"A Behavioral Theory of the Firm” M. Cyert and J. G. March Review by Tom Issaevitch Introduction As the title implies, this book is concerned with how firms behave and make decisions. The authors point out that the standard profit maximizing micro-economic prescription for firm behavior max p. y q.x K x, y , K subject to f ( x, y, K ) 0 (where p is the output price vector, y the output quantity vector, q the input price vector, x the input quantity vector and f () the production function) is untenable. Firms do not know their own production functions or their competitor’s production functions with certainty. Firms do not know the future market demand curve, prices or any other quantity of interest. Even with the recent introduction and wide scale adoption of more accurate cost allocation schemes, ERP systems, large-scale databases and, sophisticated econometric models, there remains one essential area of uncertainty (as opposed to quantifiable risk) --- the behavior of the firm’s managers. In the next section, I describe the main points of the book regarding firm behavior. In the following section, I describe the results of the author’s models. Next, I consider the implications of the author’s ideas to accounting. In the final section, I discuss how Cyert and March hold up in the year 200. Firm and Manager Behavior There are three essential components that must be present in any decision: the goal, the possible actions and the available information. It is important to note that these three components are each bi-causally connected, i.e., feedback to each other. The goals and possible actions associated with a given decision clearly dictate the type of information one would want. However, given that one almost never has all relevant information, the reverse causality also has importance. The cost and availability of information determines the types of goals and actions that a firm could consider having. In general, this is an extremely complicated problem to optimize. A main concern of The Behavioral Theory of the Firm is how the firm handles the biases, self-interest and incompleteness of the information. In effect, firms respond to this imperfect world by restricting the decision methodology, actions, actors, and goal creation. The authors use the terms “adaptively rational system” as opposed to “omnisciently rational system” to describe the firm and its reaction to shocks (which would not exist if the firm were omniscient). An important feature of this view is that firms spend less time predicting, planning and optimizing and more time achieving stability and trying to eliminate the need to predict. However, this “sub-optimal” behavior does not imply simplicity. Quite the contrary, adaptation and non-omniscience imply contingent actions, contingent goals and contingent organizational structure. With respect to the decision methodology that firms employ, firms rarely optimize. First, they generally search for alternative actions and information only if an acceptable solution is not readily apparent. With respect to goals (about which more will be discussed below), firms rarely attempt to optimize expected profit or any version of such. Instead, a great deal of focus is spent eliminating uncertainty or it’s effects, especially uncertainty that derives from the biases selfinterest introduces into the information decision-makers receive. Interestingly, this conservatism persists even though, empirically, Cyert and March provide evidence that human beings are quite good at anticipating and accounting for information biases. Because the only conscious thought arising from firms comes from the employees and because such employees often have conflicting goals, Cyert and March point out that the very idea of goal creation in an organization cannot be taken for granted. Instead, goals are arrived at through the joint negotiation of the employees of the firm. Thus, far from being static or abstract (“this firm makes cars”), organizational goals depend on the characteristics, goals and abilities of the employees. That a firm may have a long-term identity (“it makes cars”) depends on one of the goals of current employees being to choose new employees who fit in with current goals (presumably because they like making cars). Even if the individual preferences of a firm could be perfectly aligned, the presence of uncertainty and consequent need for adaptation render the notion of `firm goal’ problematic. In general, there is no simple way to operationalize an overall firm goal into a series of harmonious sub-goals. Because this decomposition is imperfect, there will inevitably be conflict between these sub-goals. For example, two possible sub goals to profits maximization are cost cutting and lowering price to increase demand. However, since cost cutting may result in lower quality, demand may actually fall. Models Despite whatever connotations the word “Behavior” in a title may invoke, this is not a touch and feely book. After describing the reasons for a rule of thumb/ain’t broke don’t fix it approach to decision-making, the authors provide two mathematical models of decision-making and support them with empirical data. The model first is a duopoly model in which two firms (an incumbent and a younger “splinter”) predict demand on competitor behavior using a “simple” auto-regression model for conjectural variations (with later knowledge, we know that such nonlinear models may behave very non-trivially, even if their simple structure would indicate otherwise). The model did