Não causalidade e não fundamentalidade em VARs fiscais para o Brasil e para os EUA

Abstract

The VAR/SVAR(Vector Autoregressive and Structural Vector Autoregressive) models are the cornerstone of the contemporaneous empirical macroeconomic research, in particular for measuring the impact of fiscal policy shocks. They may be employed as atheoretical models, as well as a mean to support the estimation and to testthe structural form of DSGE (Dynamic Stochastic General Equilibrium) models – the main theoretical tool for modern macroeconomics. Nevertheless, VAR models may be subject to pathologies, such as thenon-causality and the non-fundamentalness. They are capable of biasing the estimates in any direction or intensity. The former is related tothe existence of explosive roots in the autoregressive polynomials from stationary processesand the both refer to the insufficiency of the econometrician’s data to estimate the model’s correct parameters. This study is the first toemploy the most efficient tests– the Sahneh (2015) test for non-causality and the Forni and Gambetti’s (2014) and Canova and Sahneh (2018)’s tests for non-fundamentalness –in order to testfor these pathologies in Brazilian typical fiscal VAR model using contemporary data, as well as for US’ data, specifically the dataset used by Blanchard and Perotti (2002).That articlewas criticized by Leeper et al. (2011, 2013), as the model was found to be nonfundamental. Both models were found to be non-fundamental. Therefore, the results corroborate the critic from Leeper et al. (2011, 2013) to Blanchard and Perotti’s (2002) results.However, the solution for non-fundamentalness proposed by Leeper et al. (2011) failed to achieve the desired effect, and it was rejected as a solution for non-fundamentalness.