Previsão de vendas de veículos elétricos no Brasil: competição entre os métodos Box & Jenkins e amortecimento exponencial

Abstract

In recent years, the supply of electric vehicles—automobiles that use electric energy instead of fossil fuels, which are the main contributors to atmospheric pollution—has been increasing in several countries, including Brazil, especially in light of the country’s commitment to reduce CO2 emissions by half by 2030. In this context, the integration of econometrics and environmental economics contributes to anticipating the needs of both the private sector—such as planning infrastructure for new automakers and the installation of new charging stations—and the public sector, by providing technical support for fiscal incentive policies and measures to stimulate the adoption of sustainable vehicles. This study conducts a forecasting methods competition to analyze the behavior of the electric vehicle market in Brazil and identify the most efficient model for short-term projections. Two autoregressive forecasting models widely established in the literature were estimated: the Box–Jenkins Model (BJM) and the Exponential Smoothing Method (ESM). The methodological procedure included data organization and treatment, based on monthly electric vehicle sales from January 2020 to November 2025, as well as a comparative evaluation of the models using quantitative accuracy metrics (MAPE, adjusted R2, and GMRAE) and the Ljung– Box test for residual diagnostics. The results indicate that the BJM showed superior performance within the sample, whereas out-of-sample forecasting accuracy was higher for the ESM. Therefore, based on the ESM, the continuation of the upward sales trajectory is projected, with an estimated annual growth of approximately 21% for 2025 and 29% for 2026. Although the pace of adoption in Brazil remains lower than that observed in developed economies, the projections suggest a gradual and consistent expansion in the short term.