Knowledge Management: Organizational Economic Insights

In this section, we shall more concretely apply specific organizational economics insights to two clearly central aspects of knowledge management: knowledge creation and knowledge integration. The former category encompasses learning (by doing, using, being instructed, etc.) and innovation processes such as knowledge combination, while the latter refers to how to make best use of existing knowledge in the firm. We develop propositions based on organizational economics regarding how firms may stimulate investments by employees in firm specific knowledge, resolve incentive problems in knowledge-creating teams, and make choices between alternative means in the integration of knowledge, including knowledge sharing.

Knowledge creation

It is now almost an axiom that knowledge creation in firms lies at the heart of competitive advantage (Nonaka and Takeuchi, 1995; von Krogh et al., 2000; Nonaka and von Krogh, 2009). Expressions such as ‘firms learn’ and ‘firms know’ have become commonplace in much of the strategy and knowledge management literature.3 However, it is not firms as such that learn, and firms themselves do not possess knowledge. So-called ‘firm knowledge’ is composed of knowledge sets controlled by individual agents. We stress this admittedly basic methodological individualist point in order to emphasize the point that by focusing on the level of the individual agent, rather than the firm, organizational economics highlights questions that are neglected in the knowledge management literature because much of this literature operates on the firm level and does not have an explicitly individualistic starting point.

In particular, an organizational economics perspective directs attention to the possible incentive conflicts that may arise in connection with issues such as, for example,

Perhaps somewhat contrary to intuition, such questions are central to successful knowledge management in practice and they are particularly prone to an organizational economics treatment. This is because processes of creating knowledge—for example, in the form of innovation projects—are typically risky, unpredictable (the knowledge-to-be-created can only be partly foreseen), often long term, labor intensive, idiosyncratic (that is, hard to compare to other processes), and often require substantial human capital investments (Holmström, 1989). A number of these characteristics are the basic stuff that contracting problems are made of.4 In the following we discuss a number of ways in which firms may motivate employees to expend effort in the production of new knowledge. In this connection, we discuss how the return stream from such new knowledge is shared between the firm and the employee. Thus, the problems of motivating employees and capturing rents from new knowledge are two sides of the same coin.

We assume throughout that an asymmetric information setting prevails, and that incentive conflicts are present. To see why these assumptions are appropriate ones, consider a world where asymmetric information and incentive conflicts (agency problems, hold-up problems) are absent. Here, the interests of the various agents involved in the creation of new knowledge can be easily aligned. First, employees and employers would assess the value of new knowledge in the same way (because information about this is symmetric). Second, bargaining will take place immediately, because the symmetry of information means that there will be no strategic behavior. Third, the employee’s reward for any learning investments will be guaranteed since the employer will not attempt to hold-up the employee. In a nirvana world where both employee and employer access the same information on the value of ideas and each other’s outside options, inducing optimal human capital investment can be achieved by writing complete contracts. If more realistic assumptions are introduced, an incentive perspective on knowledge creation is particularly appropriate, because it stresses not only that agents making learning investments must somehow share in the extra surplus from those investments to be properly motivated, but also that providing such motivation is no easy matter under asymmetric information, possibly incomplete contracts, and self-interested behavior.

Earning rents from knowledge creation

The knowledge management literature seldom makes clear exactly how the mechanism from knowledge creation to new rents works. However, the resource-based view in strategic management has gone some way towards clarifying this by identifying a set of criteria that resources must meet to be sources of (sustained) competitive advantages, such as being valuable, rare, and costly to imitate (Barney, 1991). Moreover, the relevant resources should not be fully mobile (Peteraf, 1993). Knowledge assets, particularly newly created ones, are particularly likely to meet these criteria (Winter, 1987). Given this, managers may wish to induce knowledge creation by means of providing incentives to employees to upgrade their own knowledge capital and by spending corporate resources on having employees do this (e.g. training, setting up incentives, etc.). From the perspective of the firm, earning rents from employee upgrading of knowledge is far from trivial. For example, accumulation of valuable knowledge capital in firms may require employees to make asset specific learning investments. In particular, whether or not firms are likely to earn rents from employees’ knowledge depends on (1) the type of learning investment (e.g. firm specific or general knowledge); (2) the resolution of agency conflicts in firms (e.g. remuneration schemes and promotion rules); and (3) transaction costs in labor markets (e.g. signaling and screening). We consider these seriatim.

Types of learning investments

Firms’ investments in augmenting the knowledge of their employees may be of two kinds, namely general and firm-specific ones. Both may increase an employee’s productivity, but they have different implications with respect to who is likely to appropriate the returns and who will carry the costs of the investment. General learning investments may increase an employee’s productivity in a range of employment opportunities. Such general investments include the learning of languages and generic skills, such as learning word processing programs, for example, that are equally useful for current and potential employers. Becker (1962) suggests that employees will pay for their general training, because in competitive markets they are the sole beneficiaries of the improvements of their productivity. A firm will not pay for an employee’s learning of general knowledge, because of the weakness of its bargaining position after having made the investment. In contrast, the learning of firm-specific knowledge restricts an employee’s possibility to capture returns on this knowledge outside of the firm that undertakes the investment. Becker (1962) argues that to the extent that an employee’s productivity increase exceeds his or her wage increase after learning, the firm can earn rents even if it alone incurs the costs of firm specific learning investments. As far as such investments are concerned, the relative bargaining position of firms is strong because employees cannot credibly threaten to leave the firm to bargain for higher wages that reflect their productivity increase after specific learning investments. Thus, it is very likely that firms will appropriate a substantial part of the relevant rents. Of course, firms that undertake more specific learning investments will also create more rents, because the benefits (e.g. in terms of productivity or increased innovativeness) are larger to the firm in the case of specific than in general learning investments. Thus, the following refutable proposition may be put forward:

Inducing firm specific learning

Consider next the situation from the perspective of employees. From their point of view, learning is an investment of effort for which they wish to be compensated. Firms will have to provide inducements for such investments. However, as we have seen, making firm-specific learning investments restricts an employee’s outside employment options (and therefore his or her bargaining power), which will tend to reduce firm-specific learning investments below the optimal level. Due to the incentive problem, undertaking these investments means becoming more vulnerable to managerial hold-ups. Resolving this problem turns on management’s ability to credibly signal that it will not take advantage of employees who by making firm-specific learning investments have put themselves at risk. An organizational economics interpretation of (beneficial) corporate culture is that it is essentially an embodiment of such signals (Kreps, 1990). Thus, firms with corporate cultures that credibly signal that management is committed to a non-opportunistic approach in dealing with subordinates will induce higher learning investments on the part of employees. Such a corporate culture makes the provision of incentives credible, so that employees correctly believe that management will not renege on promises with respect to compensation. With respect to the issue of providing incentives for employees’ investment in firm specific knowledge, organizational economics suggests at least four solutions:

• high-powered incentives (i.e. making employees more of residual claimants);
• promotion rules; and
• conferring access to critical resources.
• making credible commitments

Consider these in turn.

High-powered incentives

High-powered incentives—often represented as the contingent portion of pay—may be used to induce contributions through providing larger shares of quasi-rents to employees (Williamson, 1996). Firm specific learning investments may be induced by providing equity to employees (e.g. in the form of stock options or equity) or other high-powered incentives, such as performance pay (Demsetz and Lehn, 1989; Williamson, 1985). However, offering such high-powered incentives may also lead to a number of distortions. This is the case, for example, when the corresponding costs (e.g. of using the firms’ assets) are not borne by those to whom high-powered incentives are offered (Holmström, 1989). Thus, as Williamson (1985) argues, this is exactly why incentives in firms are often comparatively low-powered. Another problem with high-powered incentives is that they expose employees to considerable risks. For example, performance (e.g. the value of stock options) may fluctuate for reasons beyond an employee’s control. In addition, employees may be highly dependent on the fixed, risk free part of their income if they lack alternative sources of income. Risk-averse employees may therefore shy away from high-powered incentives. On the other hand, risk estimates may be in the eye of the beholder, and more highly skilled employees may judge risk differently from other employees. Moreover, for incentive pay to be effective, either observability of output or behavior must prevail. If behaviors or output for tasks cannot be specified as cause-effect relationships are not well understood, then high performance ambiguity poses a problem because neither behaviors nor outputs can be related to specific skill acquisition with any precision. Thus, the less output and behavior can be pre-specified so as to reflect employees’ specific skill development, then the less effective high-powered incentives become (Ouchi, 1980). Thus, the following refutable proposition may be put forward:

