Recent Evolutionary Theorizing About Economic Change jel1995.pdf · economists have even tried. (For a splen-did history of evolutionary theorizing in economics, see Geoffrey Hodgson

Nelson: Theorizing About Economic ChangeJournal of Economic LiteratureVol. XXXIII (March 1995), pp. 48–90

Recent Evolutionary TheorizingAbout Economic Change

By RICHARD R. NELSON

Columbia University

An earlier version of this essay was commissioned by Neil Smelserand Richard Swedberg, for their new Handbook of EconomicSociology (1994). I want to thank them for inducing me to trysuch a broad survey, and to thank the Sloan Foundation whosegrant to the Consortium on Competition and Cooperation madethe work possible. Sidney Winter, Stanley Metcalfe, GeoffreyHodgson, and Giovanni Dosi, my close partners in developingevolutionary economic theory, gave helpful comments on an ear-lier draft. So did more than three score other scholars to whom Isent out a copy. I thank them all, and hold them blameless for theresult.

I. Introduction

The Mecca of the economist lies in economicbiology . . . . But biological conceptions aremore complex than those of mechanics; avolume on Foundations must therefore givea relatively large place to mechanicalanalogies, and frequent use is made of theterm equilibrium which suggests something ofa static analogy. (Alfred Marshall 1948, p.xiv)

THIS FAMOUS PASSAGE from AlfredMarshall’s Principles of Economics

(it first appeared in the fifth editionwhich came out in 1907) nicely bringsout two issues, which are as germane toeconomics today as they were when Mar-shall wrote. The first is the heavy reli-ance by economists in their formaltheorizing on the notion of “equilib-rium.” The other is the appeal that “bio-logical conceptions” have for many

economists, particularly when their focusis on economic change.

Marshall clearly believed that our sci-ence should aim to understand economicchange and not simply the forces mold-ing and sustaining the current configura-tion of economic variables. His “me-chanical analogies” and equilibriumconcepts included those of Newtoniandynamics, as well as those associatedwith the balancing of forces on bodies atrest. Since the time of Marshall, and fol-lowing his lead, economists have devel-oped their own equlibrium concepts.While until recently they were mostly as-sociated with analysis of situations pre-sumed to be at rest, in recent yearsmuch of economic theorizing has beenconcerned with dynamics, and the equili-bria, like those of Newtonian dynamics,are ones in which the variables under

48

study change over time. But Marshallmight observe that the equilibrium con-cept in these models still somehow has astatic feel to it.

Few economists confuse the formalstatic or dynamic equilibrium theorywith the reality. Most readily acknow-ledge that at least some economic situ-ations need to be understood as involv-ing significant elements of novelty, sothat the actors should be regarded assearching for a best action, as contrastedwith actually having found it. In theiranalysis of certain economic phenomena,for example technical advance, manyeconomists recognize that frequent orcontinuing shocks, generated internallyas well as externally, may make it hazard-ous to assume that the system ever willget to an equilibrium; thus the fixed ormoving equilibrium in the theory mustbe understood as an “attractor” ratherthan a characteristic of where the systemis.

However, until recently at least, therehas been a resistance to building thesecomplications into formal models. Partlythe reason is a belief that to do so wouldmake the models intractable, or at leastcomplex and difficult to understand. Thisseems to have been Marshall’s concern.But nowadays this predilection seemsmore than simply a matter of analytictractability and convenience. When ex-pressly doing or talking theory, unlikeMarshall most contemporary economistsseem to be drawn to equilibrium con-cepts as a matter of aesthetics. Generalequilibrium theories are seen as elegant,and theories that depart from these can-ons are seen as somewhat ad hoc.

It is interesting, therefore, that wheneconomists are describing or explainingparticular empirical subject matter in acontext that does not demand that theywrite or talk theory explicitly, they ofteneschew equilibrium language, and revealthe same inclination as did Marshall to

make use of “biological conceptions” ormetaphors. I noted above the proclivityof many economists to consider individu-als and organizations as entities thatsearch and “learn.” Industrial organiza-tion economists sometimes characterizecertain industries as “young” and othersas “mature” with the connotation thatvarious things naturally happen as an in-dustry gets older (see e.g., Dennis Muel-ler and John Tilton 1969). Similar lan-guage often is used in comparingeconomies. Evolutionary or developmen-tal language is used quite widely byeconomists to describe how the structureof an economy, or an industry, or tech-nology, or the law, changes over time.Writings in economic history almost in-variably are full of such biological meta-phors.

All this is reminiscent of Marshall. Yetwhile he was attracted to “biological con-ceptions,” it is apparent that Marshallnever had in mind simply applying bio-logical theory to economics. Indeed, thefact that he felt himself forced to fallback on “mechanical analogies” tells usthat he found it very difficult to developa formal theory, based on “biologicalconceptions,” that he thought adequatefor economic analysis.

In the years since Marshall, not manyeconomists have even tried. (For a splen-did history of evolutionary theorizing ineconomics, see Geoffrey Hodgson 1993.)Indeed, while some contemporaryeconomists continue to feel the sametension as did Marshall, that groupseems definitely a minority. One cer-tainly can rationalize the two differentstyles of economic discourse and analysisas just what one would expect, giventheir purposes. Describing, and explain-ing, in a context where it is important tobe sensitive to the details, is one thing.Theorizing is quite another.

However this proposition is problem-atic on at least two counts. First, the far-

Nelson: Theorizing About Economic Change 49

ther the language of particular explana-tion is from the logic of formal theory,the less analytic structure the latter canprovide the former. Economists whowould eschew equilibrium language, anduse “biological conceptions” in describ-ing and explaining, must pay an analyticprice. Those who do implicitly are takinga position that the analytic structure ofequilibrium theory misses elements theyregard as essential to their story, andthus are willing to pay that price.

And second, the argument draws toosharp a line between formal theorizingand verbal economic explanation. Winterand I have argued that, because the realeconomic world is so complex, theorizingabout it tends to proceed at at least twodifferent levels of abstraction (Nelsonand Winter 1982, pp. 46–48). Formaltheorizing is one level. By formal theo-rizing we mean what economists dowhen they are self consciously puttingforth a theoretical argument.

But economists also need to be under-stood as “theorizing” when they are try-ing to explain what lies behind the par-ticular phenomena they are describing,even when they are not advertising theiraccount as a “theory.” Winter and I havecalled this kind of analysis “appreciative”theorizing. While starting with the em-pirical subject matter, the accounts putforth by economists of the developmentof an industry, or the evolution of a tech-nology, focus on certain variables andignore others, just as is the case withformal theory. Quite complex causal ar-guments often are presented as parts ofthese accounts, if generally in the formof stories.

Thus the difference between the lan-guage and the logic of economists’ sto-ries about economic growth, which ofteninvolve evolutionary or developmentalconcepts, and that of equilibrium theory,is described inappropriately as a differ-ence between description and theory.

The difference is between two differentkinds of theories, in the sense that themechanisms and relationships treated ascausal are different, or at least appear tobe.

One could respond by arguing that,while the language may be different, infact the substance of theories using “bio-logical conceptions” and equilibriumtheories is not very different. In particu-lar, the theories predict much the samethings. There is no real difference be-tween saying that firms literally maxi-mize, and saying that their behaviorshave been learned through trial, error,and correction, and in some cases havebeen selected through the competitiveprocess. Thus extant actors behave “as if”they maximize. (The classic statement ofthis position is, of course, Milton Fried-man’s, 1953.)

Economists are not alone in puttingforth this argument. A number of evolu-tionary theorists in biology do also. Boththe economists and the biologists whotake this position admit that, at any par-ticular time, the actual system may notbe precisely in equilibrium, but proposethat it generally is close enough so thatthe characteristics of equilibrium tellone a lot about the actual situation.

But economists who use the languageof development and evolution in tellingtheir stories apparently do not believethat concepts like optimization and equi-librium can explain adequately the phe-nomena they are addressing, and theseeconomists have kindred souls in biol-ogy. Many students of biological evolu-tion strongly deny the proposition that“optimization” provides a meaningful ex-planation for the character of extant liv-ing forms, even when the observed con-figuration seems relatively durable andstable (see e.g., Stephen Gould 1980). Ithas been argued that the process of evo-lution is strongly path dependent andthere is no unique selection equilibrium.

50 Journal of Economic Literature, Vol. XXXIII (March 1995)

Any “optimizing” characteristics of whatexists therefore must be understood aslocal and myopic, associated with theparticular equilibrium that happens toobtain. The heart of any explanation ofextant living forms thus must be evolu-tionary analysis of how the particularequilibrium, and not a different one,came to be. Further, often there is goodreason to suspect that evolution pres-ently is going on at a relatively rapidrate, and thus equilibrium of any kind isnot an appropriate concept for analysis.

It would appear that many economistswho use developmental and evolutionarylanguage have in mind notions like these.While, as we shall see, the economistsusing evolutionary language in theirtheorizing are not of one ilk, almost allare, in effect, positing that to say thatactors behave “as if” they were maximiz-ing does not tell us much about why theyare doing what they are, and providesonly a start on any prediction of whatthey will end up doing if conditionschange. Many clearly believe that pathdependency is important in economics,and a number argue that the phenomenain which they are interested must be un-derstood as associated with continuingdisequilibrium, not equilibrium.

Until recently economists have usedthe language of evolution almost exclu-sively in their appreciative theorizing.However, in recent years evolutionaryconcepts have been employed increas-ingly in formal evolutionary theorizing.The book published by Sidney Winterand myself just over a decade ago (Nel-son and Winter 1982) has been followedby a number of others also exploring for-mal evolutionary theorizing in econom-ics. (See among others Dosi et al. 1988;Paolo Saviotti and Metcalfe 1991; PhilipAnderson, Kenneth Arrow, and DavidPines 1988; Richard Day and GunnarEliasson 1986; Norman Clark andCalestous Juma 1987; Lars Magnusson

1994.) Several recent articles have can-vassed the new writing. (See for exampleChris De Bresson 1987; and RichardLanglois and Michael Everett 1992.) Ul-rich Witt (1992) has pulled together acollection of what he regards as classicarticles in evolutionary theory. Hodgson(1993) has provided an elegant analytichistory of evolutionary theorizing in eco-nomics, and a forceful argument thatMarshall was right about Mecca. In 1991a new Journal of Evolutionary Econom-ics was founded, and several other newjournals have advertised an interest inevolutionary economics.

The recent work on formal evolution-ary economic theories has had severaldistinct, if connected, sources. One isthe influence of developments in evolu-tionary theory in biology, and sociobiol-ogy, and the attempts to extend theselines of analysis to explain the evolutionof human patterns of cooperation, co-ordination, and social behavior moregenerally. While originally an intellectualfield dominated by scholars outside ofeconomics, a number of economists havecome to be attracted both by the ques-tions, and some of the analytic ideas.(See e.g., Jack Hirshleifer and Juan Car-los Martinez-Coll 1988.)

My review will describe these develop-ments. However, I will argue that theideas developed to date in evolutionarysociobiology are not adequate to dealwith the questions of most interest toeconomists concerned with long run eco-nomic change, for example the evolutionof technologies and institutions.

The development of evolutionarygame theory has drawn extensively onthese ideas, but probably should be re-garded as a field in its own right, with itsown questions, and methods (H. PeytonYoung 1993; Michihiro Kandori, GeorgeMailath, and Rafael Rob 1993; DanielFriedman 1991). The focus is on re-peated games, and the problem of multi-


ple Nash equilibria that is characteristicof such. The analytic tack that unifies aquite diverse body of writing is to specifyan “evolutionary” process that is opera-tive on the set of employed strategies,and to explore whether or not extantstrategies converge to a steady state and,if so, the characteristics of such an equi-librium.

My review here will be concernedmostly with evolutionary analysis of longrun and continuing economic change,and thus will deal only very selectivelywith evolutionary game theory. How-ever, the notion that there may be multi-ple equilibria, and that an understandingof which one, if any, will be achievedmay require an analysis of “out of equi-librium” behavior (i.e., that equilibriumis “path dependent”) is very germane tothe review presented here.

Recent developments in under-standing of the mathematics of nonlineardynamic systems, and recognition thatmany physical systems display propertiesthat such dynamic models can explainand illuminate, is yet another stimulus toevolutionary theorizing in economics(Ilya Prigogene and Isabelle Stengers1984; David Lane 1993). While some ofthis new analytic understanding has beenemployed in evolutionary game theory,and indeed evolutionary game theorymight be considered a special case ofmodels of complex dynamic systems, twofeatures of the more general formulationare relevant here. First, for the most part(there are exceptions) evolutionary gametheory continues an older tradition ingame theory of thinking of a given finiteset of (basic) strategies, with equilibriumbeing defined in terms of these or mixesof these. In contrast, in the more generalformulation an equilibrium, if there isany such, is seen as emerging out of thedynamic process, and often involves pat-terns of behavior and activities that wereabsent early in the process. The number

and nature of possible equilibria thusoften cannot be specified ex ante. Sec-ond, while concerned with certain regu-larity properties in the time series, writ-ers who identify their work as analysis ofcomplex dynamic systems seem quiteready to believe that the system alwayswill be “out of equilibrium.”

Much of the work on complex dynamicsystems is carried out through computersimulation. The tremendous increase inthe power of computers, and the recentavailability of programming languagesand techniques that greatly facilitatesimulation of complex dynamic systems,however, should be regarded as a factorin its own right that has stimulated thedevelopment of formal evolutionarytheorizing in economics. To recall thequote from Marshall that began this es-say, the complexity of “biological con-ceptions,” in particular evolutionary pro-cesses, no longer poses the same analyticobstacles as was the case in the time ofMarshall—or even twenty years ago.

Earlier I suggested that the appeal ofequilibrium formal theorizing in eco-nomics was much more than a matter ofcomputational feasibility, but reflectedas well notions of aesthetics and ele-gance. But elegance is in the eye of thebeholder. Those working with the newmodels of dynamic complex systemsclearly are developing a sense of aesthet-ics of their own. And appreciation of adifferent kind of elegance seems to bespreading among economists.

All of the developments above havecontributed to the rise of a body of writ-ing by economists and kindred scholarswho are interested in understanding andexplaining aspects or sources of long runeconomic change, and have developedquite explicit and self consciously evolu-tionary models for that purpose. Thesewritings are the focus of this essay. I willbe concerned with evolutionary theoriz-ing that arises out of empirical research,


as contrasted with studies that developevolutionary models or arguments be-cause they are interesting in their ownright, and which bring in empirical casesmainly as examples. Much, if not all, ofthis evolutionary theorizing has been de-veloped by the authors because theyhave felt that “mechanical analogies”simply would not do for their task, andthat “biological conceptions” were moreilluminating. And in contrast with mostearlier writing, these writers have madetheir evolutionary theorizing explicit.

Like Marshall, most of these writers,while drawn to biological conceptions ormetaphors, have resisted simply transfer-ring evolutionary concepts used in biol-ogy to their area of inquiry, but ratherhave tried to analyze the evolutionary dy-namics at work there in its own right.This has not always proved easy. In manycases the processes involved appear tobe, when they are looked at closely,quite complex. Also, there still is littleexperience that can be drawn upon inconstructing an evolutionary theory ger-mane to economic change. The studies Iwill review here are highly varied reflect-ing not only their different subject mat-ter, but also the authors’ particular waysof formulating an “evolutionary” theory.All of the theories considered here areformal theories, in that they have beenexplicitly put forth by their authors as atheory to explain particular phenomena.Some are expressed mathematically;some in words. The distinction that Win-ter and I made between formal and ap-preciative theorizing did not hinge onthe media of exposition, although almostinvariably theory expressed mathemati-cally is formal theory in our terms. Inour terms the hallmark of a formal the-ory is the explicit setting out of a causalaccount, however expressed. A highlyrelevant question, of course, is the logi-cal coherence of the theoretical state-ment. Here, the use of mathematics

would seem to help, but the history ofeconomic thought displays many coher-ent verbal theories, and some incoherentmathematical ones.

The remainder of this article is orga-nized as follows. In the following sectionI draw out the similarities, and differ-ences, between the evolutionary eco-nomic theories I will be examining, andevolutionary theory in biology and socio-biology.

I turn next to the evolutionary theoriesthat are the focus of this essay. SectionIII is concerned with a group of evolu-tionary theories about particular phe-nomena associated with long run eco-nomic change: science, technology,business organization, and law. SectionsIV and V deal with evolutionary theoriesthat treat clusters of coevolving vari-ables, the former with models of eco-nomic growth driven by technical ad-vance, the latter with the coevolution oftechnology and industry structure. Sec-tions VI and VII are concerned with or-ganizations and institutions.

In the concluding section I reflect onthe present state of evolutionary theoriz-ing in economics. I also will attempt tosharpen the discussion of the ways inwhich evolutionary theorizing is differ-ent from neoclassical theorizing, and topropose some criteria that might enableone to evaluate the strengths and weak-nesses of the alternatives. Some econo-mists would argue that that issue oughtto hinge on the quality of the predic-tions, but I will suggest that the issue istoo complex for that.

