Much of the debate over the response to Hurricane Katrina has centered on the question of the division of responsibilities among the different levels of government--federal, state, and local. Concern has been expressed that for the federal government to have played a more aggressive role, for example by taking command of all response efforts and perhaps placing them in the hands of the regular army (as distinct from the National Guard, which is state rather than federal although the President is authorized to "federalize" it in situations of war, insurrection, or civil disturbance and thereby place it under federal military command), would have violated the tenets of federalism and perhaps specific provisions of the Constitution allocating powers between federal and state (including local) government, and specific statutes such as the Posse Comitatus Act of 1878, which limits military participation in law enforcement.
Issues of federalism cannot be resolved solely by reference to economic criteria. The reason is that, the Revolution having been waged by states (the former British colonies) linked in a loose confederation, the Constitution, while tightening the federation, recognizes the states as quasi-sovereign entities. Even if it would be efficient to do without states and have as centralized a government as France has, this could not be done without amending the Constitution, and indeed perhaps without replacing it with a completely new Constitution adopted in a constitutional convention.
Nevertheless, it is of course possible to analyze the economizing principles of federalism, and that is what I shall try to do in this posting.
From an economic standpoint, federalism is a scheme of decentralized governance, designed to optimize the provision of government services. In the governmental as in the private business setting, there are disadvantages to a strictly hierarchical ("U-form"--unified or unitary), as distinct from a more loosely coupled or "horizontal," method of organization ("M-form"--multidivisional). With strict hierarchy, information flows from the bottom of the enterprise up to the top echelon of management, and commands flow back down based on decisions made at the top. Inevitably, information will be filtered and otherwise lost or garbled on its way up, and as a result the top managers will perforce base their decisions on information that is frequently incomplete or inaccurate; and likewise commands will tend to be misunderstood on their way down the successive links in the chain of command. The centralizing of decisionmaking power will reduce competition, diversity, and flexibility; mistaken decisions will be more costly because they will bind the entire enterprise; and mistakes will be frequent because the top managers will not be given a full array of alternatives to choose among because their subordinates will filter out most of the alternatives on the way up in order to spare the top managers from being overwhelmed by information.
The other side of this coin, which is illustrated by the regime that preceded the Constitution--namely the Articles of Confederation, which created a very loose-knit federation of the states to conduct the Revolutionary War, and the inadequacies of which led directly to the Constitution--is that the lack of a central authority can result in suboptimal performance. Each division of the firm (or state or other regional or local government in a federal system) will tend to ignore the effects of its actions on the other divisions; each will be reluctant both to incur costs that benefit the other divisions (external benefits) or to avoid imposing costs on the others (external costs). Centralization is a way of internalizing costs and benefits throughout the enterprise by coordinating the divisions and making sure they are pulling together.
Since there are both costs and benefits to centralization, we can expect that usually the best organization will be one that has elements of both central control and divisional autonomy--one that has some hierarchy but not too much, and divisions that are only semi-autonomous. And so we observe in our federal system as a result of the provisions of the Constitution and, in particular, their (loose) interpretation by the Supreme Court. The states are allowed a considerable degree of autonomy in matters of taxation and regulation (including licensure), administration of schools and prisons, highways and other infrastructure, criminal law enforcement, etc., but Congress, the Supreme Court, and the President have a considerable override power. Congress is empowered to regulate interstate and foreign commerce; and the Supreme Court, by interpretation of the commerce power, has forbidden the states to impose tariffs and other impediments to interstate trade and travel, even if Congress fails to act. Because the states have a degree of autonomy, they function, much as the divisions of a software or pharmaceutical firm would, as laboratories for (social) experimentation. Policies invented in one state, if successful, can be copied by others. Also, people can sort themselves between states in accordance with their preferences; the right to move to a different state supplements voting power in controlling the action of government officials.
In the case of response to emergencies, one of the factors I discussed earlier--the effect of hierarchy on information flows and command responses--figures prominently, along with (depending on the scale of the emergency) externalities. The officials closest to a problem have the best information and also can act most quickly on it. We wouldn't be well served by having (only) a Federal Fire Department, so that in the event of a local two-alarm blaze the local fire chief would have to inform Washington and get permission to fight the fire. This is the point made by those who believe that the Federal Emergency Management Agency, even under competent leadership, should not be in charge of emergency responses to catastrophes.
But not all catastrophes are local. What is more, given mobility of responders, it does not make sense for every locality to invest in achieving self-sufficiency in responding to an emergency, regardless of the scale of the emergency. Suppose, as in the case of Hurricane Katrina, that the catastrophe simultaneously engulfs a large number of cities and towns in several states. Insofar as a coordinated response is optimal, and given transaction costs, which are especially high in an emergency situation, it doesn't make much sense to leave the response to state and local governments. Each state will seek to optimize its response to the damage caused it, and each locality to the damage caused that locality, disregarding the costs and benefits of its actions to the other states and localities. Moreover, as in a military situation in which one doesn't know where the enemy will attack, an effective response to an emergency requires the maintenance of reserves that can be deployed to the threatened spot, and those reserves have to be held and controlled centrally.
I conclude that while state and local government can and should be given exclusive responsibility for responding to run-of-the-mill local emergencies, the federal government should have standby responsibility for regional and (of course) national emergencies, as well as for emergencies that, as in the case of the flooding of New Orleans as a result of Katrina, wreak destruction on a scale that it would not have been efficient for the local government to prepare to meet. If you tell a city that it will receive no assistance in the event of a disaster, however great, it will overinvest in preparing to respond to disaster. Suppose for the sake of simplicity that the country has only two cities, that the cost of responding to an average disaster is 1 and to a cataclysm is 20, but that the probability that there would be two cataclysms at the same time is close to zero. Then if the federal government refuses to assist in local disasters, no matter how destructive, the two cities may incur a total cost of prevention of 40, whereas if the government invests in providing the necessary backup capability, the total cost of prevention will fall to 22 (1 + 1 + 20).
If this analysis is correct, then it was the federal government's responsibility to prepare to assist in an emergency of the scale of Hurricane Katrina. Such preparations would have been consistent with an optimal allocation of responsibilities between central and local government.