Strat. Mgmt. J., 31: 39–57 (2010) Published online EarlyView in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.799 Received 26 December 2007; Final revision received 4 July 2009
DOES FIRM PERFORMANCE REVEAL ITS OWN
CAUSES? THE ROLE OF BAYESIAN INFERENCE
YING-CHAN TANG1and FEN-MAY LIOU2*
1Institute of Business and Management, National Chiao Tung University, Taipei,
Taiwan, ROC
2Graduate Institute of Business and Management, Yuanpei University, Hsin Chu,
Taiwan, ROC
A central problem in strategic management is how the inference ‘sustainable competitive advan-tage generates sustainable superior performance’ can be put into practice. In this article we develop a theoretical framework to understand the causal relationships among (1) sustainable competitive advantage, (2) configuration, (3) dynamic capability, and (4) sustainable superior performance. We propose that a firm’s competitive advantage, resource bundle configuration, and dynamic learning capability cannot be comprehended by outsiders. Its operational performance, however, can be captured by financial indicators. We promote an inductive Bayesian interpreta-tion of the sustainable competitive advantage proposiinterpreta-tion. From this viewpoint, the presence or absence of competitive advantage may be reflected in the causal relationship between resource configuration, dynamic capability, and observable financial performance. We apply this theoret-ical framework to an example drawn from the global semiconductor industry, an area in which resource configuration and dynamic capability are essential to performance. The paper con-cludes with a summary of the proposed model and suggestions for future theoretical development of strategic management. Copyright 2009 John Wiley & Sons, Ltd.
INTRODUCTION
Academics have been debating the epistemologi-cal and methodologiepistemologi-cal status of scientific strate-gic management for the last two decades (Astley, 1985; Montgomery, Wernerfelt, and Balakrishnan, 1989; Mir and Watson, 2000; Powell, 2001; Schen-del, 1994; Seth and Zinkhan, 1991). Strategic man-agement is a science—at least in the contexts of industrial organization (Tirole, 1998) and popula-tion ecology (Hannan and Freeman, 1989).1 As
Keywords: sustainable competitive advantage; Bayesian epistemology; du Pont identity
∗Correspondence to: Fen-May Liou, Graduate Institute of
Busi-ness and Management, Yuanpei University, 306 Yuanpei St., Hsin Chu 300, Taiwan, ROC.
E-mail: mayliou@mail.ypu.edu.tw; fmay.liou@msa.hinet.net
1Industrial organization is deduced from economics, as
popu-lation ecology is deduced from biology. Both disciplines have been recognized as ‘close’ to science (Hempel, 1966).
with other sciences, it is equal parts mathematics and logic, empirical evidence and testing. Main-stream research in strategic management (Barney, 1991; Christensen and Raynor, 2003; Grant, 1991; Porter, 1980; 1985; Prahalad and Hamel, 1990) has focused on developing tools capable of prescrib-ing a particular course of action for practitioners. To this end, it deploys inductive logic to infer principles, theoretical claims, and/or ‘takeaway’ from particular cases and other empirical evidence. However, the popularity of this approach does not ensure that the generalizations procured from induction are universally tested or even broadly supported. Consider, for example, Porter’s (1985) competitive strategy:
The fundamental basis of above-average per-formance in the long run is sustainable
com-petitive advantage. Though a firm can have
a myriad of strengths and weaknesses vis-`a-vis its competitors, there are two basic types [sources] of competitive advantage a firm can possess: low cost or differentiation. The significance of any strength or weakness a firm possesses is ultimately a function of its impact on relative cost or differentiation (Porter, 1985 : 11).
Although Porter’s inference that low cost and differentiation are sources of above-average per-formance appears to be deductive in nature, com-petitive advantage is clearly not the whole picture. Scientific inquiry commonly supplements deduc-tion with hypothetical reasoning. Consider the case of China, for example, where brand names are not protected by copyright laws. Should we expect to find evidence that branding plays a role in product differentiation strategies? In this context, is superior performance more likely to derive from a firm’s relationships with govern-ment officials? Is there any other possible source of sustainable competitive advantage? This sim-ple examsim-ple shows that Porter’s generic strate-gies might not reflect genuine practices in the real business world. Perhaps the above quote is not an example of deductive inference but sim-ply a truism, or what philosophers would call a
tautology.
An alternative to Porter’s syllogistic reason-ing2 is the resource-based view (RBV). The RBV
replaces Porter’s generic strategies with general-ized VRIN (valuable, rare, inimitable, and non-substitutable) advantages. That is, ‘valuable and rare organizational resources may be a source of competitive advantage’ (Barney, 1991 : 107). The
2A syllogistic statement consists of three terms: a major premise,
a minor premise, and the conclusion. Porter’s (1985) generic argument can be represented by the following causal path:
Major premise: sustainable competitive advantage leads to above-average performance,
Minor premise: relative cost and differentiation are sources of sustainable competitive advantage,
Conclusion: relative cost and differentiation lead to above-average performance.
The RBV replaces Porter’s cost and differentiation with VRIN resources in the minor premise and promotes this statement to the major premise, thereby forming an incomplete syllogism without minor premise or conclusion.
premise VRIN advantages are sources of
sustain-able competitive advantage makes the RBV a
vir-tual tautology as well. (Or at best an analytic such as the statement a(b+ c) = ab + ac, which requires no empirical evidence outside mathemat-ics and logic.) The assumption that valuable and rare resources are a predictable source of compet-itive advantage is not empirically falsifiable (Bar-ney, 2001; Priem and Butler, 2001a, 2001b). While tautological propositions are often very plausible, they are ultimately vacuous and should never be taken as gospel (Powell, 2001, 2002, 2003).
Barney and his colleagues (Barney, 2001; Bar-ney and Arikan, 2001; Ray, BarBar-ney and Muhanna, 2004), among many others (e.g., Coff, 1999; Hoopes, Madsen, and Walker 2003; Peteraf, 1993), have expended much effort in demarcating tau-tology. Resource-based reasoning is truly real, as companies with superior resources are demonstra-bly more efficient and perform better than other companies. Likewise, critical or rare resources that can generate high value or lower costs are efficient rent-seekers. Business processes that exploit valu-able and rare resources can therefore be a source of sustained competitive advantage. These statements are certainly not mere tautologies.
Powell (2001 : 881) disputed the RBV by proposing the counterfactual condition of
com-petitive disadvantage. As noted by Powell, ‘The
two [competitive advantage and competitive dis-advantage] are quite independent—if competi-tive advantage stems from inimitable, idiosyncratic resources, competitive disadvantage is not merely the non-existence of such resources (which would create economic parity), but rather the failure even to satisfy the minimum success requirements, or “strategic industry factors” (Amit and Schoemaker, 1993), required of any firm’ (Powell, 2001 : 877). To say that a firm has an advantage is to say it has certain resources that other firms do not have. Over time, therefore, one expects the firm to exhibit above-average performance. But there is no guarantee that this must be so—a firm may fail to profit from its competitive advan-tage due to external obstacles. To address this issue, Powell suggested transforming the determin-istic, unidirectional proposition sustainable
com-petitive advantages create sustained superior per-formance into a probabilistic inference: sustain-able competitive advantage is more probsustain-able in firms that have already achieved sustained supe-rior performance. The latter proposition does not
assert that effects (evidence of superior perfor-mance) must follow from causes (sustainable com-petitive advantages). Rather, it asks us to infer the prevalence of causal factors by examining an ensemble of observable effects. By this means, Powell showed that disputes over the deterministic proposition ‘q implies p’ (competitive advantage implies superior performance) are only partially justified.
