Abstract:
This study aims to explore the interconnectedness of insurance companies through a spillover network analysis.
The methodology follows a series of operational steps. Initially, historical data of insurance companies are downloaded. Subsequently, the daily volatility of insurance company share prices is calculated; the dataset is then divided into quarterly data frames to analyse the evolution of the relationships over time.
A Bayesian Vector Autoregression (BVAR) is estimated for each quarterly data frame using daily volatility as the endogenous variable. Next, the Generalised Forecast Error Variance Decomposition (GFEVD) is calculated to decompose the variance of the endogenous variables and identify critical drivers of network dynamics. Using the output of the GFEVD, a weighted direct adjacency matrix is composed, representing the relationships and the weight of connections between insurance companies. Finally, a network graph is created using the adjacency matrix, providing a clear and intuitive visualisation of the network structure in the industry.
