The arguments in this volume can be summarized by three statements: (1) The business firm is a relevant unit for investigation. (2) It is possible to construct a theory of decision-making behavior within such a unit. (3) Such a theory must focus explicitly on actual organizational decision process. We have tried both to show how to construct behavioral models of firm decision making and to indicate the basic theoretical framework within which such models lie. We have tried to specify reasons for preferring a behavioral theory of the firm to more conventional theories for purposes of descriptive microanalysis. In this section we elaborate briefly the implications of the theory for descriptive analysis of the firm.
1. Price and output determination
We will not attempt to discuss in any detail the implications of the theory for price and output models. Most of the work reported in previous chapters has dealt specifically with that problem. A microeconomic theory of price and output has to explain the quasi resolution of conflict within business organizations, problemistic search behavior, uncertainty avoidance as a basic decision commitment, and organizational learning.
The models presented in this book represent price and output determination as an adaptively rational, multiple-objective process rather than an omnisciently rational, single-objective process. Prices and outputs change primarily in response to short- run feedback on performance, and the extent to which solutions to organizational problems are sought in price and output changes depends on past learning about the consequences of such changes. Firms do not resolve potential conflict between share-ofmarket, profit, and production stability goals primarily by a procedure of explicit mediation. Rather, they ameliorate conflict by accumulating excess resources (organizational slack), by decentralizing information, and by attending sequentially to crises. Firms attempt to avoid the uncertainties of competition by using standard industry practice and negotiated riskavoiding agreements.
2. Internal resource allocation
The procedures (e.g., budgeting, transfer pricing) by which a firm allocates resources to the various units within the organization are procedures for solving a classic economic problem — the efficient utilization of resources. By assuming the effective operation of an external price system, the theory of the firm has largely by-passed internal allocation as a focus of attention. 1 If the theory of organizational decision making that we have suggested has validity, the classical assumptions are incomplete. In fact, we should be able to identify some general implications of our theory for a model of internal allocation.
Consider the quasi resolution of organizational conflict. In budgeting (as in pricing), we expect to find that goals tend to enter as more or less independent constraints (each of which must normally be satisfied). Where an allocation plan apparently meets the constraints, we expect rather loose evaluation of the accuracy of the estimates and other assumptions on which it is based. Where resource rationing is necessary, we expect two general kinds of reactions: first, a tendency to use arbitrary allocative rules that maintain the relative positions of the members of the coalition; second, a tendency to re-evaluate those estimates that are relatively difficult to defend in terms of traditional organizational practice, standard accounting procedure, or immediacy of tangible return.
Consider search behavior. We expect to find the same general form of problem- oriented search as in the case of price and output. That is, search occurs in response to a problem facing the firm and is directed by learned rules for associating search behavior with particular problems. The search for alternative internal allocations is stimulated by failure on some criterion of relevance to a subunit of the organization. Thus, one type of search is stimulated by a need to reduce immediate expenditures, a second type by a need to improve profitability, a third type by a need to increase sales, and a fourth type by a need to maintain the position of a subgroup within the organization. Each type of search has a different organizational locus and a different set of organizational consequences. In general, for an alternative investment opportunity to have a significant chance of being introduced in the system, two conditions have to be met. A problem must be perceived by the organization, and the investment must be visible to that part of the organization in which search is stimulated by the problem.
Consider uncertainty avoidance. We expect to find extensive dependence of budgeting on standard industry and firm rules. Widely shared operative criteria (e.g., share of sales dollar invested in advertising) both standardize dimensions within the environment and, in effect, standardize decisions by permitting cross-firm comparisons. They also induce various kinds of rule-directed behavior. A case in point occurs when there is interdependence among activities. Under such conditions, evaluation depends on partially arbitrary rules for allocating common inputs and outputs. The rules tend to be biased in the direction of producing data leading to organizational acceptance of projects favored by particular subunits. For example, we expect to find a tendency to combine projects that do not meet explicit criteria with projects that more than meet them in order to obtain funds for projects desired on other grounds.
