EvolutionaryTheory in Management and Organization Theory at the Beginning of a NewMillennium: A Symposium on the State of the Art and Opportunities for FutureResearch in Evolutionary Theory.





This is a slightly editedtranscript of the symposium that took place on Monday, August 7, 2000, at theAcademy of Management Meeting in Toronto, Canada. The symposium was organizedand facilitated by JohannPeter Murmann. The speakers were HowardAldrich, DanLevinthal, and SidWinter, who each gave a 15-minute presentation and then answeredquestions from the audience. Because the audio recording equipment was not ableto pick up the questions from the audience, Johann Peter Murmann paraphrasedeach question before the panel responded.



(Aversion of this symposium was published in the Journal of Management Inquiry. Click here for thepaper.)



Johann Peter Murmann: Let me tell you a little bit about the genesis of thissymposium. The highlight of my teaching career so far happened this spring whenI had the opportunity to teach an elective Ph.D. seminar on evolutionarytheories of organization. A year ago I started my preparation for theseminar. I began to read widely in the history of evolutionary thought and inevolutionary thinking in management and organizational theory. As Istarted to put together my syllabus,I thought it would be clever to go on the web and see if someone has taught asimilar seminar which I could copy to some extent. Unfortunately, I discoveredthat nobody had done this kind of seminar I had in mind, and so I decided todevelop my own web site on evolutionary theories in the social sciences for theseminar.

Inpreparation for the seminar, I came across a wonderful writer who is not aswidely known in the Academy of Management as he should be. I want to take theopportunity to introduce him a little bit more to this audience. I am talkingabout GeoffreyHodgson, who writes a lot about the history of evolutionary thoughtin economics and in the social sciences. He recently came out with the book “Evolutionand Institutions: On Evolutionary Economics and the Evolution of Economics.” Onechapter in this book discusses in detail the history of evolutionary thought ineconomics from the writings of Thornstein Veblen in the 1890s to current times.I strongly encourage you, if you’re interested in evolutionary thought in thesocial sciences, to pay attention to GeoffreyHodgson's writings.

 As we were approaching the academy, I thought,why not capture some economies of scale. I was already developing this Ph.D.seminar, and my own survey of evolutionary thought led me to believe that manypeople in the academy would find it interesting to hear about the current stateof evolutionary thought and the challenges for evolutionary thinking in thesocial sciences and management and organization theory in particular. So Idecided to try to organize a symposium for this year’s academy. I have beenvery fortunate that I was able to recruit three leading scholars on evolutionaryideas in management, economics, and sociology. I asked Howard Aldrich, DanLevinthal, and Sid Winter to reflect not only on the achievements and thecurrent state of evolutionary thinking in management organization theory, butalso to lay out for us what they think are the key challenges that evolutionarytheory currently faces. I hope that their presentations will help us get a goodsense of what currently are the key questions that need to be addressed to make progressin applying evolutionary ideas to management.

Ihave planned the session today in this way. We will first hear from each of thethree panelists. They will speak roughly for about fifteen minutes. We want tohave ample time for discussion and debate. Later on in the session, I willinvite you to not only ask questions but also share with us your views onevolutionary thought in organization and management theory and what you thinkthe key challenges are. We hope that this second part of the session will trulybe interactive.

Ishould also note that the theme of this conference is time, and I know that wewill run out of time by 6:00 pm, and there will be still questions thathaven’t been addressed. The web site, which we created, EvolutionaryTheories in the Social Sciences (www.etss.net) has the beautifulfeature that it can overcome the limits of time. The web site has discussionforum where you can go after we run out of time at 6:00 pm to posequestions or share your ideas. So without further ado, I would like to introduceHoward Aldrich as our first speaker.


Howard Aldrich: Thank you, Peter. This is a nice opportunity for me, mainlybecause this is the fourth and last session I have today, so I’m feelingexcellent, actually, regardless of what happens in the symposium today. What Ithough I could do, in the spirit of what Peter charged us with, is lay out threeor four provocative points. For the benefit, in particular, of people out therewho are still wondering what the heck this is all about, I thought I would focuson fundamental issues. I think Dan and Sid, who’ve been doing the moretechnical side of this, will take care of the technical stuff. I want to talkabout why it’s important to look at the world through a selection lens. Andthat’s what’s important to understand. Evolutionary theory is aboutselection logic. Selection logic drives what evolutionary theorists do. Why isthat important? Well, the example that I’ve been enamored of recently is oneof multi-level selection: What does it mean to take seriously not just the ideaof multi-level analysis, but also the idea of multi-level selection?What does that mean in particular? Well, it means being concerned with the factthat some behaviors or activities on the parts of units--they can be people,groups, teams, organizations--can decrease their individual fitness within theirunit, but, in fact, increase the fitness of the unit itself. So people andorganizations, organizations and populations, can actually engage in behaviorthat lowers their fitness, that is, decreases their longevity, theirprofitability, their effectiveness as units. But, in the same instance, in thesame process, their action raises the level of fitness of that organization,team, or population. And so the problem for the evolutionary theorists then isto try to figure out what the weight of the opposing forces in this case mightbe to determine long run outcomes. Now, to give you some prosaicexamples…although Peter didn’t give the title of my talk, I asked himactually to announce that my talk was called, “Beakless chickens, celibatepriests and selection logic.”


Johann Peter Murmann: I thought that this was a joke, Howard.


Howard Aldrich: It is not a joke. Beakless chickens. Let me give you anexample. Perhaps it’s easier to grasp when you’re not worrying about humansin the equation. So, I want you to envision a lot of chickens. The problem forthe egg production industry was how you get more eggs out of a bunch ofchickens. Well, there is a classic individual level selection procedure thatthey have followed in the past; and remember, you must put this in historicalcontext. Chickens in the wild run around a lot (well, in farmyards). They runaround in flocks and they have a lot of space. They have a lot of room for eachother--there’s not territoriality as a major issue there. Modern chickens, ifyou’ve ever been to an egg production farm, are raised in cages. They’revery close together, they get in each other’s way, and, in fact, what they dois peck each other a lot. And in fact, modern chickens in these places have tobe debeaked in order to prevent them from killing each other. Nonetheless,it’s still the case that quite often the mortality rates run about 50% in aflock. So the egg production people, with the old style understanding ofselection, say “well, why don’t we raise the productivity of our flocks bysimply picking out the chickens that lay the most eggs, breeding them with otherand we’ll over time get a higher level of egg production.” Well, it turnsout that genetically they’re already at the limit of what chickens can do. Infact, their selection process at this level had a very perverse effect. Pickingthe chickens that were the greatest egg producers resulted in the long run inactually decreasing egg production. Why is that? Well, ethnographers went intothe chicken houses with notebooks and tape recorders to see what the chickenswere actually doing. What did they discover? They found that the most productivechickens could be laying more eggs because they succeed in killing the otherchickens and the other chickens can’t get food. They’re just the moreaggressive chickens. They’re the nastier chickens. And so what you get whenyou select at this level are really nasty chickens. One consequence is you getdecreased egg production. The breakthrough came when people started thinking inmulti-level selection terms. Rather than individual chickens, chicken flocks areactually the relevant unit here. They are social groups. They don’t just layas individuals, they lay in flocks. So, what the selectors began to do insteadof choosing individual chickens that were productive was to choose flocks ofchickens that had high output and bred those flocks with other flocks, and overtime, in fact, in a very short span of time, doubled egg production. And a sideconsequence, which is kind of nice for the humanist in me, was that they couldstop cutting off the beaks, because the chickens stopped killing each other. So,here we have a classic example of multi-level selection: working at the level ofindividuals you get a perverse effect at the group level, because you create agroup that’s actually less productive than before. By contrast, selecting forthe entire group produces, in this case, better chickens; better output anyway.

