【Metrics Seminar】Conditionally Optimal HAR Forecasts
2025/10/14
研討會日期 : 2025-10-14
時間 : 14:30
主講人 : Professor Andrey Vasnev
地點 : Conference Room B110
主持人 : Professor Chu-An Liu
演講者簡介 : Professor Andrey Vasnev received his Ph.D. in Economics from Tilburg University in 2006. He is currently a Professor at the University of Sydney. His research interests are Forecast Combination in Business, Finance, and Economic Applications.
演講摘要 : This study advances the investigation of the heterogeneous autoregressive (HAR) model through the lens of forecast combination, following the approach of Clements and Vasnev (2024). To enhance the prediction of realized volatility, we apply the conditionally optimal framework recently developed by Gibbs and Vasnev (2024). The traditional HAR model and its univariate components often fail to adapt swiftly to large, rapid market shifts, such as those seen during the Global Financial Crisis (GFC) in 2008 and the more recent COVID-19 pandemic in 2020. By explicitly modeling the conditional bias of the univariate model forecasts with simple autoregressive structures, we extract valuable information that can be included in the conditionally optimal weights. This method yields substantial gains in predictive accuracy across the S&P 500 and Dow Jones indices, with the most pronounced benefits observed at longer forecast horizons.
