Absorptive Capacity: ANew Perspective onLearning and Innovation
Wesley M. CohenCamegie Mellon UniversityDaniel A. LevintbalUniversity of Pennsylvania
in this paper, we argue that the ability of a firm to recog-nize the value of new, external information, assimilate it,and apply it to commerciai ends is critical to its innovativecapabilities. We label this capability a firm's absorptivecapacity and suggest that it is largely a function of thefirm's level of prior related knowledge. The discussion fo-cuses first on the cognitive basis for an individual's ab-sorptive capacity including, in particular, prior relatedknowledge and diversity of background. We then charac-terize the factors that influence absorptive capacity at theorganizational level, how an organization's absorptive ca-pacity differs from that of its individual members, and therole of diversity of expertise within an organization. Weargue that the development of absorptive capacity, and, inturn, innovative performance are history- or path-depen-dent and argue how lack of investment in an area of ex-pertise early on may foreclose the future development ofa technical capability in that area. We formulate a modelof firm investment in research and development (R&O), inwhich R&D contributes to a firm's absorptive capacity,and test predictions relating a firm's investment in R&D tothe knowledge underlying technical change within an in-dustry. Discussion focuses on the implications of absorp-tive capacity for the analysis of other related innovativeactivities, including basic research, the adoption and dif-fusion of innovations, and decisions to participate in co-operative R&D ventures.*
We appreciate the comments of KathleenCwtey, Robyn Oawes. Marie Fk^man, TomHnhoK, Sara Kiesler, Ridiard Nelson,Unda Pike, and ttwee anonynxms ref-erees. "Hie representations and condu-stons presented herein are those of theauthors. They have not been atlopted inw t n ^ or HI FMrt by tfie Federal TradeCommission, its Bureau of Economics, orany other entity wHhin ttie cx>mn^ssion.The FTCs D^dosure AvoidarKS Officerhas certified that the data ^iduded in tNsp^)er do not k ^ t i f y incfividual a»tipany
INTRODUCTION
Outside sources of knowledge are often critical to the inno-vation process, whatever the organizational ievef at which theinnovating unit is defined. While the example of Japan illus-trates the point saiiently at the national level (e.g., Westneyand Sakakibara, 1986; Mansfield, 1988; Rosenberg andSteinmueller, 1988), tt is also true of entire industries, aspointed out by Brock (1975) in the case of computers and byPeck (1962) in the case of aluminum. At the organizations!level, March and Simon (1958: 188) suggested most innova-tions result from borrowing rather than invention. This obser-vation is supported by extensive research on the sources ofinnovation (e.g., Mueller, 1962; Hamberg, 1963; Myers andMarquis, 1969; Johnston and Gibbons, 1975; von Hippel,1988). Finally, the importance to innovative performance ofinformation originating from other internal units in the firm,outside the formal innovating unit (i.e., the R&D lab), such asmarketing and manufacturing, is welt understood (e.g., Mans-field, 1968).
The ability to exploit external knowledge is thus a criticalcomponent of innovative capabilities. We argue that theability to evaluate and utilize outside knowledge is largely afunction of the level of prior related knowledge. At the mostelemental level, this prior knowledge includes basic skills oreven a shared language but may also include knowledge ofthe most recent scientific or technological developments in agiven field. Thus, prior related knowledge confers an ability torecognize the value of new infomiation. assimilate it, andapply it to commercial ends. These abilities collectively con-stitute what we cail a fimn's "absorptive capacity."
At the level of the firm—the innovating unit that is the focushere—absorptive capacity is generated in a variety of ways.Research shows that firms that conduct their own R&D arebetter able to use extemaHy available infomiation (e.g., Tifton,1971; Allen, 1977; Mowery, 1983). This implies that absorp-tive capacity may be created as a byproduct of a firm's R&Dinvestment. Other work suggests that absorptive capacitymay also be developed as a byproduct of a firm's manufac-turing operations. Abemathy (1978) and Rosenberg (1982)have noted that through direct involvement in manufacturing,a firm is better able to recognize and exploit new infomiationrelevant to a particular product market. Production experienceprovides the firm with the background necessary both to rec-ognize the value of and implement methods to reorganize orautomate particular manufacturing processes. Firms also in-vest in absorptive capacity directly, as when they send per-sonnel for advanced technical training. The concept ofabsorptive capacity can best be developed through an exami-nation of the cognitive structures that underlie leaming.
Cognitive Structures
The premise of the notion of absorptive capacity is that theorganization needs prior related knowledge to assimilate anduse new knowledge. Studies in the area of cognitive and be-havioral sciences at the individual level both justify and enrichthis observation. Research on memory development suggeststhat accumulated prior knowledge increases both the abilityto put new knowledge into memory, what we wouid refer toas the acquisition of knowledge, and the ability to recall anduse it. With respect to the acquisition of knowledge. Bowerand Hiigard {1981: 424) suggested that memory developmentis self-reinforcing in that the more objects, pattems and con-cepts that are stored in memory, the more readily is new in-formation about these constructs acquired and the morefacile is the individual in using them in new settings.
Some psychologists suggest that prior knowledge enhanceslearning because memory—or the storage of knowledge—isdeveloped by associative learning in which events are re-corded into memory by establishing linkages with pre-existingconcepts. Thus, Bower and Hiigard (1981) suggested that thebreadth of categories into which prior knowledge is orga-nized, the differentiation of those categories, and the linkagesacross them permit individuals to make sense of and, in tum,acquire new knowledge. In the context of learning a lan-guage, Lindsay and Norman (1977: 517) suggested theproblem in learning words is not a result of lack of exposureto them but that "to understand complex phrases, muchmore is needed than exposure to the words: a iarge body ofknowledge must first be accumulated. After all, a word issimply a label for a set of structures within the memorysystem, so the structures must exist before the word can beconsidered leamed." Lindsay and Nomian further suggestedthat knowledge may be nominally acquired but not well uti-lized subsequently because the individual did not already pos-sess the appropriate contextual knowledge necessary tomake the new knowledge fully intelligible.
The notion that prior knowledge facilitates the leaming of r>ewrelated knowledge can be extended to include the case rn
March t990
which the knowledge in question may itself be a set oflearning skills. There may be a transfer of teaming skillsacross bodies of knowledge that are organized and expressedin similar ways. As a consequence, experience or perfor-mance on one learning task may influence and improve per-formance on some subsequent teaming task (Ellis, 1965). Thisprogressive improvement in the performance of learningtasks is a form of knowledge transfer that has been referredto as "teaming to learn" (Ellis, 1965; Estes, 1970). Estes(1970; 16), however, suggested that the term "learning tolearn" is a misnomer in that prior experience with a learningtask does not necessarily improve performance because anindividual knows how to team (i.e., form new associations)better, but that an individuat may simply have accumutatedmore prior knowledge so that he or she needs to learn tess toattain a given tevet of performance. Notwithstanding what itis about prior teaming experience that may affect subsequentperformance, both explanations of the relationship betweenearty learning and subsequent performance emphasize theimportance of prior knowledge for teaming.
The effect of prior teaming experience on subsequentlearning tasks can be observed in a variety of tasks. For in-stance, Etiis (1965: 4) suggested that "students who havethoroughty mastered the principles of algebra find it easier tograsp advanced work in matherTiatics such as catculus." Fur-ther ittustration is provided by Anderson, Farrett, and Sauers(1984), who compared students teaming LISP as a first pro-gramming language with students learning LISP after havinglearned Pascat. The Pascal students teamed LISP much moreeffectively, in part because they better appreciated the se-mantics of various programming concepts.
The literature also suggests that probtem-sotving skitis de-velop similarly, tn this case, probtem-sotving methods andheuristics typicatty constitute the prior knowtedge thatpermits individuals to acquire retated probtem-sotving capabit-ities. In their work on the development of computer program-ming skitis, Pirotti and Anderson (1985) found that atmost altstudents devetoped new programs by analogy-to-exampteprograms and that their success was determined by how weltthey understood why these examples worked.
