Ensaios sobre o ciclo de negócios no Brasil: uma aplicação de modelos VAR identificados via heterocedasticidade

Abstract

This thesis analyzes the Brazilian business cycle and is composed of three stu- dies. In the Brazilian literature, articles aiming to obtain stylized facts using Structural Vector Autoregressive (SVAR) models adopt controversial identification hypotheses. The identification via heteroskedasticity emerges as an alternative as it eliminates the need to impose identification restrictions from economic theory. However, this approach is viable if the parameters of the SVAR model are constant, and only the variances of the SVAR residuals change over time. The first article estimates a SVAR model identified via hete- roskedasticity, as proposed by Brunnermeier et al. (2021), and finds strong evidence, for the selected variables, of regime changes in the variances of residuals, with other parame- ters remaining constant. The main results of the article are: despite the small sample size of Brazilian data, the selection of the regimes enabled the identification via heteroskedas- ticity of several macroeconomic shocks; among these, a new and relevant shock related to pessimism about the future expectations of economic activity; the article also estimates a model identified by sign restrictions, which shows impulse response functions similar to the heteroskedasticity approach. Therefore, the study shows stylized facts through a less controversial methodology, yielding results that support the economic theory used by the sign restrictions approach in Brazil. The second article aims to analyze the impact of macroeconomic shocks on the labor market by workers’ education attainment. For this purpose, a methodology is developed to identify a FAVAR model via heteroskedasticity, based on the works of Bernanke, Boivin e Eliasz (2005) and Brunnermeier et al. (2021). From a theoretical point of view, this work is the first to use this type of identifica- tion in a FAVAR. The choice of regimes follows the first article due to the similarity of the variables used in the FAVAR transition equation. Several macroeconomic shocks are identified with evidence of heterogeneity impact on the labor market since workers with higher education are the least impacted. The results support the theory of heterogeneous agents in the labor market during the business cycle. Finally, the third article aims to obtain stylized facts related to the external sector. The study analyzes the J-curve and the Marshall-Lerner condition in the trade balance of major economic categories through a SVAR identified via heteroskedasticity. Although no significant evidence of the J curve is found, there are some indications of its occurrence in the trade balance of capital goods. There are significant evidences supporting the Marshall-Lerner condition across all sectors analyzed, i.e., an positive impact on the trade balance following currency devaluations, supporting recent Brazilian literature on this subject