Effect of the U.S.--China Trade War on Stock Markets: A Financial Contagion Perspective

Abstract

In this paper, we investigate the effect of the U.S.-China trade war on stock markets from a financial contagion perspective, based on high-frequency financial data. Specifically, to account for risk contagion between the U.S. and China stock markets, we develop a novel jump-diffusion process. For example, we consider three channels for volatility contagion-such as integrated volatility, positive jump variation, and negative jump variation-and each stock market is able to affect the other stock market as an overnight risk factor. We develop a quasi-maximum likelihood estimator for model parameters and establish its asymptotic properties. Furthermore, to identify contagion channels and test the existence of a structural break, we propose hypothesis test procedures. From the empirical study, we find evidence of financial contagion from the U.S.to China and evidence that the risk contagion channel has changed from integrated volatility to negative jump variation.

Multivariate Overnight GARCH-Itô Model with Applications in Large Volatility Matrix Estimation and Prediction

“…To account for the whole-day market dynamics, Kim and Wang (2021) (Kim and Wang, 2021). Recently, to anaylze the risk contagion between the U.S. and China stock markets, Oh and Kim (2021) proposed the overnight diffusion process, which can accommodate the fact that volatility transmission occurs through overnight volatility processes. In financial applications, we are often required to explore the dynamics of the correlations among assets, which leads us to extend the univariate overnight GARCH-Itô model to a multivariate form.…”

Robust Estimation of Integrated and Spot Volatility

“…In this empirical study, we do not consider the robustness of jumps.18 We notice thatOh and Kim (2021) study the U.S.-China trade war based on a volatility contagion model, and the model is estimated by QMLE. The PaReMeDI estimator is used as an alternative measure of IV, and it is found that the QMLE and PaReMeDI estimates are quite close.19 Note that Aït-Sahalia and Jacod (2014) andLi et al (2022) only consider the regular observation scheme.…”