Note that the definitions of ‘advantage’ and ‘dis-advantage’ have always been relative, and there-fore problematic. ‘If all rivals held the same abso-lute competitive advantage then no relative advan-tage holds and competitive forces would tend to eliminate available rents’ (Arend, 2003 : 280). For example, Toyota’s lean production system is usu-ally considered a kind of ‘corporate DNA,’ very difficult to replicate or transfer and therefore a major source of sustainable competitive advantage (Spear, 2004; Spear and Bowen, 1999). Suppose that all of Toyota’s key competitors successfully implement their own version of the lean action plan (Liker, 2004; Wei, 2007). Would Toyota’s corpo-rate DNA remain a major source of competitive advantage in this situation? Moreover, can Toy-ota’s competitive advantage really be characterized as ‘sustainable’ after competitors have sought out the same rent? This fluidity and indeterminacy of
competitive advantage leads us naturally to
Pow-ell’s (2001) Bayesian epistemology. The hypothe-sis of sustainable competitive advantage (or dis-advantage) can only be confirmed by empirical evidence, but this evidence provides inconclusive support at best. The degree of competitive advan-tage conferred by Toyota’s lean production system ultimately depends on whether Toyota’s competi-tors are successful at implementing or imitating the system, and to what extent Toyota can maintain its advantage.
In this paper, we propose a framework simi-lar to Powell’s (2001) Bayesian process, which periodically updates its propositions or hypothe-ses in the face of empirical evidence. Powell laid out a syllogistic structure describing the relation-ships between competitive advantage, competi-tive disadvantage, and superior performance. The inclusion of competitive disadvantage resolves dis-putes over RBV being a tautology; the issue of competitive advantage’s many and heterogeneous sources, however, has not yet been addressed. In essence, we follow Powell (2001) and Porter
(1991) by defining superior (financial)
perfor-mance as the dependent variable, and adding an auxiliary or bridge hypothesis to help define the
causal relationship between sustainable compet-itive advantage and sustainable superior perfor-mance. Specifically, we argue that it is the firm’s unique configuration of resources (Miller, 1986; Siggelkow, 2002) that mediates between heteroge-neous sources and competitive advantages, creat-ing superior performance.
In the empirical section of this paper, we intro-duce a Bayesian discriminant model to reveal the functional dependence of superior performance on heterogeneous resource bundles. Any primary sources of competitive advantage (a unique busi-ness process such as lean production, customer relationships, etc.) are considered embedded in and inseparable from the organization itself, along with its business units and functional departments. It is assumed that the process of managing these resource bundles, variously termed configuration,
strategic fit (Siggelkow, 2001; Levinthal, 1997),
or causal ambiguity (Reed and DeFillippi, 1990; Rivkin, 2001), cannot be comprehended or imi-tated by outsiders. Realized superior performance indicators such as operating revenue, market share, stock prices, and 10-K reports, however, can be thoroughly assessed by the public.
The model is calibrated against the global semi-conductor industry, a domain in which resource configuration and dynamic capability are essential to performance. The paper concludes with a sum-mary of the proposed model and suggestions for future theoretical development of this approach to strategic management.
BAYESIAN EPISTEMOLOGY ON TAUTOLOGICAL FALLACIES AND COMPETITIVE ADVANTAGE
Ours is an uncertain world, though fortunately all things are not equally uncertain. (Howson and Urbach, 1991 : 371).
Bayesian epistemology and tautological fallacies
The logical fallacy of empirical studies in strategic management is as follows: researchers or man-agers infer the existence of competitive advantage from ex post superior performance, and conclude
that creating competitive advantages ex ante will produce sustainable superior performance (Powell, 2000). This is a circular argument, or rhetorical tautology, because both premise and conclusion are defined in the same manner (Priem and But-ler, 2001a, 2001b; Barney, 2001). To resolve this problem, Durand (2002) suggested invoking the concept of what Mackie (1965) deemed an INUS condition (named for the first letters of the itali-cized words that follow): sustainable competitive advantage is ‘an insufficient but necessary part of a condition [yielding sustained superior per-formance,] which is itself unnecessary but
suffi-cient for the result’ (Mackie, 1965 : 245, italics in
original).
Suppose, for example, that a cigarette butt ‘causes’ a forest fire. The cigarette butt by itself is not sufficient to cause a fire. An inattentive smoker, a pack of dry cigarettes, and a functional naphtha lighter are jointly sufficient for the fire.
prob(q/p)= prob(p/q)× prob(q)
[prob(p/q)× prob(q)] + [prob(p/∼q) × prob(∼q)] =
prob(p/q)× prob(q) prob(p) = (0.50)(0.10) (0.50)(0.10)+ (0.05)(0.90) = 0.05 0.095 = 0.53 (Powell, 2001 : 880) (1)
Furthermore, given the myriad necessary condi-tions associated with the cigarette butt (e.g., the smoker is in the forest, there is no rain that day, the initial fire spot is dry), we can say that the naphtha lighter is an INUS condition—the forest fire would not have happened if the lighter was not present. The complexity of such a series raises problems for the epistemology of causation.
Consider again Porter’s (1985) generic claim that differentiation is a source of sustainable com-petitive advantage, which in turn yields above-average performance. Consider, for example, the following hypothetical case. We do not know whether Alpha Company’s brand or trademark has been pirated in China, or whether its brand man-ager maintains a good relationship with local offi-cials. What we can observe is that Alpha Company has widely recognized branding, differentiating the firm from its competitors, and thus has good odds of generating above-average performance. This is
probabilistic reasoning, and can be used to
evalu-ate hypothetical claims.
To illustrate this sort of reasoning, we follow Powell’s (2001) numerical example of Bayesian analysis by defining events p and q in terms of probabilities as follows (the values are arbitrary): prob(q) = .10 (10% of all firms have sustainable
competitive advantages)
prob(∼q) = .90 (90% of all firms do not have sustainable competitive advantages)
prob(p/q)= .50 (50% of all firms that have sustain-able competitive advantages achieve sustained superior performance)
prob(p/∼q) = .05 (5% of all firms without tainable competitive advantages achieve sus-tained superior performance) (Powell, 2001 : 880).
With known priors prob(q) and prob(∼q) and conditional probabilities prob(p/q) and prob(p/∼q), the evidence probability prob(p) and posterior
probability prob(q/p) can be estimated as
Notice that the universal conditional (i.e., 100% of companies with sustainable competitive advan-tage have achieved sustainable superior perfor-mance) has been factored out. We only claim that 53 percent of firms with evidence of superior performance possess the attribute of competitive advantage. The 53 percent of superior perform-ers in this case are Powell’s ‘quadrant 1’ firms, having positive economic rents which may be monopolistic, Ricardian, or Schumpeterian (Pow-ell, 2001 : 878). This example disproves the deter-ministic claim that q logically entails p —that is, that sustainable competitive advantage always leads to sustained superior performance. The very fact that some companies with a competitive dis-advantage have achieved superior performance3 is
evidence against this claim.
3From probability axioms, we know that if p and ∼q are
independent, then the conditional probability of p given∼q is the same as the probability of p, prob(p/∼q) = prob(p). For the condition prob(p/∼q) >0 to hold (i.e., companies without competitive advantage can still achieve superior performance),
Subtracting the prior prob(q) = 0.10 from the posterior prob(q/p) = 0.53, we find that the p is
evidence for q condition prob(q/p)− prob(q) >0
is satisfied (Powell, 2001 : 879). In Bayesian lan-guage we deduce that sustained superior perfor-mance evidence confirms the probabilistic reason-ing of sustainable competitive advantage, provid-ing a 43 percent increase in probability. It follows that the sustained superior performance evidence disconfirms (or would disconfirm) the likelihoods of sustainable competitive advantage if prob(q/p) − prob(q) < 0. To put it another way, the fact that some firms have sustained superior perfor-mance can be taken as evidence for the hypothesis that competitive advantage is a source of supe-rior performance if and only if the probability that firms possess sustainable competitive advantage is greater among firms that achieved superior perfor-mance than in the overall population.
If we repeat the empirical test using 0.53 as the new prior (holding the likelihoods prob(p/q) and prob(p/∼q) constant), we get prob(q/p) = 0.50 × 0.53/(0.50× 0.53 + 0.05 × 0.47) = 0.9185. The ‘trustworthiness’ of this probabilistic reasoning q
causes p has increased by 38.85 percent. This
represents a refinement of the reasoning in light of new knowledge. If we repeat the empirical refinement one more time with 0.9185 as the new prior, we get prob(q/p) = 0.50 × 0.9185/(0.50 × 0.9185+ 0.05 × 0.0815) = 0.9912. A new round of evidence further confirms the causal role of sus-tainable competitive advantage.
