Volatility forecasting is important in portfolio and risk management because it helps portfolio and risk managers assess the potential risk and return of their investments. Accurate volatility forecasts help in setting appropriate risk limits, calculating Value-at-Risk (VaR), and managing portfolios.
Most research has focused on forecasting the point estimate or level of future volatility, with little attention given to predicting the direction of volatility, i.e., changes in volatility. Reference [1] addresses this gap by utilizing the Heterogeneous Autoregressive (HAR) model to study volatility direction changes. Specifically, the study employs three techniques to forecast volatility direction: Probit and Logit regressions, and machine learning techniques such as Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM).
To evaluate the economic performance of these prediction models, the authors implement a simple trading strategy using RV and VIX directional forecasts. They take a long position in the S&P 500 when a decline in RV/VIX is predicted and hold no position otherwise. The authors pointed out,
In terms of statistical accuracy, this study finds that classification methods consistently outperform the regression-based HAR model in forecasting volatility directions. Among these classification methods, LDA as one of the machine learning techniques consistently outperforms conventional Probit and Logit models in forecasting RV directions. For the forecasting of VIX directions, while the results demonstrate significant statistical accuracy for the classification methods, we could not find consistent evidence for the superiority of machine learning methods.
The trading performance analysis reveals that classification methods generate better trading performance compared to the regression-based HAR model. Among the machine learning methods, SVM emerges as the most consistent top performer, achieving the highest Sharpe ratios for the trading strategies based on both RV and VIX directional forecasts across all horizons. The robust and superior performance of SVM highlights its capability to capture complex data patterns, particularly in the middle percentile ranges (20%-80%) of volatility levels.
In short, trading and hedging based solely on forecasts of the direction of volatility can deliver good returns, with machine learning techniques outperforming traditional regression methods.
This is an interesting contribution to the field of portfolio and risk management. Let us know what you think in the comments below or in the discussion forum.
References
[1] Xiaodu Xie and Adam Clements, Predicting directional volatility: HAR model with machine learning integration, Applied Economics Letters, 2024
Further questions
What's your question? Ask it in the discussion forum
Have an answer to the questions below? Post it here or in the forum