Modelagem do processo de análise fundamentalista de uma empresa com utilização de vetores autoregressivos

Abstract

The study aims at forecasting a company’s stock returns and prices by simulating a fundamentalist analysis process based on a Vector Autoregressive (VAR) econometric model. To achieve this, we selected relevant fundamentalist indicators and specified a VAR model capable of simulating a fundamentalist analysis of a Brazilian publicly listed company: Sadia S/A. The VAR’s statistical validation was carried out by performing unit root tests, cointegration tests, Granger causality tests, correlation analysis, impulse-response functions, and variance decomposition. Data taken from the firm’s financial statements were used to compute seven fundamentalist indicators, which were defined as the model’s endogenous variables. The exogenous variables included in the model are the Ibovespa’s return, the Brazilian GDP, the Brazilian base interest rate (SELIC), the exchange rate, and the international price of poultry and corn. The final version of the estimated model is a Vector Error Correction Model (VECM), where cointegration relationships among the endogenous variables are taken into account. Based on this model, the company’s ex-post forecasts for the fundamentalist indicators were obtained, as well as the company’s stock return and price forecasts. The results show that the econometric model is robust as it can be seen from the quality of the projections obtained. Finally, a prognostic fundamentalist analysis was carried out based on the forecasts.