In many countries fiscal decentralization characterizes the relationship among different levels of government. In those countries, local authorities have the prerogative to tax their constituencies. However, fiscal decentralization is seldom balanced in terms of tax and expenditure assignments. In order to equalize tax capacities, to internalize spillovers or to achieve national policy objectives, central governments often provide transfers to lower levels of government. These transfers may affect the incentives to manage or to improve fiscal performance. Specifically, according to Litvack, Ahmad and Bird (1998), such transfers may induce low `tax effort' in the regions. The purpose of this paper is to investigate theoretically and empirically this relationship between intergovernmental transfers and local tax effort. An initial problem to deal with is the definition of `tax effort' in itself. First, one can associate tax effort to high tax rates. Smart (1998) asserted that such association is inadequate. Second, one can measure tax effort using actual tax revenues or the difference between actual the predicted value of tax revenues. This approach has been mainly adopted by the empirical literature on the relationship between intergovernmental transfers and local tax effort [Baretti, Huber and Lichtblau (2000), Von Hagen and Hepp (2000), Jha, Mohanty, Chattergee and Chitkara (1999), Sagbas (2001)]. Although tax revenue is an accurate and observable variable, still one can hardly say that it is a good estimate of tax effort. The reason is for a given region in a given time period tax revenue is affected by many potential variables outside the control of local governments (like idiosyncratic shocks to some specific tax bases) which are seldom well controlled for in estimates of tax capacity. In practice local tax effort encompasses a broad set of actions. One of them is clearly the battle against tax evasion. In spite of its importance, this problem has been only recently addressed by the local public finance literature. Bordignon, Manasse and Tabellini (1996), presented a model where a local government exerts costless effort to catch tax evader workers and they showed how intergovernmental transfers affect tax enforcement. The drawback of this model is that, in reality, tax enforcement is not costless and the cost depends upon other variables chosen by local authorities, like the efficiency of the local tax administration. Although Prud'homme (1995) and Tanzi (1996) have informally signaled the possible inefficiencies of the local tax administrations, this feature has not been raised by the theoretical or the empirical literature. The purpose of this paper is precisely to incorporate such dimension in the assessment of the relationship between intergovernmental transfers and local tax effort. The theoretical framework assumes that in each region there is one representative habitant and a local government. The habitant posses a low or a high-valued property. The local government maximizes tax revenues. In a first period, the local government invests resources to improve the efficiency of the tax administration or to lobby the central government in order to obtain discretionary transfers. This decision is affected by the political cost of reforming the tax administration and on the ability of the local government to negotiate with the central government. Thus, in our model, intergovernmental transfers are endogenous and simultaneously determined with the reform of the local tax system. In a second period, the local government sets the property tax schedule. But, as the local government is unable to observe the value of the property, it has to rely on the habitant announcing this value. Finally, in the third period, the local government decides to enforce the tax law by randomly auditing such announcement. If the habitant is discovered having misreported, the local government sets the corresponding property tax and imposes a penalty. We assume that audit is perfect but costly; the cost depending on the efficiency of the local tax administration. We solve the model backwards. As the local government cannot commit to the auditing probability when it designs its tax policy, the equilibrium of the audit-report game is in mixed strategies, with auditing and tax evasion. Then we find the optimal tax schedule. In order to reduce the stake for tax evasion, the local government distorts downwardly the high-valued property tax. Finally, we solve for the decision of the local government regarding how much resources to invest for improving the efficiency of the tax administration. We find that this decision is negatively associated with the domestic political costs and positively with the ability to negotiate with the Federal Government. The predictions of the model are empirically tested using data for Argentina. The theory suggests a two-step approach. In a first stage we run a probit estimation where the probability of a certain province to reform its tax system (or receiving discretionary transfers) in a given year will be correlated with domestic political variables (e.g. divided government) and also with variables describing its bargaining power vis a vis the federal authorities (e.g. political representation at the National Congress, political party of the President vis a vis that of the Governor). In a second stage, we include this exogenous instrument of tax reform in a regression where the evolution of actual provincial tax receipts are regressed against this variable plus other controls like population, density, provincial income distribution and production structure. Notice that this two stage empirical approach allow us to deal with a frequent problem encountered in the empirical literature given by the endogeneity bias affecting some of the variables of interest, like federal transfers (e.g. Jha, Mohanty, Chattergee and Chitkara (1999), Sagbas (2001)).
