Sigortacılık endüstrisinde oynaklık bağlanmışlığı ve sistemik risk

This thesis studies the systemic importance of the insurance industry in terms of volatility connectedness. We first analyze the volatility connectedness between the banking and insurance industry in the US and then we repeat the same analysis for the insurance companies from 26 countries. In both cases, we obtain both the static and dynamic total connectedness measures. Next, using pairwise directional connectedness measures, we analyze the revealed network structure. First, we display the financial networks, before and after some important systemic events, to understand the relative position of the insurance companies. Second, we use a community detection algorithm based on random walks to see how individual companies creates subgroups within the network. Additionally, we also report results showing how the contribution of the insurance industry to the systemic risk changes over time in the US. In our first result, the analysis of the US banking and insurance industry reveals that the total risk in the US financial system reaches to rather high levels once the insurance companies are included in the analysis. Second, we show that the insurance industry has become more connected globally after the global financial crisis. This is an important result because unlike the banking sector, insurance companies have no bilateral lending practices among each other. In the global insurance industry analysis, we show that the insurance companies are clustered on a geographical basis. In case of the analysis of US banks and insurers, we find that the so-called systemically important financial institutions tend to be clustered on the basis of their size and sector. On the other hand, unlike the large banks and insurers, small-sized banks and insurers tend to fall in the same community. Keywords: Financial connectedness, risk measurement, systemic risk, systemically important financial institutions, vector autoregression, variance decomposition, nonparametric estimation, lasso, adaptive elastic net, networks, communities