mance RBV (Wernerfelt 1984) views a firm as a bundle of resources and capabilities,with firms differing in their endow- these resources and capabilities. Because ments of capabilities are difficult to imitate or transfer,possessing bestows enduring competitive superior capabilities
2
advantage upon the firm (Peteraf 1993). Hence,the RBV suggests that intra-industry variations in firms performance (i.e.,firms competitive environment being the same) can be attributed to differences in their capabilities. While the RBV offers an insightful theoretical foundation,prior empirical efforts to use this framework have been plagued with problems,particularly the problem of operationalizing and measuring firms capabilities (Teece et al. 1997). paper contributes conceptually to the RBV This literature by proposing the inputoutput perspective to notion of a firms capabilities. clarify the Specifically, this approach entails modeling a firms functional activitiesviz.,marketing,R&D and operations as transformation functions that relate the productive factors/resources to its functional objectives,if the firm to deploy these resources most efficiently. were Any underattainment of the functional objective,then,is attributable to the functional inefficiency (equivalently, a lower functional capability) of the firm. The input conceptualization of a firms capabilities is output then
Marketing Science/Vol. 18,No. 4,1999
The significant intra-industry variation in firms performance is not limited to firms in high-technology markets alone but has also been empirically observed in other industries. In fact,Rumelt (1991) reports a higher intra-industry than interindustry variation in firms profitability . Boulding and Staelin (1995) do control for firm-specific factors via unobserved heterogeneity,but their emphasis is on the role of industry factors as moderating variables.
548
DUTTA, NARASIMHAN, AND RAJIV Success in High-Technology Markets: Is Marketing Capability Critical?
estimated using the stochastic frontier estimation (SFE) methodology. SFE provides the appropriate econo- technique to empirically estimate the metric efficientand hence the level of efficiency achieved frontier by various firms (equivalently,their functional the capabilities). This technique contributes methodologically to the RBV literature,in that it permits a measurement of capabilities while explicitly linking resources to capabilities . Our study contributes to the literature on market orientation (Deshpande et al. 1993,Jaworski and Kohli 1993,Day 1994) by suggesting a way to infer the market orientation of a firmnamely,by measuring its marketing capability. We suggest that a superior mar-orientation should be re ected in a higher ket marketing capability. It also adds to the literature on design for manufacturability (Hayes et al. 1988,Srinivasan et 1997) by explicating the complementarity al. between the various functional capabilities. Finally,our study on prior literature that has highlighted the builds importance of marketing and R&D and their coordination determinants of new product as important develop- success (e.g.,Gupta et al. 1987,Griffin ment and and Hauser 1996,Gatignon and Xuereb 1997) in two ways. First,we explicate and measure the impact of market-R&D; second,we measure the impact of ing on the interaction of R&D and marketing capabilities on firm financial performance. We highlight substantive findings below some of our
technologies that have applications across a range of industries . The most important determinant of a firms performance is the interaction of marketing and R&D capabilities. This supports the assertion that firms in high-technology markets need to excel at two things: the ability to come up with innovations constantly,and the ability to commercialize these innovations into the of products that capture consumer needs kinds and preferences. This finding offers further evidence on the importance of coordination between R&D and marketing,as suggested in the extant marketing literature. of the paper is organized as follows. The rest The next section discusses the resource-based view of the underlying our empirical analysis and firm develops the inputoutput conceptualization of a firms func- capabilities. Section 3 brie y discusses the tional data, variables operationalization,and empirical model specifications. We also outline the econometric methodology employed in the analysis. Section 4 presents the parameter estimates and discusses the substantive insights from the study. Section 5 highlights limita- and suggests directions for future tions research.
2. Conceptual Framework
The conceptual framework is organized as follows. 2.1 gives an overview of the RBV of Subsection the firm,defining the key concepts of resources and capabilities and linking them to sustained competitive Subsection 2.2 applies the RBV to advantage. hightechnology markets and discusses in detail the capa- that important in sustaining competitive bilities advantage in these markets. Finally,subsection 2.3 discusses the conceptual framework. operationalization of our
: A strong base of innovative technologies enhances a firms sales by favorably in uencing consumers expectations about the externality benefits associated with its product. This suggests that a past track record of consistent innovation is a credible signal to current and potential customers of the firms continued excel- in a technologically evolving lence market. Marketing capability has its greatest impact on the (quality-adjusted) innovative output for firms that a strong technological base. In other have words,firms R&D base are the ones with the with a strong most tofrom a strong marketing gain capability. Marketing capability strongly in uences the of width applicability of innovations,i.e.,a firms marketing enhances its ability to generate capability innovative
2.1. Overview of the Resource-BasedView of the Fir m The literature on the RBV of the firm (Wernerfelt 1984) attempts to identify conditions and firm-specific fac- that underlie the competitive advantage tors enjoyed by a firm. In this perspective,a firm is viewed as a bundle resources and capabilities, with of differing
549
firms
in their endowment of these resources and capabilities 3 While resources are defined as productive . that afactors uses to achieve its business firm objectives,ca- to a firms ability to deploy pabilities,refer these resources . . . to effect a desired end (Amit and 4 Thus,it is argued that for a firm Shoemaker 1993). to enjoy a competitive advantage (i.e.,a superior financial performance relative to competition),it must possess capabilities,i.e.,the ability to deploy superior resources and other productive factors more efficiently. For a firm to enjoy a sustained competitive advantage,it must be the case that these capabilities cannot be competed away, i.e., limits to competition are nec- for a firm to sustain any supernormal profits essary or rents. The RBV identifies two conditions necessary for capability to be an enduring source of a competitive advantage: imperfect mobility and imperfect imitability (Peteraf 1993). Imperfect mobility refers to the diffi- of trading in certain capabilities. This might culty be, instance,because a capability has arisen from for the complex interaction of a number of resources,and hence is firm-specific in nature. For example,it would be hard to buy firm-specific knowledge of buyers,sellers,and workers capabilities (Prescott and Visscher Imperfect imitability,on the other 1980). hand,refers to the inability of competing firms to imitate a firms distinctive capabilities. Numerous mechanisms could that a firms capabilities are imperfectly ensure imitable. Apart from such obvious reasons as patent or property rights,the inherent complexity of most capabilities makes it very hard to ascertain the exact of efficiency,thereby making imitation cause difficult (Lippman and Rumelt 1982). In summary,capabilities a that exhibit high degree of tacitness,complexity,and to be both imperfectly firm-specificity are likely imitable and imperfectly mobile.
2.2. Firms Functional Capabilities and Performance in High-Technology Markets In our conceptual framework,three critical capabilities the performance of firms in highin uence technology markets: R&D,marketing,and operations. In what follows,we show how these capabilities and their inter- bestow either demand-side and/or supplyactions side advantages on the firm. A demand-side advantage lets firm charge a higher price,relative to a competition,atof demand,or to generate a higher a given level de- (or equivalently,a higher market share) at mand given level. On the other hand,a supply-side price advantage refers to the fact that a firm enjoys a lower cost structure. We also argue why these three capabilities are imperfectly mobile and imperfectly imitable and provide a firm with sustainable competitive thus advantage. Figure 1 provides a schematic representation of the relationship between a firms functional capabilities and its financial performance. Marketing Capability. keting capabilityexhibiting A firm with a strong marsuperiority in
identifying customers needs and in understanding the factors uence consumer choice behaviorwill be that in able to achieve better targeting and positioning of its brands relative to competing brands. This higher level of product differentiation will enable the firm to enjoy margins (Kohli and Jaworski 1993,Day higher 1994) hence exhibit better financial and performance. to obtain high-quality customer This ability feedback requires skill at monitoring the environment and building strong relationships with customers,which is a complex undertaking (Deshpande et al. 1993). Such
Figure 1 Performance Firms Capabilities and
The literature on the RBV often uses the term resources ,and resources
endowments to refer to what we call
to refer ca-
to what we call
pabilities . In this paper we suggest that capabilities are imperfectly imitable and hence lead to sustained advantaged,while resources (i.e.,endowments) may be imitated by others. We use the terms resources and capabilities throughout this paper.
4
Note that resources could physicalassets) or intangible (such as goodwill).
be
either
tangible
(such
as
550
a capability,once built,is not easily imitated or trans- because it is often firm-specific and has a ferable high of tacitness (Day level 1994). R&D R&D capability is critical Capability. to achieving a superior performance in hightechnology two important reasons. First,these markets for mar- are characterized by short product life-cycles kets and a high rate of new product introductions incorporating newer generations of technology. In such markets,a a superior innovative capability will firm with enjoy consumer loyalty (Givon et al. 1995),and strong hence a demand-side advantage,because consumers will be willing to pay a premium for a firms product if they are assured that the firm will continue to dominate the market over successive generations of the product. Sec- firms superior R&D capability may also lead ond,a to competitive advantage because of supply-side factors. For instance,Japanese firms such as Sony and Hitachi have leveraged their strong capability in process innovation to dominate high-technology markets because of their favorable cost structure. An important characteristic of R&D in hightechnology markets is a significant learning-bydoing effect,which makes it very difficult for competitors to simply buy this know-how in the market and also it extremely difficult to imitate (Irwin makes and Klenow 1994). The difficulties of imitation are further exacerbated by the large tacit component of R&D. characteristics of R&D capability enable a These firm has a superior competency in R&D to that achieve sustained performance relative to superior its competition . Operations A strong operations Capability. capability in these high-technology markets entails the integration and coordination of a complex set of tasks combining components and materials from differentand industries,and with material ows sources while enabling the firm to offer its final products at a cost (Hayes et al. 1988). The great complexity lower of the operations function helps to make a superior operations capability imperfectly mobile and imperfectly imitable,thereby conferring competitive advantage upon firms that possess it. Complementarity Between the Functional Capa- In addition to each of the direct effects bilities . discussed above,capabilities can serve as important complements to each other. Such interactions can serve to
enhance performance over and above the contribution of the individual of each capabilities. a significant literature that has There is suggested that interaction between marketing and R&D can enhance a firms performance (Gupta et al. 1987,Griffin 1996) beyond their individual effects. and Hauser For instance,a strong marketing capability is better posi- to give R&D good feedback from tioned customers, turn would drive the kinds of which in innovations needed to improve the product. Similarly,a number of studies on product development have pointed to the importance of interaction between manufacturing and throughout the development process to R&D ensure and successful commercialization of speedy technolo- products at a low cost (Hayes et al. gies and 1988). Finally,prior research (e.g.,Srinivasan et al. 1997) has pointed out the high complementarity between mar- and operations capabilities,which can keting help product development by implementing the design for manufacturability concept. To summarize,in high-technology markets a firms capabilities in its marketing,R&D,and operations well as the interactions between functions,as them,are drivers of competitive advantage. critical Thus,for firms within the same industry (so that they face a competitive structure),we would expect similar inter-variations in profitability to be systematically firm related to interfirm variations in functional capabilities. conceptual framework (see Figure The proposed 1) suggests the following relationship:
Re lativ e P er fo rm ance f (R elat iv e Ma r ke ting C apa bil ity , Rel ativ e R &D C a pa bilit y ,R ela tiv e Ope ra tio ns C apa b ility , Rela tiv e M ar k etin g C a pab ility Re l ati v e R &D C apa b i l i ty , Rela tiv e R &D C a pa bilit y R elat iv e Ope ra tio ns C apa b ility , Rela tiv e M ar k etin g C a pab ility Re lativ e O per a tio ns C a pab ilit y) .
5
(1)
It is important to note here that the measure of profitability used in such an analysis should be independent of the scale of operation (and hence the amount of resource endowment). For instance,consider two firms,A and B,that differ in their marketing capabilities with firm A having a higher capability. However,firm B has larger financial resources available to it and hence spends more on mar- In spite of its relative inefficiency,firm B may have higher keting. $sales and $-profits (both measures being scale-dependent) but should have a lower profitability. In our analysis we use Tobins which is scale independent,to measure relative market performance. q,
551
2.3. Measuring Capabilities The
Firms
Functional
InputOutput
Recall Approach that a firms capability is defined as its ability the deploy to resources (inputs) available to it to achieve the desired objective(s) (output). Thus,the higher the functional capability a firm possesses,the more efficiently it is able to deploy its productive inputs to its functional objectives. Any achieve underattainment of the functional objective,then,is attrib- to the functional inefficiency of the firm. utable Evidently,the lower the functional inefficiency,the higher the functional capability of the firm. Thus,we can use inverse of a firms functional inefficiency as the the measure of its functional capability. Specifically,the inputoutput approach entails modeling a firms functional activities as efficient frontier an transformation function or (akin to the notion of production frontier/function in aeconomics; e.g., Silberberg 1990) relating productive factors/resources the by a firm to the optimal attainment of used its tional objective(s) if the firm were to deploy these resources efficiently . To illustrate,suppose that most maximizing the quality-adjusted technological output were the functional objective of the firms R&D activ- Then,the inputoutput transformation ities. function would relate approach maximu qualitythe technological output the firm m could haveadjusted achieved, amount of productive inputs/resources given the deployed if the firm were to use these resources most efficiently. The SFE methodology (see 3.3 for details) the appropriate econometric technique to provides empirically estimate the efficient frontier and hence the of efficiency achieved by the various firms in level its R&D activities. for our purposes,the inputoutput Crucially approach explicitly recognizes the linkages between resources/inputs and objective/outputs and the moderating role of capability. This is because,given identical resource/input endowments,a firm with a functional capability will be able to higher achieve a higher functional objective/output. Similarly,given identical functional capability,a firm with a larger endowment of resources/inputs will be able to achieve higher a functional objective/output. the schematic representation of Figure 2 gives the
