real options reasoning and a new look at the r&d investment

Strategic Management JournalStrat. Mgmt. J., 25: 1–21 (2004)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.358

REAL OPTIONS REASONING AND A NEW LOOK ATTHE R&D INVESTMENT STRATEGIES OFPHARMACEUTICAL FIRMS

RITA GUNTHER McGRATH* and ATUL NERKARGraduate School of Business, Columbia University, New York, New York, U.S.A.

Real options reasoning (ROR) is a conceptual approach to strategic investment that takes intoaccount the value of preserving the right to make future choices under uncertain conditions. Inthis study, we explore firms’ motivations to invest in a new option. We find, based on an analysisof a large sample of patents by firms active in the pharmaceutical industry, that their investmentsin R&D are consistent with the logic of ROR. We identify three constructs—scope of opportunity,prior experience, and competitive effects—which have an influence on firms’ propensity to investin new R&D options and which could usefully be incorporated in a strategic theory of investment.Copyright 2003 John Wiley & Sons, Ltd.

Strategic management research has long focusedon understanding differentials in performanceacross firms (Helfat, 2000). An important expla-nation for observed differentials in performanceis the ability of firms to introduce innovationsthat permit them to garner rents. Broadly con-strued, innovation consists of the development ofnew products, new processes, and/or new markets(Schumpeter, 1934). The rents attainable throughinnovative activities are often termed ‘Schumpeter-ian’ or ‘entrepreneurial’ rents, because they are therewards to firms who are prepared to act in theface of ex ante uncertainty (Knight, 1921; Rumelt,1987). While innumerable empirical studies havelinked innovation to superior performance (e.g.,Lawless and Anderson, 1996; Christensen, 1997)and many scholars have offered theoretical insightson decision making surrounding innovation withinfirms (March, 1991; Van de Ven, 1986; Dougherty,

Key words: real options; pharmaceutical industry; patent-ing; innovation*Correspondence to: Rita Gunther McGrath, Graduate School ofBusiness, Columbia University, Armstrong Hall, 4th Floor, 2880Broadway, New York, NY 10025, U.S.A.

1992), little empirical research integrates norma-tive theories developed largely in finance, andbehavioral theories emerging from organizationsresearch regarding the drivers of decisions to investin innovation.

In this paper, we use real options reasoning(ROR) to bridge these theories, by examininginnovation investment decisions that take a firminto new technological areas. We characterize theinitial foray into a new technological area as areal option, which creates a somewhat proprietaryopportunity for the investing firm to make laterdecisions, such as to further exploit or to exit thearea (Kim and Kogut, 1996). The question weaddress in this paper is: Why do firms take outreal technology options?

A real options perspective offers a comple-mentary approach to normative models of invest-ments under uncertainty borrowed from the fieldof finance (Fama and French, 1992, 1993, 1995,1996) and behavioral theories of decision making(March, 1991). Where finance theories are basedon assumptions of efficient markets and static equi-librium, ROR presumes information asymmetries,

Copyright 2003 John Wiley & Sons, Ltd. Received 17 December 2001Final revision received 9 June 2003

2 R. G. McGrath and A. Nerkar

path-dependent accumulation processes, and uncer-tainty (Miller, 1998). As many scholars haveobserved, assumptions borrowed from the field offinance are often at odds with managerial realities,leading to lack of support for hypothesized rela-tionships and to observed inconsistency in manage-rial application (Dixit, 1992). Chatterjee, Lubatkin,and Schulze (1999) suggest that a key reasonsuch inconsistencies occur is that the markets inwhich firms compete ‘are not as perfect’ as theassumptions underlying models from finance mightsuggest. Nonetheless, powerful ideas from financeoffer an economic rationale for observed manage-rial behavior.

Kogut (1983, 1991), Bowman and Hurry (1993),and Sanchez (1993) were early advocates of resolv-ing the tension between the rationality of finance-based models and the observations of behavioralresearchers through an options lens. As they pointout, options reasoning accommodates the value offlexibility, differing resource allocation horizons,the process of retrospective sense making, andpath dependence. In this paper, we offer supportfor an options perspective by showing that theinvestment strategies of a group of firms—in thiscase, pharmaceutical firms—are consistent withROR. By ‘real options reasoning,’ we mean toimply that decision-makers implicitly (or explic-itly) respond to the value of the right to preservedecision rights in the future in their investmentchoices. We believe that ROR can explain some ofthe differences between actual managerial invest-ment behavior and theorized investment behavior.Our goal is to advance a theory of investmentappropriate to the strategy field.

In short, we seek to explore whether decision-makers in pharmaceutical firms demonstrate RORin R&D investments. We follow an establishedbody of scholarly work that identifies certaininvestment decisions as amenable to ROR, thenexamines evidence for whether actual investmentpatterns are consistent with the predictions of realoptions theory. Such analyses have been conductedwith respect to joint ventures (Kogut, 1991), inter-national entry decisions (Chi and McGuire, 1996;Reuer and Leiblein, 2000), governance choices incollaboration (Folta, 1998), foreign direct invest-ment decisions (Kogut and Kulatilaka, 1994), andR&D investments (Kumaraswamy, 1996). If wefind such evidence, it suggests to us that ROR hasmuch to contribute to a theory of investment inthe field of strategy. The research reported here

builds upon a growing literature that offers empir-ical evidence to test propositions originating fromROR (such as Hurry, Miller, and Bowman, 1992;Folta, 1998; and Kumaraswamy, 1996).

We studied patterns of R&D patenting for allparticipants in the U.S. pharmaceutical industryover a period of 17 years (from 1979 to 1995),with emphasis on the strategies of the leading 31firms. Our focus was on identifying influences onthe decision of firms to invest in growth options.We operationalized the decision to take out optionsin an R&D arena by measuring patents in phar-maceutical subclasses, on the premise that with-out investment input the knowledge creation thatleads to patenting outputs would be impossible.We believe the study to be timely, for althoughROR has attracted considerable scholarly attention,empirical research is still rather sparse.

REAL OPTIONS REASONING ANDSTRATEGIC INVESTMENTS

In recent research there has been intense inter-est in understanding how ROR might usefullycomplement net present value-based approachesto investment decisions (Mitchell and Hamilton,1988; Nichols, 1994a; McGrath, 1997; Trigeor-gis, 1996; see also Bettis and Hitt, 1995). A fur-ther incentive is for scholars to develop modelsthat more closely align with managerial practice.Brennan (1995: 17) observed in a retrospective ofdevelopments in finance that a general trend in cor-porate finance is a ‘shift away from attempts toprescribe normative rules for decision-makers thatwould assist them to take decisions that are opti-mal from the point of view of shareholders andtowards attempts to describe more realistically theway that decisions are actually made’ (emphasis inthe original).

An option creates value by generating futuredecision rights. The theory of real options, inwhich the option in question is a real asset,is derived from theories originally developed infinance to account for the value of financial optionscontracts (Black and Scholes, 1973). A financialoption contract conveys the right, but not the obli-gation, on the purchaser to either buy or sell anunderlying asset at some point in the future. Byanalogy, an investment in a real option conveys theright, but not the obligation, for a firm to make fur-ther investments or defer such investments. Thus,

Copyright 2003 John Wiley & Sons, Ltd. Strat. Mgmt. J., 25: 1–21 (2004)

Real Options and Pharmaceutical R&D 3

a firm that invests in R&D can create a unit ofnew knowledge. It can secure its claim to com-mercialize this knowledge through patenting. Sub-sequently, it may elect to proceed to extend theknowledge, commercialize its knowledge, to donothing with it, or to seek to leverage the knowl-edge in some other way, for example by sharing itwith a joint venture partner or licensing it out.

Investing in real options can allow firms accessto a greater variety of opportunities than wouldbe possible if each investment represented a full-scale launch. The cost of an option on an assetis small, relative to the cost of purchasing theasset. Thus, with the same resources to spend,more opportunities can be explored using options.Once uncertainty is reduced, an investor can thenelect to exercise only those options that are ‘inthe money’ and allow the remainder to expire.By investing relatively small amounts in learningabout several promising technical directions simul-taneously, a firm can broaden the range of alter-natives it can apprehend, generating conceptualvariety with parsimony. As Gavetti and Levinthal(2000: 115–116) point out, ‘If low-outcome drawscan be costlessly discarded, then greater variancein the sample, holding the mean constant, increasesthe expected value of those draws that are adopted.This is the basic intuition behind the recent inter-est in the idea of “real options” in the businessstrategy literature.’

Expanding the number of trials that can beundertaken when uncertainty is high is valuablefrom a strategic point of view. As Chatterjeeet al. (1999: 560) point out, ‘strategy is aboutmaking resource commitments before the relation-ship between these commitments and their poten-tial performance outcomes are fully understood.’Options, they suggest, are an important mecha-nism through which firms reduce the strategic riskof making commitments. From their argument, wecan identify constructs that are relevant to a strate-gic theory of investment incorporating ROR. Here,we will focus on variance or scope of the oppor-tunity, portfolio or prior experience effects, andcompetitive effects in the R&D investment deci-sion. The logic for selecting these constructs isthat variance is relevant to the individual invest-ment proposition, portfolio effects to the activitiesof the firm as a whole, and option expiration to theviability of the underlying asset on which optionsare taken in future competition. They thus reflect

the influence of factors at three levels of analysisrelevant to the value of a real option.

