基于多层关联网络的投资组合优化研究  

Portfolio optimization based on multilayer connectedness networks

作　　者：王纲金[1] 吴昊钰 谢赤[1] WANG Gangjin;WU Haoyu;XIE Chi(Business School,Hunan University,Changsha 410082,China;School of Economics,Shanghai University of Finance and Economics,Shanghai 200433,China)

机构地区：[1]湖南大学工商管理学院,长沙410082 [2]上海财经大学经济学院,上海200433

出　　处：《系统工程理论与实践》2022年第4期937-957,共21页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(71871088,71971079);国家社科基金重大项目(21ZDA114);湖南省自然科学基金(21JJ20019)。

摘　　要：传统的单层网络往往难以全面刻画金融系统的复杂关联性,本文从多层关联网络视角进行投资组合的优化研究.首先构建包含线性关联层、非线性关联层、偏关联层和尾部关联层的股票资产的多层关联网络,然后分析其结构特征并检验投资组合中股票权重与关联网络中节点中心度的相关性,再后据此提出ρ投资策略和低中心度投资策略两种选股方法,最后运用滑窗法模拟动态投资过程并利用Sharpe比率、换手率、交易费用平衡点和alpha值评估投资组合的绩效.实证研究发现:1)各层网络间具有相似性和独特性,股票权重与节点中心度之间是负相关的;2)提出的网络组合回报高于基准组合,且这种高回报不被系统性风险因子暴露或交易费用所抵消;3)低中心度策略的等权重组合表现最好,能够为投资者提供显著高于基准组合的Sharpe比率.Since a single-layer network is often difficult to comprehensively measure the complex connectedness of the financial system,we study the asset portfolio optimization from the perspective of multilayer connectedness networks.First,we build multilayer connectedness networks(including linear connectedness layer,nonlinear connectedness layer,partial connectedness layer and tail connectedness layer)of sample stock assets.Then,we analyze the structure of multilayer connectedness networks and examine the correlation between the stock weights in the portfolio and node centralities in the network.On this basis,we propose two stock selection strategies based on the network models:p investment strategy and the low centrality investment strategy.Finally,we use the rolling window method to simulate the dynamic investment process,and employ Sharpe ratio,turnover rate,the breakeven transaction cost and alpha indicator to evaluate the portfolio performance.The empirical results reveal that:1)there exist similarity and uniqueness between each layers,and there is a negative correlation between stock weights and node centralities;2)the return of the proposed network-based portfolios is higher than that of the benchmark portfolio,and this high return is not offset by the large systematic risk factor exposure or transaction costs;and 3)the equal-weighted portfolio of the low centrality strategy has the best performance,providing investors with a significantly higher Sharpe ratio than the benchmark portfolio.

关 键 词：复杂网络 多层关联网络 投资组合优化 中心度 

分 类 号：F830.91[经济管理—金融学]