On Custom in the Economy
Chapter 14: The Division of Labour
Oxford Clarendon Press, 1998, 242-265.
The division of labour entails the problem of economic coordination. For reasons of preference and efficiency, among others, the various members of society specialize in performing distinct activities. These specialized activities must be coordinated in such a way that a useful overall result is obtained. The three modes of control discussed in the context of the firm - exchange, command, and custom - are used not only for purposes internal to the organization of the firm, but for the good of society as a whole.  Thus, the executive branch of government relies on command, and so does the military. Social interaction is governed by customary and legal rules that are the social counterparts of duty at the level of the firm. The system of markets, finally, coordinates production and consumption across firms and consumers through the mechanisms of exchange.
It is a Marxian misapprehension to think that the social division of labour is organized exclusively by the market and the division of labour within firms is organized exclusively by command.  Both the
1. See S. 13.10.
2. Marx (1873: ch. 12.4). This view is shared by many non-Marxists and is reflected, e.g., in the title Markets and Hierarchies chosen by O. E. Williamson (1975).
firm and the market rely on custom. All three control modes - custom, command, and exchange - work in various combinations, both at firm level and society level.
With regard to economic activity, the firm and the market are complementary institutions for organizing the division of labour. In order to understand the emergence of firms in a market system, it is necessary to analyse which kinds of activities are better organized within firms, and which by markets. Those kinds of coordination better performed by the market ought to be left to the market; those better performed without market interference, on the other hand, will give rise to the formation of firms. Accordingly, the fundamental coordination problem associated with the division of labour entails both the theory of the firm and the theory of the market.
This chapter is devoted to the problem of the division of labour, and the conditions under which that division of labour is better organized by the firm or by the market. Firms, as islands of custom, emerge in the market because coordination by custom is in many ways more efficient than coordination provided through the market. However, custom has a severe drawback, because the mutual reinforcement of a system of interlocking customs entails rigidity and inflexibility. The inherent inertia of coordination by means of custom renders the firm inferior to the market with regard to certain types of adaptation and change. The organizational alternative of coordinating a given task within a firm or through a market may be understood as engendering institutional competition between these two modes of organization. The institutional mix between firm and market which emerges in a society may be understood as arising from this competition. The firm thrives on the advantages of custom, and the market thrives on the flexibility attainable through standardization in an anonymous setting. In order to better understand the advantages and short-comings of custom as a device for organizing the division of labour, the subsequent sections will be devoted to looking more closely into the coordination problems posed by the division of labour.
The classic position with regard to the division of labour is encapsulated in ‘Smith’s theorem’: the division of labour is limited by the extent of the market.  The usual reading of the theorem, suggested by Smith’s own account, is, however, different. It may better be phrased as ‘The division of labour is determined by the extent of the market.’ The two versions should be distinguished.
Smith argues that specialization leads to an increase in productivity resulting from ‘increasing returns to specialization’. These result from the increasing returns to scale that occur in performing any task or sub-task. As a consequence, all specialization increases the productivity of the economy as a whole. If there are two persons who divide their time equally between producing one unit of good A and one unit of good B, the joint output will be 2 units of A and 2 units of B. If both of them specialize, however, joint output increases: the first person produces more than 2 units of A and the second person will produce more than 2 units of B. They are able to do this because of increasing returns to scale. This amounts to increasing returns from specialization.
As to the reasons for increasing returns to specialization, Adam Smith mentions four of them. The first two are straightforward. One is that specialization leads to improved dexterity:
A common smith, who, though accustomed to handle the hammer, has never been used to make nails, will scarce be able to make above two or three hundred nails a day, and those too very bad ones. A smith who has been accustomed to make nails, can seldom make more than eight hundred or thousand nails a day. I have seen several boys under twenty years of age who had never exercised any other trade but that that of making nails and who could make, each of them, upwards of two thousand three hundred nails a day. 
1. Smith (1776: 17) The name ‘Smith’s theorem’ is due to Stigler (1951), who interprets it as indicated in the following text. For an interesting discussion of the division of labour, see Leibenstein (1960: ch. 7); for a good survey of views about the division of labour, see Groenewegen (1987).
2. Adapted from Smith (1776: 7-8).
The second reason is that time can be saved that would be required ‘in passing from one sort of work to another’. 
The third reason given by Smith is more intricate. It relates to the conditions of technical progress and the introduction of machinery:
I shall only observe, therefore, that the invention of all those machines by which labour is so much facilitated and abridged, seems to have been originally owing to the division of labour. Men are more likely to discover easier and readier methods of attaining any object, when the whole attention of their minds is directed towards that single object, than when it is dissipated among a great variety of things. But in consequence of the division of labour, the whole of every man’s attention comes naturally to be directed towards some one very simple object. 
The fourth reason that Smith mentions relates to ‘differences in natural talents in different men’. He is, however quite cautious here and adds that it is in reality ‘much less than we are aware of; and the very different genius which appears to distinguish men of different professions, when grown up to maturity, is not upon many occasions so much the cause, as the effect of the division of labour’. 
Adam Smith’s ‘theorem’ results from combining the idea of increasing returns to specialization with the causal variable of the size of the market. If a market is small, specialization in either good A or good B cannot occur. The joiner will produce both window casements and doors to meet the demand of his village. Although he could produce window casements more cheaply if they were demanded in larger quantities, the demand is not large enough to permit specialization. If the market expands because, say, transportation costs are reduced, conditions may change. It may then be worthwhile for joiner A to specialize in window casements and supply customers in the surrounding villages as
1. Smith (1776: 8).
2. Smith (1776: 9) Smith develops the idea further by introducing more specialists (‘philosophers and machine-makers’), who exploit further advantages of specialization along the same lines.
3. Smith (1776: 15).
well; the other joiners then may each specialize in supplying the enlarged market with other specific items.
It is illuminating to rephrase some aspects of the reasons for increasing returns to specialization, as given by Smith, in terms of fixed costs. Fixed costs are costs that are independent of the level of production. They are setup costs incurred while establishing the possibility to produce. Many aspects of the arguments given by Smith relate to fixed costs in a straightforward way, but this view also sheds light on some other aspects of increasing returns from specialization.
Explicit training costs can be viewed as fixed costs. A modern statement stresses this aspect: ‘A specialist such as a surgeon is an indivisible resource. He has a training of 10 years and then operates for 30 years. He cannot simply be replaced by two half surgeons who each are trained for 5 years and operate on half as many patients only every other day’. 
The fixed-cost parlance would also cover costs related to machinery. A joiner who builds window frames only occasionally will not find it worthwhile to buy specialized machinery; a carpenter specializing in windows, on the other hand, would be much more likely to make regular use of such equipment. Specialization renders it worthwhile to invest in specialist machinery. This leads to additional returns from specialization that go beyond the aforementioned increase in dexterity.
Important fixed costs relate to information and innovation. A lawyer specializing in certain cases may find it worthwhile to subscribe to a specialized data bank for finding the relevant precedents quickly, but a less specialized lawyer would not find such a subscription worthwhile. Similarly, the design of a machine requires research and development. This entails fixed costs, and
1. von Weizsacker (1991: 106).
it will be worthwhile contemplating such a project only if a sufficient number of machines is to be produced. It should also be noted that the greater specialization of joiners would be likely to increase the demand for specialized machinery and would therefore make it more worthwhile to supply it.
Other important fixed costs relate to the establishment of organization and routines. Investment in ‘organization’ is indeed of the utmost importance.  It is of the same nature as any other fixed-cost investment. Firms that specialize in certain products will also have an incentive to develop routines that are specifically tailored to their production. This will lead, again, to increasing returns from specialization.
The case of on-the-job skill acquisition, as discussed by Smith, also fits into the fixed-cost framework, although on-the-job training also raises more intricate issues. Nevertheless, the costs of acquiring a skill have to be considered as fixed costs. 
Smith’s argument presupposes that gains from specialization will accrue whenever the division of labour is deepened. This is not true. Many tasks cannot usefully be further subdivided. There is no sense in subdividing the writing of a poem such that everybody specializes in certain rhymes; or in subdividing the writing of computer software such that each programmer specializes in writing one single line of code. Several factors limit the division of labour, and the extent of the market is only one of them. Another relates to the nature of the task.
Some tasks simply cannot be subdivided. This applies to tasks that are of a sequential nature, where each step requires knowledge of previous steps. Distributing steps among different workers would require the transmission of information about the previous
1. Marshall (1890: 138-9) stressed this rightly by describing ‘organization’ as one of the fundamental ‘agents of production’, along with land, labour, and capital.
2. This is elaborated in Becker (1962).
steps. This may be more difficult and more time-consuming than actually performing these steps. If a programmer has started to write a routine, it will be much easier for him to complete it than for somebody else to do so. The errors that can occur in the communication process will be avoided. Such continuity has been stressed in software engineering:
Men and months are interchangeable commodities only when a task can be partitioned among many workers with no communication among them. This is true for reaping wheat or picking cotton; it is not even approximately true of systems programming.
