Multiplicadores fiscais do governo central do Brasil: efeito de choque identificado via abordagem narrativa

Abstract

The study aims to estimate the fiscal multipliers of the central government of Brazil’s expenditures for different states of the economy and different monetary regimes in the period from 2000 to 2018. The study innovates with the use of an exogenous expenditures series built with the narrative approach proposed by Romer and Romer (2010), Favero and Giavazzi (2012) Ramey and Zubairy (2018). For the first stage, the models are: (i) Local Projection Method (LP), (ii) Bayesian VAR - BVAR and, (iii) Structural VAR - SVAR to estimate the multiplier effects with three different patterns for quarterly data: (i) exogenous expenditures, government expenditures and product, (ii) government expenditures and product, and (iii) exogenous expenditures and product. For a second stage, exogenous expenditure will be used as an instrument for estimating multipliers based on monthly data and in different states of the economy and different monetary regimes using instrumental variables and the Jordan method. Multipliers are calculated in form of integral, without using conversions in logarithm, following Gordon and Krenn (2010), avoiding bias in estimates. In general, the main results found are that the fiscal multipliers have estimates lower than unity. For estimates with quarterly data, with exogenous expenditures and without considering the state of the economy, the multipliers were 0.48 and 0.83 for the periods of one and two years, and greater than unity (1.14 and 1.31 ) for three and four years, with possible bias in these last results due to not considering the state of the economy. For government expenditure shocks, the multipliers reach 0.92 in the third year with the Jordà Local Projection method. Still, a downward bias is observed in the estimates with SVAR and BVAR models. For monthly data, the results show that the multipliers are not greater than unity in the different states of the economy or in periods when the interest rate is below its long-term trend. It is observed that only shocks in the exogenous variable are not able to increase the multiplier effects in the short term in different states of the economy, and may also not be considered different from zero in the estimates of the coefficients with the use of nominal government expenditures. Furthermore, there is evidence of an upward bias in the size of the multipliers when using primary expenditures in the estimates. As a rule, multipliers are higher (0.87) in periods of slack of the economy than in periods of expansion. For different monetary regimes, with low interest rates, the multiplier reached a maximum of 0.57. Finally, multipliers, in the vast majority, do not show statistical difference between the different states of the economy.