Celibatepriests are another puzzle. Sociobiologists are often puzzled by things likecelibate priests, because it would seem absurd for an organization or amovement—a social movement that wants to spread itself—to have key membersbe celibate. By definition, celibate priests, in theory at least, don’t leaveheirs [audience begins to laugh] …I’m not Catholic. I don’t care. Celibatepriests don’t leave heirs. This seems on its face an absurd situation. Well,it’s again, absurd only if one thinks at the level of the individuals. Thecelibate priests are decreasing their genetic fitness, that is, their gene linesin fact die out if they’re the last sons in their family. But, what arecelibate priests? Well, celibate priests at the group level are incrediblyimportant because of what do they do and what happens when others observe them.They go around doing good deeds. The good deeds, in SusanBlackmore's terms, lead other people to want to emulate them, and soit looks like it’s a good thing to be a Catholic priest because of all thesewonderful deeds that you have time on your hands to do and they are alsoavailable to proselytize. So, social movements and any organizations that areseeking to expand themselves rapidly, again, can use people like this. Peoplewhose behavior decreases their individual fitness but on the level of the groupactually makes them more effective versus the other competing groups. Now theseexamples I’ve given are very dramatic to make a point, in case you weren’taware of this multi-level selection phenomenon. We have other examples that areperhaps not quite so gory or sacrilegious. We have examples from theorganizations field. If you look at promotions’ policies within firms you cansee the same thing can be at work. Selecting on the basis of selecting andpromoting the most aggressive members of the firm can have the perverse effectof actually decreasing the fitness of the firm if that policy leads to internalconflict, an overall nasty atmosphere and people who don’t like to work thereanymore.

Foran example a level up, I turn to an example in my new book, the example ofcollective action in things like trade associations and other collective bodies.One of the things that benefit enormously a new industry is to have some of itsmembers willing to take the lead in organizing the collective action required tocreate a trade association. Again, this action is something that for theindividual firms looks like in the short run a deficit, a cost, something thatif they weren’t engaged in they would allow them to use those resourceselsewhere more productively. But it’s their activities as self-sacrificingentities that make the trade association formation possible, solidify the tradeassociation’s political power perhaps, and then benefit therefore thelegitimacy and the long run viability of the population.

Wecan go a bit further, though, and ask ourselves how do such actions spread inpopulations? It’s quite possible that organizations can learn these behaviorsin one population and spread them to another population. It’s possible alsothat the most powerful organizations engage in such actions, and that’s infact true with what happens to trade associations. In that case, they’reactually, in the long run, probably benefiting themselves a bit from it. Itdoesn’t really matter if they benefit or not, however. The major point I’mmaking here is again, they’re engaging in actions that at the collective levelhave a payoff.

So, back to my first point: multi-level selection.Multi-level selection means that we take seriously the possibility that whatbenefits, or alternatively, what decreases the fitness of a unit within a largerunit actually raises the fitness of the larger unit. And so, explanationscan’t be straightforward. It’s not a matter of adding things up. In thesecases we have to consider what the net balance is between those two levels offorces.

The second issue, is, what’s being selected. Sid [Winter]and Dick[Nelson] many years ago proposed the idea of routines andcompetencies as a unit of selection. BillMcKelvey, in his work (1984), proposed something similar called a“comp.” Think about what that means: if the argument is that what isselected are routines and competencies, then what are organizations?Organizations are just temporary repositories of routines and competencies.They’re just vehicles, or carriers. The distribution of these routines andcompetencies depends upon the selective survival and growth of organizations,but the competencies don’t care about the organizations, except insofar is itadvances their proliferation. This is some resemblance to Dawkin’s ideas, forthose of you familiar with Richard Dawkin’s idea about the selfish gene.Dawkin’s concept of selfish routines is about the same argument. From thispoint of view, coldly, calculatingly, at this level, routines and competenciesdon’t care about the organization. They don’t have to. All that matters forthose routines and competencies is that the organization lives long enough forthem to be propagated. It becomes more complicated when we start seeing thingslike epistasis with the possibility that routines and competencies don’t comein isolation, but come in bundles. What happens if it’s bundles of thesethings that matter, and not the routines or competencies as taken individually.It becomes even more complicated when we recognize that it’s not therepresentation of routines and competencies that matters, however they might berepresented inside the organization (perhaps in the form of policy manuals,norms and values that are carried by individuals as understandings). It’s notthe representation that’s important, but rather their expression. All right,selection is by consequences. It’s the expression of these routines andcompetencies, whether isolated or in bundles, that really makes the difference.It’s the consequences they have for the organization that count.

Theother possibility, then, taken up a level, is that it’s not really routinesand competencies…Let’s take seriously the resource-based view of the firm.This is something that just actually came to me recently reading NicolaiFoss'’ (1993, 1997) works, where he actually argues that aresource-based view of the firm has a lot to say to evolutionary theory.Following the resource-based view, imagine that a firm is in fact a uniquebundle—a unique bundle of routines and competencies that are completelyidiosyncratically linked to the bundle's manifestations in that organization. Inthat case, the routines and competencies rise or fall only if the organizationrises and falls. The question is, in that case, can we talk about routines andcompetencies as units of selection anymore? How can we, if they’re so tightlycoupled to the fate of the organization. If the resource-based view is reallypushed to its extreme, to the limit of the argument the strategists make, what afirm needs to do is work towards uniqueness-toward idiosyncrasy - and towardidentity that separates the firm from everyone else. Then, can we use routinesand competencies anymore as units of selection? It’s a question I leave to youguys [Dan Levinthal and Sid Winter]. Maybe you can answer that.

Onelast even more radical notion: there’s a book (1999) by Susan Blackmore – aradical book, called “TheMeme Machine. Some of you know this book – I see some headsnodding. I don’t know how you feel about this book, but it deserves reading.Let’s take the argument that, in fact, she’s right and that is that; let’stake routines and competencies as the equivalent of memes. Humans, decisions,strategies, all those things become not our focus anymore. Those are all, again,simply ways in which routines and competencies make copies of themselves. It’sa fairly radical way of thinking about selection logic. It means that if wetruly focused on routines, competencies, practices, we would NOT follow peopleanymore in our research. We wouldn’t follow people anymore, we would followhow competencies spread, replicate, and insinuate themselves into organizations.People would sort of disappear from our equations.

SoI leave those three thoughts with you. Multi-level selection logic – think ofbeakless chickens and celibate priests. And think about the possibility thatthings that benefit actors or decrease their fitness can opposite effects at thelevel of the unit they’re embedded in. Think about whether we want to haveroutines and competencies or organizations as our unit of selection. Think aboutwhat that would say for the kind of research that we would carry out if we tookit seriously.


Johann Peter Murmann:I try to be a learning individual, and so I will now announce the title of SidWinter’s talk. It is “The Killer Routine.” No, actually it is “TheProgress of Evolutionary Thinking in Economics and Management”–Sid Winter.


Sid Winter: I understand the charge here to be to look backward andthen to look forward. With a parochialism that I hope is excusable, I will lookin both directions, primarily at things that I can understand in relation to theevolutionary economics framework that Dick Nelson and I put forward in our bookin 1982. I’m going to start by talking about a number of areas where I thinkthere’s quite significant progress that is helpful to or someway supportive ofthe proposal that Nelson and I put forward. Some of the people who made thecontributions that I will note here thought they were making contributions toour endeavor. Others thought something quite different; I don’t care. This isthe selfish evolutionary theory. It gobbles up these contributions regardless oftheir intent.




Letme put one other qualification up front, that is, I’m not implying at all thatwe know enough about any of these subjects. There’s vastly more work to bedone in all of these areas, but I think if you compare where we are now withwhere we were when Dick and I wrote that book, I think there has been a verysubstantial advance.

Firstof all, on the nature of routines, and specifically of operational routines, Ithink that we really have a reasonably good picture of the way those routineswork at the operational level. That picture has developed with one veryimportant contribution at the experimental level by Cohen and Bacdayan in their1994 paper (Cohen and Bacdayan 1994), whichessentially showed that they could reproduce at the level of a dyad, and under asuitably controlled experimental situation, many of the phenomena that we thinkof as being characteristic of organizational routines. We also learned quite alot, I think, from the discussion and the literature on lean manufacturing. Thatliterature gave us quite a detailed view of differences in practices as theyexisted between the Japanese auto industry, for example, and the U.S. autoindustry before lean manufacturing diffused, and then of how those transitionsoccurred. And in particular I’m a big fan of Paul Adler’s 1993 paper on theNummi plant (Adler 1993), which has a lot of really fascinating detail that isinformative about the nature of routines. And finally, and more broadly, I’vehad an interest in quality management (Winter 1994), and I think the qualitymanagement movement has also revealed quite a bit about the nature ofoperational routines.

Iuse the term “capabilities” to refer to higher level and more significantaggregates of routines, which are more a matter of managerial discretion intheir exercise. We’ve also learned a good deal about capabilities and dynamiccapabilities of various sorts. There is a literature on new product development,for example, which is an important area of dynamic capabilities, and a famouspaper by Clark, Chew and Fujimoto (Clark, Chew and Fujimoto 1987)on that subject. There’s a very congenial paper that economic historianSteve Uselman (Usselman 1993) wrote on IBM and the origins of its capabilitiesand dynamic capabilities, which is actually written from the evolutionary theorypoint of view. And there’s a number of other contributions, and I cannot omitto mention that some of these contributions relating to capabilities haverecently appeared in a volume edited by Giovanni Dosi, Richard Nelson, andSidney Winter, entitled, “The Nature and Dynamics of OrganizationalCapabilities” from Oxford University Press.

