Non-Linear Volatility Modeling of Economic and Financial Time Series Using High Frequency Data

The current work undertakes an overview of the forecasting volatility with high frequency data topic, attempting to answer to the fundamental latency problem of return volatility. It surveys the most relevant aspects of the volatility topic, suggesting advantages and disadvantages of each alternative in modeling. It reviews the concept of realized volatility and explains why forecasting of volatility is more effective when the model contains a measure of intraday data. A discrete and a continuous time model are defined. Sampling methods at different frequencies are reviewed, and the impact of microstructure noise is considered. Details on procedures employed in the literature with respect to modeling and forecasting using realized models are discussed, while an empirical exercise will prove the advantages of using measures of high frequency data.