基于贝叶斯模型平均技术的资产波动率预测方法研究及其应用

In empirical finance, forecasting asset volatility is one of the important topics for researchers. In recent thirty years, many volatility models such as ARCH, GARCHY, SV models have been proposed to fit the asset volatility. Now, these models have widely used in asset management, risk management, option pricing, etc. However, with the development of the volatility models, a serious consequence, model uncertainty which leads to model risk, is aroused in practice. It is well-documented that this so-called model risk has caused some huge loss in finance industry. In this project, based on the Bayesian model averaging techniques, we devote to how to reduce the model risk on forecasting the asset volatility. This project has some important applications in the fields such as, future hedging, risk management, fund investment,etc. Especially, when the future trades are implemented in China so that the Chinese financial market has become a two-side market from an one-side market, this project also has some referenced implications on monitoring the financial market for the government from macroeconomical viewpoint and making investment strategies for microeconomical investors.

在实证金融研究中, 金融资产收益波动性预测研究是金融计量研究领域的一个热点研究问题。近三十年来，研究者提出了大量的波动模型如ARCH,GARCH，SVM,用以拟合金融资产收益波动率。现在，这些波动模型已经被广泛应用于资产管理、风险管理、期权定价等金融领域中。然而，伴随着模型的发展，一个重要的后果是引入了模型不确定性，从而导致了模型风险，造成了许多重大的损失。本项目基于贝叶斯模型平均技术，研究如何减少或者分散金融资产收益波动率建模的模型不确定性风险。本项目的开展，在期货对冲、风险管理、基金投资等领域都会有很大的应用价值。特别是股指期货推出后，我国金融市场由单边走向双边，这项研究在宏观上国家管理层监管金融市场，在微观上投资者制定投资策略均有重要的参考价值。

本项目主要研究内容分为四个方面：(1) 宏观和金融潜在变量模型的潜变量模型的贝叶斯假设检验方法研究(2) 基于模型不确定性的资产波动率预测方法研究(3)基于复杂随机波动率模型的波动率持续性检验方法研究(4)具有随机波动误差的贝叶斯激励期权估计方法研究。本项目共发表了6篇国际文章，其中在国际计量经济学顶级杂志Journal of Econometrics以第一作者身份发表了两篇，在其他国际SSCI杂志上发表了四篇。在理论上，部分解决了复杂宏观和金融潜在变量模型的假设检验问题，这属于理论上的原创性贡献。在实际应用中，也解决了基于随机波动模型的波动持续性问题，在风险管理，量化对冲中有一定的应用价值。