A third topic is idiosyncratic learning and thesources of variety. One of the things I think that Howard’s puzzle aboutunique bundles of routines actually alludes to is the issue of how we getdiverse solutions to the same basic task or problem in a collection oforganizations. I don’t think, Howard, that the fact that the solutions aredifferent means that we have to abandon routines as a level of analysis, but itwas an important gap in our earlier thinking that we really didn’t have a veryclear image of the nature of the variety that exists. In this case, I think thatactually the theoretical contribution has been most illuminating, and I refer inparticular to my colleague Dan Levinthal and his work on NK modeling ofcomplexity in organizations. That gives a very interesting picture, I think, ofhow complexity and different starting points and local search as mechanisms ofgeneration of organizational routines could give you very much the kind ofpicture we seem to have where we have different, more or less, equi-performingsolutions to a given problem.

One kind of glaring omission in our 1982 book wasthe failure to think about evolution, and industry evolution in particular, in ahistorical context. This is a real head- thumper kind of realization after thefact. The realization is that, if you look at historical situations, a lot ofthe struggle – survival–of-the-fittest kind of thing that goes on in anindustry is something that has a particular historical setting. And as theorganizational ecologists have shown us so clearly, there’s a lot of activityin terms of birth and death processes early on in an industry’s history. So,if you really want to understand the way the market system and the feedback fromthe marketplace provides the selective force that operates to shape what isactually happening, the examples that are really powerful are examples from ahistorical context of relatively early stages in industry evolution, or, ofcourse, later stages where there is some innovation which renews the wholeprocess. That is a very important perspective on evolutionary processes ineconomics, and there is a substantial literature now on industry evolution inwhich Steve Klepper in particular, of Carnegie Mellon, has made very importantcontributions (Klepper 1997). There have been others as well.

There’s the case of the very interestingdiscussion about the sources of variance in firm level profitability, aliterature which goes back to Dick Schmalensee’s contribution in the AER (Schmalensee 1985). It was really introduced to the strategy fieldby Dick Rumelt (Rumelt 1991) and followed up by a number of authors. My latecolleague Ned Bowman, with Connie Helfat, and Michael Porter and Anita McGahan (McGahanand Porter 1997; Bowman and Helfat 2001) and others have looked at this issue.This is friendly information as far as I’m concerned, because the generalthrust of this literature is to the effect that a lot of the variation that wesee in performance as measured by profitability (ROA) resides at deep levels inorganizations. That’s convenient because the kind of theory that Nelson and Iput forward gives you some reasons to think that there might be cross-sectionaldifferences and persistence in the way things happen at the plant level, say, orat the business unit level. But those reasons aren’t so clearly operative atthe corporate level. What the variance analysis turns up is the fact thatthere’s a lot of persistence and a lot of the variance coming from thebusiness unit level, not so much from the corporate or industry level.

Thelast item on my list of areas of progress is an area where the research programin evolutionary economics has intersected in an interesting way with the recentvogue for something called “knowledge management.” A basic idea inevolutionary economics is that firms tend to grow by doing “more of thesame.” In fact, in most of our simple models, firms employ exactly the sameoperational routines through time, in the absence of deliberate innovationefforts. Similarly, firm growth means employing the same operational routines ona larger scale; the implicit assumption is that they expect to increase profitby doing so. In the models, firms have no difficulty in replicating theirroutines either in time or in space.

Theinterest in knowledge management arises from the same motivational perspective,that firms can profit by leveraging their existing knowledge. In contrast to themodels, however, the assumption in knowledge management is that the re-use ofexisting knowledge is quite a problematic process in an organization. Consistentwith the previous observation that the sources of variety are deep insideorganizations, the empirical evidence arising from knowledge management effortsindicates that firm knowledge is far from homogeneous internally and that thereare numerous obstacles to internal transfers of routines. My colleague GabrielSzulanski has done important work illuminating those obstacles [Szulanski1999, 2000], and a lot insight into the role of routines as sources of advantagegets generated as a byproduct of this research. There remains a substantialchallenge of developing evolutionary models that correspond to the evidencebetter than the simple ones do.


Somuch for the review of progress. Let me turn now to issues for the future.Trying to think about research priorities leads me first to the observation thatthe questions that are interesting from the viewpoint of evolutionary economicsstraddle several different fields – and which of them you think are mostinteresting probably depends on the field you are coming from.

Thework Nelson and I did was conceived as a contribution to economics, particularlyrelevant to the economics of industrial organization, technological change, andeconomic growth. All we could do was to try to sketch in the picture – interms of concepts and methods as well as the substance of the phenomena ofeconomic change. Although in absolute terms there have been many contributionsthat fill in that sketch, the total is still very modest relative to the size ofthe agenda. For example, chapter 9 of the book uses simulation to explore amodel that is essentially an evolutionary version of the kind of description ofthe economy that appears in the growth theory models of orthodox economics. Someclosely related analytical tools are sketched in chapter 7. We didn’t do thejob very completely or well, and there was plenty of room for doing more andbetter work on the problem. Very little of that follow up has actually beendone. There are many other examples of parts of the evolutionary program foreconomics that require foundation work, not to speak of opportunities forapplication. But the great bulk of that will probably remain undone, short of aparadigm shift in economics that doesn’t seem to be happening.

Theimpact of evolutionary economics thinking has been greater in organizationtheory and strategic management, and hopefully that implies a greater likelihoodof some real attention to the agenda. Under the former heading, certainly theexploration of multi-level selection processes is a key heading, both forempirical work and modeling. The question of where routines and capabilitiescome from, the learning processes and contextual factors that give rise to them,deserves vastly more attention than it has received. The relevant circumstancesand processes are clearly quite diverse, so the attention has to be spread amongthe different contexts and challenges– established firms vs. new ones,“hard” technological vs. “soft” administrative routines, “new to theworld” vs. new to the market, firm, or unit, etc.

Actually,the latter sort of work is also highly relevant in the context of strategicmanagement, and some important contributions have been made in my department. Mycolleague Harbir Singh has supervised a series of dissertations that I call the“How Companies Learn to X” series. The authors and “Xs” are MaurizioZollo, now at INSEAD, on bank acquisitions, PrashantKale, at Michigan, on technical alliances, and Phanish Puranam,finishing at Wharton, on high-tech acquisitions. The empirical setting in eachcase involves companies that are facing a series of tasks of the same generaltype, and the question if whether they are learning as they go, and if so howand whether it makes a difference. Mostly they are learning, and it isinteresting to see how it happens and the difference it makes. Zollo and I havea related conceptual paper that will appear in Organization Science.

Butunder the strategic management heading, the big issue that I want to raise hasto do with the evolutionary perspective on business success. The final sectionof my presentation is devoted to that.

Howone thinks of the research priorities also depends a good deal on one’s timehorizon. As a believer in the scientific program of evolutionary social science,I feel very strongly the crying need for high quality science on foundationissues. That would include, at one extreme, careful empirical studies oforganizational learning at a very micro level, and at the other, sophisticatedmathematical modeling of evolving systems as stochastic processes. But theopportunities for application of evolutionary thinking, for example, inunderstanding the evolution of capabilities and their relationship toprofitability, are very tempting targets. I would feel somewhat irresponsible inrecommending that someone, especially a young scholar, attend to the foundationissues rather than seizing those low-hanging fruit. But I still hope that someresearchers will be attracted to the foundation issues.

Beforeproceeding to the discussion of the evolutionary perspective on success, I wantto flag two gaps in understanding that frame important research issuesregardless of the disciplinary perspective from which you approach the topic.

First,an obviously appealing empirical program for evolutionary economics is tocharacterize how firms differ in their routines, and how those differences arereflected in differences in growth and profitability. But at present, we lack anadequate approach for characterizing routines in ways that would be useful forstatistical analysis. It is important here not to be distracted by counsels ofperfection. Keep in mind the large and arguably successful literature ondiffusion of innovations, where the problem of operationalizing “innovation”exists and is very similar to the problem of operationalizing “routine”. Forthe most part, the answer provided doesn’t get much beyond the name ofinnovation i and the yes-or-no questions of whether innovation i does ordoes not exist in firm j at time t. It would be very niceto be able to do a lot better than that, but “somewhat better” would be atleast “nice”.

