Interconnectedness And Systemic Risk Of Commercial Banks In China From The Perspective Of Complex Networks

During the GFC(Global Financial Crisis)in 2007-2008,the downward risk of securitized subprime mortgages led to a worldwide catastrophe through international financial linkages and cross-border trade,while they only accounted for less than 5% of the total value of the global market.It raised the attention of both regulators and academics on systemic risk,especially the systemic risk among banks.Risk contagion and spillover are shared characteristics of all the influential system-wide events that the interconnectedness of financial institutions underlies in most cases.On 26 th November 2019,the People’s Bank of China,along with the China Banking and Insurance Regulatory Commission introduced Evaluation Criteria of Systemically Important Banks where interconnectedness is regarded as one of the first-level indexes.It shows the important role played by interconnectedness in the macroprudential supervision framework.This present study investigates different forms of banks’ interconnectedness and their effects on systemic risk,providing a better understanding of their relationships.The interconnectedness of China’s commercial banks includes interbank liabilities;the similarity of banks’ business loans;and the similarity of pledged stocks.This study constructs networks based on different forms of banks’ interconnectedness,exploring the factors contributing to the systemic risk in interbank liability networks;the reasons for banks’ similar loan structures;and the impact of holding the same pledged shares on volatility spillover effects.Moreover,the interconnectedness of different forms is related to systemic risk measures.This study quantifies the dynamic relationship between different contagion channels and the systemic risk,explaining the interaction between banks’ interconnectedness and their risk profiles,including individual risk and systemic risk.Describing the dynamics of systemic risk is of great significance to manage potential negative consequences of systemic vulnerabilities.To be specific,this study includes the following analyses:(1)Interbank exposure is one of the main channels for risk contagion especially in terms of systemic risk.However,theoretical models in this field have standardized and generalized specifications and don’t capture banks’ characteristics in reality.In this sense,this study is of practical significance as it is based on the balance sheet data of real players in the China’s financial system.Due to the lack of bilateral exposures between banks,interbank liability matrices must be estimated.The most widely used estimation methods,such as entropy maximization,give point estimators that underestimate systemic risk,while a large number of interbank liability matrices are sampled with the application of the Bayesian methodology.We also develop Gibbs samplers of three different network specifications.Besides,most of the previous literature only has listed banks or a dozen of largest banks as samples,ignoring more than one hundred smaller banks which are also indispensable for the whole banking system.We aim for shedding light on the systemic risk in China’s interbank market and for identifying susceptible or robust institutions.Based on the balance sheet data of 185 banks in China,including the “Big Six”;12 joint-equity banks;98 city banks;42 rural banks;and 27 foreign banks,we generate the default probabilities of individual banks after negative loss shocks.To the best of our knowledge,it might be the first study in the Chinese context to measure systemic risk based on the simulation of interbank liability networks.The results show that the structure of interbank liability networks can remarkably impact on the default probabilities of financial institutions.The scope of being influenced is wider if the network connection probability is in the intermediate level.In a complete network,the effect of risk sharing exceeds that of risk contagion.To conclude,the interbank exposure not only enables the institutions to share risk but also provides the channel for spreading risk.This role transformation depends on the interaction of the following factors: the characteristics of shocks,e.g.the shock size;the number of attacked banks as well as the category of attacked banks;the recovery rate of liquidation;the structure of banks’ balance sheets.(2)Most of the literature focuses on direct connections between banks,such as interbank liabilities.There are limited studies on indirect relationships.We have analyzed the sectoral similarity of banks’ business loans as a form of indirect linkages from the perspective of loan balances as well as incremental loans.With three different similarity measures,Jaccard similarity,the similarity based on Euclidean distance and cosine similarity,this study summarizes the pairwise loan similarity over time and cross-sectionally.It also explores reasons for the similar structure of commercial banks’ loans,filling the research gap in this field.Direct connections between banks might lead to domino effects in the crisis,while indirect connections could lead to popcorn effects observed as another form of risk contagion.Therefore,this study enriches the research on contagion channels,especially those of indirect linkages.We find that the sectoral similarity of banks’ business loans shows spatial effects where banks with similar operation regions have similar loan structures.The loan similarity is also related to the differences in financial indicators,such as size;profitability;liability structures;asset structures;leverage ratios;expected credit risks and loans’ sectoral concentration.All these factors can explain approximately 20% cross-sectional variation of loan similarity on average,even 33% in 2013.Among the considered factors,the differences in sizes;loans’ sectoral concentration;the ratio of loans to assets and expected credit risk significantly account for the similar business loan structure during most of the sample period.Furthermore,the loan stock similarity reached its highest in 2012,followed by a persistent decrease,and became stable after 2016;while the loan flow similarity started to increase after 2016.Compared with all the other banks,the joint-equity banks always have the highest similarity in terms of business loans’ sectoral distribution.The similarity level of the “Big Six” was relatively high but dramatically dropped since 2013 and stayed low afterwards.However,compared with the banks of the same category,the “Big Six” have the highest similarity level.(3)Pledged stocks endow financial institutions with the common exposure to the market risk on extreme market conditions.In this sense,it becomes a form of conditional indirect linkages.Most literature about stock pledge transactions studies the risk concerning firms’ valuation and operation rather than risk-taking behavior of financial institutions.This study adds conditional indirect connections based on stock pledge to the analysis framework integrating interconnectedness and systemic risk,filling the research gap.Applying the data on stock pledge transactions disclosed by listed firms,we construct networks of financial institutions according to the similarity of pledged stocks simultaneously held by different institutions,showing the dynamics of conditional indirect relationships with analytical tools in complex network theory.The changes to the interconnectedness between institutions of the same or different categories are documented,so are the time-varying relationships between the network based on shared pledged stocks and volatility spillover effects.The conclusion indicates that policy changes can significantly impact on stock pledge transactions as well as the corresponding network characteristics.Even based on the same transactions,networks and their structural features could still be quite different when we use different construction methods.Financial institutions with higher degree are more likely to put more weight on links with other institutions,especially with those showing similar degree during the time from the introduction of standardized stock pledge transactions to the implementation of more strict rules on the transaction.The financial networks based on pledged stocks in common have a significantly positive correlation with volatility spillover effects between listed financial firms under extreme market conditions.During normal times,the networks are not correlated to volatility spillover effects.Hence,only when the market value of stocks decreases and the ownership of pledged stocks could probably be transferred,can the interconnectedness shown by shared pledged stocks affect market performances of the corresponding financial institutions.(4)Since the literature usually isolates risk measures based on different contagion channels from overall systemic risk measures,this study relates the interconnectedness of banks in three different forms to systemic risks measured by SRISK and △CoVaR.Based on the principles of △CoVaR,we construct the networks of risk spillover effects to which the contributions of different interconnectedness networks are quantified in a time-varying setting.We also provide evidence on the negative externality of banks’ interconnectedness,investigating the relationship between consolidated measures of banks’ importance in the networks and bank risks,including individual risk and systemic risk.Focusing on three different forms of interconnectedness separately,we explain the mechanism in which interbank liabilities,similar loan structures,and shared pledged stocks are projected to banks’ risk profiles.The finding shows that the systemic risk is higher for banks with greater size,higher proportion of loans in assets and lower sectoral concentration of business loans.The homogeneity of banks’ business loans reduces the individual risk while it could also raise a bank’s systemic risk contribution.It is the evidence on the negative externality of loans’ similarity as a manifestation of indirect connections.The share prices are less volatile for banks that have more common pledged stocks with other institutions.However,these banks also have higher systemic risk measured by SRISK.Different forms of interconnectedness drive risk spillover effects measured by △CoVaR at different times.The significant correlation between interconnectedness and risk spillover would arise only when the risk through certain channels is considerably high.