by Winter, Sidney G. · 2003

On October 16, 2003, Richard Nelson gave a speech at conference honoring the work of Sidney Winter. The conference was held at the Wharton School of the University of Pennsylvania. On the 30th anniversary of Nelson and Winter (1982), Etss.net is pleased to publish this speech for the first time. In it, Richard Nelson sheds light on how Nelson and Winter’s landmark 1982 book came about.

A few year earlier Sid Winter already provided some insights about their collaboration. He reminisced about “The Evolution of Dick Nelson” at a conference held at Columbia.

Essay by Nelson, Richard R

Origins and Factors Shaping Our Joint Work Developing an Evolutionary Theory of Economic Change

Sid did his graduate work in economics at Yale University during the middle 1950s, as I had a few years earlier. We both came to RAND in the late 1950s, both went to work for the Kennedy administration when it came to power, and both returned to RAND in the middle 1960s. Then Sid went to the University of Michigan, and I went to Yale University. A few years later, in the mid– 1970s, Sid joined me at Yale, where we were together for about a decade. The development of our evolutionary theory took place gradually and sequentially in all of these places and contexts. We worked from a basis of shared experience that was broad and deep. When we were taking our Ph.D. training, economics was a much more eclectic, and catholic, discipline than it became later on. The environment at Yale was particularly broad. In contrast with most Economics Ph.D. programs in recent years, when we were graduate students at Yale, economic history was a required field of study. Many of us studied the history of economic thought. Economic theory was taught in that context. The orientation to economics as a discipline at Yale was very much that the goal of the subject was the understanding of real empirical phenomena, and that an important use of that understanding was to guide policy making to improve the human condition. Good economic theory was an important aspect of that understanding. But it was not all of that understanding; a good economist also ought to know a considerable amount of economic history, be a good empirical analyst, and also have a strong common–sense understanding of the economic world. There were no immutable rights and wrongs regarding the particular form a theory ought to take.

While neoclassical models based on the assumption that the individual and organizational actors understood the decision problems they faced and chose rationally were part of what we learned, we also studied theories, like that of Keynes, that did not make that assumption. At that time the absence of neoclassical behavioral foundations in Keynes was not viewed as a problem. Of course much later many economists came to view this as a major flaw. Mathematical economics was taught as a separate subject. The view of Tobin, Koopmans, Marschak, and others on the Yale faculty who were proficient in mathematics, was that mathematics often was a useful and powerful theoretical tool. There was considerable openness to other kinds of theorizing, and considerable respect for economists whose expertise consisted largely of detailed empirical and institutional knowledge. The ambience at RAND during the 1950s and 1960s was strongly consonant with this point of view. At that time RAND was marked by a very strong orientation to important policy problems and their solution, and also by a methodological openness regarding the use and development of tools and ideas. Perhaps because of the importance of the problems being worked on — principally but not exclusively related to national security — while methodological rigor was required, the researchers knew that their principal task was to get the problem right, and to illuminate real solutions to the real problem. Both Sid and I naturally took to this orientation, and we also both fell under the influence of Burton Klein. Klein was a remarkable man. When we first met him he was head of the RAND project on R & D management and strategy, and while later Burt became head of the RAND economics department, it was as a scholar of R & D and innovation that Sid and I best remember him. Both of us were very much interested in R & D and technical change. When Sid started his thesis work, he intended it to be about the determinants of R & D spending by firms. From the beginnings of my career, my central interest was in long–run economic change, and by the time I left Yale I was committed to the proposition that technological advance was the key driving force. Thus we were naturally attracted to Burt, and his projects. Perhaps more than anything else, Burt convinced us of the importance of uncertainty in the doing of R & D, in R & D planning and management, and in the processes of technological advance more generally. In all of these contexts, Burt argued that it was impossible to truly understand the full range of alternatives, the consequences of choosing any particular one, or even all of the essential values at stake. Since our mother discipline, economics, was gravitating at that time toward a theory of individual and organizational behavior in which the actors were fully rational, in the sense that they did understand all of these things, Klein’s insistence on the centrality of uncertainty cut across that grain. Sid and I imbibed Klein’s point of view on the nature and importance of real uncertainty in many economic contexts. While we didn’t fully recognize it at the time, in doing so, we clearly doomed ourselves ultimately to be viewed as heretics within our mother profession. Klein also drew us into study of Schumpeter. Schumpeter clearly believed that innovation involved uncertainty, in the sense that Klein was using the term, in an essential way. Schumpeter also provided a view of the processes of economic change, of competition in industries where innovation was important, and of the problems of decision–making in firms in such industries, that rang right. When later we began explicitly to try to develop an evolutionary theory of economic change, we recognized that we were trying to develop a rigorous version of Schumpeter, and we said so.