Thesecond of the gaps relates to a big difference in perspective betweenevolutionary economics and organizational ecology. The valuable empirical workdone in the latter field has concentrated almost entirely on countdata—births, deaths, the number of organizations in existence – and on largepopulations. Organizational forms become more prevalent when the birth rateexceeds the death rate. An organizational routine, however, can become moreprevalent through net births or the differential growth of organizationsthat possess the routine, or through changes in routine in existingorganizations, perhaps driven by imitation. (See Nelson and Winter, 1982,Chapter 7, for formal analysis. Note that this discussion presumes that we cananswer Howard’s point about recognizing “the same” routine in differentorganizations, even though no two routines are really the same.) Giventhe extreme skewness of firm size distributions, and the related fact ofrelatively high concentration in most industries, it seems likely that thelatter two mechanisms are a lot more important quantitatively than the first.This, however, is an empirical question that calls for systematic study.

Iturn now to the evolutionary perspective on business success. My first pointabout this draws on the advertising slogan of the state lotteries: Youcan’t win if you don’t play. But in competition in the market place,“playing” is a lot more complicated than buying a lottery ticket. A lot ofit is about acquiring the necessary capabilities. It is usually the possessionof the relevant capabilities that distinguishes those who are playing from thosewho are not. So, it’s relatively easy to give the reasons why a great manyfirms are non-contenders. It’s just as easy as it is to give the reasons why,although we don’t know who will win the gold medals in the upcoming Olympics,we do know that I will not be among them. Those are relatively straightforwardissues. When it comes to sorting out among the potential winners among those whohave made the investments, that’s a much more challenging task. Whereas it istrue that you can’t win if you don’t play, what the lottery slogan does notgo on to say is – you’re damned unlikely to win if you do play. Andthat’s the harder part of the problem.

Now, the evolutionary perspective says: we’reskeptical about foresight in the evolutionary school. We think people don’tsee the future very well. You should try to refrain from relying on theassumption that they see the future well when you try to explain their behavior.As you look at these examples, of who are the players and who are not players,it turns out, time after time, that the crucial investments occurred for reasonsthat were not clearly associated with a clear image of the course of the futurein which these investments were going to pay off, in which they were going to besuccessful. Very frequently you see what the biologists call preadaptation, thatis to say, that the capabilities and the orientations were put in place bycausal forces or contingencies that had relatively little relation to thesubsequent evolution in which those behaviors proved to be successful.

The final point here is this. It is true that goodguesses and judgments are important success factors, along with persistence inbuilding the necessary capabilities. But if you’re trying to understand thesources of success, the fact that it’s often hard to tell a good judgment froma good guess (i.e., good luck) is a major stumbling block.

This evolutionary perspective is a skepticalone in terms of the feasibility of deriving useful strategic guidance. And inthat sense, the evolutionary perspective might be a little bit hostile to theagenda of strategic management. I’m even more hostile than that.

Ifyou look at the literature in management, in strategic management in particular,you see that there is a lot of enthusiasm for coming up with secrets of businesssuccess. There’s a lot of pressure to come up with business success secrets.There’s a lot of pressure to claim that you did even if you didn’t. As apart of this phenomenon, we have on these slides a list of what I consider to beundesirable attributes, which are notoriously, in my view, prevalent in much ofour research. And I’m not condemning anybody in particular; I’m just sayingthere’s a lot of this. There’s a lot of ex post storytelling,there’s a lot of looking for the attributes of successful firms withoutlooking whether they are present in unsuccessful firms, or looking at the issueof success syndromes…You see people mining secrets from cases of extraordinarysuccess. A very reasonable thing to do on its face, but you should have in theback of your mind the mental model of those 10,000 people flipping coins. Afterseveral rounds of coin flipping, you’ve got somebody in the 10,000 peoplewho’s got a terrific record of coin flips. If you don’t look at the samplefrom which the extraordinary success is being drawn, you don’t actually knowwhether you have an extraordinary success or not from a statistical point ofview.


There’sa lot of flexibility in the way we measure success; there’s a lot offlexibility in the time frames that people choose for their accounts of success,and lastly, there’s one great lurking danger. The claim is often made thatwhat businesses are about is shareholder value. If shareholder value is whatbusinesses are about, then we ought to be looking at stock market valuations.There, when we discuss the persistence of success or its origins, we’re upagainst the efficient markets hypothesis and the generally strong evidence insupport of the efficient markets hypothesis, not withstanding some contraryevidence. That evidence suggests that it is very hard to come up with anunderstandable basis for predicting success, and presumably, therefore, forprescribing for success.

Ithink that both the evolutionary program and the program of research instrategy, generally, would be considerably strengthened if we could stop doingthe stuff on the previous two slides. We need higher standards for thedemonstration that there actually are such things as useful strategic advice anduseful strategies, which are truly a matter of choice and discretion, that arebeing followed by some firms and not by others. That would mean that we wouldtake the evolutionary perspective on success seriously, of course invitingchallenges. Without challenges it wouldn’t be any fun, but we need to insistthat the standards of proof for claims that there is useful advice to be givenor that there are “rules for riches” ought to be set very high. I think thefinancial economists have had a lot of success with that approach, taking theefficient markets hypotheses as a very strong null that you had to fight withvery intensely, and I think we could learn a lot from that.

Inthe end, I think, let me underscore one point, I’m not saying it’s all luck:remember that to be a player, you have to make timely investments in the rightcapabilities—that we understand. The issue is, among those who are players, dowe have a really strong basis for saying what succeeds in the sense ofdiscriminating a successful player from an unsuccessful one? I think we could doa lot to advance both the evolutionary program, refining its perspective on therole of foresight, discretion, and choice in these processes, and also advancethe program of strategic management if we would take this course.


Johann Peter Murmann: Finally, Dan Levinthal. He will talk from downstairs. Thetitle of his talk is The Evolution of Evolutionary Ideas in the Academy ofManagement.


DanLevinthal: As management theorists and social scientists more broadly,we puzzle about why things are the way they are, and my sense is basicallywe’ve come up with two big ideas. And the two big ideas are some notion ofgod, and some notion of evolutionary processes. There are a variety ofmanifestations of these basic starting points. Perhaps the most prominentmanifestation in the social sciences of the godlike idea is the neo-classicaltheory of the firm. In that theory, we don’t quite live in paradise. Indeed,much of the analysis is about exactly how far we are from paradise due to theconstraints of property rights and information sets. In the evolutionaryperspective, there’s a notion that in fact there are many possible worlds. Theworld we live in is not necessarily any more special or virtuous than theseother possibilities. How we arrive at this particular world becomes an importantquestion. In particular, ideas of path dependence have been central in much ofour thinking—the idea that the state of the system at time t constrains,informs, affects probabilities of realizations of the state at time t+1. Lesscentral, but also really quite important is the notion that selection processesonly occur over what’s out there. So in some sense, we don’t only have aproblem of survival of the fittest, we have a problem of the arrival of thefittest. Some of Howard’s [Aldrich] work in his latest book talks a bit aboutthis genesis problem, the emergence of the variety. A particular form of thisquestion of the arrival of the fittest that is quite central for biologists isthe process of aggregation. How you go, say, from a single cell entity to amulti-cellular entity. This higher-order entity then forms the grist for themill of subsequent selection processes.

As Sid already suggested, we’re supposed to bothreflect on where this intellectual lineage has been and where it might go. Onefeature of the recent history of evolutionary thought in management is itssuccess. It is the focus of increasing energy and attention. Ironically, I thinkit’s done quite well in part because it has some similar properties as one ofits foils—neo-classical economics. One of the properties they share in commonis that they’re essentially context-free. It is this property that allowseconomists to invade lots of niches. They can take over political science, andthey can push into some aspects of sociology. Similarly, evolutionary theoristscan invade a variety of niches ranging from social psychology, to game theory,to industry evolution. Another part of its appeal, again a feature that I thinkit shares with neo-classical economics, is the ability to aggregate. The abilityto go from the characterization of lower level processes to higher levelentities. Clearly, the kind of aggregation mechanism is a bit different in thetwo lines of work, with evolutionary theorists relying on selection processes asopposed to equilibrium notions, but at some broad level the intellectual enginehas some important similarities. This issue of micro/macro linkages links upwith many of the same themes that Howard started us off with and raisesquestions of what constitutes the unit of selection.