Promotion rules

The design of promotion rules is an alternative way of inducing firm specific learning investments. Consider inducing investments in firm-specific knowledge by means of ‘up-or-stay’ rules (e.g. the worker is either promoted or stays in the original job) relative to ‘up-or-out’ rules (e.g. the worker is promoted or fired) (Prendergast, 1993; Huberman and Kahn, 1988; Gibbons, 1998). Generally, when workers bear the costs of acquiring specific skills they will do so only if the wage (Ws) obtainable after skill acquisition minus their opportunity costs (Cs) exceeds current payment (Wus). The principal will pay the wage (Ws) only if the productivity difference (Ps − Pus) exceeds the difference of wages (Ws − Wus). With ‘up-or-stay’ rules principals distinguish jobs and attach different wages to them. This promotion rule creates a tension between needing a large enough wage gap to induce the worker to invest and keeping the gap small enough so that the principal is willing to promote the worker after the worker has invested (Prendergast, 1993). Gibbons (1998) illustrates this point, as follows.

For example, suppose that an untrained worker produces 10 in the easy job, that a trained worker produces 20 in the easy job and 30 in the difficult job, and that the opportunity cost of training is 15. Then training is efficient (30 − 10 < 15) but we cannot find wages that simultaneously induce the worker to invest (wage difference greater than opportunity cost, 15) and induce the firm to promote a trained worker (wage difference smaller than productivity difference, 30 − 20). As a consequence, employees’ investment in firm-specific skills may be low, although such investments would be efficient. Huberman and Kahn (1988) suggest that ‘up-or-out rules’ can solve this incentive problem. For example, with this rule the principal makes a commitment to promote the worker after a pre-specified time span or otherwise fire him (e.g. tenure in academic jobs, moving up career ladders in consultancies). The resulting rat-race creates incentives for investments in firm-specific knowledge. To illustrate, consider the example above. As before, specific learning investments only lead to firm rents when they are efficient (Ps − Pus = 15). If a worker expects promotion, he or she will invest at any wage (W*) which exceeds his or her opportunity costs plus the best alternative (e.g. W* = WALT +15). The principal promotes the worker if his or her productivity (Ps) exceeds his or her high wage (Ps = W*). Although with up-or-out rules there is always a wage (W*) that is low enough to induce the principal to promote the worker who has made sufficient investments in firm-specific capital, up-or-out rules come at a cost. Because it is not possible to keep the worker in the firm when the productivity after investment does not exceed his or her high salary, this up-or-out rule may waste investments in firm-specific skills. This is especially obvious when there are different layers where such up-or out rules apply and workers survive the first rounds but drop out at a higher level (cf. Gibbons, 1998).5 Thus, the following refutable proposition may be put forward:

Additionally, once employees have invested in firm specific capital, a firm also needs to tie employees long enough to the firm, so that firm specific human capital investments can be recouped. Turnover of key knowledge carriers is a major problem in this respect. Typically, to prevent turnover from happening firms use deferred rewards and pensions, which benefit employees only in the distant future (Milgrom and Roberts, 1990).

