Remarks

Sidney G. Winter
The Wharton School

Jaime Serra mentioned last night that it was his good fortune, when he came to Yale, to encounter Jim Tobin teaching graduate micro.

When I got to Yale almost two decades earlier, it was my good fortune to encounter Willie Fellner teaching graduate micro. I learned a lot from Jim in other settings, but not in that course. Two years later, however, I did get an important indirect benefit from the fact that Jim was then teaching micro. I found myself on a research fellowship at The Brookings Institution in Washington, puzzling with what I was learning about corporate decision making on R&D – which didn’t seem to correspond closely to the optimal decision models I had in my tool kit. At that point, somebody brought to town a copy of Jim Tobin’s then-current reading list in graduate micro. I set about reading some of the things on that list that I had not previously encountered – one of them being Armen Alchian’s article, “Uncertainty, Evolution and Economic Theory.” Reading that article, against the background of my empirical puzzle about corporate R&D, and the deeper background of Jascha Marschak’s Yale seminar on Decision and Organization, was an event that shaped my intellectual life. That shaping effect will be strongly reflected in my remaining remarks.

One of my long-standing interests is in the character of productive knowledge in organizations: what is the reality behind the everyday fact that organizations know how to do things? This theme, which is of central importance in the evolutionary economics approach developed by Dick Nelson and myself, has in recent years become increasingly prominent in business practice. First under the heading of quality management, and more recently under the broader heading of knowledge management, business firms have explored a variety of methods and organizational approaches to expanding and strengthening their productive knowledge. Both my long-standing research focus and my position in the management department of a business school have led me to attend to these developments with considerable interest. Today I want to describe how these developments in business practice fit within the general view of production taken in evolutionary economics.

For the sake of clarity, let me stick close to the orthodox language as I expound that view briefly. The basic concept is the individual technique, not the production set or function. Individual techniques are conceived to be realized through the performance of organizational routines, which are learned and practiced patterns of behavior. Some collections of routines are keyed to respond automatically to the environment, in ways that correspond to the smooth movement along a short-run production function in standard theory. Aside from that possible correspondence, however, there are no production functions or sets recognized in evolutionary theory. There are techniques and choices of technique, especially the basic comparison between the status quo technique performed through existing routines and some alternative to it. But there are no well-defined production sets and no optimizations -- unless you want to call any goal-oriented analysis of a choice between two alternatives an "optimization.

It has long been amazing to me that the production function and optimization tools are clung to so tightly by economists, given that they offer so little grip on a world where technical opportunity sets are changing rapidly. While many share this perception, it obviously has not swept through the discipline like wildfire.

In the evolutionary approach, an important question is "where is the knowledge stored?" – an issue ignored in the mainstream treatment. A full answer is inevitably complex. But an important locus of storage is the memories of the personnel who perform the relevant routines, also their machines and tools, and the physical layout of the production site. Nowadays, computer programs are an increasingly important form of storage. These loci of storage -- as contrasted with others that are not bound to the spatio-temporal setting of a particular organization -- are particularly crucial to the margins of superiority that account for competitive advantage, or for the height of the entry barriers protecting incumbent firms. Routines are “remembered by doing” in a functioning organization, not stored primarily in supplements to human memories such as files, blueprints, and manuals (although these are sometimes important).

This analysis implies that productive knowledge can be lost from an organization in various ways, and an important part is definitely lost when the routine is discontinued and personnel and machines are dispersed. Of course, it may be possible to reconstitute a very similar routine with different personnel and machines and so forth. Such an effort requires new learning and will be successful more or less quickly depending on the height of the knowledge platform from which it starts. An example of a high platform is a similar routine that is actually functioning elsewhere in the same organization; a low platform is a manual describing only the technical aspects of the routine.

An important empirical issue illuminated by the evolutionary account is the existence of very marked heterogeneity in techniques almost anywhere you look -- anywhere, that is, except where very determined efforts are made to stamp it out. It appears that substantial heterogeneity is the normal state of affairs, and approximate uniformity is achieved only with considerable effort. Why might this be? It is not so mysterious once one realizes that productive techniques are in important respects learned anew every time they are initiated in a particular locale; we sometimes say that routines are "always home grown." There is inevitably a distance between the immediately available performance associated with the best available knowledge platform and the level of performance that is attainable after many repetitions -- and the process filling that gap is idiosyncratic, path dependent learning.

One could think of the stochastic process governing performance during learning as a random walk with drift; individual steps represent solutions to the innumerable details that are settled independently across instantiations of the routine. As the process goes on, the performance realizations fan out. A more probing analysis of the sources of persistent heterogeneity recognizes the important role played by complexity, itself rooted in multiple tight complementarities among different aspects of the process. Such complexity produces “rugged landscapes”in the space of process parameters. The result is multiple local maxima in performance (“adaptive peaks”), which limits the effectiveness of local adaptation as a means for improving performance. The different learning processes of different organizations, starting from different initial conditions, can lead them to find, and stay on, different peaks.

A related question is where and when learning stops. Frequently there is no readily referenced standard for the adequacy of a particular process, other than the threshold question of whether usable output appears. Almost by default, the satisficing principle describes the general character of the stopping process. “The initial learning phase ends or fades away when performance that is deemed satisfactory is achieved... The pressure for improvement falls because performance reaches a level that satisfies criteria deriving from general considerations that are remote from the costs and benefits of further improvement in the particular routine – e.g., the satisfactory performance of the organization as a whole, or market acceptance of the final product. It is entirely possible that attractive opportunities for further improvement lie just around the corner when the search for them is abandoned.”(Winter, 1994).

