Boosting and Predictability of Macroeconomic Variables: Evidence from Brazil

Abstract This paper aims to elaborate a treated data set and apply the boosting methodology monthly Brazilian macroeconomic variables check its predictability. The forecasting performed here consists in using linear nonlinear base-learners, as well third type of model that has both components estimation history itself with lag up 12 periods. We want investigate which models for forecast horizons we have strongest performance. results obtained through different evaluation approaches point out that, on average, performance P-Splines base-learner are ones best results, especially two components: two-stage boosting. In addition, conducted an analysis subgroup available until 2022 verify validity our conclusions. also compared boosted trees other evaluated parameters cross-validation Akaike Information Criteria order robustness results.