key: cord-0856668-f6fcgped
authors: Yu, Xiaoling; Xiao, Kaitian; Liu, Junping
title: Dynamic co-movements of COVID-19 pandemic anxieties and stock market returns
date: 2021-06-08
journal: Financ Res Lett
DOI: 10.1016/j.frl.2021.102219
sha: c8f4c209006ce291ff67d79a77a03b1d7fea94ef
doc_id: 856668
cord_uid: f6fcgped

In this study, we constructed two pandemic anxiety indexes based on an assumption that people's emotions fluctuate with the COVID-19 reported cases and deaths, to examine the dynamic co-movements between these anxiety indexes and the stock markets in the BRICS and G7 countries. We found that the anxiety indexes are volatile over time but have an overall downtown trend. The correlations between stock market returns and the epidemic anxiety indexes are time varying. We found a common feature across the countries studied, namely that the correlation becomes weaker and has smaller fluctuations after the announcement of the mRNA-based COVID-19 vaccine.

Since the COVID-19 pandemic emerged in January 2020, it spread rapidly and became a human tragedy as well as both a healthcare and economic crisis around the world. To date, the total reported cases of COVID-19 exceed one hundred million, meaning that nearly 1in 70 people in the world have been infected. As the original cause has not been identified, and it is not yet known how much longer the pandemic will last, general psychological disturbances, mental anxiety, and economic shocks remain elevated. There is no doubt that the COVID-19 pandemic has been an urgent topic of research for academicians and experts in different fields. For example, there is much research being done by governments and public health experts who are responsible for making decisions to control and suppress the COVID-19 crisis and prevent future epidemics (Atalan, 2020; Kanitkar, 2020; Noorbhai, 2020) . Scholars in the field of medicine and epidemiology are studying the origin, health outcomes, treatments, and vaccine development for the COVID-19 disease (Hetkamp et al., 2020; Salameh et al., 2020; Wuyts et al., 2020) Researchers in economics are examining whether and how the pandemic has affected the economy in various ways (Allam, 2020; Ashraf, 2020a; Bernauer and Slowey, 2020; Hossain, 2021; Keogh-Brown et al., 2020) .

In this study, we examine the dynamic correlation between the COVID-19 pandemic and stock market returns in 12 countries, namely Brazil, Russia, India, China, and South Africa (often referred to as the BRICS countries) and the G7 countries, i.e., the United States, England, Canada, Italy, Germany, France, and Japan. There are already a number of studies on the relationship between the COVID-19 pandemic and financial markets. Zhang et al. (2020) analyzed the systemic risks and potential consequences of policy interventions in the global financial markets during the pandemic. Ali et al. (2020) investigated the reactions of global financial markets as the pandemic spread from Asia to Europe and then to North America. Some studies examined the stock markets' reaction in terms of volatility during COVID-19 (Baek et al., 2020; Corbet et al., 2021; Just and Echaust, 2020; Takyi and Bentum-Ennin, 2020; Topcu and Gulal, 2020; Yong and Laing, 2020; Ashraf, 2020b) . Other studies focused on the pandemic's impact on connections across international markets (Guo et al., 2021; So et al., 2021) .

Our study expands the literature on the relationship between the COVID-19 pandemic and financial markets. Our study's contributions can be summarized as follows: First, in contrast to studies that used the number of reported COVID-19 cases as a key variable (Just and Echaust, 2020; Xu, 2021; Yong and Laing, 2020) , we constructed two kinds of pandemic anxiety index based on the numbers of daily reported cases and deaths to examine the linkage between the COVID-19 pandemic and stock markets. Secondly, our approach differs from the methods used in previous studies, such as the event study, VAR models, the OLS regression approach, and Granger complex network (Baek et al., 2020; Lai and Hu, 2021; Sun et al., 2021; Topcu and Gulal, 2020; Xu, 2021) . We first employed a Dynamic Conditional Correlation-Generalized Autoregressive Conditional Heteroskedasticity (DCC-GARCH) model to capture the dynamic co-movement between the pandemic's evolution and stock market volatility. Then, we selected BRICS and G7 countries as research objects, as they are representative of emerging and developed economies, respectively.

