Banca de DEFESA: Bianca Gabriel Fellet

EMPIRICAL EVIDENCE ON ASSET PRICING MODELS APPLIED TO THE BRAZILIAN STOCK MARKET IN PERIODS OF CRISIS

Asset pricing. 5-Factor Model. APT. DMA. Financial crises. stock returns. Brazilian stock market.

This research is aimed at evaluating the explaining and predicting power of two asset pricing models, already consolidated in academic literature. The models were the Arbitrage Pricing Theory of Ross (1970) and the Fama-French 5-Factor Model (2015), by meas their application in the Brazilian stock market in periods of crisis. The research sample was composed of all companies with shares traded on “Brasil, Bolsa, Balcão” (B3), during the period from 2007 to 2021. Assets of companies in the financial sector were excluded, and the sample was divided into three sub-periods, covering the 2008 world crisis, the 2016 Brazilian political crisis, and the crisis related to the Covid-19 pandemics, respectively. The explanatory power of both aforementioned models were tested using Seemingly Unrelated Regressions with Identical Regressors (SUR). The assessment of the predictive power of such pricing models were performed using the Dynamic Model Averaging (DMA) method. The results of the empirical analysis, including explanatory and predictive power and robustness tests are performed and presented. The research contributes to the understanding of asset pricing in times of crisis, as well as the identification of factors with explanatory power of stock returns in the Brazilian stock market. It is expected that the study proves to be useful to researchers, investors and fund managers when making investment decisions by identifying factors correlated with Brazilian stock returns. The results indicate that the asset pricing models APT and 5-Factor Model have good explanatory power for stock returns in all periods defined as subsamples. However, in terms of predictive power, there is still room for the development of new models and the use of new methods, as the DMA has not been able to prove itself efficient in making predictions using either the APT or the 5-factor model.