V. What is the mechanism?

But the above discussion, though it would seem to point us in the direction of a theory of decentralization, does not explain in rigorous terms how decentralization achieves these results. Much of the argument is a priori, counterclaims are not adequately examined before being dismissed, and the conditions necessary for decentralization to succeed are not clearly delineated. What is the mechanism by which decentralization of the public economy brings about improvements in allocative efficiency? Is local government accountability an input or an output? Are local elections the only necessary legal/institutional prerequisites, or are other conditions necessary? It is to these issues that we now turn.

We begin by examining the economic literature. The most influential economic model of the local public sector is Tiebout’s (1956) A Pure Theory of Local Expenditures. This paper imagines a world where efficient local governments compete to lure perfectly informed individuals, who move costlessly between localities and choose the one which offers their optimal bundle of public services and taxes. This revelation of preferences through migration brings about a competitive equilibrium in locational decisions which ensures that no individual can make himself better off without making someone else worse off: a Pareto-efficient result. But the mechanism which Tiebout posits - individual mobility in a context of fixed public service supply - amounts to moving voters around while holding politicians constant. This directly contradicts our experience of the world, where - save for extreme cases^[9] - it is individuals who are largely fixed and governments which, via elections, change. Even in a highly mobile country such as the United States, during any given electoral cycle the overwhelming majority of the population is geographically static, and it is their governments and politicians (local, state and national) that change. Thus we must conclude that this model is fundamentally flawed: it posits a relationship which does not by and large exist in the world, and ignores a series of others (amongst voters, between voters and governors, amongst politicians vying for power) which do and may very well be important to the question of decentralization and social welfare.

Another way in which decentralization might improve public service provision involves information. Although the economics of information is a relatively new and still-developing field, the basic argument is that information about local preferences, along with technical data concerning the production and provision of public services, is so voluminous and complex that central governments are incapable of gathering it all and processing it adequately. Thus, important information will not be reported, or will be lost in transit, or will be interpreted incorrectly at the center, or some other sort of informational obstacle will arise which prevents the center from knowing what services the periphery most needs and how best to provide them. These models’ inability to explain why this is so in a precise and convincing fashion, however, amounts to an assumption that it is true. Though admittedly more elegant, this is not far removed from the operative element of the political arguments examined above. Perhaps more importantly, it is not clear in the closing years of the 20^th century that distance poses a significant obstacle for the fluid transmission of information. In an age when CNN and MTV reach most corners of the globe, it is simply not credible that data on local preferences and other relevant conditions, once accurately gathered, somehow cannot be transmitted to the center without significant distortion or cost. Indeed, the modernization of communications systems would seem to have the opposite effect; as bandwidth increases and unit costs fall, central government becomes more viable, not less.

And yet we should not dismiss this line of thinking entirely. There is intuitively some sense in which local people have easier access to better local knowledge than non-locals. But to call this an information problem is to conflate distinct concepts. For the problem does not lie in the information per se, but rather in the agent who collects it. With proper incentives and local cooperation, a non-local agent could easily gather the relevant information necessary to provide a routine public service. And indeed, a local in the same position would require comparable conditions to succeed. But the local has obvious advantages on both counts. As a local, it is she, her family, friends and community who benefit from the efficient provision of that service. Thus she has natural incentives to perform her job accurately and honestly which her colleague does not. For similar reasons, she may find obtaining cooperation less costly. And she will have obvious, though not necessarily large, advantages of familiarity with local conditions, which it would cost an outsider to learn.

This argument is completely independent of the character, quantity or quality of the information involved, and concerns instead the performance incentives which agents face. As such, it is easily generalizable to a large number of tasks incumbent upon government agents, whether central or local, and therefore to the question of central vs. local government generally. Indeed, this insight lights the way to a deeper understanding of the efficiency implications of local government. As discussed above, where economies of scale dominate, central government will enjoy productive efficiencies in the supply of public services. Local government will enjoy advantages in allocative efficiency, however, to the degree in which local officials’ professional incentives are more in line with the interests of the local population than the incentives of central government officials. This condition will obtain where local officials are fully accountable and responsible to the local population, and where electoral representation is sufficient to ensure that all groups have a voice in local affairs. Compared to such an environment, the incentives faced by national public servants, with a much broader and probably more diverse constituency, would by definition be less propitious to satisfying the needs of a given local population.

Such an argument constitutes the kernel of a theory of institutions, incentives and accountability in the provision of local public services. It comprises the heart of a political-economy model of decentralization, local government and optimal local-service provision which I am developing elsewhere, in a larger theoretical and empirical study of decentralization. I will not present a formal, fully specified version of this model here, but will instead limit myself to saying that such a model relies explicitly on local government autonomy in a context of real local power and resources, open and transparent local democracy, good (though not necessarily perfect) information on the part of voters, and some form of countervailing power or safeguard, institutional or otherwise, to protect against abuses of power by local leaders. It can be shown with this model, and in a systematic and rigorous way, that decentralization changes the incentives which public officials face by making local government accountable to its electorate. Local officials will tend to respond to their voters’ needs more than their central government peers, in the knowledge that their jobs depend on doing so. Local government will in this way produce outcomes (i.e. public services) which are more allocatively efficient than central government. The theory also predicts the possibility, given certain conditions, that decentralized governments can be more cost-effective (i.e. more outputs for a given budget) than central government as well. We now turn to the empirics of decentralization, where we shall see that both of these predictions are confirmed by data from a recent, radical decentralization reform.