Providing access to assets

Firms may positively influence learning investments by providing access to critical resources (Rajan and Zingales, 1998), such as critical knowledge resources. Access may be defined as the ability to use or work with a critical resource including other human resources. It generates an opportunity for employees to specialize relatively to these assets. We earlier analyzed this as giving rise to a potential hold-up problem, since the firm may hold-up the specialized employee. However, the other side of the coin is that specialization to a critical asset in combination with an employee’s right to withdraw his or her (also critical) human capital gives him or her considerable bargaining power with respect to the sharing of the surplus from productive activities, that is, bargain for a higher salary. It can be shown that when investments are additive (i.e. the total surplus is dependent on the sum of the investments), granting access and, as it were, giving away bargaining power, may be a superior incentive mechanism to induce firm specific learning. In contrast, when investments are complementary (i.e. the marginal return of one investment rises in the level of the other investment), which is likely to take place in team-based firms, we are back to the familiar hold-up problem (Williamson, 1985; Hart, 1995). Not only will the employee directly influence the size of the surplus if he or she withdraws their human capital; he or she will also influence it indirectly, because his or her human capital investments are complementary to the human capital investments of other employees. In this situation, it will not be advantageous to grant the employee (too much) access (see Rajan and Zingales, 1998 for details).

The three mechanisms above (incentives, promotion rules, access to resources) may be substitutes or complements, depending on the circumstances. Thus, tournaments in the form of up-or-out rules may be a substitute for performance pay when employees are sufficiently risk-averse. Access may replace incentives in the same situation. Promotion rules and incentives may be swapped for access, when giving an employee access would be giving him or her too much bargaining power. On the other hand, all three mechanisms are often seen together; for example, in consultancies partners have obtained their position through a tournament that works according to certain promotion rules, they are granted access to assets contingent on learning investments, and they are usually residual claimants. We may now put forward the following proposition:

Making credible commitments

The above analysis of firm-specific human capital has made the simplifying assumption that costs of concluding labor market transactions can be neglected. This, of course, is not the case as such costs aggravate complications of inducing firm specific investments. Asymmetric information between current and potential employers is one source of switching costs in labor markets (Akerlof, 1970). Employees must search for new job opportunities and firms must search for fitting employees. In this search process, there may be several complications. For example, a current employer usually knows more about employees’ human capital and learning ability than potential employers do (Spence, 1973, 1974). In wage negotiations employees will have to credibly signal to new employers their ability to perform. However, because some employees will overstate their ability in order to drive up wages, employers will not only incur costs of screening employees, but may also reduce wages offered to account for the risk of picking a wrong employee (i.e. a lemon). If this is the case, employees willing to switch from their current employer would find the wage offered by new employers unattractive. The higher transaction costs in labor markets are, the more difficult it is for employees to switch between employers. By implication, high transaction costs in labor markets lower incentives for employees to invest in firm specific knowledge without appropriate safeguarding and compensation. Thus, firms that operate in labor markets with high transaction costs will incur greater costs to induce employee’s firm specific learning compared to firms that do not.

One particularly interesting way to induce firm specific learning in such situations is to offer employees the possibility to engage in the acquisition of certified general knowledge such as management training or language and computer skills (Laing, 1994). Employees might face lower lock-in as a result, because the acquisition of certified general skills reduces labor market transaction costs such as screening and matching (Spence, 1974; Barzel, 1982). Nonetheless, a firm offering such general training possibilities to its employees can benefit in several ways. First, investments in general skills can increase the productivity effects of firm-specific skill investments because common knowledge between employees facilitates the combination and blending of specific skills (Kogut and Zander, 1992; Foss, 2001). Second, sponsoring general training as a form of pay also signals the commitment of employers to their employees (Kreps, 1990), and that their investments in firm-specific knowledge will not be opportunistically exploited. Thus, the following refutable proposition may be put forward:

Knowledge creation in teams

Many contributions to the knowledge management literature recommend the use of teams in the form of work groups, inter-disciplinary, and cross-functional teams to foster knowledge creation (e.g. Brown and Eisenhardt, 1995; Meyer and deTore, 1999; von Krogh et al., 2000). Teamwork may bring knowledge together that hitherto existed separately, resulting in ‘new combinations’ (Schumpeter, 1950); it may facilitate cross-functional communication, cross-fertilization of ideas, and enhance worker involvement. Through the integration of knowledge of individual members, teams may not only blend knowledge and insights beyond what individual members may achieve, but the development of new knowledge may also be stimulated by conversations and language-based learning in teams (Brown and Duguid, 1991; Nonaka and Takeuchi, 1995). However, while knowledge creation in teams has its virtues, there are special difficulties associated with aligning interests of team members (Scott and Einstein, 2001). Not only will teams be particularly prone to moral hazard, notably in the form of shirking, but the right form of incentive may also be contingent on the type of team at hand. Questions arise that remain neglected in the knowledge management literature such as: Who should be rewarded—teams or individuals? Who should evaluate contributions of team members—other team members, a specialized monitor, or an external manager? What measures of performance should be used and when? An organizational economics perspective suggests that the success of teams’ knowledge-creating efforts depend, inter alia, on (1) the size of the team, (2) trade-offs between individual and team incentives, (3) exclusion rules, and (4) matching the varying degrees of uncertainty to incentive design.6

Free rider problems and team size

Alchian and Demsetz provide a classic treatment of incentive problems in team-production—a process ‘wherein individual cooperating inputs do not yield identifiable, separate outputs’ (1972: 779). Where measuring individual input productivity and rewarding accordingly become difficult, team members may free-ride on other team-members’ contribution to knowledge creation. This is so because the benefits of withholding marginal effort accrue to each shirking member while the resulting losses accrue to the team as a whole. In principle, knowledge production in teams could be organized through a set of bilateral agreements between team members who promise best effort and ensure mutual control. However, such agreements are difficult to manage and will most likely incur large resource costs; for example, time spent on negotiation and haggling means that less time is available for knowledge creation. As teams grow in size, the larger these costs become, in fact, they increase exponentially with the number of team members (Rosen, 1991). In addition, free rider problems become more prevalent, the larger the knowledge-creating team becomes. Thus, one can derive the following refutable proposition:

Individual and/or team incentives

Team size problems are aggravated if incentives are exclusively allocated to a team as a whole rather than also considering incentives for individuals (Laursen and Mahnke, 2001). When capable and willing team members are forced to support free riders, they often withdraw effort or else leave the team. On the other hand, relying exclusively on individual incentives can inhibit cooperation in teams—especially when task performance crucially depends on the exchange of information and mutual adaptation (Thompson, 1967; Balkin and Gomez-Mejia, 1992). Nonetheless, many recommendations in the knowledge management literature are mistaken when they note that individual rewards may be the antithesis to teamwork. An organizational economics perspective urges managers not to neglect possibilities to induce individual contributions on which team performance ultimately rests.

One possibility for resolving incentive conflicts in the knowledge-creating team is that a team member specializes in monitoring other members’ contributions to generate reliable information based on which rewards may be distributed (Alchian and Demsetz, 1972). A positive effect of monitoring is that knowledge about talents is discovered which can be used to reduce shirking but can also lead to better recombination or new uses of skills and talent. However, as specialized monitors become increasingly removed from actual teamwork, possible knowledge gaps between those creating new knowledge and those specializing in monitoring may increase over time, eventually compromising effective monitoring. As an alternative, management may provide incentives for achievements of the group as a whole and let the group members distribute team rewards among themselves based on subjective performance evaluation (e.g. 360 degree reviews).7 This utilizes the fact that team members will often have information about each other’s contributions, behavior, and ability that is superior to that of external management (Gibbons, 1998). Thus, specialized incentive procedures may cope with some of the incentive problems by combining incentives to teams with incentives to individual team members. This leads us to the following refutable proposition:

Exclusion rules

We mentioned earlier that firms often use promotion rules in order to solve incentive conflicts by setting up competition between employees. Similar mechanisms may reduce incentive problems in teams. Lazear (1989) suggests that tournaments may involve self-selection and exclusion mechanisms. These drive up effort levels, because only those who believe in their survival and exercise effort and skills in a team’s knowledge creation effort are attracted (Dillard and Fisher, 1990). In particular, giving teams the right to exclude team members (Lazear, 1989; Malcomson, 1998) on the basis of subjective performance measures (e.g. peer evaluation, group leader assessment, or a combination) is clearly relevant in this context.

Setting up tournaments inside firms may be a viable control mechanism in team-based knowledge creation. But it can also be dangerous. If tournament rules cannot exclude sabotage among team members they may lead to outright breakdown of knowledge creation in teams (Lazear, 1989). An exaggerated emphasis on competition may also drive out exploration by team members who prefer to make quick wins through exploiting ideas of others rather than exploring new ideas on their own. This has two harmful effects on the knowledge-creating team (March, 1994). First, explorers benefit from developing absorptive capacity based on which they can pick up good ideas that others engaged in the same team process cannot exploit on their own. The less others involved in the knowledge-creating team are able to develop and exploit ideas themselves, the more important it becomes that others can relate to their ideas. Second, as team members increasingly engage in exploitation to the neglect of exploration, fewer ideas are available for exploitation. When competition provides disincentives for exploration and the revealing of ideas openly, the loss of relative absorptive capacity (Lane and Lubatkin, 1998) among team members diminishes the capacity for knowledge creation in the team as a whole.8 Thus, we suggest the following refutable proposition:

Uncertainty and team types

Knowledge-creating teams may operate under varying degrees of means and end uncertainty. To illustrate, the knowledge management literature distinguishes two types of knowledge-creating teams: ‘communities of practice’ and learning in ‘epistemic groups.’ The former denotes a team of peers who learn during and about the execution of pre-specified tasks with defined outcomes (Lave and Wenger, 1990; Brown and Duguid, 1991; Brown and Duguid, 1998).9 The key problem is how to create knowledge about means whose ends are well known. Examples include how to fix a working process that has broken down, how to deal with customer demands more quickly, etc. By contrast, ‘epistemic communities’ deal with knowledge creation for non-routine problems whose ends and means cannot be specified ex ante (Cohen, 1998). Here the key problem is how to discover means for ends that are unknown at the time the team starts developing knowledge. An example comes from a knowledge management team at a software security firm that described their situation as follows: ‘In 2–3 years’ time, our company will be designing security products we don’t know, incorporating technologies which haven’t been invented, made in processes yet to be defined, by people we have not yet recruited.’

One complication of means and ends uncertainty is that both complicate the provision of incentives in a team. This is because measurement bases for the provision of incentives become increasingly noisy the less means and end can be pre-specified ex ante. In other words: uncertainty leads to performance ambiguity, which complicates the provision of incentives (Ouchi, 1980). Only if performance ambiguity is low does performance pay seem effective in aligning conflicting interest. If this is not the case, variable rewards might be appropriate if pay and control can relate to specified behavior or to other forms of standardization (e.g. processes) which can serve as a basis for measuring performance. Unfortunately, to the extent that standardization of behavior or processes is prevented, such as in the case of many epistemic communities, neither behaviors nor outputs can be determined with precision. In this case, Ouchi (1980) suggests, clan control might be the solution to promote cooperation and mitigate conflict of interest: the basis of control becomes a set of internalized values and norms. It should be noted, however, that clan control can lead to normative fixation and group thinking that are both detrimental rather than conducive to knowledge creation in teams (e.g. Grandori, 2001). Comprehensive empirical research regarding managerial control dilemmas in knowledge-creating teams remains sparse and inconclusive. However, contrary to popular recommendations in the literature to abandon incentives in favor of normative control altogether,10 recent evidence shows that incentives for knowledge-creating teams seem to prevail in practice across a number of industries (Laursen and Mahnke, 2001; Laursen and Foss, 2002). An organizational economics perspective on knowledge creation would not expect otherwise. Thus, we suggest the following refutable proposition:

Nonetheless, as we move from inducing individual learning to knowledge creation in teams, complications of providing incentives have vastly increased. Given these complications of knowledge creation in teams, an organizational economics perspective suggests that team-based learning is a particularly expensive knowledge-creation mechanism that is riddled with many problems that include but are not limited to providing incentives. Seen this way, organizational economic insights might serve as reminders that knowledge creation in teams yields benefits at substantial costs. These may be compared to the benefits and costs of individual learning in firms as well as hiring of external expertise in the form of employment or contingent work, two alternative mechanisms of organizational learning (Simon, 1991).