II. What Are the Characteristics of anEvolutionary Theory?

What are the characteristics of an“evolutionary theory” of economicchange, as contrasted with theories ofeconomic change that employ “mechani-cal analogies”? In what ways are eco-


nomic evolutionary theories similar toevolutionary theories in biology and so-ciobiology, and in what ways different?These are the questions addressed in thissection.

A. Evolutionary Theory as a General Theory

One way to define evolutionary theoryin general would be to start from biol-ogy, where evolutionary theory is bestworked out, and explore where one canfind close analogies to the variables andconcepts of that theory in other areas ofinquiry—in this case economics. How-ever I believe that following this routewould tie the discussion much tooclosely to biology. After all, as Hodgson(1993) has discussed at some length, theterm “evolution” was in wide use longbefore it took on meaning as the name ofa particular theory in biology.

I believe that much of the appeal ofevolutionary language in economics isconnected with the broader use of theterm, as contrasted with its specific usein biology. Further, to start with biologyrisks getting stuck in notions that, whilesalient in biological evolution, seem ir-relevant or wrong-headed when appliedto economics. It seems more fruitful tostart with a general notion of evolution,and then examine applications in specificareas—like biology or economics—asspecial cases.

The general concept of evolutionarytheory that I propose, and employ in thisessay, involves the following elements.The focus of attention is on a variable orset of them that is changing over timeand the theoretical quest is for an under-standing of the dynamic process behindthe observed change; a special casewould be a quest for understanding ofthe current state of a variable or a sys-tem in terms of how it got there. Thetheory proposes that the variable or sys-tem in question is subject to somewhat

random variation or perturbation, andalso that there are mechanisms that sys-tematically winnow on that variation.Much of the predictive or explanatorypower of that theory rests with its speci-fication of the systematic selectionforces. It is presumed that there arestrong inertial tendencies preservingwhat has survived the selection process.However, in many cases there are alsoforces that continue to introduce new va-riety, which is further grist for the selec-tion mill.

All of the evolutionary theories of eco-nomic change I will discuss have thesecharacteristics. They also are central, ofcourse, in evolutionary theory in biology.However, biology makes heavy use ofother concepts that, by and large, are notused in in economics. The fact that sexu-ality and mating play a major role in theevolution of many species is important inbiology but seldom used in economics.The concept of generations is used in bi-ology, but does not apply easily to analy-ses of the evolution of technologies,firms, or institutions.

On the other hand, in some of thetheories considered here the new “vari-ety” that is created as grist for winnow-ing is systematically oriented toward newdepartures that seem appropriate to thecontext. That is, there is a directionallyadaptive aspect to the innovation pro-cess. Also, what entities “learn” in suchprocesses may, in some models, bepassed on to other entities. That is, someof the economic evolutionary theoriesare Lamarkian, a version of evolutionarytheory that has been discredited in biol-ogy. Some emphasize group selection.Other aspects that distinguish economicmodels from biological ones will be de-veloped along the way.

In any case, the proposed general defi-nition of an evolutionary process cer-tainly rules out certain theories ofchange, for example those that are


wholly deterministic. Thus under thisdefinition, as apparently under Mar-shall’s conception, Kepler’s laws ofplanetary motion, together with New-ton’s gravitational theory that explainsthem, would not define an evolutionarytheory. Nor would the standard neoclas-sical theory of economic growth, whichbasically presumes a moving generalequilibrium, be regarded as an evolu-tionary theory. Neither would the execu-tion of a detailed plan for the construc-tion of a building, or any realization of apre-specified blueprint, be consideredan evolutionary process.

On the other hand the definition I amproposing also rules out theories ofchange where all the action is “random,”as certain models in economics that pur-port that within an industry the growthor decline of particular firms is a randomvariable, possibly related to the size ofthe firm at any time, but otherwise notanalyzable (see e.g., Herbert Simon andCharles Bonini 1958). One can tracethrough the random processes built intosuch models and predict the distributionof firm sizes at any time, for examplethat under certain specifications it willasymtotically become log normal. Butunder the definition presented here,these models would not be consideredevolutionary models of economic change.

But revise the building constructionstory as follows. Assume that the originalhouse design is a tentative one, becausethe builder is not exactly sure how toachieve what he or she wants, and thusthe plan initially contains certain ele-ments without any firm commitment tothem, indeed that are there partly bychance. As the building gets constructedthe builder gets a better idea of what thepresent plans imply, and where the origi-nal design is inadequate, and, where con-struction in place permits, revises theplan and the path of construction accord-ingly. Revise the firm growth model as

follows. Assume that the firms differ incertain identifiable characteristics, andgrowth of those with certain ones turnsout to be systematically greater thanthose that lack these. The industrygradually develops a structure in whichonly firms with these characteristics sur-vive.

Both models now contain both randomand systematic elements. Further, inboth the systematic ones act by winnow-ing on the random ones. In the housedesign case, design elements turn out toplease or displease the builder, and areaccepted or rejected accordingly. In theindustry evolution case, the “market” orsomething is selecting on firms that havecertain attributes. A limitation of bothstories is that neither is explicit aboutwhat it is that seems to give advantage.But both give hope that the analystmight be able to find out. Perhaps it is“cost per square foot” or “nicely shapedspaces” or some combination that ex-plains why the builder revises the designas the information comes in. Perhaps it isproduction costs or ability to innovatethat is determining whether firms thriveor fail. Of course the theory has limitedexplanatory power until the question ofselection criteria gets answered. But ifthat question is answered adequately,the theory can explain, and to some ex-tent predict.

The analytic structure of these two ex-amples is reminiscent of that of evolu-tionary theory in biology, without beingclones of it. The latter, however, seemscloser to theory in biology because it re-fers to an actual population of things,while the former does not appear to, atleast at first glance. In biology the use ofthe term evolutionary nowadays is firmlyassociated with analysis of actual popula-tions of things. An embryo, or more gen-erally a living creature, usually is de-scribed as developing, not evolving. Inpart this use of language reflects a predi-


lection discussed earlier—that change“according to a plan” is usually not re-garded as evolutionary. However, it isrecognized widely that many random oc-currences will affect the development ofan embryo or a tree. The prejudiceagainst using the term “evolutionary” todescribe such biological processes stemsfrom the fact that the term has beenpreempted for use in describing anotherclass of biological phenomena. However,is it clear that prejudice should carryover outside of biology?

Consider our house builder, or an in-dividual learning to play chess, or a firmtrying to find a strategy for survival in acompetitive industry. Our house buildercan be regarded as having a number ofplan variants, or perhaps as having oneinitially in mind but being aware thatthere are a set of possible changes thatmight turn out to be desirable. One cansimilarly regard the learning chess playeror the firm. If firms, persons learning toplay chess, or housebuilders, learn fromexperience and winnow or adapt theirplans or strategies or behaviors, is it un-reasonable to think of these as evolving?In reflecting on this, one might recog-nize that the learning, or adaptation, canbe modeled in terms of a change in theprobability distribution of possible ac-tions that entity might take at any time,coming about as a result of feedbackfrom what has been tried, and the conse-quences. These “learning” equationshave basically the same form as theequations that describe the evolution ofpopulations. (See e.g., John Holland etal. 1986.)

There is no great value in extended in-tellectual haggling about the preciseboundaries which demarcate models ofchange that can be called evolutionaryfrom those that should not be. As indi-cated, I choose to use the term “evolu-tionary” to define a class of theories, ormodels, or arguments, that have the fol-

lowing characteristics. First, their pur-pose is to explain the movement ofsomething over time, or to explain whythat something is what it is at a momentin time in terms of how it got there; thatis, the analysis is expressly dynamic. Sec-ond, the explanation involves both ran-dom elements which generate or renewsome variation in the variables in ques-tion, and mechanisms that systematicallywinnow on extant variation. Third, thereare inertial forces that provide continuityof what survives the winnowing.

The variation in the theory can be as-sociated with an actual variety which ex-ists at any time—as a distribution ofgenotypes or phenotypes, or firm poli-cies. Alternately, it may characterize aset of potential values of a variable, onlyone of which is manifest at any time.Thus I would include theories of individ-ual, organizational, or cultural learningand adaptation under my umbrella, ifthey fit other characteristics. Indeed, aswe shall see, a characteristic of many ofthe economic evolutionary theories wewill examine is that individual learning,organizational adaptation, and environ-mental selection of organizations, all aregoing on at the same time.

B. Evolutionary Theory in Biology

As is the case with any active scientificfield, there is far from full agreement onall matters among modern biologists,ethnologists, paleontologists, and otherscientists concerned with biological evo-lution. However, the following sketchcaptures that part of the generallyagreed upon core that is most useful tolay out for our purposes in this essay, aswell as some of the relevant bones ofcontention. (The following draws frommany sources, but especially Richard Le-wontin 1974; Elliott Sober 1984; DavidHull 1988; and Ernst Mayr 1988.)

The theory is concerned with two ac-tual populations as contrasted with po-


tential ones. One is the population ofgenotypes, defined as the genetic inheri-tance of living creatures. The second isthe population of phenotypes, defined interms of a set of variables that happen tobe of interest to the analyst, but whichinclude those that influence the “fitness”of each living creature. These might in-clude physical aspects like size, or sight,behavioral patterns like song, or re-sponses to particular contingencies likesomething that can be eaten and iswithin reach, or a potential mate, or amember of one’s own “group” solicitinghelp.

Phenotypic characteristics are pre-sumed to be influenced by genotypicones, but not uniquely determined bythem. Modern evolutionary theory rec-ognizes that the development of a livingcreature from its origins to its pheno-typic characteristics at any time can beinfluenced by the environment throughwhich it passes—whether there was ade-quate food supply when it was young ornot, or the fact that it lost an eye in anaccident. Modern evolutionary theoryalso recognizes a variety of learning ex-periences which shape the behavior of aphenotype, including how it was taughtby its mother, whether particular behav-iors early in life were rewarded, etc.However, if we hold off for a momentconsidering evolutionary theory that rec-ognizes “culture” as something that canbe transferred across generations, thehallmark of standard biological evolu-tionary theory is that only the genes, notany acquired characteristics or behavior,get passed on across the generations.

The notion of “generations” is basic tobiological evolutionary theory. The phe-notypes get born, live, reproduce (atleast some of them do), and die (in mostspecies ultimately all of them do). Onthe other hand, the genes get carriedover to their offspring, who follow thesame generational life cycle. Thus the

genes provide the continuity of the evo-lutionary system, with the actual livingcreatures acting, from one point of view,as their transporters from generation togeneration. For species that produce thisway, sexuality provides a mechanism forcombining genotypes in a manner thatmay create new ones. Mutations also cre-ate new genotypes. On the other handselection winnows on the genetic varietythrough differential reproduction by(pairs of) phenotypes which augmentsthe relative frequency of the genes ofthe more successful reproducers and di-minishes that of the less.

In the generally held interpretation ofthis theory (there are other or morecomplex interpretations as well), selec-tion operates directly on the phenotypes.It is they, not their genes per se, that aremore or less fit. To repeat what wasstressed above, phenotypes are notuniquely determined by genotypes.However the theory assumes a strongenough relationship between the two sothat systematic selection on phenotypesresults in systematic selection on geno-types.

There are several controversial, or atleast open, aspects of this theory that aregermane to our discussion here. Foreconomists perhaps the most interestingquestion is whether, and if so in whatsense, evolution can be understood to“optimize” fitness.

The optimization notion here clearlyhas roots in Herbert Spencer’s notions(1887) of “survival of the fittest,” and theimplicit context is one in which competi-tion among members of a population issufficiently fierce that only the “fittest”survive. In recent years theorists haveformalized this idea as a game for sur-vival, and developed the concept of an“evolutionarily stable strategy” as theequilibrium solution to that game. (Seee.g., Maynard Smith 1982.) The conceptof “strategy” in these models is broad


enough to encompass any phenotypiccharacteristic that matters for survival,and the strategies that survive in equilib-rium are those that can best (at leastsmall numbers of) other pre-specifiedstrategies in the survival game.

In what sense is what survives opti-mal? The semantic correspondance be-tween between survival and optimalityseems most straightforward if the“game” is about different kinds of strate-gies passively competing for the sameenvironmental “niche,” and one typewins out. The winner might be under-stood as the most efficient forager, orsomething like that. However, things getsomewhat more complicated if the gameisn’t simply about passive competitionfor a niche, but includes some strategiesthat involve attacking competitors ofother sorts. then both efficient foragingand fighting prowess count in definingoptimal, if that term is to be used tocharacterize what survives.

Even in such simple contexts, thereare some subtleties that qualify the asso-ciation of what “survives” with “optimal.”For one thing, how a strategy fares in aseries of plays of a game depends on themix of strategies with which it competes.Thus what survives depends on what elseis competing in the game. More, if thenumber of individuals associated withany particular strategy is finite, the veryprocess of competition may eliminatealong the path to an equilibrium strate-gies that would be in a stable equilib-rium set as calculated ex ante. That is,the equilibrium may be strongly path de-pendent. (See e.g., Hirshleifer andMartinez-Coll 1988.)

Other complications come into viewwhen one recognizes that “strategies”may have many aspects, and these mayinteract strongly in determining ability tocompete and survive in a given environ-ment. Thus being an effective predatorrequires a package of attributes, ability

to get at the leaves on tall trees a differ-ent package. But then, whether a “gene”or an aspect of a strategy enhances sur-vival or not may be strongly dependentupon the other genes or aspects of strat-egy. And a “mutation” that may be lethalin one species or strategy, may be help-ful in another. Thus if strategies them-selves evolve, they likely do so in astrongly path dependent way.

There may be important interactionsacross coexisting phenotypes—strategies.The existence of giraffes provides oppor-tunites for large strong predators. Butthe number of the latter that can survivein equilibrium may depend on the num-ber of the former, and vice versa. In turnthe ecological equilibrium depends onthe number of trees and the leaves thatare available to giraffes. The emergenceof an insect whose caterpillars feed onleaves of the tall trees may bring downthe whole ecosystem.

Also, a number of students of biologi-cal evolution have argued that the selec-tion environment almost never is con-stant (see e.g., Gould 1980, 1985). Theinsect population may get large, andthen itself collapse after it has dimin-ished the population of live trees. If theselection environment is not a constant,the phenotypes extant today may bestrongly shaped by those that survived ina possibly very different environmentsome time ago (say giraffes that can eatthe leaves of low bushes that the cater-pillars do not like), and the offspringthey had, as well as recent winnowing onthe group extant yesterday. Again, theequilibrium is strongly path dependent,and today’s “optimum” may be very localand likely poor stuff compared to whatmight have been.

Gould, among others, says he readsthe evidence as indicating that the selec-tion environment not only changes, buton many occasions is relatively lax. In aloose selection environment different


phenotypes may grow in number morerapidly than others largely due to a com-bination of breeding capability and luck,rather than any special capabilities for“survival” in the environment in ques-tion. The same authors argue that the ex-tinction of particular phenotypes usuallyis the result of catastrophes that hap-pened to hit clusters of them, ratherthan the result of losing a competitionwith other phenotypes.

One also can ask what meaning thereis to the optimality concept in a contextin which mutation continues to go on,and some of the mutations enhance fit-ness, at least in prevailing environments.Modern evolutionary biology is not sim-ply about selection pressures on extantphenotypes, but also about changes thatappear from time to time in species, andalso about the origins of new species.These latter phenomena would seem torequire analysis of evolutionary pro-cesses that involve not only out-of-equi-librium behavior, but also the emergenceof novelty. (See e.g., Walter Fontana andLeo Buss 1992; and Lane 1993.)

C. Sociobiology

As indicated, animal behavior has, fora long time, been a “phenotypic” charac-teristic of interest to evolutionary theo-rists. That behavior often involves, in anessential way, modes of interaction withfellow members of one’s species. Overthe last thirty years an important subdis-cipline has grown up concerned with ex-actly these kinds of social behavior pat-terns. Much of this has been concernedwith nonprimate animals—insect colo-nies, bird families and flocks, etc. A size-able portion of it has, however, beenconcerned with humans. The part of thesociobiology literature concerned withnonhumans recognizes that learned be-havior can be passed down from genera-tion to generation, but in general haspresumed, first, that the particular capa-

bilities to learn and to transmit to off-spring are tied to genes, and second, thatthe “learning” does not progress fromgeneration to generation. To the extentthat these behaviors enhance fitness,there is selection on the genes that facili-tate them, according to the argumentssketched above. But learned behaviordoes not follow a cross generational pathof its own.

