9.2 The costs and benefits of divisional and functional organisational forms
Under the divisional structure, also known as the “M-form,” the firm is organised as a collection of self-contained divisions, each of which has full responsibility over a subset of projects, and needs to perform all tasks associated with these projects (eg, production, marketing, finance, human resources, R&D, etc.). Under a functional structure, also known as the “U-form,” the firm is organised as a collection of functional departments, each of which specializes in one task and performs it on all projects that the firm undertakes. Therefore, under the functional structure, each project is executed by a team of agents who belong to different functional departments. The study of organisation designs was pioneered by Chandler (1962), who argued that as firms like DuPont, General Motors, Sears, and Standard Oil grew and adopted more diverse product lines, the difficulties in coordinating functions across product lines induced them to switch from the functional structure (U-form) to the divisional structure (M-form). Chandler concluded that a firm's structure follows its strategy which determines the number and types of its product lines.
Several recent papers examine the choice between the functional and the divisional structures. These papers consider a firm that produces two goods which require two tasks each. The divisional structure corresponds in this framework to the grouping of agents according to products, whereas the functional structure corresponds to the grouping of agents according to tasks.
Aghion and Tirole (1995) consider a model in which the functional structure requires agents to specialize in specific tasks and hence economizes on the cost of training agents, but the divisional structure strengthens the agents' incentives to exert effort by generating better external signals about their talent. They show that as managerial work load increases, the divisional structure becomes more attractive relative to the functional structure since then, the manager relies more often on the agents' decisions and this improves their ability to signal talent to the external job market.
In Rotemberg (1999), the firm can better control agents who perform the same task under the functional structure, but cross-task coordination is more efficient under the divisional structure. He shows that the divisional structure dominates the functional structure when the number of employees is sufficiently large.
Qian et al (2006) consider a model in which the divisional structure eliminates the need for costly cross-division communication to coordinate tasks, but the functional structure economizes on the cost of coordination by coordinating tasks on a company-wide basis. The divisional structure is particularly attractive in their model when there is a need for local experimentation of uncertain innovations involving several tasks; such experimentation is inefficient under the functional structure due to the need for costly communication among different divisions that engage in different tasks.
In Maskin et al (2000), the functional structure exploits economies of scale by grouping similar tasks in the same division, but the divisional structure provides better incentives because it promotes yardstick competition among similar divisions.
Besanko et al (2005) focus on the role of risk aversion: under the divisional structure, the compensation of agents depends only on their own (risky) performance, whereas under the functional structure it also depends on the (risky) performance of other agents. Hence, agents must receive a larger risk premium under the functional structure in order to induce them to exert the same level of effort. This result may be reversed, however, if there are significant asymmetries in the contribution of the tasks to profits, or significant positive externalities across tasks.
Corts (2007) considers a model with two possible configurations of tasks: “individual accountability” (which is akin to the divisional structure), where each agent is compensated on the basis of a single (noisy) performance measure that depends only on the agent's own effort, and “teams” (which is akin to the functional structure), where compensation is based on two (noisy) performance measures which depend on the agents' joint effort. Individual accountability has the advantage of compensating the risk-averse agents on the basis of only one noisy performance measure rather than two. The disadvantage of individual accountability is that the firm uses a single performance measure to evaluate the two tasks that each agent performs, whereas under teams it uses two performance measures.
In Harris and Raviv (2002), the comparison between the divisional and functional structures depends on the likelihood that various cross-task interactions will be realized, as well as on the CEO's cost of co-ordinating company-wide interactions between all tasks. For a wide range of parameters, both the functional and the divisional structures are dominated by either the matrix form where each task is coordinated by two different middle managers, or by a flat hierarchy where only the CEO may coordinate cross-task interactions.
Berkovitch et al (2010) consider a firm that consists of a board of directors, a manager, and two agents (mid-level managers, business units or simply employees). Their theory emphasizes the interaction between organisational structure and investment decisions. The manager's role is to select projects and recommend them to the board of directors. If the board accepts the manager's recommendation, the two agents perform tasks like production and marketing on each project. In this setting, the divisional structure corresponds to the case where each agent gets the full responsibility over a subset of projects and performs all tasks on these projects, whereas the functional structure corresponds to the case where each agent specializes in one task and performs it on all selected projects.
In Berkovitch et al, the selection of projects by the manager is subject to a moral hazard problem: the manager may have a personal preference for expensive projects, even if more profitable projects available. They show that for a given set of selected projects, the divisional structure is more efficient ex post because it enables the firm to offer each agent an incentive contract that ties his compensation directly to his performance. In contrast, under the functional structure, there is a “moral hazard in teams” problem (see Chapter 8) because each project requires the joint effort of two agents. However, the ex post inefficiency of the functional structure may render expensive projects unprofitable and hence it may deter the manager from recommending them to the board of directors. Hence, the optimal organisation structure is determined by trading-off its effect on the ex ante selection of projects and its effect on the ex post implementation of selected projects.
Berkovitch et al show that, relative to firms with the functional structure, firms with a divisional structure have less restrictive standards for project evaluation, they adopt more projects, their projects are more likely to succeed and have a higher variance of gross returns, and they pay a higher expected compensation to their agents. They also show that the functional structure is more likely to dominate the divisional structure when (i) expensive projects require a larger initial investment, (ii) conditional on success, projects yield a smaller return, and (iii) the firm's technology exhibits weak economies of scope and strong economies of scale. In addition, Berkovitch et al examine how the overall profitability of the divisional and functional structures changes when firms grow and can adopt more projects, when projects become more complex and require more tasks, and when the tasks have asymmetric effects on the probability that projects will succeed.
Berkovitch et al's main insight is that organisational structures which appear to maximize firm value ex post, may not be optimal once managerial incentives are taken into account; in many cases, it is optimal to put in place an organisational structure that appears to be ex post inefficient in order to restrict the management's ability to manipulate investment decisions in the direction it likes. The idea that a firm may wish to commit itself to an ex post inefficient structure in order to enhance ex ante efficiency emerges from a model in which the choice of projects to invest in is important. Ex post inefficiency of the functional structure could actually induce the firm's management to improve its selection of projects ex ante.