First, what are we selecting? As suggested byHoward’s comments, we have to distinguish between selecting acrossorganizations versus selecting for organizations. If an organization is simply arepository of routines, there doesn’t necessarily have to be any interestingproperties about the organization itself. In some sense, what we would beselecting on is demography—the demography of routines. Some organizations mayhave a better set of routines and, in turn, a higher probability of survival,but there is nothing particularly organizational about the process. Firms getprofits and are wonderful and special because they have some interestingindividual attributes. They have a cherished brand name. They have the capacityto miniaturize electronics and so on. However, in such a perspective the firm isa bundle of these resources, these capabilities, but the firm itself doesn’treally have any meaning. So it’s important then, to think about in what waysthe organization itself might be an interesting and meaningful unit of analysisfor selection.

Here are some candidate answers to that question. Ithink an important starting point in thinking about this question is torecognize that firms receive profits and losses from the environment. Individualactivities within organizations do not. The clever proposal you sent to seniormanagement, the politicking and back biting that you engaged in late Fridayafternoon, they themselves are not directly rewarded from the environment. Acritical attribute of the organization is in some sense, its mechanisms ofcredit assignment. How can it go from some sense of overarching profit and loss,and then rain rewards, whether individual level incentive rewards or rewards inthe sense of resource allocation onto these various initiatives and actions thatare transpiring within the organization. Such a mechanism is truly anorganizational level property. Selection pressure could then operate ondifferent mechanisms of credit assignment that are out there, as opposed to thedemography of resources, or demography of relatively independent routines. Inthat sense, there is an important distinction between the natural selectionenvironment of the marketplace, versus the artificial selection environment thatmay be occurring inside the organization.

Anotherkind of basis, as it’s suggested in some of Sid’s comments, is the fact thatwe don’t just have necessarily isolated routines. We might have assemblies ofthese routines. The Fordist production system or the Toyota production system isnot a individual routine. It is an assemblage, a package, of a variety of kindof behaviors and action patterns that take place inside the organization. Thesebroader assemblages are properties of the organization and selection on theseassemblages is distinct from selection on the basis of individual traits and thedistribution of these individual traits across organizations.

Arelated notion is the idea of mutual adaptation. By being placed in the sameorganizational context, procedures and actors may become adapted to one another,resulting in heterogeneity that is truly an organizational property. Such anargument, in some sense turns transactions costs thinking on its head.Relation-specific qualities do not induce a particular governance structure;rather, the governance structure induces mutual adaptation and, in turn,relation-specific skills and capabilities.

Letme kind of conclude by addressing some of the issues that Sid raised about themanagement implications of evolutionary perspectives. Let’s consider how wemight possibly link the two columns on my fist slide—images of god-like designon the one hand and on the other hand ideas of blind evolution. Perhaps, to somedegree we can engineer these evolutionary processes. Consider the frequentlydiscussed need for organizations to balance processes of exploration andexploitation. Such manipulation is not going to be a precise engineering of theevolutionary process, but a broad awareness of the affect of alternativeorganizational policies on the dynamics of firm evolution. One particular level ofinfluence is the presence of multiple levels of adaptation and selection, whichthrough parallelism and hierarchy helps resolve the tension of theexploration/exploitation tradeoff. Experimentation can be present at some lowlevel, while the subset of these experiments that prove successful can beexploited at a higher level in the entity. This intra-organizational selectionprocess suggests at least a modest role for conscious intervention in theevolutionary process. More generally, what do the actors themselves actuallythink? We are considering the evolution of organizational entities comprised ofindividuals, not proteins. This difference may influence how we wish to modelsuch processes.


Johann Peter Murmann:What we are going to do now is togive Howard, Sid, and Dan a chance to respond to each other’s presentation,and then we’ll open up the floor for everyone else to join in. Howard’sgoing to go first.


Howard Aldrich: I just want to emphasize something I said this morning in my 8:30 session aboutthe challenge that selection logic poses to research design, and something thatDan said reminded me again of this. Particularly, what Sid said about thedangers of retrospective reconstruction, or picking what I would call outcomebased research. That is, picking winners, or picking organizations that aresurvivors at any point. Any cross sectional analysis that you do has theinherent bias of being outcome driven. Typically, a researcher using this stylesearches for—or ransacks, one could say—the organization’s history to findthe antecedents to the “success.” If one works forward, of course, you get avery different answer, using what I would call the event driven model. I canthink of very few research programs, with perhaps the exception of theecologists who used an event driven process to design their research. Butwhat’s lacking mostly is our recognition is that we actually--to construct anevolutionary explanation with data--need to have an event driven researchprocess, which means having information about the state of whatever it iswe’re observing evolve, whatever the unit might be—teams, groups,organizations, or populations—and we need to have repeated measurements over along period of time. And recognizing that after every trial, there could bedivergent events. Every event sparks not just one possible consequence, but alsomany possible consequences. That kind of research design is extremely costly andrequires us, unless we’re in the laboratory, to spend a lot of resources ondata collection. Think something like the Panel Study of Income Dynamics, orNational Institutional Survey of Youth. We don’t yet have an organizationalequivalent. We almost have, actually, with the data from the United States’Census Bureau’s BITS (Business Information Tracking Series), formerly known asthe LEEM (Longitudinal Establishment and Enterprise Microdata) files. This dataset links firms over different years in a very primitive way. We don’t havemany such research designs. So, almost everything we do in our field has someelement of an outcome bias. And I think that’s a very important thing torecognize. I am less sanguine than Sid is. He was talking as if we have moreinformation than we actually have. The best study is something like MikeCohen’s studies of people playing cards in a lab, right? It’s great stuff,but it’s just a start. So, we have to recognize the incredible challengeselection logic poses to the resources we currently put into our researchdesigns. We’re not talking here an incremental change. We’re talking a majordiscontinuity.


Sid Winter:I’ll make one comment about each of the other two presentations. First of all,Howard’s question about whether routines sort of disappear as units ofselection, if, in fact, they are unique either individually or unique as bundlesseems to me to point to a problem that is more serious for the resource basedview than it is for evolutionary theory. I think that the RBV may overemphasizethat “uniqueness” word in its account. It seems to me that if that critiquewere valid, it would pretty well dispose of genes as units of selection inbiology as well—since, apart from the case of identical twins, individuals areunique in their genetic composition once you look at all the dimensions. So, Idon’t think that is a fundamental problem. But as a matter of operationalizingthe program, it is a very real research task to understand how routines, whilenot being precisely identical from one organization to another, are similar, yetdifferent in selectively significant ways, and I think you have to able toincorporate both of those facts in a proper evolutionary theory.

Turningto Dan, I think he said something shocking, so I’m going to ask if he reallymeant that. Dan says that organizations considered as collections of routinesare not really organizations, and, to that I say, wait a minute Dan…routinesare coordination. A lot of people thought what organization was about wascoordination. Coordination is a very, very important phenomenon of the world.The complicated things that organizations accomplish for us are accomplishedwith extremely refined coordination in many cases, and those coordinationphenomena are very much illuminated by the study of routines and they’reparticularly illuminated by those people playing cards in Michael’s [Cohen]lab. So I think we have a lot of understanding about what the nature of highcoordination is, and how it arises, as I suggested in my retrospective on theissue of operational routines. I think I heard Dan say that organization isabout credit assignment. Is that what you said, Dan?


DanLevinthal: I didn’t intend to be shocking. The intent was to suggest that iforganizations were comprised of isolated routines, we might have someinteresting group selection but not necessarily interesting organizationalselection. I then tried, perhaps not coherently, to argue how we might go beyondsuch isolated routine action to assemblages of routines and broader notions oforganizational capabilities. Routines themselves are obviously aboutcoordination. They become an even more powerful source of organizationalheterogeneity when they adhere and become a broader assemblage. The morepervasive, the more systemic the routines, the greater the degree of spatialinterdependence, the stronger is the basis for group level selection. Thank youfor clarifying.


Sid Winter: Actually, as my final point, credit assignment is very important. Thismulti-level selection issue, the question of how organizations allocateresources internally, is absolutely critical. Because when you think about it,you realize that even if these evolutionary selection pressures are suchimportant drivers, they provide only a very low-dimensional feedback from actualmarket transactions. And it’s the internal workings of organizations thatelaborate that limited feedback into promoting some detailed behaviors andsuppressing others. So I think that the issue you raised there is a veryimportant one. I just think that it’s something of a luxury that we get aroundto the credit assignment problem, because it suggests that some other problemsare more or less behind us, which is good.


Johann Peter Murmann: We now open the floor for questions as well as views you want to share with us.I have to repeat your questions so they will audible on the recording of thesession.