The early work by Edward Wilson onthe biological bases of human social be-havior carried over basically this model.However, in subsequent writings byCharles Lumsden and Edward Wilson(1981), and by other scholars interestedin extending evolutionary theory in biol-ogy so as to be able to treat human cul-ture, prominently Luigi Cavalli-Sforzaand Marcus Feldman (1981), RobertBoyd and Peter Richerson (1985), andWilliam Durham (1991), human culturewas recognized as something that couldbe modified, and improved, from genera-tion to generation, and which had itsown rules of transmission. These lattermodels all do presume a basic geneticbiological capacity of humans for the de-velopment and transmission of culture.But beyond that these models treat theconnections between the evolution ofhuman behavior and culture, and geneticevolution, as something far more com-plex than that assumed in the models ofinsect and bird societies.

There are a number of important dif-ferences among these models. ThusLumsden and Wilson, and Cavalli-Sforzaand Feldman, tend to treat elements ofculture as something that directly deter-mine what people do and how effectivelythey do it, while Boyd and Richerson,and especially Durham, treat culture asprominently involving understandingsand values that, like genes, influence be-havior or capabilities, but do not directlydetermine these. Perhaps the most im-portant difference among these models


is the extent to which biology is seen asconstraining and molding culture beyondthe preconditions that all of these theo-ries recognize. Put in the terms coinedby Wilson, there are sharp disagree-ments regarding how long the “leash,”and the extent to which evolution of cul-ture itself has significantly extended thelength of that “leash.” Here Lumsdenand Wilson are far closer to the animalsociobiology models than the otherauthors.

For the purposes of the discussionhere, I want to focus on certain com-monalities of the theories in this litera-ture, which, I believe limit their range ofapplicability. In particular, all of themuse as their examples relatively simplepractices or artifacts or ideas or normswhich can easily be thought of as beingtransmitted from person to person. Eachtries to break down “culture” into small“gene like” subunits, which are assignedterms like meme, or culturgen. The sim-ple technology-artifacts and beliefs em-ployed as examples are a far distancefrom complex technologies like those as-sociated with making semiconductors oraircraft, or scientific theories like that ofbiological evolution itself, or systems likepatent law. While teachers and opinionleaders are admitted as “transmitters” or“influencers,” there is nothing in thesestudies like universities, or scientific so-cieties. Various forms of human organi-zation are discussed, but there is notreatment of organizations like industrialR & D laboratories, or business firmsmore generally, or elections or legisla-tures, or courts.

Most of the analyses clearly recognizethat in principle an element of culturecan spread for reasons that have little todo with enhancing individual biologicalfitness, in any straightforward manner,and some stress that as a general propo-sition. Boyd and Richerson even presenta model example in which the profes-

sional life of, say, a teacher, or a memberof the clergy, is assumed to carry attrac-tions of its own, but those who follow thecalling actually have a smaller number ofoffspring than those who do not. Mem-bership in the profession as a whole issustained intergenerationally by new re-cruits. However, none of these analysesattempts to come to grips with the pathsof cumulative evolution taken by culturalstructures like science, technology, thelaw, standard forms of business organiza-tion, and the like, which clearly havebeen drawn and shaped by particularvalue systems, and particular mecha-nisms for inducing and winnowingchange.

While important and interesting in itsown right, the body of writing on cul-tural evolution that traces its origins tobiological evolutionary theory, and thenmakes a sharp break, has not as yet triedto come to grips with the dynamics ofchange in modern industrial societies. Todo so requires, it would seem, buildinginto evolutionary analysis much more ofthe institutional complexity of modernsocieties than the literature above hashazarded thus far. Boyd and Richersonrecognize this explicitly when they re-mark,

Understanding the institutional complexity ofmodern societies will require the mating ofmicro-level theory like the one we have de-veloped here with the more aggregated oneof Nelson and Winter. (Boyd and Richerson1985, p. 296)

This is just what the various studies weshall consider in the following sectionshave tried to do, if with varying levels ofsuccess.

III. The Evolution of Particular Aspectsof Culture

There are three key differences be-tween the evolutionary theories I con-sider here, and in the following sections,


and those in sociobiology. First, thereare no ties whatsoever between the cul-tural selection criteria and processes andbiological fitness. Any coevolution inthese theories is not between memes andgenes, but between various elements ofculture.

Second, the authors of the theoriesconsidered here are interested in ex-plaining how and why a particular aspectof “’culture” changed over time the wayit did. Because their explanation is interms of the workings of an evolutionaryprocess, this forces them to identifysome particular characteristic of meritand selection mechanisms enforcing itthat favors certain variants over others,or which reinforces certain behaviors orinclinations and damps others. The theo-rists of biological and cultural coevolu-tion discussed above have coined theterm “cultural fitness,” but seldom havegot around to identifying it in particularcases where biological fitness is not animportant variable at stake. Third, evolu-tionary theorists, coming from sociobiol-ogy, have by and large assumed selectionmechanisms are individualistic, transmis-sion mechanisms are person to person,and that “’memes” like genes are carriedby individuals. Yet these perceptionsseem quite inadequate for analysis ofhow science or modern technologyevolves, or forms of business organiza-tion, or law.

This section will be concerned withevolutionary theories of just these ele-ments of culture, all of them major andobviously intertwined aspects of the pro-cess of long run economic change. Thetheories discussed in this section largelyrepress the intertwining. Each theorydeals with just one of these variableswhich is viewed as proceeding on itsown, as it were. In the following two sec-tions we shall consider theories in whichinterdependence and coevolution arerecognized.

The collection of theories discussed inthis section are all qualitative, and ex-pressed verbally, as contrasted with be-ing laid out mathematically. All are for-mal theories, in the sense of being putforth as self-conscious abstractions aboutwhat is driving the dynamics of the vari-ables in question. However, none is de-veloped mathematically. And some seemmuch better posed analytically than oth-ers, in that the logic seems tighter.

I also want to stress that each of thebodies of evolutionary theorizing dis-cussed in this section is very large. Mytreatment of each, therefore, must behighly selective. My particular selectionis designed not so much to be repre-sentative of the literatures involved, as tobring out some analytic issues about evo-lutionary theorizing.

A. Science

The proposition that science “evolves”has been around for some time, andthere has been and continues to be alively discussion about just how that evo-lutionary process works. For the mostpart the various theories put forth do sat-isfy my definition of what qualifies as anevolutionary theory. (For overviews seeHenry Plotkin 1982; and David Hull1988.)

Of recent writers in this vein, DonaldCampbell (1960, 1974) probably is themost cited. Using Campbell’s term, thedevelopment of new scientific hypothe-ses, or theories, is to some extent“blind,” in that their originators cannotknow for sure how they will fare whenthey are first put forth. Thus new scien-tific theories are like “mutations” in thatsome will succeed and be incorporatedinto the body of science, perhaps replac-ing older theories, or correcting them insome respects, or adding to them, andothers will not succeed. Campbell relieslargely on the ideas of Karl Popper(1968) for his “selection mechanism.”


Under Popper’s argument scientifictheories never can be proved true, butthey can be falsified. New theories thatsolve scientific problems and are not fal-sified are added to the body of science.That is, employed and “not falsified” isthe characterization of fitness in this the-ory of science. For the most part Camp-bell treats science as a relatively unifiedbody of doctrine, and his language im-plies a scientific community togethersearching after truth, that is of collectiveevolutionary learning. On the otherhand, his theory is compatible with thenotion of individual scientists puttingforth their particular theories in hope ofwinning a Nobel prize. A good case canbe made that both images of science—cooperative and competitive—are partlycorrect (see Hull 1988).

In any case the theory leaves open twoquestions. The first is what determineswhich theories are to be rigorouslytested, and what is the standing of theo-ries that have not been. “Theories” thathave not (as of yet) been subject to rigor-ous testing do not necessarily have thesame standing. Some may never bebrought to a serious test simply becausethey are regarded as irrelevant, or ontheir face absurd. Others may fit so wellwith prevailing understanding that theyare absorbed without direct testing. Thesecond question is what falsificationmeans; in many cases the conclusions ofa test may be ambiguous, or there maybe reason to question the way it was run,or whether it was appropriate. Often atheory which seems to fail a test can bepatched up with a well crafted modifica-tion or amendment. These issues openthe door to a much more complicatedtheory of the evolution of science than atleast the simple interpretation of Camp-bell’s.

The “social constructionists” recognizeand revel in these complications (seee.g., Bruno Latour 1986). They propose

that very few theories, or scientific argu-ments more generally, are ever com-pletely falsified, or even put to a test thatall would regard ex ante as conclusive.Thus scientific opinion is what mattersand, in a context where different indi-viduals and groups have different opin-ions, what is considered scientific factand is published in reputed journals,taught to graduate students, etc., islargely a matter of scientific politics.

Thomas Kuhn (1970) presents a viewsomewhat between Campbell and the so-cial constructionists. On the one handKuhn proposes that most “normal sci-ence” proceeds in almost unthinking ac-ceptance of prevailing theory, and thatthere is strong built-in disbelief of re-sults that challenge that theory. On theother hand, also central to Kuhn’s theoryof the evolution of science is that unan-swered questions or anomalies tend toaccumulate and, as they do, questions in-creasingly are asked about the adequacyof prevailing theory. A standard responseof the scientific community is to proposemodest modifications or additions to pre-vailing theory. However these may notsucceed or the developing theoreticalstructure may come to be seen as rococo.The seeds then are planted for a scien-tific revolution.

Neither Campbell nor Kuhn (in theirearlier versions) address the issue ofcompeting theories. However, such com-petition is the heart of scientific revolu-tions. Imre Lakatos (1970) proposes thatbroad theories should be regarded as de-fining research programs. These pro-grams can be judged by the communityas proceeding effectively—that is asmaking good progress—or as more orless stuck. Lakatos proposes that thereare almost always competing theoriesaround. The one that defines the moreeffective research program tends to winout. But again, one can ask what defines“effective.” A particular theory almost al-


ways points to a number of predicted im-plications, and exploring these defines avariety of puzzles and problems andtasks. A research program may be goodin dealing with some of these, and not soeffective on others. What counts?

Note that several different “theories”of the evolution of science have been de-scribed above. Some are in conflict. Inparticular the social constructionistswould seem at odds with scholars, likeCampbell, who believe that new scien-tific hypotheses, or at least those takenseriously, are subject to test, and thatenough of the tests are sufficiently ob-jective and unambiguous to monitor theenterprise. (This also clearly is Hull’sview.) Some of the theories are compat-ible, and can be joined. Thus Kuhnmight be regarded as providing an evolu-tionary theory of science within a givenresearch program, and Lakatos a theorywhich explains selection among compet-ing programs. Regarding what criteriaare used to weigh program effectiveness,later in this essay I shall suggest that partof the answer may reside in the connec-tions between science and technology.

Does science make progress? Whilethe social constructionists seem stronglyreserved about this, I think it fair to saythat most of the theorists who proposethat science evolves believe that the pro-cess does generate progress, at leastalong the lines of research pursued.(This clearly is Hull’s view.) While occa-sionally we delude ourselves that wehave understood something when we donot, and often the going toward betterunderstanding is hard, by and largethrough science we have come to knowmore and more about nature and how itworks. Or at least this is the flavor ofmost of this body of theorizing.

B. Technology

A number of analysts have proposedthat technology evolves. The analyses of

Nathan Rosenberg (1976, 1982), Christo-pher Freeman (1987), George Basalla(1988), Joel Mokyr (1990), Nelson andWinter (1977), Dosi (1988), and WalterVincenti (1990) are strikingly similar inmany respects. To keep the discussionbelow simple, I will follow the discussionof Vincenti.

In Vincenti’s theory, the community oftechnologists at any time faces a numberof problems, challenges, and opportuni-ties. He draws most of his examples fromaircraft technology. Thus, in a new paper(Vincenti 1994) he observes that in thelate 1920s and early 1930s, aircraft de-signers knew well that the standard pat-tern of hooking wheels to fuselage orwings could be improved upon, given thehigher speeds planes were then capableof with the new body and wing designsand more powerful engines that hadcome into existence. They were aware ofseveral different possibilities for incorpo-rating wheels into a more streamlineddesign. Vincenti argues that trials ofthese different alternatives were, in thesame sense put forth by Campbell, some-what blind. This is not to say that theengineers thinking about and experi-menting with solutions were ignoranteither of the technical constraints andpossibilities or of what was required of asuccessful design. Rather, his proposi-tion is that, while professional knowl-edge and appreciation of the goalsgreatly focused efforts at solution, therestill were a number of different possibili-ties, and engineers were uncertain aboutwhich would prove best, and disagreedamong themselves as to where to placebets.

This kind of uncertainty, together withthe proposition that uncertainty is re-solved only through ex post competiton,is the hallmark of evolutionary theories.In this case it turned out that having thewheel be retractable solved the problembetter than did the other alternatives ex-


plored at that time. Thus, “fitness” hereis defined in terms of solving particulartechnological problems better.

One might propose that identificationof this criterion only pushes the analyticproblem back a stage. What determineswhether one solution is better than an-other? At times Vincenti writes as if thecriterion were innate in the technologi-cal problem, or determined by consensusof a technological community who are,like Campbell’s community of scientists,cooperatively involved in advancing theart.

However, Vincenti also recognizes thatthe aircraft designers are largely em-ployed in a number of competing aircraftcompanies, whose profitability may beaffected by the relative quality and costof the aircraft designs they are employ-ing, compared with those employed bytheir competitors. But then what is bet-ter or worse in a problem solution is de-termined at least partially by the “mar-ket,” the properties of an aircraftcustomers are willing to pay for, thecosts associated with different design so-lutions, etc. In the case of aircraft, themilitary is an important customer, aswell as the airlines. Thus the evolution ofaircraft at least partially reflects militarydemands and budgets, as well as civilian.

As with the case of science, someauthors dispute that the evolution oftechnology follows a path that might beconsidered as “progress,” or even thatthere are any objective criteria for tech-nological fitness. The book by Wiebe Bi-jker, Thomas Hughes, and Trevor Pinch(1989) surveys various theories of “socialconstruction” of technology. MichaelTushman and Lori Rosenkopf (1992) de-velop a more nuanced view of social de-terminism, but one which also implicitlydenies the importance of economic effi-ciency, save as a gross screen. On theother hand, evolutionary theorists of thedevelopment of technology of the Vin-

centi camp believe strongly that there istechnological progress, and ask thereader who doubts to compare modernaircraft with those of fifty years ago,modern pharmaceuticals with thoseavailable before World War II, etc.

In recent years a particular insight orargument has somewhat complicated thisdiscussion. While those that profess thatscience “progresses” generally seem tohave in mind a unitary concept of “truth”toward which science is going, recentscholarship on the evolution of technol-ogy has proposed that there may be anumber of different evolutionary tracksthat go in quite different directions, andthat movement down one may blockmovement down another. Thus the rapidevolution of gasoline-powered automo-biles may have improved these, but atthe same time may have scotched prog-ress toward battery-powered cars. Wewill pick up this discussion in Section V.

C. Business Organization

Alfred Chandler’s research (1962,1990) has been concerned with under-standing how the complex structures thatcharacterize modern multiproduct firmscame into existence. For our purposeshis story is especially interesting, in thatit is a story of coevolution. The coevolu-tion is not of genes and memes, but oftechnology and business organizations.He argues that a variety of technologicaldevelopments occurred during the midand late 19th century which opened upthe possibility for business firms to behighly productive and profitable if theycould organize to operate at large scalesof output, and with a relatively wide ifconnected range of products. He de-scribes various organizational innova-tions that were tried, and while hiscentral focus is on those that “suc-ceeded,” it is clear from his account thatnot all did.

Arguing in a manner similar to Vin-


centi, Chandler’s “fitness criterion” isthat the new organizational form solvedan organizational problem. Presumablythe solution to that problem enabled afirm to operate at lower costs, or withgreater scale and scope, in either case,with greater profitability. Like Campbelland Vincenti, Chandler clearly sees acommunity, in this case of managers. Buthe also sees companies competing witheach other. His argument is that compa-nies which found and adopted efficientmanagerial styles and structural formsearly won out over their competitors whodid not, or who lagged in doing so.Oliver Williamson (1985), drawing fromChandler, but putting forth a much moreexplicit formal theory, proposes that arelatively sharp “fitness” criterion deter-mined which organizational forms sur-vived and which ones did not—economicefficiency.

Chandler’s and Williamson’s accountsof the development of the large multidi-mensional corporation stress the need oftop firm managers concerned with mar-ket defined efficiency somehow to de-centralize and yet still control large anddiversified bureaucracies. Marxianshighlight a different aspect of the or-ganizational forms that evolved—thatthey sharply reduced the importance ofworkers with special skills, and henceshifted power toward capital. Neil Flig-stein (1990) presents a still differentview on corporate fitness, which empha-sizes responsiveness to changed legal re-gimes, public policies, and the climate ofpolitical opinion more generally towardwhat corporate action and form oughtto be.

As with the case of technology, somerecent writing has proposed that thepath that has taken us to the modernlarge hierarchically organized corpora-tion is one we did not have to take, andthat in fact better paths existed. We willpick up this theme later.