Consider organizational learning. We expect the same general kind of behavior as for price and output. We expect changes in the goals over time, changes in the search and decision rules, and changes in the learning rules. For example, in the case of a budget allocation, we expect adjustments over time in the aspirations of the various subunits, in the criteria applied to proposals, and in the search reactions to failure. In general, we expect a behavioral model of resource allocation to be heavily history- dependent in the same way, and for the same reasons, as our suggested models of pricing.
On the whole, we think models of resource allocation that build upon these concepts of a market price system are likely to be useful first attempts at a theory of resource allocation within a firm. We think the concepts are generally consistent with descriptions of actual budget behavior. However, we are far from confident that we can identify a priori all of the major relevant factors in allocation decisions. Our only intention here is to outline some possible first approximations.
Project allocation. At least at the outset we will distinguish between allocation decisions involving projects (e.g., capital budgeting, some kinds of operating budgets) and allocation decisions involving organizational subunits without immediate reference to projects. We may wish to abandon the distinction ultimately, but it corresponds roughly to a distinction made by many organizations and thus has face validity for a process theory. It is convenient to think of project allocation decisions as resulting from three relatively independent activities: (1) the determination of a target total project budget, (2) the allocation of that budget to specific projects and subunits, and (3) the implementation of that allocation through expenditure.
The target total budget is a calculation of what the organization in some sense “can afford to spend” during the next budget period. In classic investment theory such a total is not independent of the investment opportunities, but we expect a relatively independent calculation. Presumably it is subject to some modest revision subsequently on the basis of opportunities (if the opportunities are not suppressed by the system), but the target probably depends heavily on distinctions between internal and external capital, on total revenue expectations, and on past actual total budgets. We expect the target to have two major consequences. On the one hand, it should be strongly self-confirming upward; that is, if the target exceeds the probable total normal demands of subunits for projects, the search for new projects on the part of subunits and the optimistic inflation of estimates of returns from pet projects will assure the authorization of the full budget. On the other hand, it should also be self-confirming downward — subject to some possible revision in the target if project demand exceeds target funds by a large amount. Given these self-confirming attributes of the target, the determination of total funds available becomes a critical determination in the allocation decision. It is, however, a determination that has not been studied intensively from the point of view of a positive theory.
In contrast, some preliminary work has been done on the allocation of a project budget to specific projects and subunits in the organization. On the basis of a study of actual capital-budgeting procedures and practices in a few large firms, L. R. Pondy has suggested that projects are selected by a sequence of allocations to project classes. 2 That is, the organization first allocates an approximate share of the budget to major types of projects (e.g., “normal,” “major,” etc.). The shares allocated depend on such considerations as historical legitimacy, current organizational emphasis, and presumed performance. This major allocation is taken substantially as given at the next step, and another approximate allocation is made to a more refined project class. This continues until the number of projects within a class is small enough and goals are shared enough to allow an explicit, complete ranking of individual projects. In general, different criteria are applied at the different stages; the segmentation of the decision makes final allocation decisions substantially dependent on the classification, the differences among criteria, and the nonuniqueness provided by multigoal, acceptable-level ranking procedures. Thus, a theory of project allocation would focus on the processes by which project classes are defined, class criteria for evaluation of projects are developed, and specific individual projects are assigned to classes.
Frequently a theory of organizational decision making must distinguish between the choices made by the organization and the specific action implemented. Such a distinction is especially important here because the execution of a project allocation ordinarily involves a relatively long time and the organizational separation of the “decision” from the “implementation.”
