Empirical essays on fiscal policy in Brasil

Abstract

This thesis has three essays. The first address we use unconditional wavelet to address fiscal reaction in Brazil over the period from 2001 to 2022. We identify medium-term insolvency of net debt only in the period 2003-2004.We find antiphase (short and medium term) co-movements led by the primary balance in the subprime crisis, during the 2016-2017 fiscal crisis and in the recent pandemic. There is this same evidence with low frequency between 2013 and 2019. Mainly after the change of fiscal regime in 2017, government has issued public bonds as a solution to balance public accounts. Although the current level of debt to GDP is not insolvent, the forecast of 60% at the end of 2022 and the high annual inflation level (12.13% in April 2022) require a lot of attention. The second essay we propose a discussion on the recent situation of investment by state governments in Brazil. Our main findings are based on the estimation of a dynamic balanced panel over the period from 2015 to 2021. We find significant elasticities of the investment to Net Current Revenue (NCR) in response to its own lag, the lagged external debt to NCR , the lagged domestic debt to NCR, and the lagged cash to NCR. This finding suggests that public investments have reacted to ensure its sustainability, in response to observed changes in debt and cash. The third article we exploit dataset of the government state of Ceará to search for a systemic relationship between public capital, debt and cash.We using integration vector model, an impulse-response and wavelet, find a long-term relationship characterized by a negative (positive) reaction of investments to the increase in debt (cash). Debt and cash shocks impact after seven bimesters investments in expected directions, and such responses do not dissipate in four years. We also find anti-phasic co-movements with cycles of investments led by cycles of debt and cash in years 2014 and 2015, period of intense fiscal financial crisis in Brazilian states.