D. Law

The final example treated in this sec-tion is the body of theory that proposesthat the law evolves. Donald Elliot(1985) has written a rich survey of thevarious evolutionary theories about law. Ifocus here on only a small portion of thatintellectual tradition.

In particular I will be concerned withthe body of theorizing, put forth byscholars such as Harold Demsetz (1967)and William Landes and Richard Posner(1987), that the common law evolves indirections that make it economically effi-cient. While different authors in this tra-dition have proposed different mecha-nisms, in all the decisions to litigateprovide the force that gets the law tochange. In some versions it is arguedthat litigation is more prevalent whenthe law is “inefficient” than when it isefficient, because in the latter cases con-flicts are more likely to be settled out ofcourt without any change in the law, al-though the reasoning behind that propo-sition is not clear in all accounts. Insome versions judges (juries) are in-clined to decide cases that do arise inways that are consistent with economicefficiency, and those judgments in turnmodify the common law in that direc-tion. In other versions no such inclina-tion is assumed, but rather cases willcontinue to be litigated until an “effi-cient” judgment is made, at which time itwill become precedent and litigation willdiminish.

Criticizing this simple view of legalevolution, Robert Cooter and DanielRubenfeld (1989) emphasize the com-plex nature of legal disputes and theirsettlement, involving the actions ind-viduals take that may risk suit, decisionsof potential plaintiffs to assert a legalcomplaint, bargaining regarding out-of-court settlements, and the proceedingsof cases that actually get decided in


court. They express skepticism aboutwhether there are any strong forces lead-ing to efficiency, and argue that, if thereare any such strong forces, they must bedue to the inclinations of judges. Theyare skeptical of this too, citing other le-gal values—like fairness—and also point-ing to the fact that judges may have theirown interests.

P. Ruben and M. Bailey (1992) re-cently have proposed an interesting vari-ant on this theme. They note that law-yers have a strong financial interest inthe shape of the law, and in particularbenefit when the law forces litigation.They go on to propose that the recentshift of legal precedent toward more fa-vorable reception of consumer suits re-garding products which cause them harmis, largely, the self-motivated work oflawyers.

* * *Note that the theories discussed above

are similar in certain respects, but differin others. They are similar in that theyall are concerned with a particular aspectof culture, and focus on its evolution.They are similar in proposing that theprocesses that generate new cultural ele-ments or modify old ones are to someextent blind, although the details ofthese mechanisms differ from case tocase, and in some the mutation or inno-vation mechanisms have strongly di-rected elements as well as random. How-ever in each of these theories the“selection mechanism” provides a largeshare of the explanatory power. That is,the power of these theories depends ontheir ability to specify “fitness” plausibly.

Both neoclassical economists andeconomists inclined to evolutionary theo-rizing are prone to look to a market or amarket analogue as the mechanismwhich defines what will “sell,” and to“profit” or its analogue as the reward toactors that meet the market test. Thetheories above clearly differ in the extent

to which they can be forced into thatmold.

There certainly is no real “market” outthere in Campbell’s or Kuhn’s or Laka-tos’ theory of science as an evolutionaryprocess, save for the metaphorical “mar-ket of scientific judgment.” In the casesof technology and the organization of en-terprises, a moderately persuasive casecan be made that, in many sectors atleast, real, not metaphorical, marketshave a powerful influence on what is“fit” and what is not, and that profit is animportant measure of fitness. However,as we have seen there are dissentors,mainly from outside economics. One im-portant issue is the extent to which com-petition provides a sharp fitness test insectors where markets are operative. If itdoes not there is room for a variety ofnonmarket forces to influence what “sur-vives.” Also, there are serious questionsabout the range of sectors—kinds oftechnologies and organizations—wheremarkets are strongly operative. In thecase of military or medical technologies,or military bases or hospitals, it can beargued that market forces are weak, andthat the “selection environment” is de-termined largely by professional judg-ments, and by political processes thatregulate how much professionals in thesector have to spend. The analytic prob-lem, then, is to identify how these forcesdefine “fitness.”

The dispute about what determineshow the law evolves highlights thesekinds of questions. Clearly there is noreal “market” out there, but one set ofauthors argues that market valuation ofprevailing law and its alternatives doesinfluence what the law becomes, andstrongly. Other authors are not so surethat “efficiency” in an economic senseguides the evolution of the law so muchas ideology, or “interests,” or “power.”

One could take a position that it de-pends, with sometimes one influence


prevailing, and sometimes another. How-ever, in the absence of ability to explainor predict what influence will dominatein particular cases, while evolutionarytheory may provide a useful language forhistorical discussion, the theory has littlepredictive power, and its explanationsare at least partially ad hoc.

This would seem to be a big strikeagainst an evolutionary theory of the law,or at least one that is this complex. Onthe other hand one can argue that theillumination of the complex contingentdependent process by which the lawevolves is a strength of the theory. Suchan analysis reveals the apparent strongpredictive power of a simpler theory—inthis case that the law always adjusts sothat it is maximally efficient—to be fool’sgold. I take it that this is Gould’s argu-ment against the proposition that evolu-tion optimizes biological fitness.

However, this kind of understandingof complexity that a good and well-posedevolutionary theory may yield needs tobe distinguished sharply from weak-nesses in prediction and explanation thatstem from the fact that a theory is notcoherent. Thus the argument that thelaw evolves so as to be efficient, for ex-ample, is an assertion that may or maynot be empirically correct, but whichoriginally was presented with no coher-ent evolutionary theory behind it. Theproposition that litigation stops if andonly if the law is efficient may providepart of an evolutionary theoretic basisfor such an analytic argument, but thatproposition needs more justification thanit often is given, and the general argu-ment almost surely needs some other as-sumptions as well.

In my judgment virtually all of thetheories described in this section couldbenefit from a closer scrutiny and morecareful development of their logicalstructure. I deliberately have not tiedthe term “formal theory” to the expres-

sion of a theory in mathematical orquasi-mathematical terms. Particularly ifthe logic is relatively simple, expressionin careful natural language can be rigor-ous. But when the logic becomes com-plex, the advantages of mathematical for-mulation can be substantial.

Once one is confident that the theoryput forth is logically coherent, the cen-tral question remains as to whether onefinds the theory plausible, given whatone knows about the facts of the matter.But evolutionary theories are no differ-ent in this regard than any other kind.

IV. Evolutionary Models of EconomicGrowth Fuelled by Technical Advance

The body of evolutionary theorizingconsidered in this section differs fromthat discussed above in at least three re-spects. First, the theorizing is more com-plex in the sense that it involves a num-ber of different variables, and the focusis on their coevolution. Second, the the-ory is expressed mathematically; in somecases the logical connections are devel-oped as theorems, while in others theyare explored through simulation meth-ods.

Third, the evolutionary theories pre-sented here have been put forth by theirauthors as express alternatives to anothertheory—in this case neoclassical growththeory—which they regard as too “me-chanical.” In this sense they represent adeliberate attempt on the part of theauthors to move toward Marshall’sMecca.

While concerned with economic dy-namics, the kind of analysis contained inneoclassical growth theory almost cer-tainly was among the kind Marshall hadin mind when he referred to theoriesbased on mechanical concepts of equilib-rium. Within that theory, economicgrowth is viewed as the moving equilib-rium of a market economy, in which


technical advance is continuously in-creasing the productivity of inputs, andthe capital stock growing relative to la-bor inputs. These two phenomena to-gether provide the explanation for theincrease in labor productivity and percapita income that are the standardmeasures of growth. Together, in theway the theory is put together, they ex-plain the rise in real wages that has char-acterized economic growth.

Technical advance is an essential ele-ment of the neoclassical account. Thelast few years have seen a number of in-teresting proposals to amend the simpleneoclassical growth model so as to high-light that technical advance is to a con-siderable degree endogenous. (For re-views see Paul Romer 1991 and BartVerspagen 1992.) However these “new”neoclassical models are “mechanical” inthe same sense as are the old ones. Theydo not address the problems with neo-classical growth theory felt by theauthors of evolutionary alternatives.

In particular, as we have noted, virtu-ally all serious scholars of technical ad-vance have stressed the uncertainty, thedifferences of opinion among experts,the surprises, that mark the process. Me-chanical analogies involving a movingequilibrium in which the actors alwaysbehave “as if” they knew what they weredoing seem quite inappropriate. Mostknowledgeable scholars agree with Vin-centi that the process must be under-stood as an evolutionary one. The chal-lenge faced by the authors considered inthis section has been to devise a theoryof growth in which technical advanceand capital formation together drivegrowth, as in neoclassical growth theory,and which is capable of explaining theobserved macroeconomic patterns, buton the basis of an evolutionary theory oftechnical change rather than one thatpresumes continuing equilibrium.

Without any exception I know about,

the evolutionary theories of economicgrowth that have been developed alldraw inspiration from Joseph Schumpe-ter’s Capitalism, Socialism and Democ-racy (1976, first published 1942). In thatwork, Schumpeter developed a theory ofendogenous technological advance, re-sulting from the investments made bybusiness firms to best or stay up withtheir rivals. The earliest class of formalevolutionary growth models based onthese ideas was developed by Winter andmyself (1974), and because it has pro-vided much of the base for subsequentwork, I shall concentrate on it. HoweverI also will consider variants or extensionsthat have been developed by others.

In these models, firms are the key ac-tors, not individual human beings. Ofcourse (implicitly) firms must providesufficient inducements to attract andhold the individuals that staff them. Butwithin these models, individuals areviewed as interchangeable and their ac-tions determined by the firms they arein.

In turn, the firms in these models are,from one point of view, the entities thatare more or less fit, in this case more orless profitable. But from another point ofview firms can be regarded as merely theincubators and carriers of “technologies”and other practices that determine “whatthey do” and “how productively” in par-ticular circumstances. Winter and I haveused the term “routines” to denotethese. The concept of routines is analyti-cally similar to the genes in biologicaltheory, or the memes or culturgens insociobiology.

The term “routine” connotes, deliber-ately, behavior that is conducted withoutmuch explicit thinking about it, as habitsor customs. On the other hand, withinthese models routines can be understoodas the behaviors deemed appropriate andeffective in the settings where they areinvoked. Indeed they are the product of


processes that involve profit-orientedlearning and selection. Metaphorically,the routines employed by a firm at anytime can be regarded as the best it“knows and can do.” To employ them isrational in that sense, even though thefirm did not go through any attempt tocompare its prevailing routines with allpossible alternative ones. Whether thattranslates into “optimizing” behavior de-pends on what one means by that term.(For a fine discussion of this issue in bi-ology and in economics, see Paul Schoe-maker 1991.)

These models generally involve threedifferent kinds of firm “routines.” First,there are those that might be called“standard operating procedures,” thosethat determine how and how much afirm produces under various circum-stances, given its capital stock and otherconstraints on its actions that are fixed inthe short run. Prominent among theseare technologies. Second, there are rou-tines that determine the investment be-havior of the firm, the equations thatgovern its growth or decline (measuredin terms of its capital stock) as a functionof its profits, and perhaps other vari-ables. Third, the deliberative processesof the firm, those that involve searchingfor better ways of doing things, also areviewed as guided by routines. While inprinciple within these models search be-havior could be focused on any one ofthe firms prevailing routines—its tech-nologies, or other standard operatingprocedures, its investment rule, or evenits prevailing search procedures—inpractice, in all of them search is assumedto be oriented to uncover new produc-tion techniques or to improve prevailingones. Winter and I have found it conven-ient to call such search R & D. Otherauthors of similar models have invokedthe term “learning” to describe analo-gous “improvement” processes.

Firm search processes provide the

source of differential fitness; firmswhose R & D turn up better technolo-gies will earn profits and grow relative totheir competitors. But R & D also tendsto bind firms together as a communitybecause in these models a firm’s R & Dpartly attends to what its competitors aredoing, and profitable innovations are,with a lag, imitated by other firms in theindustry.

The firm, or rather the collection offirms in the industry, perhaps involvingnew firms coming into the industry andold ones exiting, is viewed as operatingwithin an exogenously determined envi-ronment. The profitability of any firm isdetermined by what it is doing, and whatits competitors do, given the environ-ment. Generally the environment can beinterpreted as a “market,” or set of mar-kets.

Note that in the theory that has beensketched above, just as routines areanalogous to genes, firms are analogousto phenotypes, or particular organisms,in biological evolutionary theory, butthere are profound differences. First,firms do not have a natural life span, andnot all ultimately die. Neither can theybe regarded as having a natural size.Some may be big, some small. Thus inassessing the relative importance of aparticular routine in the industry mix, oranalyzing whether it is expanding or con-tracting in relative use, it is not sufficientto “count” the firms employing it. Onemust consider their size, or whether theyare growing or contracting. Second, un-like phenotypes (living organisms) thatare stuck with their genes, firms are notstuck with their routines. Indeed theyhave built in mechanisms for changingthem.

The logic of these models defines a dy-namic stochastic system. It can be mod-eled as a complex Markov process. Astandard iteration can be described asfollows. At the existing moment of time


all firms can be characterized by theircapital stocks and prevailing routines.Decision rules keyed to market condi-tions look to those conditions’ “last pe-riod.” Inputs employed and outputsproduced by all firms then are deter-mined. The market then determinesprices. Given the technology and otherroutines used by each firm, each firm’sprofitability then is determined, andthe investment rule then determineshow much each firm expands or con-tracts. Search routines focus on one oranother aspect of the firm’s behaviorand capabilities, and (stochastically)come up with proposed modificationswhich may or may not be adopted. Thesystem is now ready for next period’s it-eration.

The theory described above can beevaluated on a number of counts. One iswhether the view of behavior it contains,in abstract form, is appealing given whatit purports to analyze. The individualsand organizations in these models act, ashumans do in the models of sociobiology,on the basis of habits or customs or be-liefs; in the Nelson-Winter model allthese define routines. While firm rou-tines can be regarded as the result of alearning process, the implicit “rational-ity” in these models certainly is a“bounded” one, in the sense of Simon(1947) and James March and Simon(1958). As we shall see, it is quite possi-ble to build a certain amount of foresightinto the actors of an evolutionary theory.However if one wants a model in whichit is presumed that the actors largely un-derstand the details of the context inwhich they are operating and competing,save for the truly stochastic elements,and are able to choose their best actionin the light of this full knowledge, onemight as well use a full-blown neoclassi-cal rational choice model. This of courseis what is done in the new neoclassicalgrowth theories.

The theory can be judged by the ap-peal of the theory of technical progressbuilt into it. The view is certainly “evolu-tionary,” and in that regard squares wellwith the accounts given by scholars oftechnical advance like Vincenti, at leastin abstract forms. It must be noted that,within the theory, “evolution” is going onat several different levels. New techno-logical departures are being generatedby individual firms, which in effect “se-lect” on them deciding which to intro-duce and which not to. (For an empiricalstudy of evolution within a firm seeRobert Burgelmam 1993.) Firms alsoare, by scanning their competitors’ tech-nologies, deciding which of these to takeaboard and which not. In addition, thereis market selection on firms that are do-ing well.

Within this class of models, “profit-ability” determines the “fitness” of tech-nology, and of firms, and firms are theonly organizational actors. These obser-vations call attention to the fact thatthis theory would seem to apply only toeconomic sectors where the marketprovides the (or the dominant) selec-tion mechanism winnowing on technolo-gies and firms. It is not well suited fordealing with sectors like medical care,or defense, where professional judg-ments, or political process, determinewhat is fit and what is not. Selection en-vironments clearly differ from sector tosector, and it would seem that these dif-ferences need to be understood and builtinto sectoral level analyses. (For anelaboration of this point, see Nelson andWinter 1977.)

However the central purpose of themodels considered in this section is toexplain economic growth at a macro-economic level. Thus a fundamentalquestion about them is this. Can theygenerate, hence in a sense explain, therising output per worker, growing capitalintensity, rising real wages, and a rela-


tively constant rate of return on capital,that have been the standard pattern inadvanced industrial nations and whatneoclassical growth theory seems to ex-plain? The answer is that they can, andin ways that conform well with underly-ing appreciative theory.

Within these models a successful tech-nological innovation generates profits forthe firm making it, and leads to capitalformation and growth of the firm. Firmgrowth generally is sufficient to out-weigh any decline in employment perunit of output associated with productiv-ity growth, and hence results in an in-crease in the demand for labor, whichpulls up the real wage rate. This latterconsequence means that capital usingbut labor saving innovations now becomemore profitable, and when by chancethey appear as a result of a “search,”they will be adopted, thus pulling upthe level of capital intensity in the econ-omy. At the same time that labor produc-tivity, real wages, and capital intensityare rising, the same mechanisms holddown the rate of return on capital. If theprofit rate rises, say because of the crea-tion of especially productive new tech-nology, the high profits will induce an in-vestment boom, which will pull upwages, and drive capital returns backdown.

