Teams versus individual accountability: solving multitask problems through job design.

Many organizations are structured so that workers are jointly accountable for performance, even though there exist alternative organizational structures that align incentive compensation more closely with each worker's tasks. I develop a multitask agency model that demonstrates that such organizations may be optimal when multitask problems are severe or risk considerations are not too important. I also show that, in some circumstances, it may be optimal to share poorly measured tasks among several agents, contrary to the results of the existing multitask literature.

1. Introduction

* It seems straightforward that compensating workers on performance measures that reflect their own individual contributions is superior to compensating them on measures that confound the contributions of many other workers. However, firms often and increasingly choose to organize work in a way that makes individuals jointly accountable for the efforts of a larger team (Rynes and Gerhart, 2000; Shaw and Schneier, 1995; Bartol and Hagmann, 1992). While the agency theory literature has studied optimal contracts in partnerships (Legros and Matsushima, 1991) and in settings of team production (Holmstrom, 1982), it has emphasized the difficulties presented by these organizations when they arise as an exogenous result of the technology of production and monitoring. In this article, I argue that team-based organizations may arise endogenously as optimal organizational forms, even when there exist performance measures that reflect each worker's contribution alone. The joint accountability of teams increases the risk that each worker bears, but it also enriches the performance measures available to the manager, which helps to mitigate multitask problems. When the increased risk burden is not too costly or the multitask problem is severe, teams are the optimal organizational form.

The managerially oriented human resources literature (Shaw and Schneier, 1995; Bartol and Hagmann, 1992) and the academic management literature (Wageman, 1995) consistently characterize teams by two features: an allocation of tasks that requires collaboration and joint accountability for outcomes through incentive pay that reflects group performance rather than only individual contributions. Dunlop and Weil (1996) provide evidence of the importance of team-based pay in a study of U.S. apparel manufacturing. Among firms they study, they find that 98% of the firms using traditional assembly lines employ individual piece-rate compensation schemes and 0% use group-based incentives; in contrast, 80% of team-based manufacturing organizations use group incentives while only 30% use individual piece rates as part of their incentive scheme. My analysis of teams focuses on the role of such group incentives. Throughout the article, I assume that the contributions of each task to the principal's profit are additively separable, so complementarities and "cooperation" in the usual sense play no role. This focuses attention on the role of team-based compensation, and the allocation of tasks here serves only to alter the richness of the signals available for each agent.

In contrast, Hemmer (1995) and Besanko, Regibeau, and Rockett (2005) focus on the impact on optimal job design of spillovers between tasks. They show that these interrelationships can create an incentive to organize work in a way that induces cooperation by internalizing these spillovers. For Hemmer, this means organizing a manufacturing operation as production teams rather than an assembly line; for Besanko, Regibeau, and Rockett, it means organizing a multiproduct firm around product lines rather than functional areas. In both cases, as in my model, the more team-like structure provides richer and more complete performance measures by basing compensation on team output rather than individual output.

To foreshadow the model, consider the problem of organizing a sales force to sell industrial equipment to two customers. Suppose there are two tasks that determine the sales of the firm to any given customer: selling (providing information on product characteristics, etc.) and customer care (order processing, follow-up calls, etc.). Further, suppose that the only performance measures available are the sales to each customer; that is, revenues generated by each customer are well measured, but it is impossible to observe the extent to which these are influenced by the effort allocated to selling effort or to customer care. It is also not possible to accurately measure the precise costs of serving each customer, so the measure of sales may not accurately reflect the firm's benefit from the workers' efforts.

There are two natural alternative organizations. In the "individual accountability" configuration, each salesperson has full responsibility for the provision of selling effort and customer care to his or her designated customer. In this case, the firm can craft a compensation scheme in which each worker's pay reflects only his or her own effort because an individual customer's sales are observable. In the alternative "team" configuration, both workers are responsible for both customers. For simplicity, suppose that one worker is responsible for selling to both customers while the other worker is responsible for customer care for both customers. Because the workers share joint responsibility for each customer, the team-based incentive compensation scheme rewards each worker on the sales to both customers--that is, on a larger number of performance measures, each of which is determined by both workers' efforts. (1)

The advantage of the individual accountability model is that it compensates each worker on a smaller number of performance measures. If performance measures are noisy and workers are risk averse, this is desirable. Moreover, each worker fully determines his or her own pay (give or take some noise) because each performance measure depends on only one agent's efforts. This is desirable if workers are risk averse and unsure of other workers' ability or motivation. (2) The disadvantage of the individual accountability model is that the firm must induce each worker to perform two tasks with an incentive scheme that uses only one performance measure. There is no way to manipulate the relative effort given to selling and to customer care using only sales as a performance measure, which creates a multitask problem if sales are not perfectly aligned with the firm's profit. The team model alleviates this multitask problem by providing the manager a richer set of performance measures to work with in crafting an incentive scheme. The choice between teams and individual accountability therefore embodies a fundamental tradeoff between using more signals to fine tune incentives on multiple tasks and using fewer signals in order to mitigate risk exposure.

This article presents a model in which it is possible to analyze this tradeoff while taking into account the nature of the optimal linear compensation scheme under each regime. The optimal contract compensates for the weaknesses of each configuration of tasks, mitigating the multitask problem and the costs of exposure to risk. The analysis of this model demonstrates that the tradeoff persists under the optimal linear contract: teams are more attractive relative to individual accountability the more severe the multitask problem and the less important the costs of increased risk.

I also demonstrate that, when one set of tasks is accurately measured and the other is not, it is preferable to split the accurately measured tasks between the agents. This stands in stark contrast with the predictions of the existing multitask literature. Holmstrom and Milgrom (1991) argue that tasks should be grouped together to create relatively homogeneous groups in terms of the precision with which the tasks are measured. In their model, this ensures that some effort is devoted to the poorly measured tasks by giving them to an intrinsically motivated agent who is compensated with a flat salary. Because in my model there is no intrinsic motivation (no effort is expended in the absence of incentive compensation), it is less attractive to simply give up on incentive contracting with one agent, and splitting the easily measured tasks between the workers may be preferable.

Section 2 lays out a simple model of the choice between teams and individual accountability and establishes the basic results. Section 3 extends the analysis to an arbitrary number of tasks and agents. This sheds light on the generality of a common assumption in multitask models-specifically, the assumption that there is a single performance measure for each task and that each performance measure reflects a single task. Section 4 concludes.

2. A simple model

* This section lays out a simple model in which a principal delegates four tasks to two agents, each of which is compensated on at most two noisy performance measures. This model demonstrates the main point of the article, which is that the choice between teams and individual accountability embodies a tradeoff between solving multitask problems and minimizing the risk burden.