The Cross-region And Cross-market Risk Spillovers—Based On Complex Network From The Time-frequency Perspective

In recent years,the rise of global trade protectionism,uncertainty,and economic downturn have triggered sharp fluctuations in the international financial market.The spread of COVID-19 has further exacerbated the frequent turbulence in financial markets.These successive "black swan" events have led to the continuous transmission,superposition,evolution and upgrading of financial risks.Besides,the cross-region and cross-market resonance plummeting cases are common.Against this background,General Secretary Xi Jinping has repeatedly stressed the need to raise risk awareness,guard against the big probability of "grey rhinoceros" event and the small probability of"black swan" event,strictly prevent input risks of cross-market,cross-field and cross-border,and firmly guard against the bottom line of systemic risk.However,the domestic research on cross-region and cross-market of financial risk contagion is still in its infancy,and there is a lack of monitoring model for cross contagion of financial risks.Therefore,this paper takes the cross-region and cross-market risk contagion as the key monitoring object to guard against systemic financial risk and maintain national financial security,hoping to improve the management ability of countries,especially China.Taking the financial market risks of the group of seven and the BRICs countries as the research object,this paper focuses on the important idea of"fighting a tough battle to prevent and resolve major financial risks" in the report of the 19th National Congress of the Communist Party of China from the perspective of "strictly preventing input risks of cross-market,cross-field and cross-border",studies the four issues of "establish a model to monitor single market risk,identify risk spillovers between multiple markets,prevent crossregion and cross-market contagion,and predict economic downturn and financial crisis",and strive to improve the current situation that the existing research on cross-region and cross-market risk contagion is not in-depth and comprehensive,so as to provide an important reference basis for improving the risk early warning and prevention system of "early identification,early warning,early detection and early disposal".The main work and innovative achievements are as follows:First of all,financial risk monitoring and analysis in the single market.This paper takes the stock and foreign exchange markets of the group of seven and the BRICs countries as the research object,combines machine learning with volatility method,and further constructs GARCH-GANs-VAR model to measure time-varying risk.Compared with traditional VaR measurement methods,the unique feature of this chapter is to introduce Generative Adversarial Networks(GANs)to estimate the distribution,which can avoid the deviation caused by artificially setting parameters,and further improve the accuracy of risk measurement.The results show that GARCH-GANs-VAR model can accurately measure the financial risks of different countries and markets,identify the extreme risks brought by major political,economic,financial and health events,and capture abnormal signals.Besides,kupiec method is used to test the accuracy of GARCH-GANs-VAR model.By comparing with historical simulation method,Monte Carlo simulation method and GARCH-EVT-VAR model,it is found that GARCH-GANs-VAR model has advantages and good applicability,which can better reflect the time-varying risk of stock and foreign exchange markets.Secondly,financial risk spillover analysis among mlutiple markets.we extend the research of frequency connectedness proposed by Barunik and Krehlík(2018)to develop multiscale risk spillovers on the basis of LASSOVAR model,and further analyze the risk spillover from the time and frequency domains,static and dynamic perspectives respectively.The unique feature of this chapter is that the model can effectively deal with high-dimensional problems by introduce LASSO method.The static analysis results show that risk spillovers vary significantly across frequency domain,and the proportion of short-term risk spillovers is the largest.Overall,the stock markets of the United States,France,Germany and the United Kingdom play the role of net risk spillovers.The BRICs countries mainly act as risk recipients and bear the shock and impact from other countries.China’s stock market is primary affected by the UK stock market,and the foreign exchange market is mainly affected from the US foreign exchange market.The dynamic analysis results demonstrate that the total risk spillover shows a "W" shape,which is mainly driven by short-term shock.As for short-term,medium-term and long-term risk spillovers,they contain rich information,which help to deeply understand how financial crisis events affect risk spillover in different frequency domains.Thirdly,cross-region and cross-market risk spillover network analysis.This chapter attempts to build a time-frequency and double-layer risk spillover networks to studies the cross-region and cross-market spillover effects and spillover channels from the perspective of multi-level and multi-scale by using the complex network theory.The unique feature of this chapter is to use threedimensional network technology to depict the double-layer network,analyze the cross-region and cross-market risk spillover effects of financial risk,which is more intuitive and visual.The results show that financial risk has significant spatial dependence and cross contagion characteristics.In the cross-market risk spillover relationships,the risk spillovers among stock markets are interconnected and close,even occupy the dominant position of risk spillover.The spillover direction is mainly from the stock market to the foreign exchange market.In the cross-region risk spillover relationships,risk contagion among G7 countries is dominant,and the spillover direction is from the group of seven to the BRICs.Most G7 countries act as risk spillovers,while most developing countries are risk absorbers.As for risk spillover channels,financial risks are more likely to spill and transmit through countries with close trade relations.Countries with other substantive ties do not necessarily accelerate risk spillover.Whether in the overall,long-term,medium-term and short-term frequency domain,the abnormal fluctuation and linkage behavior of stock market and foreign exchange market will promote the cross-region and cross-market spillover of financial risk.Forthly,this paper examines forward-looking prediction ability of financial risk spillovers from two aspects of economy and finance,in order to further provide policy suggestions for preventing cross-region and cross-market risk spillovers.Firstly,we use nonlinear Granger causality test to investigate the relationship between macro-economy and time-frequency risk spillovers,as well as the relationship between financial crisis and risk spillovers.It is found that risk spillover in a single frequency domain can not become a consistent,stable and forward-looking early warning index for the economy and finance of different countries.The comprehensive indicator obtained by principal component analysis method is more suitable in predicting the macro-economy and financial crisis,compared with other dimensionality reduction methods.We next examine whether the time-frequency risk spillover indicators can improve the prediction accuracy of macroeconomic downturn.We find that timefrequency risk spillovers and comprehensive risk spillovers have a strong ability to predict macroeconomic downturn,but the density of risk spillover network in different frequency domains has significant differences in predicting macroeconomic downturn.When predicting the financial crisis,considering that the risk spillover network contains rich information,including the size of risk spillover and network topology characteristics,we use the time-frequency risk spillover networks to predict financial crisis,take the time-frequency risk spillover networks as the input variable,and incorporate it into the graph neural network model.The results show that there are differences in the early warning ability of different frequency domain risk spillover networks,and the graph isomorphism model has better prediction ability.Finally,combined with the theoretical and empirical research conclusions,this paper puts forward policy suggestions to prevent cross-region and crossmarket contagion of financial risks and maintain the smooth operation of the financial market from the following four aspects.First,establish a scientific risk monitoring system and early warning mechanism;Second,adhere to the risk supervision concept of combining dynamic regulation with key regulation;Third,establish a "firewall" to block cross-region and cross-market risk contagion;Fourth,balance risk prevention and steady growth,and prevent and resolve financial risks through high-quality development.