We argue that problem sotving and teaming capabilities are sosimitar that there is tittte reason to differentiate their modes ofdevelopment, although exactly what is leamed may differ:learning capabitities invotve the development of the capacityto assimilate existing knowtedge, white probtem-sotving skillsrepresent a capacity to create new knowledge. Supportingthe point that there is little difference between the two,Bradshaw, Langley, and Simon (1983) and Simon (1985) sug-gested that the sort of necessary preconditions for successfulteaming that we have identified do not differ from the pre-conditions required for problem solving and, in turn, for thecreative process. Moreover, they argued that the processesthemselves do not differ much. The prior possession of rele-vant knowtedge and skill is what gives rise to creativity, per-mitting the sorts of associations and linkages that may havenever been considered before. Likewise, Ellis (1965: 35) sug-gested that Harlow's (1959) findings on the development ofteaming sets provide a possible explanation for the behavioral
130/ASa March 1990
Mnorptive Capi^ty
phenomenon of "insight" that typically refers to the rapid so-lution of a problem. Thus, the psychology literature suggeststhat creative capacity and what we call absorptive capacityare quite simitar.
To develop an effective absorptive capacity, whether it be forgeneral knowledge or problem-solving or learning skills, it isinsufficient merely to expose an individual briefly to the rele-vant prior knowledge. Intensity of effort is critical. With regardto storing knowledge in memory, Lindsay and Norman (1977:355) noted that the more deeply the material is processed—the more effort used, the more processing makes use of as-sociations between the items to be learned and knowledgealready in the memory—the better will be the later retrievalof the item. Similarly, learning-set theory (Hariow, 1949, 1959)implies that important aspects of leaming how to solveproblems are built up over many practice trials on relatedproblems. Indeed, Hariow (1959) suggested that if practicewith a particular type of problem is discontinued before it isreliably learned, then little transfer will occur to the nextseries of problems. Therefore, he concluded that considerabletime and effort should be spent on early problems beforemoving on to more complex problems.
Two related ideas are implicit in the notion that the ability toassimilate information is a function of the richness of the pre-existing knowledge structure: learning is cumulative, andlearning performance is greatest when the object of learningis related to what is already known. As a result, learning ismore difficult in novel domains, and, more generally, an indi-vidual's expertise—what he or she knows well—will changeonly incrementally. The above discussion also suggests thatdiversity of knowledge plays an important role. In a setting inwhich there is uncertainty about the knowledge domainsfrom which potentially useful information may emerge, a di-verse background provides a more robust basis for learningbecause it increases the prospect that incoming informationwill relate to what is already known. In addition to strength-ening assimilative powers, knowledge diversity also facilitatesthe innovative process by enabling the individual to makenovel associations and linkages.
From Individual to Organizational Absorptive Capacity
An organization's absorptive capacity will depend on the ab-sorptive capacities of its individual members. To this extent,the development of an organization's absorptive capacity willbuild on prior investment in the development of its constit-uent, individual absorptive capacities, and, like individuals' ab-sorptive capacities, organizational absorptive capacity willtend to develop cumulatively. A firm's absorptive capacity isnot, however, simply the sum of the absorptive capacities ofits employees, and it is therefore useful to consider whataspects of absorptive capacity are distinctly organizational.Absorptive capacity refers not only to the acquisition or as-similation of information by an organization but also to the or-ganization's ability to exploit it. Therefore, an organization'sabsorptive capacity does not simply depend on the organiza-tion's direct interface with the extemal environment. It atsodepends on transfers of knowledge across and within sub-units that nnay be quite removed from the original point of
131/ASQ, March 1990
entry. Thus, to understand the sources of a firm's absorptivecapacity, we focus on the structure of communication be-tween the external environment and the organization, as wellas among the subunits of the organization, and also on thecharacter and distribution of expertise within the organization.
Communication systems may rely on specialized actors totransfer information from the environment or may involve lessstructured pattems. The problem of designing communicationstructures cannot be disentangled from the distribution of ex-pertise in the organization. The firm's absorptive capacity de-pends on the individuals who stand at the interface of eitherthe firm and the extemal environment or at the interface be-tween subunits within the firm. That interface function maybe diffused across individuals or be quite centralized. Whenthe expertise of most individuals within the organizationdiffers considerably from that of extemal actors who can pro-vide useful information, some members of the group arelikely to assume relatively centralized "gatekeeping" or"boundary-spanning" roles (Allen, 1977; Tushman, 1977). Fortechnical information that is difficult for internal staff to as-similate, a gatekeeper both monitors the environment andtranslates the technical information into a form understand-able to the research group. In contrast, if external informationis closely related to ongoing activity, then external informationis readily assimilated and gatekeepers or boundary-spannersare not so necessary for translating information. Even in thissetting, however, gatekeepers may emerge to the extent thatsuch role specialization relieves others from having to monitorthe environment.
A difficulty may emerge under conditions of rapid and uncer-tain technical change, however, when this interface functionis centralized. When information flows are somewhat randomand it is not clear where in the firm or subunit a piece of out-side knowledge is best applied, a centralized gatekeeper maynot provide an effective link to the environment. Under suchcircumstances, it is best for the organization to expose a fairlybroad range of prospective "receptors" to the environment.Such an organization would exhibit the organic structure ofBurns and Stalker (1961: 6), which is more adaptable "whenproblems and requirements for action arise which cannot bebroken down and distributed among specialist roles within aclearly defined hierarchy."
Even when a gatekeeper is important, his or her individualabsorptive capacity does not constitute the absorptive ca-pacity of his or her unit within the firm. The ease or difficultyof the internal communication process and, in turn, the levelof organizational absorptive capacity are not only a function ofthe gatekeeper's capabilities but also of the expertise ofthose individuals to whom the gatekeeper is transmitting theinformation. Therefore, relying on a small set of technologicalgatekeepers may not be sufficient; the group as a wholemust have some level of relevant background knowledge, andwhen knowledge structures are highly differentiated, the req-uisite ievel of background nnay be rather high.
The background knovvledge required by the group as a wholefor effective communication with the gatekeeper highlightsthe more general point that shared knowledge and expertise
132/ASa March 1990
Absorptive Capacity
is essential for communication. At the most basic level, therelevant knowledge that permits effective communicationboth within and across subunits consists of shared languageand symbols (Dearborn and Simon, 1958; Katz and Kahn,1966; Allen and Cohen, 1969; Tushman, 1978; Zenger andLawrence, 1989). With regard to the absorptive capacity ofthe firm as a whole, there may, however, be a trade-off in theefficiency of internal communication against the ability of thesubunit to assimilate and exploit information originating fromother subunits or the environment. This can be seen as atrade-off between inward-looking versus outward-looking ab-sorptive capacities. While both of these components are nec-essary for effective organizational learning, excessivedominance by one or the other will be dysfunctional. If allactors in the organization share the same specialized lan-guage, they will be effective in communicating with one an-other, but they may not be able to tap into diverse externalknowledge sources, tn the limit, an intemal language, codingscheme, or, more generally, any particular body of expertisecould become sufficiently overlapping and specialized that itimpedes the incorporation of outside knowledge and resultsin the pathology of the not-invented-here (NfH) syndrome.This may explain Katz and Alien's (1982) findings that thelevel of external communication and communication withother project groups declines with project-group tenure.
This trade-off between outward- and inward-looking compo-nents of absorptive capacity focuses our attention on how therelationship between knowledge sharing and knowledge di-versity across individuals affects the development of organi-zational absorptive capacity. While some overlap ofknowledge across individuals is necessary for internal com-munication, there are benefits to diversity of knowledgestructures across individuals that parallel the benefits to di-versity of knowledge within individuals. As Simon (1985)pointed out, diverse knowledge structures coexisting in thesame mind elicit the sort of learning and problem solving thatyields innovation. Assuming a sufficient level of knowledgeoverlap to ensure effective communication, interactionsacross individuals who each possess diverse and differentknowledge structures will augment the organization's ca-pacity for making novel linkages and associations—inno-vating—beyond what any one individual can achieve.Utterback (1971), summarizing research on task performanceand innovation, noted that diversity in the work setting "stim-ulates the generation of new ideas." Thus, as with Nelsonand Winter's (1982) view of organizational capabilities, an or-ganization's absorptive capacity is not resident in any singleindividual but depends on the links across a mosaic of indi-vidual capabilities.
Beyond diverse knowledge structures, the sort of knowledgethat individuals should possess to enhance organizational ab-sorptive capacity is also important. Critical knowledge doesnot simply include substantive, technical knowledge; it alsoincludes awareness of where useful complementary exper-tise resides within and outside the organization. This sort ofknowledge can be knowledge of who knows what, who canhelp with what problem, or who can exploit new information.With regard to extemal relationships, von Hippei (1988) has
133/ASa March 1990
shown the importance for innovation of close relationshipswith both buyers and suppliers. To the extent that an organi-zation develops a broad and active network of internal andextemal relationships, individuals' awareness of others' capa-bilities and knov^rledge wil! be strengthened. As a result, indi-vidual absorptive capacities are leveraged all the more, andthe organization's absorptive capacity is strengthened.