When a task cannot be partitioned because of sequential constraints, the application of more effort has no effect on the schedule... Many software tasks have this characteristic because of the sequential nature of debugging. 
But even if tasks can be subdivided, this does not necessarily increase overall output, because a finer subdivision will necessitate more coordination. This aspect has been neglected by Adam Smith, but has been recognized in the literature:
When tasks require coordination of a delicate and complicated variety, the individual worker may be more efficient if he performs a variety of tasks because the loss from non-specialisation is offset by the gains from coordination. We expect a better painting from a single good artist, than if the world’s best specialist in clouds does a portion of the picture, and the world’s best specialist in trees another portion. 
The importance of this observation is usually not appreciated fully, in spite of the literature on transaction costs. It is not of marginal concern, but rather constitutes one of the central problems posed by the division of labour. Software engineering illustrates this, once again, and in a drastic way:
In tasks that can be partitioned but which require communication among subtasks, the effort of communication must be added to the amount of work to be done. Therefore the best that can be done is somewhat poorer than an even trade of men for months.
The added burden of communication is made up of two parts, training
1. Brooks (1975: 16-17).
2. Stigler (1952: 140); see also Stigler (1966: 168).
and intercommunication. Each worker must be trained in the technology, the goals of the effort, the overall strategy, and the plan of work. This training cannot be partitioned, so this part of the added effort varies linearly with the number of workers.
Intercommunication is worse. If each part of the task must be separately coordinated with each other part, the effort increases n(n-1)/2. Three workers require three times as much pairwise intercommunication as two; four require six times as much as two. If, moreover, there need to be conferences among three, four, etc., workers to resolve things jointly, matters get worse yet...
Since software construction is inherently a system effort - an exercise in complex interrelationships - communication effort is great, and it quickly dominates the decrease in individual task time brought about by partitioning. Adding more men then lengthens, not shortens, the schedule. 
The costs involved here are sometimes immense, and the problems generated by the difficulty of coordination are equally tremendous. 
In contexts that require much interaction, the problem of the division of labour poses itself in a way that differs from what Smith has envisaged. In order to appreciate the relative advantages of custom and the market in coordinating tasks of this type, it is useful to look more closely into the issues of timing, separability, and conceptual integrity as determinants of the nature of the task.
The issue may be phrased in a slightly more general fashion by introducing a rough classification between tasks, according to their nature. To fix the ideas, let us suppose that each task may be conceived of as involving a number of actions a, b, c,.. . These actions are subdivisions of tasks. The task of screwing two pieces of wood together may be seen as a sequence focusing on the action
1. Brooks (1975: 17-19).
2. See Brooks (1975: 31). Brooks is known as the ‘father of the OS/360’. He was project manager for the development of this software system and later project manager during the design phase.
of turning the screwdriver once. By repeating this action, the task is eventually completed. ‘Action’ refers in the following to an action relative to a task. The term is not meant to denote some ‘action atom’ that cannot be subdivided further. It does seem not useful to assume that such atoms exist. The turning of the screwdriver may be viewed as a task in its own right, comprising actions of several fingers, the arm, and the body, for instance, and this could be subdivided even further.
Given that a task comprises several actions, a distinction between different tasks may be drawn according to the timing that is required to perform the actions. A sequential task requires the actions a, b, c, to be performed in sequence.  An example would be the weaving of a fabric, where one thread has to follow the other and where it is not possible (or is prohibitively costly), for technological reasons, to leave some gaps and fill them in later. Another example would be the sequence of producing and testing an item: the test cannot be performed before the production task has been completed.
A simultaneous task requires that actions a, b, c, are performed simultaneously in close coordination. Workers who want to lift a beam together have to concert their action. The performance of a piece of music by an orchestra requires much more than each player playing his part flawlessly: the musicians have to play their parts synchronously.
There are also tasks in which the actions required can be performed quite independently. These are temporarily independent tasks; entering numbers into a data bank or writing entries for a dictionary are examples.
In reality, no task fits neatly into one or the other category. Distinctions may be drawn, however, along the time dimension. The categories introduced above are intended to highlight what is involved here.
1. See also Leibenstein (1960: 111-15).
Another way of distinguishing tasks relates to intrinsic interdependence. This aspect is of great importance for the present purposes.
Consider first a sequential task. Actions a and b have been performed, and an intermediate product B is obtained. By applying action c, this intermediate product can be transformed into another intermediate product C, and so forth. Often it is unnecessary to know specifics about actions a and b in order to perform c. All information is contained in the intermediate product B. Such a task may be denoted as separable.
