贝叶斯参数化投资组合策略
Bayesian Parametric Portfolio Policies
AI总结：
本文提出贝叶斯参数化投资组合策略，通过引入先验分布修正传统策略的政策风险问题，实验证明其在风险厌恶和危机时期具有显著优势。
摘要：
参数化投资组合策略（PPP）通过最大化期望效用直接将最优投资组合权重估计为可观测信号的函数，从而避免了对资产收益率和协方差的建模需求。然而，PPP忽略了政策风险。我们证明，这种忽略是有影响的，导致期望效用被高估，投资组合风险被低估。我们开发了贝叶斯参数化投资组合策略（BPPP），通过在政策系数上设置先验分布来修正决策规则。我们推导出一个一般结果，表明PPP与BPPP之间的效用差距严格为正，并且与后验参数不确定性和信号强度成正比。在均值-方差近似下，这种修正表现为投资组合方差中的额外估计风险项，意味着PPP在信号最强和风险厌恶程度高时过度暴露。实证研究表明，在具有242个信号和六个因素的高维设置中（1973-2023年），BPPP实现了更高的夏普比率、显著更低的换手率、更大的投资者福利和更低的尾部风险，其优势随着风险厌恶程度的增加而单调增加，并在危机时期最为显著。
英文摘要：
Parametric portfolio policies map predictors into portfolio weights without modelling returns, making them a natural tool for factor timing. With many predictors, if the policy has more coefficients than observations the data cannot pin it down. We develop Bayesian Parametric Portfolio Policies (BPPP), which discipline the policy with a prior, update it with realised utility, and average the weights over the posterior. We contribute to the factor-timing literature by estimating timing as a portfolio rule rather than from return forecasts. With 242 predictors and five factors, BPPP earns 2.08 percentage points a year from timing, mostly from aggregate anomaly returns. Under a leverage limit, averaging sizes each position by its uncertainty and comes close, without a separate scale choice, to the size that maximises timing with hindsight. A turnover penalty preserves significant gross timing at lower turnover.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