Two simple case reports will illustrate this phenomenon. In a study of capital budget decisions in a large manufacturing firm, Pondy discovered that the procedures for making “technical” revisions in project plans after approval yielded somewhat different projects from the projects as approved. 3 In some cases this involved the modification or elaboration of approved projects with partial substitution of projects that either were not approved or would probably not have been approved originally. We made a similar observation in a study of capital budgeting in another large firm during a period of marked inflationary pressure on project costs. Differences between organizational rules for changing budget authorizations and organizational rules for changing project specifications predictably affected both the project characteristics and the relation between budgeted and actual project expenditures. In this situation, cost estimates were consistently low because of inflationary pressure and organizational rules required reauthorization if expenditures exceeded authorization by more than a fixed percentage. When project costs threatened to exceed authorization by more than the fixed percentage, the usual response was to redesign the project and search for alternative sources of components rather than seek a new authorization. As a result, the frequency distribution of the ratio of actual project expenditures to estimated project expenditures shows a sharp drop at the point where the reauthorization rule became operative, and the projects actually constructed were somewhat different from the projects authorized.
As in the case of price and output models, it seems probable that the first successes in models of project allocation will deal with the setting of target budgets and with the allocation of a target budget to specific projects or subunits. The implementation problem will have to await a convenient way of representing the kinds of modifications characteristically introduced at that stage. Subunit allocation. There are two important ways in which subunit allocation tends to differ from project allocation. First, because the subunit persists whereas only classes of projects persist, historical precedents for allocating resources among alternatives are likely to be of greater importance in allotments to departments than in allotments to projects. Second, because it is a more complicated mixture of activities, the subunit is likely to be harder to evaluate in widely accepted terms than is the project. Thus, it is harder to obtain agreement on procedures for ranking subunits than on procedures for ranking projects. Because of these differences, we would expect that decisions on allocations to regular subunits would be quite sensitive to past experience, to the experience of comparable subunits, and to the prima facie relevance of the subunit to other parts of the organization. We would expect learning with respect to these decisions to be rather crude and subject to substantial short- run error.Consider, for example, the allocation of resources to research and development within a business firm. Partly because it appears to represent an extreme case of the characteristics we have attributed to organizational subunits and partly because of contemporary interest in research, a number of students have considered predictive models of research and development allocation. If our general conception of an organization is correct, expenditures for research and development will be based on simple rules that change slowly over time on the basis of experience and are voided in the short run by the pressure of failure. The recent work of Seeber seems to be consistent with such a conception. Seeber’s studies of the process of budget decisions for research and development seem to indicate four important features of the process from the present point of view:
- Most organizations are aware of and probably use such simple rules as per cent of revenue as a guide to research and development allocation.
- The pressure of subunits for maintaining absolute dollar allocations, the logic of research appropriations, and the difficulties of forecasting revenues lead to considerable attempts to smooth allocations so that they vary less from year to year than do revenues.
- Target allocations are substantially influenced by estimates of allocations (per cent of sales) in other “comparable” organizations.
- Organizational failure on profit or sales goals leads to pressure to revise the allocation rules.
These gross features of the process can be considerably refined on the basis of Seeber’s research. The refinement is obviously necessary to provide specific understanding of the ways in which the demands of the various organizational subunits are mediated in the face of scarcity, the ways in which the allocation rules change over time, and the ways in which causes are attributed to failures of various kinds. For the present, however, we will simply note that (1) the results of this research on subunit allocation seem to be generally consistent with the organizational models we have outlined and (2) studies of research and development expenditures (because of the conspicuous difficulties in evaluating research and development performance) should be especially useful in providing clues to the ways in which organizations avoid uncertainty and learn in ambiguous situations.
Transfer payments. With diversification, vertical integration, and growth in firm size have come various methods for determining the payments among organizational subunits. The effect is to allocate resources; the major motivation is to allocate responsibility for aggregate organizational performance. Thus, although the term “transfer pricing” has come to have a rather specific meaning both in the literature of economics and the parlance of management, we can focus more generally on the class of allocative procedures by which the organization disaggregates performance and transfers accounting credit for performance from one divisional ledger to another. These procedures include the allocation of overhead, various preferential-treatment market systems for the purchase of goods and services by one subunit from another, organizational conventions for determining subunit responsibility for receipts or disbursements (in money or other criteria of performance), and an assortment of informal devices for exchanging budgetary allotments.
