HD28 .M414 no. Wo ALFRED P. WORKING PAPER SLOAN SCHOOL OF MANAGEMENT On the Persistence of researchers in Technological development Raghu Gajud Nfw York University Michael A. Rappa Massachusetts Institute of Technology December 1991 Sloan WP # 3395-92 MASSACHUSETTS INSTITUTE OF TECHNOLOGY 50 MEMORIAL DRIVE CAMBRIDGE, MASSACHUSETTS 02139 Massachusetts Institute of Technology On the Persistence of researchers in Technological development Raghu Ganid New Michael A. Rappa York University Massachusetts Institute of Technology December 1991 © I99I Sloan WP # 3395-92 MASSACHUSETTS INSTITUTE OF TECHNOLOGY Alfred P. Sloan School of Management Massachusetts Institute of Technology 50 Memorial Drive, E52-538 02139-4307 Cambridge, MA T 'm ! 9 1992 On the Persistence of Researchers in Technological development Raghu New GARUD Michael A. RAPPA'*' York University Massachusetts Institute of Technology December 1991 ABSTRACT Researchers often persist in their efforts to create knowledge that shapes the development of a technology. By persisting, researchers can master a knowledge and then contribute to its development. However, persistence can be detrimental to researcher's careers the organizations they to work for. This obsolescence of existing stocks is and stock of undue the well-being of because undue persistence can lead of technological knowledge. Specific hypotheses on persistence are developed and tested using the scientific and technical literature to determine the length of association of individual researchers. introduction New engage fields in of science and technology emerge through the effbns of researchers problem solving activities. These efforts lead to the creation techniques that represent technological knowledge (Allen, who of ideas and 1966; Layton, 1977:198; Rosenberg, 1982:143; Laudan, 1984). Accordingly, the study of technological change becomes one of understanding the creation and diffusion of knowledge among practitioners. A widely accepted model of the growth of knowledge is one that views it as a cumulative progression of ideas and techniques (Crane, 1972:26). Antecedent ideas and techniques influence the choice of future problems (Dosi, 1982). Success in solving these problems depends upon the acquisition of relevant experiences over time. These experiences translate into "rules of thumb" and Winter, 1982) that researchers employ (Sahal, 1981:11 1) or "search heuristics" (Nelson to address their technological problems. In this of authors is arbitrary. Raghu Garud is assistant professor of Stem School of Business (40, West 4th Street, Room 623, New York, NY 10003; tel. 212-998-4123). Michael Rappa is assistant professor of management, Massachusetts Institute of Technology, Sloan School of Management (5t) Memorial Drive, Cambridge, MA 02139; tel. 617-2533627). The authors are indebted to Susan Bjorner and Gary Quick for their assistance in creating the database and Joel Baum, Hayagreeva Rao and Robert Yaffee for their comments and guidance. ' This work management, is fully collaborative; the listing New York University, way, technological progress is a cumulative problem-solving process with only infrequent major disruptions (Rosenberg, 1982: 63). The cumulative growth of technological behavior of researchers who produce it. icnowledge has important implications for the In panicular, researchers core stock of knowledge before they can contribute to that it may take anywhere up to ten years to its become an its need to master a first growth. Herben Simon estimates That expert in an area of work. researchers have to persist in their efforts to develop a contribute to may new technology in is, order to development. Recognizing the benefits of persistence, Ziman (1987) nevertheless cautions against the decreasing marginal returns firom undue persistence. According to Chubin and Connolly (1982), researchers persisting unduly with an area of work in order to leave a research Over time, trail. of behavior leads to incremental yields from a stock of knowledge. this type The danger of undue become obsolete, thereby Indeed, many persistence is that existing researchers' stock of knowledge may undermining the well being of the organizations they work scholars have witnessed the creation of researchers from outside the system. new for. technological knowledge by These "outside" researchers offer ideas that may run new knowledge counter to existing mainstream technological knowledge. If successful, the displaces existing knowledge. What determines researchers' persistence with a technology? While an important question that has clear implications for the question has so far management of R&D not been examined empirically. In this paper, we organizations, this use the literature as a source of data on the length of researchers' association with a technology to examine persistence behavior. Persistence, collective aspects Persistence is also influenced we suggest, depends upon individual, relational of the knowledge creation process that technological change by the occurrence of important events technology that legitimize persistence, we first it. in the Before proceeding to develop and and entails. development of a test hypothesis on provide an overview of the knowledge creation process within a community of researchers. RESEARCH COMMUNITIES AND THE CREATION OF TECHNOLOGICAL KNOWLEDGE In Introducing the trade-offs between persistence and undue persistence, researchers were deliberately portrayed as individuals working independently on technological problems. However, relationships with competencies it more is in collaborative address complex indivisible technological problems (Metcalfe and Soete, 1983) that researchers individually may not able The working one another. Collaboration between researchers possessing different vital to is realistic to visualize researchers as to address. creation of knowledge by researchers engaged in collaborative relationships with others results in a steady accumulation of knowledge that other researchers can build upon. From this perspective, the creation of a new technology MacDonald's (1978) illustrate the collective knowledge creation cnuils. the it produa of three men, They point out is a colleaive endeavor. Braun and nature of technological change and the that it is unrealistic to see the transistor as or of one laboratory, or of physics, or even of the forties. Rather, the development of the transistor required contributions from hundreds of researchers, working many in different places, in Technological knowledge as many different fields, over a collective endeavor years. becomes manifest also acceptance of practices that become institutionalized within a community of inscription devices (Latour and Woolgar, 1979), These and other taken momentum (Hughes, 1987) that researchers join a technological field, there technological knowledge is is attracts other researchers. its creation benefits. As an increase in the rapidity with which created. and fraught with externalities; that its and for granted faas Technological knowledge has another collective property that has appropriate 1987), institutionalized practices legitimize a technology, thereby creating the necessary technological implications for the in researchers 1983). These practices consist of testing standards (Constant, (Jagtenberg, artifacts. many The is, diffusion. In particular, the even those who have important knowledge creation process not contributed to its creation is may presence of externalities leads to the rapid diffusion of technological knowledge. However, the presence of externalities also reduces the incentives for the private rights, produaion of knowledge. Despite mechanisms such knowledge does leak out to others. imponant property As a consequence, Nelson (1959) and Arrow (1962) have argued that the private production of knowledge community perspcaive. Under as intelleaual may be sub-optimal from a these circumstances, researchers in the public sector perform roles in the creation and dissemination of knowledge (Rosenbloom, 1966; Mueller, 1962; Utterback, 1975; Stobaugh, 1985). Besides engaging in research aaivities, researchers in public institutions identify knowledge. and fund private institutions for the creation of Rappa and Debackere (1991) introduce the notion of researcher community to integrate these various facets of the knowledge creation and diffusion process. Building Constant's (1980) work, Rappa and Debackere define the researcher of scientists and engineers who are committed technological problems. These researchers communicate community, knowledge creation is collaborative relationships determined by individual, a colleaive diffusion process within this and community. in some way with each and other. Within the endeavor driven by individuals engaged collective aspects We group organizationally dispersed in public and with others. Consequently, persistence relational, as a to solve a set of inter- related scientific may be private seaor organizations but community upon is in in some way of the knowledge creation and explore these aspects in greater detail in the next seaion. DETERMINANTS OF PERSISTENCE Analogous to Kuhn's (1970) notion of anomalies in science. Constant (1987) introduces the notion of presumptive anomalies that occur in the development of technologies. Presumptive anomalies occur when assumptions derived from science indicate that a or that a radically different technology will be conventional technology will fail, effective (Constant, 1980:15). The identification of presumptive anomalies may more result in the displacement of established technologies and the knowledge systems they entail. Existing researchers resist these discontinuous changes because of career their own new technology, researchers also commitments to technologies. Besides this instrumental reason for resisting the introduction of a resist changes for cognitive reasons. The new stock of knowledge represents a new technological paradigm (Dosi, 1982), or regime (Nelson and Winter, 1977), with its associated traditions of testability (Constant, 1987). Dosi (1982:153) points out that these paradigms have a powerftjl "exclusionary effea" rendering researchers blind to alternative technological possibilities. Confronting resistance from researchers embedded paradigm, the new technology can only grow with researchers to pursue the draw upon the skills an existing technological the efforts of a few dedicated development of the new