Based on studies developed over recent years about the use of high-frequency data for estimating volatility, this article implements the Heterogeneous Autoregressive (HAR) model developed by Andersen, Bollerslev, and Diebold (2007) and Corsi (2009), and the Component (2-Comp)model developed by Maheu and McCurdy (2007) and compare them with the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) family models in order to estimate volatility and returns. During the period analyzed, the models using intraday data obtained better returns forecasts of the assets assessed, both in and out-of-sample, thus confirming these models possess important information for a variety of economic agents.
Realized volatility; Volatility estimation; Intraday return; HAR