Governments face difficulty in inducing citizens to report truthfully the strength of their preferences for public goods financed from taxation, since they also benefit non-payers. Tiebout proposed that differentiating their supply by locality would reveal citizens' true preferences through their location decisions. In Discussion Paper No. 889, Research Fellow Paul Seabright notes that this tells us nothing about the level of government at which power should reside. He develops a model of regional and central governments as alternative means of motivating government to act in citizens' interests. He assumes that constitutional contracts are incomplete and governments' actions not observable, while the welfare that results from them and from random region-specific shocks is observable but not verifiable. Externalities between regions are verifiable and a constitution specifies the allocation of powers as a consequence of their magnitude.

Centralization allows benefits from policy coordination at the cost of diminished accountability, measured by the probability that a given region's welfare will determine the re-election of its government. A positive correlation between region-specific shocks strengthens the case for centralization but regional similarity per se does not, since a centralized government can accommodate differences in regions' observed characteristics.

Seabright also considers conflicts of interest within regions and governments that act as Leviathans by appropriating resources for their own use. Centralization can paradoxically increase accountability to interest groups that are under-represented at the regional level, while government failure may induce excessive taxation and expenditure but still provide an inadequate supply of public goods, which centralization further exacerbates. Other possible hypotheses about the direction of policy distortion such as government `capture' by interest groups with preferences for below-average taxation levels may yield predictions more favourable to centralization. The relative merits of centralization and decentralization therefore depend strongly on the details of particular policies, while clarification of the underlying theoretical relationships may provide a useful focus for future empirical work.

Accountability and Decentralization in Government: An Incomplete Contracts Model
Paul Seabright

Discussion Paper No. 889, January 1994 (AM)