The observation that the ideal knowledge structure for an or-ganizational subunit should reflect only partially overlappingknowledge complemented by nonoverlapping diverse knowl-edge suggests an organizational trade-off between diversityand commonality of knowledge across individuals. Whilecommon knowledge improves communication, commonalityshould not be carried so far that diversity across individuals issubstantially diminished. Likewise, division of labor pro-moting gains from specialization should not be pushed so farthat communication is undermined. The difficulties posed byexcessive specialization suggest some liabilities of pursuingproduction efficiencies via learning by doing under conditionsof rapid technical change in which absorptive capacity is im-portant. In learning by doing, the firm becomes more prac-ticed and hence more capable at activities in which it isalready engaged. Learning by doing does not contribute to thediversity that is critical to learning about or creating somethingthat is relatively new. Moreover, the notion of "rememberingby doing" (Nelson and Winter, 1982) suggests that the focuson one class of activity entailed by learning by doing may ef-fectively diminish the diversity of background that an indi-vidual or organization may have at one time possessed and,consequently, undercut organizational absorptive capacity andinnovative performance.
It has become generally accepted that complementary func-tions within the organization ought to be tightly intermeshed,recognizing that some amount of redundancy in expertisemay be desirable to create what can be called cross-functionabsorptive capacities. Cross-function interfaces that affect or-ganizational absorptive capacity and innovative performanceinclude, for example, the relationships between corporate anddivisional R&D tabs or, more generally, the relationshipsamong the R&D, design, manufacturing, and marketing func-tions (e.g., Mansfield, 1968: 86-88). Close linkages betweendesign and manufacturing are often credited for the relativesuccess of Japanese firms in moving products rapidly fromthe design stage through development and manufacturing(Westney and Sakakibara, 1986). Clark and Fujimoto (1987)argued that overlapping product development cycles facilitatecommunication and coordination across organizational sub-units. They found that the speed of product development isstrongly influenced by the links between problem-solvingcycles and that successful linking requires "direct personalcontacts across functions, liaison roles at each unit, cross-functional task forces, cross-functional project teams, and asystem of 'product manager as integrator' " (Clark and Fuji-moto, 1987: 24). In contrast, a process in which one unitsimply hands off the design to another unit is likely to suffergreater difficulties.
Some management practices also appear to reflect the beliefthat an ext^ssive deigree of overlap in functions may reduce
134/ASQ, March 1990
the firm's absorptive capacity and that diversity of back-grounds is useful. The Japanese practice of rotating theirR&D personnel through marketing and manufacturing opera-tions, for example, while creating knowledge overlap, alsoenhances the diversity of background of their personnel.Often involving the assignment of technical personnel toother functions for several years, this practice also suggeststhat some intensity of experience in each of the complemen-tary knowledge domains is necessary to put an effective ab-sorptive capacity in place; breadth of knowledge cannot besuperficial to be effective.
The discussion thus far has focused on internal mechanismsthat influence the organization's absorptive capacity. A ques-tion remains as to whether absorptive capacity needs to beinternally developed or to what extent a firm may simply buyit via, for example, hiring new personnel, contracting for con-sulting services, or even through corporate acquisitions. Wesuggest that the effectiveness of such options is somewhatlimited when the absorptive capacity in question is to be inte-grated with the firm's other activities. A critical component ofthe requisite absorptive capacity for certain types of informa-tion, such as those associated with product and process in-novation, is often firm-specific and therefore cannot bebought and quickly integrated into the firm. This is reflected inLee and Allen's (1982) findings that considerable time lags areassociated with the integration of new technical staff, partic-ularly those concerned with process and product develop-ment. To integrate certain classes of complex andsophisticated technological knowledge successfully into thefirm's activities, the firm requires an existing internal staff oftechnologists and scientists who are both competent in theirfields and are familiar with the firm's idiosyncratic needs, or-ganizational procedures, routines, complementary capabilities,and extramural relationships. As implied by the discussionabove, such diversity of knowledge structures must coexistto some degree in the same minds. Moreover, as Nelson andWinter's (1982) analysis suggests, much of the detailedknowledge of organizational routines and objectives thatpermit a firm and its R&D labs to function is tacit. As a con-sequence, such critical complementary knowledge is acquiredonly through experience within the firm. Illustrating our gen-eral argument, Vyssotsky (1977), justifying the placement ofBell Labs within AT&T, argued: "For research and develop-ment to yield effective results for Bell System, it has to bedone b y . . . creative people who understand as much as theypossibly can about the technical state of the art, and aboutBell System and what System's problems are. The R&Dpeople must be free to think up new approaches, and theymust also be closely coupled to the problems and challengeswhere innovation is needed. This combination, if one is lucky,wilt result in insights which help the Bell System. That's whywe have Bell Labs in Bell System, instead of having ail ourR&D done by outside organizations."
Path Dependence and Absorptive Capacity
Our discussion of the character of absorptive capacity and itsrote in assimilating and exploiting knowtedge suggests asimple generalization that applies at both the individual andorganizational levels: prior knowledge permits the assimilation
13S/ASa Man^ 1990
A similar result emerges from models ofadaptive teaming. Levitt and March (1988;322) noted that "a competency trap can 'occur when favorable perfonnance withan inferior procedure leads an organizationto accumulate more experience vwth it,thus tceepmg e}9)erience with a superiorprocedure inadequate to make ft re-warding to use."
and exploitation of new knowledge. Some portion of that priorknowledge should be very closely related to the new knowl-edge to facilitate assimilation, and some fraction of thatknowledge must be falriy diverse, although still related, topermit effective, creative utilization of the new knowledge.This simple notion that prior knowledge underlies absorptivecapacity has important implications for the development ofabsorptive capacity over time and, in turn, the innovative per-formance of organizations. The basic role of prior knowledgesuggests two features of absorptive capacity that will affectinnovative performance in an evolving, uncertain environment(Cohen and Levinthal, 1989b). Accumulating absorptive ca-pacity in one period will permit its more efficient accumula-tion in the next. By having already developed some absorptivecapacity in a particular area, a firm may more readily accumu-late what additional knowledge it needs in the subsequentperiods in order to exploit any critical external knowledge thatmay become available. Second, the possession of related ex-pertise will permit the firm to better understand and thereforeevaluate the import of intermediate technological advancesthat provide signals as to the eventual merit of a new techno-logical development. Thus, in an uncertain environment, ab-sorptive capacity affects expectation formation, permitting thefirm to predict more accurately the nature and commercialpotential of technological advances. These revised expecta-tions, in turn, condition the incentive to invest in absorptivecapacity subsequently. These two features of absorptive ca-pacity—cumulativeness and its effect on expectation forma-tion—imply that its development is domain-specific and ispath- or history-dependent.
The cumulativeness of absorptive capacity and its effect onexpectation formation suggest an extreme case of path de-pendence in which once a firm ceases investing in its ab-sorptive capacity in a quickly moving field, it may neverassimilate and exploit new information in that field, regardlessof the value of that information. There are two reasons for theemergence of this condition, which we term "lockout"(Cohen and Levinthai, 1989b). First, if the firm does not de-velop its absorptive capacity in some initial period, then itsbeliefs about the technological opportunities present in agiven field will tend not to change over time because the firmmay not be aware of the significance of signals that wouldotherwise revise its expectations. As a result, the firm doesnot invest in absorptive capacity and, when new opportunitiessubsequently emerge, the firm may not appreciate them.Compounding this effect, to the extent that prior knowledgefacilitates the subsequent development of absorptive ca-pacity, the lack of early investment in absorptive capacitymakes it more costly to develop a given level of it in a subse-quent period. Consequently, a low initial investment in ab-sorptive capacity diminishes the attractiveness of investing insubsequent periods even if the firm becomes aware of tech-nological opportunities.'' This possibility of firms being"locked-out" of subsequent technological developments hasrecently become a matter of concern with respect to indus-trial policy. For instance, Reich (1987: 64) declaims Mon-santo's exit from "float-zone" silicon manufacturing becausehe believes that the decision may be an irreversible exit froma technology, in t h a t " . . . each new generation of technology
136/ASa March 1990
M>sc»ptive Capadty
builds on that which came before, once off the technologicalescalator it's difficult to get back on."