552
proposed inputoutput conceptualization of a hightechnology firm. 2.3.1. Measuring Marketing Capability.marketing at the firm level is to the goals of enhance of the firms products in the minds of the value its current and potential customers. This goal is partly reected in enhanced sales,through a better understanding of customer needs and distinctive targeting of appropriate customers. Furthermore,increasing sales is crucial to building market share. We thus use sales revenue as the goal for marketing. A number of resources available to the firm have been mentioned in past literature. Such resources include the extent of customer awareness and liking the firms products built over the years about through its advertising effort (the carry-over effect of advertis- installed base of customers,and the ing),the expen-over the years on marketing activities like diture distribution and trade promotion efforts to build trade Similarly,the literature has suggested the loyalty. importance of investment on customer relationships (Jackson 1985). There are two additional resources that are important in high-technology markets: the firms prior base of technological know-how,and its installed base. The prior level of innovative know-how helps firms marketing in two ways. First,a large base of technological know-how helps a firm in introducing a steady stream of new products. Second,a large stock of innovative know-how makes it easier for marketing to convince its customers that the firm will be able to keep its technological leadership as technology evolves. Finally,a firms installed base would be important to the firm in where switching costs are high and markets network externality effects are strong. above discussion suggests the The ingfollowing frontier/transformation function
Sale s f (Technological Base,
One of
func-
market-
Advertising Stock,Stock of Marketing Expenditure, Investment in Customer Relationships,Installed Base).
(2) goal
2.3.2. Measuring R&D The Capability R&D is to develop high-quality technological of innovationsboth product innovations (which form the basis
Figure 2 Schematic of the InputOutput Framework
of new product introductions) and process innova- Prior tions 6 empirical literature on . distinguished two dimensions of the quality R&D/innovation has of technological output: innovativeness (Trajtenberg 1990) and width of applicability (Jaffe et al. 1993). We use maximization of quality-adjusted thus technolog- as the objective of a firms R&D ical output function.importance of learning-by-doing in The hightechnology markets immediately suggests that a firms past R&D expenditures would be an important resource available to it. In a similar vein,Cohen and Levinthal (1990) suggest that a firms prior stock of innovative technologies can enhance its ability to develop newer generations of innovative technologies. Thus,a firms technological base would be a crucial resource R&D.
6
Another key driver of a firms R&D output is its marketing capability. A strong marketing capability pro- high-quality consumer feedback to R&D vides (Griffin and Hauser 1993). Such input can help R&D come up innovations that are novel in that they with represent a substantial change over past technology. Similarly,if does a good job of scanning the marketing environment,it can suggest ideas that are of applicability towide variety of markets,within and outside a the industry . The above discussion suggests following frontier/transformation function:
Quality-Adjusted Technological Output f(Technological Base, Cumulative R&D Expenditure,Marketing Capability).
the
R& D
for
(3)
It is important,therefore,that in considering a firms innovative output we account explicitly for the quality of such innovations. This point is elaborated on in our discussion of the operationalization of quality-adjusted technological output (see 3.2.1).
2.3.3. Operations The key goal Capability of operations in high-technology markets is to produce at the lowest possible cost without compromising on product quality (Hayes et al. 1988). Thus,we adopt
553
cost minimization as the business objective of a firms operations/manufacturing function. Consistent with economic theory (Silberberg the1990), exogenous variables in the cost function are output and factor prices,i.e.,cost of capital and volume unit labor cost. Similar to the R&D specification,operations can draw on marketing capability to further its goals. Such marketing capability is exemplified in the use of de- for manufacturability concept (Srinivasan et sign al. 1997). To start with,operations can use input from marketing,derived from attribute-based models,to come up with product concepts. Furthermore,it can marketing to give it feedback on various get customerready prototypes. This enhances the likelihood of the product being acceptable to consumers while final be- produced at as low a cost as ing possible. In addition to the resources suggested above,operations can take advantage of the firms stock of innovative technologies. This stock of innovative technologies often provides the basis for process innovations which enhance the efficiency of engineering and/or manufacturing processes (Iansiti and West 1997) and lead to lower cost. The above discussion suggests the following frontier/operations transformation function:
Cost of Production Cost of Capital,Labor Cost, Technological Base,Marketing Capability). f (Output,
unique database,integrating data from multiple sources . Our focal firms are manufacturing firms whose primary business is in semiconductors (i.e.,SIC code We chose to confine our sample to one SIC 3674). code to minimize the impact of market structure factors on 7 Our sample consisted of relative firm performance. 92 publicly traded firms in this SIC code. For each firm in sample,we collected information pertaining to our the resources available to each of the three functional areas,the outputs,and firm performance from the Com- database for the years 19851994. We pustat should mention that the Compustat database was incomplete to some variables. In such cases,we with regard con- original company annual reports to get the sulted information. The Compustat database,however,did not us information pertaining to the firms give innovative output. For this we conducted a content analysis data gathered from the patent office. patent of Finally,we also collected additional information pertaining to firms primary product lines from their Web pages. estimation consisted of a random subsample Our 72 of firms,while we used a hold-out subsample of 20 firms for post-sample prediction tests. cost Furthermore,observations from the first two years we used the for initializing the stock variables used in the analysis. 3.2. Operationalization of Variables 3.2.1. Dependent (a) Sales (SALES): Variables dollar amount of actual billings for Defined as the reg- sales completed during the period,reduced ular by cash and trade discounts. (b) Quality-adjusted technological output (TECH_ OUTPUT The technological output of a firm has ) been frequently measured using patent counts (see,for example,papers in Griliches 1984),which represent the number of patents assigned to the firm over a time period. Such raw output measures have been subjected to much criticism,because they treat all patents on an
7
(4)
3. Empirical Analysis
3.1. Description of the Data Set conceptual framework offers a resourceOur based perspective on a firms capabilities and the impact of these capabilities on its financial performance. To test this,we needed data on a firms resources,its func- outputs,and its financial performance. tional Furthermore,because of the importance of stocks of various variables (e.g.,sales stock) and the lagged impact of some variables (e.g.,R&D expenditure) on performance,we required these data to be over a period of time. This posed a major challenge because no database currently exists that integrates the types of infor- we needed. To this end,we put mation together a
554
Even within four-digit SIC codes,there may exist variations in market structure. We control for this by introducing dummies that re ect variation in market structure.
Among others,Scherer (1965) has shown that the distribution of patent values is highly skewed toward the low end,with a long and thin tail toward the high-value side (Trajtenberg 1990).
8 Thus,we need to adjust raw equal footing.for quality. Recall that the literature has patent counts high- two dimensions of quality of a new lighted technology: innovativeness (Trajtenberg 1990) and width of
applicability of the innovation represented by that pat- We construct the alternative quality-adjusted ent. out-measure, put TECH_WIDTH as follows. We first , calculate the proportion of citations received by a patent firms belonging outside our focal SIC code from (3674). equal to the number of citations received This is by a patent from firms outside the focal SIC code,divided by the total number of citations received by it. The weight assigned to a firms patent,then,is the propor-outside citations for the patent,divided by tion of the sample average proportion. The sum of these proportion-of-outside citation-weighted patents,for year,for a particular firm,would be a particular the value TECH_WIDT for that firm for that year. of H We correct for truncation bias as in the case of TECH_INNV . (c) Cost of goods sold (COST). This is defined as all costs directly allocated by the company to production, such as material and overhead,and is a proxy for the average cost of production. (d) Relative profitability (REL_PROFIT). We bin q as our measure of profitability. This is Tos the capital market value of the firm divided by defined as the replacement value of its assets. q has Tobins in the literature,because of its many used widely advantages over accounting measures. It is an inherently dynamic measure in that it maximizes discounted cash Also,as pointed out by Montgomery ow. and Wernerfelt (1988),because q is a Tobins of firm rents,it implicitly uses the correct market measure riskadjusted discount rate and minimizes distortions resulting from tax laws and accounting conventions. Because our focus is on relative firm performance,we define the index q for firm i in RE _PROFI L t to the T year Tobins (i.e., qi t / Avg.q ). 3.2.2. Explanatory Variables. logical know-how (TECHBASE)
it
applicability (Jaffe et al. 1993). We discuss below our two operationalizations of the quality-adjusted innovative output TECH_OUTPUT based on these construct, , two quality dimensions. (1) Innovativeness-adjusted technological output (TECH_INNV) Consistent with the empirical R&D : literature (Trajtenberg 1990,Jaffe et al. 1993),we measure the innovativeness of technological output by measur- number of times the patents of a firm have ing the been The underlying premise here is that the cited. more innovative the technology,the higher would be its citation count. We want to emphasize that prior studies have provided empirical evidence linking patent cita- and the innovativeness of technologies tions (e.g., 9 Albert et al. 1991). construct the citation-weighted patent count We as follows. We first calculate the average number of citations received by all the patents belonging to the in our sample. The weight assigned to a firms firms patent,then,is the number of citations the patent has received,divided by the sample average. The sum of these citation-weighted patents,for a particular year, particular firm,would be the value for a of TECH_INN for that firm for that year. As pointed V in the literatureout (Trajtenberg 1990),the use of citations quality suffers from a truncation bias. to infer Because goes only to 1994,patents issued in or our data near year would not have all their citations that captured, truncation bias. We explicitly control for causing a this 10 bias while TECH_INNV calculatingWidth-of-Applicability-technological output . (2) (TECH_WIDTH) Prior literature (e.g.,Jaffe et al. . suggested that if a 1993) gets a large number has patent of citations from outside its industry,it re ects a wider
9
use
been
capital
as the ratio Tobins sample average
of q
(a) Base of techno-
Our patent search was exhaustive in that we did not confine it to All patents that had cited patents of our
any particular SIC code.
. This construct is based on (quality-adjusted) technological TECH output ( OUTPU ,measured either TECH_INN or _ TECH T by base of technological know-how is V _ WIDT ). The H computed by estimating a Koyck lag function on TECH_OUTPU ,with earlier years of innovative T technological output receiving a lower weight than later years. Formally, TECHBAS E for period t is defined as
focal firms were searched for,leading to an examination of approximately 10,000 patents. 1 0 Further details on the construction of the citation-weighted index, along with the adjustment for truncation bias,are provided in the technical appendix that is available from the authors on request.
555
kt TECHBAS dt k TECH_OUTPU t k 1 E T represents the weight attached to past values of inno- output. The higher the value vative d,the
. Here
greater
of the spillover from past levels of innovative output. Stock of marketing expenditure (MARKETING(b) STOCK . To capture the carryover effect of ) marketing expenditure/effort,we estimate the stock of marketing available to the firm using a Koyckexpenditure lag structure. MARKETINGSTOC Formally, as K is defined MARKETINGSTOC SGAEXPENSE K Here . k
t
(g) Output . This is the dollar amount (OUTPUT) of production for the firm. Cost of capital (CAPITALCOST) (h) . This refers to long-term cost of capital and represents the the average rate for long-term borrowings for that interest firm. Unit (i) labor cost . This is defined (LABORCOST) as the cost of employees wages and benefits allocated
for period t k kt c k 1 weight
to continuing number of employees
operations,divided
by
the
total . R&D ca-
. (j) Relative R&D capability (REL R&DCAP) pability R&DCA ) is calculated using the ( P inputoutput approach and is defined as the inverse of the inefficiency (expressed as a percentage of R&D
to past values administrative (SGA) expenses of the firm. While SGA also includes are not strictly within the domain of items that marketing,it is a good proxy for the amount the firm on its market research,sales effort,trade spends expenses,and other related activities. The higher the value c,the greater the spillover from past levels of of SGA . (c) Advertising stock (ADSTOCK) . To capture carry-over effect of advertising,we estimatethe the stock of advertising effort available to the firm using a Koyck-lag Formally, is defined as structure. ADSTOC t K PENSE k . The higher the value spillover of from past levels of ADSTOC Kkt x tk
k 1
c represents the attached of sales,general,and
the maximum achievable innovative TECH output, ). We define the INNV REL_R&DCA by _ index P for a certain year considering R&DCA of a firm the the P sample relative to R&DCA (i.e., R&DCA / Avg. average P P R&DCA ). P (k) Relative marketing capability (REL_MKTCAP) Marketing capability is calculated using the input- method. It is defined as the inverse of output the SALES inefficiency (expressed as a percentage of the maximum achievable SALES ). We define objective,i.e.,REL_MKTCAP the by considering MKTCAP index the of a firm for a certain year relative to the sample average MKTCAP (i.e., MKTCAP / Avg.MKTCAP . To (l) Relative operations capability (REL OPCAP) ). . Opera.
for t period ADEX greater
x ,the the advertising
expenditure. (d) Installed customer base (INSTALLEDBASE) capture the importance of installed base,we estimate of sales available to a firm using a Koyck the stock lag structure. INSTALLEDBAS for Formally, defined as
t is E t period SALES k kt INSTALLEDBAS n . t k k 1 E value The higher the n,the greater the spillover of from past sales. Receivables (RECEIVABLES) (e) . We measure the firms resources devoted to building customer relationships by the level of its receivables. This is defined as claims against others collectible in cash. (f) Accumulated R&D expenditure (CUM R&DEXPENSE . We estimate the cumulative R&D ) penditure using a exKoyck lag structure,with declining into the past. weights CUM_R&DEXPENS Formally, for kt period t k E t is defined as CUM_R&DEXPENS E The higher the value R&DEXPENS . k 1 k E of ,the greater the spillover from past levels of R&D expenditure .
556
t
tions capability OPCA ) is calculated using the ( input output method and isPdefined as the inverse of the operations inefficiency (expressed as a percentage of the minimum achievable cost, index REL_OPCA by ). We define the OPCA of a Pa certain year the firm for relative to the sample P average (i.e., OPCA / Avg.OPCA OPCA ). P 3.2.3. Variables. ment P Control P (a) Competitive environCOS T considering
). Even though we confine 17 (SUB_MARKET one industry,competitive our attention to conditions industry might vary,depending on the within the spe- product niches occupied by each firm. Each cific of these product niches can be thought of as a separate submarket,characterized by unique competitive conditions. To account for such variations and their impact of a firms sales potential and profitability,we cre- dummy variables for each of the product ated lines
that exist in our focal industry. SUB_MARKET i k is defined as:

SUB_MARKET
ik
11
The variable
in
dummy
1 if firm 0 else.
i operates submarket
k ,and
level of resource input. Thus,the further (downward the realized sales is from the sales distortion) frontier, the inefficiency of the firms marketing the higher function,and consequently the lower is the firms marketing capability. the Cobb-Douglas (C-D) formulation We use and specify the sales frontier (equivalently,marketing 12 transformation function) as follows:
k 7 k itk k m 8 m 9 mm 10 1
Thus,a firm in three product niches would have three ones and five zeroes for the eight dummies. The fact firms in our sample are indeed heterogeneous that in terms of product markets is demonstrated by the fact the least populous product category has 4 that firms, the most populous has almost while 30.(b) Business cycle effects (YEAR ). It is likely that macroeconomic conditions would have changed over 10-year time period that our data cover the
1 9
ln( SALES
) it
mm 0
SUB_MARKET
ln( ADSTOCK
it
)
it
ln ( MARKETINGSTOCK ln( TECHBASE

it it
(e.g.,the U.S. economy went through a recession in the late 1980s). We control for such business-cycle effects through variables as follows:
YEAR
tk
)
m 12
11
dummy
YEA R
1 9
,which defined
to
are
ln( RECEIVABLE S ln( INSTALLEDBASE
)
ititit
mm
g ,
(5)
1 if the observation (year 0 else.
t ) pertains year
k ,and
3.3. Empirical Model Specifications we specify the functional forms for In this section the transformation functions associated with marketing,operations activities (Equations (5)(8)) R&D,and as as provide the parametric specification of well the performance-capability model (Equation (9)). 3.3.1. Modeling Marketing Capability. section 2.3.1 we defined the marketing capability of a as its ability to deploy its resources firm In sub-
where the i represents firms t represent subscript can rewrite Equation (5) more compactly s and years. We asmmmmm mmm , ) Y it i t i t i t i t (X g f( X , ) e . i f t In addition to the marketing resources/inputs viz, TECHBASE, MARKETINGSTOCK, ADSTOCK, INSTALLEDBAS ,and RECEIVABLE we E S included the control SUB_MARKET variables reasons. As mentioned in the previous following section,even within the four-digit SIC industry classification,there could be variations in the competitive environment facing a firm. Thus,a firm facing a higher competitive intensity would be able to achieve a lower of sales revenue (relative to an equally level efficienta less competitive submarket) for a given firm in level of deployed resources. We include the dummy vari- SUB_MARKET ables 1 7 to control for the effect of market factors on a firms sales frontier. Note that,given
1 2 Note
also for 1 7 the
viz.,stocks know-how of technical TECHBAS ),marketing ex( E penditure MARKETINGSTOC ),and advertising ex( K penditure ADSTOC ) along with investment in ( K tomer relationship ( base of customers cusRECEIVABLE ) and the S installed INSTALLEDBAS )efficiently SALES ).
( to achieve maximu E possible sales level the m ( inputoutput Note that in our conceptualization,a firms marketing capability is measured by how close the realized sales is to sales frontier given the
11
certain
This categorization was done by consulting trade press reports and two independent experts. These experts were given an exhaustive list of all the products carried by the firms in our sample and asked to categorize these products independently. The categorization arrived at was consistent across the experts and consisted of eight categories .
that the Cobb-Douglas is a parsimonious specification and provides a first-order Taylor-series approximation to an arbitrary transformation function. We also used the more exible indirect translog (ITL) specificationa second-order Taylor-series approximationas an alternative specification. However,the EichenbaumHansen-Singleton (E-H-S) specification test (1983) failed to reject the C-D model in favor of the ITL model. Hence for parsimony,we re- only the C-D specification. For similar reasons we selected port the C-D specification for the R&D and operations frontiers. Technical details of the E-H-S specification test are given in the technical appendix.
557
the panel nature of our dataset,explicitly accounting for such cross-sectional (submarket differences) is important because,in the absence of such control variables,a lower sales revenue (relative to the sample), level of resources would be incorrectly for a given attributed to marketing inefficiency,even though it is a result of exogenous factors outside the firms control. interpretation of the parameters is as The m follows. The parameter represents the marginal product 8 of ADSTOC ,i.e.,the % change SALES as a result K % changein of a ADSTOC . A similar in K holds for the parameters throughmm Based on . interpretation 91 2 our discussion in 2.3.1,we would expect these parameters to be positive,i.e., through mm If there 0. 81 2 is a m certain parameter,say 1,we infer that the sales revenue exhibits diminishing marginal product with to advertising. We do not have any priors respect on relationship between the marketing resources the and (i.e.,nature of the marginal productivity of sales resources) but rather view it as an empirical issue. Equation (5),captures the intrinsic m In it randomness sales level. The error in a firms component captures the inefficiency in marketing operations resulting in subpar performance by the firm on its sales objective. Such inefficiency (downward deviation from the sales frontier) could arise because of either (a) allocative inefficiency, i.e., suboptimal allocation of marketing resources (e.g.,inappropriate targeting and segmentation),or (b) technical inefficiency, i.e.,sub- utilization of resources,possibly optimal because of agency problems. Also,denotes the composite emt i error term,including the random shock and inefficiency error,and denotes the difference between the observed mmm the sales,,andi t i Yf predicted (X , ) sales, . tmake the following distributional We assumptions mm regarding the stochastic components and . i t i The m random shock is assumed to be distributed t normal m2 with mean 0 and variance ,i.e., r2 it mm . The marketing inefficiency error component is assumed to be distributed truncated normal (i.e., 0) with mean
13
( l m , r 2 ). Thus,the l m captures gm parameter average level of marketing inefficiency (synonymously,average marketing capability) of firms in the sample. Notein this baseline set-up,all the sample firms that have the same (expected) level of capability (although the realized values of capability may vary across firms). However,it can be argued that firms could vary in their marketing capabilities because of unobserved heterogeneity (Boulding and Staelin 1995). We discuss our treatment of such unobserved (intrinsic) variations across firms in their marketing capabilities in 3.4.1.
the
3.3.2. Modeling R&D R&D ity Capability. viewed as its ability to deploy its capabilof a firm is resourcesviz.,stocks of technical knowTECH how ( ),R&D expenditure ( BASE CUM_R&DEXPENS), E along with its marketing MKTCAP )efficapability ( achieve ciently to maximu possible the m measured qualityadjusted technological output (as by either TECH_INN or TECH_WIDT ),given the level of V H deployed resources. Thus,a firms R&D the capability gmt i 0 is measured by how close the realized technological out- is to put innovation frontier. theFor parsimony (see footnote 12),we use the C-D formulation and specify the innovative frontier or R&D transformation function as follows:
ln( TECH_INNV
r 2 rr 3 it
) ln(
rr 01
TECH BASE
it
it
ln( CUM_R & DEXPENSE ln( MKTCAP