quite well in predicting the results of U.S Can (incumbent) and Continental Can (splinter) over the long stretch 1913-1956. The second model is that of price-setting and output determination under stochastic demand. The model showed remarkable ability to predict the pricing of a firm over a two-year period. I am still puzzled by this close agreement given the difficulty of the task. Impact on Accounting The accounting system provides the information on which managers are evaluated and make decisions. Cyert and March demonstrate that actual firm decision-making differs significantly from the economic-based ideal of profit maximization. It therefore follows that economic-based accounting systems (if present) may need to be modified to support the satisficing decision style. Moreover, since this change in decision method stems primarily from the information quality due to manager behavior (and not the accounting system --- if it were the accounting system, one would change the accounting system rather than change the decision goals and methods), the accounting system must have a strong manager monitoring function. As mentioned above, adaptation and non-omniscience imply contingent actions, contingent goals and contingent organizational structure. Thus, the organization is in constant flux and needs to contain an extensive infrastructure to support and monitor the changes conditionality implies while, at the same time, providing mechanisms that promote long-term stability. The AIS is a crucial tool in this change infrastructure. For control and monitoring, we have on the fly cost allocation, variance analysis and all other aspects of managerial control. For learning, the AIS constitutes the historical repository of all past budgets (i.e., a monetary view of past decisions) and transactions (i.e., the consequences of prior decisions) and studying the relationship between the two facilitates learning. Finally, with regard to stability, historical data can yield long-term benchmarks and standards, i.e., an “operating point” around which a stabilization goal can be formed. The nature of organizational goals also has a significant effect on the requirements of an AIS. In particular, the accounting system must allow progress towards goals to be measured. One cannot cut costs without first knowing what costs are. Recent Developments This book was first published in 1963. Since then, a profound information revolution has taken place, and one may wonder how well Cyert and March hold up. Two aspects hold up quite well: human interactions are still well beyond the capabilities of any accounting or information system and rules of thumb remain a crucial element of control under uncertainty. On the other hand, we now have a better understanding of just how complicated rules of thumb can be and it is not clear (to me, at least) that implementing such rules is easier or superior to economic-based optimization approaches. I now address the issues of control under uncertainty and rules of thumb in more detail. Spurred by engineering and the military, a systems revolution and manner of thinking began in the 1950’s and it is clear that this influenced Cyert and March. By the 1960’s, control of parametrically known stochastic linear systems was well understood. Moreover, the algorithms derived had significant robustness properties that led to optimistic beliefs concerning the control of complicated (i.e., large, nonlinear and possibly uncertain) systems. However, this optimism (particularly that which was placed in artificial intelligence) was misplaced. Over the last three decades, “adaptive control” has emerged the main tool used to control complicated uncertain systems. One of the interesting ingredients of adaptive control is the wide-spread use of ad-hoc rules in lieu of a more rigorously based strictly Bayesian approach. This has been necessitated by the extreme complication of the Bayesian approach and fits Cyert and March’s insight to a tee. Thus, even the mathematical end of the spectrum has adopted a rule of thumb approach. Even more support for Cyert and March’s position comes from the field of robust control (now often viewed as a sub-field of adaptive control) in which the focus of controlling uncertain plants shifts from optimality to stability. On the other hand, rules of thumb have their own problems --- particularly when, as is done often in The Behavioral Theory of the Firm, the decision variables become discrete. In general, discrete decisions are more difficult than their continuous counterparts (often NP hard, in fact) and many procedures that work on small toy problems fail spectacularly on larger problems. While I do not argue with Cyert and March’s call to keep things simple, I merely point out that choosing the right (or even reasonably good) rules of thumb may prove to be just as hard as profit maximization. Another difficulty with using nonlinear models (and the threshold models the authors use are nonlinear) is that of structural instability and chaos. The former problem makes model selection difficult. The later problem is probably generic when one considers that the models are designed to operate under fluctuating demand. It is quite common that so-called driven nonlinear systems behave chaotically and management science has certainly become aware of this fact. Finally, I give a few comments on the relation between firm structure and the need for adaptability. It is well known that most firms are hierarchical. The common explanation, hierarchies division of labor (with particular emphasis on division of skills and power) and span of control, is a static one. However, because hierarchies are well suited to adaptation, one could also provide a dynamic explanation. By well suited to adaptation, I mean in a structural sense --hierarchies are tree-structures and the easy addition or deletion of nodes in a tree (i.e., changes in the tree structure such as adding a subdivision or merging two subdivisions) has been much exploited by computer science. “Complex Organizations” C. Perrow Introduction This book is concerned with how the interaction between individual behavior, organizational structure and, the environment leads to firm behavior in complex organizations (although much of the discussion concerns corporations, other types of organizations are considered). A major theme of the book is the importance of considering all three levels and the interaction between them. But the author is also concerned with within level descriptions and spends a good deal of time describing and strongly criticizing the many prior management theories that focus attention on single given level or on two adjacent levels. As the book progresses, there is less criticism and more Perrow, but it is only in the final chapter that Perrow is mostly presenting his own theory. The rest of this review is organized as follows. In the next section, I discuss Perrow’s thesis in more detail. I conclude with a brief section analyzing the book in light of what we have learned in the three decades since it’s initial publication. Book Outline Perrow starts off with a general introduction to the types of transaction and agency problems that bureaucracy solves. First, by bringing external functions under internal control, one avoids transaction costs. Once internalized, anonymity, independence and professionalism of hiring reduce the problems of nepotism (which leads to power cliques through employees owing their hirer a debt) and bigotry. Perrow also describes incentives and the consequences of poor contract design. An interesting example is that of outsourcing tax collecting and allowing the agent to keep collections above a certain level. This obviously leads to an incentive to overtax mitigated only by the tax collectors fear of retribution. A contract that avoids this problem is the straight salary --- the entrepreneur (tax collector) is now a bureaucrat. Perrow then describes various features of organizations and institutions (henceforth often referred to as firms for brevity). They have hierarchical structure and control, overt and covert rules and procedures that operationalize firm knowledge and learning. Firms also have individuals with their own aspirations. While firms have the capacity to act rationally in an uncertain environment (Bernard would say they are the only elements of society capable of such rationality), this does not mean that they are optimally efficient or that individuals cannot subvert them to their own ends. In particular, there is a well-caricatured tendency of the “ins” to preserve their status even at the expense of overall firm wellfare. Perrow spends some time defending hierarchy. First, hierarchy is essential for task specialization (at least for human agents). Perrow also points out that hierarchy does not preclude decentralized decision and, in fact, can respond to external events flexibly with, e.g., task groups. He also makes an interesting observation that tall thin hierarchies actually may have more decentralized decision-making. Essentially structure (hierarchy) follows function, rather than the converse, and so flat hierarchies are flat precisely because it is easy for a manager at one level to control many employees at the next level down. This implies not only that the tasks these employees perform must be simple and easily monitored but also that it is not possible for these employees to have much intellectual input into the firm (who has the time to listen?). Part of the image problem hierarchies face is caused by the fact that their wide-spread introduction and the invention of scientific management went hand in hand. Since scientific management was a main impetus for the wide adoption of hierarchies (of course, hierarchies existed before scientific management) hierarchies may be tainted with the worst aspects often attributed to scientific management, e.g., worker as robot or Bernard’s belief that they required little pay. Thus, as a counter point, Perrow maps out in detail the evolution of views of the worker from robot to intelligent human. Hierarchy is still needed. After this introduction, Perrow proceeds to analyze a number of different organizational and managerial theories. In general, as he progresses further through the book, Perrow’s distaste for the theories he describes decreases. His progression also follows the three divisions he studies: human, structure and the environment. The models or schools of thought he reviews are: the human relations model; the neo-Weberian model; the institutional school. The Human Relations School The human relations school may be caricatured as “happy workers are productive workers”. Perrow goes to great lengths to point out that all evidence put forth to link worker satisfaction to productivity fails on closer inspection (including a number of early metaanalyses). There are simply too many confounding variables. More importantly, the human relations school suffers from its too narrow focus. The confounding variables are not random, but rather determined by firm