The remainder of the study is organized as follows: Section 2 introduces our methodologies, Section 3 describes and provides some basic statistics about the data used in the study, Section 4 details the empirical results, and Section 5 concludes the study.

Following Salisu et al. (2020) ,we construct two types of pandemic anxiety indexes as proxies for measuring sentiments about the spread and severity of COVID-19, based on an assumption that people's anxiety emotions fluctuate with the changes of the COVID-19 reported cases and deaths. The first index, called COVID-19 Anxiety Index (AI), is similar to the approach in Salisu and Akanni (2020) and was constructed using daily reported cases and deaths (Salisu and Akanni, 2020) . The second index, called COVID-19 Rolling Anxiety Index (RAI) was constructed based on the total number of reported cases and deaths in the 14-day incubation period. The two pandemic anxiety indexes can be specified as below:

(1)

where C i,t denotes the number of COVID-19 reported cases at dayt for country i; D i,t denotes the number of COVID-19 reported death at dayt for country i;C i,t − 14 andD i,t − 14 are reported cases and deaths, respectively, in country i at the beginning of the incubation period. ∑ 14 k=1 C i,t− 14+k and ∑ 14 k=1 D i,t− 14+k denote the total number of reported cases and deaths, respectively, in the 14-day incubation period including dayt; while ∑ 13 k=0 C i,t− 14+k and ∑ 13 k=0 D i,t− 14+k respectively denote the total number in the previous 14-day incubation period.

After constructing the pandemic anxiety indexes, we next take Granger causality test to examine whether changes in the level of anxiety associated with the COVID-19 pandemic causes a change in stock market returns. The Granger causality test equations are specified as follows:

where r i,t is the one-day log stock return at time t for country i; and AI i,t and RAI i,t are the two COVID-19 pandemic anxiety indexes constructed above. p is the lag order which is selected by BIC information criterion. We calculated the daily log return at time t for country i as r i,t = 100*log(r i,t /r i,t − 1 ).The null hypothesis of the Granger causality test is β i,t− 1 , …, β i,t− p = 0. Rejecting the null hypothesis means that the anxiety due to the pandemic causes a change in stock market.

Next step, we further constructed a bivariate DCC-GARCH (1,1) model to examine the dynamic correlation between stock volatility and COVID-19 pandemic anxiety indexes. Engle (2002) introduced the dynamic conditional correlation-Generalized Autoregressive Conditional Heteroskedasticity (DCC-GARCH) model to examine time-varying correlations between assets. The bivariate DCC-GARCH (1,1) model is specified below:

where r t = (r stock ,r efi ) ′ ,r stock is the stock market return and r efi is the change of pandemic anxiety index; z t is the distributed error vector with z t = (z t,stock , z t,efi ) ′ ; ε t is the standardized residual vector of the mean equation with ε t = (ε t,stock , ε t,efi ) ′ ; H t is the dynamic conditional

; ω, α and β are the parameters of GARCH (1,1) model while a DCC and b DCC are non-negative scalar parameters of the DCC procedure, where a DCC + b DCC < 1. R t is the dynamic correlation coefficient matrix and D t is the conditional standard deviation matrix;Q t is a symmetric positive definite matrix.

We collected daily stock market data for the BRICS and G7 countries from the CEIC database. For each country, we selected one Fig. 6 in Appendix). Fig. 2 plots the trends of absolute number of daily COVID-19 reported cases and deaths for each country. The mean daily reported cases for Brazil, Russia, India, China, South Africa, the United States, England, Canada, Italy, 3 . Trends in the two daily COVID-19 anxiety indexes for the BRICS and G7 countries. Note: The pandemic AI is plotted on the left axis, and the pandemic RAI is plotted on the right axis.

