Signalling and the design of delegated management contracts for public utilities.

We analyze the shape of contracts between local governments and the contractors they hire to run public facilities on their behalf. Governments are privately informed about the quality of the facility, while risk-neutral contractors undertake a nonverifiable operating effort. The design of the contract signals the quality of the facility in such a way that the better this quality, the greater the share of operating risk kept by the government. This feature reduces the agent's marginal incentives, creating a tradeoff between signalling and moral hazard. We provide extensions of our framework in several directions, allowing for risk aversion on the agent's side, double moral hazard, and political delegation. The model is supported by some stylized facts from the water industry.

1. Introduction

* Consider a local government seeking to procure a public good or service on behalf of its citizens. On top of choosing the exact nature of the good or service provided, the government also has to decide the production mode. It may either manage the service itself under public ownership or outsource the activity to the private sector. In that case, a private company is selected following a bidding procedure. A partnership is then developed between the local government and the company that provides the service. The goal of this article is to understand the contractual forms that such delegated management may take when the quality of the infrastructure is key to the social value of the service.

Although the scope of our analysis is more general, the water industry provides a meaningful illustration of this problem. Local governments indeed have to decide whether to delegate water production to the private sector or to maintain it under their own control. Partnerships between the public and the private sectors for distributing drinking water or treating waste water differ greatly from one municipality to the next. In France, for instance, although some localities still manage production through public agencies--the so-called regies municipales--the dominating contractual form over the past century has been leasing contracts or "affermage." (1) In this case, a private entity, independent of the local government, is hired to manage the service and operate facilities. These facilities may have been built and operated in the past by the municipalities themselves. In the course of such operations, municipalities acquire private knowledge about the quality of the existing assets. (2) Within this class of leasing contracts, various kinds of arrangements can be found that differ in terms of how financial and operating risks are shared. Those contracts typically involve a lump-sum payment, a fixed regulated price for consumers, and a share of the realized profit. Depending on the share left to the contractors, the contracts may embody high-power incentives to exert effort in operating and managing assets, or instead implement much lower-power incentives.

One critical feature of the water industry is the fact that, although the quality of the infrastructure drives the price paid by consumers, the quality of the service, here water supply, is easily verifiable. Indeed, either in the United States with the Safe Drinking Water Act or in France with the 1992 Water Act, the minimal water quality standards are the same for all water providers and are quite easily checked by regulators. A second important feature of this activity, as for many public services, (3) is that rather small and poorly diversified local governments delegate to private contractors--typically, large private companies operating in several markets--the management of assets on which the local governments have a priori private information.

Starting from those specific features, in this article we present a theory explaining the shape of delegation contracts and the kinds of public/private partnerships that can emerge in practice. When a local government (hereafter, the principal) is privately informed about the quality of the infrastructure that a potential service provider (the agent) will use, the design of concession contracts acts as a signal of the facility's quality to the private sector. We analyze how this signalling issue interacts with the moral hazard problem that the local government faces when it delegates production to the private sector.

Local governments are assumed to be risk averse. In a regulatory context, this assumption has been defended by Lewis and Sappington (1995), who argue that it can be viewed as a convenient shortcut to model financial constraints, especially when those regulators represent small localities. Had the quality of the infrastructure been common knowledge, risk aversion on the principal's side would justify shifting all risk to the private sector, not only for insurance reasons but also because it provides full incentives when the agent's effort in running assets is nonobservable in a moral hazard environment.

Instead, when the local government is privately informed on asset quality, contract design results from the interaction between two major forces. On the one hand, by keeping a share of the financial and operational risk, the local government credibly conveys information on the quality of the infrastructure. On the other hand, by keeping some risk, the local government reduces the agent's incentives to exert an effort. There is thus a tradeoff between signalling and moral hazard. At a separating equilibrium of our game of contractual offers, local governments with a high-quality infrastructure are ready to pay a premium to separate themselves from those with a low-quality one.

Our model provides a number of sharp predictions. First, in the fully separating perfect Bayesian equilibrium we select, the amount of risk kept by the local government increases with the quality of the infrastructure. The private sector is made residual claimant only for the worst infrastructures. When profit results both from the intrinsic quality of the infrastructure and from the contractor's effort, profit is a U-shaped function of the infrastructure quality. Second, full privatization emerges only for the worst-quality infrastructures, whereas the best ones remain under public management. Indeed, because sometimes profits are not easily verifiable, profit-sharing schemes between the public and the private sectors cannot always be written. In this case, the only signalling tool available to the local government is whether it retains asset ownership or not. Fully separating allocations can certainly no longer occur in such environments because the crude allocation of property rights cannot account for the continuum of possible values of asset quality. A semi-separating equilibrium may nevertheless arise where public ownership and low-power incentives come with the high-quality assets, whereas private ownership and high-power incentives prevail otherwise.

The empirical evidence on the impact of ownership on performance is hard to interpret, since most of the information is rather partial and often suffers from an endogeneity bias--ownership being itself something to be explained. Nevertheless, the few studies taking into account this bias are consistent with our main prediction, i.e., the positive relationship between public ownership/low-power incentives and infrastructure quality.

Let us now review the relevant literature. Although our primary concern is on the shape of concession contracts, this article is also more broadly related to the literature on privatization and to the lively debate on the boundary between the private and the public sectors. An important benchmark is given there by the "Irrelevance Theorem" of Sappington and Stiglitz (1987). This theorem establishes that, provided contracting takes place ex ante (i.e., before any uncertainty is resolved) between a risk-neutral private firm and a government, delegating production to the private sector comes with no loss even though control of the productive assets may give an informational advantage to the operating firm. We depart from their assumptions by assuming that the principal already has private information at the time of contracting with the private sector. The principal then becomes an active player with his own incentives to manipulate information. This cost of signalling makes full privatization less attractive than in Sappington and Stiglitz (1987) and provides a fresh perspective on the make-or-buy problem. (4) Other authors, such as Hart, Shleifer, and Vishny (1997) and Hart (2003), have analyzed the impact of allocating residual rights of control a la Grossman and Hart (1986) on incentives to perform specific investments. For instance, Hart, Shleifer, and Vishny investigate the costs and benefits of private ownership in a framework where the quality of the service is nonverifiable. They highlight a tradeoff between an excessive tendency to reduce both quality and cost under private ownership and an incentive to reduce only quality under public ownership. Although it applies reasonably well for some public services (prison management, for instance), this tradeoff does not seem so relevant for other services such as water supply, waste disposal, and transportation, where the major concern is with the existing quality of the infrastructure and not with the quality of the service itself, which is observable and heavily regulated.