Thus, the cumulative quality of absorptive capacity and its rolein conditioning the updating of expectations are forces thattend to confine firms to operating in a particular technologicaldomain, if firms do not invest in developing absorptive ca-pacity in a particular area of expertise early on, it may not bein their interest to develop that capacity subsequently, evenafter major advances in the field. Thus, the pattern of inertiathat Nelson and Winter (1982) highlighted as a central featureof firm behavior may emerge as an implication of rational be-havior in a model in which absorptive capacity is cumulativeand contributes to expectation formation. The not-invented-here syndrome, In which firms resist accepting innovativeideas from the environment, may also at times reflect whatwe call lockout. Such ideas may be too distant from the firm'sexisting knowledge base—its absorptive capacity—to be ei-ther appreciated or accessed. In this particular setting, NIHmay be pathological behavior only in retrospect. The firmneed not have acted irrationally in the development of the ca-pabilities that yields the NIH syndrome as its apparent out-come.
A form of self-reinforcing behavior similar to lockout may alsoresult from the influence of absorptive capacity on organiza-tions' goals or aspiration levels. This argument builds on thebehavioral view of organizational innovation that has beenmolded in large part by the work of March and Simon (1958).In March and Simon's framework, innovative activity is insti-gated due to a failure to reach some aspiration level. De-parting from their model, we suggest that a firm's aspirationlevel in a technologically progressive environment is notsimply determined by past performance or the performanceof reference organizations. It also depends on the firm's ab-sorptive capacity. The greater the organization's expertise andassociated absorptive capacity, the more sensitive it is likelyto be to emerging technological opportunities and the morelikely its aspiration level will be defined in terms of the oppor-tunities present in the technical environment rather thanstrictly in terms of performance measures. Thus, organiza-tions with higher levels of absorptive capacity will tend to bemore proactive, exploiting opportunities present in the envi-ronment, independent of current performance. Alternatively,organizations that have a modest absorptive capacity will tendto be reactive, searching for new alternatives in response tofailure on some performance criterion that is not defined interms of technical change per se (e.g., profitability, marketshare, etc.).
A systematic and enduring neglect of technical opportunitiesmay result from the effect of absorptive capacity on the orga-nization's aspiration level when innovative activity (e.g., R&D)contributes to absorptive capacity, which is often the case intechnologically progressive environments. The reason is thatthe firm's aspiration level then depends on the very innova-tive activity that is triggered by a failure to meet the aspirationlevel itself. If the firm engages in little innovative activity, andis therefore relatively insensitive to the opportunities in theexterna! environment, it will have a low aspiration level withregard to the exploitation of new technology, which in tum
137/ASa March
This argument that sucix reactive andproactivB behavior nnay coexist in an in-dustry over the tong run assumes thattnere is sladc in the selecticm environmentand that te^m^ogicatty progressive be-havior is not essential to survh^l. One can,attematrv^, identify a nuniber of indus-tnes, such as semicor^uctors, in which ite ipears that o r ^ flnns tiiat aggte^ivetyfficpkxt technical cq: x>rUir»ties surwe.
implies that it will continue to devote little effort to innovation.This creates a self-reinforcing cyde. Likewise, if an organiza-tion has a high aspiration level, influenced by exterrially gen-erated technical opportunities, it will conduct more innovativeactivity and thereby increase its awareness of outside of Dor-tunities. Consequently, its aspiration level will remain high.This argument implies that reactive and proactive modes offirm behavior should remain rather stable over time. Thus,some organizations (like Hewlett-Packard and Sony) have therequisite technical knowledge to respond proactively to theopportunities present in the environment. These firms do notwait for failure on some performance dimension but aggres-sively seek out new opportunities to exploit and develop theirtechnological capabilities.*
The concept of dynamically self-reinforcing behavior that maylead to the neglect of new technological developments pro-vides some insight into the difficulties firms face when thetechnological basis of an industry changes—what Schum-peter (1942) called "the process of creative destruction." Forinstance, the change from electromechanical devices to elec-tronic ones in the calculator industry resulted in the exit of anumber of firms and a radical change in the market structure(Majumdar, 1982). This is an example of what Tushman andAnderson (1986) termed competence-destroying technicalchange. A firm without a prior technological base in a partic-ular field may not be able to acquire one readily if absorptivecapacity is cumulative. In addition, a firm may be blind to newdevelopments in fields in which it is not investing if its up-dating capability is low. Accordingly, our argument impliesthat firms may not realize that they should be developing theirabsorptive capacity due to an irony associated with its valua-tion: the firm needs to have some absorptive capacity alreadyto value it appropriately.
Absorptive Capacity and R&D Investment
The prior discussion does not address the question ofwhether we can empirically evaluate the importance of ab-sorptive capacity for innovation. There is a key insight thatpemnits empirical tests of the implications of absorptive ca-pacity for innovative activity. Since technical change within anindustry—typically incremental in character (Rosenberg andSteinmueller, 1988)—is often closely related to a firm's on-going R&D activity, a firm's ability to exploit external knowl-edge is often generated as a byproduct of its R&D. We maytherefore consider a firm's R&D as satisfying two functions:we assume that R&D not only generates new knowledge butalso contributes to the firm's absorptive capacity.' If absorp-tive capacity is important, and R&D contributes to it, thenwhatever conditions the firm's incentives to leam (i.e., tobuild absorptive capacity) should also influence R&Dspending. We may therefore consider the responsiveness ofR&D activity to leaming incentives as an indication of the em-pirical importance of absorptive capacity. The empirical chal-lenge then is to ur>derstand the impact of the characteristicsof the leaming environment on R&D
of the theoretic^ and st^^equent empir-ical an^f^ md rasute to Cohmi « id te-vmthal (1989a), from n^w^h the fc^lowngdtoiission is drami.
We constfuct a simple static nnodet of firm R&D intensity,which is defined as R&D divided by sales. Normalization ofR&D by film sales controis for the effect of firm size, which
138MSa March 1990
This secoTKl dimension is incorpwated inthe modet developed in Cohen and te-wrtivi (19898). We do not incorporate thiswcond dfetwnsion in tf» present moctelbecause all the qui^itative tfwOTetical ande n w i c ^ rmults associated with ttiissewnd (fimen»an of tedwolos^cai t^ipor-u » ^ ate the same as ^tose associatedwith the first consid««d here.
affects the return per unit of R&D effort. This model is devel-oped in the broader context of what applied economists havecome to believe to be the three classes of industry-level de-terminants of R&D intensity: demand, appropriability, andtechnological opportunity conditions (Cohen and Levin, 1989).Demand is often characterized by the level of sales and theprice elasticity of demand. The latter indicates the degree towhich a firm's revenue will increase due to a reduction inprice. For example, in the case of a process innovation thatreduces the cost of production and, in turn, the product price,the price elasticity of demand reflects the associated changein total revenue that influences the economic return to inno-vative effort. Appropriability conditions refer to the degree towhich firms capture the profits associated with their innova-tive activity and are often considered to reflect the degree towhich valuable knowledge spills out into the public domain.The emphasis here is on valuable knowledge, because if acompetitor's knowledge spills out but the competitor has al-ready exploited a first-mover advantage in the marketplace,this knowledge is no longer valuable to the firm and does notconstitute a spillover by our definition. The level of spillovers,in tum, depends on the strength of patents within an industry,the efficacy of secrecy, and/or first-mover advantages. Tech-nologicai opportunity represents how costly it is for the firmto achieve some normalized unit of technical advance in agiven industry. As typically conceived, there are two dimen-sions of technological opportunity (Cohen and Levin, 1989).The first, incorporated in our model, refers simply to thequantity of extraindustry technological knowledge, such asthat originating from government or university labs, that ef-fectively complements and therefore leverages the firm'sown knowledge output. The second dimension of technolog-ical opportunity is the degree to which a unit of new knowl-edge improves the technological performance of the firm'smanufacturing processes or products and, in turn, the firm'sprofits. For example, given the vitality of the underlyingscience and technology, an advance in knowledge promisesto yield much larger product-performance payoffs in thesemiconductor industry than in steel.*
The basic model of how absorptive capacity affects the de-termination of R&D expenditures is represented diagramati-cally in Figure 1. We postulate that leaming incentives willhave a direct effect on R&D spending. We also suggest thatwhere the effect of other determinants, such as technologicalopportunity and appropriability, depend on the firm's or rivals'assimilation of knowledge, absorptive capacity—and there-fore learning incentives—will mediate those effects. Finally,we suggest that the effect of appropriability conditions (i.e.,spillovers) will be conditioned by competitor interdependence.In this context, we define interdependence as the extent towhich a rival's technical advances diminish the firm's profits.
