A Novel Cluster HAR-Type Model for Forecasting Realized Volatility

Abstract

This paper proposes a cluster HAR-type model that adopts the hierarchical clustering technique to form the cascade of heterogeneous volatility components. In contrast to the conventional HAR-type models, the proposed cluster models are based on the relevant lagged volatilities selected by the cluster group Lasso. Our simulation evidence suggests that the cluster group Lasso dominates other alternatives in terms of variable screening and that the cluster HAR serves as the top performer in forecasting the future realized volatility. The forecasting superiority of the cluster models are also demonstrated in an empirical application where the highest forecasting accuracy tends to be achieved by separating the jumps from the continuous sample path volatility process.

Forecast Evaluation in Large Cross-Sections of Realized Volatility

“…More specifically, the authors focus on the identification of rare sparse predictors within particular intervals in order to examine the predictive accuracy of a Cross-Section of lagged returns for rolling one-minute-ahead return forecasts using the Lasso shrinkage methodology. Further applications similar to our study include the paper of Yao et al (2019) in which the authors introduce a group cluster Lasso specification for the Augmented HAR model.…”

Forecasting Bitcoin volatility: A new insight from the threshold regression model

“…Here, we extend our statistical results to longer forecasting horizons. Following related studies on volatility forecasting (see, e.g., Demirer et al, 2020; Liu et al, 2018; Shen et al, 2020; Yang et al, 2019; Yao et al, 2019), we consider out‐of‐sample predictive horizons of 5, 10, and 22 days.…”

A moving average heterogeneous autoregressive model for forecasting the realized volatility of the US stock market: Evidence from over a century of data

“…Due to the significant benefit of accurate forecast of realized assets volatility, a number of studies have been conducted on realized volatility in recent time. These include studies on realized volatility of US stock market such as; Bauer and Vorkink (2011), Chen et al (2019), Gong and Liu (2019), Gupta et al (2018), Wu and Wang (2019) and Yao et al (2019). Other studies such as Mei et al (2017), Ping and Li (2018), Peng et al (2018) and Zou et al (2019) have investigated realized volatility of Chinese stock market, while Degiannakis (2017;, Sharma and Vipur (2016) investigated realized volatility of stock markets of selected multiple countries, while Salisu and Ogbonna (2018) investigated the importance of time variation in the stochastic volatility component of G7 countries.…”

“…As evident from the Monte Carlo simulation conducted by Motegi et al (2019), MAT-HAR was found to have higher forecast accuracy than the HAR baseline model. Earlier studies on the realized volatility of US stock market such as Bauer and Vorkink (2011), Chen et al (2019), Gong and Liu (2019), Gupta et al (2018), Wu and Wang (2019) and Yao et al (2019) have not considered moving average threshold effect as a modification to the conventional HAR model. While recent studies such as Gupta et al (2018) and Wu and Wang (2019) appear to account for time variation in the predictive model, they assume unknown threshold; which complicates both estimation and inferences (Motegi et al, 2019).…”