RESOLVING THE COMMITMENT VERSUS FLEXIBILITY TRADE … 2019/Pacheco et al (2008).pdfpreempted by...

RESOLVING THE COMMITMENT VERSUSFLEXIBILITY TRADE-OFF:

THE ROLE OF RESOURCE ACCUMULATION LAGS

GONCALO PACHECO-DE-ALMEIDANew York University

JAMES E. HENDERSONIMD International

KAREL O. COOLINSEAD

We examine how time-consuming resource accumulation influences the classic strat-egy trade-off between commitment and flexibility. In particular, using 1975–95 datafrom the worldwide petrochemical industry, we study the impact of new plants’time-to-build on firms’ decisions to invest under uncertainty. Our results suggest anontrivial positive effect of resource accumulation lags on investment. Contradictingconventional wisdom, we show that competition may be fiercer in industries in whichfirms accumulate resources more slowly and that uncertainty is not always a disin-centive for investment. The robustness of these results is only diminished for extremelylong resource accumulation lags.

In December 2000, the management of Airbusannounced its plans to spend $11.9 billion tolaunch a new super jumbo jet, the A380. This com-mitment entailed a ten-year investment of financialand organizational resources in the long-haul high-capacity aircraft market, a market that is big enoughto allow only one firm to make a profit. The credi-ble commitment to the A380 made by Airbus pre-empted Boeing from competing in the super jumbojet market. Indeed, a few months later, Boeing’smanagement announced that it would cancel aproject to build a high-capacity plane, a “stretched”version of its popular B747. Airbus management’sdecision was made at a time of substantial uncer-tainty about the future of air travel and the com-mercial viability of a super jumbo jet. The companycould have remained flexible by waiting until themarket and technological investment conditionsgrew more certain, but it would have risked being

preempted by Boeing (Besanko, Dranove, Shanley,& Schaefer, 2004; Esty & Ghemawat, 2002).

This high-profile example illustrates the funda-mental strategic trade-off between commitment andflexibility that managers face when deploying firmresources to establish product-market positions.Commitment and flexibility lie on opposite ends ofa firm’s investment spectrum, and scholars havehistorically been divided as to which of the twostrategies is the main driver of investment value.On the one side, Stigler’s early contribution (1939)and recent research on real options (Adner &Levinthal, 2004; Dixit & Pindyck, 1994; Kogut,1991; Kulatilaka & Perotti, 1998; Luehrman, 1998;McGrath, 1997, 1999; McGrath, Ferrier, & Mende-low, 2004; Trigeorgis, 1996) have largely empha-sized the value of flexibility. As demand or techni-cal uncertainty increases, keeping options open bypostponing strategic investments and waiting foruncertainty to subside may be the optimal strategy.On the other side, work coming from mainstreameconomics and strategy has stressed the value ofinflexibility (Dixit & Pindyck, 1994; Fudenberg &Tirole, 1983; Ghemawat, 1991; Gilbert & Lieber-man, 1987; Lieberman, 1987a; Spence, 1979). Theargument in this stream of research is that makingearly irreversible commitments may secure futuremarket space and discourage rivals from investing;the inflexibility of such commitment has value byshaping rivals’ future behavior. Given these two

For their helpful comments, we wish to thank JuanAlcacer, Pankaj Ghemawat, Javier Gimeno, Duane Ire-land, Will Mitchell, Lars-Hendrik Roller, Rachelle Samp-son, and Peter Zemsky; several seminar participants atNew York University, London Business School, and the2005 Atlanta Competitive Advantage Conference; tworeviewers for this journal; four Academy of Managementreviewers; and two Stategic Management Society review-ers. All errors remain our own. We gratefully acknowl-edge financial support from Fundacao para a Ciencia e aTecnologia (grant number BD/11189/97).

� Academy of Management Journal2008, Vol. 51, No. 3, 517–536.

517

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download or email articles for individual use only.

opposing views, it is often not obvious which of thetwo drivers—flexibility or inflexibility—will ulti-mately prevail in imperfectly competitive anduncertain business situations.

Interestingly, this debate has overlooked the im-pact of resource accumulation on the trade-off be-tween commitment and flexibility, despite thesubstantial attention resource accumulation has re-ceived in the resource-based view literature (e.g.,Barney, 1991; Dierickx & Cool, 1989). In our initialexample, would Airbus management have commit-ted so early to the risky A380 project if aircraftdevelopment took six months instead of ten years?

Intuitively, the time required to accumulate re-sources should impact the relative attractiveness offlexible strategies. The more time it takes to accu-mulate resources prior to entry into a market, themore slowly firms will enter that market. Thus, ifmanagers decide not to invest ex ante and there isdemand ex post, their firms will have relinquishedprofits for the period in which they were out of themarket. In such a circumstance, waiting or post-poning investments in order to remain flexiblebecomes a less valuable option. Indeed, longerperiods of resource accumulation may offset uncer-tainty, leading more firms to invest earlier in newmarket opportunities. Thus, contrary to conven-tional wisdom, competition may be fiercer in in-dustries in which firms accumulate resources moreslowly. Certainly this view is consistent with sys-tematic empirical evidence of chronic excess sup-ply in industries such as commercial real estate,electricity generation, and bulk chemicals, in eachof which it takes a long time to accumulate re-sources and bring investments on line (Ghemawat,1984; Henderson & Cool, 2003b; Kling & McCue,1987; MacRae, 1989).

In this study, we examined the role of resourceaccumulation in an effort to better understand thecommitment versus flexibility trade-off. In particu-lar, we looked at how the time-consuming nature ofresource accumulation (one of the most fundamen-tal principles of the resource-based view) impactsthe decision to invest and the value of the “optionto wait.” By doing so, we directly extend the exist-ing literature on real options and strategic commit-ment (for a review, see Besanko et al. [2004:232–258]).

The empirical setting is the global petrochemicalindustry during the period 1975–95. In this indus-try, managers must make frequent, important deci-sions about whether to invest in plants to expandcapacity in the midst of demand uncertainty. As aresult, petrochemical firms are particularly proneto the commitment-flexibility dilemma. Further-more, the time it takes to plan and bring invest-

ments (such as new plants) on line is publiclyreported and varies substantially over regions andproduct categories.

THEORY AND HYPOTHESIS DEVELOPMENT

Resource accumulation is a central tenet of theresource-based view of the firm. According to thisview, a firm’s sustainable competitive advantagestems not from privileged product-market posi-tions, but from valuable and rare firm-specific re-sources deployed to support those product-marketpositions (Barney, 1991). These firm-specific re-sources cannot be bought on strategic factor mar-kets; they must be internally accumulated by firmsover time (Barney, 1986; Dierickx & Cool, 1989;Grant, 1991; Peteraf, 1993; Rumelt, 1984). For ex-ample, to implement product-market strategies,firms need employees with firm-specific skills andspecialized labor that cannot be rented on the mar-ket but have to be developed through on-the-joblearning and training (e.g., Schroeder, Bates, & Junt-tila, 2002).

The period of sustainability of a firm’s competi-tive advantage depends on the extent to whichthese firm-specific resources are difficult for com-petitors to imitate. Rivals’ attempts to catch up byspending money to develop a strategic resourcequickly are often unsuccessful because of timecompression diseconomies, resource “stickiness,”causal ambiguity, or lack of complementary re-sources (Barney, 1991; Dierickx & Cool, 1989; Lipp-man & Rumelt, 1982; Mishina, Pollock, & Porac,2004; Reed & DeFillippi, 1990).1 These barriers toimitation lead to extensive periods of resource ac-cumulation, which in turn sustain competitive ad-vantage (Cohen, Nelson, & Walsh, 2000; Lieberman& Montgomery, 1988, 1998).

In summary, the time firms take to internallyaccumulate firm-specific resources to follow prod-

1 Time compression diseconomies imply that speed-ing up resource accumulation more than proportionatelyincreases costs (Mansfield, 1971; Pacheco-de-Almeida &Zemsky, 2007; Scherer, 1967). Preliminary estimates oftime compression diseconomies have suggested that a 1percent reduction in the time taken to develop a resourcemay inflate development costs up to 2 percent (Graves,1989). Resource accumulation may also slow down withresource stickiness (Mishina, Pollock, & Porac, 2004;Penrose, 1959; Szulanski, 2003). If resources are sticky,their specialized nature makes them less useful if the taskat hand changes. Therefore, firms cannot leverage theirexisting resource bases and quickly convert resourceslack to alternative uses. In other words, adjusting anorganization’s stocks of sticky factors takes time.

518 JuneAcademy of Management Journal

uct-market strategies is central to the field of strat-egy and to the ex post sustainability of a firm’scompetitive advantage. Yet understanding of howtime-consuming resource accumulation affectsfirms’ ex ante strategic decisions, in particular in-vestment timing, is much more limited. We focuson this question in this section, and begin by de-fining resource accumulation lags.

Resource Accumulation Lags

We define a resource accumulation lag as thetime a firm takes, on average, to accumulate theresources to produce one unit of output in a prod-uct-market of interest.2 There are two reasons toadopt this functional definition of resource accu-mulation lags. First, it links internal resource de-velopment to external product-market positioning,answering recent calls in the resource-based viewliterature for an integrated analysis of resourcesand products (Barney, 2001; Priem & Butler, 2001a,2001b). Second, the definition obviates the mea-surement problem associated with most resourceaccumulation processes: their ongoing, unstruc-tured, intangible, and seldom empirically observ-able nature (Hall, 1992; Itami, 1987; Villalonga,2004). The accumulation of resources with the pur-pose of producing a certain quantity of output in aproduct-market tends to be limited in time, care-fully planned, and often externally visible (Cool,Almeida Costa, & Dierickx, 2002; Dierickx & Cool,1994). For example, investments in new produc-tion facilities are externally visible because theyoften require a substantial commitment of bothphysical and intangible resources to an industryand are often indicative of new market entry (e.g.,Lowe, 1979). The case in point in this article is thepetrochemical industry, where all firms’ invest-ments in new plants are systematically reported inthe Oil and Gas Journal (OGJ), the industry’s maintrade journal.