These deductions of evolutionarygrowth theory would not surprise an ad-vocate of neoclassical theory. On the sur-face they appear similar to those of neo-classical growth theory. Indeed forevolutionary theory to have credibilitythese predictions had better be similar,because any broad growth theory needsto be consistent with the basic empiri-cally documented broad features of eco-nomic growth as we have experienced it.However, while at first glance the mech-anisms explaining these patterns havea certain surface similarity in evolution-ary and neoclassical theory, if one looks

beneath the surface one can see thatthe mechanisms in fact are very differ-ent. In particular, one theory is based onthe “conception” of a moving equilib-rium, and the other most emphatically isnot.

And if one takes a closer look, it be-comes clear that evolutionary theory en-ables one to see, to expect, phenomenato which neoclassical theory is blind, ordenies. At the same time that the modelgenerates “macro” time series that re-semble the actual data, beneath the ag-gregate at any time there is considerablevariation among firms in the technolo-gies they are using, their productivity,and their profitability. Within this modelmore productive and profitable tech-niques tend to replace less productiveones, through two mechanisms. Firmsusing more profitable technologies grow.And more profitable technologies tendto be imitated and adopted by firms whohad been using less profitable ones. Thusthe theory is consistent with both thelarge body of empirical work that hasdocumented considerable and persistentintra-industry inter-firm dispersion (e.g.,Richard Rumelt 1991 and Mueller 1989)and with what is known empiricallyabout the diffusion of new techniques(see e.g., Stanley Metcalfe 1988). Neo-classical growth theories have troublebeing consistent with these elements ofeconomic growth as we have experiencedit.

Luc Soete and Roy Turner (1984),Metcalfe (1988, 1992), and Metcalfe andMichael Gibbons (1989) have developedevolutionary growth models focusing ondiffusion, in the sense above. Theseauthors repress the stochastic element inthe introduction of new technologiesthat was prominent in the models de-scribed above and, in effect, work with agiven and fixed set of technologies. How-ever, within these models each of the in-dividual technologies may be improving


over time, possibly at different rates. Atthe same time, firms are tending to allo-cate their investment portfolios moreheavily toward the more profitable tech-nologies than toward the less. As a re-sult, rising productivity in the industry asa whole, and measured aggregate “tech-nical advance,” is the consequence oftwo different kinds of forces. One is theimprovement of the individual technolo-gies. The other is the expansion of use ofthe more productive technologies rela-tive to the less productive ones.

Both groups of authors point out thatthe latter phenomenon is likely to be amore potent source of productivitygrowth when there is prevailing largevariation in the productivity of technolo-gies in wide use, than when the besttechnology already dominates in use.Thus the aggregate growth performanceof the economy is strongly related to theprevailing variation beneath the aggre-gate.

The model by Gerald Silverberg, Dosi,and Luigi Orsenigo (1988) develops thebasic theoretical notions introduced inthis section in another direction. In theirmodel there are only two technologies.One is potentially better than the other,but that potential will not be achievedunless effort is put into improving pre-vailing practice. Rather than incorporat-ing a separate “search” activity, in Silver-berg et al. a firm improves its prevailingprocedures (technologies) through learn-ing associated with operation. What afirm learns is reflected in its increasedproductivity in using that technology, butsome of the learning “leaks out” and en-ables others using that technology to im-prove their productivity for free, as itwere.

In contrast with the other models con-sidered in this section where firms donot “look forward” to anticipate futuredevelopments, in the model of Silver-berg et al., firms, or at least some of

them, recognize that the technology thatinitially is behind in productivity is po-tentially the better technology, and alsothat they can gain advantage over theircompetitors if they invest in using andlearning with it. In contrast with theNelson-Winter model, a firm may em-ploy some of both technologies, andhence may use some of its profits fromusing the prevailing best technology toinvest in experience with presently infe-rior technology that is potentially thebest. If no firm does this, then of coursethe potential of the potentially bettertechnology never will be realized.

An early “innovator” may come out awinner, if it learns rapidly, and little ofits learning “spills out,” or its competi-tors are sluggish in getting into the newtechnology themselves. On the otherhand, it may come out a loser, if itslearning is slow and hence the cost ofoperating the new technology remainshigh, or most of its learning “spills out”and its competitors get in in a timelymanner, taking advantage for-free of thespillover.

Several other evolutionary growthmodels have been developed. GunnarEliasson and colleagues have been con-structing over the years a very detailedevolutionary model calibrated on theSwedish economy. (See Eliasson’s chap-ter in Day and Eliasson 1986.) FrancescaChiaramonte and Dosi (1993) recentlyhave blended into the Silverberg-Dosi-Orsenigo model elements of the Nelson-Winter assumptions about stochasticsearch for new techniques. KatsuhitoIwai (1984) and John Conlisk (1989) alsohave published models in this class.There clearly is a lot of richness in these“Schumpeterian” models of economicgrowth, and I believe a lot of potential.It remains to be seen how many econo-mists studying economic growth usingthe “old” theoretical technology will beattracted to gamble on the new.


V. Path Dependencies, DynamicIncreasing Returns, and the Evolution of

Industry Structure

The models considered in Section IVgo a certain distance toward consistencywith the appreciative theoretic accountsof long run economic change, but therestill is a lot that is “mechanical” aboutthem. Certain variables grow over time,in particular output per worker and realwages. Others remain more or less con-stant, like the rate of return on capitaland factor shares, or at least show no sys-tematic drift. However by and largenothing goes on that could be called “de-velopment.” While industry may becomemore concentrated over time, there areno major changes in industry structure ofthe sort often highlighted in economichistories. No radically new technologiesemerge, no new institutions.

The evolutionary theories consideredin this section have more of a develop-mental flavor. They involve path depen-dencies, dynamic increasing returns, andtheir intereaction.

Path dependencies are built into all ofthe models considered above, and dy-namic increasing returns (which is oneway path dependency may emerge) intosome. Thus in virtually all of the models,the particular firms that survive in thelong run are influenced by events, to aconsiderable extent random, that happenearly in a model’s run. To the extent thatfirms specialize in particular kinds oftechnology, what technologies survive isinfluenced similarly by early randomevents. In some of the models, “dynamicincreasing returns” makes path depen-dency particularly strong. Thus in Silver-berg, Dosi, and Orsenigo the more afirm uses a technology the better it getsat that technology. More, some of thelearning “spills over” to benefit otherfirms using that particular technology.Thus the more a technology is used,

the better it becomes vis àvis its com-petitors.

But while path dependencies and dy-namic increasing returns are built intomost of the models we already have con-sidered, this was not the center of atten-tion of the authors. Over the past fewyears, however, a considerable literaturein evolutionary economics has grown upfocused on these topics. The works ofBrian Arthur (1988, 1989) and PaulDavid (1985, 1992) are particularly inter-esting. My treatment here will aim togeneralize the issues they address.

A. Technology Cycles and Dominant Designs

I begin by considering models that fo-cus on competition among technologies.Students of technical advance long havenoted that, in the early stages of a tech-nology’s history, there usually are a num-ber of competing variants. (For a finediscussion and a number of illustrativecase studies see James Utterback 1994.)Thus in the early history of automobiles,some models were powered by gasolinefuelled internal combustion engines,some by steam engines, some by batter-ies. As we know, gradually gasoline fu-elled engines came to dominate and theother two possibilities were abandoned.

The standard explanation for this isthat gasoline engines were the superiormode, and with experience that wasfound out. The Silverberg, Dosi, andOrsenigo paper contains a model of thismechanism. In their analysis a poten-tially superior new alternative requiressome development—learning—before itslatent superiority becomes manifest. Itcan take time before that developmentoccurs and, with bad luck, it even is pos-sible that it never occurs. But by andlarge the potentially better technologywill win out.

In the Arthur and David models, onecan see a different explanation for why


the internal combustion engine won out.It need not have been innately superior.In these models there are dynamic in-creasing returns, in that the more a par-ticular technology is employed, thegreater its attractiveness relative to itscompetitors. Thus in the case in ques-tion, all that would have been requiredfor the gasoline engine to come to domi-nate was a run of luck. For some chancereason it gained an initial lead, and thisstarted a rolling snowball mechanism.

What might lie behind an increasingreturns rolling snowball? Arthur, David,and other authors suggest several differ-ent possibilities.

One is that each of the competingtechnologies involved is what Winter andI have called a cumulative technology. Ina cumulative technology, today’s techni-cal advances build from and improveupon the technology that was available atthe start of the period, and tomorrow’sin turn builds on today’s. The cumulativeeffect is like the technology specificlearning in the Silverberg et al. model.

Thus according to the cumulativetechnology theory, in the early history ofautomobiles, gasoline engines, steam en-gines, and electrical engines, might allhave been plausible alternative technolo-gies for powering cars. While we nowknow that gasoline engines becamedominant, according to this theory thismight have been simply a matter of luck.By chance inventors tended to concen-trate on it, or by chance big advanceswere made. However, once the gasolineengine had been developed to a pointwhere it was significantly superior to ex-tant steam or electrical engines, invest-ing time and resources to advance theseother technologies came to appear a badbet, because such a large gap in perfor-mance needed to be made up beforethey would be competitive.

There are two other dynamic increas-ing returns stories that have been put

forth. One stresses advantages to con-sumers or users if different individualsbuy similar, or compatible products—this has been called network externali-ties—which lend advantage to a variantthat just happens to attract a number ofcustomers early. The other stresses com-plements, for example where a particularproduct has a specialized complementaryproduct or service, whose developmentmay lend that variant special advantages.Telephone networks, in which each useris strongly interested in having other us-ers have compatible products, is themost commonly employed example ofthe first case. Video cassette recorderswhich run cassettes that need to be spe-cially tailored to their particular design,or computers that require compatiblesoftware, are often used as examples ofthe second. (For a very good general dis-cussion and review of the literature onboth of these stories, see Michael Katzand Carl Shapiro 1994.)

However, while the stories are differ-ent, the mathematics used to formalizethem tends to be the same. (See e.g., Ar-thur 1988.) Also, the phenomena oftenare intertwined, and also linked with theprocesses involved in the development ofcumulative technologies, as in David’s(1985) example of the QWERTY type-writer keyboard.

Thus to return to our automobile ex-ample, people who learned to drive intheir parents’ or friends’ car powered byan internal combustion engine almostcertainly were drawn to similar carswhen they themselves came to purchaseone, in part to avoid the new learningand potential surprises that would be in-volved if they bought a steam or electricpowered one. At the same time the as-cendancy of automobiles powered by gasburning internal combustion enginesmade it profitable for petroleum compa-nies to locate gasoline stations at conven-ient places along highways. It also made


it profitable for them to search for newsources of petroleum, and to developtechnologies that reduced gasoline pro-duction costs. In turn, this increased theattractiveness of gasoline powered carsto car drivers and buyers.

Note that, for those who consider gasengine automobiles, large petroleumcompanies, and the dependence of alarge share of the nation’s transportationon petroleum, a complex that lies behindmany social ills, the story spun out aboveindicates that “it did not have to be thisway.” If the roll of the die early in thehistory of automobiles had come out an-other way, we might today have hadsteam or electric cars. A similar argu-ment recently has been made about thevictory of A.C. over D.C. as the “system”for carrying electricity (for an open-minded discussion see David 1992). Thestory also invites consideration of possi-bly self-interested professional judg-ments or political factors as major ele-ments in the shaping of long runeconomic trends, a subject we will pickup shortly. After all, under these theo-ries all it takes may be just a little push.

On the other hand, other analysts maysee the above account as overblown.Steam and battery powered car engineshad major limitations then and still donow; gasoline clearly was better. A.C.had major advantages over D.C., and stilldoes. According to this point of view dy-namic increasing returns is an importantphenomenon, but it is unlikely that it hasgreatly influenced which technology wonout, in most important cases. I predictthat this issue will be a lively topic ofempirical research and argument overthe coming years.

There also is a more general openquestion about the range of technologieswhere a “dominant design” emerges, forany reason. The various dynamic increas-ing returns stories seem plausible forsome product classes, but not for others.

And in some product class areas differ-ent user needs may tend to prevent aparticular product from coming to domi-nate the market, even if there are dy-namic increasing returns. Pharmaceuti-cals, the value of which are extremelysensitive to both the particular diseaseand the particular characteristics of thepatient, are a good case in point.

B. Firm and Industry Structure

I turn now to a different but relatedbody of evolutionary writings—that con-cerned with the evolution of industrystructure as a technology develops. It istied to the notion that in most technolo-gies after a period of time a dominantdesign emerges, but is not committed toany particular theory of how that hap-pens, whether because the truly bettervariant is finally found and consensus de-velops around it or because of dynamicincreasing returns phenomena. In anycase, within this body of evolutionarytheorizing, the establishment of a domi-nant design has important implicationsregarding the subsequent nature of R &D, and for industry structure.

The basic argument would appear tohave two sources. The first is Muellerand Tilton (1969), based on their specu-lations about patterns of industry evolu-tion they were observing. The secondwas William Abernathy and Utterback(1975) based on their detailed study ofautomobiles. Because the Abernathy andUtterback story is closely linked to an in-teresting theory of what happens to R &D as a dominant design emerges, I willfollow it.

The basic proposition is that, prior tothe emergence of a dominant design,there is little R & D directed toward im-proving production processes, becauseproduct designs are unstable, and themarket for any one is small. With theemergence of a dominant design, theprofits from developing better ways of


producing it become considerable. Oftenthe development of better productionprocesses will involve the exploitation oflatent scale economies, and the estab-lishment of capital intensive modes ofproduction. In turn, the improvementof production processes that are specificto a particular broad design further locksit in, and disadvantages competing de-signs.

The argument then is that this patternof technological evolution causes a par-ticular pattern of evolution of firm andindustry structure. In the early stagesof an industry—say automobiles—firmstend to be small, and entry relativelyeasy, reflecting the diversity of technolo-gies being employed, and their rapidchange. The industry consists of a num-ber of smallish firms, but with a lot ofentry and exit. As the quality of theproducts improve, and the market grows,so do the number of firms active in theindustry. However, as a dominant designemerges, and specialized production pro-cesses are developed, barriers to entrybegin to rise as the scale and capitalneeded for competitive productiongrows. Also, with the basic technologyset, learning becomes cumulative, andincumbent firms are advantaged relativeto potential entrants for that reason aswell. After a shake out, industry struc-ture settles down to a collection of estab-lished largish firms.

When this theory was first put forth,there was only limited data supportingit. Since that time Michael Gort andSteven Klepper (1982), Klepper andElizabeth Graddy (1990), Utterback andFernando Suarez (1993), Utterback(1994), and Franco Malerba andOrsenigo (1993, 1994) have providedconvincing evidence that this pattern ofevolution in fact holds over a wide rangeof industries.

A recent formal model developed byKlepper (1993) accepts the broad em-

pirical story, but puts forth a differentevolutionary theory to explain it. InKlepper’s model the investments madeby a firm in product innovation are inde-pendent of firm size, but investments inprocess innovation are positively relatedto firm size. As in the more standardstory, in the early days of a technology’shistory, firms are small, for that reasonlittle process R & D is done, and entrybarriers are low. The presence of manyfirms makes for rapid product innova-tion. But as profitable extant firms growand invest more in process innovation,entry barriers rise. Shake out occurs be-cause of rivalry among the extant firms,increasingly competing on the basis ofcost. No dominant design emerges in theKlepper model, but as the number of ex-tant firms dwindles, product innovationslows.

C. Supporting Institutions

The writing on the coevolution oftechnology and industry structure tendsto define industry structure rather con-ventionally. However, there are a num-ber of studies which define industrystructure more broadly, or look outsidethe industry, narrowly defined, and areconcerned with the coevolution of atechnology and industry with varioussupporting institutions.

As an industry becomes established,one frequently observes not only the de-velopment of technical and product stan-dards, but also the emergence of stan-dard patterns of interaction moregenerally between firms, suppliers, andcustomers, and across firms in the indus-try. Economic relations become embed-ded in social ones, along the lines de-scribed by Mark Granovetter (1985), andpeople become conscious that there is anew industry, and that it has collectiveinterests and needs. Michael Hannanand Glenn Carroll discuss in some detailthese processes of “legitimation” and


their consequences. See also BennettHarrison (1992). Industry or trade asso-ciations form. These give the industry arecognized organization that can lobbyon its behalf for regulation to its liking,for protection from competition fromoutside the group, for public programs tosupport it, etc. This is another feature ofan industry’s evolution that can lock inthe status quo.

More generally, while the formal evo-lutionary growth models of Section IV,and the dynamic increasing returns mod-els discussed at the beginning of this sec-tion, take the basic parameters of the“selection environment” (usually treatedas a market) as given, many of the soci-ologists studying industry evolutionstress that the industry itself stronglymolds its own selection environment. Itdoes so through the rules of behaviorand interaction among firms that evolvespontaneously, through the formation ofa variety of industry-related organiza-tions that decide matters like standards,and through political action. (See e.g.,Michael Tushman and Elaine Romanelli1985; and Rosenkopf and Tushman1994.) In turn such action may be centralin determining what design or systemturns out to be dominant. (For such adiscussion bearing on electric power sys-tems, see Patrick McGuire, Granovetter,and Michael Schwartz 1993.)

