The business cycle in Brazil: identification via heteroskedasticity

This article analyzes the Brazilian business cycle from Jan 2000 to February 2020 using a structural vector autoregression (SVAR) model. In Brazilian literature, articles aiming to obtain stylized facts using SVAR models adopt controversial identification hypotheses. The identification via heteroskedasticity emerges as an alternative, eliminating the need for such restrictions. Despite the limited sample size of Brazilian data, we exogenously select regimes with sufficient changes in the variance of the residuals over time. This allows us to identify a SVAR model via heteroskedasticity, as proposed by Brunnermeier et al. (Am Econ Rev 111(6):1845–79, 2021), using a different set of macro variables. The results from this approach are similar to those of a model identified via sign restrictions, providing support for the economic theory used by the latter to identify the shocks. The agnostic approach of identification via heteroskedasticity has enabled the identification of a new shock. We interpret it as a “pessimism” shock, particularly related to future expectations about economic activity.