The validity of this account of production is attested by a number of management techniques that have become prominent in recent years and that address issues flagged by evolutionary theory. Benchmarking, for example, involves the systematic attempt to learn and improve by systematic comparison of an organization’s way of performing a specific task with the method employed by another organization -–typically one that is more strongly focused on that task, or otherwise believed to be superior at it (and typically not a rival). Internal benchmarking, or internal transfer of best practices, is the same technique applied within the same organization. Internal benchmarking is, according to one survey, the most widely used of all knowledge management practices. These techniques seek to exploit for learning purposes the natural heterogeneity of problem solutions that are independently arrived at, heterogeneity often magnified by differential incentives to attend to solution quality. The connection to the evolutionary emphasis on variety should be obvious.

The quest for continuous improvement is essentially an organizational device for challenging the hegemony of the satisficing principle in governing learning. Through close quantitative monitoring of performance details, “root-cause” analysis of recurrent problems, and the continuing upward adjustment of goals, it is often possible to keep process efficiencies moving upward for a sustained period – the stock of problems to solve does not seem to get depleted very fast.

An important part of the continuous improvement/quality management approach is the training and empowerment of line employees to analyze process problems, make local adjustments, and propose broader adjustments. Fundamental to this approach is the idea that “the people who know the work best are those who perform it,” an idea that is fully consistent with the evolutionary economics image of the distributed character of knowledge in an organization – but quite inconsistent with the standard economics view that treats available production techniques as the property of the firm per se and confines economic analysis to incentive issues. Whereas many economists continue to analyze the employment relation largely in terms of mitigating “shirking” through monitoring and appropriate incentives, human relations experts preach the doctrine of a committed workforce, and many firms demonstrate the effectiveness of that doctine.

What are the implications of this analysis – apart from the point about the limited usefulness of textbook economics in understanding productive organizations? Let me make two points.

First, at the broadest level the evolutionary analysis of production underscores very strongly the role of path dependence -- the fact that the technology and organization of today are the result of a historical process of accumulation in which a myriad of solutions were found to a myriad of small problems. Thus, if we think about the problems of controlling greenhouse gases, or delivering quality health care at reasonable cost, or using computers effectively in education, it is important to realize that the consequences of a major shift in attention and incentives accumulate over time, more or less indefinitely, in a fashion hardly suggested by a price line rolling around a transformation curve. If it is very hard to actually predict where such change processes will take us, it should be somewhat less hard to stop pretending that we can predict successfully by fitting production functions to past data.

Second, it is obvious that improvements in technique at a very micro level are a source of productivity change and contribute to a positive residual in a growth accounting exercise. Apart from that presumed appearance in the residual, however, there is little quantitative trace of this activity. It is not typically measured as R&D expenditure, nor does it involve accumulation of physical or human capital as those are conventionally measured. There is even less direct trace of meta-level innovations like statistical process control or total quality management or modern human resource management approaches that underlie the concrete activities of organizational problem solving. Still more elusive are the even higher level sources of progress, such as generally shared managerial aspirations involving more ambitious ideas of what it means to manage well.

So it is difficult to make a systematic and convincing case for the importance of this activity. Impressionistically, it seems to me quite important, and is perhaps a significant factor in the recent remarkable performance of the American economy. While I began my career with the notion that business decision making practice was often much simpler than economists believed, I now am often impressed with the opposite observation: business practice often reflects a subtler understanding of the problems than is found in economics – based in a very different conception of what the problems are. When I encounter what appears to be obvious mismanagement in the practices of business firms, as I often do, my comment is different. I used to say, “there is the reality of business decision making,” but I now say “why haven’t they got the word yet?

Related Reading

Adler, P. (1993), "The learning bureaucracy: New United Motor Manufacturing, Inc.,"Research on Organizational Behavior, 111-194.

Argote, L. and E. Darr (1999 forthcoming), "Repositories of knowledge in franchise organizations: individual, structural and technological," in The Nature and Dynamics of Organizational Capabilities. G. Dosi, R. R. Nelson and S. G. Winter, Eds. New York: Oxford University Press.

Levinthal, D. (1997), “Adaptation on rugged landscapes,” Management Science, 43, 934-950.

Nelson, R.R. (1980), "Production sets, technological knowledge and R and D: fragile and overworked constructs for analysis of productivity growth?," American Economic Review, 70, 62-67.

Nelson, R. R. and S. G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, MA: Harvard University Press.

Szulanski, G. (1996), “Exploring internal stickiness: impediments to the transfer of best practice within the firm,” Strategic Management Journal, 17, 27-43.

Tyre, M. J. and W. J. Orlikowski (1994), “Windows of opportunity - temporal patterns of technological adaptation in organizations,” Organization Science, 5, 98-118.

Winter, S.G. (1982), "An essay on the theory of production," Economics and the World Around It, S.H. Hymans, ed., Ann Arbor: University of Michigan Press.

Winter, S. G. (1994), "Organizing for continuous improvement: evolutionary theory meets the quality revolution," Evolutionary Dynamics of Organizations. J. A. C. Baum and. J. V. Singh, Eds. New York: Oxford University Press.