technology. These pioneering researchers and knowledge of researchers from other peripherally connected with the development of the connected researchers in may make without committing themselves to new disciplines who may be technology. These peripherally transient contributions to the emerging technology its long term development. Transient contribution behavior from these peripherally connected researchers its own value during early periods of technology development. However, sustained contributions from a growing commit themselves number of researchers new technology technology. As long as the to to developing it. Constant (1987) offers the new technological field commitment. new as one approach by which technology. Traditions of testability consist technological practice. Akin to the formation of a the new of testing techniques and normative values that sustains and defines the traditions of testability help researchers new lacks legitimacy, researchers will be unwilling notion of traditions of testability legitimacy can be established for the the set are necessary for the success of the Events that legitimize the are therefore critical to induce researchers' of a may have new vocabulary and a specific grammar, these communicate with one another and to legitimize technology. Technologies also can gain legitimacy from the actions and statements of powerful institutional organizations. from include: seed capital Examples of events that serve federal to legitimize technologies funding agencies, endorsements from regulatory bodies, and the formation of professional organizations. In combination with the formation of traditions of testability, these events that lend legitimacy to the increasing the H 1 : commitment of researchers Researchers are more to the new technology result in development of the technology. likely to persist in their efforts to develop a technology in a legitimized field. Besides these events that legitimize the new technological field, the presence of a critical mass of researchers may be required to establish technological momentum. Size confers power and community's prestige to the emerging community. Power and prestige ability to generate resources researchers in their problem solving once the community size increases result in the from the environment that sustains individual efforts. Thus, researchers are more beyond a threshold limit as a likely to persist bandwagon effea occurs (Rappaand Debackere, 1989). H2: Researchers are more likely to persist in their development of a critical mass of researchers. Over time, praaices followed by community and begin efforts to these researchers develop a technology after the become institutionalized within this dictating the direction of future research. Representing collective specialized stocks of knowledge, these practices become accessible only to those who undergo professionalized education and training. As investments required to enter the community increases, it becomes more who contribution behavior. As a result, researchers community anempt will for researchers to difficult join a professionalized to leverage their entry investments with H3: Researchers' persistence likely to increase is exhibit transient research ongoing contributions. with the accumulation of knowledge within a technological field. Individual researchers emerging paradigms may pursue (Dosi, different technological trajectories within these Trajectories 1982). represent directions specific of technological change based on researchers' theoretical assumptions and the physical they employ. Early during the development of a technology, researchers hold artifacts about "what different beliefs is feasible or at least worth attempting" (Nelson and Winter, 1982:258-259). Because of the uncertainty and ambiguity associated with technological choices, researchers "place their bets" Over build up on different trajectories. time, researchers develop technology specific competencies. These competencies in a path dependent manner (Cohen and Levinthal, 1990; Arthur, 1988; David, 1987) as earlier technological choices direa future options and solutions. As competencies become specialized, researchers find it increasingly difficult to re-deploy them to pursue other trajectories, or to pursue other technological paradigms. As a consequence, there are powerful incentives for researchers to persist with a trajectory. Thus, an important determinant of persistence technology itself Gieryn's (1978) work supports is the longevity of association with the this hypothesis. Studying the published work of American astronomers over the period 1950-1975, Gieryn found who had published a paper on a particular specialized subjea were likely to continue publishing in the same area for length of that researchers many years. Similarly, Mullins (1972) found that the mean membership of a Phage group of researchers was about three six years if those who years, but doubled to published only one paper on the subject were excluded. H4: The likelihood ofpersistence in the development of a technology increases with duration of researcher's association with a technology. So far, we have examined individual and collective dimensions that influence persistence of researchers. Besides