Guilherme Lindenmeyer

This paper aims to elaborate a treated data set and apply the boosting methodology to monthly Brazilian macroeconomic variables to check their predictability. The forecasting performed here consists of using linear and nonlinear base-learners, as well as a third type of model that combines both linear and nonlinear components. The models estimate variables using historical data with lags of up to 12 periods. Results, evaluated through multiple approaches, indicate that boosting models using P-Splines as base-learners achieve the best performance, especially the two-stage boosting methodology. Robustness checks further validate these findings.

This paper tests the validity of component-wise boosting as a forecasting tool for regional economic series during recessions. Using 822 predictors, we forecast the monthly electricity consumption of the Brazilian state Rio Grande do Sul. The dataset spans 190 observations during Brazil's 2015 economic crisis. Boosting models effectively identify key predictors, including seasonal lags, national electricity consumption trends, and unemployment rates, outperforming SARIMA benchmarks. Robustness is validated using alternative exercises such as k-fold cross-validation and quantile boosting. The results highlight the utility of boosting algorithms for forecasting regional series during economic downturns.

This thesis investigates the impact of the 9-Euro Ticket, a low-cost monthly public transport ticket available in Germany from June to August 2022, on hotel pricing. The analysis focuses on whether the ticket increased competition between hotels in city centers and those located in peripheral areas. Using granular pricing and occupancy data, the study employs econometric models to capture spatial competition dynamics influenced by the ticket policy.