it i t it it
) MKTCAP
it
) ln(
rr
ln( TECHBASE
g ,
(6)
g N (0, r gmt i gmt i

m it
it
where the i represents firms t represent subscript can rewrite Equation (6) more compactly s and years. We as: Y rrrrr i t i t i (X , ) g f( X , ) e . it ti rrri t f t The interpretation of the parameters is as )13 follows. r The parameter represents the marginal product 1 of TECHBAS ,i.e.,the % change TEC _INNV as a reE of a % change in sult TECHBAS H . A similar in E interprerr tation holds for the parameters through . r Conceptually,the parameter captures the 1 learning- effect while the parameter captures by-doing
24 r 3
lm
0 and ,i.e.,
2 variance rg g
We later extend the SFE formulation to allow for byheteroskedasticity allowing the variance of the random shock to vary across firms, with the variance assumed to be proportional to the mean sales,i.e., 22 rr SALES h . See 3.4.2. for additional details. Similar ,i i extensions are made for the R&D and operations capabilities models.
the voice-of-the-customer impact of marketing (Griffin and Hauser 1993) on the quality of technological output across both dimensions. Similarly,parameter
r 4
558
captures the role of marketingR&D interaction in enhancing a firms quality-adjusted technological output. on our discussion in 2.3.2,we would Based expect rr through 0 . Similar to marketing capability,the error term the intrinsic randomness in a firms captures innova- output 1 4 while m tive gcaptures the 0 with, ciency in operations.also We R&D ineffiit 14 r it
ln( COST
it
) ln(
aoo 01
OUTPUT ) ln( ) ln(

o 3 o o 5 oo
it
)
i t)
ln( LABCOST ln( TECH BASE
it
CAPCOST MKTCAP
it
i t it i t
(7)
estimated the innovative frontier,Equation (6),with TEC _WIDT instead TEC _INNV H H of H measure for (quality-adjusted) technological output. from Recall 3.2.1 that these two operationalizations two quality dimensions of correspond to the technol- innovativeness and width of applicability. ogy,viz. Essentially,both measures represent the (weighted) total number of patents filed by a firm in a year,with
where the i represents firms t represent subscript can rewrite Equation (7) more compactly s and years. We as: Y ooooo t i t i(t X , ) g f(X , ) e . it i ooot i f it The interpretation of the parameters is as follows. 0 represents the economies of size,so that the production technology exhibits increasing,constant or scale according as is 1/ decreasing returns to
o 1
as the
greater than,equal to,or less than 1. The parameters and o ,being cost elasticities,represent the marginal
3
o 1 o 2
the weighting schemes re ecting the two quality dimen- As sions. TEC _INNV ,we would expect the with H parr rameters through in the case 0 TEC 14 of H WIDT . H 3.3.3. Modeling Operations Capability. ations capability of a firm is viewed as its ability to its deploy resources LABCOS viz.,labor ( T ital CAPCOST ) along with its stocks of Oper) and cap-
impact of LABCOS and CAPCOST on COST and are exoo pected toTbe 0. The parameters and capture 45 the impact of technical know-how (pool of process inno- and marketing capability on a firms vation) opera- capability and are expected to be 0 tional (i.e.,reducing cost of production). to Similar marketing and R&D capabilities,cap-randomness in the production tures the intrinsic process,and 0 captures the inefficiency in the go production process resulting in higher-than-optimal cost. 3.3.4. Modeling the Relationship Between Functional Capabilities The andProfitability. ship between a firms functionalrelation(R&D,marketing,and and financial performance operations) capabilities is specified as follows:
k 7 kk
o it
it
( technical )efficiently to achieve know-how TECHBAS ( E minimum possible level the cost of of COS production ( capability is measured T Thus,a firms operations by close the realized cost of production is to how the frontier/function. 15 As before,we use the C-D formulation and thespecify cost frontier/operations transformation function as follows :
).
cost
ln( REL_PROFIT
k 16
) it
f0 f
SUB_MARKET
f kit,k
k 8
YEAR
7 17
f ) ) ln(
ln( REL_MKTCAP f19
it
) )
f1 8
1 4 These
ln( REL_R&DCAP ln( REL_MKTCAP ln( REL_OPCAP
it
ln( REL_OPCAP REL_R&DCAP ) ),
it
random shocks could arise for many reasons. Innovative activity is inherently stochastic in nature,with serendipitous discoveries punctuated by long periods of low output. Furthermore,a host of macroeconomic variables could affect innovative output. Apart from obvious policy variables,external market conditionscould play a big role (e.g.,the recent crisis in Southeast Asia has forced Intel to put off a manufacturing plant in Malaysia,in turn affecting its incentive to come up with cost-reducing innovations particular to those conditions). that cost function is the dual counterpart of function
f2 0 f2 1
it it
it it it
) ln(
REL_R&DCAP
(8)
where the i 1,...,92represents firms and subscript t 1,...,10represents years 19851994. We control for cross-sectional variations in firms competitive environment and longitudinal variations from macroeconomic resulting uctuations through the use of dummy SUB_MARKET variables, YEA . R
559
1 5 Note
productionfrontier/
that measures maximum level of production that a the firm can achieve,given the level of the deployed resources.
and
The interpretation of the parameters is as follows. The f 17 throug f 19 capture the parameters a firms functional capabilities on its h marginal impact of profitability. For f 17 denotes the % change in instance, a firms profitability (relative to the sample average) as result of a % improvement in its marketing a capability (again,relative to the sample average). Because the parameters f throug f (being elasticities) h independent,a comparison 9 scale- magnitude of their yields into the relative role of the three insights functional in bestowing competitive advantage. capabilities The parameter f 2 0 captures the interaction between R&D and marketing capabilities,so f 0 that that a firm with a higher R&D capability can suggest lev- its high marketing capability more than a erage firm a lower with Similarly, tures marketing capability. f 21 f 2 1 cap0 implyR&Doperations interaction,with profitability is enhanced by ing that a firms better R&Doperations coordination. 3.4. Econometric Methodology Stochastic Frontier Estimation Consistent with economic theory,our conceptual assumes an optimization behavior for framework the firm in its R&D,marketing,and operations functions. Thus,the objectives of R&D and marketing functions were postulated as attainment of maximum qualityadjusted technological output and sales,respectively, of the operations function was while the objective postulated as attainment of minimum cost of production level of deployed resources. In for a given reality,a fail to attain optimal results because of firm may inefficient deployment of resources (allocative ineffi- and/or inefficient utilization of resources ciency) (tech- inefficiency). In fact,the proposed nical conceptual frameworkviz.,the input-output approach explicitly recognizes the existence of these inefficiencies and it to the notion of a firms capability links postulated in the RBV literature (e.g.,Amit and Shoemaker 1993). detailed in 3.1 and 3.2,our data set As contains observations on the realized output levels and the lev- of els deployed resources relating to R&D,marketing, and operations functions for a sample of 92 firms (in 3674) for the period 19851994. The SIC econometric is task threefold: calibrate the marketing,R&D,and To operations
560
20 17 1
transformation functions (Equations (5)(7)) linking the functional resources/inputs and outputs estimate the functional capabilities To (equivalently,inefficiencies) displayed by these firms To measure the relative impact of these on capabilities financial performance (Equation the firms (8)). Given the importance of estimating functional inefficiency,it is crucial that we use the appropriate meth- for the task. In the literature,there are odology two approaches to estimating economic efficiency: data envelopment analysis (DEA) and (2) StochasticEstimation (SFE). DEA uses linear Frontier
16
are
(1)
would
programming techniques to construct economic frontier and estimate technical and allocative inefficiencies. The main advantage of DEA is that,being a nonparametric method,no explicit functional form needs to be im- on the data. However,the main drawback posed of DEA is that the economic frontier is assumed to be deterministic or nonstochastic,so the estimated frontier may be warped if the data are contaminated by statistical noise (Bauer 1990). In contrast,the SFE approach explicitly allows for the existence of inherent randomness in sales,innovation,and production processes (resulting from events outside the firms control) allowing certain types of specification besides error omitted variables uncorrelated with the and regressors al. 1977; see Ferrier and Lovell 1990 for (Aigner et additional discussion on the relative advantages of the two approaches). In this paper,we use the SFE methodology to implement the proposed input-output framework. Below,we provide the specifics of the SFE formulation optimal behavior when the maximization entails function (as in the case of marketing objective and functions,where the objectives are to attain R&D maximum sales and quality-adjusted technological output, 1 7 This basic SFE formulation is due respectively) .
16
of an
to
Note that using a linear model formulation for the transformation function (estimated through OLS) would recover only the average linkage between the resources and output rather than the frontier/
optimal relationship because the linear model implicitly assumes that firms are operating on the efficient frontier. Furthermore,be- are assumed to correspond to the efficient cause the data frontier, no inefficiency is allowed for,and as such,OLS cannot be used to estimate a firms functional capabilities. 1 7 Analytical details of the SFE formulation corresponding to the
min-
Stevenson (1980). Sections 3.4.1 and 3.4.2 extend the basic formulation to allow for unobserved heterogeneity and heteroskedasticity. SFE Formulation for the Maximization Problem Consider the frontier transformation . function Y i t i t it i t i t(X , ) g f(X , ) e ,(9) i f t where Y i t denotes the appropriate function of the out-TECH_INN put ),in the case of R&D it (e.g.,ln( for V sample firm, fron- 1,2,..., tier) ith i N ,in tth the the time t 1,2,..., T ; X is the vector of period,functions of inputs/resources associated appropriate with the ith sample firm in tth time period; and is theof the coefficients for the associated vector the independent variables in the transformation function impacting innovative output. Thus,in Equation (9), represents the deterministic component of the efficient frontier and represents the maximum expected output given that i employ X level firm s resources efficiently . Let represent the purely stochastic error it component (random shocks) impacting output,assumed to independent and identically distributed be as (0, r 2 ) . N gi t represent the inefficiency Further,let error component in the transformation process adversely affecting the output,assumed to be an independent and identically distributed nonnegative random variable,by the truncation (at zero) of the defined N (l , r 2 ) g distribution with l 0.1 We further assume 8 mode the random ,and the that inefficiency g ,are shock, independent,i.e. E[error,] 0,and that these g it it , error components are distributed independently of the independent variables in the E [X i t model,i.e., E[X g ] Given0. a sample of
imization
it it it it it it
each firm,it can be shown (Battese and Coelli 1988) sample likelihood function for the SFE that the formulation corresponding to Equation (9) is given by
NT
L
i 1 t 1
1 rr
22 g
1U
rg [Y itfi t
gg
(X , )]
rl
22 rr rrrr 2 2 g
Y ift i t
(X , ) r 22 r g
l l
1U,
(10)
where (.) and f(X i t , )
U (.) denote the standard normal
den- distribution sity and functions,respectively. likelihood (ML) estimates Consistent maximum theof model parameters can be obtained by maximizing the sample likelihood function (equation 10). Using the parameter estimates,we can ,compute the rela- functional capabilities of tive firm given in the Appendix. i for year t, i t,as
of
3.4.1. Controlling for Heterogeneity in Firms Capabilities andTechnologyMSM Estimation.for the output frontier (Equation Recall that 9), notes the expected level of the inefficiency error component,which is assumed the same across all the firms sample. In effect,this corresponds to the in the assumption that the firms are identical in terms of their expected functional capabilities,although they may in terms of the realized values of their differ capabili- is a restrictive assumption. Unobserved ties. This heterogeneity in firms capabilities may exist,for instance, of differences in their managerial because capabilities and (Boulding Staelin 1995). Another implicit assumption in the above formulation is that the model parameters are assumed to be same across the firms. Again,this is a the potentially assumption. For instance,consider the restrictive marketing capability model,Equation (5). The impact of marketing inputs/resources such as advertising and installed base is likely to vary across firms,dependingof their product lines. One can on the nature similarly argue for potential unobserved heterogeneity across with respect to Equations (6) to firms (8).
561
l de-
it
N firms with
T observations for
problem (as in the case of operations where the objective is to attain minimum cost of production) are omitted for brevity. Details are given in the technical appendix,which is available from the authors upon request.
18
Note that while the random i t can take any positive or negshock ative value,the inefficiency error git can take only component It is this difference in their supports of distribution positive values. that allows for identification. Furthermore,it is the fact g is it that defined only on the positive orthant that allows us to interpret it as the inefficiency component.
It has been pointed out in the literature (Chintagunta et al. 1993) that failure to control for unobserved heterogeneity may lead to inconsistent parameter esti- To allow for variations in firms functional mates. capabilities and elasticities,we use a parametric random specification (Gonul and Srinivasan 1995). effects Specifically,we capture unobserved heterogeneity across firms in their marketing/R&D capabilities (Equation positing that the (9)) by l follows a gamma parameter distributio lC (s ,w ),which is a reasonably n distribution. To control exible for heterogeneity on elasticities,we assume that Allowing for unobserved N ( , R ).
r 22 Yh ,where Y T 1 T 1 Y i t is the average it i i r level (sales and quality-adjusted technological output outputs for the marketing and R&D frontiers,respec- i. tively) for Note h 0 would correspond firm the case. that homoskedastic to
4. Results andDiscussion
4.1. Specification Tests, Comparison with Nested Models and Parameter Estimates For each of the functional capability models (Equations first test entailed comparing the (5)(7)) the CobbDouglas (C-D) specification with the indirect translog (ITL). 20 In each case,the EichenbaumHansenSingleton (E-H-S) specification tests (1983) failed to re- the nested C-D specification in favor of the ject more ITL specification. Having selected the general C-D specification,we then compared the full model with several nested models. The full model had: (1) unobserved heterogeneity in capabilities,(2) unobserved heterogeneity in the marginal products of inputs,and (3) heteroskedasticity of the random shock term. This compared with models with one or more of was these features absent to see which specification fit best.
unobserved heterogeneity,however, makes ML estimation computationally infeasible be- of the high order of integrals involved. cause There- obtain consistent parameter estimates,we fore,to use method of simulated moments (MSM). The the intuition behind the MSM procedure (McFadden 1989) is if,for instance,the proposed empirical R&D that capability specification is a true characterization of the linkage between innovative output and R&D resources,then the conditional moments implied by the model (Equation (6)) must match the observed sample counterparts asymptotically. To set up the orthogonality conditions,we use the moment,Y 2 X , ],respectively and E[|
it it
first
19
E[Y i t | X i t , ]
. for
3.4.2. Controlling Heteroskedasticity. Another implicit assumption
Us- the E-H-S test results,we selected the model ing with simplest structure consistent with data. the Furthermore,using J test (1982) for Hansens identifying restrictions,we tested for the fit of the data with the selected model. Sales Frontier: Based on the E-H-S selected model entailed a C-D specification with: (1) unob- heterogeneity in mean inefficiency,(2) served unobserved heterogeneity in marginal products/elasticities, and (3) heteroskedasticity of the random shock. MSM estimates of the sales frontier are given in Table 1. All the parameters are significant,except for STOCK ,and of the expected sign. The most resource significant MARKETINGSTOC ,followed
over-
of the SFE specifications discussed previously is homoskedasticity of the composite error e t ,so that its i component, assumed to be the same across all the variance, 22 r ,is r the sample. Given the differences in the scale firms in of operations (in terms of sales,number of patents,volume of production) of these firms,it seems reasonable heterogeneity in the variance of the to allow for com- error component and then test for posite homoskedas- of imposing it a priori ticity,instead (Kumbhakar account for heteroskedasticity 1997). To parsimoniously,we allow for firm-specific variance for the ran- shock with variance proportional to the dom mean (Baltagi and Griffin 1988). Specifically,we output let
19
tests,the
r2 e
AD -
TECH is K by BASE and RECEIVABLE . The elasticities of the are positive S and less than inputs one,indicating a diminishing marginal productivity.
20
Analytical details are appendix,available from the authors upon request.
given
in
the
technical
For both the C-D and the ITL models,we assumed homoskedasticity and no unobserved heterogeneity. Details of the E-H-S and Hansen J tests (along with the test statistics) are provided in s the technical appendix.
562
Innovative Frontier: Based on the E-H-S tests,the selected model entailed a C-D specification with: (1) unobserved heterogeneity in mean inefficiency,(2) no unobserved heterogeneity in marginal products,and (3) heteroskedasticity of the random shock. MSM no estimates of the innovative frontier are given in Table 2. results are consistent with our The expectations. Based on the magnitude of the coefficients,we find that TECHBAS seems to be the most important E input,followed CUM_R&DEXPENS and MKTCAP . by E effects action MKTCAP and TECHBAS Interare
l
Table 1 Parameter Estimates (MSM) of Marketing Capability Model

Pa ra m e te r E s t ima t e s (S t a nda rd Erro r) Po pul a t io V a ria b le s A n e ra ge Ef f e c v t Va ri a nc e of Un obs e rv e t He d e rog e ne it y C ompo ne nt
I mpa c t of I np ut s : (A DST OC 0 .2 157 (0. 13 ln ) m 8 K 92) m ln (MA RK ETI NGS TOC) 0 .6 815(0. 20 16) 9 K ** ln (TE CHB A) 0 .4 118 (0. 18 m 10 SE 52)* * ln (REC EI V A BL 0 .3 752 (0. m110 ) 1 ES 97)* * m ln (I NS TA LLE DBA S 0 .2 747 (0. 09 ) 12 E 15)* * I ne f fic i e nc y Err or: Mod e of I ne f fic i e nc y Erro r Te rm 45 .0 492 (9. 72 m 16)* * V a ria n c e o f I n e ffi c ie n c y or T er m Err
V r( ) V r( ) V r( ) V r( ) V r( )
m a 8 m a 9 m a 10 m a1 1 m a2 1
0. 09 15 (0 .0 71 1) 0. 33 51 ( 0. 102 5)** 0. 27 04 ( 0. 153 6) 0. 05 12 (0 .0 46 2) 0. 07 49 ( 0. 028 3)**
between E positive and significant. Cost Frontier: Based on the E-H-S tests,the selected model entailed a C-D specification with: (1) unob- heterogeneity in mean inefficiency,(2) no served unobserved heterogeneity in marginal products,and (3) heteroskedasticity of the random shock. MSM no estimates of the model parameters are given in Table 3. The effect of on cost significant,suggesting the positive impact of operationsmarketing coordination (design for manufacturability). Capabilities-Profitability Model: The E-H-S indicated the presence of significant unobserved heterogeneity in the model f across MKTCAP is
r2
gm
V a m ) 7. 04 92 ( 3. 281 l r( 5)** 1 0. 921 4 (4. 03 58) **
Ra nd om s ho c k V:a ria n c e o f R a ndo m Sh oc k 5 7. 710 3 (1 5. 229 r2 3)* * He t e ros k e d as t i c it y Pa r a me t er Mi nim iz e d C rit e ri on F un c t ion v2 S t a t is t ic f or H a ns e nT e s t (d. f J s .)
m
1 .6 294 (0. 38 94) ** 0 .0 021 4 1. 177 (40 ) 2
test
21
**: significant at 1% significance level : significant at 5% significance level
the parameters firms. We conducted the Breusch-Pagan test (Green 1997,p. test for heteroskedasticity. 552) to v2 test The of 0.08 with 3 d.f. is less than the critical value of statistic 0.12,the homoskedasticity assumption is not so rejected at 1% significance level. Similarly,Breusch-Godfrey tests (Greene 1997,p. 595) for autocorrelation (AR(P) and MA(P) processes) failed to reject the hypothesis of no autocorrelation at 1% level of significance. Table 4 reports the MSM parameter estimates. theAllparameter estimates are positive and significant. The interaction effect REL_R&DCA between REL_MKTCAP P is found to be the most important determinant of inter-firm variations in profitability,followed REL_R&DCA and REL_MKTCAP . by Furthermore,to test the predictive validity of P the proposed specification (Equation (8)),we conducted the Hoffman-Pagan post-sample prediction test (1989)
2 1 As
on a holdout sample of 20 firms (we used 72 firms for estimation sample). the v2 test statistic of 8.72 The is less than the critical value of 19.92 at 1% with 40 d.f. significance level,indicating that the model fits the hold-sample well. An additional measure of out goodness2 of-fit is provided by the pseudovalue R . pseudo-R 2 value of 0.92 suggests a good predictive validity for the proposed model. 4.2. Discussion of Substantive Insights and Managerial and Implications Our findings offer a number of substantive insights and managerial implications. Sales Frontier: Our first substantive insight to the importance pertains MARKETINGSTOC and of K frontier (Table 1: m CEIVABLE in the sales 9 S and 0.3752). 0.6815 that the stock of m Recall marketing
2 2 The
22
Th e
RE -
11
with the capability models (Equations (5)(7)) where we allow the parameters to be randomly distributed across firms,we allow the f to be randomly distributed parameters as necessitates the use of MSM estimation methodology.
pseudo-R
is computed exactly like the R
N ( f,R ). This
f
linear model. However,because the model is nonlinear,the pseudoR2 is not constrained to lie between 0 and 1 (Greene 1997). We thank an anonymous reviewer for suggesting this measure to us.
measure the
for
563
Table 2 Parameter Estimates (MSM) of R&D Capability Model Parameter Error) Variables TECH_INNV TECH_WIDTH Estimates (Standard
Table 3 Parameter Estimates (MSM) of Operations Capability Model

P a ra m e te r E s t ima t e s (S t a nda rd Erro r) V a ri a nc e of Uno bs e rv e Hedt e roge n e it y Co mpo ne nt
V a ri a ble s P opu la t io n A v e ra ge Ef f ect C os t El a s t ic i t ie s : n(OUT l P) UT l n(L AB C OS ) T l n(C A PC OS ) T l n(T EC HB AS ) E l n(M KT CA) P
Impact of Inputs:
r 1 r 2 r 3 r 4
[ln( TECHBASE )] [ln( CUM_R&DEXPENSE [ln( MKTCAP )] [ln( MKTCAP ) )]
0.8713 (0.2738)** 0.8258 (0.2910)** 0.7512 (0.1725)** 0.6829 (0.3004)** 0.2594 (0.1783) (0.1326) 0.3217
o 1 o 2 o
3
o 4 o 5
ln( TECHBASE )] 0.5681 (0.1972)** 0.5902 (0.2141)** Inefficiency Error: l [Mode of Inefficiency
r
0 .9 138 (0 .3 825 )** 0 .2 429 (0 .1 382 ) 0 .3 183 (0 .0 947 ) ** 0. 13 80 (0. 11 73) 0. 27 05 (0. 07 28) **
V a r( o) 0. 091 5 (0. 07 1 11) V a r( o) 0. 132 6 (0. 10 2 25) V a r( o) 0. 227 4 (0. 08 26) 3 ** V a r( o) 0. 088 3 (0. 05 4 62) V a r( o) 0. 166 2 (0. 04 73) 5 **
I n ef fi c ie nc y E rror : M ode of I ne f fi c ie nc y ror T e rm 1 8. 231 5 (6 .6 821l Er ) ** V a ri a nc e of I ne f fic i e nc y ror T e rm Er
Error Term] 3.5271 (0.0862)** 3.6483 (0.0715)** r 2 [Variance of Inefficiency

gr
r2
go
V al o ) 6. 15 42 (4 .2 r( 815 ) 3 .5 118 (1. 00 27)* *
Error Term] 1.0527 (0.0227)** 1.0391 (0.0385)**
Random Shock: r 2 [Variance of Random Shock]

r
Ra n dom s h oc k V:a ri a nc e of Ra nd om S hoc k 14. 04 51 ( 3. 347 2)** He t e ro s k e da s t ic i t y P a ra me t er M ini miz e d C ri te r ion F unc t i on v 2 St a t is t i c fo r Ha ns e n s J Te s t (d .f . )
r2
o
1.2518 (0.0489)** 1.0463 (0.3258)**
1 .8 512 (0. 29 31) ** 0 .0 018 2 1. 001 (16 ) 6
Minimized Criterion Function 0.00153 0.00117 v 2 Statistic for J Hansens Test (d.f.) 0.8415 (35) 0.6435 (35)
expense includes incentives to salespersons,trade incentives,and customer incentives over the years,while proxy a firms ability to offer longer receivables credit terms to its customers. In many high-technology markets,and especially in semiconductors,a firm sells its product to an OEM,who in turn combines this product number of other products and sells the with a systemend consumer. In such an industrial market to the setting,trade incentives and incentives to salespersons importance. Also,the importance of are of great loose terms and other such expenditure can now credit be readily understood as a device to satisfy OEMs as much as possible. The significance of the OEM as an intermediary also explains why advertising is not of great significance in this industry. OEMs are generally more knowledgeable than the average end consumermuch less likely to be in uenced by and are advertising than salesforce.
23
Table 4 Parameter Estimates (MSM) of Capabilities-Performance Model