Scope of opportunity

Strategic decisions with respect to entry into newmarkets have at their core firm-specific risk. Thecore question is whether an investment or entryinto a new area has the potential to help a firmprotect its future earnings streams from industryand macro-economic pressures reflected in mar-ket portfolio returns. If such entry contributes tocompetitive advantage, investment in the area isvaluable, a conclusion that is not controversial(Porter, 1980). However, it is important to considerthe scope of opportunity, potential access to otheravenues of growth or, in other words, the varianceunderlying such investments. For financial options,an increase in the scope of the opportunity repre-sented by the volatility of the stock on which anoption contract is written leads to an increase inthe value of the option. This occurs because theinvestment in the option is fixed at the price of theoption, giving the investor access to a greater rangeof potential outcomes on the upside, while con-taining exposure on the downside. This effectivelytruncates the left-hand tail of a performance distri-bution, creating a performance distribution curvethat is skewed to the right, yielding asymmetricpay-offs. The real options analogue is that providedthe downside loss an organization would sustain ifit elects to stop further investment in a technologyarea is contained, its investments increase in valuewith increases in variance of results (Mitchell andHamilton, 1988). Thus, the presence of poten-tially high variance in performance outcomes foran investment should be positively associated withinvestment using ROR, where they might depressthe value of an investment using other approaches(Morris, Teisberg, and Kolbe, 1991).

In an argument consistent with this line ofthought, Dixit and Pindyck (1994) have suggestedthat a test for the presence of ROR in R&D strategyis whether those making investment choices favorexploratory research, with a scope well beyondthat of current activities. Exploratory research islikely to show the potential for opening up newlines of scientific inquiry in a significant manner,increasing the potential variance of returns. Suchincrease in scope can potentially lead to a radi-cal departure from prevailing practice. The greater



the scope of an opportunity area, the greater is thepotential upside of an uncertain investment.

The following hypothesis is consistent with ourexpectations that the presence of ROR in invest-ment decisions will be reflected in a propensity toengage in more exploratory R&D:

Hypothesis 1: There will be a positive associ-ation between a firm’s propensity to take outgrowth options in a new technological area andthe scope of the opportunity area.

Prior experience

In large pharmaceutical firms, the R&D functionencompasses many projects across many therapeu-tic and scientific areas. Decisions about makinginvestments in R&D are therefore always madein the context of a portfolio of other, compet-ing investments. As we observed earlier, makinginvestments in options implies that an investor canexplore a larger number of possibilities than wereeach exploratory foray to be a full launch. But doesROR offer any insight into trade-offs decision-makers must make when considering alternativeinvestment proposals?

Conventional investment theory suggests thatfirms should proceed to invest in all projects witha positive net present value. If resources for invest-ment are limited, as would be the case undercapital rationing, the conventional wisdom is toestablish a discounted cash flow valuation for eachproject independently. The next step is to eval-uate which combinations of affordable projectswould yield the firm the highest total net presentvalue, then invest in this combination of projects.The assumptions underlying this approach are thatprojects can be evaluated independently, and thattheir joint effect on future value is largely additive(although correct application of the NPV rule isbased on incremental cash flows).

ROR allows instead for an interaction amonginvestments that changes the value not only ofa given investment but also of the other optionswithin a firm’s R&D portfolio. Because optionsinteract, making a decision with respect to oneoption affects the value of other options. Forinstance, subsequent option investments in R&Darenas can increase the value of options openedearlier, because they lead to increase in the valueof the underlying knowledge as such knowledge istypically created in a path-dependent cumulative

manner (Nelson and Winter, 1982). Lack of con-tinued investment in an arena can also diminishthe value of earlier options because this terminatesthe learning process, effectively extinguishing theearlier option (Dierickx and Cool, 1989; Trige-orgis, 1996: 257–258). Because options withina portfolio may not be additive, and becauseopening options consumes resources, adding moreoptions to an existing portfolio does not necessar-ily increase the value of the portfolio. Real optionsthat are compound options often involve nestedinvestments such as entry into new technologi-cal areas (Miller and Folta, 2002). For instance,patents in new areas can be considered as com-pound options that require additional investmentsbefore the value of the original option is real-ized. This is in stark contrast to the net presentvalue approach, which largely assumes indepen-dence and additivity for each project.

Present value analysis also does not explicitlyrecognize the often positive effects on the valueof an asset of deferring investment or of the riskof loss of value due to option expiration. As wementioned, the decision to open or not open anoption has value under conditions of uncertainty.The value of waiting, however, is sensitive to theextent to which contingencies are highly uncertain,and the claim of a firm on the underlying real assetcan be maintained. As uncertainty is reduced overtime, the benefits of waiting relative to investingor abandoning decrease. Trigeorgis (1996: 241),for instance, suggests that the value of the option ifnot exercised is subject to diminishing returns withthe passage of time. This insight has importantportfolio implications. For a decision-maker usingROR, the value of a new option will be influencedby the presence of preexisting options. Moreover,its contribution to a firm’s existing portfolio ofR&D opportunities depends to a large extent onthe firm’s prior experience and the declining valueof waiting to exercise previously created options.

Thus, if a firm has made extensive investmentsin opening options, its managers should not beanxious to open still more options. Rather, theywill be inclined to invest in better understandingthose options they have already created, becausethe value of their option to defer exercise willbegin to decline if they do not. Regardless of theinherent attractiveness of a new opportunity, if afirm’s resources are already engaged in the pursuitof other attractive opportunities, it is likely to valuenew ones less. This is consistent with March’s



(1991) proposition that firms need to incorporatea balance of exploration and exploitation-orientedinvestments in their knowledge-creating strategies.We thus suggest that extensive prior commitmentis likely to dampen enthusiasm for subsequentinvestment, as in this hypothesis:

Hypothesis 2a: There will be a negative associ-ation between firm propensity to take out growthoptions in new technological areas and theextensiveness of its commitments to new areasin the past.

A second, portfolio-related characteristic consistsof the firm’s experience or previous track recordof investing in new areas. Strong incentives existfor learners to deepen their understanding of areasin which they have gained initial experience. Thisis due to increases in absorptive capacity thatmake the assimilation of each new unit of knowl-edge easier, faster, and less expensive (Cohen andLevinthal, 1990). Further, self-reinforcing behav-iors cause the level of competence with an exist-ing technology to exceed the level of compe-tence a learner could quickly achieve with a newtechnology, however potentially superior the newone might be (Levitt and March, 1988; Lane andLubatkin, 1998). With respect to R&D invest-ments, Sørensen and Stuart (2000) find empiricalevidence supportive of the idea that as firms ageand gain experience with the knowledge genera-tion process they pursue more innovations. Otherresearchers have suggested that this effect is mostpronounced when the new innovations sustainprevious technological trajectories for establishedfirms (Christensen, 1997; Henderson and Clark,1990).

Extending the argument we developed above, afirm that has opened options on an area will haveincentives to persist in making investments in thatarea. Because the value of an option is subject todiminishing returns, a firm cannot indefinitely putoff making a decision with respect to the new areawithout the value of its existing options eroding.Further, the decision to divert resources from anoption (once opened) to pursue another opportu-nity effectively decreases the value of the originaloption in two ways. First, it truncates the poten-tial of the investing firm to idiosyncratically reduceuncertainty and claim an underlying asset for itself.Second, the decision to divert resources impliesthat the routines accumulated to date in exploring

the option will be forgotten, essentially leadingthe option to expire. The effect is to diminish thefirm’s incentive to invest in attractive new areasbecause it is already committed to areas it had pre-viously explored. We can state this argument as ahypothesis:

Hypothesis 2b: There will be a negative asso-ciation between a firm’s propensity to take outgrowth options in a new technological area andits cumulative investment experience in previoustechnological areas.

Competition

The third construct we consider in our explorationof ROR concerns the effects of competition on afirm’s propensity to invest in new options. Attrac-tive opportunities tend to attract competition. Thiscreates a tension for the decision-maker. On theone hand, options on more attractive spaces arelikely to be more valuable while, on the other,competitors are likely to also perceive the oppor-tunity and seek to take out their own options.Further, a key premise of ROR is that investmentsare sequential. Failing to take out an investmentwhen others are investing creates the potential forsubsequent lock-out.

The use of ROR should be reflected in the wayin which this tension is resolved (Dixit, 1992).Decision-makers from different firms are likelyto have different expectations with respect to theworth of an emerging arena. The behavior of com-peting firms is then interpreted as either supportingor disconfirming these expectations. If a decision-maker observes that no other firm is taking outoptions in an area that he or she has independentlyconcluded is attractive, one reasonable inferenceis that the judgment was excessively optimistic,which decreases the incentive to invest. If noother organizations take out options, it makes senseto postpone investment. Conversely, let one firminvest and decision-makers from other firms arelikely to revise their expectations upward, creatingwhat Dixit (1992: 119) calls a ‘bunching’ of invest-ment. Investment by one thus provokes investmentby many (see also Trigeorgis, 1991).