The OGJ follows stage-by-stage the developmentof every plant construction project. It also includesinformation on delayed, suspended, and aban-doned expansions. Investments in new chemicalproduction facilities usually undergo four mainphases: study, planning, engineering, and construc-tion. According to the OGJ, the first two phases(study and planning) capture the accumulation ofintangible resources such as managerial and tech-nological knowledge, demand information, com-petitive intelligence about rivals’ planned in-

vestments, government support, and employeeoperational expertise. This capital investment inintangibles represents the investment in organiza-tional capabilities (Maritan, 2001; Schroeder et al.,2002). The third and fourth phases of investment innew production facilities (engineering and con-struction) pertain mostly to the accumulation ofphysical resources immediately prior to production(e.g., equipment and plant facilities). Firms deployall these intangible and physical resources accumu-lated during the investment process to product-markets when production and market operationscommence. Table 1 provides an overview of each ofthese plant expansion phases and their averagelength based on information collected by the OGJfor plants built in the United States, Europe, andJapan during the period 1975 to 1995.

In most manufacturing industries, the time firmstake to plan and build new production facilities isempirically observable and clearly defined. Fre-quently, it also includes the intraorganizational ac-cumulation of both intangible and physical re-sources required to enter and compete in a certainmarket. Therefore, the “time-to-build” of a newplant (also referred to as the “investment lag”) is agood proxy for empirical measurement of resourceaccumulation lags.

Time-to-build is the product of multiple factors,typically including access to suppliers, governmen-tal policies, technology and product features, andother industry structural characteristics (Porter,1980). In prior studies, preliminary data on time-to-build has varied widely across industries (Ko-eva, 2000; Krainer, 1968; Mayer, 1960; Mayer &Sonenblum, 1955), ranging from two years for anew plant in the petrochemical industry (Lieber-man, 1987b), to as many as ten to develop a newaircraft (Esty & Ghemawat, 2002), to almost imme-diate set-up in some Internet-based businesses. Ko-eva (2000) found that time-to-build can be as low as13 months for simple, commodity products such asrubber and more than double that for more techno-logically advanced goods (e.g., 28 months for trans-port equipment). These results, although based on asmall number of time-to-build observations per in-dustry (five data points, on average), suggest thatresource accumulation lags may differ substantiallyover industries.

Despite their empirical importance, resource ac-cumulation lags have been underexamined in strat-egy research. Dierickx and Cool’s (1989) study isprobably the only seminal work that directly dis-cusses the impact of time lags on the ex post sus-tainability of competitive advantage. Moreover,very little has been said about the effect of resourceaccumulation lags on managers’ ex ante strategic

2 We use one unit of output as an (arbitrary or normal-ized) point of comparison between industries.

2008 519Pacheco-de-Almeida, Henderson, and Cool

decisions.3 In particular, analyses of managers’ in-vestment timing decisions have not accounted fortime-consuming resource accumulation. We nowturn to this ex ante investment problem by revisit-ing the commitment versus flexibility debate.

Resource Accumulation LagsFavoring Commitment

Intuitive interpretations of extant theory wouldsuggest that resource accumulation lags decreasethe likelihood of commitment. According to re-source-based and organization theory, such lagsslow down imitation and increase inertia, whichhinders firms’ investments in new market opportu-nities (Barney, 1991; Dierickx & Cool, 1989; Han-nan & Freeman, 1984; Hannan, Polos, & Carroll,

2004). According to the real options literature, un-certainty will exacerbate the negative effect of timelags on investment (e.g., Trigeorgis, 1996). The gen-eral real options perspective is that longer invest-ment projects are inherently highly uncertain (i.e.,have a high probability of producing extreme fu-ture market outcomes), which discourages firminvestment. However, in this subsection we con-test each of these two conventional views one at atime, starting with the resource-based/orga-nization theory perspective. Given the smallamount of research to be found in the strategyand management literatures on resource accumu-lation lags and investment timing, we draw sub-stantially on the financial economics literature tocomplement our understanding of firms’ invest-ment timing decisions.

Two theoretical modeling papers have studiedthe impact of resource accumulation on the relativebenefits of commitment and flexibility (Bar-Ilan &Strange, 1996; Pacheco-de-Almeida & Zemsky,2003). Their findings suggest that time-consumingresource accumulation has two fundamental impli-cations for managers’ investment decisions that arerobust to alternative theoretical model specifica-tions (e.g., different types of industries and varyingdegrees of strategic interaction between firms).

First, in industries in which it takes a long timeto accumulate resources, managers are unable toquickly adjust their strategies to new market andcompetitive information. When making an invest-ment decision, managers weigh the benefits and

3 Exceptions include prior work on strategic flexibil-ity, which is defined as the capability to quickly respondto changing competitive conditions (Hitt, Keats, & De-Marie, 1998; Shimizu & Hitt, 2004). On a similar note,other authors have “referred to this ability to [quickly]achieve new forms of competitive advantage as ‘dynamiccapabilities’. . . . The term ‘dynamic’ refers to the capac-ity to renew competences so as to achieve congruencewith the changing business environment . . . requiredwhen time-to-market and timing are critical, the rate oftechnological change is rapid, and the nature of futurecompetition and markets difficult to determine” (Teece,Pisano, & Shuen, 1997: 515). One could argue that theability to rapidly accumulate resources is one type ofdynamic capability.

TABLE 1Plant Expansion Phases in the Petrochemical Industrya

Project Phase

Average Time to Build

Phase DescriptionMonths Percentages

1. Study 1.5 5 Market and competitive information analysisFinancing, capital budgeting and risk managementSecuring government approval

2. Planning 9.1 31 Technology R&D or technology supplier selectionSchematic and developed project designPreliminary schedule and permitting

3. Engineering 5.5 19 Detailed engineering and process designSystem specifications (electrical, mechanical, piping)Final cost estimates

4. Construction 13.3 45 Procurement, bid analysis, equipment purchase ordersMaterial fabricationPhysical construction of the plant

Total 29.4 100

a Source: Oil and Gas Journal information on plants built in the United States, Europe, and Japan 1975–95.


costs of early investment. The opportunity cost ofwaiting is the forgone income from a project, whichdepends on the price the product commands dur-ing the delay. If a firm can enter a market immedi-ately in situations of strong demand or high prices,the opportunity cost of waiting to invest is limited.However, if there is an interval of time between adecision to invest and receipt of a project’s firstrevenues, the opportunity cost of postponing in-vestment can be substantially higher, including theprofits forsaken while the firm is accumulating re-sources to enter the desired market (Bar-Ilan &Strange, 1996; Pacheco-de-Almeida & Zemsky,2003). Indeed, long resource accumulation lags areknown to create barriers to imitation and sustaincompetitive advantages (Barney, 1991; Dierickx &Cool, 1989). These resource accumulation lags cre-ate a grace period with mild competition and ex-traordinary rents for early mover firms as their laterivals struggle to catch up. If an early mover’s mar-ket strategy proves to be winning (e.g., if there ismarket demand), a price premium is expected dur-ing the resource accumulation lag because compe-tition and supply are low. Ultimately, the anticipa-tion of such favorable ex post conditions, which areamplified by longer investment lags, triggers earlyinvestment.

To summarize, as resource accumulation lags in-crease in length under a constant (low or high)level of uncertainty, firms may be more likely toinvest, contrary to popular wisdom. The followinghypothesis expresses this main direct effect of timeon the likelihood of investment:

Hypothesis 1. Ceteris paribus, resource accu-mulation lags have a positive effect on a firm’slikelihood of investment.

Second, in industries in which resource accumu-lation is more time consuming, strategic invest-ment decisions must be based on projections andforecasts for a more distant future. This require-ment increases uncertainty, or the likelihood ofextreme future market outcomes. However, firmsalso have more time to exercise the option of aban-doning their investment projects; thus, their profitsare better safeguarded in unfavorable market con-ditions. As a result, longer resource accumulationlags increase the profit potential of an investmentproject more than its loss potential. This relation-ship has been shown to increase the overall ex-pected value of ex ante investment (Bar-Ilan &Strange, 1996), propelling managers to invest underuncertainty with a higher probability and suggest-ing a positive interaction effect between uncer-tainty and time lags. Stated differently, the abilityto abandon a project and exit a market when prices

are low means the downside of the investment istruncated. An increase in uncertainty, therefore,raises the expected profits over the period of thedelay and may lead to earlier investment. This ef-fect should be more salient when investment lagsare longer, as managers necessarily make projec-tions for a more distant and, thus, more uncertainfuture. In short, time is expected to have a moder-ating effect on the negative impact of uncertaintyon investment:

Hypothesis 2. Ceteris paribus, resource accu-mulation lags reduce the main negative effectof uncertainty (the option value of waiting) ona firm’s likelihood of investment.

Hypotheses 1 and 2 may explain the hypercom-petitive behavior and chronic excess capacity ob-served in capital-intensive industries that ex-perience long resource accumulation lags (e.g.,commercial real estate, electricity generation, pet-rochemicals, semiconductors). They may also shedlight on the aggressive investment policy describedby Intel’s CEO Andy Grove, of “building factoriestwo years in advance of needing them, before . . .having the products to run in them, and before . . .knowing the industry’s going to grow” (Kirkpatrick,1997: 61). In addition, the reasoning leading to ourhypotheses may influence firms’ patterns of invest-ment in new markets.4

Resource Accumulation Lags Favoring Flexibility

As an industry’s resource accumulation lags in-crease, at the limit, two mechanisms may reduce oreven reverse the aforementioned positive effect ofresource accumulation lags on investment (com-mitment), making waiting (flexibility) more likely.

