The Brazilian budgetary process is characterized by heavy central government control with wide ranging discretionary authority. State and municipal governments may try to exploit the discretionary power of the central government by choosing fiscal actions that will induce the latter to help them. This behavior might result in an inefficient budget structure in which local governments would try to act as free riders on federation resources.

This dissertation models the Brazilian budgetary process in two different frameworks. The First considers an altruistic central government that aims to maximize the sum of the state utility functions. The second maximizes the federal government's probability of being reelected. In a sequential game in which local governments play first and the central government plays afterwards, the signs of the derivatives of the federal government's discretionary resources distribution reaction functions are evaluated. The theoretical results are similar for both models in that there appears a soft budget constraint for decentralized governments.

These reaction functions are then empirically evaluated and negative correlations between state GDP and discretionary transfers, and between compulsory grants and discretionary transfers are found. The tests could not detect any significant relation either between discretionary transfers and local tax collection or between discretionary transfers and local debt service payments. In other words, no fiscal behavior exploiting a soft budget constraint was found for Brazilian states and municipalities. A positive and significant relationship was detected between discretionary transfers and the political power of states. This finding runs counter to the hypothesis of a purely benevolent federal government; it is consistent with the use of those transfers to support the re-election of the incumbent federal government. In addition, the findings indicate that the federal government has used discretionary grants to offset, at least in part, the redistributive effect of the equalization grants.