|
Article Excerpt
This article studies hierarchical organizations where concerns for fast execution are important and employees must be coordinated to avoid wasteful duplications of effort. Simple conditions are provided for the time spent on coordinating subordinates to be increasing and the span of control to be decreasing as one goes up the hierarchy, with equalities holding if delay is all that matters. When returns to specialization are substantial, the span of control also tends to widen and the hierarchy to flatten as urgency increases. The model suggests that concerns for first execution may be key in explaining recent trends toward decentralization and delayering in firms.
1. Introduction
One of the fundamental tasks facing management is to ensure that the activities within an organization are properly coordinated. This article contributes to the understanding of this issue by studying how concerns for fast execution affect organizational choices, in particular the allocation of coordination responsibilities across hierarchical levels.
Considerations of delay have long been recognized as key in shaping organizations. Rajan and Wulf (2006) report that Jeffrey Immelt, the CEO of General Electric, motivated the decision to shorten the chain of command at GE as follows: "The reason for doing this is simple--I want more contact with the financial services teams.... With this simplified structure, the leaders of these four businesses will interact directly with me, enabling faster decision making and execution." (1) Boeing is another case in point. According to Galbraith (1977), "after 1964 the problem facing Boeing was not to establish a market but to meet the opportunities remaining as quickly as possible.... Now a delay of a few months would result in canceled orders and fewer sales." Galbraith reports that, to respond to competitive time pressure from Douglas, Lockheed, and the British-French Concorde, Boeing was forced to drastically reduce the time devoted to product development and design. The initial product development phase, for instance, was compressed from four years on the 727 to four months on the 747. As one would expect, this exacerbated coordination problems and created information overload. Boeing's primary response to these problems was to allocate "more resources, liaison and task forces, to coordination through local relations.... The lateral relations allowed a temporary decentralization of influence. The organization was still a functionally dominant one with an integrating role between functions" (Galbraith, 1977).
We study how concerns for fast execution interact with organizational choices in a simple model of an organization that must carry out a task. The task could be to build an airplane, to develop software, or to test a large number of chemical compounds in search of new drugs. For expositional ease, however, we will often refer to the task simply as information to be processed. We distinguish between managers, whose job is to plan and coordinate the allocation of duties among their direct subordinates, and workers, who engage in production at the bottom of the hierarchy. Coordination is important to reduce "the almost inevitable duplication in information processing that accompanies the use of multiple managers" (Geanakoplos and Milgrom, 1991). Specifically, we posit that by spending time on coordination activities such as work scheduling and the exchanging of information with colleagues, managers can reduce wasteful overlapping between tasks and thus the amount of information that must be processed by the organization. Examples include the chemical department at Du Pont, which by the end of 1920 took on the responsibility of coordinating the different manufacturing departments "so that overlapping of the research programs ... may be avoided as completely as possible" (Chandler, 1990), and the coordination department at Jersey, which was responsible for advising the board so as "to avoid any needless duplication of equipment" (Chandler, 1962). (2)
Multilayer hierarchies can arise in this model for two reasons. First, when a task is delegated to multiple subordinates, a finer division of labor results, which may induce workers to specialize and thereby become more efficient. Second, delay may be reduced through parallel information processing. The main restriction placed on the hierarchy is that it must be balanced, so that all the managers at the same level must have the same number of subordinates (span of control) and spend the same time coordinating. This is essentially equivalent to assuming that tasks are divided evenly among subordinates, so that all the managers at the same level face exactly the same problem.
Production and coordination activities are assumed to be costly both in terms of wages and delay. The objective of the organization is to minimize a weighted average of the wage bill and the cost of delay. Thus, the weight attached to delay captures in a simple way the need for fast execution, or "urgency." The choice variables are the number of layers L in the organization, the time t = ([t.sub.1], ..., [t.sub.L-1]) managers spend coordinating at each layer of the hierarchy, and the span of control s = ([s.sub.1], ..., [s.sub.L-1] at each layer of the hierarchy. We stress that, as in the information- processing literature, no conflict of interest is assumed between the organization and its members, and therefore the agents will simply maximize the organization's objective.
The analysis of the model yields three main results. First, a simple log-supermodularity condition is provided which ensures that, in balanced hierarchies, coordination time is increasing and the span of control is decreasing as one travels up the hierarchy. When this condition is fulfilled, senior managers spend more time planning and coordinating than junior managers and supervise fewer (direct) subordinates than their lower-level counterparts. Loosely speaking, the log-supermodularity condition requires coordination and the span of control not to be "too complementary" (in relative terms) in reducing duplications. Indeed, if the span of control has to be larger near the bottom of the organization, then it cannot be that the incentives to coordinate increase too sharply with the number of subordinates, or otherwise junior managers would spend more time planning and communicating than their superiors.