P aram et er E st im at es (S t and ard Er ror ) P op ul at io n Av erag e Ef f ect V ari an ce of U nob serv ed He t erog ene it y
V ar iab les
M ain Ef f ect s : l n( REL _M K TC A P ) 0. 4 79 1 f7 . 09 92 ) (0 1 ** l n( REL _R &D CA ) 0. 5 10 4 (0f8 16 09 ) . 1 P ** l n( REL _O PC A) 0. 3 20 6 (0f9 18 . 1 P 37 ) I n t eract i on Ef f ect s:n( REL _M K TC A l P) l n( REL _R &D CA) 0. 7 38 2 (0 . 28 11 ) f 20 P ** l n( REL _O PC A) P l n( REL _R &D CA ) 0. 1 21 72 1(0 . 03 15 ) f P ** M in im iz ed Cr it er io n Fu nct i on v 2 S t at ist i c f or Han sen s J T est (d . f .)
V ar(f 1 7 ) 0. 1 52 7 ( 0. 09 81 ) V f 1 8 ) 0. 2 71 8 ( 0. 08 25 ) ar( ** V f 1 9 ) 0. 1 00 4 ( 0. 07 ar( 51 )
V f ) 0. 4 13 2 ( 0. 22 ar( 2 0 15 ) V f 2 1 ) 0. 0 21 7 ( 0. 01 79 ) ar( ** 0. 00 13 6 0 . 74 25 (5 0)
by
an
excellent
23
The result on the nonsignificance of advertising needs to be
564
The importance MARKETINGSTOC of K CEIVABLE leads to the following managerial S implication for the semiconductor industry and related markets. The focus in such markets should be on building relationships with customers (OEMs). long-term To end,it is important to have salespeople who this are technically competent because the aim is to convince OEMs. Similarly,great care needs to be taken in the selection and training of distributors,who in uence the purchase decisions in such markets crucially. Such distributors have to be wooed with appropriate trade incentives to ensure that they push the product enthusiastically. The importance of such relationship mar- has already been pointed out for a host keting of markets (Jackson 1985),but its significance in hightechnology markets has rarely been recognized. Our second substantive insight relates to the importance of prior stock of knowTECHBAS how ( uencing sales (Table 1: 0.4118). mConsider E the of a customer (i.e.,OEM) in a highcase technology market. High switching costs in such markets mean that OEMs would wish to go with a firm that is
10
and
RE -
AMD to the recent move of Vinod Dham,the engineer chie y responsible for the Pentium design,from Intel to AMD. Salespersons also need to be trained to high- a firms technological excellence in their light dealings with customers. In short,managers need to give a lot thought to configuring their marketing of activitiesthe common goal of communicating to around customers the technological excellence of their firm. Custom- to be informed of the innovative ers need technologies possesses and of the future R&D that the firm initia- undertaken by tives it. R&D Frontier : Our results on the R&D frontier lead to a couple of substantive insights. Our first insight to the impact of marketing on a firms relates innova- We find that marketing capability has tiveness. its greatest impact on the quality-adjusted output of firms have a strong technological base (Table which
r 4
)inin-
2: 0.5681 & 0.5902). Thus,in addition to the importance of interfunctional coordination suggested in prior literature (Gupta et al. 1987),we also suggest that mar- capability has a disproportionate interactive keting ef- The higher a firms technological fect: base,the is the impact of a higher marketing stronger capability on the R&D productivity. This insight translates to a important managerial implication: It is very precisely firms are good technologically that would get that bang from buck if they improved their the most marketing capability . The second substantive insight relates to the of impact marketing capability impact on the extent to which the innovative technology is applicable across a wide of industries (Table 2: 0.3217).r Recall range 3 that of marketings tasks is to listen to the one consumer up with a pool of ideas. A strong and come marketing would imply that this pool of ideas is capability wider, spanning a number of applications. Consequently,innovations from R&D that rely on this pool will have applications across a wider range of industries. This carries a strong message for managers: One result of most fecund sources of ideas for innovation is the the results of marketing activity. Thus,marketing needs to involved from the beginning of the be innovation process . Relative Profitability : Our results relating to firm performance suggest relative one important insight. The important determinant of firm performance is most the
likely a technology leader in the future. In such a to be situation,the focal firm needs to signal the likelihood of it a technology leader to in uence customer being expectations appropriately. Given the importance of in uencing customers, managers need to tailor their marketing activitiesthe need to inform customers of the around technological excellence of their firm. A consistent theme needs pursued in all interactions with the to be customer, through salespersons or through whether promotions. Thus,customers need to be informed of the innovative that the firm possesses and of the technologies future initiatives undertaken by it. Similarly,any R&D potential applications of innovative technology developed firm,and of technologies under by the development, should be emphasized to customers. The hiring of star scientists or engineers should be widely publicized case in point is the publicity an excellent given by
qualified. The industry itself seems to be changing,as suggested by the success of the Intel Inside campaign,followed by a highprofile TV campaign by Cisco Systems. Thus,advertising might well be growing in importance,even in the semiconductor industry.
565
interaction of marketing and R&D capabilities (Table 4: f 2 0 0.7382). This supports the assertion that firms in high-technology markets need to excel at two things: the ability to come up with innovations constantly,and the ability to commercialize these innovations into the of products that capture consumer needs kinds and preferences. This finding offers further evidence on the importance of R&Dmarketing coordination,as sug- in the marketing literature (e.g.,Griffin gested and Hauser 1996) and is applicable across a wide range of markets.
For example,operations could be concerned with on- delivery or with a low defect rate. time Marketing could be concerned with service quality or with cus- satisfaction. Similarly,for the R&D frontier tomer we could have used objectives such as time to commer- of a new product (Griffin and Hauser cialization 1996) number of new products introduced. It or the would be interesting to see what new insights result from the changing of objectives or the combining of a number of them.
24
5. Conclusions
This paper proposes a conceptual framework with the resource-based view (RBV) of the firm as its theoretical underpinning,to explain interfirm differences in firms profitability in high-technology markets in terms of differences in their functional capabilities. Specifically, we suggest that marketing,R&D,and operations capabilities,along with interactions among these capabilities,are important determinants of relative financial performance within the industry. The paper contributes to a number of different literatures. First, it contributes to the RBV literature by proposing an conceptualization of a firms input-output capabilitiesthen operationalized using the SFE which is methodology. Methodologically,the use of the SFE is an important step that permits measurement of a firms
g
Appendi x
Estimatin REL_MKTCAP g Note that in Equation (9),
it
and REL_R&DCAP
it
git is the composite error term, including the random shock and the inefficiency error component, and represents the difference between the observed output Y i t and the predicted output Y it f( X it ,where ), denotes parameter estimate. A consistent estimate of the inefficiency for firm t is given by (Battese and Coelli 1988): gE
i t i t it it i t
ei t i t
the i in period
[g | e e r2
* 22
] l i* i t t rr
**
l* l*
1
1 U ,(A.1)
where lr r2 2 22
g [ Y ift it 2 22 r 2rrr gg
l i* and t
( X ,)]
r2 *
rr g
. (A.2)
capabilities,using archival data,while explicitly linking productive resources to the attainment of the firms its objectives . Our study contributes to the literature on market orientation by suggesting that a stronger market orien- of a firm should get re ected in a higher tation mar- capability. It also adds to the literature keting on design for manufacturability by explicating the complementarity among the various functional capabilities empirical evidence on their relative and offering importance in in uencing a firms performance. represents a first cut at assessing This paper the role of firm-specific capabilities in intra-industry vari- in firms performance in high-technology ation mar- and has a number of limitations,which kets suggest avenues for further research. Thus,one could consider of different objectives for the three a host capabilities.
566
e e
A consistent estimate of is inefficiency (as a % of firm maximum achievable output,given the level of resources) in t is period given by As/ Y . mentioned earlier,we take the inverse of it i t inefficiency as the measure of the firms capability. Thus,the consistent estimate of capability of i in t,is given by CAP Y / g , which it itit firm period is scaleindependent. We estimate the sample average capability as follows. Let ( , ,..., e ,..., , ,..., ,) be the vector of sample 1 1 12 1T N1 N1 NT values of the composite error components i 1,2,..., N firms for for time period t 1,2,..., T . Then the consistent estimate of the s sample average inefficiency is given by
24
The authors gratefully acknowledge the insightful comments and suggestions of Rick Staelin,the area editor,and three anonymous reviewers on an earlier draft,and thank Prokriti Mukherji for research assistance and Anant Kumar for expert programming. This work was partly supported by a grant from the Marketing Science Institute; Faculty Research Fund at the Marshall School of Business, University of Southern California (to the first author); and the Beatrice Companies Faculty Research Fund at the Graduate School of Business,University of Chicago (to the third author).
E [g| e e r 2 22 **
l ** l ** rr
** **
Dutta,S.,A. Weiss. 1997. The relationship between a firms techno-logical innovativeness and its pattern of partnership l **
1
1 U ,(A.3)
agreements Management Sci. 43 343356. . Eichenbaum,M. S.,L. P. Hansen,K. J. Singleton. 1988. A time series analysis of representative agent models of consumption and leisure choice under uncertainty. Quart.J.Economics 103 5178. J.Royal
where lr r2 2
1
l **
r2 2 e gg
NT NT
22 rr g
;
NT
Farrell,M. J. 1957. The measurement of productive efficiency. Statist.Soc. (Series 120 253281. A) Ferrier,G. D.,C. A. K. Lovell. 1990. Measuring cost efficiency in
it
r 211
**
22
NT r
; and,
g
e NT e
. (A.4)
1t
banking: econometric and linear evidence. Econometrics 46 229245.
programming
J .
Thus,a consistent estimate of the sample average capability is given by CAP where refers to the sample Y Y / g average of output, and is given by of i in firm period [g / g ].
it
Gatignon,H.,J-M. Xuereb. 1997. Strategic orientation of the firm and new product performance. J.Marketing Res. 34 7790. Givon,M.,V. Mahajan,E. Miller. 1995. Software estimation of lost sales and the impact on software diffusion, 59 (1) 2937. piracy: J.Marketing
NT i 1
t is thus given by
Y . The relative capability it REL_CAP [Y / it it Y
Analytical details are given in the technical appendix.
Reference s Aigner,D.,C. A.
esti-
K. Lovell,P. Schmidt. 1977. Formulation and J.Econometrics
mation of stochastic production frontier models. 6 2137. dation of citation counts as indicators of industrial patents. Policy 20 251259.
Gonul,F. F.,K. Srinivasan. 1993. Modeling unobserved heterogeneity in multinomial logit models: methodological and managerial Marketing Sci. 12 213229. implications. Greene,William. Econometric Analysis. Prentice Hall,NJ. 1997. Griffin,A.,J. R. Hauser. 1993. The voice of the customer. Marketing Sci. 12 (1) 127. ,. 1996. Integrating R&D and marketing: a review and analysis of literature. 215. Griliches,Z. (ed.). 1984. the J.Product Innovation Management R&D, Patents, and Productivity. 13 191 University of
Albert,M. B.,D. Avery,F. Narin,P. McAllister. 1991. Direct valiRes.
Amit,R.,P. J. Shoemaker. 1993. Strategic assets and organizationalStrategic Management J. rent. 14 (1) 3346. Baltagi,B. H.,J. M. Griffin. 1988. A generalized error component model with disturbances. 745753. heteroskedastic Inter.Econ.Review 29
Chicago Press,Chicago,IL. Gupta,A. K.,S. P. Raj,D. Wilemon. 1987. Managing the R&D-marketing interface. Res.Management 30 (2) 3843. Gupta,Sachin,Pradeep K. Chintagunta. 1994. On using demo-
Battese,G. E.,T. J. Coelli. 1988. Prediction of firm-level technical efficiencies with a generalized frontier production function and panel data. J.Econometrics 38 387399. Bauer,P. A. 1990. Recent developments in the econometric estimation of J.Econometrics 46 3956. frontiers. Boulding,W.,R. Staelin. 1995. Identifying generalizable effects of strategic actions on firm performance: the case of demandside returns to R&D spending. Marketing Sci. 14 (2) 222236. Business Week. 1998. Innovation: you aint seen nothing yet. August 31,60 63. Chintagunta,P. K.,D. C. Jain,N. J. Vilcassim. 1993. Investigating heterogeneity in brand preferences in logit models for panel data. J.Marketing Res. 28 (4) 217228. Cohen,W.,D. Levinthal. 1990. Absorptive capacity: a new perspective on innovation. 128152. learning and Administrative Sci.Quart. 35
graphic variables to determine segment membership in logit mixture J.Marketing Res. 31 (1) 128136. models. Hansen,L. P. 1982. Large sample properties of generalized method of moments Econometrica 50 1029 estimators. 1054. Hayes,R. H.,S. C. Wheelwright,K. B. Clark. 1988. Dynamic Manufacturing. The Free Press,New York. Hoffman,D.,A. Pagan. 1989. Post-sample prediction tests for generalized method of moments Oxford Bull.Economics estimators. and Statist. 51 (3) 333343. Iansiti,M.,J. West. 1997. Technology integration: turning great research into great products. 79. Irwin,D. A.,P. J. Klenow. 1994. Learning-by-doing spillovers in the semiconductor industry. J.Political Econ. 102 (6) 1227. Jackson,B. B. 1985. Build customer relationships that last: how close can industrial marketers get to their customersand for how long Harvard Bus.Review 63 120128. ? Jaffe,A. B.,M. Trajtenberg,R. Henderson. 1993. Geographic localization of knowledge spillovers as evidenced by patent citations. Quart.J.Econ. 108 (3) 577598. J.MarJaworski,B. J.,A. K. Kohli. 1993. Market orientation: antecedents and consequences. J.Marketing 57 (3) 5371. J . Kumbhakar,S. C. 1987. The specification of technical and allocative inefficiency in stochastic production and profit frontiers. Econometrics 34 (3) 335348. 1200 Harvard Bus.Review MayJune,69
Cornwell,C.,P. Schmidt,R. C. Sickles. 1990. Production frontiers with cross-sectional and time-series variation in efficiencylevel. J.Econometrics 46 185200. Day,G. 1994. The capabilities organizations. 58 3752. keting of market-driven
Deshpande,R.,J. U. Farley,F. E. Webster,Jr. 1993. Corporate culture, customer orientation,and innovativeness in Japanese firms: a quadrad J.Marketing 57 (1) 2337. analysis.
567
. 1997. Efficiency estimation with heteroskedasticity in a panel data Appl.Econ. 29 379386. model. Lippman,S.,R. P. Rumelt. 1982. Uncertain imitability: an analysis of interfirm differences competition. Econ. 13 418453. in efficiency under Bell J.
Prescott,E.,M. Visscher. 1980. Organizational capital. Econ. 88 446 Rumelt,R. P. 1991. 461. much does industry matter. How agement J. Silberberg,E. 1990. York. 12 (3) 167185. The Structure of Economics.
J.Political Strategic ManMcGrawHill,New
McFadden,D. 1989. A method of simulated moments for estimation of discrete response models without numerical integration. Econometrica 57 (5) 995 1026. Montgomery,C. A.,B. Wernerfelt. 1988. Tobins tance of focus performance. 246251. in firm
Srinivasan,V.,W. S. Lovejoy,D. Beach. 1997. Integrated product dethe 78 (1) sign for marketability and J.Marketing Res. manufacturing. 34 (1) 154 163. Stevenson,R. E. 1980. Likelihood functions for generalized stochastic frontier J.Econometrics 13 5766. estimation. Trajtenberg,M. 1990. A penny for your quotes: patent citations and the value of innovation. Rand J.Econ. 21 (1) 172187. Teece,D.,G. Pisano,A. Shuen. 1997. Dynamic capabilities and strategic Strategic Management J. 18 (7) 50933. management. Wernerfelt,B. 1984. A resource-based view of the StrategicManfirm, agement J. 5 171 180. Birger Wernerfelt.
q and imporAmer.Econ.Review
Padmanabhan,V.,S. Rajiv,K. Srinivasan. 1997. New products,upgrades,and new releases: a rationale for sequential product introduction. J.Marketing Res. 34 (4) 456482. Peteraf,M. A. 1993. The cornerstones of competitive advantage: a resource-based Strategic Management J. 1 (3) 179 view. 191. Porter,M. E. 1980. Competitive Strategy. Free Press,New York.
This paper was received January 13, 1998, and has been with the authors 11 months for 3 revisions; processed by
568