A nice illustration of separable tasks is given by Demsetz: ‘The economical use of industrial chemicals by steel firms does not require transfer of knowledge of how these chemicals are produced; similarly, the use of steel by industrial chemical firms does not require transfer of knowledge of how steel is produced.’ 
Sequential tasks may, however, be far from separable. Adding another line of computer code requires knowledge of the preceding code. All preceding actions a and b must be known in order to perform action c. If the actions of a sequential task depend on each other in this way, the task is termed an interdependent task.
The distinction between separable and interdependent sequential tasks is not a sharp one. In a sense, the computer program with completed parts a and b may contain all the information that is necessary to perform action c in an appropriate manner. The intermediate product B contains that information, and in a formal sense the task may appear additive. In order to cope with this difficulty, interdependence may be conceived as a matter of degree. The larger the information needed about previous production steps, the stronger is this interdependence.
But even if all information required for taking step c were to be contained in the intermediate product B, the worker in charge of
1. Demsetz (1988: 159). Demsetz (1988: 160) has stressed this informational aspect in his view of the firm: ‘Roughly speaking... the vertical boundaries of a firm are determined by the economies of conservation of expenditures on knowledge.’
taking step c would have to retrieve the relevant information about the previous actions from the intermediate product B. For instance, he would have to try to comprehend the code written by the previous programmers. In such a case, it might be cheaper to have the worker who had performed the previous steps also perform the next step c. This would save the costs that arise from retrieval, and would reduce the danger of errors. This may be considered as still another form of interdependence.
It may be the case, however, that the intermediate product does not contain enough clues to retrieve the information that is necessary to undertake step c. It may even be cheaper to perform steps a and b again rather than to try unravel the plan that gave rise to the intermediate product B. This too would be a case of interdependence.
Consider next simultaneous tasks. Again, these may be either separable or interdependent. A separable task is one where each action can be performed without any technical knowledge of the other actions; action c can be executed without knowing precisely how a and b are performed. Measurements performed simultaneously at different points of the globe are separable. In such cases coordination is simple, as only a timing device is needed. Each measurement can be made independently, and the way in which the measurements are made is inessential. It is only important that each observer has an accurate watch, and that he is instructed about the time of measurement.
Simultaneous tasks may, however, be strongly interdependent. A piece of chamber music requires each player to react instantaneously to the playing of his partners. Each nuance of articulation has to be perceived by each player, and each has to adjust his playing in order to achieve a perceptual unity and coherence. It takes many years of continuous practice for an ensemble to perfect such a degree of coordination.
Software engineering is no less demanding. Large projects have to be split up into modules that are developed simultaneously. If module a uses subroutines of module b and module b uses sub-routines of module a, separability may be achieved by standardi-
zation. The standards must, however, be established in such a way that they can be implemented. Typically, they have to be changed to incorporate new developments. This cannot be done without intense interaction.
With regard to temporarily independent tasks, and in order to exhaust the classification, there may again be either separability or interdependence. If each individual action does not require knowledge about the other actions, the task is separable; if each task requires knowledge about the other tasks, the task is interdependent. In many cases interdependence will lead to simultaneity, but this will not always be the result. The task of filling three circles with three different colours is, in a way, both temporarily independent - because the sequence does not matter - and interdependent - because the filling of each circle requires knowledge about the colours of the circles already being filled. Yet each worker can do his job whenever he likes.
Interdependence may also be generated beyond the temporal dimension through the requirement of conceptual integrity. This requirement has been stressed in particular in software engineering, but it is of general relevance:
Most European cathedrals show differences in plan or architectural style between parts built in different generations by different builders. The later builders were tempted to ‘improve’ upon the designs of earlier ones, … So the peaceful Norman transept abuts and contradicts the soaring Gothic nave .
Against these, the architectural unity of Reims stands in glorious contrast. The joy that stirs the beholder comes as much from the integrity of the design, as from any particular excellencies. As the guidebook tells us, this integrity was achieved by the self-abnegation of eight generations of builders, each of whom sacrificed some of his ideas so that the whole might be of pure design...
Even though they have not taken centuries to build, most programming systems reflect conceptual disunity far worse than that of cathedrals. Usually this arises not from a succession of master designers, but from the separation of design into many tasks done by many men.
I will contend that conceptual integrity is the most important consideration in system design... Simplicity is not enough. Mooers’s TRAC language and Algol 68 achieve simplicity as measured by the number of distinct elementary concepts. They are not, however, straight-forward. The expression of the things one wants to do often requires involuted and unexpected combinations... It is not enough to learn the elements and rules of combinations; one must also learn the idiomatic usage, a whole lore of how the elements are combined in practice. Simplicity and straightforwardness proceed from conceptual integrity. Each part must reflect the same philosophies and the same balancing of desiderata. Every part must even use the same techniques in syntax and analogous notions in semantics. Ease of use, then, dictates unity of design, conceptual integrity. 
Overall clarity is, thus, an important functional requirement for a piece of software. This requirement ties the different parts together and creates strong interdependence in programming.
It should not be surprising that clarity requirements enter here. The principles that govern human behaviour in playing a game like Eleusis and those that underlie the formation of custom are equally relevant for dealing with a computer language, or any other advanced item.  It is for this reason that clarity is an important aspect of functionality. With respect to everyday staple products such as cars, computers, or television sets, this may just appear as ‘good handling’ and straightforward operation to the consumer, but it is important for assembly and maintenance. Unclear design will translate into high costs of repair and unreliability after repair. It will make for a poor product.
Consider now the organization of the division of labour. The fundamental question is, which tasks are most economically coordinated through the market, and which are more logically coordinated within firms?
For tasks that involve substantive interdependence, the market is
1. Brooks (1975: 42-4).
2. The card game Eleusis has been described in S. 7.2.
a poor instrument for coordination. ‘Jointness of effort and activities... involves more cooperation among individuals than that of simple market exchange.’  While the interdependence may be reduced by defining standards for interfaces between production steps, this will not always be feasible, and it may be costly. Yet even such standards will not necessarily lead to the formation of markets. Decentralization of interdependent tasks may also occur on a large scale through custom. The rules of traffic illustrate a decentralization of interconnected activities without a firm or a market.
But even separable tasks may be organized more easily within a firm than through a market. All the incentives that the market provides could also, theoretically speaking, be provided within the firm, while the overhead of additional marketing costs could be saved. Coordination could only be improved, because the firm as an organization offers a vastly richer menu of coordination devices than does the market.  It is easier to develop specific routines, or particular languages, within the firm than in the market, and informal ways of cooperation are more readily available.  Inserting a market between two steps introduces marketing costs into the production process and drastically reduces the possibilities of coordination by other means.
Thus, while the market has strong drawbacks with regard to interdependent tasks, it is also difficult to pinpoint the advantages of market organization for separable tasks. Instead of asking ‘why firms?’ the difficult question to answer is ‘why markets?’ 
1. Alchian (1991: 233). Demsetz (1988: 157-62) has articulated this view in terms of information costs.
2. ‘Perhaps the most distinctive advantage of the firm, however, is the wider variety and greater sensitivity of control instruments that are available for enforcing intrafirm in comparison with interfirm activities’. (Williamson 1971: 113).
3. This has been stressed in particular by Hirschman’s (1970) insistence that ‘voice’ be granted an important role in the theory of the firm.
4. This has been stressed by various authors, e.g. Coase (1937: 43: ‘Why is not all production carried on by one big firm?’); 0. E. Williamson (1985: 132-8: ‘A Chronic Puzzle’), Von Weizsacker (1991: 100: ‘It is difficult to prove, rather than simply state, that there are functions performed better by the market rather than any possible alternative.’). The problem is sometimes referred to as ‘Williamson’s puzzle’ or the ‘centralization paradox’; see Stiglitz (1991: 18). O. E. Williamson (1985: ch. 6) and, more recently, Putterman (1995) provide excellent discussions of various facets of the problem. My concern here relates to more general aspects and the nature of custom.
In answering the question ‘why firms?’, Coase rejected the idea that people might prefer to work in firms rather than for the market: ‘Such individuals would accept less in order to work under someone, and firms would arise naturally from this. But it would appear that this cannot be a very important reason, for it would rather seem that the opposite tendency is operating if one judges from the stress normally laid on the advantage of “being one’s own master”.’  This suggests that people may have a preference for market transactions. However, at least if real-world experience may be taken at face value, this does not seem to be true. The misleading conclusion arises because Coase stresses command rather than custom as the distinguishing mark of the firm. A job offers a range of competencies and responsibilities, and the worker is, within the confines of his job, his ‘own master’. He may actually perceive more autonomy in performing his job than he would have in organizing his work as an entrepreneur. If the firm were his own, he would then also have to be alert to changes in circumstances and market conditions. The desire of housewives to obtain jobs seems to flow from the perception of personal autonomy in a job. Further, the firm provides an environment for social interaction which is sometimes valued highly; the arguments against ‘telecommuting’ (working with a computer at home) illustrate this. In the end, there seems to be a preference for working in firms. This renders the question ‘why markets?’ even harder to answer.
