Abstract: Ever since the 2008 global financial crisis, the supervision of systemic financial risk has been a hot topic in the field of academic and policy-making departments, both at home and abroad. Especially since 2012, financial system reform began to accelerate, and investment constrain have been gradually deregulated. The extensive relevance and intersectionality of the financial services business brought about significant changes in the financial sector, which led to a substantial increase in systemic financial risk.The multivariate quantile regression model provides a good tool for analyzing systemic risk. Considering the deficiency of original MVMQ-CAViaR model ignores the asymmetric impacts of positive and negative shock. In this paper, it is extended to asymmetric MVMQ-CAViaR model and joint asymmetric MVMQ-CAViaR model. Subsequently, these models are used to study China's financial industry risk transmission effect between different sectors. Then both Kupiec LR(likelihood ratio) test and dynamic quantile test are used to backtest the prediction performance of these models.
The results show that:Banks have significant spillover effects on securities and insurance sectors, while securities can just unidirectional absorb other sectors' risk;The impacts of good and bad news exhibit leverage effect to some extent to their own as well as other sectors. In general, negative shock has greater effect than positive effect. Furthermore, joint negative impact will amplify the current risk level;Two newly constructed models can significantly improve the risk prediction accuracy, and joint asymmetric MVMQ-CAViaR model is relatively more competitive.
Important practical and social implication are suggested.First of all. Regulators should pay special attention on strengthening the disclosure system of bank risk and the transparency of bank financial information. Then policy makers should strengthen the macro-prudential regulatory requirements and build good co-operation relationship between different industries in order to deal with emergency warning system.

Key words: risk transmission, MVMQ-CAViaR model, asymmetric characteristics, systemic risk