An Econometric Panel-Midas Model of Asset Returns in the Brazilian Stock Market

Abstract

The paper describes the specification, estimation, and testing of an unrestricted structural econometric model design to explain and forecast individual returns of securities listed on the Brazilian stock market. The model's explanatory variables include macroeconomic, fundamental and behavioural variables which are sampled at different frequencies, since they are informed within different time periods, i.e. daily, monthly or quarterly. Accordingly, the model is specified and estimated using the MIDAS (Mixed Data Sampling) regression methodology, which supports estimation of regressions with variables sampled at different frequencies. The sample includes non-financial institutions listed in the Brazilian stock exchange from 2010 to 2016. The results indicate that the model is robust in explaining and forecasting quarterly returns of individual shares listed in that market.