to enable governments to make good commitments while preventing them from making bad ones.

I have used throughout the example of government regulation of a monopoly. Yet the same considerations apply to other forms of economic intervention. They also apply to "social" regulation, for example of health, safety, the environment, and employment (Baron, 1995). State intervention can be superior to nonintervention when governments have some information about private agents, when they have legal or fiscal instruments to regulate, and when the institutional framework allows credible commitments.

Yet none of these conditions guarantees that governments will intervene in the public interest. The very capacity of the state to intervene makes it an attractive target for influence by private interests, and the very ability to commit opens the possibility of collusion. Hence, there are reasons to expect that the quality of state intervention in the economy depends on the internal organization of the state—in particular, on the relations between politicians and bureaucrats—and on the design of the democratic institutions that determine whether citizens can control politicians.

In a democracy, the authority of the state to regulate the life of the society coercively is derived from elections. Yet many of the functions of the state and all of the services the state supplies to citizens are delegated by the elected representatives to someone else, specifically to the public bureaucracy. Delegation is inevitable. As Kiewiet and McCubbins (1991:3) observe, "desired outcomes can be achieved only by delegating authority to others."

Delegation raises the standard principal-agent problems. Since it is impossible to write legislation that would fully specify the actions of agents under all contingencies, the executive and administrative agencies are left with a significant degree of discretion.⁸ But the objectives of bureaucrats need not be the same as those of citizens or of the elected politicians who represent them. Bureaucrats may want to maximize their autonomy or the security of their employment, render clientilistic favors to friends and allies, shirk their duties in office, aggrandize their budgets, (Niskanen, 1971), or simply get rich—all at the expense of the public. Again, they have private information concerning the benefits and costs of their actions, and they undertake actions that cannot be observed directly, but only inferred from outcomes or monitored at a cost. Hence delegation must inevitably give rise to agency costs.