The firm, as a system of highly integrated routines, reward systems, and command structures, benefits from integration. The productivity gains obtainable are due not only to improved coordination by means of specialized and well-matched routines, languages, and tacit ways of cooperation, but also to the strong impact on motivation arising from the formation of a group. The
1. Coase (1937: 38).
firm generates group forces that emanate from the same clarification processes that lead to interlocking customs. A set of interlocking customs provides many diffuse reasons for working and thinking in a certain way. Self-attribution will not be able to link behaviour to such a diffuse pattern of causes. Thus, self-attribution will focus on intrinsic motivation and will generate corresponding attitudes and motives.  In this way, ‘corporate culture’ - or, better, ‘corporate custom’ - shapes cognition and motivation. It is a productive asset and arises from the same forces that shape different cultures. Such group forces can be fairly strong. (They occasionally cause war.) Firms make use of them. 
But a system of highly integrated routines, rewards, and command structures has drawbacks because it is highly integrated. While integration shapes motivation and bundles information in a most parsimonious manner, it ties everything together and in that way creates rigidity. What has been said about interlocking customs applies here, too.  Coordination by routines may be efficient, but it requires the routines themselves to be inflexible.
Such inflexibility relates to the rules rather than to the actions that they generate. A firm may respond very quickly to changes in circumstances as long as this change can be handled by the prevailing routines and command structures. An army is organized in order to respond quickly, but it does not use the market mechanism to achieve that aim: rather, the organization is designed for that purpose. The routines and command structures are themselves fairly rigid.
Integration binds together everything within an organization. Consider pay. All textbooks on compensation stress the
1. See S. 9.7.
2. On the foundation of groups, see Asch (1952: p. III and V). Such group phenomena are obviously important and have been emphasized in quite early writings, as can be seen from the quotation of Marx given in fn. 1, p. 241 above. Yet the theory of the firm has not taken account of this, and has not even tried to refute group effects on factual, rather than paradigmatic, grounds.
3. See S. 10.8.
overwhelming importance of ‘internal consistency’, and practitioners agree. The following quotation from a textbook on compensation illustrates this:
Internal consistency refers to pay relationships among jobs or skill levels within a single organization. It is one of the basic compensation policies. It involves equal pay for jobs of equal worth and acceptable pay differentials for work of unequal worth. But internal consistency involves more than the pay structure. Often called internal equity, a policy that emphasizes internal consistency places importance on the inner workings, the relationships and pressures found within an organization. Consequently it includes concerns for the fairness of the procedures used to establish the pay structure, as well as the structure itself. 
The phenomenon itself hardly needs further elaboration. The various techniques used to determine compensation demonstrate the practical importance of these considerations. The equity standards are undeniably of great importance within organizations. They cannot be interpreted merely as surface appearances of underlying economic considerations. Clarity requirements are not reducible to scarcity arguments or to the logic of rationality. A 50-50 split of outcomes, as used in the Halsley method of employee compensation, can hardly be explained in terms of scarcity and incentives alone:  it relates to symmetry and clarity. Similar observations apply to a host of other compensation methods.
These remarks should not be taken to suggest that incentives and scarcity do not play a role: incentives and scarcity are obviously of the greatest importance. The point is rather that clarity requirements constrain the possibilities of obtaining the full and complete return from these instrumental considerations. The costs of custom emerge, just as do the benefits, from the need for coherence.
Many examples illustrate this feature. Wages for the same type of work vary systematically across industries and firms.  Trans-
1. Adapted from Milkovich and Newman (1984: 31-2).
2. Milkovich and Newman (1984: 343). See also the criticism of game-theoretical approaches to custom in S. 10.2.
3. These ‘firm effects’ and ‘industry effects’ may, however, also have other explanations; see Oi (1990), Krueger and Summers (1988), Blackburn and Neumark (1992).
portation may be contracted out in order to avoid compensating the truck drivers according to the firm’s own corporate culture;  cleaning, maintenance, and service may be contracted out for similar reasons. Some takeovers occur because the merger largely voids the compensation practices of the firm that is swallowed up. This may be a profitable and painless way of replacing one set of constraints for another. 
Other constraints resulting from integration relate to standards of quality. Work routines are geared to such standards. Typically, each firm predominantly serves a particular quality segment of the market even in cases where substantial economies could be expected from integrating high-quality production lines with low-quality products. 
These consistency constraints are well known, but less well understood. Why, it may be asked, are they so important? It has been shown, after all, that people get used to nearly everything. Psychological studies of adaptation
suggest that any stable state of affairs tends to become acceptable eventually, at least in the sense that alternatives to it no longer readily come to mind. Terms of exchange that are initially seen as unfair may in time acquire the status of a reference transaction... The gap between what people consider fair and the behaviour they expect in the marketplace tends to be rather small. 
Custom is adaptive, after all. This argument is misleading. Even if workers can adapt to the most unintuitive regulations, this attempt is bound to create control problems. If the set of routines is augmented by a routine that embodies a different philosophy and style, each of the earlier established routines will lose some of its imperative character and a part of its impact on behaviour. Similar effects are well-known in software engineering. An entire program may suffer if features are added that do not ‘fit’; then,
1. Kubon-Gilke (1997: 5. 5.3.6).
2. O. E. Williamson (1985: 158) reports on an acquisition that went wrong because of internal equity constraints.
3.Wagner (1994: 107-25).
4. Kahneman et al. (1986b: 731-2); see also Major and Testa (1989).
whenever a further feature is added, it is uncertain whether it follows the old style or the new one. The delays and errors induced by this uncertainty may easily outweigh the benefits of the added features.  Similarly, the rule to do one job quickly and the other with the utmost scrutiny will induce workers to perceive the approach to work as somewhat arbitrary. The labour force will not easily develop a coherent style, an appropriate set of habits, and a functional structure of motivation. Furthermore, a mixing of principles of compensation will put the entire compensation policy of the firm at risk. Each worker will perceive the benefits that would accrue to him if he were treated like the others in some particular aspects. He will not consider that he may be benefiting in other dimensions. This creates internal stress and is detrimental to corporate custom.
The market has been interpreted as a system that enables economic agents to make use of widely dispersed knowledge in a parsimonious manner. Prices serve as signals that encode all relevant information about scarcities and make this knowledge available throughout the social and economic system. It is, thus, not necessary to enquire further about specific scarcities, wants, or production conditions. 
The earlier discussion of the role of routines within firms has stressed efficiency aspects of coordination by custom in a parallel manner.  Routines are maintained by following them; what is not repeated is forgotten. This implies that only useful knowledge survives. The routines embody that knowledge, even if the individuals are not aware of the fact. By allocating responsibilities to job holders, the idiosyncratic expertise of these workers can be used to the benefit of the organization. In this sense, a firm-specific set of
1. See S. 14.7.
2. Hayek (1945).
3. See S. 13.11. Demsetz (1988: ch. 9) has argued e.g. that firms can thrive in a market environment because they are more efficient in economizing on information flows; see also the discussion of separability and independence in S. 14.6 above.
customs may be conceived as a mechanism that enables the members of the firm to make coordinated use of widely dispersed knowledge.
This observation holds true in society as well. There are national economies that are coordinated mainly by custom. People follow the established behavioural patterns, and an overall coordination is achieved without any explicit awareness. Each individual needs only to know how to perform his allotted task and to follow the role prescription. His entitlements and obligations, his duties and routines transmit all the information that is necessary for him to coordinate social and economic activity. The market mechanism is not unique in this respect. Hayek, who (following Menger) developed the view of the price system as making use of dispersed information, was quite clear about this:
We make constant use of formulas, symbols and rules whose meaning we do not understand and through the use of which we avail ourselves of the assistance of knowledge which individually we do not possess... The price system is just one of those formations which man has learned to use. 
Custom is another formation of this kind. If the market is a marvel, custom is a marvel, too.
Furthermore, command may also be understood as economizing on information. By centralizing the power of decision-making, it is possible to set up an efficient means of gathering information, and to distribute to the individual actors only those pieces of information that are necessary for them to take the appropriate action.
Thus, all three control modes - exchange, command, and custom - can be viewed as economizing on the use and distribution of knowledge. The market is not necessarily the most efficient way to deal with informational issues. Rather, it is conceivable that there are several kinds of information, and that the different control modes have specific advantages or drawbacks with respect to gathering, channelling, and distributing that intelligence.
1. Hayek (1945: 528).
It is, however, misleading to put exclusive stress on the informational aspects of economic allocation mechanisms. Motivation is equally important in the economy.
It has been argued that the problem of economic coordination arises only under conditions of change.  Under stationary conditions, all transactions are repeated over and over, and the task of coordination vanishes. Once a transaction pattern is established, no further need for coordination arises. Mere repetition will suffice to organize the division of labour.
The organization of the division of labour becomes crucial - and becomes intricate under conditions of change. Change necessitates adjustment. It therefore becomes important to understand how the economic organization of society generates and digests the change that is needed.
The division of labour within a given society will rely on certain organizational structures which are fixed in the short run. This holds true for markets, hierarchies, and custom. Each of these control modes is characterized by a certain rigidity in the underlying rule system. This renders it possible to cope with certain types of change but not with others.
Consider markets. They work best with standardized products, because standardization renders it unnecessary to evaluate and communicate quality parameters. More generally, it has been argued that ‘for anything approaching perfect competition to exist, an intricate system of rules and regulations would normally be needed’.  Commodity and stock exchanges provide examples.
Any hierarchical system of command, moreover, depends on a certain institutional structure. Competencies must be defined, and chains of command determined. In a similar way, control by custom relies on fixities of behavioural patterns.
1. Hayek (1945: 524), von Weizsacker (1991: 102-3,109-10).
2. Coase (1988: 9).
Thus, all three modes of control exhibit rigidity in some dimensions and flexibility in others. This suggests that each of the control modes may enjoy a differential advantage in coping with certain types of change. Markets, for example, can cope more easily with changing demands and supplies, whereas hierarchies and customary forms of organization may find it more difficult to digest such changes.  Changing traffic flows can be better coordinated by a system of traffic rules rather than by the auctioning of priority rights. Fires and other disasters can be better dealt with by a hierarchical organization relying on command than by prices and markets designed to allocate rescue forces.
The organization of interdependent tasks poses difficulties for the market. The set of instruments available for direct coordination within and between firms is much richer than that available in the market. On the other hand, the market, in spite of informational shortfalls, is better for organizing independent tasks because it unleashes the forces of competition. Through competition, it overcomes the rigidity of custom that impedes dynamic advance in closely knit organizations. 
Yet the perceived interdependency of tasks can attenuate over time. Once routinization takes place, it leads to a standardization of task interfaces. Standardization greatly reduces the need for personal communication. Each task module becomes more separable (and, possibly, internally more interdependent). In this way, the use of the market as a coordination device is rendered feasible.
The picture emerging from this discussion is as follows. New technological developments are organized first within firms or as joint ventures between firms, without strong reliance on the market. This is because it is difficult to use the market for coordination purposes for entirely non-standardized and volatile interdependent tasks. In the course of time, the project takes shape, routinization develops and takes hold, and standardization becomes possible. This reduces task interdependencies, which in
1. However, the internal organization of firms may respond to different challenges. Burns (1963) has argued, for instance, that change requires an ‘organismic’ rather than a ‘mechanistic’ organization structure.
2. See S. 14.8-14.10.
turn renders it increasingly advantageous to introduce the market as a coordination device for organizing the division of labour. There emerges a tendency to organize non-standardized change within firms, and standardized change through the market.  Firms are, so to speak, the nurseries for new projects. Once the projects have achieved an independent identity, they will spin off as separate firms. Those firms will then develop a special organizational culture and a special set of customs of their own.
The clarity view thus implies a view about the division of labour between the firm and the market. Both the firm and the market rely on exchange, command, and custom, but the firm permits tight integration and the development of specialized customs. This eases production and enhances motivation. Yet the tight integration within the firm comes at the cost of internal rigidity. The mutual adjustment of routines, command structures, and internal pricing schemes renders it difficult to implement change. Market interaction is less specialized and unleashes competitive forces to overcome organizational inertia.
The argument comes down to explaining the organization of the division of labour within firms and across the market in terms of the costs and benefits of integration. Where the benefits of integration outweigh the costs, integration will occur; where the costs outweigh the benefits, transactions will be carried out in the market.
The last statement is, by itself, as empty as the statement that transaction costs determine integration. Yet the view expounded here puts things differently. It stresses that the functioning of institutions is not reducible to their various components in an additive way. The stress on ‘chemical’ interaction is directed
1. A related view has been outlined by von Weizslcker (1991: 109-13). He argues that firms are better in handling slow change, and markets are better in handling fast change, and that the incentive to innovate falls as market concentration increases, giving rise to an ‘equilibrium theory of innovation’.
against nominalistic views of the firm, the market, and other institutions. The tight alignment of routines, pricing procedures, and command structures, both across the market and within firms, emanates from the same tendencies of the human mind that fashion customs in all spheres of life. They relate to processes of overall clarification.