Audience Question1 (Jill Shepherd, Strathclyde University, Glasgow, Scotland): I amskeptical of the idea that there are multiple levels of selection. Doesn’t SusanBlackmore's work show that there is only one level of selection?


Howard Aldrich:A couple weeks ago I posed the celibate priest example to her, and sheacknowledged that it made sense. (Even though Richard Dawkins of course wouldpass out if he heard her say this, please don’t pass this on, of course, toRichard Dawkins.) But what she actually said that yes, it does make sense totalk about selection at the group level, as well as the individual level.


Audience Question 2 [KathyEisenhardt]: What are themain differences between evolutionary theory and the resource-based theory?


Sid Winter: I have to choose my words carefully for this one. The RBV is a very flexibletheory. [Audience laughs] I think its popularity and the scope of itsapplication has something to do with that flexibility. Evolutionary theory isstill pretty flexible, but not that flexible. I think that in the basic view ofthe firm, a very important question is whether the category of resources isunderstood to include processes or routines, as opposed to things that are morein the nature of tradable assets. Some of the analysis offered by the RBV makesperfectly good sense if resources are understood to be tradable assets. It’sperhaps…well, it depends on which article you read, but some articles are veryclear about the speculative aspect of that theory. That is to say, you onlyprofit from having these unique resource configurations if you bought theresources cheap, for one reason or another. If it’s confined to an analysiswhere you could think of firms just as portfolios of tradable assets, thenit’s quite a narrow theory. If you broaden it so that it includes things likeroutines, then a lot of the analysis about the nature of profitability orcompetitive advantage begins to look more suspect because very frequently therearen’t markets for those things, and you can’t understand advantage in thesame terms when you have idiosyncratic firm processes among the things that arecharacterizing the firm. I had to try at sorting out some of this territory inmy paper on the four R’s of profitability, which is in the volume edited byCynthia Montgomery (Winter, 1995) and also I have long been an admirer of theDierickx and Cool (1989) paper on asset accumulation and sustained competitiveadvantage, and I think that paper essentially lays down a foundation where youcould understand that evolutionary theory was from that point of view a kind ofa branch of the RPV with a particularly strong emphasis on certain kinds ofinternally created competencies.


Audience Questions 3: How does the dynamic capability theory differ from the resource-based theory?How is it different from evolutionary thinking?


Sid Winter: Well, the dynamic capabilities theory is a very prominent illustration of astanding major problem in developing this evolutionary program. That problem--already alluded to here, especially in Dan’s comments—is the problem ofunderstanding what is going on in behaviors that on the one hand are clearlypatterned and persistent, and systematic in certain kinds of ways, and yet inother ways are addressing and creating novelty, and displaying human creativity.Dynamic things like new product development processes are very much of thatcharacter—organizations have their ways of doing these things. They arepersistent things—you can tell one style from another. In lots of respectsthose are routine-like features of dynamic capabilities. But on the other handthe whole point is creating something new. Facilitating creativity is a big partof the story and I think it’s a very great challenge to understand how thosethings fit together. One other thing I will say about that is that for somereason, perhaps it has something to do with the Teece, et. al., article [1997],I think the term “dynamic capabilities” has acquired a certain aura. As Isaid before, I don’t believe that dynamic capabilities are the key in thesense that if we pursue this idea we will all learn how to make above averageprofits. But, nevertheless, I think there is still a real phenomenon there, andone that is worthy of a lot of study and certainly it has been getting some ofthat.


Audience Question 4:What does the panel think Maturana’s ideas of self-organizing systems? How docomplexity theories fit into the evolutionary paradigm?


Johann Peter Murmann:Who wants to take this? … I guess I have to assign it to someone: Howard!


HowardAldrich:As I said, this is my fourth session today, and although my eyes are open, Ithink the light is not on. I have just one thing to say…you reminded me ofsomething I was going to say in response to Sid. I think it’s difficult to useevolutionary ideas to look at large firms, and to get the kind of purchase thereyou could get if you looked at smaller firms. My argument is that we can learn aheck of a lot more by looking at the fruit flies and not at the condors, or theeagles or whatever other “big” species you could name. This morning Imentioned the redwoods versus the crabgrass. You can look at things likeyou’re describing in a more detailed way in smaller firms. In my book I talkabout a lot about the movement from nascent entrepreneurs to new organizationsand the problem of coherence and boundedness and how boundaries emerge throughcommunities of practice. Those kinds of things you can see if you followorganizations up from the ooze to coherence. It becomes a lot harder to see suchtransformations when you switch over to the kinds of firms I think Sid playswith, the kind the Wharton School people are interested in, “real firms” andwhen they say “firms,” they mean Fortune 500 companies. They’re talkingthe .000000001% that get onto the Fortune 500 list. It’s a lot harder in thosecases to see this stuff, so I think certainly looking at smaller firms gives usinsight to these kinds of processes, and the coherence problems, and that’swhere I would say my own efforts are certainly going to go in the future. Beyondthat, your question went off in ways that my brain couldn’t follow.


Sid Winter:I’m not sure I understand the question either. But let me just say that itseems to me a lot of the complexity theory ideas that you reference there mayhave their proper role in understanding the origins of patterned behavior, whichis certainly one very important part of the story. On the other hand,persistence is another very important part of the story, and at least among thelarge organizations that Howard has just disparaged, persistence is a veryinteresting and significant feature. The AT&Ts, the IBMs, the Boeings, etc.,of the world have been around a long time. For a large fraction of thetotal time that has elapsed since their principal technologies were invented,the people have come and gone, but those organizations have existed aspersistent structures that are the society’s repositories of particular,technological and organizational capabilities. The processes by which theysustain themselves and remain able to command those types of capabilities are aworthy subject of study too. I don’t think those bottom-up complexity ideasare too helpful with this question.


DanLevinthal: Just as a footnote in response to Howard. If you move beyond birth and death asdependent variables, and look as Sid has at differential growth rates and worryabout the size distribution of firms, then there actually is a lot of action,even amongst these fairly large entities. So, while their probability of deathis miniscule, shifts in the size distribution are actually non-trivial evenamong that sub-population.


Audience Question 5 [PhilBromiley]: Sid don’t you thinkthat you are a bit too pessimistic about the role of managers in making a firmsuccessful.


Johann Peter Murmann:Maybe I can tell a little anecdotethat speaks about the difficulty of understanding the role of managers. When Iwas a little kid, I went with my grandfather on a ferry. I saw the captain steerthe boat and I really wanted to do it too. So I asked him if I could take overthe helm for a while. The captain was nice and let me do it. He gave me veryclear instructions: always point the ship to the big tower that you see far awayon the horizon; this will get us where we want to go. And so I was steering butafter a while it got really boring. You know, these big ships don’t move atall. Just holding the helm and steering the ship in the same direction prettyquickly loses its attraction for a little boy. My grandfather came by and Iasked him if could he take over for me. His sense of duty made him stand in forme. But after and hour and a half, granddaddy was bored too. He ran tocaptain’s suite and demanded that someone take over steering the ship. Thecaptain told him in a calm voice that there was no need to worry. The ship hadbeen on autopilot all the time. But granddaddy and I had been under theimpression that we were actually steering the ship. I think this little anecdoteillustrates quite nicely what kind of situation managers can be in.


Sid Winter: I’m not sure it’s possible to improve on that. That was pretty good. WellPhil, I tried to cover the issue that I think you’re raising by saying, ofcourse, you can’t win if you don’t play, and playing is a more complicatedmanner in business than it is in the lottery, and maybe you even learn somethingabout how to play by taking strategy courses. I think that’s quite possible.The issue though is more like this: Phil, if you think that there is usefuladvice to be conveyed, then here is the thought experiment: you tell me what youneed to run this experiment. You want one hour with the CEO? Five hours? A week?You want a research grant for three years before that? Will you bet some of yourportfolio by putting it in the stock of the company you’re advising becauseyou’re confident that if you can only have that audience with the CEO that youcan actually change things in the direction that will make that company moreprofitable? If you try to think in those terms, which is to think anexperimental intervention to see whether a certain kind of advice makes adifference, then I think it is doubtful. Even in the case of lean manufacturing,where the differences in performance were very real, whether the fact that theAmerican automobile companies were “slow” in doing that was basicallyattributable to their ignorance of what was in strategy journals, or in strategycourses, I think is very doubtful. I think there are other reasons why they wereon the course and on autopilot in the way in that Peter just suggested. So Ihave in mind a rather high standard for demonstrating the power of strategicadvice. But even if I’m right in that part of my pessimistic view, thatdoesn’t mean we have nothing to say. I still think we have a lot to say.


DanLevinthal: I’ll add a little bit to that. Actually, I think, Sid, that some of your ownwork on replication, and some of the work GiovanniGavetti and I have done (2000) has given us the beginnings of thatmiddle ground that I think Phil maybe hunting for. The notion is that we mayhave crude templates as to what constitutes useful information. Of course, ifthat’s all you knew you would likely experience quite poor performance. Thesearticulated, codified cognitions aren’t enough to take you very far. However,they might set you on what might ultimately emerge as an interesting path. Ofcourse, you might die and be weeded out before you get to effectively executethis new initiative. I think we can begin to explore a little bit the kinds ofwildly under-specified suggestions that might affect behavior in possiblydysfunctional and potentially useful ways. Consistent with the lottery imagery,I will not have great confidence necessarily in the expected value of thoseefforts because they are so under-specified. Ultimately, however, my interest isin the social science problem of understanding how belief structures interactwith experiential learning processes, rather than the management problem ofexactly how good on average such journeys are likely to be.


Audience Question 6: The question to Sid is—has anything happened since 1982? [Editorial Note: Thisis the year when the Nelson and Winter book came out.]


Sid Winter: There’s this new book (Dosi, Nelson, and Winter, 2000) that just cameout…Well, if you ask if anything has happened that will prevent you from beingmugged by economists, the answer is probably no. If you can really get them in acorner, you might be able to embarrass them with the sort of behavioral decisiontheory results that are very strong and very reproducible. I’ve seen DickThaler do his number on economists, running the experiment in the room with theeconomists as the subjects, and having the standard results come out. But, ofcourse, that doesn’t make them change their spots the next day.


DanLevinthal: A minor addition. I don’t think anything will prevent the mugging, but I thinkalong the lines of Sid’s comment that recent work in behavioral game theory isa particularly powerful illustration of the role of bounded rationality becauseof the presence of competition in those settings. These sorts of experimentsdiffer from early market games which often were such structured contexts thatalmost any behavior would lead to a competitive equilibrium. Indeed, Sunder andGoode (1993) have a paper in which they show that in the setting of a simplecompetitive market, random bidding behavior, subject to budget constraints,leads to the competitive equilibrium. It is not until you explore games wherethere’s not a trivial interdependence among the actors that outcomes depend onthe particular nature of people’s learning process. However, I have to confessas to not have much optimism regarding the avoidance of a mugging.


Audience Question7  [JanRivkin]: What would a national income accounting of evolutionary economics look like?


Sid Winter: The national income accounting part was a metaphor I think, right? That is, weneed some way to operationalize key variables…. and right, that’s exactlywhat we need. We need some way to operationalize key variables. I think we needthat to advance the science. We need more careful detailed studies oforganizational routines as complex structures above the laboratory level, and weneed to address this problem of quantifying the diversity in families of similarroutines. If you could do that, then you would have the beginnings of an engineon which you could do statistical analysis and try to relate performanceoutcomes to the characterization of the routines. So, I think that’s one veryimportant part of the program. I also think coming back to some of Dan’sthemes that we know remarkably little about how internal resource allocationsprocesses in large organizations operate as a level of selection. We knowsomething about those processes, but not a lot, and it’s not been placed inthis type of framework. That’s another area that’s really very important andcalls for very close examination. I’m sure I could come up with a number ofothers. You know my e-mail address, I think.


DanLevinthal: I think one of the challenges is how context dependent are these measures goingto be. For instance, in the card game of Michael Cohen’s work, it is quiteclear what a routine is. But for the national income account you want, some setof context free measures must be developed. For instance, in general how dofirms manage interdependency on the value chain? Expressed this way, thechallenge of such “national accounts” is quite daunting. Maybe some folkswill come up with some taxonomies of this relatively generic sort. But I thinkthat it’s important to recognize that a key part of the patterned behavior ofa routine is the context dependence of these behaviors. So I’m not offering apossibility or impossibility statement, but I am flagging how Spartan theindicators are as we rise to these more aggregate levels.


Howard Aldrich: If you’re compiling a laundry list, I could add a couple of more. One of themis something that ArtStinchcombe mentioned on areview he did on the website that Peter runs – etss.net. Hisargument is that we need more organizational physiologists or I think he woulduse a better term, such as “organizational ethnographers.” We need peoplewho spend lots of time in organizations in a disciplined way, and collectsystematic information with a reliable accounting scheme. It might go some waytoward what we’re talking about here. We don’t encourage such research amongour graduate students. I don’t know why economists figure so heavily into thisdiscussion today. How many economists are actually in this room? [About threepeople raised their hands.] Yeah, that’s what I thought. Ok. We’re inmanagement departments, for goodness sakes! Think about the kind of training ourstudents get. What you should be doing is teaching your students how to beethnographers --some of your students ought to be doing fieldwork. They ought tobe learning anthropology. They ought to be going to anthropology departments andlearning how to do fieldwork and do it in a disciplined and systematic way. Notwalking into organizations, spending a couple of hours there, and walking outwriting a case report. That’s one thing you can do. A second thing iscomputational modeling. I’m looking again at my friends Alessandro Lomi andEric Larson (Lomi and Larsen, forthcoming) out there. They have a nice book,coming out on computational modeling, looking at some of these issues. Youdon’t have to actually get field data. You don’t have to get census data.You can actually build models with realistic assumptions. Some of them comingfrom the lab experiments that Mike Cohen and his colleagues have done. Socomputational modeling is one possibility. As I mentioned before, just torepeat, we don’t have national accounting data on organizations. It’s themost scandalous thing, I think, about our national data. We have an enormousamount of information collected every year from firms, even the big firms, thevery biggest firms, and small firms. And that information is stored in such away that it makes it impossible for us to link it up over time and so we don’thave the ability to create decent pictures of how the economy has evolved. Wehave cross-sectional snapshots, we have the biggest firms cataloged very well,the publicly traded firms. Well, that’s again, the 20,000 or so publiclytraded firms, the firms that the business schools feature in their teaching. Butthe larger population of organizations is recorded and then filed away andforgotten, so we need to put more money into building these data sets that allowus to link firms over time. So those are three other possibilities. This is piein the sky, of course, but actually, the part that’s not pie in the sky, isyou training more people to do field work. That you could do. You all haveanthropology departments on your campus. Send a few of your brightest studentsover there and have them learn fieldwork techniques.


Audience Question 8 [IvoZander]: Is there a theory of thefirm in evolutionary economics?


Sid Winter: Well, the aforementioned 1982 book did not say that it was advancing a theory ofthe firm. So, the question is appropriate, I guess. My problem with the questionthough is that I don’t think anybody has a theory of the firm. The reason Ithink that is that, to my mind, one of the most salient characteristics of thatpuzzle is the fact that firms, as legal entities, can be put together by mergersaccording to the impulse of investment bankers and the marketing of investmentbankers. There is an industry of putting firms together and taking them apartagain. So, if by a firm you meant the results of that process, then you wouldhave to understand how the firm-putting-together-and-taking-apart industry doesits work, which is not a subject as far as I can see that’s being addressed byany of the alleged contenders for the theory of the firm. Now, on the otherhand, if you don’t mean to refer to the reality that is out there, the thingsthat these investment bankers put together for longer or shorter periods oftime, then you ought to be clear about that. You could have a perfectlyrespectable scientific approach that says that I’ve got a theory of the firmhere, but City Group isn’t an example. It’s not one of those firms that mytheory is about, that’s something the investment bankers did. That would be arespectable approach to the problem: to have an idea about a firm in which notall the commonly recognized instances out there in the economy were actuallyexamples of firms as understood in the theory. But again, I don’t see peopledoing that. So I don’t know where they stand on that basic issue. And I thinkit’s a very fundamental issue, and from an evolutionary point of view, theevolutionary point of view as contrasted to neo-classical or rationalistic pointof view, any dumb thing can happen for a while. Insofar as investment bankersdecide to do dumb things, those things happen for a while, and the question isnot whether there are some dumb things out there in the world, but how long theylast, and what are the processes that eliminate them. So that’s a somewhatdifferent agenda from the standard theory of the firm question.


DanLevinthal: Along those lines, and what I tried to point to in my initial remarks, I wouldphrase the question somewhat differently–is the firm an important unit ofselection? Is evolutionary theory about the selection of particular routines ortraits; and are firms simply carriers of these properties or is what is beingselected truly a group-level property?


Audience Question 9: How come the speakers have spent the last hour and fifteen minutes on selectionand not on the more important process of variation?


