The Structure of Evolutionary Models

The decision rules employed by firms form a basic operational con­ cept of our proposed evolutionary theory as well as contemporary orthodoxy. However, we rej ect the notion of maximizing behavior as an explanation of why decision rules are what they are; indeed, we dispense with all three components of the maximization model-the global objective function, the well-defined choice set, and the maxi­ mizing choice rationalization of firms’ actions. And we see “decision rules” as very close conceptual relatives of production “techniques,” whereas orthodoxy sees these things as very different.

Our general term for all regular and predictable behavioral pat­ terns of firms is “routine.” We use this term to include characteristics of firms that range from well-specified technical routines for pro­ ducing things, through procedures for hiring and firing, ordering new inventory, or stepping up production of items in high demand, to policies regarding investment, research and development (R&D) , or advertising, and business strategies about product diversification and overseas investment. In our evolutionary theory, these routines play the role that genes play in biological evolutionary theory. They are a persistent feature of the organism and determine its possible behavior (though actual behavior is determined also by the environ­ ment); they are heritable in the sense that tomorrow’ s organisms generated from today’ s (for example, by building a new plant) have many of the same characteristics, and they are selectable in the sense that organisms with certain routines may do better than others, and, if so, their relative importance in the population (industry) is aug­ mented over time.

Undoubtedly, there is a great deal of business behavior that is not, within the ordinary meaning of the term, “routine.” Equally clearly, much of the business decision making that is of the highest impor­ tance, both from the point of view of the individual firm and from that of society, is nonroutine. High-level business executives do not, in the modern world, spend humdrum days at the office applying the same solutions to the same problems that they were dealing with five years before. We do not intend to imply any denial of these propositions in building our theory of business behavior on the no­ tion of routine. For the purposes of economic theorizing, the key point is somewhat different. It is that most of what is regular and pre­ dictable about business behavior is plausibly subsumed under the heading “routine,” especially if we understand that term to include the relatively constant dispositions and strategic heuristics that shape the approach of a firm to the nonroutine problems it faces. The fact that not all business behavior follows regular and predictable patterns is accommodated in evolutionary theory by recognizing that

there are stochastic elements both in the determination of decisions and of decision outcomes. From the point of view of a participant in business decision making, these stochastic elements may reflect the result of tumultuous meetings or of confrontations with complex problems under crisis conditions; but from the viewpoint of an ex­ ternal observer seeking to understand the dynamics of the larger system, the significant point about these phenomena is that they are hard to predict. Conversely, if they were not hard to predict, the ob­ server would be inclined to interpret the tumult and the sense of crisis as some sort of organizational ritual-a part of the routine.

Our use of several different terms for different types of routines is meant to convey our appreciation that, for some purposes, it is im­ portant to distinguish between a production technique whose opera­ tion is tightly constrained by machinery or chemistry and procedures for choosing what technique to employ at a certain time, and also between a relatively low- order procedure or decision rule (for ex­ ample, the way a new order is handled or an inventory decline recog­ nized and responded to) and a higher-order decision rule or policy (for example, a rule to switch from use of oil to natural gas as fuel when the relative price ratio hits a certain level, or the custom of keeping advertising expenditures roughly in proportion to sales) . But, as the use of the common term “routine” indicates, we believe that these distinctions are subtle and continuous, not clear and sharp. Orthodox theory makes a sharp distinction between the choice set and choosing- between what is involved in operating a particular technique and what is involved in deciding what tech­ nique to use. In our evolutionary theory we see strong similarities in these. In mixing up batches of raw materials, decisions have to be made as to whether the composition and temperature are right or not, and, if not, what to do. If there is a rationale for orthodoxy’S pol­ icy of denying theoretical recognition to this element of choice in firm behavior by including it in the description of technique, it pre­ sumably has to do with the fact that the choices are made in a routin­ ized manner, and perhaps also that they are not an important source of variability in the firm’s profits . But empirical studies of pricing behavior, inventory management, and even advertising policies re­ veal a similar “by-the-rule” character of fi rm decision making in these arenas. In some cases, though not in all, routinization holds sway in particular deci sion-making arenas because the important ac­ tion is elsewhere-perhaps in finance, R&D policy, or coping with regulati on.4 Thus, orthodoxy’s unwillingness to give parallel treat­ ment to the similar forms of routinized behavior involved in “doing” and “choosing” remains a puzzle and will be a recurring theme in this book.