Germany, France, and Japan are 27,263,11,555, 32,485,61,4357, 76,830,11,344,2307,7613, 6644,9331, and 1149, respectively . The corresponding mean daily reported deaths are 667, 219, 467, 5, 131, 1290, 313, 59, 264, 169, 225 , and 17, respectively. As we know, the corresponding total population of these 12 countries are 210. 000, 144.478, 1324.000, 1400.050, 56.52, 330.000, 65.580, 37.059, 60.800, 82.930, 66 .987 and 124.776 million, respectively. If we standardize the daily numbers of reported cases by using the total population in each country, the corresponding infection destiny are 129. 824, 79.978, 24.535, 0.044,77.088,232.818,172.980,62.252, 125.214,80.116,139.296 and 9 .209 per million, respectively. This shows that among the countries included in this study, the U.S. has been most heavily affected by the pandemic (the infection destiny is 232.818 per million), while the impact on China has been the least severe (the infection destiny is 0.044 per million). Fig. 3 plots trends of the COVID-19 pandemic anxiety indexes. We observed a downward trend in both the anxiety indexes as a whole, despite various phases of increases and decreases in reported cases and deaths across all the countries in our sample. In particular, in the second half of 2020, although the daily number of reported COVID-19 cases and deaths hit an all-time high (except in China), the anxiety indexes were still lower than at the beginning of the COVID-19 pandemic. Table 2 summarizes the descriptive statistics for the stock market returns and COVID− 19 pandemic anxiety indexes of each country. All series, except market returns for China and Russia, are skewed and have excess kurtosis, indicating a leptokurtic distribution. As shown in Table 3 , the Augmented Dickey-Fuller (ADF) test results for the pandemic anxiety indexes, except for the AI in China, the United States, Germany, and France, show that we cannot reject the null hypothesis of a unit root, implying the time series are nonstationary. However, all the ADF test results for the differences in stock market returns and pandemic anxiety indexes (Δr, ΔAI, Notes:This table shows descriptive statistics of stock market returns and pandemic anxiety indexes for the BRICS and G7 countries. In order to observe whether the skewness and kurtosis of the stock returns have changed from the pre-pandemic period, we also report the skewness and kurtosis for the same period prior to our sample in the right part of this table. Our sample period spans from March 20, 2020 to January 29, 2021 while the benchmarked pre-sample period covers May 9,2019 to March 19, 2020, whereby including 226 trading dates for each period. As shown in this table, the stock returns, except for India, are more skewed and have fatter tails during the pandemic than under "normal" times in all cases. The COVID-19 AI and RAI are shown in Eqs. (1) and (2), respectively. We calculated daily anxiety indexes from March 20, 2020 to January 29, 2021, using data from Monday to Friday each week to match stock trading dates. ADF test r − 11.174*** − 10.824*** − 13.004*** − 9.8768*** − 9.7788*** − 10.299*** − 9.5583*** − 13.215*** − 9.0987*** − 9.716*** − 9.7338*** − 9.1193*** Notes: This table reports the results of stationarity and ARCH tests for stock market returns and pandemic anxiety indexes. r, AI, and RAI represent stock market returns, the pandemic anxiety index, and the rolling anxiety index, respectively. Δr, ΔAI, and ΔRAI denote the corresponding differential series, Δr t =r t -r t− 1 , ΔAI t = 100*(AI t -AI t − 1 ) and ΔRAI t = 100*(RAI t -RAI t − 1 ). The p-values are presented in the parentheses. *** , and *** indicate significance at the 10%, 5%, and 1% levels, respectively. and ΔRAI) are significant at the 1% level. Moreover, the corresponding ARCH test results for the differential series ΔAI, except for India, and for the series ΔRAI, except for Germany, are statistically significant. This highlights that the data is suitable to estimate a GARCH model using the changes in stock market returns and the pandemic anxiety indexes, namely Δr, ΔAI, and ΔRAI.

Table 5 summarizes the unconditional correlation matrix and the results of the Granger causality tests as shown in Eqs. (3) and (4) for each country, where the optimal lag orders are selected based on the Bayesian Information Criterion (BIC) shown in Table 4 . As shown on the left side of Table 5 , the Chinese stock market has the weakest correlation with its domestic pandemic anxiety index (C Δr-ΔAI = 0.0127; C Δr-ΔRAI = 0.0040), while India has the strongest unconditional correlation with its domestic pandemic anxiety index (C Δr-ΔAI = − 0.5729; C Δr-ΔRAI = − 0.6318).

Tables 6 and 7 show the estimated parameters of the bivariate DCC-GARCH model based on Eqs. (5) to (10). Table 6 reports the Notes: This table reports the results of lag selections for bivariate VAR model based on the BIC criterion. The selected lags will be used to take Granger tests for whether the change in pandemic anxieties causing change in stock market returns and also will be used in the DCC-GARCH model hereinafter. The left part shows the results for pandemic anxiety index AI and stock market returns while the right part shows the results for pandemic rolling anxiety index RAI and stock market returns in each country. * presents the optimal lag order. The results of the Granger causality tests shown in the right side of Table 5 indicate that the null hypothesis is rejected for 8 out of 12 selected countries, namely, Brazil, Russia, India, South Africa, the United States, Canada, Germany and Japan, indicating that the COVID-19 Anxiety Index changes, causing stock market returns in these eight countries. Interestingly, among these eight countries, the unconditional correlations between the changes in stock market returns and changes in the COVID-19 Anxiety Index are positive for Brazil, South Africa, the United States, Canada and Germany, while the correlations are negative for Russia, India and Japan. It highlights that the impact of the COVID-19 pandemic on stock market are not consistent for different countries. Note:The left side of this table presents unconditional correlations between stock market returns and COVID-19 anxiety indexes for each country; the right side shows the results of the Granger causality tests. C Δr-ΔAI denotes the correlation coefficient between daily changes in stock market returns (Δr) and pandemic anxiety index (ΔAI), C Δr-ΔRAI is the correlation coefficient between daily changes in stock market returns (Δr) and pandemic rolling anxiety index (ΔRAI), while C ΔF -ΔRAI denotes the correlation between AI and RAI. Table 8 and 9, respectively. ***, and *** indicate significance at the 10%,5%, and 1% level, respectively. results from the model for stock market returns and the AI, while results for the model involving stock returns and the RAI, are presented in Table 7 . We concentrate on the estimated parameters of the dynamic conditional correlation based on Eq. (9), shown in the columns labeled a DCC and b DCC . For the DCC between stock market returns and the AI, the estimated results of a DCC are significant for the United States, Germany, France and Japan, while the estimates for b DCC are significant in all cases except for China, England, Canada, Germany and France, as shown in Table 6 . The estimated parameters a DCC and b DCC for the dynamic conditional correlation between stock market returns and the RAI, are shown in Table 7 . The results for a DCC are significant for India, Canada and Italy, while the estimates for b DCC are 1) for RAI. a DCC and b DCC are parameters of the DCC procedure. Standard errors are presented in parentheses. ***and *** indicate the significance level at 10%,5% and 1%, respectively. statistically significant for Brazil, Russia, India, South Africa, the United States, England, Canada, Germany, France and Japan. The time-varying conditional correlations between the stock market and AI and RAI for each country are depicted in Figs. 4 and 5, respectively. We found that the conditional correlations between the stock market and the two pandemic anxiety indexes were time varying in all cases. With respect to the range of dynamic conditional correlations, the links between the stock market and both the Fig. 4 . Dynamic conditional correlations between stock market returns and the COVID-19 AI for each country. Note: Graphs are portioned into two phases by a dotted line, depicting the co-movements of stock markets and the AI before and after the announcement of the mRNA-based COVID-19 vaccine. Table 5 . The range of correlations between the stock market and AI is largest in India (− 1.0~0.5), while the range of correlations for the stock market and RAI is strongest in the United States (− 1.0~1.0). It seems that the lower the pandemic AI, the less correlated it is with the stock market returns. A decrease in the anxiety index could be attributed to greater knowledge of the COVID-19 virus, more preventive measures, and especially, the availability of a vaccine. On November 18,2020, Pfizer Inc. and BioNTech SE announced the results of the Phase 3 clinical trial of their mRNA-based COVID-19 vaccine, which showed a efficacy rate of 95% (p < 0.0001).