This convergence of probabilistic reasoning, or what Powell calls the merging of ‘sense-making relations’ (Powell, 2003 : 287), depicts an onto-logical belief change in the Bayesian scheme. Even widely discrepant ‘sense-making machiner-ies’ about sustainable competitive advantage will almost surely be driven to a consensus after a sufficiently long period of learning, experiment-ing, and knowledge sharing. This is one way of stating the well-known ‘washing out of priors’ phenomenon in the Bayesian literature (Edwards, Lindman, and Savage, 1963 : 201): people will the following two conditions must be met. (1) p and∼q are independent, so competitive advantage does not cause superior performance and vice versa. The Chinese market mentioned above, where a good relationship with the government can yield superior performance, serves as an example. (2) prob(p) >0, meaning a superior performance outcome is inevitable for some firms without any conditional. This relates to Powell’s (2001 : 878) ‘quadrant 3’ firms, which can achieve superior performance even through wrongdoing or inaction.
eventually assign nearly the same posterior proba-bility to a hypothesis even if they started out with very different priors. In other words, people ratio-nally respond to newly acquired evidence from reality by revising their ontological beliefs (the pri-ors) over time. In the numerical example demon-strated above, evidence supporting the ‘false’ the-ory (i.e., that firms without sustainable competitive advantage have achieved sustained superior per-formance) becomes ‘swamped’ or ‘washed out’ (decreasing from 0.05× 0.90 to 0.05 × 0.47 to 0.05× 0.0815) as the value of the ‘true’ theory increases.
Suppose for a moment that such an equifinality of diverging theories exists. That is, suppose that such probabilistic reasoning provides the whole truth on the causal role of sustainable competi-tive advantages. The end state represents a scenario where the scientific paradigm has greatly advanced (Kuhn, 1962; Lakatos, 1978), or where researchers have arrived at a utopia (Laudan, 1984). The ‘per-fect’ hypothesis would predict the observed evi-dence completely, such that prob(p/q) = 1.4 That
is, whenever we examine a firm with sustain-able competitive advantage we will find it achiev-ing sustained superior performance. Now let us assume an extremely low value for the prior, such as prob(q) = 0.000001, so that very few firms have a sustainable competitive advantage in the first place. Plugging these numbers into Bayes’ equation, we get
prob(q/p)= 1× .000 001
(1× .000 001) + (0 × .999 999) = 1 The above calculation demonstrates that when prob(p/q) = 1 holds, the prior probability doesn’t matter. Whether prob(q) is low or high, we always get a posterior probability of 1. Briefly, almost every new piece of evidence will confirm our theory and the accumulated weight of past data will appear incontrovertible. This is Hempel’s (1945) famous ‘all ravens are black’ paradox: if every sighting of a black raven confirms our theory, so does every sighting of a non-black non-raven.
However, the state of strategic management research is not so bleak. After all, we have not
4prob(p/q)= 1 implies prob(q/p) = 1. We know prob(q) = 1−
prob(∼q). Plugging this into Bayes’ equation and simplifying it with the prob(q) term, we get prob(q/p) = 1. In this scheme, conclusive confirmation implies prob(q) = 1 while conclusive refutation implies prob(q)= 0.
yet studied the entire body of empirical evidence available on all firms and industries. The proposi-tion ‘all ravens are black’ can easily be falsified5
by the discovery of a newborn white raven. By the same token, our probabilistic reasoning can be decisively refuted if the prediction ‘sustainable competitive advantage leads to sustained superior performance’ turns out to be false. If we consider Toyota’s lean production system its sole source of sustainable competitive advantage, then find that all its competitors have successfully implemented lean production systems, Toyota might not be able to generate superior performance. Similarly, Porter’s generic strategy claim would be refuted if the prediction ‘differentiation leads to superior per-formance’ (Porter, 1985 : 120) turns out to be false. Microsoft Office and CISCO ISO serve as counter-factual examples to Porter’s generic strategy: both are highly differentiated products which have trou-ble generating sales in China despite their unique brand names, source codes, and status as industry standards.
While no finite amount of empirical evidence can verify that the competitive advantage reason-ing is truthful, the idea that empirical evidence can easily falsify this reasoning is likewise flawed. As far as theory falsification is concerned, Bayesian epistemology is very sensitive to the pre-evidential prior probability assumed. A slightly different ver-sion of the theory (e.g., ‘a copyrighted brand is the source of sustainable competitive advantage’ or ‘all ravens are brown’) must be treated as a new and distinct theory if it differs in its treatment of the original evidence (‘some brand names are not protected by trademark law in China’ or ‘some
prob(q/Y)= prob(Y/q)× prob(q)
prob(Y/q)× prob(q) + prob(Y/∼q) × prob(∼q) =
prob(Y/q)× prob(q)
prob(Y) (2)
ravens are black’) and the alternative (Porter’s [1985] generic strategy, or Hempel’s [1945] ‘all ravens are black’) can easily be falsified.
5Suppose a white raven is found (a firm has sustainable
com-petitive advantage, but does not achieve superior performance). This will happen if prob(p) = 0. This indicates that the the-ory is independent of the evidence, prob(p/q)= 0. In this case the posterior probability will be zero for any values of prob(q), prob(∼q) and prob(p/∼q).
Thus, according to the arguments laid out above, the falsification of a strategic management theory depends largely on what value is assigned to the pre-evidential prior probability. This number will be different for each proposition or theoretical claim under consideration.
Consider again Powell’s (2001) numerical exam-ple. A scientific analysis of the probabilistic rea-soning will ask why the pre-evidential prior prob-ability prob(q) is set to 0.10. Why not 0.13, 0.50, 0.94, or some other number? The answer is that its initial value depends on the researchers’ subjective choices —in particular, which proxies for compet-itive advantage (unique corporate DNA, differenti-ation, cost leadership, copyrighted brand, etc.) they would like to work with. In strategic management, there may well be an infinite number of variants. Testing every possible proposition is tedious and impractical, but Bayesian inductive logic offers a very elegant way to reconcile them by incorporat-ing empirical observations and test cases.
Suppose the number of sources of sustainable competitive advantage is finite, but covers a wide range of probabilities that superior performance will be generated. We can generalize Powell’s sin-gle event q in equation (1) to the vector q, which represents an exhaustive set of mutually incompat-ible competitive advantage hypotheses or theories, and extend the other scalar p to represent a col-lective set of empirical performance indicators Y. In epistemological terms, this revision promotes Powell’s constant probability values to a higher normative theory of competitive advantage. The Bayesian analysis can be written in the following aggregate form:
The epistemological significance of the Bayesian process is that it provides a dynamic and amor-phous view about the reasoning of competitive advantage. If prob(Y)= 0, then the posterior prob-ability prob(q/Y) in Equation (2) is undefined. The inquiry requires no empirical evidence, meaning the unconditional probability prob(q) is defined on the basis of a tautology or an analytic state-ment. If prob(Y) > 0, however, we can infer and
revise our belief or hypothesis concerning the
pos-sible sources of competitive advantage (q) in light of new evidence (Y). For the present scientific inquiry, we are interested in characterizing the pat-terns of strategic reasoning employed by success-ful firms and providing ‘rational constructions’ of their sustainable competitive advantage given the performance evidence.
Bayesian inference and sustainable competitive advantages
Replying to Priem and Butler’s (2001a : 29) con-tention that ‘unique firms possess competitive advantages’ is tautological, Barney (2001 : 46) recast the problem in terms of strategic equifi-nality6–if equifinality exists, then firms are not
unique and therefore competitive advantages can-not exist. Barney redefined equifinality as a
substi-tutability characteristic: ‘if a resource is valuable,
rare, and costly to imitate, . . . [but] has strategi-cally equivalent substitutes that are themselves not rare or not costly to imitate, then it cannot be a source of sustained competitive advantage’ (Bar-ney, 2001 : 47).
The strategic equifinality debate raises prob-lems for the epistemology of causation among VRIN resources, organizational configuration, and sustainable competitive advantage. For example, can a firm’s distinctive accumulation of valuable and inimitable resources be considered a source of competitive advantage even if the firm has achieved value and/or performance similar to that of its competitors? Or to phrase it differently, can the imitated configuration of resource bundles that generate different value or different performance level be considered sources of competitive advan-tage for the focal firm?