Johann Peter Murmann:Howard can take this because in his new book he talks a great deal aboutmechanisms that generate variation.


Howard Aldrich: I think we are talking about random event and then path dependence today.That’s the answer for why today looked like it did. I think selection logic iswhat most cleanly separates evolutionary theory from the other kinds of thingsyou’re going to hear at the Academy of Management. The papers you’re goingto hear in the other sessions will talk about variation, and, in fact, most ofthem will talk about retention. They’re talking about management systems.They’re talking about retention, persistence, and reproduction replication.That’s a fairly common theme. Variation is also a fairly common theme. Whatyou don’t hear so much of is selection. So when I talk to other people, peoplewho aren’t evolutionary in their thinking, the first thing I stress to them isthe importance of selection logic. I tell them a story like my chicken beakstory, just to get them to think about what it would mean to take a selectionargument seriously. There’s no evolutionary theory without a theory ofvariation. It’s very clear. You don’t have something to select unlessyou’ve got coherent units that are out there to be selected for, or selectedwithin, or whatever. Don’t take our discussion today as any sign of animbalance of thinking by evolutionary theorists; it’s a chance event.


Sid Winter: From the point of view of the economics discipline, you know, you raise a veryimportant point. There’s a huge difference between the neo-classical orthodoxperspective in economics on what the market system is doing for you and theperspective in evolutionary theory. In evolutionary theory the story is: themarket system is a device such that, when new opportunities and new problemscome up, a bunch of different actors have incentives to think up solutions andtry them. Some of those solutions are better than others and the market systemfunctions in such a way as to promote, at least in a crude sense, superiorsolutions, and suppress the less desirable solutions. That is a story about howsociety goes about solving its ever-changing economic problems; that is acompletely different story than the story that orthodox economics tells aboutefficient allocations. In the evolutionary story, obviously, the experimentationand the generation of variety is a very important element. Organizationalarrangements that might look entirely sensible in a world in which you thoughtthere were no more new problems to solve, look terrible in the perspective of aworld in which there are always new problems to solve and which you needexperimentation and you need to have people with strong incentives to tryto come up with new solutions. I think that it’s a very good thing that youraised that issue.


Audience Question 10: If the efficient market process selects only the best players, how is that ideaconsistent with what Howard articulated in the very beginning of our sessionthis afternoon that sometimes individual fitness may be inconsistent or may workagainst the fitness of the group? And if the group is the important economicunit, how would that square with the market doing its work properly?


Sid Winter: I think if I understood your question correctly, the reference to efficientmarket mechanisms goes back to what I just said, which is, the market sorts outthe more desirable solutions to problems, and suppresses the undesirable ones,as opposed to other connotations of the term “efficient markets”, fromfinance or elsewhere. On that assumption as to what the question is about, themulti-level selection issue is quite relevant in that, for example, orthodoxeconomics has a long tradition of hostility to industry level organization,seeing cartelization everywhere, basically. It says, since we know from thetextbook theorems that all we need is independently acting competitive firms inorder to achieve nirvana, it must be that anytime you see industry levelorganization that this is perverse and undesirable. I have to say that as apart-time political analyst, I think the standard orthodox view on industrylevel organization should not be neglected. It may well be cartelization indisguise. But I think nevertheless that it’s pretty clear that in many casesindustries have been made to function better because some check was imposed by ahigher level of organization, whether it was in setting standards or settingrules for fair competition, and so on. So I think, in fact, the evolutionaryperspective is open to a constructive role for more levels of organization thanis characteristic in the orthodox theory.


Audience Question 11: Is social engineering of fitness landscape in fact possible given there may notbe enough information, or good information, to actually do this. Is it possibleto engineer a coherent system of incentives?


DanLevinthal: It’s a good comment. The notion is that you’re likely to get it wrong, boththe boundedly rational social engineer and the boundedly rational lower agentresponding to these structures. What I’m suggesting is that we consider issuesof mechanism design by and for boundedly rational actors.


Johann Peter Murmann:It is 6 pm and we’ve got to stop, but as I said before, if you have furtherquestions or viewpoints you can post them in the discussion forum on theetss.net website. So thank you very much for coming and please give a big handfor our panel.






Adler,P. (1993). "The Learning Bureaucracy: New United Motor Manufacturing,Inc." Research on Organization Behavior: 111-194.


Aldrich,Howard (1999). Organizations Evolving. London, SAGE Publications.


Aldrich,H. and C. Fiol (1994). “Fools rush in? The institutional context of industrycreation.” Academy of Management Review 19: 645-670.


Axelrod,R. and M. Cohen (2000). Harnessing Complexity: Organizational Implications ofa Scientific Frontier. New York, The Free Press.


Blackmore,S. J. (1999). The Meme Machine. New York, Oxford University Press.


Bowman,E. H. and C. E. Helfat (2001). “Does corporate strategy matter?” StrategicManagement Journal 22: 1-23.


Clark,K., W. B. Chew and T. Fujimoto (1987). “Product Development in the World AutoIndustry.” Brookings Papers on Economic Activity 3(0): ages 729-71..


Cohen,M. and P. Bacdayan (1994). “Organizational routines are stored as proceduralmemory.” Organization Science 5: 554-568.


Dawkins,R. (1976). The Selfish Gene. New York, Oxford University Press.


Dierickx,I. and K. O.Cool (1989). “Asset stock accumulation and sustainability ofcompetitive advantage.” Management Science 35(December): 1504-1511.


Foss, N. 1993.“Theories of the Firm: contractual and competence perspectives.” Journal ofEvolutionary Economics. 3: 127-144.


Foss,N. 1997. “Evolutionary and Contractual Theories of the Firm: How do theyRelate?” Rivista Internazionale di Scienze Sociali, 3 (Luglio-Settembre):309-337


Gavetti,G. and D. Levinthal (2000). “Looking Forward and Looking Backward: Cognitiveand Experiential Search.” Administrative Science Quarterly.


Gode,D. and S. Sunder (1993). “Allocative Efficiency of Markets withZero-Intelligence Traders: Market as a Partial Substitute for IndividualRationality.” Journal of Political Economy 101(1): ages 119-37.


Hodgson,G. Martin (1999). Evolution and institutions: on evolutionary economics andthe evolution of economics. Northampton, MA, E. Elgar Publishing.


Klepper,S. (1997). “Industry life cycles.” Industrial and Corporate Change 6:145-181.


Lomi, A.and E. Larsen. (Editors). Forthcoming. Dynamics of Organizational societies:models, theories and methods. Menlo Park, CA and Cambridge, Massachusetts:AAAI Press/MIT Press


McGahan,A. M. and M. E. Porter (1997). “How much does industry matter really?” StrategicManagement Journal 18(Summer Special Issue): 15-30.


McKelvey,B. (1982). Organizational Systematics: Taxonomy, Evolution and Classification.Berkeley, CA, University of California Press.


Rumelt,R. R. (1991). “How much does industry matter?” Strategic ManagementJournal 12: 167-185.


Schmalensee,R. (1985). “Do markets differ much?” American Economic Review 75:341-351.


Szulanski,G. (1996). "Exploring internal stickiness: Impediments to the transfer ofbest practices within the firm. Strategic Management Journal: 17(Winter Special Issue): 27-43. 1996.


Szulanski,G. (2000). " Appropriability and the Challenge of Scope: Banc OneRoutinizes Replication." In The Nature and Dynamics of OrganizationalCapabilities, G. Dosi, R. Nelson and S. Winter, eds. Oxford: Oxford UniversityPress: 69-98.


Teece,David J., Gary Pisano and Amy Shuen (1997). “Dynamic Capabilities andStrategic Management.” Strategic Management Journal 18(7): 509-533.


Usselman,S. W. (1993). “IBM and its imitators: organizational capabilities and theemergence of the international computer industry.” Business and EconomicHistory 22(Winter 1993): 1-35.


Winter,S. G. (1994). "Organizing for continuous improvement: Evolutionary theorymeets the quality revolution." In Evolutionary Dynamics of Organizations.J. A. C. Baumand J. V. Singh, Eds. New York, Oxford University Press: 90-108.


Winter,S. G. (1995). "Four Rs of Profitability: Rents, Resources, Routines andReplication." Resource-Based and Evolutionary Theories of the Firm:Towards a Synthesis. C. Montgomery, Eds. Boston, Kluwer Academic Publishers.


Zollo,M. and S. G. Winter (forthcoming). “From Organizational Routines to DynamicCapabilities.” Organization Science.



Returnto ETSS Home.