these dimensions, persistence relational aspects that capture the extent to is also determined by which researchers associated with the development of a technology collaborate with one another. Collaboration represents the joint creation of ideas by individuals. In a recent study of a bio-medical research community, Small and Greenlee (1986) found that collaboration between researchers increased steadily over the ten year period that they studied. Beaver and Rosen (1979) found panern of increased collaboration from a similar concluded growth the that of collaboration to They their study of scientists. be natural a consequence of professionalization. ProfessionaJization implies the division of labor and specialization The specialized division labor of requires collaborative efforts where among researchers. different skills can be brought to bear to address complex technological problems (Ziman, 1987; Jagtenberg, 1983). Collaboration thus increases the reach of individual researchers, thereby allowing them to penist with their contributions. H5: Researchers are more likely to persist in their efforts to develop a techno loQi if they collaborate with others. RESEARCH SITE We chose the field of cochlear implants as an illustrative case for the present analysis. Although studies of the elearical stimulation of the ear have a long the 1950s that researchers began the systematic investigation of might provide of sound. a One means result history, how it was not electrical stimulation for enabling people with sensorineural deafness to gain of this research until some sense was the advent of the cochlear implant, a device that uses elearical stimulation of the cochlea to provide a sense of sound to profoundly deaf people. One of the first trials 1961 by William House, a of a rudimentary cochlear implant devices was performed clinical physician, who major center for cochlear implant research. At later first, founded the House Ear Institute, a few researchers took an interest cochlear implants, largely due to the intense controversy the field encountered. considered the device to lack a scientific basis and believed human experimentation. It was not until the early cochlear implants emerged as a viable research that galvanized the burgeoning field. it in in Many was too premature for 1970s that the controversy subsided and field. Several meetings were held in 1973 These meeting included a workshop held as part of the American Otological Society meeting (a group that was adamantly opposed to the topic in previous years) One and the study of importance first is international conference Bilger's work on the initiated in subject. 1975 on the comparative performance of cochlear implants. Funded by the U.S. National Institutes of Health and released in 1977, the "Bilger report" (1977) showed that although the early claims regarding performance were exaggerated, cochlear implants did have merit. conclusion, the report lent hundred individuals active organization world wide. S. sorely needed legitimacy number of researchers the door for a greater located in the U. some in By 1991, its and thereby opened to the field to participate. In there were about four the field employed in nearly one hundred different About community forty percent of the cochlear implant Although tremendous progress has been made over the years, is and about three thousand patients have received the cochlear devices, the cochlear implant remains largely an experimental procedure with many challenging problems yet to be overcome. DATA COLLECTION AND METHODS Gathering longitudinal data on researcher persistence First, it requires some degree of historical investigation. If presently works in a particular field the experience of many how individuals Second, the number of researchers who have who have who ask everyone it who we would overlook over the years. participated in a field can be quite large and very hard to ascertain the extent it become appealing participates in a field conference papers and patents in a given articles, we simply come and gone already of their participation. Confronted with these constraints, the literature as a source of data about not a simple task. is long they have participated, can be widely dispersed geographically thereby making Journal in a field and for how to look to long. field represent a detailed, self reported archival record of the effon generated by researchers to solve the scientific and technological problems confronting them. Furthermore, the literature is an appealing source of data in several respects: the publication conventions ensure a level of quality and authenticity; the data can be collected unobtrusively; the findings can be replicated tested for reliability; When and the data taken together, the literature can be viewed researchers to establish a involved, with, are publicly available how new field, as and not very expensive and can provide information about the researchers long they were involved, where they were employed, difficult to to collea. a unique chronology of the efforts of what problems they pursued, and when they were would be and who they collaborated active in the field. Clearly, it match the comprehensive scope and longitudinal nature of the literature using other data collection techniques. Data collection Five commercial electronic databases (covering the medical . and engineering literature) were used to identify publications related to the sciences field of cochlear implants. be The commonly used be either in the lexicon of cochlear research bibliography The earlier to implant researchers. These key terms might or classification terms of a document. in the title, abstract was derived from an known databases were searched using a set of key terms that are The search strategy one used by the National Library of Medicine to create a on cochlear implants. data colleaion procedure resulted in the identification of 1329 unique documents relating to cochlear implants published used to identify each researcher This procedure yielded a who total between 1973 and 1989. The data base was then contributed to the field over the seventeen year period. of 1,257 researchers. A statistical database was created containing several individual-, relational-, and population-level variables for each researcher that was derived ft^om information obtained from published documents. Dependent variable publications in a field . The number of years spanning — that is the "contribution span" measure of researchers persistence The primary — some methodological issue serves as a of concern since the individual have not yet left the field, is some minimum value for this, survival analysis statistics (that and known last unique and usefiil is it who presendy that for those researchers is is only is issues that arise that require are active in cochlear implants, the overall length of contribution contribution first Although calculating the contribution span in a field. relatively straightforward, there are fiinher explanation. a researcher's known is indeterminate: that is, that the length of their the entry year to the present year). To account were used to analyze the data. Such techniques take into consideration precisely this kind of problem in the calculations with a procedure that adjusts for the biases that right censored data create. Determining reason for this is if a researcher is that researchers still may active can be difficult in certain conditions. The not publish every year. Therefore, a researchers' contribution span in the field can be charaaerized by gaps of several years in duration in which there are no publications to their credit. publication records raise the issue of how The frequendy a researcher must publish be considered an aaive contributor to the field. proper censoring scheme to use in the analysis. this question is is it The is order to question important in determining the arises: How long reasonable to assume they are no longer in the field? necessary in order to determine in Thus, understanding the nature and the prevalence of gaps in a researchers' contribution span ceases to publish existence of discontinuities in who has exited the field and after someone The answer who persists. to 10 The cochlear implant data shows that 242 (19%) researchers have a gap in their contribution spans. Among those who have a gap in their contribution spans, four out of five on researchers have a gap of three years or less in duration. Based that researchers who had this evidence, published within the past three years of the last we decided year of the data set (1989) would be censored. Only number (3.7%) of a small the researchers have a gap between publications of all longer than three years in duration. These sparse contribution spans individuals who do having left the field begin a new if Explanatory We treated these researchers as they again started contributing. variables among field. be indicative of they did had contributed during a three year period, and as having when cycle heterogeneity not contribute continuously to the may . Several These include the kind of organization researchers. researcher was employed. were created that might account for variables These researchers were coded according to in which each whether they reside in an academic, industrial, government or independent research laboratories. Besides these variables, other explanatory variables were created to test the four hypotheses developed on contribution spans. To test for the hypothesis on legitimacy, a variable was created to distinguish between a pre-legitimacy period (considered to be prior to the publication of the Bilger repon in 1978) and a post-legitimacy period after 1978. To test for the hypothesis on was created. Population published in the to capture field in a size critical mass, a variable that measured the population size measured is given year. A in number of terms of the researchers second order population variable was any quadratic associations between population size who also created and contribution spans. Besides these two variables, a third variable at the population level was created to control for possible influence of dispersion of researchers persistence. This dispersion variable among measured the extent different organizations to which the community on is concentrated in a few organization or spread across many. For this purpose, a Hirfindahl determined by calculating the sum of the squared share of researchers statistic, which aiFiliated with each organization A variable is is used. was created that represented the accumulated knowledge determined by the cumulative number of publications to of knowledge in the field. To test test for in the field as the hypothesis on growth the hypothesis at the individual level, a variable that was 11 the time of researchers' association with the technology was created. Chubin and Connolly (1982) suggest a potential alternative explanation for individual persistence. They point out that persistence can be a result of positive feedback relationship between and researchers from the environment. their ability to gather additional resources we developed control for this alternative explanation, successftjl To measure of the cumulative a productivity of individual researchers. Collaboration was measured by the associated with as a co-author on number of unique To publications. individuals that a researcher had control for other potential group level explanations, two group level variables were created to reflea the size and produaivity of The a researchers' institutional affiliation. of the nimiber of individuals affiliated Cumulative group produaivity size with that organization measured is of a researcher's group as who the cumulative is measured terms in also publish in the field. number of cochlear implant publications by individuals affiliated with a researchers' group. MODELING STRATEGY AND RESULTS Model Event : history methods were used implies a lower probability of exiting a field conditional Persistence association. The instantaneous exit conditional probability of exiting the field rate, also known as the r.(t)= exit rate at time t X(t) is lim rate, upon earlier the exit rate. is The is: P^(t,t^At/t) At was modeled by estimating the following specification: ri(t) Where hazard ^->0 ' The to study persistence behavior of researchers. = exp [aiX(t)] a vector of the values of the covariates at time t, and ai is a vector of parameters. In this model, hazard rates are postulated to be log-linear functions of the In order to incorporate time variations in the explanatory variables, a variables X. multiple-spells formulation of this each period exiting. is broken down model was used. In the multiple into one-year observations in Each of these annual spells is spells formulation, which the researcher is at risk of treated as being right censored unless the researcher 12 exits. RATE (Tuma, 1980) was method of maximum Results . used to estimate the vector of parameter estimates by the likelihood. The model was estimated in a sequence of steps by adding sets of explanatory variables into the equation (see Table among included to control for variations The second model affiliation. First, variables 1). pertaining to researcher type were researchers' persistence because of organizational included the length of researchers' association along with the variable on cumulative productivity. This model was estimated persistence may phenomenon age dependent be an to identify controlling after for whether possible heterogeneity of the sample, and controlling for researcher success and reputation as measured by the cumulative produaivity tested hypotheses were variables The also 1 As explained introduced at rises covariates from 18 on the Model exit rate. Model 6 even though The in For results in Models 3 to 6 added variables that the methods section, appropriate control each step. results indicate that the The X^ among to 4. variable. fit 1 this of the models improved to 614 paper Model in we 6. as covariates Model 6 focus our attention were included: displays the effeas of on the all results displayed in of other models are provided for a comparison. results indicate that the hypothesis relating to collaboration the five that did not receive support. One reason for this is was the only one that collaboration as measured by the number of cumulative co-authors does not adequately capture the entire spectrum along which researchers collaborate. Nevertheless, the lack of significance suggests that relational aspects are less of a predictor of persistence. This conclusion bolstered by the lack of significance of other group level variables, including group is size and group productivity. The results do support other hypotheses. For instance, the positive sign of the legitimacy variable suggests that the likelihood of persistence increased after 1978 with the publication of the Bilger Report that served to legitimize the cochlear implant field. Similarly, there is support for the critical mass hypothesis. The positive coefficient of the population variable combined with the negative coefficient of the second order term implies an inverted U-shaped relationship between population size Critical and the hazard mass appears to have been accomplisheed with a population researches (which occurred in 1983) at which point the hazard size rate decreases. rate. of around 213 13 In support for the hypothesis relating to knowledge accumulation and persistence, the variable that reflects the cumulative publications in the field was with a negative knowledge The That sign. to be significant the risk of leaving the field decreases as the accumulated increases. significant negative sign of the length of association variable indicates that researchers may have a tendency to productivity of researchers. time, is, found we conduaed To persist with time despite controlling for the cumulative further explore individual determinants of persistence over a non-parametric analysis of the hazard flinaion based of researchers' association. For procedure LIFETEST this (see Figure 1). purpose, we used on the duration the lifetable approach in SAS (v5.18) 14 CONCLUSION The objeaive of paper was to empirically explore persistence behavior of this development of a technology. Non-parametric estimates of researchers associated with the the analysis of persistence through a hazard analysis shows that the risk of exiting the field is About 40-percent of the greatest in the first year of their contribution. researchers one year associated with cochlear implants demonstrate transient persistence for only a The duration. hazard rate declines sharply after the which until the sixth year after it and by the thereafter for more fertile territory. program might be a research data collection and analysis. by the prospea of making an at either, the researcher But if established; By to a field might It is once in place it would require the end of the sixth year, a among exit the field shortly the researcher can initially "survive" in the critical which time the success of the program leads the researcher to remain probability of ever leaving. of the research possibility of attracting the resources to underwrite the of his or her work. Barring success cost in a behavioral characteristic For example, a researcher might be drawn imjX)rtant contribution year and remains fairly constant continues to decline rapidly. This result might have an explanation process. first field, several years point is of reached, at in the field with linle these latter category of researcher that we would look for undue persistence. Modeling the instantaneous exit rate with parametric statistics provided other insights about persistence. Early during the development of a technology, persistence associated with the legitimacy of the technology During this stage, the creation and the of technological size momentum is strongly of the research community. has very much to do with whether or not individuals are ready to commit themselves to the emerging technology to sustain its growth. Over time, a reversal occurs wherein accumulated knowledge results in creating strong incentives for researchers to persist. What is collaboration perhaps most interesting is that relational aspects including direct and other forms of collaboration including group membership and group produaivity do not appear to be key determinants of persistence. Persistence thus appears to be with determined by individual its an emerging professionalized field concomitant norms for the creation and diffusion of technological knowledge opposed to intermediary This efforts to contribute in is relational aspects as of the knowledge creation process. a preliminary analysis that sets the base for the exploration of more fine- grained processes connected with researcher persistence. For instance, data from the 15 cochlear implant literature have been gathered with respea to the nature of the conduaed by each clinical research), researcher (such as basic research, applied research, development, or and the panicular technological trajeaory each researcher has pursued (such as single-channel versus multi-channel devices). funding levels conduaing work being Other kinds of data such as annual and the extent of the market commercialization have been collected a subsequent analysis. for REFERENCES Allen, T.J. 1966. "Studies of the problem-solving process in engineering design," Transactions on Engineering IEEE Managemmt, EM- 18: 72-83. Arrow, K.J. 1962. "Economic welfare and the allocation of resources for innovative aaivity," in R.R, Nelson (ed.) The Rate and Direction of Inventive Activity. Princeton: Princeton University Press. Arthur, B. 1988. "Self-reinforcing mechanisms in economics," in P Anderson Economy as an Evolving Complex Beaver D. and Rosen R. 1978. "Studies in scientific collaboration: Pans Scientometrics, 1: 65-245. Bilger, R.C., et al. et al.. The System. Reading, Mass.: Addison-Wesley. 1, 2 &: 3" 1977. "Evaluation of subjects presently fined with implanted auditory prosthesis," Annals of Otology, Rhynology, and Laryngology, 86, 3, Pan 2, Suppl. 38. Braun, E. and MacDonald, S. 1978. Revolution in Miniature: The History and Impact of Semiconductor Electronics. Cambridge: Cambridge University Press. Chubin, D.E. and Connolly, T. 1982. "Research trails and science policies: Local and extra-local negotiations of scientific work" in Elias, N., et al. (eds.) Scientific Establishments and Hierarchies. Dordrecht: Reidel. Cohen, W. M. and Levinthal, D. A. 1990. "Absorptive capacity: A new perspective on learning and innovation," Administrative Sciences Quarterly, 35, 1: 128-52. Constant, E.W. 1980. The Origins of the Turbojet Baltimore: John Hopkins University Press. Constant. E.W. 1987. "The social locus of technological praaice: organization?," in Systems. W.E. Bijker, et Cambridge, Mass.: MIT al. (eds.) Community, system, or The Social Construction of