Multilayer network analysis of investor sentiment and stock returns

Abstract

We propose multilayer networks, including an investor sentiment layer and a stock return layer, to study similarities and differences between investor sentiment connectedness and stock return connectedness. Using investor sentiment scores and daily returns of 227 Chinese stocks during 2013–2020, we construct static and dynamic multilayer networks and analyze their topological characteristics at the system, firm, and sector levels. We find that (1) at the system level, the connectedness of the investor sentiment layer is weaker than that of the stock return layer; (2) at the firm level, a firm with high net-connectedness often holds large market capitalization on the investor sentiment layer, but this phenomenon is different for the stock return layer; and (3) at the sector level, the finance and real estate sector acts as a risk-emitter on two layers. Using the quantile-on-quantile approach, we find that the relationship between two layers’ connectedness displays quantile specific patterns.