Informal Authority in Organizations

We assert that decision rights in organizations are not contractible: the boss can always overturn a subordinate’s decision, so formal authority resides only at the top. Although decision rights cannot be formally delegated, they might be informally delegated through self-enforcing relational contracts. We examine the feasibility of informal authority in two informational environments. We show that different information structures produce different decisions not only because different information is brought to bear in the decision-making process, but also because different information creates different temptations to renege on relational contracts. In addition, we explore the implications of formal delegation achieved through divestitures. Informal Authority in Organizations George Baker, Robert Gibbons, and Kevin J. Murphy Authority is the defining feature of hierarchy. The boss can restrict the subordinate’s actions, overturn his decisions, and even fire him (unless the boss’s boss objects, in which case the boss herself may be fired). Tracing this chain of authority up the hierarchy, we eventually reach a person (sole proprietor) or group (shareholders) who can be thought of as owning all the decision rights in the organization. In short, formal authority resides at the top. Of course, few organizations are run by tyrants who actively exercise their ownership of all the decision rights in the organization. To the contrary, many middle managers wield substantial authority. But we assert that such authority is always informal, in the sense that it can be retracted by those higher up the hierarchy, ultimately by those at the top who hold the formal authority. That is, we see all subordinates’ decision rights as loaned, not owned. Given that formal authority resides at the top of organizations, when and how will bosses delegate informal authority to subordinates? To begin to study these questions, we develop a simple model in which a subordinate develops a project to propose to a boss. The boss has the formal authority over whether to ratify the project (i.e., over whether to allow the project to be implemented), but the boss may informally delegate this ratification authority to the subordinate. We see this ratification decision as a metaphor for a wide range of decisions that might be informally delegated, including decisions about human resources (hiring, training, job design), decisions about production (sourcing, capital and operating expenditures), decisions about competition (pricing, advertising, product design), and so on. We analyze our model in two environments: in the first, the boss has the information necessary to assess the project before it is ratified; in the second, the boss does not have this information. For certain parameters in the “informed boss” model, the boss informally delegates authority by promising to ratify all projects that the subordinate proposes, even if they are not in the boss’s (or even the firm’s) best interest. If this promise is believed, it Baker, Gibbons, and Murphy 2 November 3, 1998 Informal Authority in Organizations induces superior effort from the subordinate in the initiation stage (i.e., in searching for and developing projects). These benefits from increased effort can outweigh the expected costs of the poor projects that are sometimes ratified, in which case delegation is efficient. But the boss has the information to assess a particular project before it is ratified, and so will be tempted to renege on the promise by rejecting a project that is not in her (or the firm’s) interest; that is, delegation may not be feasible, even if it is efficient. As in other repeatedgame models, informal delegation is feasible in our informed-boss model if the boss values her reputation for delegating authority more than she would save by reneging on her promise to ratify all proposals. A different problem arises in the “uninformed boss” model. Here the boss does not have the information necessary to assess a proposed project before it is ratified, but she does observe the results from the project after it has been ratified and implemented. Therefore the boss may informally delegate authority by promising to ratify all projects that the subordinate proposes, but also threatening to retract the subordinate’s future authority if the eventual results from the current project are sufficiently poor. In this model it is the subordinate’s reputation for using his authority appropriately that is at stake, whereas in the informed-boss model it is the boss’s reputation for delegating authority that is on the line. We believe that these models capture two familiar and important features of authority in organizations. First, a boss frequently finds herself choosing not to overturn a subordinate’s bad decision because doing so would reduce the subordinate’s effort and enthusiasm in the future. Our model thus captures an important aspect of delegation: bosses often feel “regret” as they knowingly allow bad decisions to be made. Second, subordinates frequently face the opportunity to abuse their authority (for instance by using an expense account for unauthorized expenditures) but opt not to because such abuse would lead to the retraction of authority in the future. This model thus captures the tension felt by subordinates in exercising their authority “responsibly.” At a more abstract level, our two models differ in who is tempted to renege on whom. This distinction produces a key insight from our analysis: different information structures 1 The analogies to parenting and childhood, although not lost on us, are beyond the scope of this paper. Baker, Gibbons, and Murphy 3 November 3, 1998 Informal Authority in Organizations produce different decisions not only because different information is brought to bear in the decision-making process but also because different information creates different reneging temptations. As an example of this idea, we show that costlessly moving the boss from no information to full information can make the parties worse off, because it changes who is tempted to renege and how sorely. Thus, our analysis begins to illuminate not only when informal delegation should be observed (e.g., when the relevant party is patient, as in other repeated-game models) but also how informal delegation should be achieved (e.g., in tandem with what information structure). Our modeling of informal delegation draws on two existing arguments concerning formal delegation. The first is due to Fama and Jensen (1983), who decompose decision rights into a four-step process: initiation, ratification, implementation, and monitoring. Fama and Jensen argue that organizations in which decision rights are held by non-owners will always separate the first and third of these steps (the decision management rights) from the second and fourth (the decision control rights). Our model assumes a simplified version of this separation, focusing on only the first two steps: the subordinate searches for and develops a project proposal (initiation) but the boss retains the formal authority over whether the project will be implemented (ratification). We ignore the latter two steps (the implementation of the project once it is ratified and monitoring of its results), which are the focus of the classic agency