Network-augmented time-varying parametric portfolio selection: Evidence from the Chinese stock market

Abstract Recently, the connection between assets in a portfolio has attracted widespread attention. How to improve the performance of a large portfolio selection from the perspective of network is still a challenging but meaningful work. To this end, we propose a novel network-augmented time-varying parametric portfolio selection model labeled as NA-TVPP. First, we construct a financial network using the least absolute shrinkage and selection operator-vector autoregression (LASSO–VAR) approach. Then, we extract two network topological characteristics and incorporate them into the time-varying parametric portfolio selection (TVPP) model to improve its performance. Finally, we apply it to construct a portfolio using the constituent stocks from the Shanghai Stock Exchange (SSE) 50 Index of China from 2010 to 2019. The empirical results illustrate the effectiveness of the NA-TVPP model in two aspects. To be specific, the NA-TVPP model outperforms several conventional portfolio selection models in terms of standard deviation, Sharpe ratio, and efficient frontier. Additionally, the stock network topological characteristics, such as degree centrality ( D C ) and eigenvector centrality ( E C ), are significant to portfolio selection through the negative effect on the weights.