An Early Warning Test for the Brazilian Inflation-Targeting Regime: An Application to the COVID-19 Pandemic

We estimate in this paper a mixed causal noncausal model for Brazilian inflation year-over-year (YoY) and ask the question of whether it could serve as an early-warning system for the Brazilian Central Bank during the COVID-19 pandemic era. We focus on forecasting inflation, and the probability of staying within the bounds of the Inflation-Targeting Regime during the Covid-19 pandemic and its aftermath – namely, the sample from January 2020 to December 2022. We estimate a high probability thatBrazilian inflation will leave the tolerance bounds of the Inflation-Targeting System in March 2021, using information up to February 2021. This is one month in advance compared to the Consensus of experts in the Focus database. For point forecasts we show that the mixed causal noncausal MAR(1,1) model has a significant improvement for 1 and 3-months ahead horizons compared to the forecast of these experts. This is an interesting finding, since our model only requires the estimation of a linear model with leads and lags under non-Gaussian disturbances. Although simple to estimate, it has the important feature of being a forward-looking model.