Portfolio tail risk forecasting for international financial assets: A GARCH-MIDAS-R-Vine copula model

Abstract

The increasingly complex international environment poses more challenges in accurately forecasting the portfolio risk of international financial assets. Therefore, this paper proposes a generalized autoregressive conditional heteroscedasticity mixed data sampling (GARCH-MIDAS)-R-Vine copula model to forecast the portfolio tail risks, Value at Risk (VaR) and Expected Shortfall (ES), of international financial assets by comprehensively considering the internal complex dependences and external impact of low-frequency macroeconomic factors. Based on the daily prices of Bitcoin, crude oil, gold, seven international stock assets, one global and seven specific monthly economic policy uncertainty (EPU) indexes ranging from January 2011 to August 2022, we find that the proposed model could increase the forecasting accuracy of portfolio tail risk under the optimal information ratio (IR) criterion. Internal high-dimensional dependences can be captured by the flexible R-Vine copula model with 16 kinds of bivariate copula functions, and the external EPU factors observe a significant impact on the corresponding financial assets. Moreover, the CAC 40, the DAX, and the S&P 500 are three dominant financial assets, and Bitcoin and gold are suitable for risk investment and risk hedging assets respectively. These results are beneficial for both risk management and portfolio optimization in the global financial market.