Success in High-Technology Markets: Is Marketing Capability Critical?

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Success in High-Technology Markets: Is Marketing Capability Critical?

Hochgeladen von

Copyright:

Verfügbare Formate

http://books.google.com.pk/books?

Success in High-Technology Markets: Is Marketing Capability Critical?

SUCCESS IN HIGH-TECHNOLOGY MARKETS: IS MARKETING CAPABILITY CRITICAL?

endowments to refer to what we call

Note that resources could physicalassets) or intangible (such as goodwill).

Marketing Science/Vol. 18,No. 4,1999

2.3. Measuring Capabilities The

Advertising Stock,Stock of Marketing Expenditure, Investment in Customer Relationships,Installed Base).

Figure 2 Schematic of the InputOutput Framework

Marketing Science/Vol. 18,No. 4,1999

Marketing Science/Vol. 18,No. 4,1999

as the ratio Tobins sample average

(a) Base of techno-

any particular SIC code.

Marketing Science/Vol. 18,No. 4,1999

for period t k kt c k 1 weight

to continuing number of employees

total . R&D ca-

c represents the attached of sales,general,and

for t period ADEX greater

x ,the the advertising

that exist in our focal industry. SUB_MARKET i k is defined as:

ln ( MARKETINGSTOCK ln( TECHBASE

ln( RECEIVABLE S ln( INSTALLEDBASE

1 if the observation (year 0 else.

also for 1 7 the

Marketing Science/Vol. 18,No. 4,1999

ln( CUM_R & DEXPENSE ln( MKTCAP

g N (0, r gmt i gmt i

OUTPUT ) ln( ) ln(

ln( LABCOST ln( TECH BASE

ln( REL_MKTCAP f19

ln( REL_R&DCAP ln( REL_MKTCAP ln( REL_OPCAP

ln( REL_OPCAP REL_R&DCAP ) ),

Marketing Science/Vol. 18,No. 4,1999

Marketing Science/Vol. 18,No. 4,1999

where (.) and f(X i t , )

U (.) denote the standard normal

Marketing Science/Vol. 18,No. 4,1999

3.4.2. Controlling Heteroskedasticity. Another implicit assumption

Analytical details are appendix,available from the authors upon request.

Marketing Science/Vol. 18,No. 4,1999

Table 1 Parameter Estimates (MSM) of Marketing Capability Model

0. 09 15 (0 .0 71 1) 0. 33 51 ( 0. 102 5)** 0. 27 04 ( 0. 153 6) 0. 05 12 (0 .0 46 2) 0. 07 49 ( 0. 028 3)**

V a m ) 7. 04 92 ( 3. 281 l r( 5)** 1 0. 921 4 (4. 03 58) **

1 .6 294 (0. 38 94) ** 0 .0 021 4 1. 177 (40 ) 2

**: significant at 1% significance level : significant at 5% significance level

is computed exactly like the R

Marketing Science/Vol. 18,No. 4,1999

Table 3 Parameter Estimates (MSM) of Operations Capability Model

V a ri a ble s P opu la t io n A v e ra ge Ef f ect C os t El a s t ic i t ie s : n(OUT l P) UT l n(L AB C OS ) T l n(C A PC OS ) T l n(T EC HB AS ) E l n(M KT CA) P

[ln( TECHBASE )] [ln( CUM_R&DEXPENSE [ln( MKTCAP )] [ln( MKTCAP ) )]

I n ef fi c ie nc y E rror : M ode of I ne f fi c ie nc y ror T e rm 1 8. 231 5 (6 .6 821l Er ) ** V a ri a nc e of I ne f fic i e nc y ror T e rm Er

Error Term] 3.5271 (0.0862)** 3.6483 (0.0715)** r 2 [Variance of Inefficiency

V al o ) 6. 15 42 (4 .2 r( 815 ) 3 .5 118 (1. 00 27)* *

Error Term] 1.0527 (0.0227)** 1.0391 (0.0385)**

Random Shock: r 2 [Variance of Random Shock]

1.2518 (0.0489)** 1.0463 (0.3258)**

1 .8 512 (0. 29 31) ** 0 .0 018 2 1. 001 (16 ) 6

**: significant at 1% significance level : significant at 5% significance level

**: significant at 1% significance level : significant at 5% significance level

Table 4 Parameter Estimates (MSM) of Capabilities-Performance Model

0. 09 15 (0 .0 71 1) 0. 33 51 ( 0. 102 5) 0. 27 04 ( 0. 153 6) 0. 05 12 (0 .0 46 2) 0. 07 49 ( 0. 028 3)

V a m ) 7. 04 92 ( 3. 281 l r( 5) 1 0. 921 4 (4. 03 58)

Error Term] 3.5271 (0.0862) 3.6483 (0.0715) r 2 [Variance of Inefficiency

Error Term] 1.0527 (0.0227) 1.0391 (0.0385)

1.2518 (0.0489) 1.0463 (0.3258)