Title: A network centrality approach to stock returns: Sectoral interdependences, communities, and volatility transmission
Authors: Sebastiano Michele Zema - Scuola Superiore Sant'Anna di Pisa (Italy) [presenting]
Abstract: Simple sectoral influence strength measures are provided for equity returns which yield a hierarchical taxonomy between the sectors of the economy. Covering mega, large, and midcap public companies stably listed in the US during the period 2000-2018, financial companies are found to be central in the network and the main contributors to the market mode. The stocks' tendency to cluster in communities is thus explored by means of the Louvain algorithm over different periods and for different thresholds imposed on the correlation matrix, showing the topological collapse of the stock market during the crises. The identified communities follow the GICS classification partially and only out of turbulent periods. The hierarchical structure detected in the stock market at the micro level is exploited, showing how few stocks, topologically central, are statistically significant in a Granger causality framework for the realized volatility of the aggregate, while stocks in the periphery do not possess any forecast power. This few confirm the rules behavior to be robust under different network metrics, providing useful insights regarding the role played by market actors in shaping aggregate volatility.