Structure, Stability, Persistence and Entropy of Stock Networks During Financial Crises

Abstract

We investigate the network structures of stocks in SET100, NASDAQ100, and FTSE100 from 2006 to 2022, using the correlation distance and the time-space average of correlations as a threshold for connectivity of two stocks. Structure, stability, multifractality, and entropy of the networks are investigated to compare their behaviors before and after financial crises. The results show that during high volatility periods, such as the global financial crisis in 2008 and the COVID pandemic in 2020, the network characteristic path length decreases, while the clustering coefficient increases, suggesting that the network has shrunk in size, and stocks become tightly linked, similar to trends of price and return behaviors observed in many stocks during financial crises. Furthermore, the minimal level of network entropy implies that the market network stability decreases, and each sector has lost its ability to perform independently. We also find that the persistence of the network structure and the network entropy in SET increase during a period of high volatility as evident by a significant increase of the Holder exponent, while results from NASDAQ and FTSE do not exhibit such pronounced behavior, possibly due to having higher market fluctuation. Network features of SET and FTSE show recovery of same values after the 2008 crisis faster than NASDAQ, and in less than 100 trading days; however, they exhibit slower recovery, except for the network entropy, from the COVID-19 pandemic.