First, since resource accumulation lags enhancefirms’ disposition toward investing (Hypotheses 1and 2), very long lags may ultimately generatewidespread “bandwagons,” wherein a large num-ber of firms in an industry simultaneously investearly (Pacheco-de-Almeida & Zemsky, 2003). If ahigh number of firms invest early, it will be moredifficult for any particular company to get ahead of

4 Hypotheses 1 and 2 also formalize the following in-tuition expressed by Porter: “Long lead times [in invest-ment] require firms to base their decisions on projectionsof demand and competitive behavior far into the future orpay a penalty in not capitalizing on opportunity if de-mand materializes. Long lead times increase the penaltyto the firm who is left behind without capacity, andhence may cause risk-averse firms to be more prone toinvest even though the capacity decision itself is risky”(1980: 329).


the competition by making an investment largeenough to preempt rivals. In other words, the syn-chronicity of early investments reduces the oppor-tunity for timing- and volume-based first moveradvantages such as large strategic commitments toblock competitors (Ghemawat, 1991). Therefore,with very long resource accumulation lags, manag-ers have more incentives to make smaller-scale in-vestments (Pacheco-de-Almeida & Zemsky, 2003)or to wait to invest. Indeed, the prospect of wide-spread investment bandwagons may eventuallydiscourage investment altogether as managers an-ticipate problems of industry overcapacity (Hen-derson & Cool, 2003b). As a result, waiting isfavored.

Second, the rate of return required to make aninvestment worthwhile may be higher for projectswith very long resource accumulation lags becauseof short-term stock market pressures. Indeed, a sub-stantial amount of research in finance and econom-ics (e.g., Narayanan, 1985a, 1985b; Stein, 1988,1989) has suggested that incentive schemes or thefear of losing control often lead the managers ofpublicly traded companies to overemphasize theimportance of their firms’ short-run stock pricemovements. These short-term objectives may ulti-mately lead to underinvestment in very long-runprojects. If amassing the resources required to pur-sue market opportunities takes an extremely longtime, managers may prefer not to invest and insteadmay turn to alternative projects with shorter pay-back periods. Projects that yield quick cash recov-ery enhance the reinvestment possibilities for firms(Ross, Westerfield, & Jaffe, 1996). As a result, therate of return required to render longer-termprojects economically attractive tends to be higherthan the comparable rate for short-term projects(Ainslie & Haslam, 1992), making long-term invest-ment less likely. It has been empirically shown thatsuch short-term bias increases exponentially withtime lags, an effect that naturally imposes an inflec-tion point on the positive effect of time for verylong resource accumulation lags (Miles, 1993).

Hypothesis 3. Ceteris paribus, very long re-source accumulation lags have a negative ef-fect on a firm’s likelihood of investment.

In conclusion, the likelihood of commitment isexpected to increase with resource accumulationlags (Hypotheses 1 and 2). This effect will becurbed or even reversed for very long resource ac-cumulation lags (Hypothesis 3), making flexibilitymore likely.

RESEARCH DESIGN

Sample Data

The empirical analysis was carried out in thepetrochemical industry in the United States, Eu-rope, and Japan during the period 1975–95. Theindustry spans several industrial classificationcodes and includes numerous products, such ascommodity chemicals, plastic resins, synthetic rub-ber, and fibers (for a basic review, see Chapman[1991]).

Several traits make the petrochemical industry aparticularly appropriate setting for studying the ef-fect of resource accumulation lags on strategic in-vestment. First, industry players make constantcapacity expansion decisions and are thus con-sistently faced with the commitment versus flexi-bility dilemma. Second, resource accumulationlags in the form of the time it takes firms to planand bring investments (such as new plants) onlineare publicly reported and vary substantially overregions and product categories. Third, the variousregions and products (i.e., petrochemical subindus-tries) can be studied simultaneously while main-taining a homogeneous sample. Finally, there is alarge body of empirical work in strategy availableon the petrochemical industry, which we are ableto build upon.

The industry evolution can be divided into threemain life cycle periods: 1850–1950, 1950–1975,and 1975–2005. The most recent period was themost suitable for our study for multiple reasons(our data set covers the years 1975–95). By this latestage, firms’ expansion decisions were mainly de-termined by purely profit-maximizing criteria, asgovernments offered fewer subsidies to render un-economic projects financially viable. In addition,this third stage of the industry life cycle is when theassumptions behind the (game-) theoretical modelsthat generate our hypotheses hold the best. As thefollowing quotation from an industry periodicalsuggests, during this period “information asym-metries” (and “expectation asymmetries”) amongfirms about future market conditions and competi-tors’ moves were reduced to their lowest level sincethe industry’s birth: “Demand forecasts now reflecta strong consensus, there being a difference be-tween the highest and lowest of less than 10%. Thiscompares with the 100% difference which wasprevalent in the 1970s” (Chemical Insight, 1987: 2)The period also saw improved communication andinformation about the investment plans and strate-gic commitments of different players in theindustry.

Two data sets were used in this article. The firstconsists of 556 total time-to-build observations corre-


sponding to 556 plant construction projects (i.e., ex-pansions) that we manually collected from trade jour-nal articles in the Oil and Gas Journal (OGJ) andcomplemented with field interviews.5 We used thesedata only to generate measures of resource accumu-lation lags. All other data on firms’ investments camefrom our second data set, which we built from mul-tiple data sources, including the Tecnon ConsultingGroup, Datastream International, Chemical Insight,Compustat, Moody’s, the Organisation for EconomicCooperation and Development (OECD), DeWitt &Company, and Chem Systems. This is a very rich dataset containing performance and investment informa-tion at both the firm and product-market levels (Hen-derson, 1998). In this study, we employed data onproduct-market capacity at the firm level and data onproduction and consumption at the product-mar-ket level. Our subsample included a total of 5,848investment observations on 879 different plantconstruction projects.

Both data sets used in this study included informa-tion on nine major commodity petrochemicals (eth-ylene, HDPE, LDPE, LLDPE, PP, styrene, PS, PVC,and VCM), which together account for over 50 per-cent of the industry’s total sales volume.6 The time-to-build and investment data sets included 166 and116 firms, respectively, with most firms present inboth sets. Eleven firms exercised the option to aban-don, leaving 13 plant construction projects uncom-pleted; most decisions to abandon projects alreadyunderway took place in Europe. Table 2 summarizesthe observations in both data sets for each of thepetrochemical subindustries.

The purpose of this study was not to match ormerge the two data sets, but to use the time-to-buildobservations to construct accurate measures of in-dustry average resource accumulation lags that, inturn, could help explain the investment patternsevidenced in the second database. Merging the twodatabases would have been further unadvisable be-cause a substantial number of investment observa-tions, lacking corresponding time-to-build data,would be lost. Indeed, the time-to-build data on

individual construction projects proved the mostdifficult piece of information to collect: data wereavailable for 63 percent of the 879 investmentprojects reported in the larger investment data set.

Dependent Variable

Our hypotheses concern the impact of resourceaccumulation lags and uncertainty on firms’ likeli-hood of investment. We operationalized firms’ in-vestments in capacity expansion in the petrochem-ical industry as has been done in prior work on thetopic (Gilbert & Lieberman, 1987; Henderson &Cool, 2003a, 2003b; Lieberman, 1987a, 1987b). Fora given observation year, investment was equal to 1for all observations in which a firm expanded ei-ther by adding a new “greenfield” plant or by in-creasing at least one existing plant’s productioncapacity by more than 10 percent. Otherwise, it wasequal to 0. Debottlenecking, representing only 0–10percent of capacity expansion per year, does notreflect a major capacity expansion decision; thus, itwas not included in the analysis.

Explanatory Variables

Resource accumulation lag. In the second sec-tion of this article, we argued that the time firms taketo plan and build new production facilities is gener-ally a good proxy for resource accumulation lags. Thetimes-to-build of plants are empirically observable,clearly defined, and frequently include the intraor-ganizational accumulation of both intangible andphysical resources prior to market entry. This is cer-tainly the case in the petrochemical industry. There-fore, we used the time-to-build of plants to measurethe firms’ resource accumulation lags. Our data setcontained 556 time-to-build observations.

We calculated the time-to-build of each individualproject as follows. The official start (end) of a plantexpansion was assumed to be the date on which theproject was first (last) reported in the OGJ minus(plus) 90 days. The 90-day lag resulted from the factthat the OGJ only reports the status of each plantexpansion twice per year, in April and October. Thus,if a plant expansion appeared for the first time in oneissue of the journal, we could only infer that theproject started sometime after the prior issue andbefore the current one. For simplicity, we assumedthat each plant expansion started exactly in betweenthe two consecutive issues of the OGJ and thus usedthe 90-day lag (three months). A similar logic appliedto the official end date of the plant expansion, unlessan expected completion date for the project was re-ported, in which case the latter was assumed to be theofficial end date of the expansion.

5 Twice per year (in April and October), the OGJ re-ports on the development of every plant constructionproject. On average, each project was reported 3.1 timesin the OGJ during its lifetime, which added up to a totalof 1,721 data points. Therefore, our time-to-build data setcontained 1,721 single observations (or data entries), butonly 556 project time-to-build observations.

6 The product names were abbreviated as follows:HDPE is high-density polyethylene; LDPE is low-densitypolyethylene; LLDPE is linear low-density polyethylene;PP is polypropylene; PS is polystyrene; PVC is polyvinylchloride; and VCM is vinyl chloride monomer.


Using these 556 individual project time-to-buildobservations, we measured the resource accumula-tion lag variable as product-region averages perunit of output. Since larger petrochemical projectstake longer to build, simply averaging times-to-build without adjusting for an equivalent amountof capacity would have generated industry averageresource accumulation lag measures that were notcomparable for the purposes of this study. In otherwords, we were interested in how long, on average,it took firms to accumulate resources to produceone unit of output in each petrochemical subindus-try. Several alternative operationalizations of theresource accumulation lag variable can be found inthe robustness checks section; all measures pro-duced similar final econometric results. Thus,

Resource accumulation lag � �e � 1

n (Te/Ke)n

,

where Te represents the time-to-build of expansione in the petrochemical subindustry defined by one

product-region combination, Ke is the expansioncapacity, and n is the total number of expansions inthat subindustry over 1975–95. We standardizedresource accumulation lag to reduce multicol-linearity between its main and squared terms in thefinal estimated regression model.

Table 3 summarizes the average resource ac-cumulation lags for the 27 petrochemical subin-dustries (i.e., product-region combinations) thatwe studied. These are time-to-build averages inmonths. For simplicity of interpretation, all mea-sures were scaled up to 100,000 metric tons (MT)per year of equivalent capacity (the median capac-ity expanded in the time-to-build data set). Severalinteresting empirical patterns emerge from thisdata. First, the average time-to-build of a plant isapproximately 29 months, 36 percent consisting ofintangible resource accumulation (the study andplanning phases of each investment project) and 64percent, of physical asset accumulation (the engi-neering and construction phases). Second, resourceaccumulation takes, on average, 17 percent longer

TABLE 2Observations Included in the Two Data Sets

Product Region

Number of Observations Number of Expansions Number of Firms

Total Capacity(103 MT/year)a

Time-to-BuildData

InvestmentData

Time-to-BuildData

InvestmentData

Time-to-BuildData

InvestmentData

Ethylene United States 164 315 64 39 28 19 1,040HDPE United States 68 221 25 51 14 14 435LDPE United States 42 182 20 12 11 11 295LLDPE United States 12 63 5 16 4 8 281PP United States 124 220 47 38 24 13 327PS United States 45 227 13 39 9 14 193PVC United States 64 252 25 38 13 15 432Styrene United States 32 170 15 15 13 10 537VCM United States 37 157 12 16 9 10 632Ethylene Europe 249 404 73 57 39 24 901HDPE Europe 137 264 31 64 20 16 311LDPE Europe 73 281 24 35 17 18 400LLDPE Europe 29 71 14 21 9 8 164PP Europe 168 276 55 85 25 16 401PS Europe 53 314 17 53 13 20 187PVC Europe 98 284 23 54 19 17 383Styrene Europe 43 174 13 19 11 11 570VCM Europe 19 270 19 18 15 16 500Ethylene Japan 82 221 14 27 9 13 585HDPE Japan 23 176 5 28 3 9 168LDPE Japan 29 173 5 14 5 10 174LLDPE Japan 9 64 3 10 2 7 73PP Japan 58 220 15 43 12 13 186PS Japan 21 267 6 29 5 16 114PVC Japan 8 193 3 24 3 12 135Styrene Japan 19 150 6 21 6 9 326VCM Japan 15 239 4 13 4 14 187

Total 1,721 5,848 556 879

a 1995 data.


in Japan than in Europe, and 49 percent longer inEurope than in the United States. European coun-tries and Japan are typically more stringent aboutenvironmental issues than the United States, whichslows down plant approval and construction. Third,resource accumulation is faster for some petro-chemical products than for others. Average time-to-build ranges from a minimum of 8.5 months to amaximum of 71.1 months and is shorter for primaryand intermediate chemicals such as ethylene,VCM, and styrene than for commodity plasticssuch as HDPE, LDPE, PVC, PS, or PP. These geo-graphic and product differences are all statisticallysignificant (all p’s � .01).

Demand uncertainty. The second explanatoryvariable we sought to operationalize was uncertainty,which usually is the variance of a variable. In thisstudy, we focused on the most prominent driver ofuncertainty in the petrochemical industry: demand.The corresponding measure was defined as the stan-dard deviation of four years’ worth of industrial pro-duction prior to the year under consideration. Thefour-year period prior to the focal year was chosen tomatch the operationalization of similar variables inprior empirical work in this industry (e.g., Lieber-man, 1987a). We computed demand uncertainty foreach geographic region (U.S., Europe, and Japan).This measure was standardized.7

Control Variables

The controls consisted of a few theoretically rel-evant variables that prior empirical studies of

firms’ investment decisions in the petrochemicalindustry have repeatedly shown to be significantand signed as predicted. These variables can bedivided into two distinct categories: nonstrategicand strategic. Independently of rivals’ competitivemoves, three variables matter for firms’ expansiondecisions: historical demand growth rate, invest-ment “lumpiness,” and industry capacity utiliza-tion/excess capacity. Two strategic investmentvariables should also be considered: relative mar-ket share and rivals’ expansion. These variables aredefined below in this same order.

Demand growth. Predictably, the demand growthrate has been shown to have a positive impact onthe probability of expansion (Gilbert & Lieberman,1987; Henderson & Cool, 2003a, 2003b; Lieberman,1987a, 1987b). Demand growth was the four-yearhistorical compound annual growth rate of produc-tion for a product-region. Although firm-levelgrowth in production would have been more indic-ative of demand, data were not available. Priorwork has used this product-region proxy.

Excess capacity. As has been observed in previ-ous studies, excess capacity discourages further ca-pacity expansions (Gilbert & Lieberman, 1987; Hen-derson & Cool, 2003a, 2003b; Lieberman, 1987a,1987b). It was measured as the amount of oversup-ply, taken as a percentage of total industry capacity,which is the inverse of capacity utilization. Capac-ity utilization was calculated as a product-marketaverage rate of capacity utilization over the twoprevious years.

Investment lumpiness. Like excess capacity, in-vestment lumpiness hinders investment (Gilbert &Lieberman, 1987; Henderson & Cool, 2003a, 2003b;Lieberman, 1987a, 1987b). The larger the plant sizerequired to reach the minimum efficient scale(MES), the fewer incentives firms have to expand.This is because large amounts of capacity are risk-ier to add than small ones. Lumpiness should becalculated as the percentage of the market capacitycovered by the capacity of an MES plant, but datawere not available. Instead, we constructed aproxy: average plant size divided by total product-market production.

Market share. A firm’s market share is expectedto have a strongly positive effect on its expansionprobability (Gilbert & Lieberman, 1987; Henderson& Cool, 2003a, 2003b). For multiple reasons, largerpetrochemical firms may have a competitive ad-vantage in capacity expansion over smaller firms(e.g., more investment experience, better access tosuppliers, stronger brand name). At the same time,expansions for larger firms represent smaller incre-ments of their total capacity. Market share was afirm’s share of total product-market capacity.

7 Standardization reduces multicollinearity with thecontrol variable demand growth. Multicollinearity re-sults from the fact that demand growth and demanduncertainty were defined as the mean and the variance ofthe same empirical measure (industrial production) andsimultaneously included as independent variables in themodel.

TABLE 3Average Resource Accumulation Lags in Months

per 100,000 Metric Tons per Year of Capacity

Product United States Europe Japan

Ethylene 14.7 19.6 17.7HDPE 24.3 34.1 34.2LDPE 15.3 29.3 51.6LLDPE 13.8 26.2 41.4PP 29.3 28.5 40.6PS 40.1 47.0 71.1PVC 30.4 40.0 37.7Styrene 13.8 24.9 25.3VCM 8.5 20.3 14.9


Rivals’ expansion. Finally, rivals’ expansionshould also have a clear, positive impact on the prob-ability of investment: firms have been found to bemore likely to invest when they see their rivals in-vesting (for a review, see Henderson and Cool [2003a,2003b]). This may happen because of managementmyopia, defective corporate governance systems, orinformation asymmetries between firms, wherebycompetitors’ expansions signal expectations aboutfuture market conditions. The variable rivals’ expan-sion was constructed as the percentage of total capac-ity added simultaneously by rivals in a product-market during an observation year.

Estimation Methods

A logit regression analysis tests the hypothesesabout the effect of industry resource accumulationlags on firms’ likelihood of investment. The modelis estimated using both random and fixed effects(firm, product, region, and year) in the followinggeneral form:8

P(investment � 1�X, �) � ��

� �1demand growth � �2excess capacity

� �3investment lumpiness � �4market share

� �5rivals’ expansion

� �6demand uncertainty

� �7resource accumulation lag

� �8(resource accumulation lag

� demand uncertainty)

� �9(resource accumulation lag2) � ��.

RESULTS

Descriptive Statistics

Table 4 summarizes the key statistics for thecentral variables used in the estimations. All ofthem show substantial variation. On average, ap-proximately 15 percent of the firms in the sampleinvested in capacity in any given year, if oneassumes projects are equally distributed amongfirms. Incremental expansions accounted forabout 71 percent of these new investments, withan average of 63,295 MT per year in capacity.Greenfield projects represented the remaining 29

percent of investments, with an average of151,800 MT per year in capacity. Total capacityadded was approximately 81,090 MT per year,with a maximum of 680,000 MT per year for anew ethylene cracker. The historical demandgrowth rate was positive at some points and neg-ative at others, and the industry’s excess capacityvaried widely, both reflecting the strong industrycyclicality experienced during the 1975–95 pe-riod. The table presents the descriptive statisticsfor the uncertainty and resource accumulationlag variables prior to standardization. The peak ofuncertainty was reached at 4.19 standard devia-tions above the mean, an occurrence with lessthan 2.85 percent probability in a two-tailed dis-tribution (after standardization according to Che-byshev inequality). Finally, variation in the ac-cumulation lag measure is also considerable, aspredicted.

Logit Analysis

This study examined the influence of time-consuming resource accumulation on firms’ in-vestment strategies on the trade-off between com-mitment and flexibility. Accordingly, our logitregression analysis addressed the extent to whichfirms exploit the simplest type of flexibility—theoption to wait— by delaying investment.

Table 5 summarizes the results for the random-and fixed-effects models. As is conventional, weperformed a likelihood ratio test comparing therestricted or pooled with the unrestricted logit re-gressions to determine the joint significance of allpossible fixed effects (firm, product, region, andyear). All dummies came out significant (regions:

8 The function �(.) denotes the logistic cumulativedistribution.

TABLE 4Descriptive Statistics, Selected Variablesa

Variables Mean s.d. Minimum Maximum

Capacity expansion 0.15 0 1Greenfield expansion 0.04 0 1Incremental expansion 0.11 0 1Total capacity added 81.09 107.23 1 680.00Demand growth 3.54 4.19 �7.33 28.98Excess capacity 17.16 12.13 0 58.99Investment lumpiness 4.38 2.70 1.10 18.75Market share 0.83 0.83 0 6.37Rivals’ expansion 3.77 3.86 0 24.85Demand uncertaintyb 3.39 1.47 1.48 9.55Resource

accumulation lagb, c29.59 13.22 8.52 71.10

a n � 5,848.b Nonstandardized variable.c Months per 100,000 metric tons per year.