The intellectual ideas that were bubbling up at Carnegie Tech during the 1950s and 1960s, under the leadership of Herbert Simon, Richard Cyert, and James March, also were part of our shared intellectual heritage. For a variety of obvious reasons, there was considerable interest at RAND in understanding how large organizations behaved. The Carnegie school theories about this provided the intellectual basis for an informal organization theory seminar at RAND that met regularly during the middle 1960s. The theory of the firm later developed in our book rested heavily on the ideas of Simon, Cyert, and March. A key element of the Carnegie theory was that a large share of organizational behavior was guided by the decision rules —routines— that were operative at any time. There was little conscious deliberation involved in routine guided action. Both of us were strongly impressed both by the evidence provided by the Carnegie school researchers of the existence and nature of such routines, and by the ability of a routine oriented theory of organizational behavior to explain even the details of things like firm pricing. There clearly was a very different theory of behavior here than that provided by rational choice theory, and it worked empirically. Sid later adopted and put forth the pregnant idea that, regarding firm behavior, routines are genes. For me, the organizational routines idea meshed well with a conception of the nature of technologies I had been developing in my work. Technologies could be characterized as recipes. What Cyert and March gave to me was the notion that those technology recipes were organizational routines. One can see clearly the fruits of what we had learned, and the seeds of our evolutionary theory, in two papers written just before we left RAND. Sid’s paper, called “Toward an Neo– Schumpeterian Theory of the Firm,” reflected his continuing central interest in firm behavior.

