Previsões sobre os lucros por ação dos bancos comerciais brasileiros de capital aberto: uma análise de acurácia

Abstract

The purpose of this study is to carry out a retrospective analysis (backtest) of different models for estimating profit growth in the banking sector, focusing on the main Brazilian banks: Banco Itaú, Bradesco, Santander and Banco do Brasil. Three distinct periods (quarterly, annual and biennial) were considered in order to evaluate the predictive capacity of the models in different time horizons. The models used were the estimation by Monte Carlo simulation, the Fundamentalist model and the Naive model, which assumes that the best estimate for the next data is the current data itself. After analyzing the results obtained and evaluating the performance of each model, statistical tests were applied, such as the modified Diebold-Mariano test and the Harvey-Leybourne-Newbold test, to evaluate the significance and statistical coverage between the models, respectively.