Because knowledge development is a cumula-tive process, competitive entry into an area is notonly a signal of its attractiveness, but actually doesmake the arena more valuable by increasing thetotal investment in knowledge creation across all



actors in the area. Hambrick, MacMillan, and Bar-bosa’s (1983) conclusion that total investment mat-ters more than relative investment to technologicalsuccess is supportive of this argument. When anew technological area is just emerging, it is inthe collective interest of all participants to investto create a substantive body of underlying scientificknowledge, and even to share this knowledge withone another (see McGrath and McGrath, 2001, fora recent review of the issues involved in knowl-edge spillovers).

Singly and collectively, firm-level investmentsin R&D thus help to deepen knowledge in a field,reducing uncertainty. Dixit and Pindyck (1994)refer to the taking of such options in uncertain newareas as investments to reduce ‘technical uncer-tainty.’ This uncertainty cannot be reduced bypostponing investment, unlike uncertainties thatmight have to do with purely exogenous factorssuch as the cost of raw materials inputs. High lev-els of technical uncertainty thus create pressure toinvest.1 We capture this set of arguments in thefollowing hypothesis:

Hypothesis 3: There will be a positive associa-tion between firm propensity to take out growthoptions in new technological areas and compe-tition in those areas.

RESEARCH METHODS

Research setting: The pharmaceutical industry

Investment in pharmaceutical R&D has often beencharacterized in the literature as investment in thecreation of real options (Dixit and Pindyck, 1994;Trigeorgis, 1996: 1–30; Mitchell and Hamilton,1988; McGrath, 1997). Indeed, it is precisely theuncertainty, long time horizons and asymmetricpay-off distributions to investments in R&D thathave prompted considerable scholarly and prac-tical interest in furthering the development ofreal options theory for pharmaceutical R&D (forinstance, at Merck; Nichols, 1994b).

The pharmaceutical industry is attractive as asetting in which to test ideas about ROR. R&D

1 Increasing competition typically leads to decreases in profitpotential in the context of product markets (Porter, 1980) butnot necessarily in technology markets (Arora, Fosfuri, and Gam-bardella, 2001). In such cases, using real options as a preemptivemechanism might be a useful strategy. See Miller and Folta(2002) for a detailed exposition of these and related issues.

investment in pharmaceuticals can be assessedby reference to published patents. Patents are animportant source of technological advantage ingeneral but more specifically in the pharmaceuticalindustry (Levin et al., 1987), with the consequencethat it is a reasonable assumption on our part thatinvestment in an area is reflected by the patentsgranted to a firm in that area. Successful patentingis a precursor to product development in pharma-ceuticals. The upside of a successful discovery thatbecomes a new drug is assured, to some extent, bythe intellectual property protections in the patent.We use patent data to assess the propensity to makeR&D investments by firms, consistent with theresearch efforts of other scholars who have usedpatents to measure R&D knowledge (Ahuja, 2000;Silverman, 1999; Dutta and Weiss, 1997; Hender-son and Cockburn, 1994; Jaffe, Trajtenberg andHenderson, 1993; Engelsman and van Raan, 1994).

There is considerable precedent for our treat-ment of patents as real options. Pakes (1986),for instance, uses option valuation to assess thepropensity of patent holders to exercise or allowtheir patents to expire. Patents align with the cri-teria for a real option. Because the pay-off tolater commercialization is uncertain, their perfor-mance distribution is unknown. Because taking outa patent does not commit the firm to follow-oncommercialization, a firm can control the poten-tial downside loss and make sequential decisions,creating the asymmetric distribution of potentialreturns that is characteristic of real options. Inshort, a patent confers on the firm the right but notthe obligation to make further investments, culmi-nating in a decision whether to commercialize itsknowledge or not. Investments made towards com-mercializing the knowledge underlying the patentare analogous to the exercise price on the realoption.

Operationalizing options in R&D investmentdecisions

A core issue for our empirical research is distin-guishing among those decisions that might resultfrom random chance or luck and those that resultfrom strategic decision making (Alchian, 1950;Barney, 1986). We have operationalized invest-ment in an option as the granting of a sec-ond patent to a firm in a technological area thatwas previously new to it. One could argue thatobtaining a first patent in a new area could arise



because of luck. The patent might simply be aproduct of the uncertain stochastic processes thatunderlie success and failure in R&D (Levinthal,1990; Nelson and Winter, 1982). Irrespective ofwhether taking out a first patent can be viewed asan option or not, a second patent in a new areais prime evidence of the initiation of a patternof investment in that area, indicates a firm-levelcommitment to that area and is a much strongerindicator of a deliberate choice to focus there thana first patent.

We collected data for the period 1971–95, butwe used only the data in our sample from 1979 to1995, a period of 17 years, to test our hypothe-ses. The data we collected from 1971 to 1979were used to help alleviate the problem of leftcensoring by ensuring that we did not systemati-cally omit patents whose precursor investment tookplace before our period of observation began, orcount third patents as second patents because afirst patent was granted in the period before ourobservation period began.

We defined a technological area in a way that isconsistent with U.S. Patent Office Classificationsand earlier research (Shane, 2001; Ahuja, 2000;Rosenkopf and Nerkar, 2001). Each patent con-tains extensive information about the inventor, thecompany to which the patent is assigned, the tech-nological antecedents of the invention in the formof other patents that it cites, and the claims eachpatent proposes for its technical contribution. Theabove information can be accessed in computer-ized form. Every patent is assigned to a three-digittechnical class. Within each three-digit class thereare many subclasses. In this study, we define atechnical area as the three-digit subclass withinthe 514 area (i.e., drugs and bio-affecting compo-sitions) as the pharmaceutical class (Penner-Hahn,1998).

The unit of analysis is the individual patent andits associated content, and the level of the analysisis the firm. We consider only patents filed in theUnited States. The sources for this informationinclude the U.S. Patent and Trademark ExaminersOffice and online databases such as Lexis-Nexis

and Derwent.

Focal organizations

Our research question relates to the strategicchoices made by firms. In the pharmaceuticalindustry, 31 firms account for 80 percent of

all drug development and patenting in the phar-maceutical sector. The remaining actors includesmaller start-up organizations, universities, inde-pendent researchers, and a variety of other firms(for instance, those not competing in pharmaceuti-cals at all). For purposes of our study, the invest-ment decisions of greatest interest are those involv-ing the 31 major players in pharmaceuticals. Thedata set on which we test our hypotheses consistsof 45,757 patents established by these 31 firmsfrom the 17-year period from 1979 to 1995. Eachof the 31 firms lists as their primary StandardIndustrial Classification (SIC) code Number 2834(pharmaceutical) and each had at least 100 patentsgranted to them during the time frame of the study.By examining these firms, we capture the major-ity of activity taking place in the pharmaceuticalsector.

The results of our analysis may not generalizeacross smaller firms with fewer than 100 patents. Itis important for the reader to realize this, becausesmaller firms are regarded as likely to undertakemore exploratory forays in the development ofnew technical knowledge, have fewer resources,and are apt to be more limited in the extent towhich they open new options. During the periodof the study, based on the 47,757 patents, thefirms in our sample took out 7654 first patentsand 4013 second patents in the same technicalarea (patent subclass). Our sample of options thusconsists of 4013 observations of second patents intechnological areas new to the firms taking out thepatents. A sample data point is shown in Figure 1.

Dependent variable

The dependent variable for this study is the propen-sity of a firm to take out a new option, which isoperationalized as the instantaneous probability orhazard rate of taking out a second patent in a patentsubclass that is new to the firm (i.e., it has onlyone previous patent in a pharmaceutical subclassthat it had not patented in before).

Independent variables

Scope of opportunity

The scope of an investment in a second patentis measured by the potential of the opportunitypresented by the first patent for a firm. The greater



1971

1979

1995

1st p

aten

t in

clas

s 51

4#22

0D

ec 1

2, 1

980

2nd P

aten

tcl

ass

514#

220

June

26,

198

1

Pate

nts

exam

ined

in th

is p

erio

d to

ens

ure

that

all

entr

ies

afte

r 19

79 a

re ‘

real

’ en

trie

s

All

inde

pend

ent v

aria

bles

for t

his

obse

rvat

ion

are

mea

sure

d at

this

poi

nt o

f tim

e :

- Sc

ope

of o

ppor

tuni

ty-

Prio

r E

xper

ienc

e-

Eff

ects

of

com

petit

ion

Dep

ende

nt v

aria

ble

is th

e lik

elih

ood

ofth

e fi

rm f

iling

a 2

nd p

aten

t giv

en th

at it

file

d th

e 1st

one

on

Dec

. 12,

198

0

Obs

erva

tion

Figu

re1.

Illu

stra

tive

data

poin

t



its scope, the more likely that it has applications inmultiple technological areas. We use two measuresof scope: claims made in the patent and overallpotential.

Number of claims. While the number of classesindicates the categorization of knowledge withinthe patent into different areas, the entire patent isconstructed around the claims it makes (Tong andFrame, 1994). A claim is the actual contributionmade by the invention. For instance, a patent filedby a firm with respect to a cure for arthritis mighthave two claims. The first one deals with thecompound, while the second describes the mannerin which it is administered. Firms are only allowedto file claims that are considered nontrivial andnonobvious by the patent examiner. In additionto potential within and outside pharmaceuticals,the number of claims made by a patent indicatesthe areas of impact. More claims increase thepotential application areas for a patent (Lanjouwand Schankerman, 1999), increasing variance inperformance outcomes and hence the option valueof pursuing a second patent. To measure breadthof potential impact we measured the number ofclaims granted in the first patent.

Overall potential. Following Lerner (1995), Ros-enkopf and Nerkar (2001), and Shane (2001), whosuggest that patents classified in more technologi-cal areas are broader in scope, we operationalizedthe scope of a patent by measuring the number oftechnological classes into which it is categorized.

Prior experience

Extensiveness of previous commitments was mea-sured by counting the cumulative number of newareas entered by the firm. First we determinedthe total number of all new technical areas (newsubclasses) that the firm has entered in the past,measured at the time that the first patent in thecurrent new technical area was granted. The moreoptions it has taken out in the past, the more thefirm needs to be concerned with expiration of thoseoptions.

Cumulative investment experience: here we usedtwo measures.

Pharmaceutical patents. This is the cumulativenumber of patents the firm has been granted in

the pharmaceutical sector, calculated at the timewhen the first patent in a new technical area wasgranted. This is a measure of the degree to whichthe firm may be concerned with exploitation ofexisting options.

Non-pharmaceutical patents. This is measured asthe cumulative number of patents that the firm hasbeen granted in other sectors, calculated at the timewhen the first patent in a new technical area wasgranted. This is another, even broader measure ofthe degree to which the firm may be concernedwith the exploitation of existing options.

Effects of competition

We used two measures for effect of competition.

Competitor commitment. This is measured as theaverage number of patents granted per competitorin the new technical area for the firm, measured atthe time the first patent was granted to the firm inthat new technical area.2 Very low commitment bycompetitors indicates limited opportunity for firmsseeking patents, signaling a sparse environmentunattractive to potential new entrants. A largenumber of patents per competitor signals a hotlycontested environment, also unattractive to newentrants.

Number of competitors. This is measured as thenumber of competitors in the technical area at thetime the first patent in the new technical area wasgranted to the firm. Competitors were defined asany other organization with patents in the technicalarea of concern. This might include not-for-profitorganizations (such as universities).

Figure 2 depicts the operationalization of ourtheoretical model developed in Hypotheses 1, 2a,2b, and 3.

Analytical techniques

Following Podolny and Stuart (1995) and Podolny,Stuart, and Hannan (1996) we use event history

2 We also ran regression models with number of patents insteadof patents per competitor and the results were similar. However,we prefer using patents per competitor as the number of patentsis highly correlated with number of competitors.



Scop

e of

Opp

ortu

nity

(H1)

1. N

umbe

r of

cla

ims

with

in a

pat

ent

2. O

vera

ll po

tent

ial

Ext

ensi

vene

ss o

f Pri

or E

xper

ienc

e (H

2a)

3. C

umul

ativ

e nu

mbe

r of

new

are

as e

nter

ed

Cum

ulat

ive

Inve

stm

ent

Exp

erie

nce

(H2b

)

4. P

aten

ts in

the

phar

mac

eutic

al a

rea

5. P

aten

ts in

non

-pha

rmac

eutic

al a

reas

Eff

ects

of C

ompe

titi

on (H

3)

6. C

ompe

titor

com

mitm

ent

7. N

umbe

r of

com

petit

ors

in a

rea

Inve

stm

ent P

rope

nsit

y:L

ikel

ihoo

d of

taki

ng o

utan

opt

ion

(i.e

. sec

ond

pate

nt in

a n

ew te

chni

cal

area

)

H1:

Pos

itive

H2a

and

H2b

:N

egat

ive

H3:

Pos

itive

Figu

re2.

Con

stru

cts

and

mea

sure

s



analysis to model investment behavior. Hazard ratemodels are used as they incorporate information onboth censored and uncensored cases, i.e., whethera firm files a second patent in the area. If T is theduration since the first patent was granted, thenthe instantaneous (hazard) rate of a second patentbeing filed at time t is defined as

r(t) = lim�t → 0

Pr(t ≤ T < t + �t)

�t

We modeled the hazard rate using semiparametricCox models (Cox, 1972; Kalbfleisch and Prentice,1980; Allison, 1995). The equation that we esti-mate takes the following specification:

r(t) = h(t) exp{Xβ}

where r(t) is the transition rate or hazard rate of asecond patent being filed after the grant of the firstpatent, h(t) is an unspecified baseline rate for thetransition, X is a matrix of time-constant covariatesand β is a vector of unknown regression parameter.Because h(t) is an unspecified step function, theCox model offers an extremely flexible means formodeling time dependence (Sørensen and Stuart,2000). In this case X consists of the set of controlsand the independent variables.

The models also include dummy variables tocontrol for fixed firm and year effects that accountfor factors inside the firm and time-changing fac-tors that may have affected the pharmaceuticalindustry. These dummy variables control for omit-ted variables that may have constant effects onthe firms in the sample but vary over time. Ouranalysis also includes the number of options takenout in the past (second patents in new areas) asan explanatory variable. By including the num-ber of times the dependent variable has previouslyoccurred for each firm we control for unobservedheterogeneity (Heckman and Borjas, 1980). AsStuart (2000) indicates, by including such vari-ables we can control for the time constant effects ofunobserved factors that produce variance in firms’abilities or dispositions to patent in new areas.

RESULTS

Table 1 provides information on the patentingactivity of the 31 focal firms included in thesample. Table 2 presents simple statistics and a

Table 1. Statistics for firms in sample: 1979–95

Firm name Censoreda Entries Total

Abbott Laboratories 121 86 207American Cyanamid 155 138 293Allergan 106 77 183American Home Products 115 66 181Alza 52 16 68Bristol Myers Squibb 129 199 328Boehringer Mannheim 108 153 261Ciba Geigy 134 169 303Roussel 130 193 323Sanofi 122 98 220Fujisawa 127 135 262Glaxo 111 136 247Hoechst 131 125 256Janssen 54 79 133Eli Lilly 140 154 294Merck 132 217 349Monsanto 82 30 112Pfizer 102 173 275Pharmacia 68 23 91Upjohn 114 52 166Roche 150 197 347Rhone Poulenc Rorer 135 195 330SmithKline Beecham 108 131 239Sandoz 135 109 244Searle 127 178 305Schering Plough 130 238 368Syntex 90 115 205Wellcome Laboratories 135 88 223Warner Lambert 176 239 415Yamanouchi 89 55 144Zeneca 133 149 282

Total 3641 4013 7654

a Censored entries are first patents that did not lead to secondpatents by the focal firm during the observation period 1979–95.

partial correlation matrix for the 31 organizationsincluded in the sample.

Table 3 presents the results of the proportionalhazards Cox regression of the likelihood of tak-ing out a successful second patent in a new area.The first column reports the baseline log likeli-hood of such an event. The second column reportsthe log likelihood with firm and year fixed effectsincluded. These effects are significant in all themodels that we analyzed. To test for the indepen-dent effects of each of the constructs, we includethe variables measuring each construct separatelyto the baseline fixed effects specification in Models1 through 4 in Table 3.

Model 1 includes the variables measuring theattractiveness of the opportunity, i.e., overall



Tabl

e2.

Sim

ple

stat

istic

san

dpa

rtia

lco

rrel

atio

nm

atri

x

Var

iabl

eN

Mea

nS.

D.

Min

Max

(1)

(2)

(3)

(4)

(5)

(6)

(1)

Num

ber

ofC

laim

s76

5415

.647

14.1

811

237

1.00

0(2

)O

vera

llPo

tent

ial

7654

1.68

31.

440

09

0.16

9∗∗∗

1.00

0(3

)N

umbe

rof

Are

asE

nter

ed76

5416

6.88

695

.442

142

5.5

−0.0

19∗

−0.0

181.

000

(4)

Phar

mac

eutic

alPa

tent

s76

5439

5.82

530

7.46

16

2256

−0.0

40∗∗

∗−0

.019

∗0.

524∗∗

∗1.

000

(5)

Non

Phar

mac

eutic

alPa

tent

s76

5495

4.72

110

9228

7192

0.03

5∗∗∗

0.02

1∗∗0.

096∗∗

∗0.

387∗∗

∗1.

000

(6)

Com

petit

orC

omm

itmen

t76

541.

861

0.71

71

8.55

60.

024∗∗

0.03

4∗∗∗

−0.0

48∗∗

∗−0

.043

∗∗∗

−0.0

54∗∗

∗1.

000

(7)

Num

ber

ofC

ompe

titor

s76

5436

.671

30.2

101

284

−0.0

020.

030∗∗

∗−0

.063

∗∗∗

−0.0

67∗∗

∗−0

.048

∗∗∗

0.54

4∗∗∗

∗∗∗p

<0.

01;

∗∗p

<0.

05;

∗p

<0.

1


Real Options and Pharmaceutical R&D 13Ta

ble

3.Pr

opor

tiona

lha

zard

sC

oxre

gres

sion

mod

els

oflik

elih

ood

ofop

enin

gan

optio

n(2

ndpa

tent

inne

war

ea)

Var

iabl

ede

scri

ptio

nB

asel

ine

Fixe

def

fect

sM

odel

1M

odel

2M

odel

3M

odel

4M

odel

5

Scop

eof

oppo

rtun

ity

Num

ber

ofC

laim

s0.

0040

∗∗∗

0.00

37∗∗

∗

(0.0

010)

(0.0

010)

Ove

rall

Pote

ntia

l0.

0386

∗∗∗

0.03

39∗∗

∗

(0.0

109)

(0.0

110)

Ext

ensi

vene

ssof

prio

rex

peri

ence

Num

ber

ofA

reas

Ent

ered

−0.0

035∗∗

∗−0

.002

6∗∗∗

(0.0

007)

(0.0

008)

Cum

ulat

ive

prio

rin

vest

men

tPh

arm

aceu

tical

Pate

nts

−0.0

007∗∗

∗0.

0000

(0.0

002)

(0.0

002)

Non

-Pha

rmac

eutic

alPa

tent

s−0

.000

1∗−0

.000

1∗

(0.0

001)

(0.0

001)

Eff

ects

ofco

mpe

titi

onC

ompe

titor

com

mitm

ent

0.21

41∗∗

∗0.

2033

∗∗∗

(0.0

232)

(0.0

234)

Num

ber

ofC

ompe

titor

s0.

0063

∗∗∗

0.00

62∗∗

∗

(0.0

007)

(0.0

007)

Firm

effe

cts

(om

itted

firm

—Z

enec

a)Si

gnifi

cant

Sign

ifica

ntSi

gnifi

cant

Sign

ifica

ntSi

gnifi

cant

Sign

ifica

ntY

ear

effe

cts

(om

itted

year

—19

95)

Sign

ifica

ntSi

gnifi

cant

Sign

ifica

ntSi

gnifi

cant

Sign

ifica

ntSi

gnifi

cant

N76

5476

5476

5476

5476

5476

54C

enso

red

3641

3641

3641

3641

3641

3641

Eve

nts

4013

4013

4013

4013

4013

4013

Deg

rees

offr

eedo

m46

4847

4848

53−2

log

likel

ihoo

d67

624.

5267

221.

5467

190.

4067

192.

6767

201.

5966

938.

4666

890.

88Im

prov

emen

tIn

log

likel

ihoo

d40

3.01

∗∗∗

31.1

4∗∗∗

29.1

4∗∗∗

19.9

5∗∗∗

283.

08∗∗

∗33

0.66

∗∗∗

Com

pari

son

Bas

elin

eFi

xed

eff.

Fixe

def

f.Fi

xed

eff.

Fixe

def

f.Fi

xed

eff.

∗∗∗p

<0.

01;

∗∗p

<0.

05;

∗p

<0.

1.V

alue

sin

pare

nthe

ses

are

stan

dard

erro

rs.



potential and number of claims. The increase inlog likelihood is statistically significant as com-pared to the model that included only the firm andyear fixed effects. Also, the parameter estimates ofboth these variables are positive and statisticallysignificant, offering strong support for Hypothesis1. These results are consistent with ROR as thereis an increasing propensity to take out a secondpatent, or real option, with corresponding increasesin potential and claims.

Model 2 includes only the variable for cumula-tive number of areas as a measure for the exten-siveness of prior experience. This variable is neg-ative, statistically significant, and in the hypothe-sized direction, offering support for Hypothesis 2a.The improvement in log likelihood as comparedto the baseline model with fixed effects is alsosignificant. A similar result supporting Hypoth-esis 2b is seen in Model 3, where only vari-ables measuring cumulative prior investments areincluded. These results generally support the argu-ment that the R&D investment decisions of thefirm will be influenced by past option investmentdecisions, rather than showing no impact from pre-vious investments, as the net present value modelmight suppose.

The independent effects of competition are testedfor by including the two variables of competitorcommitment and number of competitors in Model4. Both these variables are positive, statisticallysignificant, and in the hypothesized direction. Theimprovement in log likelihood over the baselinefixed effects model is statistically significant. Theresults of Model 4 generally support the ROR argu-ment that the R&D option decisions of the firmwill be influenced by the signals of option poten-tial emanating from the activities of competition.The pattern of our results suggests that both thenumber of competitors and their patenting recordwill have a positive effect on a firm’s propensityto take out an option.

Model 5 is the full model and a completespecification that includes all variables measuringthe different constructs described in the differ-ent hypotheses. All the independent effects exceptthose represented by the variable measuring cumu-lative prior investments continue to be signifi-cant in this model, offering strong support forHypotheses 1, 2a, and 3, and partial support forHypothesis 2b.

Comparing the results across the four sets ofconstructs from Model 5 offers some insight into

their relative strengths in terms of their impact onlikelihood of taking out an option. An increaseof 10 percent in the overall potential measurefrom its mean level leads to a 6.48 percentincrease in the hazard of taking out an option3

(exp[0.034 × 1.8515] = 1.0648). In contrast, anincrease of 10 percent in the mean commitmentexhibited by competitors leads to an increase of51.6 percent in the hazard of taking out an option(exp[0.2033 × 2.047] = 0.1516). The correspond-ing percentage increase or decrease in the hazardof taking out a second patent with 10 percentincrease in mean value of the various constructs isas follows: claims (6.56), cumulative areas entered(37.9), pharmaceutical patents is (−1.72), non-pharmaceutical patents is (−9.48) and number ofcompetitors (28.31). In other words, the effect ofthe average commitment exhibited by competitorsis nearly eight times the magnitude of the effectof the average potential exhibited by the oppor-tunity area. The above comparison also suggeststhat competitive effects have the greatest influ-ence on the likelihood of taking out an option,followed by prior options taken out and the scopeof the opportunity. These differences in magnitudeare consistent with research that technical attrac-tiveness is not the only variable that drives entrydecisions (Podolny and Stuart, 1995).

Effects of competition: Testing for curvilineareffects

As uncertainty is reduced, however, the valueof options decreases, and the potential value ofthe underlying knowledge asset becomes easier toassess. The implication is that, in the early stageof a new technological arena, the more competitorstake out options, the greater will be the incentivefor additional firms to take out options. As the areamatures, however, continued investment will onlybe attractive for those firms who perceive that theywill be in a good position to exercise their options.For the rest, allowing their options to expire (or

3 To examine the multiplier effect of each variable on the hazardrate, keeping all other variables at their mean, one needs tocompute the exponential function of the product of the new valueof the focal variable and its parameter estimate. For instance,in our case the 10 percent increase in mean value of overallpotential leads to a value of 1.8515. On taking the exponentialfunction of this value multiplied by the parameter estimate (0.34)we get the increase in the hazard rate, i.e., 1.0648. For a moredetailed explanation of how multiplier rates are computed seeappendix B of Haveman and Cohen (1994).



Table 4. Proportional hazard Cox regression models of likelihood of opening an option: testingfor curvilinear competition effects

Variable description Model 6 Model 7 Model 8

Scope of opportunityNumber of Claims 0.0036∗∗∗ 0.0038∗∗∗ 0.0037∗∗∗

(0.0010) (0.0010) (0.0010)Overall Potential 0.0333∗∗∗ 0.0332∗∗∗ 0.0329∗∗∗

(0.0110) (0.0110) (0.0110)Extensiveness of prior experienceNumber of Areas Entered −0.0026∗∗∗ −0.0025∗∗∗ −0.0025∗∗∗

(0.0008) (0.0008) (0.0008)Cumulative prior investmentPharmaceutical Patents −0.0001 −0.0001 −0.0001

(0.0002) (0.0002) (0.0002)Non-Pharmaceutical Patents −0.0001∗∗ −0.0001∗∗ −0.0001∗∗

(0.0001) (0.0001) (0.0001)Effects of competitionCompetitor commitment 0.5397∗∗∗ 0.1836∗∗∗ 0.4455∗∗∗

(0.0839) (0.0242) (0.0859)Competitor Commitment2 −0.0606∗∗∗ −0.0461∗∗∗

(0.0152) (0.0151)Number of Competitors 0.0052∗∗∗ 0.0116∗∗∗ 0.0096∗∗∗

(0.0007) (0.0015) (0.0016)Number of Competitors2/1000 −0.0408∗∗∗ −0.0315∗∗∗

(0.0102) (0.0102)Firm effects (omitted firm—Zeneca) Significant Significant SignificantYear effects (omitted year—1995) Significant Significant SignificantN 7654 7654 7654Censored 3641 3641 3641Events 4013 4013 4013Degrees of freedom 54 54 55−2 log likelihood 66870.67 66870.87 66859.56Improvement over full Model 6 20.20∗∗∗ 19.99∗∗∗ 31.31∗∗∗

Chi square 753.86∗∗∗ 753.66∗∗∗ 764.97∗∗∗

∗∗∗ p < 0.01; ∗∗ p < 0.05; ∗ p < 0.1

trading them with one of the more advantagedfirms) makes more sense. The options story is thusnot so much about competition creating resourcepressure upon rivals (as in the ecological litera-ture) but rather about strategic redirection of effort,based on the joint effects of uncertainty reduc-tion and the potential of heterogeneous resourcecombinations. Both the positive effects of collab-oration and positive spillovers from competitorsdoing R&D and the negative effects of rivalryamongst competition are possible on likelihood ofentry. Thus the curvilinear results for competitionare worth further exploration. To do this we ana-lyzed three models, shown in Table 4, where weinclude the squared terms of the variables measur-ing effects of competition.

Models 6 and 7 show support for the indepen-dent curvilinear effects of competitor commitment

and number of competitors respectively. Model 8includes both curvilinear effects and the resultscontinue to support a curvilinear relationship bet-ween the likelihood of taking out an option andcompetition. To find the point at which the impactof number of competitors on the hazard of takingan option is at a maximum, we differentiate theresults for Model 8 with respect to number ofcompetitors, and set the result equal to zero:4

This gives: 0.0096 − 2 × (0.0315)

× number of competitors/1000 = 0

Solving gives the number of competitors as 152.6.This suggests that the deterrent effects of excessive

4 In the absence of curvilinear effect of competitor commitmentthe number of competitors at which the hazard is at a maximumis 142.15.



competition on a firm’s propensity to invest are attheir maximum when 152 competitors have enteredthe area. Only after as many as 152 competitors arein the market do firms actually find the competi-tiveness discouraging, but the evidence of curvi-linearity suggests a negative marginal effect. Thepositive effect of increase in competition beginsto tail off as the number of competitors increases.Examination of the basic statistics in Table 1 sug-gests that very few of the data points in our samplehave a number of competitors exceeding 152 (themean number of competitors is 36.8 while the stan-dard deviation is 30.28).5 This result needs to beinterpreted with care, as these 152 competitors arenot competitors in the product market context butin the resource or R&D market. A brief exami-nation of a few entries in our data set suggeststhat early entrants in new areas are universities,research organizations, and independent inventorswho are less likely to be interested in the commer-cial aspects of the knowledge they generate. Futureresearch could examine the nature of competitionin R&D markets to explore whether signals fromentry of different competitors are equal.

To find the point at which the impact of numberof patents per competitor on the hazard of takingan option is at a maximum we differentiate theresults for Model 8 with respect to competitorcommitment, and set the result equal to zero:6

This gives: 0.4455 − 2 × 0.0461

× number of patents = 0

Solving gives the number of patents per com-petitor as 4.83. This further supports the RORargument—only after as many as 4.83 patents percompetitor do firms reduce their patenting. Againstatistics from Table 1 suggest very few cases inour sample have competitor commitment beyond4.83 (mean is 1.86 and standard deviation is 0.72),but again the evidence of curvilinearity suggests adecreasing positive effect as the number of patentsper competitor increases. In both cases the concavefunction obtained strongly suggests that the rate ofincrease in the log likelihood of a firm taking outan option goes down with increases in the number

5 We are thankful to an anonymous reviewer for identifying thispattern in our data.6 In the absence of curvilinear effects of number of competitorsthe competitor commitment at which the hazard rate is maximumis 4.45 patents per competitor.

of competitors and competitor commitment. Thusfirms in our sample are initially increasingly, thendecreasingly inclined to take out options in newtechnical areas as the opportunity space becomesmore competitive.

We see that the effect of patent/competitor ismuch more than that of number of competitors.This is consistent with the argument that com-petitive entry has an important signaling effecton the propensity to enter. Signs that competitorsare making significant headway in the new area,as evidenced by their success at patenting there,however, can be expected to act as a deterrent,consistent with our results.

DISCUSSION AND IMPLICATIONS

Our study of R&D investments in the pharmaceuti-cal industry suggests that strategic decision-makersdo either intuitively or explicitly use ROR whenmaking investment choices under uncertainty. Weidentified three constructs that are derived fromROR but that at present have not been incorporatedinto a theory of investment for the field of strat-egy, namely the potential for increased variance,effects of previous portfolio investments, and thetension created by competitive entry.

Rather than being deterred in making investmentby their potential variance enhancement, we foundthat scope (potential of the previous patent in a newarea) had a positive effect on firms’ propensity toinvest in R&D. One implication is that variance,measured as the scope of the opportunity area,rather than being a negative that would requiregreater risk premiums from investors, is positivewhen the investment in question is an option. Animplication for the strategic theory of investmentis that different standards should apply to evalu-ation for scope-increasing options as opposed tocommitments. It may make sense to allocate dif-ferent resource pools for the different kinds ofinvestments, because different assessment criteriaare appropriate and because different discount ratesare also appropriate. Such a distinction, based onoptions logic, was urged by McGrath and MacMil-lan (2000: 163–196).

A criticism of net present value-based invest-ment models is that they tend to underestimatefuture cash flows that are generated from oppor-tunities discovered at a later point in time thanwere obvious at the point of initial investment.



Scholars have argued that ROR, because it allowsfor consideration of future opportunities, miti-gates undervaluing investments with latent poten-tial. Our findings suggest that decision-makersdo appear in practice to take future opportuni-ties into account, at least when they invest inoptions. Firms were more likely to invest in newoptions in areas revealed to be radically differ-ent in scope. Those options with greater scopeare also those which are more likely to lead topotentially valuable future growth opportunities.By implication, the taken-for-granted assumptionthat variance-enhancing potentially radical oppor-tunities are likely to be unattractive to establishedfirms should be questioned. Established firms areoften characterized as risk-averse, rigid and resis-tant to change. Our study suggests that investingin change through an options lens may offer apath to greater innovation, mitigating organiza-tional rigidity.

We also found evidence that decisions to investin new options were influenced by the state of theexisting portfolio of options in a firm. The pri-mary implication of this finding is that assessment(and logically, quantitative evaluation) of any sin-gle option is interdependent with the rest of thefirm’s portfolio. Failing to invest in more optionsin an area diminishes the worth of all previousoptions in that area because new discoveries withgrowth potential are less likely.7 Attractiveness ofan established area diminishes a firm’s propen-sity to invest in additional new areas because ofdiminishing returns to postponement of exercisein the established area. Because of these effects,investing in new options is not necessarily addi-tive. It is conceivable that adding more optionsto an existing portfolio can actually decrease itsvalue, not increase it as a simple additive modelwould assume. Thus, in our thinking and modelingwe need to go from assumptions of independenceand additivity of new investments to assumptionsof interdependence and nonlinear effects.

Portfolio effects offer some interesting points ofintersection between ROR and theories of orga-nizational learning in strategy. In a highly influ-ential article, March (1991) argued that ‘balance’between exploration and exploitation was essen-tial to organizational viability. Our study con-tributes to the burgeoning literature sparked by

7 However, real options are not additive in nature and the dropdue to investment may not be linear.

March’s paper. We found two different mecha-nisms that appear to influence firm investments innew areas. The scope of a new area triggered anexploratory investment response. In contrast, thesuccessful results of past exploration influencedfirms to respond with less exploratory investing,presumably to focus on the learning they had donethat was most vulnerable to diminishing returnsand eventual erosion. To the extent that previouslyopened options have not been exercised or allowedto expire, a firm is rational to divert resources fromthe exploration of new options to concluding itsinvestigations of old ones. Options reasoning thusoffers insight into the mechanism through whichprescriptions for firms to achieve balance in knowl-edge creation can be translated to actual strategicdecision making.

We also found explanations consistent with RORin the way in which firms responded to com-petitive entry. In the early period of discoveryof a new technological area, competitive entryinduced investment. The observation that competi-tors have found the arena attractive is likely tolead a decision-maker to adjust expectations forthe attractiveness of the area upward as well, mak-ing investments in the area more likely. As playersenter, all of them benefit from the net investmentinto the area and the concomitant reduction in tech-nical uncertainty that this produces. With reductionin uncertainty, however, comes evidence of whichoptions will be ‘in the money’ and which will not.Not all players will be in a position to exercisethose options that they have opened. Firms with-out a potential advantage are then likely to exitas a matter of strategic choice (exercising theiroption to stop or to redirect). Although this patternhas been identified in previous empirical studies,in which firms have entered, faced stiff competi-tion, and exited from a field (for instance, Sahlmanand Stevenson, 1985), this has mostly been exam-ined at the level of a line of business. Researchersfocusing on the business level of analysis havethus developed a different rationale for the pat-tern—namely, entry prompted by high (and oftenunrealistic) expectations of the carrying capacityof an attractive area, with exit prompted by theinability to sustain the business in the face of com-petition.