If the technology on which the indus-try is based has novel characteristics,new technical societies and new techni-cal journals, tend to spring up. In somecases whole new fields of “science” maycome into being (Rosenberg 1982, ch. 7;and Nelson and Rosenberg 1993). Thusthe field of metallurgy came into exist-ence because of a demand for better un-derstanding of the factors that deter-mined the properties of steel. Computerscience was brought into existence bythe advent of the modern computer.Chemical engineering and electrical en-

gineering rose up as fields of teachingand research because of industry de-mand for them that occurred after thekey technological advances that launchedthe industries. Earlier I noted the appar-ent blindness of much of the writing onhow science evolves to the use of sciencein technology. The technology-orientedsciences directly provide a “market like”environment stimulating research onvarious topics and also a stringent testenvironment for new scientific theoriesand other published findings.

The emergence and development ofthese technology-oriented sciences tendto tie industries to universities, whichprovide both people trained in the rele-vant fields, and research findings whichenable the technology to advance fur-ther. The development of these sciencesnaturally lends extra strength to prevail-ing technologies. On the other hand thepresence of university research tends todilute the extent to which firms in beinghave knowledge advantages over poten-tial entrants. Also, research at universi-ties just may become the source of radi-cally different technological alternatives.

Recognition of the role of technical so-cieties and universities in the develop-ment of modern technologies opens thedoor to seeing the wide range of institu-tions that may co-evolve with a technol-ogy and an industry. Often legal struc-tures need to change. Thus there may beintellectual property rights issues thatneed to be sorted out—bio-technology isa striking contemporary case in point.There almost always are issues of regu-lation, as was prominently the case inradio and, in a different manner, bio-technology again. Hughes (1983) hasdescribed in great detail the wide rangeof legal and regulatory matters that hadto be decided before electric powercould go forward strongly, and how theparticular ways they were decided af-fected the evolution of the technology


and the industry. The coevolution of lawand technology and industry structurehas been only lightly touched in the writ-ings on how the law evolves.

In many cases new public sector activi-ties and programs are required. Thusmass use of automobiles required thatsocieties organize themselves to buildand maintain a system of public roads.Airplanes required airports. The devel-opment of radio, and of commercial tele-vision, required mechanisms to allocatethe radio spectrum.

These examples indicate that the evo-lution of institutions relevant to a tech-nology or industry may be a very com-plex process involving not only theactions of private firms competing witheach other in a market environment, butalso organizations like industry associa-tions, technical societies, universities,courts, government agencies, legisla-tures, etc. In turn, the way these otherorganizations evolve and the things theydo may profoundly influence the natureof the firms and the organization of in-dustry. Thus Michael Piore and CharlesSabel (1984) have proposed that the or-ganization of manufacturing activitythrough large vertically integrated firms,which came to be the norm in many U.S.industries in the early decades of thetwentieth century, was not inevitable,but was drawn by the broader institu-tional context of the U.S. We might in-stead have organized production in manyof these industries through a more fluidstructure of networked small and me-dium-sized firms. (See also Sabel andJonathan Zeitlin 1993.) In the view ofthese authors, the U.S. might be in bet-ter shape now had the latter been thecase.

VI. Responding to the Winds of Change

Evolutionary theory in biology pro-vides a sharp answer to the question of

how life responds in situations wheremajor environmental changes make exist-ing dominant life forms ill adapted. Tothe extent that better adapted life formsare present in at least small numbers,these and their similar offspring willthrive and multiply, and their now poorlyadapted peers will tend to die out. Somenew varieties created through mating ormutation that would have had no chancein the old regime, may do well in thenew one. Others that would have pros-pered now may have no chance.

How is it in economics? If one consid-ers “firms” or other organizations as car-riers of basic practices—earlier I calledthem routines—what happens when themarket or something else changes? Afundamental difference between organi-zations and organisms, of course, is thatthe former are not stuck with their rou-tines but can change them, while the lat-ter cannot change their genes. Thus,unlike the case in biology, it is meaning-ful in economic evolutionary theory toask about the extent to which significantadjustment to changed environmentalconditions—for example a sharp changein patterns of consumer demands, or fac-tor availabilities and prices, or the ad-vent of radically new technology—isachieved largely by old organizationslearning new ways, or requires the deathof old organizations, and the birth of newones.

Some of the organizational ecologymodels developed by sociologists take aposition that firms are like biological or-ganisms. Thus Michael Hannan and JohnFreeman (1989) posit (for the purposesof their formal theorizing) that organiza-tions cannot change their ways at all. Un-der this view society’s ability to respondto change depends entirely on the pres-ence at any time of a variety of organiza-tions, or the generation of new ones.(For more eclectic surveys of sociologi-cal approaches to the evolution of or-


ganizations, see Howard Aldrich 1979;W. Richard Scott 1992; and Joel Baumand Jitendra Singh 1994.)

While this position may sound bizarreto many economists, a number of carefulstudents of firm behavior have been im-pressed that the set of things a firm cando well at any time is quite limited, andthat, while firms certainly can learn to donew things, these learning capabilitiesalso are limited. Thus Mueller (1989),Karel Cool and Dan Schendel (1988),and Rumelt (1991) have shown that,within an industry, there tends to be per-sistent differences across firms in profit-ability or productivity. While “imitation”is an important economic phenomenon,there would appear to be durable firmdifferences, associated with unique re-sources or competences. Dosi, DavidTeece, and Winter (1992) have devel-oped an argument that, to be effective, afirm needs a package of routines, includ-ing those concerned with learning andinnovation, that are “coherent.” But thatcoherency, on the other hand, entails acertain rigidity.

Paul Milgrom and John Roberts(1990), commenting on a wide range ofrecent literature on firm competences,have stressed that competences tend tocome in strongly complementary pack-ages of traits. As Daniel Levinthal (1994)argues, this undoubtedly is an importantreason that successful firms often aredifficult to imitate effectively, becauseto do so requires that a competitor adopta number of different practices atonce. It also is an important reason whyfirms who do well in one context mayhave great difficulty in adapting to a newone.

Winter and I (1977, 1982) and Dosi(1982, 1988) have used the concept oftechnological regime or paradigm to re-fer to the set of understandings about aparticular broad technology that areshared by experts in a field, including

understandings about what a firm needsto be doing to operate effectively in thatregime. Tushman and Anderson (1986)have coined the term “competence de-stroying technical advance” to charac-terize new technologies when the skillsand understandings needed to deal withthem are significantly different fromthose relevant to the old. There is nowconsiderable evidence (see e.g., Tush-man and Anderson 1986; Utterback1994; Clayton Christensen and RichardRosenbloom, forthcoming; RebeccaHenderson and Kim Clark 1990; andHenderson 1993) that when such a newtechnology comes along, the old entrybarriers fall down, new companies enter,and many old ones fail. Thus or-ganizations may be more like organismsthan many economists are wont to be-lieve, and significant economic changelike significant biological change may in-volve large elements of creative destruc-tion.

What about the institutions that sup-port a particular industry or technology?Can the old ones change to meet thechanged needs, or must a basically newset come into existence? If the latter isthe case, does this tend to involve theascendancy of new regions or nations,and the decline of the old? William La-zonick (1990) among others has arguedthat the broad organization of work andinstitutions for training labor thatworked so well for British industry in thelate nineteenth century became a handi-cap in the twentieth. Thornstein Veb-len’s famous essay (1915) on the rise ofGermany as an economic power stressesmore generally that British industry wassorely handicapped in adopting the newtechnologies that were coming into placearound the turn of the century by an in-terlocking set of constraints associatedwith her institutions and past invest-ments, whereas Germany could workwith a relatively clean slate.


Recently these ideas have been revis-ited by Carlotta Perez (1983) and C.Freeman (1991), who have developedthe concept of a “techno-economic para-digm.” Their argument starts along linesdeveloped by Schumpeter many yearsago: different eras are dominated by dif-ferent fundamental technologies. Theythen propose that to be effective withthose technologies a nation requires a setof institutions compatible with and sup-portive of them. The ones suitable for anearlier set of fundamental technologiesmay be quite inappropriate for the new.Perez and Freeman propose that the pe-riod since 1970 has seen the rise of “in-formation technologies” as the new basisof economic effectiveness, and arguethat effective accommodation requires avery different set of institutions thanthose required in the earlier era. Japanthey see as coming closest to havingthem. Other but related explanations ofthe rise of Japan, and the decline of theU.S., but focusing on characteristics ofJapanese firms, and their determinants,have been put forth by Masahiko Aoki(1990) and Ronald Gilson and Mark Roe(1993). The Piore and Sabel (1984) argu-ment about the institutional forces thatled to the particular structuring ofAmerican firms is about the other side ofthis story.

Over the last several decades a num-ber of biological evolutionary theoristshave proposed that in biology evolutionoften follows the pattern of “punctuatedequilibrium.” Periodically there arebursts of mutations that somehow takehold, and a new species emerges. Therefollows a period during which the spe-cies evolve rapidly into a form that, then,seems to stabilize. Then in some cases, anew species emerges that replaces theold. The foregoing analysis suggests that,like species, the pattern of evolution oftechnology linked institutional formsoften is that of punctuated equilibrium.

VII. Economic Institutions and TheirEvolution

Two somewhat different intellectualstreams have fed into the renewed inter-est economists have taken in institutions.Economists long have looked to differ-ences across nations in their basic insti-tutions as an explanation for differencesin economic performance and livingstandards (see Hodgson 1988, for a finereview of the “old” institutional econom-ics), and in recent years that interest hasintensified. For the most part until re-cently research along these lines hasbeen empirical, with the theorizing ap-preciative. Recently that empirically mo-tivated theorizing has become more for-mal. Also, over the past fifteen years orso, game theorists have come to be inter-ested in “institutions,” associating themtheoretically with a particular solution ofgames that have multiple Nash equili-bria. That is, the pattern of behavior as-sociated with an equilibrium is seen as“institutionalized.” In turn, this intellec-tual development has had a strong influ-ence on the empirically oriented theoriz-ing.

One issue that has plagued both oldand new research on institutions andtheir evolution has been how to defineinstitutions. The term has been used tocover a grab bag of varied things. Somewriters, particularly the older generationof institutional economists, have usedthe term to refer to what the theorists ofcultural evolution, discussed above,would call “culture,” or more specificallyto those aspects of culture that affect hu-man and organizational action. Underthis perspective institutions refer to thecomplex of socially learned and sharedvalues, norms, beliefs, meanings, sym-bols, customs, and standards that deline-ate the range of expected and acceptedbehavior in a particular context. Thisview of institutions is alive and well in


modern sociology. (See in particularWalter Powell and Paul Di Maggio1991.)

The “new” economic institutionalistshave a different intellectual startingplace, and as noted above have borrowedextensively from game theory. ThusDouglass North (1990) has proposed thatinstitutions are “the rules of the game.”(See also Thrainn Eggerston 1990.) Theargument then is that, given the motiva-tions of individuals and organizationsand technological or other constraints,“the rules of the game” determine howand why it is played as it is. AndrewSchotter (1981), recognizing that gamesmay have multiple equilibria, has sug-gested that institutions define “How thegame is played” (see also Robert Sugden1989). Thus the concept here includesnot only the rules, but also the standardand expected patterns of actual play thathave evolved, which define the con-straints and expectations of the presentplayers. It is this concept of institutionsthat has become prevalent in evolution-ary game theory. , While developed in adifferent way, the game theoretic view ofinstitutions has much in common withthe sociological. What is different is thestress by sociologists on norms and beliefsystems rationalizing action in a givencontext, whereas the emphasis in gametheory is on the self-enforcing nature ofinstitutionalized behavior. It should benoted that North is very close to the so-ciologists here.

On the other hand, most historical ac-counts of institutions refer to more con-crete things: the form of the modern cor-poration, or the modern researchuniversity, the financial system, and theparticular kind of money in use, thecourt system, a nation’s basic legal code,etc. Alessandra Casella and Bruno Frey(1992) use the term “institution” to referto particular structures and bodies of lawlike GATT, which define a kind of public

order. How do these two apparently dif-ferent notions about institutions relate?

It is not totally clear, but North makesa distinction between what prevailing in-stitutions in his broad sense allow or re-quire, and particular realizations withinthe set of the institutionally possible.Thus I understand him as taking a posi-tion akin to Durham’s regarding culture,that institutions, as he defines them, in-fluence and constrain, but leave consid-erable room for variability in the way so-ciety actually organizes itself. Alongthese lines one can, as Williamson(1985), see the M form of organization ofmulti-product firms as, during the periodfrom 1920 to 1970, a prevailing institu-tion, defining the expectations andnorms for such firms. IBM and GM wereorganizational exemplars of these norms,and as such often referred to as institu-tions. However, according to this inter-pretation the “institution” is really de-fined by the pattern and the norms.Similarly, during the 1960s and 1970sHarvard and the University of Californiawere institutional exemplars of what waswidely accepted research universitiesshould be. In the same vein, the IMF,and GATT, were particular organiza-tions, but also representing institutions,in being the particular manifestation of aset of norms and beliefs.

I confess uneasiness at the broad androomy definition of institutions invokedby the old institutionalists, and my un-easiness here carries over to the modernpractice of calling any widespread prac-tice, that can be interpreted as the equi-librium of a game, an institution. Buthere I put these concerns aside and re-flect on the proposition that, howeverthey may be defined, the institutions wenow have came about as a result of anevolutionary process.

Abstracting from the enormous diver-sity of things that have been called insti-tutions, there are several key matters


that I believe any serious theory of insti-tutional evolution must address. One ispath dependency. Today’s “institutions”almost always show strong connectionswith yesterday’s, and often those of acentury ago, or earlier.

A second is that it almost certainly isnecessary to think of evolutionary “pro-cesses” in the plural. Different kinds ofinstitutions evolve in different ways. Theearlier generation of institutional econo-mists tended to stress the role of expresscollective decision making. The presentgeneration tends to stress unplannedself-organization. In many cases the evo-lutionary processes at work seem to in-volve a blend of market, professional,and political processes, and it is likely anenormous task to sort these out and getan accurate assessment of operative “fit-ness” criteria and selection mechanisms.

One virtue of recognizing evolutionarytheorizing as a class is that this encour-ages the application of what is learned inanalysis of one topic to analysis of others.The intellectual traverse taken by Northis quite interesting in this regard. In hisearly work on economic institutions(Lance Davis and North 1971), North’sposition was that, despite the fact thatinterested parties often differed in theirgoals, and despite the fact that collectivepolitical processes often were involvedcentrally in the process of institutionalevolution, evolution did assure some-thing like optimality. On the other hand,in his recent writings (North 1990), hedraws lessons from the above learningand distances himself sharply from anyposition along side Pangloss. His centralargument is this. First, the major differ-ences among nations in economic per-formance largely are due to differencesin their institutions and how they haveevolved. While nowhere can they be re-garded as optimal, in some countriesthey have evolved in a way that is favor-able to economic progress and in other

countries not. Second, the advanced in-dustrial nations have been extremely for-tunate in this regard; one cannot attrib-ute their relative well being to anyspecial virtue and wisdom but rather tocultural and political contingencies.

Friedrich Hayek (1988) long hasstressed the evolutionary character ofthe way modern economic institutionshave developed, using the following ar-gument. The structure of prevailing in-stitutions is far too complex for humanbeings to comprehend, hence there is noway people could actually have designedthem. More, to think that we could, orthat we can scrap them and replace themwith something we can plan that wouldbe better, is a “Fatal Conceit.” Hayek isfar too sophisticated a scholar to betarred as arguing that existing institu-tions are optimal. Nor, while conserva-tives appropriately place him in theirPantheon, does he deny that consciouspublic action has played an importantrole in structuring the institutions wepresently have. Rather, his central pointis that our present institutional struc-tures must be interpreted as largely theresult of a process involving somewhatblind variation and social selection.

However, for reasons he is unable orunwilling to state, Hayek does not layout exactly “How the West Grew Rich,”to borrow a term from Rosenberg andLuther Birdzell (1986). There is littlediscussion in Hayek about the actualmechanisms that have “selected” the in-stitutions we now have, only some asser-tions that what we have is the result ofsocial learning. He says virtually nothingabout how that occurs, or how it works inthe benefit of the society as a whole, ascontrasted with favoring individual inter-ests that, when they are aggregated, aredestructive of everyone. Yet somehow(he implicitly argues) what we haveachieved works pretty well (this isNorth’s point), and in any case messing


with it in any radical way almost surelywill make things worse.

Rosenberg and Birdzell argue a variantof this theme. It is that “the West GrewRich” because the societies broke loosefrom the norms and constraints of old in-stitutions, and kept political processfrom doing too much, and let the “mar-ket” work.