A second result is that coordination times and spans of control will typically be equalized across layers when concerns for fast execution are paramount. Thus, if senior managers coordinate more than junior managers (as the previous result suggests), then a shift toward granting junior managers broader authority in deciding how work is organized should be expected as urgency increases. Intuitively, we show that whereas communication overloads at the top levels of the hierarchy help reduce wage costs, bottlenecks are also created, and these are very costly when decisions need to be taken quickly. Thus, as urgency increases, a more decentralized allocation of responsibilities will often emerge. (3)
A third result relates to the optimal number of layers in a hierarchy. An unappealing feature of early information-processing models is that because delegation is only driven by the need to reduce delay, the number of levels in a hierarchy tends to increase with urgency (see, e.g., Keren and Levhari, 1979). In a simplified version of the model with uniform span of control across layers, we show that if returns to specialization are substantial, that result can be reversed. In particular, sufficient conditions are provided for the span of control to widen and the hierarchy to flatten as urgency increases. Empirically, the model suggests that if the main concern of an organization is to coordinate a specialized workforce cost efficiently, relatively tall hierarchies are optimal. However, a transition toward broader and flatter organizations should be observed as competitive time pressure grows.
These results may shed some light on dispersed observations about organizations. Many commentators have argued that managerial jobs entail a continuing effort to coordinate activities within the organization, and that planning and coordinating receive the greatest emphasis within top management (see, e.g., Sayles, 1964; Mahoney, Jerdee, and Carroll, 1965; Guetzkow, 1981). Starbuck (1971) also claims that "the span of control was supposed to be smaller near the top of the management hierarchy than near the bottom, because there was greater need for coordination near the top." Galbraith (1977) provides some support for that view by showing that in the production departments of U.S. and Canadian oil refineries, the span decreases as one passes from the second level (foreman) to the third (general foreman), and then again from the third to the fourth (superintendent). The present framework is consistent with these observations, provided of course that the log-supermodularity condition holds. (4)
In recent years, a trend toward the empowerment of lower-level managers and the decentralization and flattening of the firm has also been documented. Rajan and Wulf (2006) show that in the last 20 years, U.S. corporations have experienced a reduction in the number of formal layers and an increase in the number of managers directly reporting to the top management, and find a positive correlation between delayering and empowerment. Building on that work, Guadalupe and Wulf (2008) study the effect of trade liberalization between the United States and Canada on various measures of organization design. Their results suggest that greater international competition leads to flatter and broader firms. Similarly, Acemoglu et al. (2007) document a robust positive correlation between product market competition (measured as the inverse of the Lerner index) and various measures of decentralization in three data sets of French and British firms. Mendelson (2000) uses data collected from 63 business units in the information technology (IT) industry to test the idea that the choice of organizational form depends on the dynamism ("clockspeed") of the environment. He defines a measure of organizational IQ which combines several specific variables measuring the degree to which decision rights are decentralized, the importance of incentive pay, the adoption of focus strategies, and the like. He shows that this measure is strongly positively associated with profitability and growth and, more importantly, that organizational IQ has a stronger effect in the faster-moving segments of the IT industry than in the slower-moving ones. The latter result is interpreted as evidence that the adoption of a more "organic" form of organization is driven by a fast clockspeed. Overall, therefore, the evidence seems consistent with the view that concerns for fast execution (or competitive time pressure) may be an important determinant of the recent trends toward decentralization and delayering in firms.
Related literature. This article contributes to a literature that studies the internal organization of firms from a viewpoint that abstracts from incentives. We share with the information-processing literature the view that, whereas individuals are limited in their ability to process information, organizations can help them overcome these limitations (see Van Zandt, 1998b, for an excellent survey). The seminal paper in this tradition is Radner (1993). Radner develops an explicit model of information processing based on the assumption that agents cannot process information instantaneously and shows that decentralizing operations through the use of a hierarchy is valuable to reduce delay. Bolton and Dewatripont (1994) also emphasize the benefits of hierarchical communication, but in a setting where the goal is to exploit returns to specialization (this is the second motive for delegation in the present article). Van Zandt (1998a) and Orbay (2002) extend Radner's framework to situations where new information arrives periodically, whereas Van Zandt (1999) considers scenarios where information arrives in real time. In all these models, information is hierarchically aggregated by boundedly rational agents. By contrast, the focus of this article is on how tasks should be subdivided when coordination problems are present. In particular, we allow for coordination activities (not only information processing) to take time, and study how such activities should be allocated across hierarchical levels. (5)
Keren and Levhari (1979) present a precursor to the information-processing literature which is closely related to the present work. They also study how concerns for fast execution affect the design of organizations and obtain results about the span of control that can be seen as a special case of ours. However, because there are no coordination activities in their model, they cannot highlight the role of complementarities between coordination and the span of control or study the decentralization of coordination responsibilities. Furthermore, greater urgency always leads to taller and narrower hierarchies in their model, whereas the opposite obtains here when returns to specialization are substantial. On a more technical side, this article differs from theirs (but also Keren and Levhari, 1989; Qian, 1994) as we do not rely on continuous approximations...
|
