Big data analysis in financial networks: An econometric approach for the detection of SIFIs and the measurement of systemic risk = Big data elemzés pénzügyi hálózatokban: Egy ökonometriaimegközelítés a SIFI-k azonosítására és a rendszerkockázat mérésére

Absztrakt (kivonat)

The financial crisis (2007¡2009) shed light on the vulnerability of the financial system. The bankruptcy of Lehman Brothers in 2008was a severe signal, which uncovered that dense networks exist in the financial sector, which can pose a systemic risk for the stability of the whole system. Initially, the banks were identified as the primary transmitters of spillovers, but the bailout of American International Group (AIG) demonstrated that insurance companies could also be vulnerable. However, the insurance branch deserved less interest in the past decade despite its potential riskiness. The sector was handled as a homogenous industry, which radically contradicts the diversity of insurance branches. Different underwriting activities can be characterized, like life, accident and health, property and casualty, surety and financial guarantee insurance, and reinsurance. This division of traditional business models is essential, thus the various covered risks have a considerable impact on the exposure of the insurers. My research focuses on the interconnectedness of insurance companies operating in different branches and the banking sector, including three levels of aggregation: institutional, subsectoral, and sectoral. The methodology used in this work utilizes the pairwiseGranger-causality approach based on the idea of Hué et al. (2019) and extended by Song and Taamouti (2019) involving principal component analysis to filter out indirect and spurious edges from the network. This makes it possible to measure real connections among companies, sectors, and industries. The main finding of this research was that filtering out non-real connections resulted in a significant decrease regarding the number of edges in the insurance and banking industry during the financial crisis (2007¡2009). However, the financial sector demonstrated growing connectedness from 2001 to 2019. At an institutional level, some small banks were identified as SIFIs, which highlights the problem of concentrating only on big companies and can cause distortions in vulnerability rankings. At a sectoral level, the banks with high market capitalization were a dominating part of the network,while in the insurance sector, themost edgeswere detected on theNorthAmerican market; first of all, the P/C insurerswere deeply embedded in the network. Nevertheless, fast growth of interconnectedness was found in the European life insurance and reinsurance sectors.