The main objective of this research is to analyse the serial dependence of high frequency data for G7 stock indices. The authors use two different periodicities, and with linear and nonlinear approaches, they evaluate the stock markets' behaviour and conclude about the higher or lower dependence levels of the stock markets in the periods before and after the COVID-19 pandemic declaration. They use mutual information and the global correlation coefficient based on that measure, comparing results with the linear coefficient. The results are clear, showing that nonlinear dependence exists and could be an important factor in terms of historical information, especially for very high frequency data. Results are mixed in regard to the effect of the pandemic declaration in the dependence of stock markets.
TopIntroduction
In the current health crisis, we are witnessing a world that changes its behavior in all areas of citizens' lives. At the social and economic level, the impact of the pandemic is evident and dramatic. Accompanying this anomalous behavior, financial markets also showed abrupt reactions, with investor distrust taking root. It is in this context of crisis and attempted recovery that we consider it essential to understand the individual behavior of financial markets that are fundamental to the world economy.
The main objective of this research is to analyze the effect of Covid-19 on the serial dependence of high frequency data for G7 stock indices. In this data, we will use two different frequencies: 5 and 30 minutes, and through linear and nonlinear approaches we will try to evaluate the stock markets’ behavior and conclude on each market’s greater or lesser proximity to efficiency in different sub-periods.
In order to evaluate the effects of the Covid-19 crisis, we split the period under analysis in 2 sub-periods: pre-pandemic (from 1 August 2019 till 31 March 2020) and post-pandemic (from 12 March 2020 till 1 June 2020).
Most prior studies generally found that financial markets could be described by random walks (see, for example, Fama 1963, Osborne 1964 or Granger and Morgenstein 1963, among others). One possible explanation could be related to the use of linear methodologies to analyze dependence, which could fail to detect other kinds of dependence, such as nonlinear ones. In this case, the use of linear approaches is not sufficient (see, for example, Granger et al., 2004 and Krause et al. 2018). According to Campbell (1987), Wen et al. (2018) and Nam and Seong (2019), linear serial dependence can quickly disappear, while other kinds of dependence could be present in data and not detected by those linear methodologies.
In order to achieve our goal, we will use mutual information and the global correlation coefficient, comparing results with the linear coefficient.
The concept of mutual information (MI) used in this research is based on entropy differences from the theory of communication (Shannon, 1948). The main feature of MI is the capacity to measure common information in a global way between two or more random variables, as used in several studies to examine dependence structures or global correlation (see for example Darbellay and Wuertz, 2000; Menezes et al., 2012; Ferreira and Dionísio, 2014, 2017; Mothi et al., 2019, Nam and Seong, 2019).
In the context of the Covid-19 crisis, the literature contains several studies analyzing the effects of the pandemic on financial markets, including the effects on stock markets (see, for example, Zhang et al. 2020; Aslam, Nogueiro, Brasil, Ferreira, Mughal, Bashir & Latif, 2020; Haroon & Rizvi, 2020) but also in other financial assets such as exchange rates (Iyke, 2020), commodities (Apergis & Apergis, 2020; or Devpura & Narayan 2020) or even insurance (see Wang, Zhang, Wang & Fu, 2020), among others.
Our results point to high and significant levels of global autocorrelation in the 5-minute frequency, with these autocorrelations increasing in the post-pandemic period. For the 30-minute frequency, the results are somewhat ambiguous, showing some similar patterns in European stock markets and NYSE.