Integrating knowledge: insights from organizational economics

Organizational economic insights (Coase, 1937; Demsetz, 1988; Jensen and Meckling, 1992; Williamson, 1985) have already substantially fertilized the literature on knowledge in organizations that characterizes the firm as a knowledge-integrating institution (Conner and Prahalad, 1996; Grant, 1996; Kogut and Zander, 1992, 1993).11 Therefore, this section is restricted to briefly reviewing key insights on knowledge integration needs and mechanisms.12

Specialization of tasks leads to focused learning in narrowly defined domains (Smith, 1978). However, because the division of tasks also leads to the division of knowledge, knowledge integration may be required when several activities are interdependent and individuals need to adapt their action to each other (Thompson, 1967). If individuals are specialized in different knowledge domains this will limit the rate at which knowledge that lies outside a narrow specialization can be assimilated, accumulated, and applied (Simon, 1991; Lane and Lubatkin, 1998). Three coordination mechanisms may be conducive to address such knowledge-integration problems—direction, common knowledge, and autonomous adaptation—but their efficacy may vary with varying task dependencies at hand.

Autonomous adaptation is the marvel of market. As Hayek (1945) argues, markets (be they between or in companies) make individuals do desirable things without anyone having to tell them how to do them. While the price mechanism economizes on investments in common knowledge, it only facilitates thin communication among individuals that co-ordinate their tasks and action. Its applicability may also be limited to situations where task-coordination is signified by low uncertainty and low interdependence between tasks that make autonomous adaptation possible (Grandori, 2001). Moreover, pricing knowledge in exchange faces a fundamental paradox: the value of knowledge to a purchaser is not known until after the knowledge is revealed; however, once revealed, the purchaser has no need to pay for it (Arrow, 1984). Second, Arrow also argues that ‘authority, the centralization of decision-making, serves to economize on the transmission and handling of knowledge’ (Arrow, 1974: 69). Demsetz (1988) agrees when he suggests that ‘[d]irection substitutes for education (that is, for the transfer of the knowledge itself).’ For example, employees transfer reports and memos rather than the knowledge on which they are crafted; superiors give advice on what to do and intervene at times rather than transfer the knowledge on which their judgment is based. Building on this argument, Conner and Prahalad (1996) stress that authority not only provides a low cost method of communicating, but also allows the flexible blending of expertise when contingencies emerge that were not foreseeable when, for example, an employment contract was concluded. This nicely corresponds to Coase (1937) who makes co-ordination by entrepreneurial direction based on employment contracts the distinguishing mark of the firm as an institution. Like price coordination, direction economizes on investments in common knowledge. In addition direction saves communication cost not because communication is restricted to thin communications as was the case with price coordination, but because communication (be it thin or thick) is restricted to top-down interaction on particular occasions. However, the application of top-down direction to coordinate knowledge finds its limits when superiors do not understand what and how results are achieved at a lower level—as is often the case with knowledge work (Foss, 1999, 2002). Finally, common knowledge (Grant, 1996) in the form of combinative capabilities, routines, shared context or codes, or social capital (Kogut and Zander, 1992, 1993; Nelson and Winter, 1982; Nahapiet and Ghoshal, 1998) may ease coordination, particularly when tasks are highly interdependent. However, as a discussion of knowledge-codification tools illustrates, investments in common knowledge and knowledge sharing—both in terms of managerial effort (see Zollo and Winter, 2002) and aligning diverging interest (Mahnke, 1998)—is particularly expensive. Thus, an organizational economics perspective suggests: