(汉青经济与金融高级研究院发布于：2019-12-09 15:58:42)
汉青论坛第297期：12月11日周浩博士

本期主题：Incorporating Model Uncertainty and Model Instability in Dynamic Financial Time Series Prediction – a Probabilistic Approach

摘要：
Model instability determines how long a predictive relationship can stay in effect. Incorporating model instability is extremely important to assure the accuracy of asset pricing models in an increasingly dynamic economic environment. We propose an innovative approach to handle model instability: in particular we use a data driven approach to detect structural break/regime switching in the time series data. We then adopt Bayesian posterior probability to handle the uncertainty in selecting the models and the change points. Previous approaches in the literature are targeted to model gradual change (“Estimating the Dynamics of Mutual Fund Alphas and Betas” Mamaysky, Spiegel and Zhang， Review of Financial Studies, 21(1), 2008; “Predictive regressions with time-varying coefficients” Dangla and Halling, Journal of Financial Economics 106(1), 2012 ). Our approach is more flexible as it can model structural break/regime switching either with a step change or a gradual change totally from a data driven perspective. We demonstrate our approach’s effectiveness in bond and stock indices’ return prediction, inflation rate and industry production growth rate forecasting and hedge fund return analytics.

报告人：周浩原安邦资产管理有限责任公司宏观策略部副总经理和权益投资部高级投资经理
时间：12月11日（周三） 12：15
地点：明德主楼515教师交流室

报告人简介：
原安邦资产管理有限责任公司宏观策略部副总经理和权益投资部高级投资经理，负责安邦资产的量化投资。曾在中国国际期货， 美国道富银行和摩根大通银行从事量化投资工作。曾任美国奥本海默基金机构资产管理公司量化股票投资部总监。他在美国德雷塞尔大学获得工商管理博士学位和计算机硕士学位，哈尔滨工业大学获得管理信息系统硕士和学士学位。熟悉金融市场主要的风险管理工具。熟悉投资管理和金融市场的最新理论和实践。有丰富的研究、分析和评估投资策略的经验。有丰富的建设和管理投资组合，在严格的风险管理下达到预期投资回报目标的经验。