TA

BL

E5

Res

ult

sof

Log

itA

nal

ysis

ofE

xpan

sion

Pro

babi

lity

a

Var

iabl

esan

dM

ean

Val

ues

Mod

el1

Mod

el2

Mod

el3

Mod

el4

Mod

el5

Mod

el6

Mod

el7

Mod

el8

Con

stan

t�

0.24

***

(0.1

2)�

0.24

***

(0.1

2)�

0.24

***

(0.1

2)�

0.24

***

(0.1

2)�

0.22

***

(0.1

2)�

0.22

***

(0.1

2)�

0.23

***

(0.1

2)�

0.22

*(1

.06)

Dem

and

grow

th(0

.04)

0.62

***

(0.0

1)0.

67**

*(0

.01)

0.58

***

(0.0

1)0.

64**

*(0

.01)

0.54

***

(0.0

1)0.

54**

*(0

.01)

0.59

***

(0.0

1)0.

51**

*(0

.01)

Exc

ess

cap

acit

y(0

.17)

�0.

05(0

.00)

�0.

04(0

.00)

�0.

08*

(0.0

0)�

0.07

†(0

.00)

�0.

11**

(0.0

0)�

0.11

**(0

.00)

�0.

10*

(0.0

0)�

0.10

*(0

.00)

Inve

stm

ent

lum

pin

ess

(0.0

4)�

0.75

***

(0.0

2)�

0.77

***

(0.0

2)�

0.59

**(0

.02)

�0.

62**

(0.0

2)�

0.43

*(0

.02)

�0.

44*

(0.0

2)�

0.47

*(0

.02)

�0.

38†

(0.0

2)

Mar

ket

shar

e(0

.83)

0.02

***

(0.0

5)0.

02**

*(0

.05)

0.02

**(0

.05)

0.02

***

(0.0

5)0.

02**

*(0

.05)

0.02

***

(0.0

5)0.

02**

*(0

.05)

0.01

**(0

.05)

Riv

als’

exp

ansi

on(0

.04)

0.78

***

(0.0

1)0.

79**

*(0

.01)

0.77

***

(0.0

1)0.

78**

*(0

.01)

0.71

***

(0.0

1)0.

70**

*(0

.01)

0.71

***

(0.0

1)0.

65**

*(0

.01)

Dem

and

un

cert

ain

ty(0

)�

0.01

*(0

.04)

�0.

01†

(0.0

4)�

0.01

†(0

.04)

�0.

00(0

.04)

Res

ourc

eac

cum

ula

tion

lag

(0)

0.02

**(0

.04)

0.01

**(0

.04)

0.03

***

(0.0

6)0.

03**

*(0

.06)

0.03

***

(0.0

6)0.

03**

*(0

.06)

Res

ourc

eac

cum

ula

tion

lag

squ

ared

(1)

�0.

01**

*(0

.03)

�0.

01**

*(0

.03)

�0.

01**

*(0

.03)

�0.

01**

*(0

.03)

Res

ourc

eac

cum

ula

tion

lag

�d

eman

du

nce

rtai

nty

(0.0

2)0.

01†

(0.0

4)0.

01*

(0.0

3)0.

01†

(0.0

4)

Nu

mbe

rof

obse

rvat

ion

s5,

848

5,84

85,

848

5,84

85,

848

5,84

85,

848

5,59

0

Log

-lik

elih

ood

chi-

squ

are

�2,

389.

90�

2,37

9.00

�2,

376.

30�

2,37

4.70

�2,

368.

50�

2,36

7.00

�2,

365.

00�

2,24

9.13

Ch

i-sq

uar

efo

rm

odel

147.

86**

*15

0.77

***

155.

57**

*15

8.05

***

169.

21**

*17

1.82

***

173.

55**

*33

5.77

***

Ch

i-sq

uar

efo

rfi

rms

Sig

nif

ican

t

†p

�.1

0*

p�

.05

**p

�.0

1**

*p

�.0

01

�2[2] � 16.53; products: �2[8] � 37.73; years:�2[16] � 73.09; firms: �2[98] � 168.48).9

The firm dummies were included in the esti-mations and are reported in the last column ofTable 5. Although the year, product, and regiondummies were also significant, their inclusion inthe estimation rendered much of the analysis un-interesting. With the year dummies, the variablescapturing cyclicality (excess capacity and de-mand uncertainty) unsurprisingly turned insig-nificant. Since there was no other loss in signif-icance in the variables of interest and the controlvariables adequately captured the circumstancesunique to particular points in time, the time dum-mies were not kept in the estimations. The prod-uct and region dummies naturally absorbed partof the variance in the dependent variable ex-plained by the product-region accumulation lagmeasures (and demand uncertainty), which coun-tered the purpose of this study. As a result, al-though we carefully examined them during es-timation, we do not report these aggressivefixed-effect models below.

The first column in Table 5 lists the independentvariables and their values at the sample mean. Theother columns report the parameter estimates andstandard errors. Each parameter estimate is the par-tial derivative of the probability of expansion withrespect to the independent variable, with the logis-tic cumulative distribution evaluated at the samplemeans of the data.10 Under model 1, the basemodel, we only report the estimates for the controlvariables. In models 2, 3, and 4 we add the maineffects of uncertainty and accumulation lag. Mod-els 5 and 6 introduce the squared and interactioneffects of resource accumulation lag on the likeli-hood of investment. Model 7 is the full model.Finally, although models 1–7 are random-effectsmodels, model 8 reports the estimation results withfirm fixed effects. The usual tests for multicol-linearity, heteroscedasticity, and serial correlationwere performed. No major problems were found,and the estimation proceeded via the conventionalregression techniques described above.11

Table 5 deserves several comments. First, thechi-square tests for the models are significant, al-lowing us to reject the null hypothesis that allcoefficients with the exception of the intercept arezero. Second, the control variables are generallysignificant and with the expected signs. Rapid de-mand growth signals the need for investment inadditional plant capacity and, thus, has a positiveimpact on investment likelihood. The level of idleor excess capacity in the industry was expected todissuade investment since it is often associatedwith industry oversupply, price declines, and in-cumbent retaliation responses to new market entry.Accordingly, the corresponding coefficient is con-sistently negative and significant in the random-effects models. The effect of investment lumpinessis also straightforward: if capacity can only beadded in large increments, firms have fewer incen-tives to invest. The likelihood of investment de-pends positively on firms’ relative market shares,reflecting the idea that large firms have a competi-tive advantage over small firms. A strong case forinvestment bandwagon behaviors can also bemade, with a positive and significant coefficient forrivals’ expansion, the percentage of total capacityadded simultaneously by rivals.

Third, supporting the conventional option theoryview of investment, demand uncertainty doeshinder investment in the random-effects models, asexpected.

Fourth, the logit analysis indicates clear supportfor all the hypotheses on resource accumulationlags favoring commitment developed herein. In-deed, the coefficient associated with the main re-source accumulation lag variable is consistentlypositive and significant in all the estimated models,as predicted in Hypothesis 1. In particular, model 7suggests that a marginal six-month increase overthe mean in the average time it takes firms to accu-mulate resources (from 29.59 to 35.59 months)—anincrease that may be due, for instance, to the intro-duction of stricter governmental regulation or morecomplex technologies—raises firms’ predicted like-lihood of investment by 1.24 percent (from 12.50%

9 In the degrees of freedom, three years and 17 firmswere dropped because of perfect collinearity.

10 Following Ai and Norton (2003), we computed themagnitude of the interaction effect. We found that therewas not, in fact, much difference between the marginaleffect of the interaction term (reported in Table 5) and itsmagnitude.

11 See Appendix B for the Pearson correlation coeffi-cients. The unstructured within-panel error correlationmatrix did not suggest any discernible AR1 pattern. For

heteroscedasticity, we assumed that the coefficients werebiased and thus estimated the models using the sand-wich or robust estimator of variance (i.e., when thesource of the bias is unknown). Four estimations of thefull model were conducted: the original model; robuststandard errors for the whole sample; robust standarderrors within each region, company, product combina-tion; and robust standard errors within each companycluster of observations. However, the changes to the stan-dard errors were minor, leaving the significance of thecoefficients intact.


to 13.74%). This increase corresponds to five newfirms investing each year, resulting in at least 89new projects during the period of analysis (i.e., a 10percent increase in the total number of capacityexpansions recorded in the database, assuming thatinvestment projects are equally distributed amongfirms). These numbers are particularly striking in amature industry with chronic excess capacity suchas petrochemicals. These results reflect the fact thatlonger resource accumulation lags decelerate firms’reactions to new market and competitive informa-tion, increasing the potential forgone profits iffirms wait to invest, and thereby inducing them tocommit. Furthermore, the coefficient associatedwith the interaction between resource accumula-tion lag and demand uncertainty is consistentlypositive and significant, as predicted in Hypothesis2. Contrary to conventional wisdom, in industrieswith sufficiently time-consuming resource accu-mulation, an increase in uncertainty may ulti-mately encourage, rather than dissuade, invest-ment. An increase in uncertainty raises theexpected profits over the period of delay because offirms’ greater ability to abandon an investmentwith longer resource accumulation lags, whichleads to commitment. For example, since the mainand interaction terms of demand uncertainty havesimilar coefficients of opposite signs in model 7, ifthe resource accumulation lag variable is greaterthan 1 (greater than 42.37 months prior to standard-ization), increasing uncertainty raises the probabil-ity of investment.

Fifth, the data also support Hypothesis 3, onresource accumulation lags favoring flexibility. Inindustries in which resource accumulation is verytime-consuming, the risk of widespread investmentbandwagons and industry overcapacity is espe-cially high. Also, short-term stock market pressuresdiscourage investment in long-run projects that re-duce firms’ cash recovery and reinvestment possi-bilities. The negative and highly significant coeffi-cient of the squared resource accumulation lagvariable confirms that, if it takes very long to amassresources prior to market entry (more than 42.83months for a mean level of uncertainty in model 7),firms start requiring higher expected rates of returnon investment and, in turn, they favor waiting.

Figure 1 is a graphical representation of the jointeffect of industry resource accumulation lags anduncertainty on the predicted probability of invest-ment, using the coefficients estimated in model 7.Within (at least) one standard deviation away fromthe mean (29.59 � 13.22 months), slower resourceaccumulation lags have a positive impact on invest-ment likelihood, after which this effect is curbed.Uncertainty, through its interaction with invest-ment lags, skews the probability curve to the left:uncertainty encourages investment when firmshave enough time to exercise the option of aban-doning a project (i.e., when accumulation lags aresufficiently slow). In particular, with uncertainty atits maximum (level 4 in figure 1), slowing down thespeed of resource accumulation almost always in-creases firms’ investment probability, and the fac-

FIGURE 1Industry Resource Accumulation Lags and Predicted Investment Probabilitya

a Graph depicts logit model 7.


tors curbing the effect of time only kick in forextremely long resource accumulation lags (greaterthan 63.42 months with level 4 uncertainty).12

These results are robust to the alternative esti-mation with firm fixed effects (model 8). All co-efficients had the expected sign and, with the ex-ception of uncertainty (which had a negative,insignificant coefficient), they were significant.One possible explanation for this finding is thatmost of the uncertainty variance is region-basedand firms tend to invest within a single region (theUnited States, Europe, or Japan). Hence, includingfirm dummies serves as a natural surrogate for un-certainty’s role in capturing region-based variance.This conjecture finds some support in the fact thatincluding region dummies produces a similar out-come on the uncertainty coefficient.

Robustness Checks

Several checks confirmed the robustness of ourfindings. First, pooled and population-averagedlogit models produced identical results. Second,using an alternative operationalization of demanduncertainty, four years’ worth of production of pet-rochemicals (also standardized), did not change theestimations.

Third, three other possible measures of the re-source accumulation lag variable (see Appendix A)left the sign and significance of the model coeffi-cients intact, with the exception of the squaredterm of resource accumulation lag, as follows. Thesquared term coefficient was only negative and sig-nificant for the alternative measure of the resourceaccumulation lag variable that also included datafor the petrochemical subindustry with the slowestspeed of resource accumulation (71.1 months in thesubset for polystyrene in Japan). With the two otheroperationalizations (where data were missing forthis subindustry), the squared term came out insig-nificant. The reason for this result is that thesquared resource accumulation lag term curbs thepositive effect of time lags precisely for observa-tions on the upper tail of the distribution (e.g., the

Japanese polystyrene subindustry). Note, however,that all four resource accumulation lag measuresare highly correlated (� � 0.8). All these resultsremain unchanged if we use the industry-averagetime-to-build for the average amount of capacityadded in each subindustry instead of averages perunit of output.

Fourth, other operationalizations of the controlvariables (for a review, see Henderson [1998]) weretested with no major impact on the estimationresults.

Finally, several robustness checks were run ondifferent, though related, dependent variables: (1)the total number of investments per year (excludingthe rivals’ expansion independent variable, for ob-vious reasons), (2) the total amount invested incapacity (in a Tobit regression), and (3) the likeli-hood of investment in different types of capacityexpansions (in an ordered logit regression).13 Find-ings were identical.

DISCUSSION AND CONCLUSIONS

This study is the first in the strategy field to givea systematic empirical account of resource accumu-lation lags. According to the resource-based view,the speed at which firms accumulate resources isan important determinant of the length of imitationlags, the scope of lead-time advantages, and thesustainability of competitive advantage. Borrowingthe control variables of the solid empirical work onthe petrochemical industry (Gilbert & Lieberman,1987; Henderson & Cool, 2003a, 2003b; Lieberman,1987a, 1987b), we tested the robustness of a novelset of competitive strategy hypotheses concerningthe effect of time-consuming resource accumula-tion on firms’ investments under uncertainty and,in particular, the trade-off between commitmentand flexibility.

All the hypotheses we developed in this studyfound support in the data. Industry resource accumu-lation lags were found to have a nontrivial impact onfirms’ strategies. The longer it takes, on average, to

12 All curves in Figure 1 intersect when industry re-source accumulation lags are such that changes in uncer-tainty have no impact on the predicted probability ofinvestment. Because the main and interaction effects ofdemand uncertainty have similar coefficients of oppositesigns, the overall effect of uncertainty cancels out whenthe resource accumulation lag measure is close to 1 (i.e.,one standard deviation above the mean). More precisely:�6 �8 � resource accumulation lag � 0 N resourceaccumulation lag � –�6 / �8 � 0.97 (or 42.37 monthsprior to standardization; coefficients from model 7).

13 In the petrochemical industry, firms can expandcapacity in two main ways: by adding a greenfield plantor by adding a new incremental unit to an existing plant.In the investment data set, greenfield and incrementalexpansions represented 29 and 71 percent, respectively,of the total 879 plant construction projects observed in1975–95. Expansion type was set equal to 1 if a firmexpanded incrementally in a given year and equal to 2 if,alternatively, that firm expanded with a greenfield plantin that same year. Otherwise, the variable assumed thevalue 0. The ordered logit estimation confirmed the re-sults of the logit analysis.


accumulate resources to compete in a certain market,the more likely firms are to invest under uncertainty.This is because slower resource accumulation decel-erates firms’ reactions to new market or competitiveinformation, thereby increasing the potential forgoneprofits if firms wait to invest and market conditionsturn out to be favorable. At the same time, with longerresource accumulation lags, investment decisions de-pend on projections for a more distant future, a con-dition that not only increases the likelihood of ex-treme future market outcomes, but also givesmanagers more time to exercise the option of aban-doning their investment projects. Profits becomemore effectively safeguarded under poor businessconditions (e.g., demand or prices are low), and thenet present value of investing ex ante increases, pro-pelling firms to commit (invest).

A more subtle consequence of this line of rea-soning also found empirical support: with ex-tremely long resource accumulation lags, the an-ticipation of sufficiently large forgone profits expost increases the risk of early widespread in-vestment bandwagons, making it more difficultfor any particular company to get ahead of thecompetition by making an investment largeenough to preempt rivals. This situation ulti-mately reduces the incentives for commitmentand, thus, makes flexibility (waiting) more likely.Short-term stock market pressures reinforce thiseffect: if it takes too long to amass resources priorto market entry, managers will consider not in-vesting as alternative projects with shorter pay-back periods may be more financially attractive.Indeed, projects that yield quick cash recoveryenhance the reinvestment possibilities for firms.The positive effect of resource accumulation lagson commitment is thereby curbed, and the prob-ability of flexibility is increased, as evidenced inthe data. These empirical findings are consistentwith prior theoretical work (Bar-Ilan & Strange,1996; Narayanan, 1985a, 1985b; Pacheco-de-Almeida & Zemsky, 2003; Stein, 1988, 1989).

In short, resource accumulation lags generallyfavor commitment; flexibility is only preferred insituations in which resource accumulation isvery time-consuming (more than one standarddeviation above the mean in our sample).

Strategy Implications

Time is an important element of industry struc-ture. Often the structural characteristics of anindustry dictate the speed of firms’ resource accu-mulation; clearly, in some industries firms canaccumulate resources faster than in others. The rea-son to measure resource accumulation lags in a

study of industry analysis a la Porter (1980) isthreefold.

First, in industries with lengthy resource accu-mulation lags, competitive advantage may behard to attain. Contrary to conventional wisdom,competition may be fiercer in markets in whichfirms accumulate resources slowly than in thosein which accumulation is relatively rapid. Longresource accumulation lags generally offset un-certainty and reduce inertia, inducing firms tocommit and start racing for new market opportu-nities early. In addition, because time-consumingresource accumulation sustains competitive ad-vantage (Barney, 1991; Dierickx & Cool, 1989),firms face a critical trade-off: competitive advan-tage is harder to create precisely in those indus-tries in which it is easier to sustain. In otherwords, the benefits of slow imitation and leadtime advantages in industries with long resourceaccumulation lags are partly offset by the risk ofstiff competition. This finding is consistent withsystematic empirical evidence of chronic excesssupply in industries characterized by long invest-ment lags, such as commercial real estate, elec-tricity, and chemicals (Ghemawat, 1984; Hender-son & Cool, 2003b; Kling & McCue, 1987; MacRae,1989).

Second, our results contradict the view thatuncertainty is always a strong disincentive forinvestment. We have shown empirically that anincrease in uncertainty may encourage ratherthan dissuade commitment, owing to the positivemoderating effect of resource accumulation lagson uncertainty. Indeed, the longer it takes to planand bring new investments online, the more timefirms have to learn about conditions in a market(demand, prices, competition, technology, regu-lation, etc.) before commencing operations inthat market. Long resource accumulation lags areoften associated with staged investment projectsthat, should new unfavorable market informationbe revealed, firms can abandon before movinginto the next stage of the investment. Because theinvestment can be abandoned at multiple stages,its downside is truncated. An increase in uncer-tainty, therefore, augments the upside potentialof a project more than its downside, boosting itsexpected profits ex ante and leading not only to agreater but also to an earlier likelihood ofinvestment.

Third, the nonlinear inverted U-shaped effectof resource accumulation lags on commitmentimplies that flexibility is more valued in indus-tries in which resource accumulation is eithervery time-consuming or virtually instantaneous.In these settings, managers are more likely than


they are in other settings to exploit the option towait, postponing investments; thus, preemptivestrategies are less frequent.

Several normative managerial implications canbe derived from these empirical results. Contraryto conventional wisdom, our study suggests thatslowing down, rather than accelerating, firms’resource accumulation may lead to hypercompe-tition. Slower resource accumulation reduces im-itation pressures, but it also substantially in-creases the number of competitive investments inan industry. Managers should carefully weighthis trade-off between the pace of imitation andthe intensity of competition when decidingwhether to enter slow- or fast-moving industries.Managers of industry-incumbent firms shouldalso anticipate the competitive effects of struc-tural changes on the pace of an industry. Forinstance, stricter governmental regulations ormore complex technologies often slow downfirms’ resource accumulation, thereby increasingcompetition.

In addition, to the extent that the patterns ourresults reveal influence firms’ investment in newmarkets, the study suggests that first mover advan-tages may be easier to establish in fast-moving in-dustries. With faster resource accumulation, morefirms follow a wait-and-see strategy by postponinginvestments if initial market uncertainty is highand quickly catching up when market conditionsare favorable. This pattern of investment concedesa (small) timing advantage to the few firms thatdecide to enter an industry early. These early en-trants can preempt quick imitation by late moversby investing in excess capacity, patenting, andproduct proliferation. Preemptive strategies are lesseffective in slow-moving industries because morefirms enter the market early.

These implications for strategic managementalso apply in the context of international diver-sification, as it is reasonable to expect substantialdifferences in resource accumulation lags acrosscountries, likely related to variation in regulatoryenvironment. Governments have a major influenceon firms’ total times-to-market and, thus, can criti-cally influence the level of competition (e.g., over-capacity) in some industries over time.

Generalization and Limitations

The aforementioned implications of our resultsdo not apply uniformly to every industry. We stud-ied the petrochemical industry, a mature industrywith chronic excess capacity, and the reported ef-fects of resource accumulation lags on firms’ in-vestment strategies may be typical of such indus-

tries. In emergent and growing markets, potentialforgone profits from delaying investment tend to begreater, and the positive impact of resource accu-mulation lags may be even stronger. It is unclearhow other structural characteristics of an industry(e.g., level of product differentiation, rate of tech-nological breakthroughs, degree of investment se-crecy) change the effect of time lags on firms’ strat-egies under uncertainty. This observation maymotivate future work.

Another interesting direction for future re-search would be to examine how differences infirm speed within one industry change our keyresults regarding investment strategies. Moredata on time-to-build would need to be gathered,to increase variation in resource accumulationlags among firms and over time.

Finally, finding measures of the resource accu-mulation lag other than the time needed to buildproduction facilities might capture aspects of re-source accumulation that this study has over-looked. In manufacturing industries with rela-tively homogeneous products, establishing aproduction facility is often all a firm needs tocompete in existing markets. The key elements offirms’ strategies (such as technological innova-tion and production efficiency) are usually im-portant components of the investment process. Incontrast, in some service industries, the construc-tion of “production facilities” is not as vital forfirms’ strategies; in such industries, strategiesmight depend on the long-term cultivation of in-tangible assets such as corporate reputation orbrand image. For some industrial contexts, there-fore, the time-to-build of new plants is obviouslya less appropriate proxy for resource accumula-tion lags, and alternative measures need to beidentified.

REFERENCES

Adner, R., & Levinthal, D. A. 2004. What is not a realoption: Considering boundaries for the applicationof real options to business strategy. Academy ofManagement Review, 29: 74–85.

Ai, C., & Norton, E. C. 2003. Interaction terms in logit andprobit models. Economic Letters, 80: 123–129.

Ainslie, G., & Haslam, N. 1992. Hyperbolic discounting.In G. Loewenstein & J. Elster (Eds.), Choice overtime: 57–92. New York: Russell Sage Foundation.

Bar-Ilan, A., & Strange, W. C. 1996. Investment lags.American Economic Review, 86: 610–622.

Barney, J. 2001. Is the resource-based “view” a usefulperspective for strategic management research? Yes.Academy of Management Review, 26: 41–56.


Barney, J. B. 1986. Strategic factor markets: Expectations,luck, and business strategy. Management Science,32: 1231–1241.

Barney, J. B. 1991. Firm resources and sustained compet-itive advantage. Journal of Management, 17(1): 99–120.

Besanko, D., Dranove, D., Shanley, M., & Schaefer, S.2004. Economics of Strategy (3rd ed.). New York:Wiley.

Chapman, K. 1991. The international petrochemical in-dustry: Evolution and location. Oxford, U.K.: BasilBlackwell.

Chemical Insight. 1987. April, Mike Hyde Publications,London.

Cohen, W. M., Nelson, R. E., & Walsh, J. P. 2000. Protect-ing their intellectual assets: Appropriability con-ditions and why U.S. manufacturing firms patent(or not). NBER working paper no. 7552, NationalBureau of Economic Research, Cambridge, MA.

Cool, K., Almeida Costa, L., & Dierickx, I. 2002. Con-structing competitive advantage. In A. Pettigrew, H.Thomas, & R. Whittington (Eds.), Handbook of strat-egy and management: 55–71. London: Sage.

Dierickx, I., & Cool, K. 1989. Asset stock accumulationand sustainability of competitive advantage. Man-agement Science, 35: 1504–1511.

Dierickx, I., & Cool, K. 1994. Competitive strategy, assetaccumulation and firm performance. In H. Daems &H. Thomas (Eds.), Strategic groups, strategic movesand performance: 63–80. Oxford, U.K.: PergamonPress.

Dixit, A. K., & Pindyck, R. S. 1994. Investment underuncertainty: Princeton, NJ: Princeton UniversityPress.

Esty, B. C., & Ghemawat, P. 2002. Airbus vs. Boeing insuperjumbos: A case of failed preemption. HBSStrategy Unit working paper, Harvard BusinessSchool, Boston.

Fudenberg, D., & Tirole, J. 1983. Capital as a commit-ment: Strategic investment to deter mobility. Journalof Economic Theory, 31: 227–250.

Ghemawat, P. 1984. Capacity expansion in the titaniumdioxide industry. Journal of Industrial Economics,33(2): 145–163.

Ghemawat, P. 1991. Commitment. New York: Free Press.

Gilbert, R. J., & Lieberman, M. 1987. Investment and coor-dination in oligopolistic industries. RAND Journal ofEconomics, 18: 17–33.

Grant, R. M. 1991. The resource-based theory of compet-itive advantage: implications for strategy formula-tion. California Management Review, 33(3): 114–136.

Graves, S. B. 1989. The time-cost tradeoff in research anddevelopment: a review. Engineering Costs and Pro-duction Economics, 16: 1–9.

Hall, R. 1992. The strategic analysis of intangible re-sources. Strategic Management Journal, 13: 135–144.

Hannan, M. T., & Freeman, J. 1984. Structural inertia andorganizational change. American Sociological Re-view, 49: 149–164.

Hannan, M. T., Polos, L., & Carroll, G. R. 2004. Theevolution of inertia. Industrial and CorporateChange, 13: 213–242.

Henderson, J. E. 1998. Learning to coordinate and timecapacity expansion decisions: A longitudinal studyof the worldwide petrochemical industry, 1975–1995. Unpublished doctoral dissertation, INSEAD,Fontainebleau, France.

Henderson, J. E., & Cool, K. 2003a. Corporate governance,investment bandwagons and overcapacity: An anal-ysis of the worldwide petrochemical industry, 1975–95. Strategic Management Journal, 24: 349–373.

Henderson, J. E., & Cool, K. 2003b. Learning to timecapacity expansions: An empirical analysis of theworldwide petrochemical industry, 1975–95. Stra-tegic Management Journal, 24: 393–413.

Hitt, M. A., Keats, B. W., & DeMarie, S. M. 1998. Navi-gating in the new competitive landscape: Buildingstrategic flexibility and competitive advantage in the21st century. Academy of Management Executive,12(4): 22–42.

Itami, H. 1987. Mobilizing invisible assets. Cambridge,MA: Harvard University Press.

Kirkpatrick, D. 1997. Intel’s amazing profit machine. For-tune, 135: 60–66.

Kling, J. L., & McCue, T. E. 1987. Office building invest-ment and the macroeconomy: Empirical evidence.Journal of the American Real Estate and UrbanEconomics Association, 15(3): 234–255.

Koeva, P. 2000. The facts about time-to-build. IMFworking paper, International Monetary Fund, Wash-ington DC.

Kogut, B. 1991. Joint ventures and the option to expandand acquire. Management Science, 37: 19–33.

Krainer, R. E. 1968. The time profile of capital accumu-lation in the automobile industry. Journal of Politi-cal Economy, 76: 1049–1057.

Kulatilaka, N., & Perotti, E. C. 1998. Strategic growthoptions. Management Science, 44: 1021–1031.

Lieberman, M. B. 1987a. Excess capacity as a barrier toentry: An empirical appraisal. Journal of IndustrialEconomics, 35: 607–627.

Lieberman, M. B. 1987b. Postentry investment and mar-ket structure in the chemical processing industries.RAND Journal of Economics, 18: 533–549.

Lieberman, M. B., & Montgomery, D. B. 1988. First-moveradvantages. Strategic Management Journal, 9(spe-cial issue: Strategy content research): 41–58.


Lieberman, M. B., & Montgomery, D. B. 1998. First-mover(dis)advantages: Retrospective and link with the re-source-based view. Strategic Management Journal,19: 1111–1125.

Lippman, S. A., & Rumelt, R. P. 1982. Uncertain imita-bility: An analysis of interfirm differences in effi-ciency under competition. Bell Journal of Econom-ics, 13: 418–438.

Lowe, P. 1979. Investment for production: Managingthe plant investment process. New York: Wiley.

Luehrman, T. A. 1998. Strategy as a portfolio of realoptions. Harvard Business Review, 76(5): 89–99.

MacRae, K. M. 1989. Critical issues in electric powerplanning in the 1990s. Calgary, AB: Canadian En-ergy Research Institute.

Mansfield, E. 1971. Industrial research and technolog-ical innovation. New York: Norton.

Maritan, C. A. 2001. Capital investment as investing inorganizational capabilities: An empiricallygrounded process model. Academy of ManagementJournal, 44: 513–531.

Mayer, T. 1960. Plant and equipment lead times. Journalof Business, 33(2): 127–132.

Mayer, T., & Sonenblum, S. 1955. Lead times for fixedinvestment. Review of Economic Statistics, 37:300–304.

McGrath, R. G. 1997. A real options logic for initiatingtechnology positioning investments. Academy ofManagement Review, 22: 974–996.

McGrath, R. G. 1999. Falling forward: Real options rea-soning and entrepreneurial failure. Academy ofManagement Review, 24: 13–30.

McGrath, R. G., Ferrier, W. J., & Mendelow, A. L. 2004.Real options as engines of choice and heterogeneity.Academy of Management Review, 29: 86–101.

Miles, D. 1993. Testing for short termism in the UK stockmarket. Economic Journal, 103: 1379–1396.

Mishina, Y., Pollock, T. G., & Porac, J. F. 2004. Are moreresources always better for growth? Resource sticki-ness in market and product expansion. StrategicManagement Journal, 25: 1179–1197.

Narayanan, M. P. 1985a. Managerial incentives forshort-term results. Journal of Finance, 40: 1469 –1484.

Narayanan, M. P. 1985b. Observability and the paybackcriterion. Journal of Business, 58: 309–323.

Pacheco-de-Almeida, G., & Zemsky, P. 2003. The effectof time-to-build on strategic investment under un-certainty. RAND Journal of Economics, 34: 166 –182.

Pacheco-de-Almeida, G., & Zemsky, P. 2007. The timingof resource development and sustainable competi-tive advantage. Management Science, 53: 651–666.

Penrose, E. T. 1959. The theory of the growth of thefirm. New York: Wiley.

Peteraf, M. A. 1993. The cornerstones of competitiveadvantage: A resource-based view. Strategic Man-agement Journal, 14: 179–191.

Porter, M. E. 1980. Competitive strategy. New York: FreePress.

Priem, R., & Butler, J. 2001a. Is the resource-based “view” auseful perspective for strategic management research?Academy of Management Review, 26: 22– 40.

Priem, R., & Butler, J. 2001b. Tautology in the resource-based view and the implications of externally deter-mined resource value: Further comments. Academyof Management Review, 26: 57–65.

Reed, R., & DeFillippi, R. J. 1990. Causal ambiguity, barriersto imitation, and sustainable competitive advantage.Academy of Management Review, 15: 88–102.

Ross, S. A., Westerfield, R. W., & Jaffe, J. 1996. Corporatefinance. Mahwah, NJ: Irwin.

Rumelt, R. P. 1984. Towards a strategic theory of the firm.In R. B. Lamb (Ed.), Competitive strategic manage-ment: 556–570. Englewood Cliffs, NJ: Prentice-Hall.

Scherer, F. M. 1967. Research and development resourceallocation under rivalry. Quarterly Journal of Eco-nomics, 3: 367–391.

Schroeder, R. G., Bates, K. A., & Junttila, M. A. 2002. Aresource-based view of manufacturing strategy andthe relationship to manufacturing performance.Strategic Management Journal, 23: 105–118.

Shimizu, K., & Hitt, M. A. 2004. Strategic flexibility:Organizational preparedness to reverse ineffectivestrategic decisions. Academy of Management Exec-utive, 18(4): 44–59.

Spence, A. M. 1979. Investment strategy and growth in anew market. Bell Journal of Economics, 10: 1–18.

Stein, J. C. 1988. Takeover threats and managerial myo-pia. Journal of Political Economy, 96: 61–80.

Stein, J. C. 1989. Efficient capital markets, inefficientfirms: A model of myopic corporate behavior. Quar-terly Journal of Economics, 104: 655–669.

Stigler, G. J. 1939. Production and distribution in theshort run. Journal of Political Economy, 47: 305–327.

Szulanski, G. 2003. Sticky knowledge: Barriers to theknowing firm. Thousand Oaks, CA: Sage.

Teece, D. J., Pisano, G., & Shuen, A. 1997. Dynamiccapabilities and strategic management. StrategicManagement Journal, 18: 509–533.

Trigeorgis, L. 1996. Real options. Cambridge, MA: MITPress.

Villalonga, B. 2004. Intangible resources, Tobin’s Q, andsustainability of performance differences. Journalof Economic Behavior and Organization, 54:205–231.


APPENDIX A

Alternative Measures of Resource Accumulation Lag

Different projects were initially reported in the OGJin varying stages of completion.14 Directly averaging aproject’s time-to-build per product-region irrespectiveof the phase during which the project was first re-ported would have biased the industry averages. Ob-servations for projects that were initially in the studyor planning phase were more complete than those inwhich time-to-build was not measured until the expan-sion had reached the engineering or constructionstages. A “right-censoring” problem in the time-to-build measures was associated with the 257 projects

first reported in the engineering and constructionphases. Classifying these right-censored data as miss-ing observations was also not advisable: the missingsubsample was not random, and several time-to-builddata points would have been lost.15 Hence, we com-puted several measures of the resource accumulationlag, values for which are shown in the tables below. Weused (1) only the 97 time-to-build observations first

14 The OGJ reports on projects twice per year. Whenthe 556 expansion projects were reported in the OGJ forthe first time, 97 were under study or being planned, 85in the engineering stage, and 172 already under construc-tion. Of the remaining 202 expansions, only 28 includedsome information on project status in a later issue of thejournal.

15 A dummy variable with two groups, cases withright-censored time-to-build observations and cases withclean time-to-build observations, was constructed. Weperformed parametric and nonparametric tests of meandifferences for the remaining continuous data set vari-ables; results allowed us to reject the null hypothesis ofequal means. Thus, the right-censored subsample wasnot random, and omitting the limit observations wouldhave created biases.

TABLE A1Resource Accumulation Lags for Projects First

Observed at a Study or Planning Stage


Ethylene 20.5 19.6 52.8HDPE 21.0 65.1 57.6LDPE 28.3 40.1 82.9LLDPE 8.9 23.2 39.3PP 61.4 64.4 43.5PS 43.7 124.2PVC 29.1 48.5 73.0Styrene 5.0 34.4 24.8VCM 8.4 30.8 33.2

TABLE A2Resource Accumulation Lags for All Projects

with Initial Information on Project Status, withAdjustment for Right Censoring


Ethylene 22.1 22.0 16.0HDPE 22.4 48.5 49.6LDPE 18.3 38.2 51.6LLDPE 16.2 32.5 41.4PP 36.1 38.4 48.8PS 66.6 51.1 71.1PVC 59.7 53.2 96.2Styrene 22.7 30.0 28.6VCM 10.1 27.6 20.4

TABLE A3Resource Accumulation Lags for the Fitted

Values of a Right-Censored Regression


Ethylene 29.9 40.9 27.5HDPE 61.4 81.1 97.4LDPE 25.3 67.4 83.4LLDPE 33.8 23.0 46.5PP 37.6 57.6 67.1PS 43.7 103.5PVC 71.6 78.9 96.2Styrene 21.4 30.2 27.7VCM 10.1 30.9 34.0

APPENDIX BPearson Correlation Coefficients of the

Independent Variables

Variable 1 2 3 4 5 6

1. Demand growth2. Excess capacity �.15*3. Investment

lumpiness.17* .07*

4. Market share �.06* .04* �.06*5. Rivals’ expansion .33* �.16* .07* �.07*6. Demand

uncertaintya.20* .08* .08* .04* .06*

7. Accumulation lagb .02* .20* �.16* .02 �.02 .02

a Standardized variable; no multicollinearity problems withthe interaction or squared terms.

b Months per 100,000 MT/year.* p � .05


reported when they were in study or planning phases(Table A1) and then (2) only the 354 time-to-buildobservations in which we adjusted the 257 right-cen-sored data points by adding the average time spent inthe study and planning stages in the correspondingsubindustry (Table A2). The third measure calculatedwas the one discussed in the body of the article andreported in Table 3; it consists of the complete 556time-to-build observations in which the 257 right-cen-sored observations were corrected in the same way asthose in Table A2. A final, fourth data sample includedthe 354 fitted values of a right-censored exploratoryregression of the determinants of time-to-build at thefirm level (Table A3).16 For simplicity, all measureswere scaled up to 100,000 MT/year of capacity.

Goncalo Pacheco-de-Almeida ([email protected])is an assistant professor of management and organiza-tions at the Leonard N. Stern School of Business, NewYork University. His research focuses on the foundationsof strategy dynamics and the effect of time-based compe-tition and the timing of resource development on firmcompetitive advantage. He received his Ph.D. in strategicmanagement at INSEAD.

James E. Henderson ([email protected]) is a pro-fessor of strategy at the International Institute for Man-agement Development (IMD) based in Lausanne. His re-search focuses on strategy formulation and execution incapital-intensive and high-technology industries. Hismost recent studies concern timing and bandwagon ef-fects in capacity expansion decisions in the petrochem-ical industry, corporate venture capital management andperformance in the telecommunications and pharmaceu-tical industries, and cluster competitiveness in the wineindustry. He received his Ph.D. in strategic managementat INSEAD.

Karel O. Cool ([email protected]) is the BP ChairedProfessor of European Competitiveness and a professor ofstrategic management at INSEAD, Fontainebleau, and avisiting professor at Northwestern University. His re-search focuses on problems of competitive analysis (e.g.,industry overcapacity, critical mass races, the accumula-tion dynamics of resources, and the economics of re-source and product substitution). He received his Ph.D.in strategic management from Purdue University.

16 The regression equation used the 354 under-study,planned, engineering, and construction observations. Weregressed time-to-build on a measure of project capacityand other controls (firm expansion experience, expan-sion type, a year trend, and product-region dummies,among others), taking into account which observationswere right-censored and that the true censoring valuewas unknown and could change from observation toobservation. This regression was run using the “intreg”command in Stata 8.2.


Date post:	01-Apr-2021
Category:	Documents
Upload:	others
View:	3 times
Download:	0 times

RESOLVING THE COMMITMENT VERSUS FLEXIBILITY TRADE … 2019/Pacheco et al (2008).pdfpreempted by...

Documents