Melike E. Bildirici, Yildiz Technical University, Istanbul, Turkey

Abstract: This study was developed for two purposes. The first one is detection of the existence of chaotic structure and uncertainty behaviour of the total number of people infected with the COVID-19 outbreak, and the returns of precious metals and oil by using the Lyapunov exponent, Kolmogrov and Shannon entropy tests. The additional aim was to analyze the co-movement and contagion behavior among COVID-19 infected persons, returns of precious metals and oil for the period of between December 30, 2019 and October 26, 2020 by MSGARCH-copula and MSGARCH-multi-layer perceptron methods. Confirmation of the results was provided by using the Diebold-Mariano (DM) tests and Wilcoxon signed rank (WS) tests. Accordingly, the empirical findings of this paper are as follows: The presence of chaotic structure and uncertainty behavior in the variables were determined by Lyapunov exponent, Kolmogorov and Shannon entropy tests. Additionally, co-movement and contagion behavior among the analyzed variables was established by the MS-GARCH-MLP copula method. Following, in the context of the results of confirmation, it has been determined that MSGARCH-MLPst has the best prediction performance under all criteria compared to MSGARCHst, or GARCHst. As a further result, COVID-19 had considerable effects on the returns of precious metals and oil prices, and there was co-movement and contagion behavior between COVID-19 and the returns of precious metals and oil.

Keywords: COVID-19, precious metals, oil price volatility, Markov-switching, GARCH, neural network, MLP