For causal inference and reasoning, including the concept of equifinality requires us to assign auxiliary hypotheses or bridge principles (Hempel, 1966 : 72–75) to mediate between competitive advantage and superior performance. Barney (2001 : 42) would call this parameterizing the causal relation. The auxiliary hypotheses have
6The equifinality thesis originated in General Systems Theory
(von Bertalanffy, 1968), and was adopted in organizational theory to indicate that similar performance outcomes might be achieved by very different typologies, causal paths, or dynamic learning (Doty, Glick, and Huber, 1993; Gresov and Drazin, 1997; Payne, 2006).
been given many names: organizational
configura-tion (Miles and Snow, 1978; Miller and Mintzberg,
1983; Miller, 1986, 1996), dynamic capability (Teece, Pisano, and Shuen 1997; Eisenhardt and Martin; 2000; Zollo and Winter, 2002),
orga-nizational routines (Nelson and Winter, 1982;
Rivkin, 2001; Winter and Szulanski, 2001; Zan-der and Kogut, 1995), and causal (paths) ambiguity (Reed and DeFillippi, 1990; Black and Boal, 1994; Rivkin, 2000, 2001) in the literature on organi-zational management. Their goals, however, are the same: to assert a relationship between ‘ratio-nal constructions’ and the unobserved properties of behavioral theories, and to derive an instance of the relationship based on empirical data that are easier to observe and measure.
In essence, we propose extending the causal relation between competitive advantage and supe-rior performance to a strategy-configuration-performance causal series. At first glance this approach may seem overwhelming, but it can quickly be comprehended by recognizing that cer-tain of the constructs percer-tain to theoretical claims that we have already discussed. For instance, if the ‘configuration’ is left out, we can define focus or
differentiation as a source of competitive strategy
and transform this reasoning series into Porter’s (1985) competitive strategy proposition. Similarly, skipping the ‘superior performance’ result and treating competitive advantage as the dependent variable leads to Barney’s resource-based view (Peteraf and Barney, 2003; Ray et al., 2004).
Given that heterogeneous performance deduc-tively entails different configurations, the poste-rior probability prob(q/Y) in Equation (2) can be extended to the general conditional prob(q, ψ /Y), where ψ is an auxiliary equifinality proposition representing a mixture of heterogeneous resource bundles x and their associated weights l, ψ =
(x, λ). The causal series can be extracted by the Bayesian discriminant model7(Sivia, 1996), which
assumes that the population of firms is composed of two unaffiliated factions: those with compet-itive advantage and those without (i.e., having
7Using the probability product rule, Equation (3B) can be
rewritten as prob(q, ψ/Y)= prob(ψ/q, Y)× prob(q/Y). Set-ting the right-hand side of Equation (3B) equal to this new expression yields prob(q/ψ, Y)× prob(ψ/Y) = prob(ψ/q, Y)× prob(q/Y). Rearranging the left-hand side gives Bayes’ theorem, prob(q/ψ, Y)=prob(ψ/q, Y)prob(ψ/Y)× prob(q/Y), a generalized form of Powell’s (1985) numerical example.
competitive disadvantage).
prob(q/Y)+ prob(∼q/Y) = 1 (3A) prob(q, ψ/Y)= prob(q/ψ, Y)
× prob(ψ/Y) (3B)
The probabilities of the competitive advantage hypotheses q are straightforward. Statistical infer-ence of competitive advantages (and competi-tive disadvantages) comes from induccompeti-tive rea-soning based on the unobserved configurations
of heterogeneous resource bundles ψ and the empirical evidence of superior performance Y. Bayesian reasoning generates one possible ‘ratio-nal construction’ of sustainable competitive advan-tage, which is depicted in Figure 1.
For instance, when trying to determine sources of competitive advantage that in turn cause supe-rior financial performance, we may need to con-sider quite a long list. The result of testing these alternative hypotheses might depend on several of the researcher’s choices: (1) how the evidential outcome Y is assessed; (2) the means by which
Sustainable competitive advantage
Sustainable superior performance
Adv- advertising expenses; AR- accounts receivable; AP- accounts payable; CGS- cost of sales; Dep- depreciation and amortization; FA- fixed assets; IC- invested capital; NOPLAT- net profit less adjusted tax; R&D- research and development expenses; ROIC- return on invested capital; SG&A- selling, general, and administration expenses; Tax- corporate income tax.
FA IC AP Cash Inventory AR R&D NOPLAT Tax Sales Adv SG&A Dep Sales Operating efficiency Capital leverage CGS (−) (−) Intellectual property R&D/sales SG&A/sales Supplier relationship AP turnover Inventory turnover CGS/sales Fixed asset management Fixed asset turnover Depreciation/ sales Customer relationship Adv/sales AR turnover Configurations
Du Pont identity: ROIC = IC NOPLAT = S NOPLAT × IC S Dynamic capability Resources bundles
the outcome will be performed; (3) the various paths, devices, or linkages leading to the evidential outcome; and (4) alignments of elements that lead to sustainable competitive advantage. In our case, (1) the return on invested capital (ROIC) is used to measure a firm’s sustainable superior perfor-mance and/or value creation (Appendix I); (2) the firm’s resource bundles x, such as advertising and accounts receivable, are treated as driving elements of ROIC; (3) the configuration weights l repre-sent dynamic linkages such as operating efficiency and capital leverage that interconnect resource bundles; and (4) configuration of interconnected resource bundles such as customer relationships, intellectual property, and fixed asset management that might lead to sustainable competitive advan-tage.
As Durand (2002) suggests, competitive advan-tage may be a necessary but insufficient condition for performance outcome. If competitive advan-tage is about value creation and value capturing (Porter, 1980; Brandenburger and Stuart, 1996; Lippman and Rumelt, 2003; Coff, 1999), then we need an acting agency (in our case, the organiza-tion) to be responsible for the outcomes. We argue in Figure 1 that a firm’s configuration, which we consider the causally relevant condition, is a nec-essary element among a set of conditions (includ-ing resource bundles, organizational routines, and operating efficiency) that are jointly sufficient for the superior performance outcome.
The problem of identifying the causally relevant conditions of an observed outcome, or in epis-temological terms the ‘cause-in-evidence,’ is not new in the RBV literature. Barney (1997) makes a similar point by changing his VRIN model into a VRIO model, with organization replacing non-substitutable resources. Peteraf and Reed (2007) take a similar approach by introducing the man-ager, an actor responsible for bringing the con-figuration into efficient alignment (or ‘fit’). The main difference of Bayesian epistemology is that the causal role of organization is probabilistic. Superior performance thus may still occur in the absence of effective organization, or fail to occur in its presence. Organization is considered a cause if (and only if) it significantly increases the prob-ability of ensuing superior performance.
We now turn the discussion to an empirical test of this Bayesian inference model on sustainable competitive advantage.
EMPIRICAL STUDY: THE SEMICONCUCTOR INDUSTRY
In this section we address the concept of sus-tainable competitive advantage in the worldwide semiconductor industry. The semiconductor mar-ket experienced both downward and upward cycles from 2000–2005 (Semiconductor International, 2005). This is an interesting period to study because many semiconductor manufacturers com-menced operations in 2000 but then had to face industry-wide problems such as new product tran-sitions, design patent protection, production over-capacity, price erosion, shorter technology life cycles, global supply chain problems, and other logistical issues. The fierce battleground that resulted is ideal for testing whether superior per-formance can be traced to sustainable competitive advantages.
There are 208 semiconductor companies (with Standard Industrial Classification code 3674) in our sample, contributing a total of 1,248 records to Standard and Poor’s COMPUSTAT database from 2000 to 2005. Sixty-one companies were excluded from our dataset. Ten of these provided fewer than three years of data, while 41 lacked data on vari-ous expenditure components (research and devel-opment [R&D], selling, general, and administra-tive expenses [SG&A], cost of goods sold [CGS], depreciation [Dep.], and Tax). In addition, compa-nies were excluded if any of their financial indica-tors (excluding ROIC) were outliers8by more than
three standard deviations from the industry mean. This criterion identified 10 more companies, five with positive ROIC and five with negative ROIC. The final dataset contains 147 companies and 786 firm-year observations. Of these, 118 compa-nies are located in developed countries (the United States, within Europe, and Japan). The other 29 are in the Asia/Pacific region.