model. Thus, Fama and Jensen analyze a richer environment (one that concatenates our model and the classic agency model), but they argue that decision management will always be separated from decision control. We accept their assertion that decision management will be formally separated from decision control, but ask when and how the boss can informally delegate her control rights (ratification) to induce the subordinate to make better use of his management rights (initiation). We also draw on Aghion and Tirole (1997), who examine the distinction between formal and real authority and ask how delegation affects incentives for information gathering by the subordinate. Unlike Fama and Jensen, Aghion and Tirole argue that bosses may choose formal delegation in order to give subordinates stronger incentives. That is, by giving the subordinate formal authority over both initiation and ratification, the boss may greatly improve the subordinate’s incentive to search for and develop projects, and these benefits can Baker, Gibbons, and Murphy 4 November 3, 1998 Informal Authority in Organizations outweigh the costs of the poor projects that are sometimes implemented. As noted above, informal delegation also has this effect in our model. But the fact that informal delegation can be retracted in our models leads to different predictions for when delegation will be feasible. We elaborate on such differences below. In sum, we see this paper as a simple but realistic reconciliation of the contradictory theoretical arguments by Fama-Jensen and Aghion-Tirole concerning delegation of authority in organizations. Our assumption that the boss always has the formal authority to ratify projects matches not only Fama and Jensen’s theoretical argument but also countless descriptions of real organizations. But our focus on the boss’s delegation of ratification authority to the subordinate matches not only Aghion and Tirole’s theoretical argument but yet more descriptions of real organizations: middle managers often have substantial authority. We reconcile the two by emphasizing that delegation in organizations is informal. Although we emphasize throughout the paper that subordinates’ authority is informal, we conclude by briefly sketching a third model in which we reconsider formal delegation. Following Grossman and Hart (1986), we argue that there is a way to achieve formal delegation, but it occurs between rather than within organizations: for a boss to convey formal authority to a subordinate, she must make the subordinate the owner of the assets that the subordinate would otherwise have merely managed. To be concrete, we envision an R&D lab that generates new products. If the parent company owns the lab then the parent has formal authority over which products come to market; if the lab is divested then it has that formal authority. In either case, however, there can be a useful relationship (formally, a repeated-game equilibrium) between the parent and the lab. Analogous to our related work on internal versus external supply relationships (Baker, Gibbons, and Murphy, 1997), formal authority is allocated through asset ownership, but informal authority can still be delegated in a repeated game, and the allocation of formal authority affects the set of equilibria in the repeated game. We hope to pursue a model along these lines in future work. Baker, Gibbons, and Murphy 5 November 3, 1998 Informal Authority in Organizations II. The Economic Environment We study an infinitely repeated game involving a boss and a subordinate. Both players are risk-neutral and live forever (in the usual repeated-game sense). Each period, the subordinate investigates a potential project and makes a recommendation to the boss, who either accepts or rejects the recommended project. Potential projects offer noncontractible benefits of X and Y to the subordinate and the boss, respectively. The subordinate discovers a project’s payoffs by investigating the project. For simplicity, we assume that the benefits take only two values, positive or negative: XH > 0 > XL and YH > 0 > YL. Thus, it is in the subordinate’s private interest to recommend projects yielding XH and ignore projects yielding XL, and it is in the boss’s private interest to accept projects yielding YH and reject projects yielding YL. The subordinate can take actions that affect the probability of discovering a project he likes, such as searching across more potential projects and considering alternative ways to implement a given project. We define the subordinate’s “search intensity” as the probability of discovering a good project, a ≡ Prob(XH). The conditional probability that the boss’s payoff is YH when the subordinate’s payoff is XH is p ≡ Prob(YH|XH); the conditional probability that the boss’s payoff is YH when the subordinate’s payoff is XL is q ≡ Prob(YH|XL). Thus, given the subordinate’s search intensity, the probability of each (X,Y) state is: (1) Prob(XH,YH) = ap, Prob(XL,YH) = (1-a)q, Prob(XH,YL) = a(1-p), Prob(XL,YL) = (1-a)(1-q). We assume the following timing each period. First, the boss chooses an amount s to pay the subordinate (where s < 0 means the subordinate pays the boss), so that the subordinate’s utility equals or exceeds his reservation utility. Second, the subordinate chooses an unobservable search intensity at cost c(a) = γa, thereby discovering a project. The subordinate observes the project’s payoffs (X,Y). If the project yields XL, the subordinate ignores the project, and its existence is not disclosed to the boss. If the project 2 That is, the event X=XL can be interpreted as the subordinate not discovering a project. Alternatively, we could assume that the boss knows that a project was discovered and rejected by the subordinate, but that its Baker, Gibbons, and Murphy 6 November 3, 1998 Informal Authority in Organizations yields XH, the subordinate recommends the project to the boss, who in turn decides to either accept or reject the recommendation. In our informed-boss model, this recommendation reveals the boss’s payoff to her, YH or YL. In our uninformed-boss model, however, the boss lacks the prior information necessary to discern her payoff from the subordinate’s recommendation. To pave the way for our repeated-game analysis of informal authority, we begin with two static analyses as benchmarks. First, we consider the static version of our informed-boss model, which we label “informed centralization” (or simply “centralization” where there is no risk of confusion). Second, even though the major premise of our repeated-game analysis is that decision rights in organizations must be loaned rather than owned, we consider a static model in which the boss can transfer formal authority to the subordinate; we label this second benchmark model “contractible delegation” (or simply “delegation”). INFORMED CENTRALIZATION In the static case of our informed-boss model, the subordinate would like the boss to ratify projects yielding either (XH,YH) or (XH,YL), but the subordinate will rationally anticipate that the boss will reject projects yielding YL < 0. Because only projects yielding YH > 0 will be accepted, the subordinate will choose the search intensity that maximizes the expected utility (2) MAX a s + apXH – c(a). The subordinate’s optimal search intensity under centralization therefore solves: