In this paper, we introduce a method to quantify the deviation of local policymaking
from expected norms based on local characteristics. We begin by computing the cosine similarity of municipal public budgets between pairs of municipalities to generate a measure of similarity in policymaking. This measure strongly correlates with differences in local characteristics, such as geographical distance, size and demographic composition, and the political traits of local officials. Next, we use a fixed-effect model of municipality pairs to predict, out-of-sample, their budget similarity based on local characteristics.

The prediction errors (residuals) measure fiscal divergence, indicating how much a municipality’s behavior deviates from expectations. We apply and test this method using data from Italian municipalities, focusing on how electoral accountability affects policy decision-making. Our findings show that fiscal divergence tends to decrease in the year preceding local elections, illustrating the impact of electoral cycles on political behavior and providing a practical application of the method.

Keywords: Fiscal Policy, Policymaking, Italy, High-Dimensional Data

JEL Classification: H72, H20, H30