There are two factors that will affect a firm's incentives tolearn, and, therefore, its incentives to invest in absorptive ca-pacity via its R&D expenditures. First, there is the quantity ofknowledge to be assimilated and exploited: the more there is,the greater the incentive. Second, there is the difficulty (or,conversely, the ease) of leaming. Some types of infomnationare more difficult to assimilate and use than others. We inter-
T990
Hgw* 1. Model erf absorptive capacity and RftD incentives.
TechnologicalOpportunity
CompetitorInterdependence
Appropriability
AbsorptiveCapacity
R&D Spending
pret this to mean that per unit of knowledge, the cost of itsabsorption may vary depending on the characteristics of thatknowledge. As learning is more difficult, more prior knowl-edge has to have been accumulated via R&D for effectiveteaming to occur. As a result, this is a more costly learningenvironment. In such a setting, R&D is more important tobuilding absorptive capacity and the more R&D effort the firmwill need to have expended to achieve some level of absorp-tive capacity. Thus, for a given level of a firm's own R&D, thelevel of absorptive capacity is diminished in environments inwhich it is more difficult to learn. In addition, we are sug-gesting that a more difficult teaming environment increasesthe marginal effect of R&D on absorptive capacity. In con-trast, in environments in which learning is less demanding, afirm's own R&D has little impact on its absorptive capacity, Inthe extreme case in which external knowledge can be assim-ilated without any specialized expertise, a firm's own R&Dwould have no effect on its absorptive capacity.
We have argued that the ease of learning is in tum deter-mined by the characteristics of the underlying scientific andtechnological knowledge. Although it is difficult to specify apriori all the relevant characteristics of knowledge affectingthe ease of learning, they would include the complexity of theknowledge to be assimilated and the degree to which theoutside knowledge is targeted to the needs and concerns ofthe firm. When outside knowledge is less targeted to thefirm's particular needs and concerns, a firm's own R&D be-comes more important in permitting it to recognize the valueof the knowledge, assimilate, and exploit it. Sources that pro-duce less targeted knowledge would include university labsinvolved in taasic research, while more targeted knowledgemay be generated by contract research labs, or input sup-pliers. In addition, the degree to which a fietd is cumulative, orthe field's pace of advance, should also affect how criticalR&D is to the development of absorptive capacity. The morethat findings in a field build on prior findings, the more neces-sary is an understanding of prior research to the assimilationof subsequent findings. The pace of advance of a field affectsthe importance of R&D to devetoping absorptive capacity be-cause the faster the pace of knowtedge generation, the larger
140/ASQ, March 1990
Abso^ptlva Capacity
the staff required to keep abreast of new developments. Fi-nally, following Nelson and Winter (1982), the less explicit andcodified the relevant knowledge, the more difficult it is to as-similate.
To structure the analysis, we assumed that finns purposefullyinvest in R&D to generate profit and take into account R&D'sdual role in both directly generating new knowledge and con-tributing to absorptive capacity. Knowledge is assumed to beuseful to the firm in that increments to a firm's own knowl-edge increase the firm's profits while increments to rivals'knowledge diminish them. We posit a simple model of thegeneration of a firm's technological knowledge that takes intoaccount the major sources of technological knowledge uti-lized by a firm: the firm's own R&D knowledge that originateswith its competitors' R&D, spillovers, and that which origi-nates outside the industry. Figure 2 provides a stylized repre-sentation of this model in which, first, the firm generatesnew knowledge directly through its own R&D, and second,extramural knowledge, drawn from competitors as well asextraindustry sources such as government and universitylabs, also contribute to the firm's knowledge. A central fea-ture of the model is that the firm's absorptive capacity deter-mines the extent to which this extramural knowledge isutilized, and this absorptive capacity itself depends on thefirm's own R&D. Because of this mediating function, absorp-tive capacity influences the effects of appropriability andtechnologicai opportunity conditions on R&D spending. Thus,the effects of appropriabiiity and technological opportunity arenot independent of R&D itself.
Figure 2. Model of sources of a firm's technical Itnowiedge.
Own R&D
Absorptive Capacitv .
\
Technical' Knowledge
Spillovers of Competitors' KnowledgeExtraindustry Knowledge
A key assumption in the model is that exploitation of com-petitors' research findings is realized through the interactionof the firm's absorptive capacity with competitors' spillovers.This interaction signifies that a firm is unable to assimilateextemally available knowledge passively. Rather, to utilize theaccessible R&D output of its competitors, the firm invests inits absorptive capacity by conducting R&D. Figure 2 also illus-trates that, like its assimilation of competitors' R&D output, afirm's assimilation of extraindustry knowledge—the dimen-sion of technological opportunity considered here—is con-strained by its absorptive capacity. According to our model,therefore, the factors that affect learning incentives (i.e., theease of learning and the quantity of available knowledge) in-fluence the effects of appropriability and technological oppor-tunity conditions on RM).
U1/ASQ, March 1990
Direct effect of ease of learning. As shown formally inCohen and Levinthal (1989aK this model implies that as theease of leaming diminishes, teaming becomes more depen-dent on a ftrm's own R&D, and R&D spending increases be-cause of two effects. First the marginal impact of R&D onabsorptive capacity is greater in more difficult learning envi-ronments. As the leaming environment becomes more diffi-cult, however, there is a second, more subtle effect. Since,ceteris paribus, a more difficult leaming environment lowersfirms' absorptive capacities, R&D activity becomes more of aprivate good in the sense that competitors are now less ableto tap into the fimn's R&D findings that spill out.
Technological opportunity. We predict that an increase intechnological opportunity—the amount of available relevantextemal technical knowledge—will elicit more R&D in moredifficult teaming environments. Greater technological oppor-tunity signifies greater amounts of external information,which increase the firm's incentive to build absorptive ca-pacity, and a more challenging learning environment in-creases the level of R&D necessary to build absorptivecapacity.
Appropriabiiity. We predict that spillovers will provide, inpart, a positive incentive to conduct R&D due to the interac-tion of spillovers with an endogenous absorptive capacity.Traditionally, spillovers have been considered only a deterrentto R&D activity (e.g.. Nelson, 1959; Arrow, 1962; Spence.1984). In the standard view, a firm's incentive to invest inR&D is diminished to the extent that any findings from suchactivities are exploited by competitors and thereby diminishthe innovator's own profits. In our framework, however, thisnegative appropriability incentive associated with spillovers iscounterbalanced by a positive absorptive-capacity-building in-centive. The more of its competitors' spillovers there are outthere, the more incentive the firm has to invest in its ownR&D, which permits it to exploit those spillovers.
We have shown elsewhere (Cohen and Levinthal, 1989a) thatwhen this absorption incentive is large, as when leaming isdifficult, spillovers may actually encourage R&D. The relativemagnitude of the absorption incentive is greater when firmswithin an industry are less interdependent in the sense thatrivals' technical advances have less of an effect on the firm'sown profits. With tess interdependence, the degree to whichrivals gain from the firm's R&D spillovers at the firm's ex-pense diminishes relative to the benefit of being able to ex-ploit the rivals' spillovers. Either a more competitive marketstructure or a higher price elasticity of demand for the firm'sproduct can diminish interdependence in an industry.
METHODS
Data and Measures
To test the predictions of our framework for R&D activity, weused cross-sectional survey data on technological opportunityand appropriability conditions in the American manufacturingsector collected from R&D tal? managers by Levin et al. {1983,1987), and the Federal Trade Commission's Line of BusinessProgram data on business unit sales, transfers, and R&D ex-penditures. The dependent variable, R&D intensity, was de-
142/ASa Man^ 1990
A t t h o i ^ geology v ras dassed as a basicscience bv Levin et ai.. we classed it asan ^ ^ ^ed science because of its mductivemethodology and intensive use by finnsin tiw extractive sector.
Mssorptiva C^MCilv
fined as company-financed business-unit research anddevelopment expenditures, expressed as a percentage ofbusiness unit sales and transfers over the period 1975through 1977. The data on interindustry differences in tech-nological opportunity and appropriability are industry (line ofbusiness) mean scores computed as an average over all re-spondents within a given industry. The sample consists of1,719 business units representing 318 firms in 151 lines ofbusiness.
The data pose two estimation issues. First, some 24 percentof the firms performed no R&D in at least one year. If the in-dependent variables reflect both the probability of conductingR&D, as well as the amount of R&D spending, then a Tobitanalysis would be appropriate. Alternatively, a firm may re-quire some initial level of absorptive capacity before it is in-fluenced by the characteristics of the learning environment. Inthis case, the variables reflecting the ease of learning only af-fect the amount of R&D conducted by firms engaging inR&D activity and not the probability of engaging in R&D ac-tivity. In light of the uncertainty over the appropriate estima-tion technique, we explored the robustness of the results byanalyzing a Tobit and an OLS (or GLS) specification. Thesecond estimation issue is the presence of heteroscedasti-city. We found the assumption of homoscedasticity to be vio-lated, with the logarithm of the error variance being a linearfunction of the exogenous variables and the number of re-spondents to Levin et al.'s (1983, 1987) survey. Unless other-wise noted, the results we report in this section reflectrobust effects that hold across three different estimationmethods, including ordinary least squares (OLS), generalized(east squares (GLS) in which we adjust for heteroscedasticity,and Tobit, which was used when we included the observa-tions for which R&D expenditures were zero.
We tested our predictions in the context of an empiricalmodel of business unit R&D intensity in which technologicalopportunity, appropriability, and demand conditions are con-sidered as the principal industry-level determinants of firms'R&D spending. While data constraints do not permit observa-tion of the direct effect of the ease of leaming or its deter-minants on firms' R&D spending, we were able to examinehow these variables condition the influence on R&D of tech-nological opportunity and appropriability conditions.
Technological opportunity was assessed with variables mea-suring the "relevance" or "importance" for technologicalprogress in each line of business of what are considered tobe two critical sources of technological opportunity—thescience base of the industry and extraindustry sources ofknowledge (Cohen and Levin, 1989). These measures aredrawn from Levin et al.'s survey, in which R&D managers in-dicated on a 7-point Likert scale the relevance of eleven basicand applied fields of science and the importance of externalsources of knowledge to technological progress in a line ofbusiness. The basic fields of science include biology, chem-istry, mathematics, and physics, and the applied fields ofscience include agricultural science, applied math/operationsresearch, computer science, geology, materials science,medical science, and metallurgy." The five extraindustry
143/ASQ, March 1990
sources of knowledge considered here included equipmentsuppliers (EOUIPTECH), materials suppliers (MATERIAL-TECH), downstream users of the industry's products (USER-TECH), government laboratories and agencies (GOVTECH),and universities (UNIVTECH). We interpreted the measures ofthe relevance or importance of each field or knowledgesource to index the relative quantity of knowledge generatedby that field or source that is potentially useful. We then dis-tinguished across the eleven scientific fields and the five ex-traindustry knowledge source variables on the basis of theease of leaming associated with each. We suggested abovethat one important determinant of the ease of leaming is thedegree to which outside knowledge is targeted to a firm'sneeds and concerns. One can readily distinguish among boththe eleven fields and the five extraindustry knowledgesources on that basis. The knowledge associated with thebasic sciences is typically less targeted than that associatedwith the applied sciences. We also distinguished among theextraindustry knowledge sources on the same basis. A priori,we ranked university labs, government labs, materials sup-pliers, and equipment suppliers as providing increasinglymore targeted knowledge to firms. We did not rank the rela-tive effect of knowledge originating from users because, assuggested by von Hippel (1978), users will often provide aproduct idea to potential suppliers, but the informativeness ofthe "solution concept" is quite variable. Therefore, the tar-geted quality of the information is variable as well.
To represent intraindustry spillovers of R&D, we employedmeasures from Levin et al.'s survey of the effectiveness ofsix mechanisms used by firms to capture and protect thecompetitive advantages of new processes and new products:patents to prevent duplication, patents to secure royalty in-come, secrecy, lead time, moving quickly down the leamingcurve, and complementary sales and service efforts. We em-ployed the maximum value of the effectiveness scores at-tained by these mechanisms as our measure of appropriabilityor spillovers, and label this variable APPROPRIABILITY; a highleve! of APPROPRIABILiTY reflects a low level of spillovers.
In our theory, we predicted an interaction effect by which, asthe ease of leaming diminishes, or firms become less inter-dependent, the effect of spillovers on R&D spending shouldbecome more positive (or less negative), in the absence ofany direct measure of the ease of leaming, we distinguishedcategorically between those industries in which basic sciencewas more relevant to technical progress than the relativelymore targeted applied sciences and assumed that leamingwas generally less difficult in industries that fell into the lattercategory. Thus, we created a dummy variable, DUMBAS,that equals one when the average value of the relevancescores associated with the basic fields exceeds that asso-ciated with the applied fields and that equals zero otherwise.We specified the dummy variable, DUMAPP, analogously. Tocapture the interdependence of firms, we employed mea-sures of industries' competitiveness as represented by eachindustry's four-firm concentration ratio (C4) and industry-levelestimates of the price elasticity of demand (PELAS).
To further control for industry demand conditions, we usedindustry estimates developed by Levin (1981) of price elas-
144/ASa March 1990
Absoipiive CiM»^ty
ticity (PELAS) and income elasticity (INCELAS) and a demandtime-shift parameter (DGROWTH). Finally, we included an-other control variable that may also reflect technological op-portunity, industry maturity. We used a somewhat crudemeasure of industry maturity, NEWPLANT, that measures thepercentage of an industry's property, plant, and equipmentinstalled within the preceding five years.
RESULTS
Technological opportunity. Our theory suggests that whenthe targeted quality of knowledge is less (i.e., leaming is moredifffcult), an increase in the relevance (i.e., quantity) of knowl-edge should have a more positive effect on R&D intensity.Therefore, the coefficient estimates of the variables mea-suring the relevance of the four basic scientific fields shouldexceed those of the variables measuring the relevance of theseven applied scientific fields. Confirming the prediction.Table 1 indicates that the estimated coefficients for the ap-plied sciences are, with the exception of computer science,lower than that for the basic sciences. The similarity of theestimate of the effect of the relevance of computer science,an applied science, to those of some of the basic sciencessuggests that the assumption may not be correct that onlyone determinant of the ease of learning, the targeted qualityof the field, varies systematically across the fields of appliedand basic science. Another determinant of the ease oflearning postulated above is a field's pace of advance, wherefaster pace should require more R&D to permit assimilation,and the pace of advance in computer science has been rela-tively rapid over the past two decades.
To further test the prediction that the coefficient values of theless targeted, basic science field variables would exceedthose of the applied fields, we estimated a specification, oth-erwise identical to the first, in which we constrained the co-efficients of the basic sciences to be the same and thecoefficients of the applied sciences to be the same. Thisshows the effect on R&D spending as the overall technolog-ical opportunity associated with basic science and appliedscience, respectively, change. The constrained coefficient es-timates of the effect of the technological opportunity asso-ciated with the basic and applied sciences are significantlydifferent (at the p < .01 level) across all estimation methods,with the former equal to .189 and the latter equal to - .080 inthe GLS estimation. Therefore, relative to the effect of an in-crease in the technological opportunity associated with ap-plied science, an increase in that associated with basicscience elicits more R&D.
Our predicted ranking of the coefficient magnitudes asso-ciated with the extraindustry sources of knowledge, reflectingincreasingly targeted knowledge from these sources, islargely confirmed. The coefficient estimate for the importanceof knowledge originating from universities exceeds that forgovemment labs, which, in tum, is greater than that for ma-terials suppliers, which exceeds that for equipment suppliers.The difference between coefficient values is statistically sig-nificant in the case of govemment sources versus materialssuppliers for both the OLS and Tobit results (p < .01) and inthe case of materials suppliers versus equipment suppliers in
U5/ASQ, March 1990
Table 1
Anatysis of m o fentansity*
Variable
Intercut
APPROPRIABUmr X C4
APPROPRIABILITY x PELAS
APPROPRtABtLITY x DUMAPP
APPROPRlABiLITY x DUMBAS
USERTECH
UNIVTECH
GOVTECH
MATERIALTECH
EOUiPTECH
Biology
Chemistry
Math
Physics
Agricultural Science
/^ptied Math/Operations Research
Computer Science
Geology
Materials Science
Medical Science
Metallurgy
NEWPLANT
PEUS
INCELAS
DGROWTH
FP
• p < .05;-p<.01.• R^roduced from Cohen and Lewnthal (1989a:
OLSW = 1302)
- 5 . 1 8 4 -(1.522}
.213(.128)
-.192(-106).448*
(.202).302
(.208)4 7 0 -
(.104).374-
(.131).221*
(.106)- . 2 5 8 -
(.098)_ . 4 0 1 -
(.111).314-
(.102).289-
(.084).184
(.131),373-
(.117)- . 4 4 1 -(.088)
-.237(.148).294*
(.124)- . 3 6 3 -
(.084)-.110(.t25)
-.179{.0931
- . 3 1 5 -(.077).057-
(.008).936
(.611)1.077-(.170).068
(.090).287
590-591,569-596).
Regrassion CoefUcientGLS
{N = 1302)
-2.355*(1.037)
.342-(.103)
- ,200-(.091).248
(.143).174
(.144).397-
(.069).318-
(.091).069
(.079)-.074
(.070)- . 4 8 4 -
(.077).185-
(.071).081
(.062).151
(.097).323-
(.091)- . 2 7 3 -(.064)
-.117(.102).116
(.090)- . 2 4 0 -(.061)
-.150(.095)
- .133(.070)
- . 1 9 5 -(.053).049-
(-006)1.082*(.627).587-
(.131)-.074
(.063)
Standard errors are in parentheses
Tobit{/V = 1719)
- 4 . 0 8 6 -0.461)
.368-(.130)
- .176(.103).211
(,194).094
(.206).612-
(-107).395-
(.147).137
(.107)- . 3 0 3 -
(.100)- . 5 7 4 -
(.117).276*
(.114).191-
(.088).123
(.143).310*
(-128)- - 3 0 8 -(,099)
-.366*(.152).433-
(,122)- .365-(.097).116
(.118)-,133
(103)- . 3 9 3 -
(.0^).045-
(.007).892
(-573)1.112-(.188).004
(.105)
the GLS results (p < .01). While we had no prediction re-garding the coefficient value for USERTECH, the consistentlyhigh value of the coefficient estimate may reflect some ele-ment of demand conditioos. Consistent with this, we haveobserved the variable USERTECH to be significantly corre-lated with measures of the importance of product differentia-tion {cf. Cohen and Levinmal, 1989a).
146/ASQ, Match 1990
AbsOTptiv* Capacity
Appropriability. The results largely support the predictionthat the ease of leaming conditions the effect of knowledgespillovers. The effect on R&D intensity of increasing appro-priability (i.e., diminishing spillovers) was significantly greater(p < .05) in those industries in which the applied sciences aremore relevant to innovation than the basic sciences. This re-sult suggests that the positive absorption incentive asso-ciated with spillovers is greater in industries in which thedifficulty of learning is greater. Second, there is a significantpositive effect (p < .01) of the interaction between marketconcentration and the appropriability level. As market concen-tration increases (indexing a diminution in competitiveness),the positive effect of a given appropriability level on R&D in-tensity increases, as predicted. Likewise, the effect of the in-teraction of the price elasticity of demand and the level ofappropriability is negative (but only significant at p < .05 inthe GLS estimate), providing additional support for the propo-sition that the positive effect of spillovers will increase in in-dustries in which firms are less interdependent. The resultssuggest that the learning environment affects the impact ofspillovers on R&D spending and that the importance of thepositive absorptive-capacity-building incentive relative to thatof the negative appropriability incentive is conditioned by thedegree of competitor interdependence.
While we have shown that the learning environment modifiesthe effect of appropriability conditions, the question remainswhether spillovers may, on balance, actually encourage R&Din some industries. To explore this possibility, we examinedthe effect of spillovers in the four two-digit SIC code level in-dustries for which our sample contains enough lines of busi-ness to permit separate industry regressions. These includeSICs 20 (food processing). 28 (chemicals), 35 (machinery),and 36 (electrical equipment). Due to the reduction in the de-grees of freedom for industry-level variables, we simplifiedthe estimating equation to consider only the direct effect ofAPPROPRIABILITY. and the science field variables weresummarized as the maximum relevance scores attained bythe basic and applied fields, respectively. In SICs 28 and 36,the effect of the APPROPRIABILITY variable was negativeand significant at conventional levels, implying that R&D in-tensity rises with spillovers. In the Tobit results, the sign wasalso positive for SICs 28 and 36. but the coefficient esti-mates were not quite significant at the .05 confidence ievei.Thus, in SICs 28 (chemicals) and 36 (electrical equipment),R&D intensity rose with spillovers when we controlled forother industry-level variables conventionally thought to driveR&D spending, including technological opportunity and de-mand conditions. Although the analyses showing a positiveeffect of spillovers in these two industry groups do not repre-sent a direct test of our model, the results suggest, particu-larly when considered with the interaction results, that thepositive absorption incentive associated with spillovers maybe sufficiently strong in some cases to more than offset thenegative appropriabiltty incentive.
JMPLICATIONS FOR INNOVATIVE ACTIVITY
Drawing on our prior vtrark (Cohen and Levinthal, 1987,1989a), we offer some impHcations of absorptive capacity for
147/ASa Maroh 1990
Markets for information often fail becausethey inherency represent a situation of in-formation asymmetry in which the less in-formed party cannot property value thetnforma^n he or she wishw to purchase,and the more informoj party, acting self-interestedly. attempts to e?q]loit itiat in-abittty (WWKBIISCWI, 1975).
the analysis of other innovative activities, including basic re-search, the adoption and diffusion of innovations, and deci-sions to participate in cooperative R&D ventures, that followfrom the preceding analyses.
The observation that R&D creates a capacity to assimilate andexploit new knowledge provides a ready explanation of whysome firms may invest in basic research even when the pre-ponderance of findings spill out into the public domain. Spe-cifically, firms may conduct basic research less for particularresults than to be able to provide themselves with the generalbackground knoyvledge that would permit them to exploitrapidly useful scientific and technological knowledge throughtheir own innovations or to be able to respond quickly—be-come a fast second—when competitors come up with amajor advance (see also Rosenberg, 1990). In terms of ourdiscussion of the cognitive and organizational aspects of ab-sorptive capacity, we may think of basic research as broad-ening the firm's knowledge base to create critical overlapwith new knowledge and providing it with the deeper under-standing that is useful for exploiting new technical develop-ments that build on rapidly advancing science and technology.
This perspective on the role of basic research offers a ratherdifferent view of the determinants of basic research than thatwhich has dominated thinking in this area for the thirty yearssince Nelson's (1959) seminal article. Nelson hypothesizedthat more diversified firms will invest more heavily in basicresearch because, assuming imperfect markets for informa-tion, they will be better able to exploit its wide-ranging andunpredictable results. Nelson thus saw product-market diver-sification as one of the key determinants of basic research.®Emphasizing the role of basic research in firm learning, ourperspective redirects attention from what happens to theknowledge outputs from the innovation process to the natureof the knowledge inputs themselves. Considering that ab-sorptive capacity tends to be specific to a field or knowledgedomain means that the type of knowledge that the firm be-lieves it may have to exploit will affect the sort of researchthe firm conducts. From this vantage point, we would conjec-ture that as a firm's technological progress becomes moreclosely tied to advances in basic science (as has been thecase in Pharmaceuticals), a firm will increase its basic re-search, whatever its degree of product-market diversification.We also suggest, with reference to all firm research, not justbasic research, that as the fields underlying technical advancewithin an industry become more diverse, we may expectfirms to increase their R&D as they develop absorptive ca-pacities in each of the relevant fields. For example, as auto-mobile manufacturing comes to draw more heavily on newerfields such as microelectronics and ceramics, we expect thatmanufacturers will expand their basic and applied researchefforts to better evaluate and exploit new findings in theseareas.
The findings on the role of absorptive capacity and the waysin which it may be developed also have implications for theanalysis of the adoption and diffusion of innovations. Our per-spective implies that the ease of leaming, and thus tech-nology adoption, is affected by the degree to which aninnovation is related to the pre-existing knowledge base of
148/ASa March 1990
Misorptive Capadty
prospective users. For example, personal computers diffusedmore rapidly at the outset among consumers and firms whohad prior experience on mainframes or minicomputers. Like-wise, software engineering practices seem to be adoptedmore readily by programmers with previous Pascal ratherthan Fortran experience because the structure of Pascal moreclosely reflects some of the underlying principles of softwareengineering (Smith et a!., 1989). Our argument also suggeststhat an innovation that is fully incorporated in capital equip-ment wilt diffuse more rapidly than more disembodied inno-vations that require some complementary expertise on thepart of potential users. This is one of the anticipated benefitsof making computers more "user friendly."
