This article reports on tests of rational political business cycle (PBC) models using an extensive new data set covering all Portuguese mainland municipalities. With a panel of observations for budget balances and expenditure items over the 1979-2000 periods, it is possible to examine the fiscal choices of local governments over a number of electoral cycles.
Thus, we can check whether incumbent politicians increase municipal spending in pre-election periods. Although most previous studies of political business cycles have employed macro-level time series data, the examination of municipal government decisions is motivated by rational political business cycles of the Rogoff and Sibert (1988) type.
In those models, incumbents increase spending relative to taxes in pre-election periods as a signal of competence in the presence of imperfect information. This signaling motivation should apply to incumbent politicians at local, as well as national levels. Using this data, we can also investigate if expenditure choices are affected by the timing of national elections and if the opportunistic cycle in spending is influenced by the mayor's ideology, the support she enjoys at the municipal assembly, and by her decision to run for another term in office.
Use of data for Portuguese municipalities1 is motivated by the fact that they constitute a very good laboratory to test for the existence of rational political business cycles. First, data on public expenditures are very detailed, allowing for tests of PBC on particular expenditure categories. Second, the institutional structure of local governments and the policy instruments available are the same for all localities, making this panel preferable to one composed of several countries, or states, with different institutions and policy instruments.
Third, election dates are fixed and defined exogenously from the perspective of the local authorities, and all municipalities have elections in the same day. Finally, because the data set is large (with a maximum of 278 cross-sections and 22 years of observations) , inferences are likely to be more revealing that those obtained with smaller panels of countries and/or states.