structure and the environment. Despite Perrow’s objections, the human relations model still (as far as I can tell) holds considerable importance management theory. In part, this is due to the improved results that have been made possible by recognizing the importance of controlling for environmental factors. The Neo-Weberian School The role of the individual in firms is an important focus of the early part of the book. Perrow dismisses the Bernardian view that leaders get their authority and worker cooperation from moral authority. He also dismisses the Weberian view that workers are simply cogs in the firm’s hierarchical machine of structure and rules and for whom cooperation is not an issue. The neo-Weberian school attempts to fill in the machine with human beings. In Simons model, we have introduced the now famous “boundedly rational” man. The firm’s structure and standards (a “smoothly oiled machine”) combined with the firm’s goals and the instructions of it’s leaders provide aids that allow managers to cope and make satisficing decisions. In particularly, stability (with some slow adaptation allowed) becomes an important goal in the face of environmental uncertainty. Goal alignment is achieved through bargaining among the managers and the corresponding proper incentives. Most control is unobtrusive, in part passively enforced by the firm’s structure and, in part enforced by controlling the way employees think. The neo-Weberian school goes a long way towards a theory of organizational behavior. However, for the most part, the neo-Weberian school does not consider environmental factors. There are exceptions, e.g., Cyert and March, who recognize the importance of an uncertain environment in leading to organizational adaptation and the need for an appropriate structure and behavior. However, at the level of their analysis, the environment is just some random black box. They do not classify or study a range of environments. The Institutional School The last set of models Perrow critiques belong to the institutional school. The main idea is that the function of the organization determines its structure. Perrow points out the danger of this feature of firms. If too much function has been passively relegated to letting the structure handle it (i.e., the firm operates largely on autopilot), the firm will be slow to recognize and adapt to change. Thus, in particular, unobtrusive controls are not enough. Perrow focuses a good deal of attention on organizations that have “taken on a life of their own”. The institutional school backs up its main idea with an extensive array of field data documenting organizational failures arising from structure. It also makes an effort to identify the important components of firm morphology. The final contribution of the institutional school that Perrow chooses to highlight is the identification of the importance of the environment. Not all firms ossify and fail to adapt to environmental changes --- some adapt quite well. Probably a good deal of the difference is caused by the uncertainty in the environment at the time the firms structure was established (an unchanging original environment will probably result in an in-adaptable structure) and how well the firm adapted it’s original structure to the original environment (a close match will also probably lead to poor firm adaptability). While Perrow applauds this understanding of the importance of the environment in understanding structure, he faults the school for not recognizing the role organizations have on the environment. This fact, which every marketing major would applaud, is the subject of Perrow’s final chapter. The Environment Perrow does not offer a general theory of environment --- structure --- human interactions. Instead, his chapter on the environment starts with of a number of industry case studies. However, he then goes on to study more abstractly a specific type of environment --- the network of firms that have developed under post war international capitalism. Networks, in the form of economy-wide input-output models, were nothing new to economics in 1972. However, these network models generally stayed at the level of the network and did not consider the effects of inter-firm networks on firms or visa-versa. Perrow emphasizes the interaction (feedback, basically) between the two different levels. The Current State The importance of the overall environment is now recognized in a number of different areas. It may even represent a paradigm shift in artificial intelligence (AI) in which decades of failure are now believed to be explained by the fact that AI had ignored context. Likewise, the application of a broad arsenal of graph-theoretic techniques to problems of social science has been a minor industry over the last few decades. Most of these analyses have been static ones, however, (e.g., clique decompositions, percolation thresholds) and I suspect Perrow would not be satisfied (nor, for that matter, Cyert and March with their emphasis on adaptation). Studying dynamic networks has been a pet project of mine and I believe such studies will find many new insights. Ironically, recognizing the importance of environmental factors has allowed the human relations school to improve its methodology in searching for motivating factors. Likewise, behavioral research (neo-Weberian, to Perrow) has had some success in identifying and understanding the heuristics and biases used by managers (and others). It is clear that the field of social science has benefited from Perrow’s work.