Construction and robustness of directed-weighted financial stock networks via meso-scales

In order to investigate the evolution of stock market in any period of time, in this paper, combined with sliding window method, Granger causality test and cointegration test, a novel algorithm of constructing directed-weighted financial stock networks via meso-scale. We use the proposed algorithm on the financial stock data of the Shanghai Stock Exchange from January 1st, 2016 to December 31st, 2020 divided at the meso-scale level to construct some directed-weighted financial stock networks with fixed meso-scale and different meso-scale. Analyzing the topology and robustness of these constructed networks indicates that the robustness of financial stock networks with fixed meso-scale (assuming one year) is generally poor. By comparison, in 2018, the financial stocks were closely in contact with each other while the stock market was relatively stable. We also noted that the financial stock networks with different meso-scale show a high level of robustness when being attacked according to descending arrangement of in-degree or out-degree. They are however less robust against attacks according to descending arrangement of degree. Besides, the networks perform best for a meso-scale of 190. Hence, long-term investors are advised to consider a 190-day holding.(c) 2022 Elsevier B.V. All rights reserved.

Automated summary

In this paper, a novel algorithm of constructing directed-weighted financial stock networks via meso-scale is proposed. The topology and robustness of these constructed networks are analyzed, and it is found that networks with larger meso-scale show better stability. The results suggest that long-term investors may consider a holding period of 190 days.