Technological Press. Crane, D. 1972. Invisible Colleges: Diffusion of Knowledge in Scientific Communities. Chicago: University of Chicago Press. David, P. 1985. "Clio and the economics of QWERTY" Economic History, 75: 227-332. Dosi, G. 1982. "Technological paradigms and technological trajeaories," Research Policy 11: 147-162. Gieryn, T.F. 1978. "Problem retention and problem change in science," Sociological Inquiry, AS: 96-115. Hull, D.L. 1988. Science as a Process: An Evolutionary Account of the Social Development of Science. Chicago: University of Chicago Press. and Conceptual 17 Hughes, T.P. 1987. "The evolution of large technological systems," in W.E. (eds.) The Social Construction of Technological Systems. Cambridge, Mass.: Bijker, et MIT al. Press. Jagtenbcrg, T. (1983), The Social Construction of Science: A Comparative Study of Goal Direction, Research Evolution and Legitimation. Boston: Reidel Publishing.Co. Kuhn, T.S. 1970. The Structure of Scientific Revolutions. Chicago: University of Chicago Press (2d ed.). Latour, B. and S. Woolgar. 1979. Laboratory Life: The Social Construction of Scientific Facts. Beverley Hills: Sage Publications. Laudan, R. 1984. The Nature of Technological Knowledge. Dordrecht: Reidel. Layton, E. 1977. "Technology as knowledge". Technology Mansfield, E. 1985 Economia, 34, 2: "How and Culture, 15: 31-41 rapidly does a technology leak out?" Journal of Industrial 217-223. J. S. and Soete, L. 1983. "Notes on the evolution of technology and international competition," a paper presented at the workshop on Science and Technology Policy, University of Manchester. Metcalfe, Mullins, N.C. 1972. "The development of a scientific speciality: The phage group and the origins of molecular biology," Minerva, 10: 51-82. Mueller, W.F. 1962. "The origins of the basic inventions underlying product and process innovations, 1920-1950," in R. Nelson Du (ed.) Pont's major The Rate and Direction of Inventive Activity. Princeton: Princeton University Press. Nelson, R. 1959. "The simple economics of basic scientific research". Journal of Political Economy, 6: 297-306. Nelson, R. and S.G. Winter. 1977. "In search of a useful theory of innovation," Research Policy. 6: 36-76. Nelson, R. and S.G. Winter. 1982. An Cambridge, Mass.: Harvard University Evolutionary Theory of Economic Change. Press. R&D communities: Implications for technology Rappa, M.A. 1989. "The dynamics of strategy," working paper, MIT Sloan School of Management. and Debackere, K. 1991. "R&D communities and technological development," working paper, MIT Sloan School of Management. Rappa, M.A. Rosenberg, N. 1982. Inside the Black Box: Cambridge University Technology and Economics. Cambridge: Press. Rosenbloom, R.S. 1966. "Product innovation in a scientific age," Successjul Marketing. Chicago: Proceedings of the Marketing Association, Chapter, 23. in New Ideas for 1966 World Congress, American Sahal, D. 1981 Patterns of Technological Innovation. Reading, Ma5S.: Addison-Wesley Publishing Small, 2): Company. H. and Greenlee, E. 1986. "Collagen research in the 1970s," Scientometrics, 10 (1- 95-1 17. Stobaugh, R. 1985. "Creating a monopoly: Product innovation in petrochemicals," in R. Rosenbloom New Tuma, N.B. Tuma (ed), Research on Technological Innovation, Management and Policy, Vol 2. York: JAI Press (81-112). 1980. Invoking RATE, (2nd ed.). Menio Park, Calif.: SRI International. N.B. and Hannan, M.T. 1984. Social Dynamics: Models and Methods. Orlando: Academic Press. Utterback, J.M. 1974. "Innovation in industry and the diffusion of technology," Science, 183: 620-26. Ziman, J. 1987. Knowing Everything About Nothing: Specialization Careers. Cambridge: Cambridge University Press. and Change in Scientific 19 TABLE ML Estimation Modell Intercept -1.5 1 of Contribution Spans Mgdcl2 -.7 (0.06) Models -.7 (0.08) (0.08) Mociel4 -2.5 (0.45) Models -1.77 (0.47) Model6 -1.8 (0.47) Typt ef OrsjtnfsatJQn Industrial Research Government .001 .06 .1 .1 .07 .04 (0.26) (0.26) (0.26) (0.26) (0.26) (0.26) .13 .012 .03 -.01 -.01 .007 (0.1) (0.1) (0.1) (0.1) (0.1) 1.5 (0.38) 1.11 1.16* .71 .58 (0.58) (0.58) (0.58) (0.58) (0.1) .47 (0.58) Geographic location USA -.03'" -.22 (0.09) (0.085) Length of association Cumulative researcher productivity Legitimacy of -.22 (0.08) -.16 -.15 (0.09) (0.09) (0.086) -.27"' -.27'" (0.05) (0.04) (0.05) -.1" -.1" -.11'" -.09" (0.033) (0.03) (0.03) (0.04) .48"" 1.33"' 2.17'" 2.15'" (0.14) (0.03) (0.36) (0.003) .03"- .03'" .03"' (0.00) (0.00) (0.009) -.09"' -.07'" (0.01) (0.01) (0.004) .001 -.or -.009" (0.002) (0.002) (0.001) -.28 -.28'" (0.046) (0.046) -.10" (0.033) field -.02 Population Popularion^/lOOO Dispersion of researchers -.29 -.0024" Cumulative knowledge (0.000) Group -.08'" -.003'" (0.003) -.01 size (0.012) -.0007 Cumulative group productivity (0.04) -.017 Cumulative co-authors (0.02) P-leve! Chi square NOTE: number of spells = 2976; .000 .000 .000 .000 4 6 7 10 11 14 18.0 263.7 273.8 540.7 610.0 614.1 number of exits = 604. 1.) Overall 2.) Figures in parenthcics arc standard errors of estimates 3.) Significance level: ' < .05; " .000 .000 df < .01; "' < .001. J b5b 2u H'UBRARIEs 3 DUPL ^oao oo7sbflis Date Due f-?-^ Lib-26-67 MIT LIBRARIES 3 TDfiD DD75t.fil.5 ^ i I tf*iMBtMMMM _^ tmttmk I