Much of the theory of organizational capabilities and behavior put forth in our book grew out of this paper. My paper was called “A Diffusion Model of International Productivity Differences.” It presented a model in which economic growth was a disequilibrium process, in which superior new technology gradually replaces lower productivity, older technology. A good portion of the industry and economy–wide dynamics that we developed in our book was anticipated in this paper. In the late 1960s Sid went to Michigan, and I went to Yale. By that time we were talking seriously about pulling together his developing theory of dynamic firm behavior and my developing theory of technological change and productivity growth into “An Evolutionary Theory of Economic Capabilities and Behavior.” Sid, please correct me if I am wrong, but I think that was the first title we gave to the opus we planned to write. Our writing started then, and intensified when Sid came to Yale. Let me pull together and extend some of the shared intellectual commitments that went into the development of our evolutionary theory of economic change. First of all, economic theory is, or ought to be, about real phenomena. The development of good theory depends on theorists having a strong empirical understanding of the phenomena they are theorizing about. The notion that one theorizes first, and then tests ones theory against the empirical data, tends to put the cart before the horse, or at least to repress the interactive back and forth between conceptualization and observation that is essential in the development of reliable knowledge. Second, the assumption that actors are fully rational, built into neoclassical theory, often gets in the way of understanding. Sid had found this in his early attempts to develop a theory of the determinants of firm R & D spending. Early in my own work studying technical advance, I came to believe that the phenomena I was observing could not be understood within the intellectual structure of a theory that assumed that the actors correctly understood the pros and cons of different alternative paths, or even that they all saw the same set of paths. Third, both Sid and I, but particularly Sid, had reflected a lot on Milton Friedman’s argument that, while the actions of firms often seemed to be generated by more or less mechanical rules of thumb, it was good theory to assume that firms behave “as if” they maximized profits – good in the sense that that theory generates the right predictions of what the firms in fact do. We both came to the position that this argument is a snare and a delusion. The fact that a particular theory successfully predicts or explains certain things provides no reason to believe that that theory provides a correct explanation. There well may be other theories that explain these phenomena equally well, and are better grounded. And while Friedman argued that it is not important that a theory about the outcomes of a decision process specify the process correctly, if the process is in fact misspecified, that indicates clearly that there is something wrong about the theory, and that holding it can get in the way of understanding. More, if the basic causal explanation is not correct, there are good reasons to fear that the light cast by the theory may be distorted in significant ways. Both Sid and I were well aware of Tjalling Koopmans’ argument to this effect, and his proposition that if Friedman and other economists really believed that it was competitive selection that was explaining the observed behavior of surviving firms, they ought to develop an expressly evolutionary theory to explain that behavior. We agreed with Tjalling’s suggestion that, contrary to Friedman’s position, economic theory ought to be built on a specification of process that was at least roughly consisted with what was known about the actual processes at work. We also believed that there was a considerable amount of empirical data that could be used to identify, at least roughly, those processes. Here we were strongly influenced by the Carnegie School work, particularly the book by Cyert and March on A Behavioral Theory of the Firm, in which in a number of cases the researchers seemed to have identified the actual processes going on. And models that specified these processes seemed to do a very good job at generating, and explaining, behavior. We thought there was a general lesson here. Fourth, following along these lines, we felt the theoretical proposition that behavior is largely determined by a set of routines would provide a solid intellectual foundation for an economic evolutionary theory. Because we were interested in phenomena at the level of an industry or the economy as a whole, we knew that such a theory would need to incorporate the kind of routines built into the Cyert and March formulation in stylized and simplified form. And given our objectives, those routines had to be treated as endogenous rather than exogenous. In the first place, we wanted to address directly the question of whether the possession by firms of reasonably effective routines could be explained as the result of an evolutionary selection process. But also, we needed a theoretical structure in which change could be continuing if we were to include “technologies” among our routines. Thus we needed a theory of how new routines get generated, and how selection processes operate to eliminate certain routines, and increase the use of others. With this block of conceptions, we were on our way to developing an evolutionary theory of economic change. Both our personal central interests, and the particular building blocks we put in place as a start on such a theory, led us to focus much of the book on three broad (and connected) topics. One was a dynamic theory of firm capabilities and behavior; this was Sid’s center of interest. A second was an evolutionary theory of technological advance and of economic growth driven by technological change; this is where my interest was focused. The third topic combined these two elements into a theory of Schumpeterian competition and industrial dynamics in sectors where competition is largely through innovation. Both Sid and I find it highly gratifying that in each of these areas, an evolutionary perspective has taken deep root. Many of you have been in the forefront carrying on the research. I want to call attention to the fact that in each of these areas, what we wrote added to and strengthened a body of theorizing that already was attracting many scholars, and an evolutionary perspective was natural and welcomed. As I noted earlier, the Simon, Cyert, March tradition already had set the foundations for a serious theory of firm behavior. We helped to place that theory in a dynamic context, and to focus attention on the issue of firm capabilities. Scholars at the Science Policy Research Unit at the University of Sussex, and kindred analysts, already were developing a strong body of understanding regarding the processes of technological advance, within an analytic framework that was implicitly evolutionary. We made the evolutionary perspective explicit, sharpened it, and built it into a theory of economic growth. A number of economists long had believed that Schumpeter’s view of dynamic competition in industries where innovation was important was much more relevant than the view in neoclassical theory. We provided a theoretical structure within which those scholars, and their successors, could proceed confidently. I confess disappointment, however, and some hurt, at the hostility —disinterest would be much too mild a description— that our economic evolutionary theory has met within the general academic economics community. Maybe I should have expected that reaction, but I confess I didn’t. I am uncertain whether or not Sid did. With the wisdom of hindsight, it is clear that I should not have been surprised by the hostility we met. While the Yale economics department, in the years Sid and I were there as faculty, still had much of the pragmatism that it had when we were students, both of us suspected then that Yale was quite rare in its oneness to the kind of theorizing we were espousing. We certainly had a feeling that in much of economics that oneness was closing. The assumption that economic actors fully understood the context they were in, and that their “uncertainty” amounted to lack of full knowledge about the values that were or would be taken on by a few key variables, was in much of the discipline becoming the only legitimate way to theorize about behavior. But while I knew that, sort of, I still had faith in the pragmatism, the openness, regarding economic theory that I had imbibed in my days as a student at Yale, and as a member of the Yale economics faculty. If economic theory were regarded as a collection of models and tools that are useful in different applications, certainly we had provided some useful new ones. At the least we had, building on the Carnegie school, provided a theory of the firm that enabled economists to really study and come to understand firm behavior, if they chose. We offered a theory of technological advance and productivity growth that clearly was far superior to what neoclassical theory was offering, and a theory of Schumpeterian competition and industrial dynamics, areas where neoclassical theory was mute. But what we offered was not viewed in that light. The reason was that by that time economists had moved away from the “useful tool kit” view of theory to a position that economic theory ought to provide a consistent view of economic activity generally, and that neoclassical theory provided that view. It also is true that Sid and I did not present our work as merely a collection of models useful in particular contexts. We were, and are, advocating a more general philosophy about theorizing in economics. The heart of that philosophy is that as a general matter it is important to recognize the “bounded” aspect to human rationality. Where human action is particularly effective, this is a matter to be explained through processes of individual, group, and cultural learning, which themselves may need to be analyzed, rather than simply assumed. And in many cases behavior may be ineffective or worse. This is certainly not a “logical” (within the theory) impossibility, but a phenomena to be watched for. I am not sure of Sid’s view on this, but in my view, in addition to providing a superior way of conceptualizing economic dynamics, we demonstrated that theorizing about general economic phenomena could proceed quite effectively without the heavy baggage of the assumptions that the actors were fully rational, and the system as a whole was at or close to an equilibrium. The standard “reduced forms” in economics, like demand and supply curves, do not need neoclassical underpinnings. These, or equivalent concepts can be well understood in terms of behavioral concepts. This notion was not new. It was quite standard in economics prior to the 1970s. And this notion, I believe, has been the one most strongly resisted by contemporary economic theorists. I would like to say that my attitude toward this is “so much the worse for economics,” and it is in part. However, many of the topics dealt with by economics as a discipline are extremely important, and they are only partially treated in disciplines and communities of research that have taken aboard a behavioral and evolutionary perspective. And many young scholars taking their training in economics are interested in topics where an evolutionary perspective clearly is much more illuminating and powerful than a neoclassical one. So, while I could not be more gratified by the positive reception and impact our evolutionary theory has had in many places, I also am saddened that it has been thus far blocked out from our mother discipline, at least in the United States. I gather Sid will have something to say about this later in this conference. I look forward to hearing him.