In  any  case,  evolutionary  modeling  highlights  the  similarities among different sorts of routines. At any time, a firm’s routines de­ fine a list of functions that determine (perhaps stochastically) what a firm does as a function of various external variables (principally market conditions) and internal state variables (for example, the firm’ s prevaili ng stock of machinery, or the average profit rate it has earned in recent periods) . Among the functions thus defined might be one that relates inputs required to output produced (reflecting the firm’ s technique), one that relates the output produced by a firm to market conditions (the supply curve of orthodox theory), and one that relates variable input proportions to their prices and other vari­ ables . But whereas in orthodox theory the available techniques are a constant datum, and decision rules are assumed to be the conse­ quence of maximization, in evolutionary theory they are treated as simply reflecting at any moment of time the historically given rou­ tines governing the actions of a business firm.

Although the routines that govern behavior at any particular time are, at that time, given data, the characteristics of prevailing routines may be understood by reference to the evolutionary process that has molded them. For the purposes of analyzing that process, we find it convenient to distinguish among three classes of routines.

One of these relates to what a firm does at any time, given its pre­ vailing stock of plant, equipment, and other factors of production that are not readily augmented in the short run. (In effect here we are defining the basic unit “period” in our evolutionary modeling, as a counterpart to Marshall’s “short run.”) These routines that govern short-run behavior may be called “operating characteristics.”

A second set of routines determine the period-by-period augmen­ tation or diminution of the firm’s capital stock (those factors of pro-duction that are fixed in the short run). The extent to which actual in­ vestment behavior follows predictable patterns probably varies a good deal from one situation to another. In some cases the decision making surrounding the question of whe ther to build a new plant may not be much different in kind from the decision making regarding whether or not to continue to run a particular machine that has been operating roughly, or to stop it and call in the maintenance crew. In other cases, the new plant decision may be more like a deci­ sion to undertake a major R&D program on a recently opened tech­ nological frontier, a problem without real precedent that is dealt with through improvised procedures. Which of the two patterns obtains probably depends importantly on the size of the investment project relative to the existing activity of the firm. As suggested above, this spectrum of realistic possibilities corresponds in evolutionary theory to a range of differing roles for stochastic elements in the represen­ tation of investment decision making. In the particular models we shall develop later in this volume, the investment rule used by firms will be keyed to the firm’s profitabili ty, and perhaps to other vari­ ables. Thus, profitable firms will grow and unprofitable ones will contract, and the operating characteristics of the more profitable firms therefore will account for a growing share of the industry’s activity.

The selection mechanism here clearly is analogous to the natural selection of genotypes with differential net reproduction rates in bio­ logical evolutionary theory. And, as in biological theory, in our eco­ nomic evolutionary theory the sensitivity of a firm’s growth rate to prosperity or adversity is itself a reflection of its “genes. “

Finally, we view firms as possessing routines which operate to modify over time various aspects of their operating characteristics. In a sense, the model firms of evolutionary theory can be thought of as possessing market analysis departments, operations research shops, and research and development laboratories. Or there may be none of these organizational devices built into a fi rm, but at least from time to time some people within the firm may engage in scrutiny of what the firm is doing and why it is doing it, with the thought of revision or even radical change. We propose that these processes, like other ones, are “‘rule guided.” That is, we assume a hierarchy of decision rules with higher-order procedures (for example, scrutiny of the currently employed production technique, or the undertaking of a study of a range of possible modifications in advertising policy) which act occasionally to modify lower- order ones (the techniques used to make a particular part, or the procedure determining the mix of raw materials employed, or current decision rules regarding ad­ vertising expenditure). And there may even be procedures of a still higher order, such as occasional deliberations regarding the ade­ quacy of present research and development policy, or of the method­ ological soundness of the marketing studies being used to guide ad­ vertising policy.

These routine-guided, routine- changing processes are modeled as “searches” in the following sense. There will be a characterization of a population of routine modifications or new routines that can be found by search. A firm’s search policy will be ch aracterized as deter­ mining the probability distribution of what will be found through search, as a function of the number of variables -for example, a firm’s R,&D spending, which in turn may be a function of its size. Firms will be regarded as having certain cri teria by which to evaluate proposed changes in routines: in virtually all our models the crite­ rion will be anticipated profit. The particular model we shall employ for search will depend on the question we are probing.

Our concept of search obviously is the co unterpart of that of muta­ tion in biological evolutionary theory. And our treatment of search as partly determined by the routines of the fi rm parallels the treatment in biological theory of mutation as being determined in part by the genetic makeup of the organism.

As in orthodoxy, the characterization of individual firms in evolu­ tionary theory is primarily a step toward analyzing the behavior of industries or other large-scale units of economic organization. The models in this book are of “industries”- that is, situations in which a number of broadly similar firms interact with one another in a market context characterized by product demand and input supply curves. In modeling these situations we often find it convenient to assume that “temporary equilibrium” is achieved-to abstract from such short-run dynamic processes as those that establish a single price in the market in a single period . However, we emphatically do not assume that our model industries are in long-run equilibrium, or focus undue attention upon the characteristics of long-run equi­ libria.

The core concern of evolutionary theory is with the dynamic process by which firm behavior patterns and market outcomes are jointly determined over time. The typical logic of these evolutionary processes is as follows. At each point of time, the current operating characteristics of finns, and the magnitudes of their capital stocks and other state variables, determine input and output levels. Together with market supply and  demand conditions that are ex-ogenous to the firms in question, these firm decisions determine market prices of inputs and outputs.6 The profitability of each indi­ vidual firm is thus determined. Profitability operates, through firm investment rules, as one maj or determinant of rates of expansion and contraction of individual firms. With firm sizes thus altered, the same operating characteristics would yield different input and out­ put levels, hence different prices and profitability signals, and so on. By  this  selection  process,  clearly,  aggregate  input  and  output  and price levels for the industry wo uld undergo dynamic change even if individual firm operating characteristics were constant. But operating characteristics, too, are subj ect to change, through the workings of the search rules of firms . Search and selection are simul­ taneous, interacting aspects of the evolutionary process: the same prices that provide selection feedback also influence the directions of search. Through the joint action of search and selection, the firms evolve over time, with the condition of the industry in each period bearing the seeds of its condition in the following period.

Just as some orthodox ideas seem to find their most natural mathe­ matical expression in the calculus, the foregoing verbal account of economic evolution seems to translate naturally into a description of a Markov process- though one in a rather complicated state space. The key idea is in the final sentence of the preceding paragraph: the condition of the industry in each time period bears the seeds of its condition in the following period. It is precisely in the character­ ization of the transition from one period to the next that the main theoretical commitments of evolutionary theory have direct applica­ tion.  However,  those commitments include the idea that the process is not deterministic; search outcomes, in particular, are partly sto­ chastic. Thus, what the industry condition of a particular period really determines is the probability distribution of its condition in the following period. If we add the important proviso that the condition of the industry in periods prior to period t has no influence on the transition probabilities between t and t + I, we have assumed pre­ cisely that the variation over time of the industry’s condition -or “state”- is a Markov process.

Of course,  a vast array of particular models can be constructed within the broad limits of the theoretical schema just defined. Each particular model defines a particular Markov process, which may be analyzed with the aid of the mathematical propositions relating to Markov processes in general. For such analysis to reach conclusions of economic interest,  however, there must be a lot of specific eco-nomic content in the model. General theorems about Markov pro­ cesses are not themselves of economic interest; they are just tools that are useful in attempting to extract the conclusions that have been in­ troduced into the model through its specific assumptions . For ex­ ample, it may be possible to show that the industry approaches a “long-run equilibrium/’ which may be either a static condition or a probability distribution of the industry state that applies (approxi­ mately) to all dates in the remote future . And if an approach to such an equilibrium is in fact implied in the model’s assumptions, it will ordinarily be possible to describe some properties of such an equilibrium- for example, to describe the operating characteristics of firms that. survive.

An important determinant of the success of efforts to extract such conclusions is the complexity of the model. This brings us to an im­ portant point regarding the scope of evolutionary theory and, more particularly, of the class of Markov models of industry evolution. At an abstract level , this modeling schema has enormous generality. We may think of a “firm state” as comprising descriptions of the firm’s physical state (plant and equipment), information state (contents of file drawers and human memories), operating characteristics, invest­ ment rules (affecting transitions of physical state) , recording rules (affecting transitions of information states), and search rules (af­ fecting transitions· of operating characteristics, recording rules, and search rules) . All of these descriptions could in principle be highly detailed. We can think of an “industry state” description as in­ volving the list of all firm state descrip tions, for all firms in being and also for potential or deceased firms, together with a list of environ­ mental variables that may be determined as given functions of time and/or as functions of the firm states. The transition rules for this complex industry state description are largely implicit in the descrip­ tion itself. Operating characteristics map physical and information states into current actions . Current actions and the date determine the environmental variables. Firm by firm, the current firm state and values of environmental variables are mapped into a new firm state by application of investment, recording, and search rules. And the process continues .

There is nothing wrong with the foregoing as an abstract concep­ tualization. However, the point of a modeling effort is not just to describe a system, but to describe it in such a way that its behavior may in some degree be understood. It is for this reason that the models that appear later in this book are very simple examples within the abstract scheme just described. Like most of our orthodox colleagues, we distinguish sharply between the power and general-ity of the theoretical ideas we employ and the much more limited re­ sults that our specific modeling efforts have yielded thus far.

Source: Nelson Richard R., Winter Sidney G. (1985), An Evolutionary Theory of Economic Change, Belknap Press: An Imprint of Harvard University Press.

Leave a Reply

Your email address will not be published. Required fields are marked *