But this will not quite do as a coherenttheory. The “market” here is not just themarket for goods and services or newtechniques of production or modes of or-ganizing private production. Rosenbergand Birdzell also are concerned with theinstitutions of modern science, bodies oflaw and mechanisms to enforce law andmake new law, etc. It probably is use-ful to posit that these “institutions”“evolved.” One can even speak of a “mar-ket” for institutional changes. We sawthis earlier in the discussion of theoriesthat proposed the law evolved to en-hance economic efficiency. But in factthere is no real “market” that sorts outamong proposed changes in the law.Rather there is a set of economic and po-litical interests, professional and lay be-liefs about what the law should be, and adiverse set of mechanisms, some ex-pressly political and some not, throughwhich these interests and norms influ-ence the evolution of the law. And thesame is true for most other things thatwe lump under the term “economic insti-tutions.” We have very little under-standing of how this kind of a selectionenvironment works, and how it defines“fitness.” (For a similar view see MaryDouglas 1986.) We have no reason to be-lieve that such selection environmentsare stringent, or stable, much less thatthey select on “economic efficiency.”

And yet, it is arguably the case that thenow advanced industrial nations haveachieved dramatic economic progress (inmost if not all dimensions) over the lastcentury and a half. As argued in an ear-

lier section, development of new tech-nology certainly has been the primaryforce, but institutional structures haveevolved to enable new technologies tooperate relatively effectively. Indeed,the broad form of the modern corpora-tion with R & D laboratory, and themodern university, which have becomethe major sources of technological ad-vance, themselves have coevolved withtechnology.

It is clear that, somehow, in the nowadvanced industrial nations, there havebeen mechanisms that have made thecoevolution of technology, industrialorganization, and institutions morebroadly, move in directions that have ledto sustained economic progress. Privateactions leading to “self organization”have been part of the story, but collec-tive action has been as well. It is absurdto argue that processes of institutionalevolution “optimize“; the very notion ofoptimization may be incoherent in a set-ting where the range of possibilities isnot well defined, even if the issue of dif-ferent interests could be resolved in thisterminology (as through the set conceptof Pareto optimality). However, thereseem to be forces that stop or turnaround particular directions of institu-tional evolution that, pursued at greatlength, would be disastrous. And strongshifts in the needs of large and powerfulgroups tend to be followed by shifts inthe direction of institutional evolutiontoward ones that better reflect theirchanged needs. I can conjecture plausi-ble models that yield these results. How-ever, to date they have not been ex-plored analytically with any rigor.

Undoubtedly part of the problem re-flects the still primitive state of our abil-ity to work with cultural evolutionarytheories. In this particular case I am sureit also stems from an overly broad andvague concept of the variable in ques-tion—institutions—which is defined so


as to cover an extraordinarily diverse setof things. Before we make more head-way in understanding how “institutions”evolve we may have to unpack and dras-tically disaggregate the concept. But ourdifficulty also may signal the limits of thepower of economics or social sciencetheory more generally to comprehend aset of processes as complex as those be-hind economic growth as we have knownit.

VIII. Reprise

This essay has aimed to provide anoverview of recent writings by econo-mists, and some other social scientists,who have put forth express theoreticalarguments that the variables the authorsare examining change through evolution-ary processes. I have concentrated onworks where empirical subject matter isthe focus of attention, and an evolution-ary theory is invoked to explain the ob-served or alleged pattern of change, andlargely have neglected works where theformal aspects of an evolutionary theoryare central and empirical subject matterbrought in mostly as stylized examples.However a unifying characteristic of thewritings surveyed here is that the evolu-tionary theorizing is set out explicitly, ascontrasted with coming in mostly as away of talking about the empirical sub-ject matter.

As I argued in the introduction, thelatter long has been common in econom-ics. It is the express evolutionary theoriz-ing that is relatively new.

The theoretical arguments I have sur-veyed range from quite precise and for-mal, to storytelling. Virtually all of thethem, however, are put forth by theirauthors to provide a different, and in theauthor’s view a better, theory than onewhich uses the conventional assumptionsof “equilibrium” theorizing.

This of course raises the question of

what one might mean by “better.” Moreaccurate prediction? “On the button”prediction never has been a hallmark ofeconomic analysis, and it is unlikely thatpredictions motivated by an evolutionarytheoretic framework are systematicallybetter or worse than those motivated bya neoclassical theory. The heart of quan-titative prediction making in economicslies in the details of the prediction equa-tions, and these almost always reflectjudgment of the particular context asmuch as formal theory.

Better explanation? If by “better” onemeans statistical “better fits” in varioussenses, again the heart of the exercise isin the details of the equations that arefitted, and those details are as much amatter of art as of broad formal theory.Indeed formal general theory usuallyprovides only loose constraints on mod-els designed to fit particular bodies ofdata.

On the other hand if by “better expla-nation” one means one that is consistentwith informed judgments as to what re-ally is going on, that is exactly the casefor evolutionary theory put forth bythose that advocate it. In general thoseinformed judgments reflect inferencesdrawn from a broad and diversified bodyof data. Thus evolutionary theories ofproductivity growth at a macroeconomiclevel feel right to their advocates notsimply because they can be tuned to fitthose particular data pretty well, but alsobecause the evolutionary explanation isconsistent with observed differences inproductivity and profit among firms,with the fact that even obviously supe-rior new technology usually diffusesslowly, and like observations, that ittakes more strain for neoclassical theoryto encompass.

And that, I would argue, is an impor-tant part of what the “betterness” crite-rion ought to be. Does the explanationring right to those who know the details


of the field? It would seem that this isthe issue Marshall had in mind when hewrote the sentences that began the es-say. Mechanical theories did not ringright with him.

But he also raised the issue of com-plexity. If there is value in formal theo-rizing in economics it lies in the abilityto work through complex causal argu-ments, but if the complexity is too greatone either may lose ability to understandwhat the theory is doing—what leads towhat conclusions—or to check the logicfor accuracy, or both. For all the reasonsdiscussed in the introduction, econo-mists now are far better able to deal withanalytic complexity in general, and thecomplexity of evolutionary models inparticular, than we were twenty yearsago, much less in Marshall’s time. Thereis no doubt, however, that evolutionarytheories still tend to be complex.

Thus those who are attracted to devel-oping and employing them to addressthe phenomena in which they are inter-ested are making an intellectual bet thatthe price of added complexity is worthpaying to buy the better ability to deviseand work with a theory that rings right.The bet is that evolutionary theory opensup a productive research program, to useLakatos’ idea, that is foreclosed or moredifficult if one stays with mechanicalanalogies.

The use of formal evolutionary theoryin economics is still new, and the propo-nents of evolutionary theory are strug-gling with both techniques and stan-dards. It is clear that a number of theevolutionary theories put forth by econo-mists in recent years are difficult to fol-low in terms of their cause—effect logic,and some may be logically incoherent.Merely adopting evolutionary theoreticlanguage does not automatically lead to alogical model. But a number of the newevolutionary theories do seem coherent,and analytically powerful. The coherence

and power of evolutionary theorizing ob-viously depends on the skill and dili-gence of the theorist. There would ap-pear to be nothing different herebetween neoclassical and evolutionarytheorizing.

This said, it is clear that one of theappeals of evolutionary theorizing abouteconomic change is that that mode oftheorizing does seem better to corre-spond to the actual complexity of theprocesses, as these are described by thescholars who have studied them in detail.There is no question that, in takingaboard this complexity, one often endsup with a theory in which precise predic-tions are impossible or highly dependenton particular contingencies, as is thecase if the theory implies multiple orrapidly shifting equilibria, or if under thetheory the system is likely to be far awayfrom any equilibrium, except under veryspecial circumstances. Thus an evolu-tionary theory not only may be morecomplex than an equilibrium theory. Itmay be less decisive in its predictionsand explanations. To such a complaint,the advocate of an evolutionary theorymight reply that the apparent power ofthe simpler theory in fact is an illusion.

A good case can be made that the top-ics and sectors where evolutionary theo-ries that have been developed to date arenotably weak regarding prediction, andsomewhat ad hoc on explanation, arethose where standard neoclassical theo-ries have great difficulties also. They areareas where there is no real market, orwhere market selection is strongly mixedwith political or professional influences.The problem in theorizing here clearlylies not in the evolutionary art form, butin the complexity of the subject matter.

Many years ago Veblen (1896) asked,“Why Is Economics Not an EvolutionaryScience?” In my view economics wouldbe a stronger field if its theoreticalframework were expressly evolutionary.


Such a framework would help us see andunderstand better the complexity of theeconomic reality. That, I think, is itsgreatest advantage. But it will not makethe complexity go away.

REFERENCES

ABERNATHY, WILLIAM AND UTTERBACK, JAMESM. “A Dynamic Model of Process and ProductInnovation,” Omega, 1975, 3(6), pp. 639–56.

ALDRICH, HOWARD. Organizations and environ-ments. Englewood Cliffs, NJ: Prentice Hall,1979.

ANDERSON, PHILIP W.; ARROW, KENNETH J. ANDPINES, DAVID, eds. The economy as an evolvingcomplex system. Redwood City, CA: Addison-Wesley Pub. Co., 1988.

AOKI, MASAHIKO. “Toward a Model of the Japa-nese Firm,” J. Econ. Lit., Mar. 1990, 28, pp.1–27.

ARTHUR, W. BRIAN. “Self-reinforcing Mecha-nisms in Economics,” in PHILIP W. ANDERSON,KENNETH J. ARROW, AND DAVID PINES, eds.1988a, pp. 9–31.

———. “Competing Technologies: An Overview,”in GIOVANNI DOSI ET AL., eds. 1988b, pp.590–607.

———. “Competing Technologies, Increasing Re-turns, and Lock-In by Historical Events,” Econ.J., Aug. 1989, 99(394), pp. 116–31.

BASALLA, GEORGE. The evolution of technology.Cambridge: Cambridge U. Press, 1988.

BAUM, JOEL A. C. AND SINGH, JITENDRA V. Evo-lutionary dynamics of organizations. New York:Oxford U. Press, 1994.

BIJKER, WIEBE E.; HUGHES, THOMAS P. ANDPINCH, TREVOR J. The social construction oftechnological systems. Cambridge, MA: MITPress, 1987.

BOYD, ROBERT AND RICHERSON, PETER J. Cul-ture and the evolutionary process. Chicago: U.of Chicago Press, 1985.

BURGELMAN, ROBERT A. “Fading Memories: AProcess Study of Strategic Business Exit in Dy-namic Environments.” Stanford BusinessSchool xerox, June 1993.

CAMPBELL, DONALD. “Blind Variation and Selec-tive Retention in Creative Thought as in OtherKnowledge Processes,” Psychological Rev.,1960, 67, pp. 380–400.

———. “Evolutionary Epistemology,” in The phi-losophy of Karl Popper. Ed.: PAUL A. SCHILPP.LA SALLE, IL: Open Court, 1974, pp. 413–63.

CASELLA, ALESSANDRA AND FREY, BRUNO S.“Federalism and Clubs: Towards an EconomicTheory of Overlapping Political Jurisdictions,”Europ. Econ. Rev., Apr. 1992, 36(2–3), pp. 639–46.

CAVALLI-SFORZA, LUIGI L. AND FELDMAN, MAR-CUS W. Cultural transmission and evolution: Aquantitative approach. Princeton: Princeton U.Press, 1981.

CHANDLER, ALFRED D. Strategy and structure:Chapters in the history of industrial enterprise.Cambridge: Harvard U. Press, 1962.

———. Scale and scope: The dynamics of indus-trial capitalism. Cambridge: Harvard U. Press,1990.

CHIAROMONTE, FRANCESCA AND DOSI,GIOVANNI. “Heterogeneity, Competition, andMacroeconomic Dynamics,” Structural Changeand Economic Dynamics, 1993, 4, pp. 39–63.

CHRISTENSEN, CLAYTON AND ROSENBLOOM,RICHARD. “Explaining the Attacker’s Advan-tage: Technological Paradigms, OrganizationalDynamics, and the Value Network,” ResearchPolicy, forthcoming.

CLARK, NORMAN AND JUMA, CALESTOUS. Longrun economics: An evolutionary approach toeconomic growth. London: Pinter Pub., 1987.

COHEN, WESLEY M. AND LEVINTHAL, DANIEL A.“Innovation and Learning: The Two Faces of R& D,” Econ. J., Sept. 1989, 99(397), pp. 569–96.

CONLISK, JOHN. “An Aggregate Model of Techni-cal Change,” Quart. J. Econ., Nov. 1989, 104(4),pp. 787–821.

COOL, KAREL AND SCHENDEL, DAN. “Perform-ance Differences Among Strategic Group Mem-bers,” Strategic Manage. J., 1988, 9(3), pp. 207–23.

COOTER, ROBERT D. AND RUBINFELD, DANIELL. “Economic Analysis of Legal Disputes andTheir Resolution,” J. Econ. Lit., Sept. 1989,27(3), pp. 1067–17.

DAVID, PAUL A. Technical choice, innovation andeconomic growth. Cambridge: Cambridge U.Press, 1975.

———. “Clio and the Economics of QWERTY,”Amer. Econ. Rev., May 1985, 75(2), pp. 332–37.

———. “Heroes Herds and Hysteresis in Techno-logical History,” Industrial and CorporateChange, 1992, 1(1), pp. 129–79.

DAVIS, LANCE E. AND NORTH, DOUGLASS C. In-stitutional change and American economicgrowth. Cambridge: Cambridge U. Press, 1971.

DAY, RICHARD H. AND ELIASSON, GUNNAR. Thedynamics of market economies. Amsterdam:North Holland, 1986.

DE BRESSON, CHRIS. “The Evolutionary Para-digm and the Economics of TechnologicalChange,” J. Econ. Issues, June 1987, 21 (2), pp.751–62.

DEMSETZ, HAROLD. “Toward a Theory of Prop-erty Rights,” Amer. Econ. Rev. , May 1967,57(2), pp. 347–59.

DOSI, GIOVANNI. “Technological Paradigms andTechnological Trajectories: A Suggested Inter-pretation of the Determinants and Directions ofTechnical Change,” Research Policy, 1982,11(3), pp. 147–62.

———. “Sources, Procedures, and MicroeconomicEffects of Innovation,” J. Econ. Lit., Sept. 1988,26(3), pp. 1120–71.

DOSI, GIOVANNI ET AL., eds. Technical changeand economic theory. London: Pinter Pub.,1988.


DOSI, GIOVANNI; TEECE, DAVID AND WINTER,SIDNEY. “Towards a Theory of CorporateChange: Preliminary Remarks,” in Technologyand enterprise in a historical perspective. Eds.:GIOVANNI DOSI, RENATO GIANNETTI, ANDPIER ANGELO TONINELLI. Oxford: ClarendonPress, 1992, pp. 185–211.

DOUGLAS, MARY. How institutions think.Syracuse: Syracuse U. Press, 1986.

DURHAM, WILLIAM H. Coevolution: Genes, cul-ture, and human diversity. Stanford: StanfordU. Press, 1991.

EGGERTSSON, THRAINN. Economic behavior andinstitutions. Cambridge: Cambridge U. Press,1990.

ELLIOT, DONALD. “The Evolutionary Tradition inJurisprudence,” Columbia Law Rev., Jan. 1985,85, pp. 38–94.

FLIGSTEIN, NEIL. The transformation of corpo-rate control. Cambridge: Harvard U. Press,1990.

FONTANA, W. AND BUSS, LEO W. “What WouldBe Preserved if the Tape Were Played Twice.”Santa Fe, NM: Santa Fe Institute mimeo, 1992.

FREEMAN, CHRISTOPHER. The economics of in-dustrial innovation. London: Penguin, 1974.

———. “The Nature of Innovation and the Evolu-tion of the Productive System,” in Technologyand productivity. Paris: OECD, 1991, pp. 303–14.

FRIEDMAN, DANIEL. “Evolutionary Games inEconomics,” Econometrica, May 1991, 59(2),pp. 637–66.

FRIEDMAN, MILTON. Essays in positive econom-ics. Chicago: U. of Chicago Press, 1953.

GILSON, RONALD J. AND ROE, MARK J. “Under-standing the Japanese Keiretsu: Overlaps Be-tween Corporate Governance and IndustrialOrganization,” Yale Law J., Jan. 1993, 102(4),pp. 871–906.

GORT, MICHAEL AND KLEPPER, STEVEN. “TimePaths in the Diffusion of Product Innovations,”Econ. J., Sept. 1982, 92(367), pp. 630–53.

GOULD, STEPHEN J. The panda’s thumb: More re-flections on natural history. New York, Norton,1980.

———. The flamingo’s smile: Reflections in natu-ral history. New York: Norton, 1985.

GRANOVETTER, MARK. “Economic Action and So-cial Structure,” Amer. J. Sociology, Nov. 1985,91(3), pp. 481–510.

HANNAN, MICHAEL T. AND CARROLL, GLENN R.Dynamics of organizational populations. NewYork: Oxford U. Press, 1992.

HANNAN, MICHAEL J. AND FREEMAN, JOHN. Or-ganizational ecology. Cambridge: Harvard U.Press, 1989.

HARRISON, BENNETT. “Industrial Districts: OldWine in New Bottles,” Reg. Stud., 1992, 26(5),pp. 469–83.

HAYEK, FRIEDRICH. The fatal conceit: The errorsof socialism. Chicago: U. of Chicago Press,1988.

HENDERSON, REBECCA. “Underinvestment and

Incompetence as Responses to Radical Innova-tion: Evidence from the PhotolithographicAlignment Equipment Industry,” Rand J. Econ.,Summer 1993, 24(2), pp. 248–70.

HENDERSON, REBECCA AND CLARK, KIM. “Archi-tectural Innovation: The Reconfiguration of Ex-isting Product Technologies and the Failure ofEstablished Firms,” Admin. Sci. Quart., 1990,35(1), pp. 9–30.

HIRSHLEIFER, JACK, AND MARTINEZ-COLL, JUANCARLOS. “What Strategies Can Support theEvolutionary Emergence of Cooperation,” J.Conflict Resolution, June 1988, 32(2), pp. 367–98.

HODGSON, GEOFFREY M. Economics and institu-tions. Cambridge: Polity Press, 1988.

———. “Economic Evolution: Intervention Con-tra Pangloss,” J. Econ. Issues, June 1991, 25 (2),pp. 519–33.

———. “Thorstein Veblen and Joseph Schumpe-ter on Evolutionary Economics.” ZIF ResearchGroup, Biological Foundations of Human Cul-ture. U. of Bielefeld, Germany, 1991/92.

———. Economics and evolution: Bringing lifeback into economics. Cambridge: Polity Press,1993.

HOLLAND, JOHN H. ET AL. Induction: Processes ofinference, learning, and discovery. Cambridge:MIT Press, 1986.

HUGHES, THOMAS P. Networks of power: Electri-fication in Western society, 1880–1930.. Balti-more: Johns Hopkins U. Press, 1983.

HULL, DAVID. Science as a process. Chicago: U.of Chicago Press, 1988.

IWAI, KATSUHITO. “Schumpeterian Dynamics PartI,” J. Econ. Behav. Organ., June 1984, 5(2), pp.159–90.

———. “Schumpeterian Dynamics Part II,” J.Econ. Behav. Organ., Sept.-Dec. 1984, 5(3–4),pp. 321–51.

KANDORI, MICHIHIRO; MAILATH, GEORGE J.AND ROB, RAFAEL. “Learning, Mutation, andLong Run Equilibria in Games,” Econometrica,Jan. 1993, 61(1), pp. 29–56.

KATZ, MICHAEL AND SHAPIRO, CARL. “SystemsCompetition and Network Effects,” J. Econ.Perspectives, 1994, 8, pp. 93–116.

KLEPPER, STEVEN. “Entry, Exit, Growth, and In-novation Over the Product Cycle.” Carnegie-Mellon U., unpublished manuscript, 1993.

KLEPPER, STEVEN AND GRADDY, ELIZABETH.“The Evolution of New Industries and the De-terminants of Market Structure,” Rand J. Econ.,Spring 1990, 21(1), pp. 27–44.

KUHN, THOMAS S. The structure of scientific revo-lutions. Chicago: U. of Chicago Press, 1970.

LAKATOS, IMRE. “Falsification and the Methodol-ogy of Scientific Research Programmes,” inCriticism and the growth of knowledge. Eds.:IMRE LAKATOS AND ALAN MUSGRAVE. NewYork: Cambridge U. Press, 1970, pp. 91–196.

LANDES, WILLIAM M. AND POSNER, RICHARD A.The economic structure of tort law. Cambridge:Harvard U. Press, 1987.


LANE, DAVID. “Artificial Worlds and Economics,”J. Evolutionary Econ., 1993, 3 (11), pp. 89–108.

LANGLOIS, RICHARD N., ED. Economics as a pro-cess: Essays in the new Institutional Economics.New York: Cambridge U. Press, 1986.

———. “Transaction Cost Economics in RealTime,” Industrial and Corporate Change, 1992,1 (1), pp. 99–127.

LANGLOIS, RICHARD AND EVERETT, MICHAEL.“What is Evolutionary Economics.” Storrs, CT:U. of Connecticut, unpub. ms., 1992.

LATOUR, BRUNO. Science in action. London: Mil-ton Keynes Press, 1986.

LAZONICK, WILLIAM. Competitive advantage onthe shop floor. Cambridge: Harvard U. Press,1990.

LEWONTIN, RICHARD C. The genetic basis of evo-lutionary change. New York: Columbia U.Press, 1974.

LUMSDEN, CHARLES J. AND WILSON, EDWARDO. Genes, mind, and culture. Cambridge: Har-vard U. Press, 1981.

MAGNUSSON, LARS, ed. Evolutionary and neo-Schumpeterian approaches to economics. Lon-don: Kluwer Academic Pub., 1994.

MALERBA, FRANCO AND ORSENIGO, LUIGI. “In-dustry Evolution and Artificial Worlds.” Paperpresented for the Workshop on Artificial WorldModels. Santa Fe Institute. Santa Fe, NM, Nov.11–14 1993.

———. “The Dynamics and Evolution of Indus-tries.” Ms. Dept. of Economics, Bocconi U.,1994.

MARCH, JAMES G. AND SIMON, HERBERT A. Or-ganizations. New York: Wiley, 1958.

MARSHALL, ALFRED. Principles of economics. 8thed. London: Macmillan, 1948.

MAYNARD SMITH, J. Evolution and the theory ofgames. Cambridge: Cambridge U. Press, 1982.

MAYR, ERNST. Toward a new philosophy of biol-ogy. Cambridge: Harvard U. Press, 1988.

MCGUIRE, PATRICK; GRANOVETTER, MARK ANDSCHWARTZ, MICHAEL. “Thomas Edison and theSocial Construction of the Early Electrical In-dustry in the United States,” in Explorations ineconomic sociology. Ed.: RICHARD SWEDBERG.New York: Russell Sage, 1993.

METCALFE, STANLEY. “The Diffusion of Innova-tion: An Interpretative Survey,” in GiovanniDosi et al. eds. 1988, pp. 560–89.

———. “Variety, Structure, and Change: An Evo-lutionary Perspective on the Competitive Pro-cess,” Revue D’Economie Industrielle, 1992,(59), pp. 46–61.

METCALFE, STANLEY AND GIBBONS, MICHAEL.“Technology, Variety and Organization,” in Re-search on technological innovations, manage-ment and policy. Eds.: RICHARD S. ROSEN-BLOOM AND ROBERT A. BURGELMAN. JAIPress, 1989, 4, pp. 153–93.

MILGROM, PAUL AND ROBERTS, JOHN, “The Eco-nomics of Modern Manufacturing,” Amer. Econ.Rev., June 1990, 80(3), pp. 511–28.

MOKYR, JOEL. The lever of riches. New Yord: Ox-ford U. Press, 1990.

MUELLER, DENNIS. Public choice II. Cambridge:Cambridge U. Press, 1989.

MUELLER, DENNIS AND TILTON, JOHN. “Re-search and Development Costs as Barriers toEntry,” Can. J. Econ., Nov. 1969, 2(4), pp. 570–79.

NACHBAR, JOHN. “Evolution in the Finitely Re-peated Prisoner’s Dilemma,” J. Econ. Behav.Organ., Dec. 1992, 19(3), pp. 307–26.

NELSON, RICHARD R. AND ROSENBERG,NATHAN. “Technical Innovation and NationalSystems,” in National innovation systems: Acomparative study. Ed.: RICHARD R. NELSON.New York: Oxford U. Press, 1993, pp.

NELSON, RICHARD R. AND WINTER, SIDNEY.“Neoclassical vs. Evolutionary Theories of Eco-nomic Growth: Critique and Prospectus,” Econ.J., Dec. 1974, 84(336), pp. 886–905.

———. “In Search of a Useful Theory of In-novation,” Research Policy, 1977, 6(1), pp. 36–76.

———. An evolutionary theory of economicchange. Cambridge: Harvard U. Press, 1982.

NORTH, DOUGLASS C. Institutions, institutionalchange, and economic performance. Cambridge:Cambridge U. Press, 1990.

PENROSE, EDITH. “Biological Analogies in theTheory of the Firm,” Amer. Econ. Rev., Dec.1952, 42(5), pp. 804–19.

PEREZ, CARLOTTA. “Structural Change and theAssimilation of New Technology in the Eco-nomic and Social System,” Futures, 1981.

PERROW, CHARLES. Complex organizations: Acritical essay. 3rd ed. Glenview, IL: ScottForesman, 1986.

PETROSKI, HENRY. The evolution of useful things.New York: Alfred Knopf, 1993.

PIORE, MICHAEL J. AND SABEL, CHARLES F. Thesecond industrial divide: Possibilities for pros-perity. New York: Basic Books, 1984.

PLOTKIN, HENRY C. Learning, development, andculture: Essays in evolutionary epistemology.New York: John Wiley, 1982.

POPPER, KARL R. Conjectures and refutations:The growth of scientific knowledge. New York:Harper Torchbooks, 1968.

POWELL, WALTER W. AND DIMAGGIO, PAUL J.,EDS. The New institutionalism in organiza-tional analysis. Chicago: U. of Chicago Press,1991.

PRIGOGINE, ILYA AND STENGERS, ISABELLE. Or-der out of chaos. London: Fontana, 1984.

ROMANELLI, ELAINE. “The Evolution of New Or-ganizational Forms,” Annual Review of Sociol-ogy, 1991, 17, pp. 79–103.

ROMER, PAUL. “Increasing Returns and New De-velopments in the Theory of Growth,” in Equi-librium theory and applications. Eds.: WILLIAMBARNETT ET AL. Cambridge: Cambridge U.Press, 1991, pp. 83–110.

ROSENBERG, NATHAN. Perspectives on technol-ogy. Cambridge: Cambridge U. Press, 1976.


———. Inside the black box: Technology andeconomics. Cambridge: Cambridge U. Press,1982.

ROSENBERG, NATHAN AND BIRDZELL, LUTHERE. How the West grew rich. New York: BasicBooks, 1986.

ROSENKOPF, LORI AND TUSHMAN, MICHAEL.“The Coevolution of Technology and Organiza-tion,” in JOEL A. C. BAUM AND JITENDRA V.SINGH 1994.

RUBEN, P. AND BAILEY, M. “A Positive Theory ofLegal Change.” Dept. of Economics, Emory U.,Atlanta, GA. Unpub. ms., 1992.

RUMELT, RICHARD P. “How Much Does IndustryMatter?” Strategic Management J., 1991, 12(3),pp. 167–85.

SABEL, CHARLES AND ZEITLIN, JONATHAN. “Sto-ries, Strategies, Structures: Rethinking Hist-orical Alternatives to Mass Production.” ms.,1993.

SAHAL, DEVENDRA. Patterns of technologicalinnovation. Reading, MA: Addison-Wesley,1981.

SAVIOTTI, PAOLO AND METCALFE, J. STANLEY,EDS. Evolutionary theories of economic andtechnological change. Reading, MA: HarwoodAcademic Pub., 1991.

SCHOEMAKER, PAUL J. “The Quest for Optimality:A Positive Heuristic of Science?” Behavioraland Brain Sciences, June 1991, 14(2), pp. 205–45.

SCHOTTER, ANDREW. The economic theory of so-cial institutions. Cambridge: Cambridge U.Press, 1981.

———. “The Evolution of Rules,” in RICHARDLANGLOIS, ed. 1986, pp. 117–33.

SCOTT, W. RICHARD. Organizations: Rational,natural, and open systems. 3rd ed. EnglewoodCliffs, NJ: Prentice Hall, 1992.

SCHUMPETER, JOSEPH A. Capitalism, socialism,and democracy. 5th ed. London: George Allen& Unwin, [1942] 1976.

SILVERBERG, GERALD. “Technical Progress, Capi-tal Accumulation, and Effective Demand: ASelf Organizing Model,” in Economic evolutionand structural adjustment. Eds.: DAVID BAT-TEN, JOHN CASTI, AND BORJE JOHNSEN. Ber-lin: Springer-Verlag, 1987, pp. 116–44.

SILVERBERG, GERALD; DOSI, GIOVANNI ANDORSENIGO, LUIGI. “Innovation, Diversity andDiffusion: A Self Organizing Model,” Econ. J.,Dec. 1988, 98(393), pp. 1032–54.

SIMON, HERBERT. Administrative behavior. NewYork: Free Press, 1947 .

SIMON, HERBERT AND BONINI, CHARLES. “TheSize Distribution of Business Firms,” Amer.Econ. Rev., Sept. 1958, 48, pp. 607–17.

SMELSER, NEIL AND SWEDBERG, RICHARD,EDS. Handbook of economic sociology. Prince-ton: Princeton U. Press, 1994.

SOBER, ELLIOTT, ED. Conceptual issues in evolu-tionary biology. Cambridge: MIT Press, 1984.

SOETE, LUC AND TURNER, ROY. “TechnologyDiffusion and the Rate of Technical

Change,” Econ. J., Sept. 1984, 94(375), pp.612–23.

SOLOW, ROBERT. “A Contribution to the Theoryof Economic Growth,” Quart. J. Econ., Feb.1956, 70, pp. 65–94.

SPENCER, HERBERT. The factors of organic evolu-tion. London: Williams & Norgate, 1887.

SUGDEN, ROBERT. “Spontaneous Order,” J. Econ.Perspectives, Fall 1989, 3(4), pp. 85–97.

TUSHMAN, MICHAEL L. AND ANDERSON, PHILIP.“Technological Discontinuities and Organiza-tional Environments,” Administrative SciencesQuart., 1986, 31, pp. 439–65.

TUSHMAN, MICHAEL L. AND ROMANELLI,ELAINE. “Organizational Evolution: A Meta-morphis Model of Convergence and Reorienta-tion,” in Research in organizational behavior.Eds.: B. M. STAW AND LARRY L. CUMMINGS.Greenwich, CT: JAI Press, 1985, 7, pp. 171–222.

TUSHMAN, MICHAEL AND ROSENKOPF, LORI.“Organizational Determinants of TechnologicalChange: Toward a Sociology of TechnologicalEvolution,” Research in organizational behav-ior, 1992, 14, pp. 311–47.

UTTERBACK, JAMES. Mastering the dynamics ofinnovation. Boston: Harvard Business SchoolPress, 1994.

UTTERBACK, JAMES AND SUAREZ, FERNANDO.“Innovation, Competition, and Market Struc-ture,” Research Policy, 1993, 22(1), pp. 1–21.

VEBLEN, THORSTEIN. “Why Is Economics Not anEvolutionary Science?,” Quart. J. Econ., 189812, pp. 373–97.

———. Imperial Germany and the industrialrevolution. New York: Macmillan, 1915.

VERSPAGEN, BART. “Endogenous Innovationin Neo-classical Growth Models: A Survey,”J. Macroeconomics, Fall 1992, 14, pp. 631– 62.

VINCENTI, WALTER. What engineers know andhow they know it? Baltimore: Johns HopkinsPress, 1990.

———. “The Retractable Airplane Landing Gearand the Northrup Anomaly: Variation-Selectionand the Shaping of Technology,” Technologyand culture, 1994, 35, pp. 1–33.

WILLIAMSON, OLIVER E. The economic institu-tions of capitalism. New York: Free Press, 1985.

WINTER, SIDNEY G. “Economic ‘Natural Selec-tion’ and the Theory of the Firm,” Yale Econ.Essays, 1964, 4(1), pp. 225–72.

———. “Schumpeterian Competition in Alter-native Technological Regimes,” J. Econ. Be-hav. Organ., Sept.-Dec. 1984, 5(3–4), pp. 287–320.

———. “The Research Program of the BehavioralTheory of the Firm: Orthodox Critique andEvolutionary Perspective,” in Handbook of be-havioral economics. Volume A. Eds.: BENJAMINGILAD AND STANLEY KAISH. Greenwich, CT:JAI Press, 1986, pp. 242–45.

———. “Comments [Rationality of Self and Oth-ers in an Economic System] [Adoptive Behavior


and Economic Theory],” in J. Bus., Part 2. Oct.1986, 59(4), pp. S427–34.

WITT, ULRICH. “The Evolution of Economic Insti-tutions as a Propagation Process,” PublicChoice, Aug. 1989, 62(2), pp. 155–72.

———, ed. Evolutionary economics. London: Ed-ward Elgar, 1993.

YOUNG, H. PEYTON. “The Evolution of Conven-tions,” Econometrica, Jan. 1993, 61(1), pp. 57–84.


Documents

Recent Evolutionary Theorizing About Economic Change jel1995.pdf · economists have even tried. (For a splen-did history of evolutionary theorizing in economics, see Geoffrey Hodgson