When Does Coordination Require Centralization?

“Our intelligence community needs better coordination of operations and exchange of information, and that’s why we need an overall director of national intelligence and a national counterterrorism center.” (U.S. Senator Jim Ramstad, on the post-9/11 need for a ‘spy-czar’)

“The same folks who gave us Homeland Security … now want to give us a czar of national intelligence. … A certain amount of competition and independence is a good thing in the intelligence world.” (Joe Galloway, military consultant for the former Knight-Ridder chain of newspapers)

The question of how large, multidivisional entities should organize has a long pedigree. Often relevant to governments, armies, corporations, and nongovernmental organizations, the topic has been debated for centuries; Plato’s Republic, Sun Tzu’s Art of War, and Aesop’s fables all addressed various forms of this problem. The most significant issue in this debate is whether more efficiency and better decision making are achieved in a centralized setup (where the divisions report to headquarters, which makes the decision), or in a decentralized setup (where the divisions exchange information among themselves before making their decisions independently). Academics, practitioners, and management gurus have all vacillated on the subject, and organizations occasionally have been restructured to follow current management trends.

Strategizing by biased division managers can mean that local information remains a powerful force for decentralization.This problem is complex. On the one hand, synergies can result from a centralized coordination. On the other, the restrictions such coordination places on individual divisions can keep them from best adapting to the specific conditions they face. Further, each division’s profitability is affected by (1) how well its decisions are adapted to local operating conditions and customers—-poor adaptation, intuitively, hurts profits; and (2) how coordinated the decisions of the two divisions are—-a lack of coordination adversely affects profits due, for example, to a loss from unrealized synergies or from unexploited economies of scale. Niko Matouschek, an assistant professor in the Kellogg School’s Management and Strategy Department, and his co-authors Ricardo Alonso (University of Southern California; a graduate of the Kellogg School’s Managerial Economics and Strategy program) and Wouter Dessein (University of Chicago), use tools from game theory to bring fresh insights in examining the nature of this trade-off.

Adaptation and Coordination Matter

To analyze this problem, Alonso, Dessein, and Matouschek constructed a formal model in which they considered a company with two operating divisions. Think, for example, of a multinational corporation operating both in the United States and in Japan. Each division faces unique operating conditions. Further, each division’s profitability is affected by 1) How well its decisions are adapted to local operating conditions and customers. Poor adaptation, intuitively, hurts profits; and 2) How coordinated the decisions of the two divisions are; a lack of coordination adversely affects profits due, for example, to a loss from unrealized synergies or from unexploited economies of scale.

Each division has a manager who is informed of his or her department’s operating conditions but not those of the other. Both managers would like to maximize profits, but each values his or her own division’s profits more than those of the other. In other words, both managers are biased toward their own division. In this setting, the authors compare two possible strategies for the organization:

• Centralization: Both division managers report to a superior (vertical communication), who, after receiving their reports, makes decisions for both departments to try to maximize the profits of the firm as a whole.
• Decentralization: The two division managers share information about their local conditions with each other (horizontal communication), and then individually decide on the actions their divisions should take.

In both configurations division managers will communicate strategically. To analyze communication in each setup, the authors used a game-theory tool called “cheap talk,” which examines how strategic agents (in these cases, the division managers and headquarters) communicate with respect to their incentives. In a context in which one party needs to convince the other that the situation warrants taking a certain action but cannot objectively prove his or her claims nor commit in advance to a specific action, the messages sent by one party to the other are costless—‘talk is cheap’ (Crawford and Sobel, 1982). In the absence of external influences, the information communicated by a division manager might be vague, possibly causing the superior to make decisions that are inferior to ones that could have been made based on relevant information. The authors build on this to analyze the loss from strategic communication in their setting.

Consider the centralization scenario in which each division manager is biased toward his or her own division but also knows that the superior cares only about the sum of both divisions’ profits. In an attempt to influence decisions that would favor the department’s profits, each manager could misreport the conditions faced by his or her division; this could result in a loss of information. This strategizing hurts the quality of communication, potentially leading to decisions that may be not the best for the firm as a whole. In the decentralization scenario, similarly, both division managers might try to mislead each other regarding the local conditions.

As one would suspect, the choice between centralization and decentralization strategies depends critically on the strength of the potential synergies that could result from coordination (or on the significance of losses from lack of coordination). The choice also depends on the degree to which each division manager favors his or her own department. Alonso, Dessein, and Matouschek show that regardless of these parameters, division managers will always report more information to headquarters than to each other. This communication advantage vanishes as coordination becomes important; managers who are mindful of that importance will report nearly as much to each other as to a superior.

Coordination and Communication Correlate

The quality of vertical communication can also deteriorate as the need for coordination becomes stronger. In this case, headquarters can become focused on making decisions that do not take into account the individuality of each division. Anticipating this, each division manager might exaggerate the idiosyncrasies of his or her situation to influence headquarters’ decision. This implies, rather surprisingly, that decentralization is preferable even if coordination is critical and that centralization is only optimal when divisions are quite heterogeneous and/or when managers are strongly biased toward their own divisions.

This idea that decentralization can be important is part of the received knowledge in economics, with many researchers pointing toward the presence of local knowledge as a reason to decentralize decision-making. The thinking behind this, however, traditionally held that relaying such information to the decision-makers was difficult due to the constraints on physical communication. Matouschek and his co-authors show that in the current age, when technological advancement has driven down communication costs tremendously, strategizing by biased division managers can mean that local information remains a powerful force for decentralization.

This paper adds to the body of thought that suggests that decentralization is often the best thing a firm can do, in keeping with the idea that decisions should be made by the people with the most pertinent information. It suggests also that decentralization may be significantly more robust than was previously thought. As the authors point out in their paper, several notable practitioners such as Alfred Sloan, the longtime president and chairman of General Motors, have suggested that decentralization can be helpful even when coordination is important to a system. The authors’ research provides a rationale for this belief; the quality of communication in centralized and decentralized modes of organization is affected by the relative strengths of the managers’ biases and the gains from coordination. Therefore, even when coordination is important, decentralization can be the right thing to do.

Further Reading

Crawford, Vincent P. and Joel Sobel (1982). “Strategic Information Transmission,” Econometrica, 50(6): 1431-1451.

Sloan, Alfred P. (1964). My Years with General Motors. New York: Double Day.