8Hansen, Perry, and Reese (2004) argue that the central problem
with the RBV is that it relies on extraordinary performers (positive outliers) instead of averages. Their argument, however, confounds the relationship between exceptional performers and statistical outliers. A statistical outlier is defined as an extremely unlikely data point on either side of a stochastic distribution. The exceptional performer may appear similar, but its appearance is not random—it is determined by the ‘rational construction’ of the firm. Hansen et al.’s (2004) study selected firms that had recently appointed CEOs for empirical calibration—a procedure that screens out all statistical outliers to begin with. Thus, it is similar to RBV in that it relies on exceptional performers.
Table 1. Principal component analysis of financial indicators and the resulting resource configurations
Financial Resource configuration
indicators Factor1: Relationship advantage Factor2: Management ability Factor3: Knowledge management
Accounts receivable turnover 0.578 −0.085 0.338
CGS/sales −0.677 −0.204 −0.417
Inventory turnover 0.595 0.053 −0.033
Accounts payable turnover 0.684 0.008 0.043
R&D/sales 0.238 0.046 0.859
SG&A/sales −0.063 −0.184 0.812
Depreciation/sales 0.034 0.870 0.014
Tax/sales 0.568 −0.229 −0.379
Fixed assets turnover 0.017 −0.793 0.101
Eigen value 2.36 1.56 1.45
Accumulated variance (%) 0.26 0.43 0.60
Bold numbers indicate a high correlation between the common factor and the corresponding financial indicator (greater than 0.5).
Unraveling sustainable competitive advantages As discussed above, it might not be possible to directly observe a firm’s sustainable competitive advantage or its efficient alignments responsible for the same. Certain effective configurations of observable traits, however, can be inferred from the firms’ financial performance data. To begin with, principle component analysis (PCA)9 was
conducted on the financial indicators10 to
iden-tify these configurations. After applying a varimax rotation to the eigenvectors and retaining those with eigenvalues greater than one, we obtained three principal components that together account for 60 percent of the total variance. Table 1 shows these three components or factors (represented as configurations of financial indicators) and the loadings associated with each variable. Significant loadings (0.55 and above) are printed in boldface. In Factor 1, all significant financial indicators are related to relationship management. This factor includes customer relationship management (accounts receivable turnover), three variables related to supplier relationship management
9The purpose of principal component analysis is to identify the
most parsimonious groupings or configurations of variables that account for observable performance. It is based on the linear equation c= la + ee, where c contains observable financial indicators, a is the ‘latent structure’ of the strategic configura-tions, and l are the factor loadings connecting financial indica-tors and resource configurations. The bridge hypothesis eeis the ‘causal ambiguity’ projection, and encapsulates the maximum explainable variation in the relationship.
10Advertising expenditures are not included here due to data
constraints.
(accounts payable turnover, inventory turnover, and CGS/sales) and one variable associated with the government (tax to sales ratio). Thus, this factor illustrates the sustainable competitive advantage of firms that skillfully manage their upstream (suppliers), downstream (customers), and governmental relationships. There is also a negative correlation between CGS/sales and Factor 1 (−0.677), indicating that good relationship management can pay off with respect to a lower CGS. The semiconductor/IC industry has developed several partitions over the years, with firms dealing in intellectual property (NXP and IBM), integrated circuit (IC) design (Qualcomm and NVIDIA), wafer foundry (Taiwan Semiconductor Manufacturing Co. [TSMC]), and IC assembly (Advanced Semiconductor Engineering). The form of Factor 1 indicates that all these firms are highly interdependent—each has to ally with both upstream and downstream members of the industry.
Factor 2 consists of indicators related to a firm’s fixed asset managing capability, including Dep./sales ratio and fixed assets turnover. The neg-ative correlation between fixed assets turnover and Factor 2 (−0.793) indicates that firms exhibiting greater competence in assets management gener-ate revenue at a lower unit historical cost. It is imperative in the semiconductor industry that firms fully utilize their fixed assets in a short period of time. The high correlation between Dep./sales and fixed asset management capability (0.870) reveals another unique feature of this capital- and
equipment-intensive industry: that effective asset management is associated with low asset depre-ciation. This result underlines the importance of ‘light’ asset operation in the semiconductor indus-try.
Factor 3 consists of indicators related to
knowledge management, including R&D/sales and SG&A/sales. Both ratios measure a firm’s effectiveness in resource deployment. The high correlations between Factor 3 and R&D/sales (0.859) and SG&A/sales (0.812) indicate that lower unit costs are associated with efficient management.
Principal component analysis thus confirms our proposition that the resource configurations and management capabilities of firms can be inferred from their observable financial indicators. We will examine the reliability and validity of this infer-ence in the following section.
Segregating competitive advantage and competitive disadvantage
To infer sustainable competitive advantage, it is necessary to investigate sources of competitive advantage and the valuation of sustained supe-rior performance on a deeper level. We follow Porter (1985), Hunt (2002), and Priem and But-ler (2001b) in defining competitively advantaged firms as those whose financial performance is supe-rior to the industry average. Companies with a high ROIC typically attract competition, so this ratio is taken as the appropriate indicator of financial per-formance. Furthermore, companies that have built up a sustained competitive advantage should gen-erate a consistent or increasing ROIC over a long period of time. As Spanos and Lioukas (2001) have noted: ‘The time period . . . is admittedly short (i.e., previous three years) to account for any business cycle effects or transient problems. It is impor-tant to note, however, that a longer time-frame (e.g., five instead of three years) could endanger the reliability of responses’ (Spanos and Lioukas, 2001 : 923). Thus, only firms having a three-year average ROIC above the industrial level are con-sidered to have observable superior performance.
Out of 147 firms, 138 provided information on all 11 of the financial indicators we require. Appendix 2 provides some descriptive statistics of the sample companies. The ROIC ratios of individual firms range from −43 percent to 45 percent, with an average of 4 percent. Their assets
range from US$9 million to US$47,867 million (Intel). Table 2 ranks the top 15 semiconductor firms in terms of ROIC, and lists their resource-related financial ratios during 2003–2005.
The IC design houses Novatek Microelectronics (Taiwan), Mtekvision (Korea), and Memc Elec-tronic Matrials (United States) command the high-est ROICs in the industry. Two indicators confirm the existence of sustainable competitive advan-tage in the top two companies: (1) they have both high fixed assets turnover and low accounts payable turnover, indicating an ability to par-lay their unique technologies into cost-effective design and manufacturing processes; and (2) their SG&A and R&D expenditures are low relative to sales, indicating effective knowledge management. Mtekvision also has a high accounts receivable turnover ratio (9.10), evidence of a strong relation-ship with its customers (manufacturers of mobile phones, smart phones, PDAs, digital cameras, MP3 players, and voice recorder products). In contrast, Memc (a global leader in the manufacture and sales of wafers and related intermediate products to the semiconductor and solar industries) has a low fixed assets turnover ratio (2.48). On the other hand, it has very effective knowledge management and strong customer relationships. It is noteworthy that all three high-return companies operate on a rather small scale in terms of total assets, compared to the industrial average.
Two global leaders, Intel and TSMC, command the highest Dep. to sales ratios. They also have relatively low fixed assets turnover ratios due to their ‘heavy’ asset investments. Among the top 15 firms, Intel, the personal computer central process-ing unit leader, commands the highest R&D/sales (0.14) and SG&A/Sales ratios (0.14). TSMC, the pure-play foundry business leader, has the high-est Dep/sales ratio (0.278). Thus, neither company generated sustained competitive advantage from asset and/or knowledge management; competitive advantage instead derived from their management of supplier relationships. The CGS-to-sales ratio is very low in both firms, yielding high gross margins capable of subsidizing their high R&D and SG&A expenses. Furthermore, high inventory turnover compensates for low fixed asset turnover in both companies. The sustainable competitive advantages of these two companies, which have quite different configurations, are not based upon a single source, but rather an amalgamation of sources.
T able 2 . T he top 1 5 semiconductor fi rms during 2003 – 2005, ranked b y R OIC Company A rea R OIC T A A R T CGS/S A PT INVT R&D/S S G&A/S D ep/S F A T T ax/S Novatek M icroelectronics T aiwan 0.45 426 4. 19 0.70 6 .08 11 .60 0 .05 0 .04 0 .005 31.08 0 .000 M tekvision Korea 0 .42 9 8 9 .10 0 .66 8 .75 8 .60 0 .07 0 .02 0 .004 20.93 0.003 Memc Electronic M atrials U SA 0.38 962 7.91 0.63 9 .00 8 .14 0.04 0.07 0.044 2.48 0 .002 Sitronix T echnology T aiwan 0.37 54 4.82 0.65 11 .78 8 .89 0 .09 0 .05 0 .004 20.35 0 .014 Melexis N v U SA 0.31 151 4.91 0.52 29 .82 5 .64 0 .14 0 .07 0 .074 4.20 0 .030 Intel U SA 0.23 47,867 10.50 0.28 17 .6 8 1 2 .48 0.14 0.14 0.135 2.08 0 .082 National S emiconductor U SA 0.22 2,432 11.92 0. 36 21 .16 1 0 .85 0.17 0.14 0.093 3.14 0 .045 Hynix S emicondutor Korea 0 .22 8 ,698 6.04 0.44 11 .77 1 0 .04 0.06 0.08 0.213 1.12 − 0.0 2 5 On Semiconductor U SA 0.22 1,140 8.45 0.60 10 .20 6 .74 0 .08 0 .12 0 .089 2.57 0 .005 Xilinx Inc USA 0 .20 3 ,050 6.80 0.34 22 .38 1 0 .23 0.19 0.18 0.035 4 .52 0 .050 Samsung T echwin K orea 0.20 1,645 7.32 0.80 13 .77 8 .98 0 .03 0 .07 0 .041 3.66 − 0 .000 T aiwan Semiconductor M fg Co. T aiwan 0 .19 14,354 7.93 0.29 27 .51 1 6 .10 0.05 0.05 0.278 1.01 0 .007 Silicon Laboratories Inc USA 0 .19 492 7.61 0. 41 9 .73 13 .28 0 .19 0 .14 0 .042 11.91 0 .061 Diodes Inc USA 0 .19 194 4.22 0.61 8 .23 8 .52 0.02 0.14 0.068 3.01 0 .028 E2V T echnologies Plc U K 0 .18 162 4.31 0. 61 9 .37 5 .06 0.04 0.17 0.052 4.64 0 .022 Industrial average 0.04 2,243 6.27 0.56 12 .55 8 .53 0 .14 0 .15 0 .09 5 .69 0 .019 ROIC: return o n invested capital; AR T: accounts receivable turnover ratio; C GS: C ost o f goods sold; S : annual sales; A PT: accounts p ayable turnover ratio; INVT : inventor y tur nover ratio; S G& A: selling, general and adm inistration expenditure; F A T : fi xe d assets turnover ratio; and T A : total assets in m illion U S dollars.
Table 3. Discriminant analysis on advantaged and disadvantaged firms Standardized canonical coefficients F value Prob > F ART −0.0517 1.08 0.3014 CGS/S −0.2782 0.15 0.6958 APT −0.1720 0.66 0.4184 INVT 0.1710 3.74 0.0552∗ R&D/S −0.2515 23.85 <0.0001∗∗∗ SG&A/S −0.2504 37.89 <0.0001∗∗∗ Dep/S −0.3525 2.01 0.1587 FAT −0.1317 0.14 0.7136 Tax/S 0.2282 19.35 <0.0001∗∗∗ ∗p <0.10;∗∗p <0.05;∗∗∗p <0.01.
Eigenvalue Canonical correlation Likelihood ratio F value Prob > F
1.1121 0.725628 0.47346 15.82 <0.0001
Classification results used for cross-validationa
Groups Competitive advantage Competitive disadvantage Total
Competitive advantage 64 13 77
Competitive disadvantage 15 46 61
aCross-validation is done by recalculating the discriminant function for all firms other than the validated firm. b88.4% ((70+ 52)/138) of firms are correctly classified.
c79.7% ((64+ 46)/138) of the cross-validated firms remain correctly classified.
Discriminant function analysis11 (DFA) is
applied to identify the underlying resource con-figurations that best distinguish the 138 firms, all of which are classified as having either competi-tive advantage or competicompeti-tive disadvantage by the three-year ROIC criterion mentioned above. DFA computes the posterior probability prob(q/ψ, Y) (cross-validated hit ratio) that financial indica-tors are associated with the competitive advantage and competitive disadvantage groups, given group-specific density estimates prob(ψ/q, Y)(the canon-ical coefficients in Table 3) and unconditional den-sity estimates prob(ψ/Y) (the prior probability is set to 58% initially, since 80 of the 138 firms have three-year ROICs above the industry aver-age). Table 3 presents the results of our two-group
11Discriminant function analysis (DFA) is an empirical
ver-sion of Bayes’ theorem that transforms the prior probabilities of the advantaged and disadvantaged groupsprob(q)= prob(θ, ∼θ) into posterior group memberships. The performance like-lihood function (based on financial indicators) must be known beforehand. The ‘hit ratio,’ or Powell’s (2003 : 287) merging of ‘sense-making relations,’ measures the accuracy of the prior and posterior membership transformations.
discriminant analysis. An examination of the group means shows immediately that ROIC discrimi-nates the groups more effectively than any other indicator. In addition, SG&A/sales (S), R&D/S, and Tax/S all demonstrate significant (p < 0.01) power to separate the two groups. Table 3 also presents the classification accuracy of the discrim-inant function. Our results show that 90.9 percent of competitive-advantage firms and 85.3 percent of competitive-disadvantage firms are correctly clas-sified, for an overall accuracy of 88.4 percent (>75%). Leave-one-out cross-validation correctly classifies 79.7 percent of firms (>58%). Evidently, financial resource bundles (Penrose, 1959; Rumelt, 1984) can be used to distinguish between competitive-advantage and competitive-dis-advantage groups, given some knowledge of their configurations.
DISCUSSION
Much of past RBV research has focused on resource bundles, groups of indicators sharing
common themes such as culture (Barney, 1986), organizational routines, management skills, and socially complex resources (Barney, 1991, 1997). Similarly, modern research on fit and organiza-tional alignment (Porter, 1996; Siggelkow, 2001, 2002), including work based on the NK model (epistatic interactions among N organizational attributes in a K-dimensional fitness space) (Kauff-man, 1993; Levinthal, 1997), really considers these bundles the source of competitive advantage. The problem is that resource and activity bundles are notoriously hard to dismantle, since they include complex linkages, complementarities (Milgrom and Roberts, 1990, 1995), and tacit dimensions (Nelson and Winter, 1982; Reed and DeFillippi, 1990). This is a particularly serious problem for variants of the RBV, such as the knowledge-based view (Kogut and Zander, 1992; Liebeskind, 1996) and the works on dynamic capabilities (Teece
et al., 1997; Eisenhardt and Martin, 2000). Not
only are the bundles hard to unpack in these mod-els, but it becomes very difficult to test the pre-diction that such bundles provide a competitive advantage or affect long-run (or even short-run) performance.
There have been some efforts in this direction. Before the RBV made much headway, there was a stream of research connecting broad archetypes or configurations (Miles and Snow, 1984; Miller, 1986, 1996) to performance. Since Milgrom and Roberts (1990, 1995), there has also been some empirical work suggesting that complementari-ties do provide efficiency advantages (e.g., Ich-niowski, Shaw, and Prennushi, 1997). Peteraf and Reed (2007) took this analysis to a higher level, examining how bundles of capabilities were man-aged and showing that proper management can lower firm costs, thereby creating value. Connect-ing bundles to competitive advantage and sustain-able performance is a longstanding and impor-tant problem. Advancing the RBV, the dynamic RBV (Helfat and Peteraf, 2003), and research on dynamic capabilities (Teece et al., 1997) depends critically upon achieving this milestone. Similarly, advancing Porter’s (1996) claim that ‘fit’ leads to sustainable competitive advantage depends on our ability to unfold the strategy-configuration-performance causal dependencies.
The resource configurations based on financial indicators introduced in this paper provide an effective way to value and predict the existence of competitive advantage within an industry. All the
financial indicators of Figure 1 are related to value creation, either on the willingness-to-pay side or on the cost side (Hoopes et al., 2003). This approach has been shown to work well in several Harvard Business Cases and teaching notes (e.g., Gilson and Cott, 1997; Ghemawat, 2004; Ghemawat and Nueno, 2006; Rivkin and Porter, 1999). The DFA analysis (with its underlying Bayesian understand-ing) provides prima facie evidence that companies with a track record of sustainable profitability (not just a lucky year) are more likely to have a com-petitive advantage in terms of value. The PCA analysis reveals causal linkages among resource bundles, efficient alignments, and dynamic capa-bilities that indicate that competitive advantage causes superior performance. By combining these calibration tools, we can find out which potential routes to competitive advantage yield long-term payoffs in performance and profitability given a specific context, and which resource bundles really matter.
CONCLUSIONS
To resolve disputes over the resource-based tautol-ogy, Powell (2001) suggested adopting Bayesian probabilistic reasoning as a means of distinguish-ing sustained competitive advantage from sus-tained superior performance. This paper advances Powell’s idea by proposing that particular resource configurations mediate between the two. Through a discussion of Bayes’ theorem, we describe how empirical data on past financial performance in a population of firms can be used to generate the posterior probability of sustainable competi-tive advantage, given the prior probabilities of both competitive advantage and competitive disadvan-tage. The financial drivers of the du Pont identity are taken as a basis to derive relevant configura-tions of resource bundles.
Three configurations of ‘resource bundles’ were identified in an example drawn from the semicon-ductor industry: upstream and downstream rela-tionship management, management of intellectual property, and fixed asset management. We con-clude that superior financial performance arises from a firm’s unique resource configuration and management capability. Since financial data is easy to access, this theoretical framework is very useful for investigating the competitive advantage propo-sition in a systematic and extensive manner.
Taking the process a step further, we can bring other midrange theories such as strategic archetypes (Miles and Snow, 1978; Miller and Friesen, 1978; Hambrick, 1984), causal ambiguity (Lippman and Rumelt, 1982; Reed and DeFillippi, 1990), strategic equifinality (Payne, 2006), and contingency theory as conditional statements or auxiliary hypotheses to the competitive advantage and superior performance dyad. These possibilities certainly merit further investigation, and form a promising area for future research.
ACKNOWLEDGEMENTS
We thank the two anonymous referees for their thoughtful and constructive comments, which have substantially improved the exposition of the arti-cle. This research is financially supported by the National Science Council of the Republic of China (NSC 96-2416-H-264-008-MY2 and NSC 98-2410-H-009-029)
.
REFERENCES
Amit R, Schoemaker PJ. 1993. Strategic assets and organizational rent. Strategic Management Journal 14(1): 33 – 46.
Arend RJ. 2003. Revisiting the logical and research considerations of competitive advantage. Strategic
Management Journal 24(3): 279 – 284.
Astley WG. 1985. Administrative science as socially constructed truth. Administrative Science Quarterly 30(4): 497 – 513.
Barney JB. 1986. Organizational culture: can it be a source of sustained competitive advantage. Academy
of Management Review 11(3): 656 – 665.
Barney JB. 1991. Firm resources and sustained com-petitive advantage. Journal of Management 17(1): 99 – 120.
Barney JB. 1997. Organization economics: understanding the relationship between organizations and economic analysis. In Handbook of Organization Studies, Clegg S, Hardy C, Nord W (eds). Sage Publishers: London, UK; 115 – 147.
Barney JB. 2001. Is the resource-based ‘view’ a useful perspective for strategic management research? Yes.
Academy of Management Review 26(1): 41 – 56.
Barney JB, Arikan AM. 2001. The resource-based view: origins and implications. In The Blackwell Handbook
of Strategic Management , Hitt MA, Freeman RE,
Jeffrey S (eds). Blackwell Business: Oxford, UK; 124 – 188.
Besanko D, Dranove D, Shanley M, Schaefer S. 2007.
Economics of Strategy (4th edn). John Wiley & Sons:
Hoboken, NJ.
Black JA, Boal KB. 1994. Strategic resources: traits, configurations and paths to sustainable competitive advantage. Strategic Management Journal , Summer Special Issue 15: 131 – 148.
Brandenburger AM, Stuart HW. 1996. Value-based busi-ness strategy. Journal of Economics and Management
Strategy 5(1): 5 – 24.
Cao B, Jiang B, Koller T. 2006. Balancing ROIC and growth to build value. McKinsey on Finance 19: (Spring): 12 – 16.
Christensen CM, Raynor ME. 2003. The Innovator’s
Solution: Creating and Sustaining Successful Growth.
Harvard Business School Press: Boston, MA. Coff RW. 1999. When competitive advantage doesn’t
lead to performance: the resource-based view and stakeholder bargaining power. Organization Science 10(2): 119 – 133.
Doty DH, Glick WH, Huber GP. 1993. Fit, equifinality, and organizational effectiveness: a test of two configurational theories. Academy of Management
Journal 36(6): 1196 – 1250.
Durand R. 2002. Competitive advantages exist: a critique of Powell. Strategic Management Journal 23(9): 867 – 872.
Edwards W, Lindman H, Savage LJ. 1963. Bayesian statistical inference for psychological research.
Psychological Review 70(3): 193 – 242.
Eisenhardt KM, Martin JA. 2000. Dynamic capabilities: what are they? Strategic Management Journal , October– November Special Issue 21: 1105 – 1121. Firer C. 1999. Driving financial performance through the
du Pont identity: a strategic use of financial analysis and planning. Financial Practice & Education 9(1): 34 – 45.
Ghemawat P. 2004. ZARA: fast fashion. Harvard
Business School Teaching Note 703-496, Harvard Business School: Boston, MA.
Ghemawat P, Nueno JL. 2006. ZARA: fast fashion. Harvard Business School Case 703-479, Harvard Business School: Boston, MA.
Gilson SC, Cott J. 1997. Scott Paper Company. Harvard Business School Case 296 – 048, Harvard Business School: Boston, MA.
Grant RM. 1991. The resource-based theory of competi-tive advantage: implications for strategy formulation.
California Management Review 33(3): 119 – 135.
Grant RM. 2008. Cases to Accompany Contemporary
Strategy Analysis. Blackwell: Malden, MA.
Gresov C, Drazin R. 1997. Equifinality: functional equivalence in organization design. Academy of
Management Review 22(2): 403 – 428.
Hambrick DC. 1984. Taxonomic approaches to studying strategy. Journal of Management 10(1): 27 – 41. Hannan MT, Freeman J. 1989. Organizational Ecology.
Harvard University Press: Cambridge, MA.
Hansen MH, Perry LT, Reese CS. 2004. A Bayesian operationalization of the resource-based view.
Strate-gic Management Journal 25(13): 1279 – 1295.
Helfat CE, Peteraf MA. 2003. The dynamic resource-based view: capability lifecycles. Strategic
Hempel CG. 1945. Studies in the logic of confirmation (l and ll). Mind 54(213): 1 – 26, 97 – 121. Reprinted in Hempel’s 1965 Aspects of Scientific Explanation and
Other Essays in the Philosophy of Science, Free Press:
New York; 3 – 46.
Hempel CG. 1966. Philosophy of Natural Science. Prentice-Hall: Englewood Cliffs, NJ.
Hoopes DG, Madsen TL, Walker G. 2003. Guest editors’ introduction to the special issue: why is there
a resource-based view? Toward a theory of
competitive heterogeneity. Strategic Management Journal , October Special Issue 24: 889 – 902.
Howson C, Urbach P. 1991. Bayesian reasoning in science. Nature 350: 371 – 374.
Hunt S. 2002. Foundations of Marketing Theory: Toward
a General Theory of Marketing . M.E. Sharp: New
York.
Ichniowski C, Shaw K, Prennushi G. 1997. The effects of human resource management practices on productiv-ity: a study of steel finishing line. American Economic
Review 87(3): 291 – 313.
Kauffman SA. 1993. The Origins of Order . Oxford: New York.
Kogut B, Zander U. 1992. Knowledge of the firm, combinative capabilities, and the replication of knowledge. Organization Science 3(3): 383 – 397. Koller T, Goedhart M, Wessels D. 2005. Valuation:
Measuring and Managing the Value of Companies (4th
edn). Wiley: New York.
Kuhn TS. 1962. The Structure of Scientific Revolutions. University of Chicago Press: Chicago, IL.
Lakatos I. 1978. Mathematics, Science and Epistemology:
Philosophical Papers (Volume 2). Cambridge
Univer-sity Press: Cambridge, MA.
Laudan L. 1984. Science and Values: The Aims of Science
and Their Role in Scientific Debate. University of
California Press: Berkeley, CA.
Levinthal D. 1997. Adaptation on landscapes.
Manage-ment Science 43: 934 – 950.
Liebeskind JP. 1996. Knowledge, strategy, and the theory of the firm. Strategic Management Journal , Winter Special Issue 17: 93 – 107.
Liker JK. 2004. The Toyota Way: 14 Management
Principles from the World’s Greatest Manufacturer .
McGraw-Hill Professional: Columbus, OH.
Lippman SA, Rumelt RP. 1982. Uncertain imitability: an analysis of interfirm differences in efficiency under competition. Bell Journal of Economics 13(2): 418 – 438.
Lippman SA, Rumelt RP. 2003. A bargaining perspective on resource advantage. Strategic Management Journal 24(11): 1069 – 1086.
Mackie JL. 1965. Causes and conditions. American
Philosophical Quarterly 2(4): 245 – 264.
Miles RE, Snow CC. 1978. Organizational Strategy,
Structure and Process. McGraw Hill: New York.
Miles RE, Snow CC. 1984. Fit, failure and the hall of fame. California Management Review 26(3): 10 – 28. Milgrom P, Roberts J. 1990. The economics of modern
manufacturing: technology, strategy and organization.
American Economic Review 80(3): 511 – 528.
Milgrom P, Roberts J. 1995. Complementarities and fit strategy, structure, and organizational change in manufacturing. Journal of Accounting and Economics 19(2 – 3): 179 – 208.
Miller D. 1986. Configurations of strategy and structure: towards a synthesis. Strategic Management Journal 7(3): 233 – 249.
Miller D. 1996. Configurations revisited. Strategic Management Journal 17(7): 505 – 512.
Miller D, Friesen PH. 1978. Archetypes of strategy formulation. Management Science 24(9): 921 – 933. Miller D, Mintzberg H. 1983. The case for configuration.
In Beyond Method: Strategies for Social Research, Morgan G (ed). Sage: Beverly Hills, CA; 57 – 73. Mir R, Watson A. 2000. Strategic management and the
philosophy of science: the case for a constructivist methodology. Strategic Management Journal 21(9): 941 – 953.
Montgomery CA, Wernerfelt B, Balakrishnan S. 1989. Strategy content and the research process: a critique and commentary. Strategic Management Journal 10(2): 189 – 197.
Nelson RR, Winter SG. 1982. An Evolutionary Theory
of Economic Change. Harvard University Press:
Cambridge, MA.
Payne TG. 2006. Examining configurations and firm performance in a suboptimal equifinality context.
Organization Science 17(6): 756 – 770.
Penrose E. 1959. Theory of the Growth of the Firm. Wiley: New York.
Peteraf MA. 1993. The cornerstones of competitive advantage: a resource-based view. Strategic
Manage-ment Journal 14(3): 179 – 191.
Peteraf MA, Barney JB. 2003. Unraveling the resource-based tangle. Managerial and Decision Economics 24(4): 309 – 323.
Peteraf MA, Reed R. 2007. Managerial discretion and internal alignment under regulatory constraints and change. Strategic Management Journal 28(11): 1089 – 1112.
Porter ME. 1980. Competitive Strategy Techniques for
Analyzing Industries and Competitors. Free Press:
New York.
Porter ME. 1985. Competitive Advantage: Creating and
Sustaining Superior Performance. Free Press: New
York.
Porter ME. 1991. Towards a dynamic theory of strategy.
Strategic Management Journal , Winter Special Issue 12: 95 – 117.
Porter ME. 1996. What is strategy? Harvard Business
Review 74(6): 61 – 78.
Powell TC. 2000. Falibilism and organizational research: the third epistemology. Journal of Management
Research 4: 201 – 219.
Powell TC. 2001. Complete advantage: logical and philosophical considerations. Strategic Management
Journal 22(9): 875 – 888.
Powell TC. 2002. The philosophy of strategy. Strategic
Management Journal 23(9): 873 – 880.
Powell TC. 2003. Strategy without ontology. Strategic
Prahalad CK, Hamel G. 1990. The core competence of the corporation. Harvard Business Review 68(3): 79 – 91.
Priem R, Butler J. 2001a. Is the resource-based ‘view’ a useful perspective for strategic management research?
Academy of Management Review 26(1): 22 – 40.
Priem R, Butler J. 2001b. Tautology in the resource-based view and the implications of externally deter-mined resource value: further comments. Academy of
Management Review 26(1): 57 – 66.
Ray G, Barney JB, Muhanna WA. 2004. Capabilities, business processes, and competitive advantage: choosing the dependent variable in empirical tests of the resource-based view. Strategic Management
Journal 25(1): 23 – 37.
Reed R, DeFillippi RJ. 1990. Causal ambiguity, barriers to imitation, and sustainable competitive advantage.
Academy of Management Review 15(1): 88 – 102.
Rivkin JW. 2000. Imitation of complex strategies.
Management Science 46(6): 824 – 844.
Rivkin JW. 2001. Reproducing knowledge: replication without imitation at moderate complexity.
Organiza-tion Science 12(3): 274 – 293.
Rivikin JW, Porter ME. 1999. Matching Dell. Harvard Business School Case 799 – 158, Harvard Business School: Boston, MA.
Rumelt RP. 1984. Towards a strategic theory of the firm. In Competitive Strategic Management , Lamb RB (ed). Prentice-Hall: Englewood Cliffs, NJ; 556 – 570. Schendel D. 1994. Introduction to special issue:
competitive organizational behavior: toward an organizationally-based theory of competitive advan-tage. Strategic Management Journal , Winter Special Issue 15: 1 – 4.
Semiconductor International. 2005. After the blink: a kinder, gentler downturn. 1 January. http://www.reed-electronics.com/semiconductor/article/CA490075 (29 May 2007).
Seth A, Zinkhan G. 1991. Strategy and the research process: a comment, Strategic Management Journal 12(1): 75 – 82.
Siggelkow N. 2001. Change in the presence of fit: the rise, the fall, and the renaissance of Liz Claiborne.
Academy of Management Journal 44: 838 – 857.
Siggelkow N. 2002. Evolution toward fit. Administrative
Science Quarterly 47(1): 125 – 159.
Sivia DS. 1996. Data analysis: a Bayesian tutorial. Oxford University Press: Oxford, UK.
Spanos YE, Lioukas S. 2001. An examination into the causal logic of rent generation: contrasting Porter’s competitive strategy framework and the resource-based perspective. Strategic Management Journal 22(10): 907 – 934.
Spear S. 2004. Learning to lead at Toyota. Harvard
Business Review 82(5): 78 – 86.
Spear S, Bowen HK. 1999. Decoding the DNA of the Toyota production system. Harvard Business Review 77(5): 96 – 106.
Teece DJ, Pisano G., Shuen A. 1997. Dynamic capabili-ties and strategic management. Strategic Management
Journal 18(7): 509 – 533.
Tirole J. 1998. The Theory of Industrial Organization. MIT Press: Cambridge, MA.
von Bertalanffy L. 1968. General System Theory. George Braziller: New York.
Wei J. 2007. Examining the lean action plan with the Nissan revival plan: a case study. Proceedings of the 6th International Conference on Information and Management Sciences: Lhasa, China. 1 – 6 July: 212 – 217.
Winter SG, Szulanski G. 2001. Replication as strategy.
Organization Science 12(6): 730 – 743.
Zander U, Kogut B. 1995. Knowledge and the speed of the transfer and imitation of organizational capabilities: an empirical test. Organization Science 6(1): 76 – 92.
Zollo M, Winter SG. 2002. Deliberate learning and the evolution of dynamic capabilities. Organization
