基于多元驱动因素的主权CDS利差预测研究

Forecasting Sovereign CDS Spreads Based on Multiple Determinants

自从欧洲主权债务危机爆发以来,主权风险引发了业界和学术界越来越多的关注.本文基于多元驱动因素视角,采用多种机器学习算法对中美两国主权CDS利差展开预测,并结合平均影响值(MIV)算法分析了中美主权CDS利差驱动因素的异质性.研究发现SVR模型在中美两国的主权CDS利差预测中处于优势地位.而且,中美两国的主权CDS利差的重要驱动因素存在显著差异,本国性因素(例如标普500指数和美元指数)对美国主权CDS利差相对重要,而中国主权CDS利差基本上由全球性因素所决定.研究结论对于国际金融监管协调机构维护全球金融稳定、跨国企业海外投资以及国际投资者保证资产安全具有重要的意义.

Since the outbreak of the European sovereign debt crisis,sovereign risk has attracted increasing attention from industry and academia.Based on the perspective of multiple determinants,this paper adopts several machine learning algorithms to predict sovereign CDS spreads of China and the United States,and combines the mean impact value(MIV)algorithm to analyze the heterogeneity of determinants of sovereign CDS spreads of China and the US.We find that SVR model is in dominant position in the prediction of sovereign CDS spreads of China and the US.Moreover,there are significant differences in driving factors of sovereign CDS spreads of China and the US.Domestic factors(such as S&P 500 index and US dollar index)are relatively important to sovereign CDS spreads of the US,while China’s sovereign CDS spreads is basically determined by global factors.Research conclusions are of great significance to the international financial regulatory coordinating agency to maintain global financial stability,overseas investment of transnational corporation,and international investors to ensure asset security.