基于两步波动率的量化投资策略

Quantitative Investment Strategy Based on Two-Step Volatility

摘要: 本次研究从波动率具有长短期成分的经济学直觉出发，将市场波动视为“稳态–偏离–回归”的动态过程，采用长期视角对波动率趋势进行常态化分析，并通过短期视角捕捉波动率对稳态的暂时性偏离，进而为投资者提供详细的决策指引。在方法上，本研究与Component GARCH、HAR-RV等多成分波动率模型在思想上有共通之处，但更强调通过EWMA与GARCH的分立结构生成连续交易信号，服务于策略执行而非仅提升预测精度。具体而言，EWMA模型用于拟合波动率变化趋势，作为判断市场波动处于“均衡水平”的动态标准；GARCH模型用于滚动样本外预测，精准获得短期波动状态。通过计算短期预测与长期基准的相对偏离度，构建连续的波动率状态识别信号。以沪深300ETF期权为样本，设置清晰交易机制并选择两个基准策略进行对比，实证结果表明，本文基于两步波动率构建的策略在累计收益、夏普比率、最大回撤等核心绩效指标上均呈现较强优越性。

Abstract: This study begins with the economic intuition that volatility consists of long- and short-term components, viewing market volatility as a dynamic process of “steady state-deviation-reversion.” It adopts a long-term perspective for routine analysis of volatility trends while capturing temporary deviations from the steady state through a short-term lens, thereby providing investors with detailed decision-making guidance. Methodologically, while this study shares conceptual common ground with multi-component volatility models such as Component GARCH and HAR-RV, it places greater emphasis on generating continuous trading signals through a decoupled structure combining EWMA and GARCH, serving strategy execution rather than merely improving predictive accuracy. Specifically, the EWMA model is used to fit the trend in volatility, serving as a dynamic benchmark for determining whether market volatility is at an “equilibrium level”; the GARCH model is applied for rolling out-of-sample forecasts to precisely capture short-term volatility dynamics. By calculating the relative deviation between short-term forecasts and the long-term benchmark, a continuous signal for identifying volatility states is constructed. Using CSI 300ETF options as the sample, a clear trading mechanism is established, and two representative benchmark strategies are selected for comparison. Empirical results demonstrate that the strategy constructed on the basis of the two-step volatility approach exhibits strong superiority across key performance metrics such as cumulative return, Sharpe ratio, and maximum drawdown.