中国房地产企业间系统性风险溢出效应分析——基于尾部风险网络模型

本文以2007—2021年沪深A股上市房企为样本,首先基于SIM单指数分位数回归技术提出测量系统性风险的新指标SIM-CoVaR,并结合前沿的TENET网络模型,构造跨房地产企业风险动态传染的尾部风险网络,然后采用块模型探究房地产市场系统性风险溢出的聚类性、触发机制及传播路径,最后考察网络整体结构和宏观经济变量对房地产市场系统性风险溢出的影响。研究表明:(1)我国房地产企业间存在明显的系统性风险联动性和溢出效应,在市场动荡时期房地产部门是金融风险溢出的放大器;(2)评估系统重要性节点企业时,除考虑企业规模等内部属性,还应考虑房企间关联结构,利用系统性风险指数可有效捕捉网络中系统重要性节点;(3)跨房企的系统性风险溢出具有显著的聚类特征,尾部风险网络可被划分为4个不同的功能模块,各模块的成员及其角色呈现明显的时变特性,监管部门可据此从供给端“因企施策”;(4)网络聚集性、网络效率和网络匹配性的降低能显著降低房地产市场的系统性风险溢出效应。本文从企业微观层面探讨房地产市场风险的形成机制,为促进房地产业健康发展和防范化解宏观层面的系统性金融风险提供参考。

Analysis of Systemic Risk Spillover Effects among China's Real Estate Companies Based on the Tail Risk Network Model

At present, China's economic development faces the threefold pressure of demand contraction, supply shock, and expected recession. The liquidity risk, delivery risk, debt default risk, and credit risk caused by ruptures in the capital chain of real estate companies may not only hinder economic growth but also induce systemic financial risks. Therefore, to control financial risks and ensure stable economic growth, it is important to accurately measure real estate companies' financial risk and systemically identify companies in need of supervision or rescue. This is also in line with the goal of “promoting a virtuous circle and healthy development of the real estate industry” set out at the 2021 Central Economic Work Conference.
This paper uses data on listed real estate companies in China's A-share market. First, based on the SIM single-index quantile regression method, a new systemic risk indicator SIM-CoVaR is proposed, and the systemic risk of listed real estate companies is measured. Next, by combining it with the forward-looking TENET model, we construct dynamically evolving real estate market tail risk networks and reveal the time-varying spillover effect of systemic risk in the real estate market through the dynamic network topology. Second, we calculate the systemic risk contribution index and exposure index, and we capture systemically important companies with strong tail risk correlation through the index ranking. Furthermore, through the block model method, we explore the clustering of systemic risk spillover effects and the contagion path between blocks. Finally, we investigate the influence of the overall network structure on systemic risk spillovers in the real estate market.
The empirical results show the following. (1) There is an extensive systemic risk co-movement and spillover effects across real estate companies, and the tail risk network presents heterogeneity. A diversified business structure strengthens the diversified relationship between real estate companies, broadens the channels for dispersing risks, and further reduces the risk spillover level of the overall real estate market. (2) The risk spillover intensity between real estate companies during the 2008 financial crisis was greater than that during the 2015 domestic stock market crash. This difference is affected by the external economy, related policy, and internal linkage, and the real estate industry has become an amplifier of systemic risk spillovers during economic recessions. (3) The systemic risk spillover effect has significant clustering characteristics in the tail risk network. The tail risk network can be divided into four blocks with different functions. In the process of systemic risk transmission, the roles and functions of each block show clear time-varying characteristics. (4) From the perspective of the impact mechanism of spillover effects, the reduction of network aggregation, network efficiency, and network matching can significantly reduce systemic risk spillover. In addition, during periods of economic prosperity and steady development, the real estate sector tends to hide potential risks, and the spillover effects of systemic risks are weakened.
The study makes the following contributions. First, TENET incorporates more companies into the regression through the SIM selection technology and captures the nonlinear risk spillover relationships between real estate companies in a complex higher-dimensional environment, thereby overcoming the limitations of existing models such as high-dimensional interconnection and multivariate variable selection. Second, this paper uses the block model from social network theory to investigate the aggregation, trigger mechanism, and propagation path of risk spillover in the real estate market. Third, this paper discusses the formation mechanism of real estate market risks from the micro-level of companies, which is in line with the actual needs of real estate companies facing a capital dilemma and deeper supply-side regulation in this real estate cycle. The findings provide a useful reference for multi-level real estate market supervision and regulation, and for promoting the healthy development of the real estate industry.