Some important caveats to our conclusions areother explanations that could exist for our resultsand which should be addressed in future research.For instance, past research has shown that some



technologies are more accessible than others, lead-ing to simultaneous inventions by multiple inven-tors (Merton, 1972). Our data set does not per-mit us to analyze such effects but future research,by gathering detailed information on the natureof each sub-area, could begin to examine theseeffects—specifically, the extent to which knowl-edge in the area is easily articulated, codifiable,and accessible (Winter, 1987) and the extent towhich there are network effects on the likelihoodof innovation (Powell, Koput, and Smith-Doerr,1996). For instance, in the business methods areaor software arena where knowledge is articulated,codified, and accessible, and where the need forcomplementary assets such as R&D facilities isminimal, one would expect the likelihood of entryto be driven by the extent to which firms canquickly appropriate rents from such inventions(Bagby, 2000).

Another avenue for research is to develop bet-ter measures of value and variability of value fromentering a new area. Our measure of the scope ofthe opportunity, while a reasonable proxy for theunderlying variance, is still crude. Future researchcould develop alternate measures based on licens-ing or royalty agreements. Finally, we motivate ourstudy by offering a complementary approach tonormative investment models borrowed from thefinance field. Our intention was to provide evi-dence that investment strategies of pharmaceuticalfirms are consistent with ROR. An important topicfor future research is to compare the efficacy ofthese two approaches. Qualitative evidence gath-ered from our field studies suggests that the realoptions approach resonates with the thinking of atleast a few R&D managers (Nichols, 1994b). How-ever, more rigorous empirical testing of these twoapproaches in the R&D context is required beforefurther comparisons are made.

ROR, and our study, suggests a different logicfor investment within the firm than that reflect-ing conventional assumptions. Based on this logic,we see entry prompted indeed by expectations forfuture success. Significant new investment, how-ever, actually does make the area more valuableby leading multiple firms to reduce uncertaintyand establish a base of knowledge. Entry eventu-ally slows and ceases, not necessarily as a conse-quence of resource pressures, but because the areais now revealed to be differentially attractive todifferent players. Those for whom the upside is nolonger substantial stop making further investments

in options in it. An implication for scholars is thatthe concave function that underlies the relationshipbetween competitive entry and propensity to investhas a different causal explanation for options thanfor businesses (which are by definition more likelyto represent commitments, not options). The chal-lenge for researchers is not to stop at identifyingthe pattern and attributing causality (as is all toooften done) but at identifying the logic underlyingthe pattern.

CONCLUSION

In this research, we explored whether actual firminvestment behavior appears consistent with ROR.We thus examined the outcome of firm-levelinvestment decisions in the pharmaceutical sectorover a long period of time (17 years). Our findingssupport the premise that decision-makers eitherimplicitly or explicitly utilize ROR. Our resultssuggest that the options concept could prove to bea fruitful approach to developing theories of invest-ment in the strategy field that are more consistentwith observed managerial behavior than those the-ories that rely purely on the assumptions of financeor economics. Specifically, we found that decisionsto pursue an option on a new technological area(measured by successful second patents filed inthose areas) are influenced by the scope of thetechnological opportunity, the competition in thearea, and a firm’s past investment behavior. Toour knowledge, this is one of the first studies toempirically test the options premise so extensively.

The options approach has implications for im-proving theories in strategy of how managers mat-ter to the creation of competitive advantage. A coreidea in ROR is that of making limited-downside,relatively small investments until sufficient uncer-tainty has been reduced to make commitmentswith greater confidence. One effect is to increasethe number of areas that can be explored whiledecreasing the cost of each exploratory foray.Other things being equal, one might thereforeexpect that under high levels of uncertainty firmsmanaged with an options sensibility have greaterinternal variety than those that do not, improv-ing their chances for both survival and advan-tage under rapidly changing external circumstances(Ashby, 1956).

Options reasoning provides an economic ration-ale for the emergence of performance heterogeneity



subsequent to the investment in options. By observ-ing how firms take out options, thus beginningthe developmental path for what may turn outto become valuable capabilities, we have speci-fied and measured a point at which managementdecisions directly influence capability formation.Moreover, options theory offers an economic ratio-nale for this behavior, tying together powerful con-cepts of future value developed in finance and theobserved incremental-investment behavior of firmsoperating under uncertainty. As organizations seekto cope with ever more rapidly moving compet-itive environments, creating new capabilities anddoing so in a cost-effective way will become acompetitive necessity. By examining the process ofinvesting in real options, such as patents in a newpharmaceutical subclass, we provide a point ofdeparture for better understanding resource-basedcompetitive advantages, because at the time ofinvestment the future benefit is not yet known.Thus, the research presented here offers an exam-ple of an empirical approach strategy researchersmight use to respond to the criticism that the fieldrelies on post hoc rationalization (see Cockburn,Henderson, and Stern, 2000).

A further theme that is central to ROR butnot yet to our theories of strategic investment isthe idea that investments in uncertainty reductionerode in value over time. To use the language ofoptions, they expire. Future research on the erosionof option value and the expiration of real optionsshould test whether firms adopting ROR for invest-ments do let options expire as uncertainty associ-ated with the underlying option is reduced. Fur-ther, if existing resource combinations that under-lie technology investments become commoditized,and options expire at a rapid rate, firms must beable to rapidly go from exploration of new pos-sibilities to exploiting the commercial potential ofthose opportunities. In established organizations,such rapid learning and responsiveness is difficultto create as established systems are constructed toyield reliable performance rather than engage ininnovative behavior. Our finding of evidence sup-portive of option expiration, when coupled with thetendency for the products of existing competencesand routines to become commoditized, suggestswhy organizational learning and knowledge man-agement have become a central focus for scholarlyattention.

ROR offers a perspective from which to developideas that are relevant to the problems facing

decision-makers in established firms. We demon-strated that it is possible to derive testable hypothe-ses grounded in ROR and to collect data that canbe used to test the hypotheses. The support for ourhypotheses suggests that decision-makers in prac-tice (at least in the U.S. pharmaceutical industry)deploy a pattern of investment choices that is con-sistent with ROR. As we conclude, it is our hopethat real options concepts might offer a useful per-spective for those seeking to develop a theory ofinvestment for the field of strategic management.

ACKNOWLEDGEMENTS

The data for this paper form part of a largerdata set collected by the second author for hisdissertation. Both authors acknowledge researchsupport provided by the Wharton EntrepreneurialCenter.

REFERENCES

Ahuja G. 2000. The duality of collaboration: inducementsand opportunities in the formation of interfirm linkage.Strategic Management Journal 21(3): 317–343.

Alchian AA. 1950. Uncertainty, evolution and economictheory. Journal of Political Economy 58: 211–222.

Allison PD. 1995. Survival Analysis Using the SASSystem . SAS Institute: Cary, NC.

Arora A, Fosfuri A, Gambardella A. 2001. Markets forTechnology . MIT Press: Cambridge, MA.

Ashby WR. 1956. An Introduction to Cybernetics .Chapman & Hall: London.

Bagby JW. 2000. Business method patent proliferation:convergence of transactional analytics and technicalscientifics. Business Lawyer 56(1): 423–458.

Barney JB. 1986. Strategic factor markets: expectations,luck and business strategy. Management Science 32:1231–1241.

Bettis RA, Hitt MA. 1995. The new competitivelandscape. Strategic Management Journal , SummerSpecial Issue 16: 7–19.

Black F, Scholes M. 1973. The pricing of options andcorporate liabilities. Journal of Political Economy 81:637–654.

Bowman EH, Hurry D. 1993. Strategy through the optionlens: an integrated view of resource investmentsand the incremental-choice process. Academy ofManagement Review 18: 760–782.

Brennan MJ. 1995. Corporate finance over the past25 years. Financial Management 24: 9–22.

Chatterjee S, Lubatkin MH, Schulze WS. 1999. Towarda strategic theory of risk premium: moving beyondCAPM. Academy of Management Review 24(3):556–567.



Chi T, McGuire DJ. 1996. Collaborative ventures andvalue of learning: Integrating the transaction cost andstrategic option perspectives on the choice of marketentry modes. Journal of International Business Studies27: 285–307.

Christensen CM. 1997. The Innovator’s Dilemma: WhenNew Technologies Cause Great Firms to Fail . HarvardBusiness School Press: Boston, MA.

Cockburn I, Henderson R, Stern S. 2000. Untanglingthe origins of competitive advantage. StrategicManagement Journal , Special Issue 21(10–11):1123–1145.

Cohen WM, Levinthal DA. 1990. Absorptive capacity:a new perspective on learning and innovation.Administrative Science Quarterly 35: 128–152.

Cox DR. 1972. Regression models and life tables.Journal of the Royal Statistical Society B34: 187–220.

Dierickx I, Cool K. 1989. Asset stock accumulation andsustainability of competitive advantage. ManagementScience 35: 1504–1511.

Dixit A. 1992. Investment and hysteresis. Journal ofEconomic Perspectives 6: 107–132.

Dixit AK, Pindyck RS. 1994. Investment Under Uncer-tainty . Princeton University Press: Princeton, NJ.

Dougherty D. 1992. Interpretive barriers to successfulinnovation in large firms. Organization Science 3(2):179–202.

Dutta S, Weiss AM. 1997. The relationship betweena firm’s level of technological innovativeness andits pattern of partnership agreements. ManagementScience 43: 343–356.

Engelsman EC, van Raan AFJ. 1994. A patent basedcartography of technology. Research Policy 23: 1–26.

Fama EF, French KR. 1992. The cross-section ofexpected stock returns. Journal of Finance 47:427–465.

Fama EF, French KR. 1993. Common risk factors in thereturns on bonds and stocks. Journal of FinancialEconomics 33: 33–56.

Fama EF, French KR. 1995. Size and book to marketfactors in earnings and returns. Journal of Finance 50:131–155.

Fama EF, French KR. 1996. Multifactor explanationsof asset pricing anomalies. Journal of Finance 51:55–84.

Folta TB. 1998. Governance and uncertainty: the trade-off between administrative control and commitment.Strategic Management Journal 19(11): 1007–1028.

Gavetti G, Levinthal DA. 2000. Looking forward andlooking backward: cognitive and experiential search.Administrative Science Quarterly 45: 113–137.

Hambrick DC, MacMillan IC, Barbosa RR. 1983. Busi-ness unit strategy and changes in the product R&Dbudget. Management Science 29: 757–769.

Haveman HA, Cohen LE. 1994. The ecological dynamicsof careers: the impact of organizational founding,dissolution and merger on job mobility. AmericanJournal of Sociology 100(1): 104–152.

Heckman JJ, Borjas GJ. 1980. Does unemploymentcause future unemployment: definitions, questions,and answers from a continuous time model of

heterogeneity and state dependence. Econometrica 52:909–938.

Helfat CE. 2000. Guest Editors’ introduction to theSpecial Issue: Evolution of firm capabilities. Strate-gic Management Journal , Summer Special Issue21(10–11): 955–959.

Henderson RM, Clark KB. 1990. Architectural innova-tion: the reconfiguration of existing product technolo-gies and the failure of established firms. AdministrativeScience Quarterly 35: 9–30.

Henderson RM, Cockburn I. 1994. Measuring com-petence? Exploring firm effects in pharmaceuticalresearch. Strategic Management Journal , Winter Spe-cial Issue 15: 63–84.

Hurry D, Miller AT, Bowman EH. 1992. Calls onhigh-technology: Japanese exploration of venturecapital investments in the United States. StrategicManagement Journal 13(2): 85–101.

Jaffe AB, Trajtenberg M, Henderson R. 1993. Geo-graphic localization of knowledge spillovers as evi-denced by patent citations. Quarterly Journal of Eco-nomics 108: 577–675.

Kalbfleisch JD, Prentice RL. 1980. The StatisticalAnalysis of Failure Time Data . Wiley: New York.

Kim D, Kogut B. 1996. Technological platforms anddiversification. Organization Science 7(3): 283–301.

Knight F. 1921. Risk, Uncertainty and Profit . Universityof Chicago Press: Chicago, IL.

Kogut B. 1983. Foreign direct investment as a sequentialprocess. In The Multinational Corporation in the1980s , Kindleberger CP, Audretsch DB (eds). MITPress: Cambridge, MA; 38–56.

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

Kogut B, Kulatilaka N. 1994. Operating flexibility, globalmanufacturing, and the option value of a multinationalnetwork. Management Science 40: 123–139.

Kumaraswamy A. 1996. A real options perspective onfirms R&D investments. Doctoral dissertation, NewYork University.

Lane PJ, Lubatkin M. 1998. Relative absorptive capac-ity and interorganizational learning. Strategic Manage-ment Journal 19(5): 461–477.

Lanjouw JO, Schankerman M. 1999. The quality ofideas: measuring innovation with multiple indicators.NBER working paper no. W7345 (http://www.nber.org/w7345).

Lawless MW, Anderson P. 1996. Generational techno-logical change: effects of innovation and local rivalryon performance. Academy of Management Journal39(5): 1185–1217.

Lerner J. 1995. Patenting in the shadow of competitors.Journal of Law and Economics 38: 463–495.

Levin RC, Klevorick AK, Nelson RR, Winter SG. 1987.Appropriating the returns from industrial research anddevelopment. Brookings Papers on Economic Activity3: 783–831.

Levinthal DA. 1990. Random walks and organizationalmortality. Administrative Science Quarterly 36:397–420.

Levitt B, March JG. 1988. Organizational learning.Annual Review of Sociology 14: 319–340.



March JG. 1991. Exploration and exploitation inorganizational learning. Organization Science 2:71–87.

McGrath B, McGrath RG. 2001. Spillovers and strategy:implications of the network economy. Working paper,Columbia Business School, New York.

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

McGrath RG, MacMillan IC. 2000. The EntrepreneurialMindset: Strategies for Continuously Creating Oppor-tunity in an Age of Uncertainty . Harvard BusinessSchool Press: Boston, MA.

Merton RK. 1972. The Sociology of Science. Universityof Chicago Press: Chicago, IL.

Miller K. 1998. Economic exposure and integrated riskmanagement. Strategic Management Journal 19(5):497–514.

Miller KD, Folta TB. 2002. Option value and entry tim-ing. Strategic Management Journal 23(7): 655–665.

Mitchell GR, Hamilton WF. 1988. Managing R&D asa strategic option. Research-Technology Management27: 15–22.

Morris PA, Teisberg EO, Kolbe AL. 1991. When choos-ing R&D projects, go with long shots. Research-Technology Management January–February: 35–40.

Nelson RR, Winter SG. 1982. An Evolutionary Theoryof Economic Change. Belknap Harvard: Cambridge,MA.

Nichols NA. 1994a. Medicine, management, and merg-ers: an interview with Merck’s P. Roy Vagelos. Har-vard Business Review 72(6): 105–114.

Nichols NA. 1994b. Scientific management at Merck: aninterview with CFO Judy Lewent. Harvard BusinessReview 72(1): 89–99.

Pakes A. 1986. Patents as options: some estimatesof the value of holding European patent stocks.Econometrica 54: 755–784.

Penner-Hahn JD. 1998. Firm and environmental influ-ences on the mode and sequence of foreign researchand development activities. Strategic ManagementJournal 19(2): 149–168.

Podolny JM, Stuart TE. 1995. A role based ecology oftechnological change. American Journal of Sociology100: 1224–1260.

Podolny JM, Stuart TE, Hannan MT. 1996. Networks,knowledge and niches: competition in the worldwidesemiconductor industry, 1984–1991. American Jour-nal of Sociology 102: 659–689.

Porter ME. 1980. Competitive Strategy: Techniques forAnalyzing Industries and Competitors . Free Press:New York.

Powell WW, Koput KW, Smith-Doerr L. 1996. Interor-ganizational collaboration and the locus of innovation:

networks of learning in biotechnology. AdministrativeScience Quarterly 41: 116–145.

Reuer JJ, Leiblein MJ. 2000. Downside risk implicationsof multi-nationality and international joint ventures.Academy of Management Journal 43: 203–214.

Rosenkopf L, Nerkar A. 2001. Beyond local search:boundary spanning, exploration, and impact in theoptical disc industry. Strategic Management Journal22(4): 287–306.

Rumelt RP. 1987. Theory, strategy and entrepreneurship.In The competitive Challenge: Strategies for IndustrialInnovation and Renewal , Teece DJ (ed). Harper &Row: New York; 137–158.

Sahlman WA, Stevenson H. 1985. Capital market myopia.Journal of Business Venturing 1: 7–30.

Sanchez R. 1993. Strategic flexibility, firm organization,and managerial work in dynamic markets: astrategic options perspective. In Advances in StrategicManagement , Shrivastrava P, Huff A, Dutton J (eds).JAI Press: Greenwich, CT; 251–291.

Schumpeter J. 1934. The Theory of Economic Develop-ment . Harvard University Press: Cambridge, MA.

Shane S. 2001. Technological opportunities and new firmcreation. Management Science 47: 205–220.

Silverman B. 1999. Technological resources and thedirection of corporate diversification: toward anintegration of the resource-based view and transactioncost economics. Management Science 45: 1109–1124.

Sørensen JB, Stuart TE. 2000. Aging, obsolescenceand organizational innovation. Administrative ScienceQuarterly 45: 81–112.

Stuart TE. 2000. Interorganizational alliances and theperformance of firms: a study of growth andinnovation rates in a high-technology industry.Strategic Management Journal 21(8): 791–811.

Tong X, Frame DJ. 1994. Measuring national technolog-ical performance with patent claims data. ResearchPolicy 23: 133–142.

Trigeorgis L. 1991. Anticipated competitive entry andearly preemptive investment in deferrable projects.Journal of Economics and Business 43: 143–156.

Trigeorgis L. 1996. Real Options: Managerial Flexibilityand Strategy in Resource allocation . MIT Press:Cambridge, MA.

Van de Ven AH. 1986. Central problems in themanagement of innovation. Management Science32(5): 590–608.

Winter SG. 1987. Knowledge and competence asstrategic assets. In The Competitive Challenge:Strategies for Innovation and Renewal Teece DJ (ed).Ballinger: Cambridge, MA; 186–219.


real options reasoning and a new look at the r&d investment

Documents