In recent years these procedures have been elaborated by the development of transfer pricing schemes and by the demonstration of various theorems about the efficiency of such schemes. 6 Some form of modified market system has been used by a number of firms to allocate charges for key goods and services. More ancient procedures for cost and revenue allocation have an apparent purpose that is quite similar — if somewhat cruder. They attempt to make a “fair” division among the subunits.
If, however, we are primarily interested in a positive theory of allocation and if we view the organization as a coalition of participants in which conflict is only partially resolved, the concepts of “efficiency” and “fairness” have limited utility. In fact, we expect to find that organizational participants view the rules for making transfer payments as largely arbitrary (at least within wide limits). From the point of view of the subunits, performance is determined partly by the return from the external environment and partly by the transfer payment rules they can arrange by bargaining with the other parts of the organization. Subunit success involves dealing effectively with the environment and negotiating effectively with the organization on accounting conventions.
If our model of an organization is correct, we should find that transfer payment rules result primarily from a long-run bargaining process rather than a problem-solving solution. We should find that renegotiation of transfer payments will be viewed as a primary activity for solving subunit problems, especially where the subunit does not perceive the external environment as being susceptible to further exploitation. We should find that the same factors that produce organizational slack with respect to the external environment will produce slack with respect to transfer payments. Subunits that have been successful will be less active in seeking new transfer rules than will units that have been unsuccessful. In general, we should find that transfer payments are made on the basis of a few simple rules that (1) have some crude face validity and (2) have shown some historic viability. We should find that they are the focus of conflict among subunits in the same way as other allocative devices.
Wage payments. One way of approaching the allocation to wage payments would be to focus explicitly on the wage decision and build a model roughly comparable to the pricing models outlined in this book. That is, we would view the firm as “solving” the wage decision problem the same way it “solves” the pricing problem — by comparing wage payments with aspiration levels, searching when the aspiration levels are not being met, applying simple decision rules on the basis of short-run feedback, and adapting gradually to experience. Perhaps the major way in which the theory would have to be structurally different would be in the relative independence of the labor union, as compared with the functional subunits considered in the price and output models, from the other subunits in the organization. The independence has two consequences. First, it would probably be necessary to pay greater attention to the internal processes of conflict resolution within the subunit ( labor union) than we have previously done. In general, we have been able to ignore the fact that our subunits are also rather complex organizations; we probably could not do so here. The union also “solves” the wage decision, presumably following the same general kind of decision process. Second, we would have to attend to the difference between ritual bargaining (i.e., bargaining between two groups in which conflict of interest is explicitly conceded although agreement is required by both) and informal bargaining (i.e., bargaining between groups in which conflict of interest is implicit). Because of the traditions of the relations between unions and management, most wage bargaining is ritual bargaining. Some of the procedures for the quasi resolution of conflict we suggested earlier are more relevant for informal bargaining than they are for ritual negotiations.
Where union negotiation is not involved, we would expect most wage decisions to be based on the same types of rules of thumb we have observed in other contexts. The organization faces a series of constraints on its decisions. For example, there might be constraints on the total salary budget, on the salary differentials among positions, on the minimum or maximum amounts payable to some specific individuals or positions, and on consistency within positions. The organization considers whether the existing allocation (or a proposed alternative) satisfies the constraints. If it does, fine; if not, the organization searches for an alternative. The constraints and the search procedures are subject to adjustment in the face of experience, but in the short run they dominate the decision. Thus, Simon has argued that much of the observed variance in executive compensation is consistent with an extremely simple organizational constraint model. 8 Basically, the model assumes a market determination of lowestlevel executive salaries and an organizational constraint on salary differentials within the system. Although Simon does not claim much for the model as an actual representation of salary determination within a firm, it is not hard to see why simple internal consistency rules might easily control both salary aspirations and salary levels.
Stockholder payments. Periodically, a firm allocates resources to stockholders. These resources generally are in the form of money, stock, stock options, or other closely related variants. We can outline roughly a process for determining stockholder payments. Suppose we imagine an organization with a set of financial goals of an aspiration-level type. The firm wishes to maintain at least a certain liquidity position; it wishes to maintain a particular stock value relative to the value of other stocks in the industry, to avoid governmental intervention in stockholder allocation, to avoid a stockholder revolt, and so on. The firm obtains information on how well it is doing vis-à-vis its goals. If there is a problem, search is undertaken to determine a new strategy. So long as no problems arise, some simple computational rules are used to determine the appropriate payments. These rules emphasize historical legitimacy and conformity to simple partial-ordering constraints (e.g., if profits are up, dividends should not go down). Studies by Lintner, Walter, and others indicate that some such process seems reasonable as a first approximation.
3. Innovations
One persistent problem in the development of a theory of the firm is the problem of dealing with innovations. Technology changes, and, for purposes of analysis over time, the production function must be viewed as adaptive rather than simply given. Products and product preferences change, and therefore the product mix and business of a firm change. We would like to predict who will introduce what kind of innovation when. We would like to know how to stimulate or discourage particular classes of innovations.
At one level, it appears that our general theory — especially the concept of problemistic search — is of considerable relevance to the prediction of innovations. We have argued that failure induces search and search ordinarily results in solutions. Consequently, we would predict that, everything else being equal, relatively unsuccessful firms would be more likely to innovate than relatively successful firms. Such a prediction is a legitimate derivation from the theory we have outlined where “innovation” means a new solution to a problem that currently faces the organization. Unfortunately, the evidence does not support the prediction for major technological changes. Data collected by E. Mansfield do not support the innovation- in-the-face-of-adversity hypothesis with respect to the introduction of twelve different innovations in four industries.
It is possible that this means our theory cannot be used to predict innovations. We prefer initially, however, to modify the theory to accommodate both what we think is an accurate portrayal of the mechanisms involved and the data on acceptance of new technological innovations. To do this we need to reconsider our discussion of organizational slack. As we pointed out in Chapter 3, slack is the difference between the payments required to maintain the organization and the resources obtained from the environment by the coalition. In general, success tends to breed slack. One of the main consequences of slack is a muting of problems of scarcity. Subunit demands are less closely reviewed (since they are less likely to conflict with other demands). Resources are more likely to be allocated if they are sought strongly by a single subunit. Thus, distributed slack is available for projects that would not necessarily be approved in a tight budget. We have also argued that the criteria of acceptance for organizational courses of action are heavily influenced by traditional procedures and historical rules of thumb. In general, therefore, the tighter the budget, the more expenditures will be controlled by essentially conservative rules. Slack provides a source of funds for innovations that would not be approved in the face of scarcity but that have strong subunit support. As we observed in our study of the magnetic controllers in Chapter 4, such innovations typically include improvements in the technology (better engineering, newer forms of products). These are not problem-oriented innovations. In the short run they contribute mostly to subunit goals (professional status, subunit prestige, and so forth), but some portion of them turn out to be major technological improvements. As a result, when we study the firms that have made specific significant technological improvements, we find that they were made by firms with substantial slack (and thus mostly successful firms).
Since we now have a prediction that firms will innovate both when successful and when unsuccessful, we need some operational criteria for distinguishing between the kinds of innovations in the two cases. The distinction is not hard to make in principle. Problem-oriented innovation will tend to be justifiable in the short run and directly linked to the problem. Slack innovation will tend to be difficult to justify in the short run and remotely related to any major organizational problem. The classification of innovations is, however, difficult. Consider, for example, the introduction of a computer to process a payroll. We might suspect that in most cases this represents a case of slack innovation, but we can imagine circumstances under which specific problems with labor costs or labor inaccuracies would make it a problem-oriented innovation.
We think that this view of innovations is a reasonable approach, if the problem of distinguishing between the two types of innovations is not insuperable. It will become an even more plausible approach if we can devise a measure for slack that will distinguish between distributed and undistributed slack. In the theory the two types of slack are clearly distinct, but they are difficult to distinguish empirically.
Source: Skyttner Lars (2006), General Systems Theory: Problems, Perspectives, Practice, Wspc, 2nd Edition.