The importance of absorptive capacity also helps explainsome recent findings regarding firms' cooperative researchventures. First, Link (1987) has observed that cooperative re-search ventures are actually found more typically in industriesthat employ more mature technologies rather than in indus-tries in which technology is moving ahead quickly—as seemsto be suggested by the popular press. Second, it has beenobserved that cooperative ventures that have been initiated topursue basic research, as well as more applied research ob-jectives, have been subject over the years to increasing pres-sure to focus on more short-term research objectives(Mowery and Rosenberg, 1989). The simple notion that it isimportant to consider the costs of assimilating and exploitingknowledge from such ventures provides at least a partial ex-planation for these phenomena. Many cooperative venturesare initiated in areas in which the cost to access the output ofthe venture is low, or they often gravitate toward such areasover time. Conversely, those who are attempting to en-courage cooperative research ventures in quickly advancingfields should recognize that the direct participation in the ven-ture should represent only a portion of the resources that itwill take to benefit from the venture. Participating firms alsomust be prepared to invest internally in the absorptive ca-pacity that will permit effective exploitation of the venture'sknowledge output.
CONCLUSION
Our empirical analysis of R&D investment suggested thatfirms are in fact sensitive to the characteristics of the leamingenvironment in which they operate. Thus, absorptive capacityappears to be part of a firm's decision calculus in allocatingresources for innovative activity. Despite these findings, be-cause absorptive capacity is intangible and its benefits are in-direct, one can have little confidence that the appropriatelevel, to say nothing of the optimal level, of investment in ab-sorptive capacity is reached. Thus, while we have proposed amodel to explain R&D investment, in which R&D both gen-erates innovation and facilitates learning, the development ofthis model may ultimately be as valuable for the prescriptiveanalysis of organizational policies as its application may be asa positive model of firm behavior.
An important question from a prescriptive perspective isWhen is a firm most likely to underinvest in absorptive ca-pacity to its own long-run detriment? Absorptive capacity ismore likely to be developed and maintained as a byproduct of
149/ASa March 1990
routine activity when the knowledge domain that the firmwishes to exploit is closety related to its current knowledgebase. When, however, a firm wishes to acquire and useknowledge that is unrelated to its ongoing activity, then thefirm must dedicate effort exclusively to creating absorptivecapacity (i.e., absorptive capacity is not a byproduct). In thiscase, absorptive capacity may not even occur to the firm asan investment altemative. Even if it does, due to the intan-gible nature of absorptive capacity, a firm may be reluctant tosacrifice current output as well as gains from specialization topennit its technical personnel to acquire the requisite breadthof knowledge that would permit absorption of knowledgefrom new domains. Thus, while the current discussion ad-dresses key features of organizational structure that deter-mine a firm's absorptive capacity and provides evidence thatinvestment is responsive to the need to develop this capa-bility, more research is necessary to understand the decisionprocesses that determine organizations' investments in ab-sorptive capacity.
Abemathy, William J.1978 The Productivity Dilemma. Bal-
timore: Johns Hc^kins Univer-sity Press.
Mien, Thomas J.1977 Managing the Flow of Tech-
nology. Cambridge, MA: MITPress.
Alien, Thomas J., and Stephen D.
"tnfomiatlon flows in R&Dlabs." /Wmintstrative ScienceQuarterly, 20: 12-19.
Anderson, John R., Robert Farrelt,and Ron Sauers1984 "Learning to program in USP."
Cognitive Science, 8: 87-129.
Arrow, Kenneth J.1962 "Economic welfare and the al-
location of resources for in-vwition." in R. R. Nelson {ed.).The Rate and Direction of In-ventsve Acti\flty: K)9-625.Princeton, NJ: Princeton Uni-versity Press.
Bower, Gordon H., and Ernest R.Hiigard1^1 Theories of Leaming. Engle-
wood Cliffs, NJ: Prentice-Hall.
Bradshaw, Gary F., Patrick W.Langley, and Herbert A. Simon1363 "SttKfymg sdentffic discovery
by computer simulation."Science, 222:971-975.
Brock, Gerald W.1975 The U.S. Computer (rtdustry.
Camt»1dge, MA: ^(linger. •
Bums, T<»ii, and George M. StalkerThe Man^ement of Innova-tiwi. LorwitMi: Javistodk.
Clark, Kim B., and TakahiroFujimoto1^7 "Overlapping problem solving
in product development."Technical R^xart, HarvardBusiness School.
Cohen, Wesley M., and Richard C.Levin1989 "Empirical studies of innova-
tion and market structure." InR. C. Schmalensee and R,Willig (eds). Handbook of in-dustrial Organlza^on:1059-1107. Amsterdam: El-sevier.
Cohen, Wesley M., and Daniel A.Levinthal1^7 "Participation in cooperative
research ventures and the costof learning." Technical Report,Dept. of Soda! and DecisicmSciences, Came^e MellonUniversity.
1989a "inrKivation and leaming: Thetwo faces of B&D." Eco-nomic Joumal, 99: 569-596.
tevin, Ridiard C, Alvin K.Klevorick. Richard R. Nelson, andSidney G. Winter1983 "Questionnaire on industrial
researdi and development."Dept. of Eojnomics, Yale Uni-versity.
1987 "Appropriating the returnsfrom industrial R&D."Brooicir>gs Papers on EconomicActivity, 783-820.
Levitt, Barbara, and James G.March1988 "Organizational leaming." An-
nual Review of Sociology, 14:319-340.
Lindsay, Peter H., and Donald A.Norman1977 Human Information Pro-
cessing. Orlando, FL: Aca-demic Press.
Link, Albert N.1987 "Cooperative research activity
in U.S. Manufacturing." Tech-nical Report, University ofNorth Carolina, Greensboro.Final report submitted to theNational Science Foundationunder grant PRA 85-212664.
ments and TechnologyTransfers: An Empirical Studyof Dynamic Competitive Ad-vantage, The Case of Elec-tronic Calculators. Lanham,MD: University Press ofAmerica.
Mansfield, Edwin1968 Economics of Technological
Char>ge. New York: Norton.1988 'The speed and cost of indus-
trial innovation in Japan andthe United States: Extemal vs.Intemal technology." Manage-ment Science, 34(10):1157-1168.
March, James G., and Herbert A.Simon1958 Organizations. New York:
Wiley.
Mowery, David C.1983 "The relationship between in-
trafirm and contractual fomisof industrial research in Amer-i c a manufacturirtg,1900-1940." Explorations inEconomic History, 20:351-374.
Mowery, David C, and NathanRosenberg1989 Tachnt^ogy and the Pursuit of
Economic Growth. New York:University Press.
Mueller, Witlard F.1962 "The origins of the basic in-
ventions underlying DuPont'smajor product and process in-novations, 1920 to 1950." InR. R. Nelson (ed.). The Rateand Directicm of Inventive Ac-tivity: 323-358. Princeton:Princeton University Press.
Myers, Sumner, and Donald C.Marquis1969 "Successful industrial innova-
1978 "Technical communication inR&D laboratories: The Impactof project work character-istics." Administrative ScienceQuarteriy. 21: 624-644.
Tushman, Michael L., and PhilipAnderson1986 "Technological discontinuities
and organizational environ-"ments." AdministrativeScience Ouarterly, 31:439-465.
Utteiiiack, James M.1971 "The process of technological
innovation within the fi im."Academy of MartagementJoumal, 12: 75-88.
von Hippei, Eric1978 "Successful industrial products
from customer ideas." Journalof Marketing, 42: 39-49.
1988 The Sources of Innovation.New York: Oxfort UniversityPress.
Vyssotsky, V. A.1977 "The innovation process at Bell
Labs." Technical Report, BellLalx}ratories.
151/ASa March 1990
Westney, D. Eleanor, and Kiyonori Williamson, Oliver E. Zenger, Todd R., and Barbara S.^ 1975 Markets arKl Hierarchies: Lawrence
'The rde of Japan-based R&D ^jwlysis and ^ t i t rus t (mplica- 1989 "Organizaticmal demography:in global technology strategy." tions. New York: Free Press, The differential effects of ageIn M. Hurowitch (ed.). Tech- and tenure distributions onnology in the Modem Corpora- technical communication."tion: 217-232, London: Acactemy of ManagementPergamon. Joumal, 32: 353-376.