As shown in Figs. 4 and 5, after the COVID-19 vaccine was announced, the co-movements between the stock markets and pandemic anxiety indexes weakened and fluctuations in the correlations declined in the twelve countries, despite the fact that the number of reported cases and deaths (except in India and China) hit their highest levels during this period (see Fig. 2 ). In addition, except in India, the highest correlations occurred in the first half of 2020, corresponding to the beginning or the first small surge in the pandemic for each country.

The COVID-19 pandemic spread around the world in 2020, and the link between the pandemic and financial markets is worth studying empirically. In this study, we constructed two anxiety indexes (AI and RAI) related to the COVID-19 pandemic based on daily reported cases and deaths and used them to examine the co-movements between the pandemic and stock market returns in the BRICS and G7 countries, employing a bivariate DCC-GARCH model over the period from March 23, 2020 to January 29, 2021.

We showed that the COVID-19 pandemic anxiety indexes were volatile over the aforementioned period but displayed a downtown trend over time for each country. Even when the number of daily reported cases and deaths hit the highest level in the second half of 2020, the pandemic anxiety indexes were still below the previous level reached in the early stages of the pandemic. We also showed that the conditional correlations between stock market returns and the two pandemic anxiety indexes fluctuate over time for each country, experiencing phases of increased and decreased correlation. A common feature of the co-movements between stock market returns and the pandemic anxiety indexes, found for all of the BRICS and G7 countries, is that the correlations weakened and had smaller fluctuations after the COVID-19 vaccine results were announced. Moreover, compared with other countries, the results show that China's stock market has the weakest correlation with its domestic COVID-19 anxiety indexes.

One weakness of our study is the fact that the empirical results show that the correlations between changes in stock returns and changes in the COVID-19 pandemic anxiety are not consistent and stable for different countries, sometimes positive for some countries and sometimes negative for the others, but we cannot figure out the specific reasons behind these phenomena, in that there are many other sources of uncertainty or outcomes (such as economic policy, pandemic prevention measures and the United States Presidential election, etc.) that affect stock returns. Secondly, we construct the anxiety index based on the assumption that people's emotions (3), with the standard errors presented in the parentheses. ***, and *** indicate significance at 10%, 5%, and 1% levels, respectively.

fluctuate with the changes of COVID-19 reported cases and death. However, the actual sentiment changes are more complicated than that in real life. In addition, we just select 12 countries but not all countries in the world for this study. Some other countries successfully suppressing the spread of COVID-19 (e.g., New Zealand, Vietnam) could have been useful for comparing China's results. As part of future analysis, it would be interesting to extend our study to resolve these limitations. returns for the BRICS and G7 countries. φi,α i,1 , α i,2 , α i,3 , α i,4 ,β i,1 , β i,2 , β i,3 and β i,4 are the parameters based on Eq. (4) with the standard errors presented in parentheses. ***, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 

Coronavirus (COVID-19)-An epidemic or pandemic for financial markets

Chapter 9 -Oil, Health Equipment, and Trade: revisiting Political Economy and International Relations During the COVID-19 Pandemic

Economic impact of government interventions during the COVID-19 pandemic: international evidence from financial markets

Stock markets' reaction to COVID-19: cases or fatalities?

Is the lockdown important to prevent the COVID-19 pandemic? Effects on psychology, environment and economy-perspective

COVID-19 and stock market volatility: an industry level analysis

COVID-19, extractive industries, and indigenous communities in Canada: notes towards a political economy research agenda

Pandemic-related financial market volatility spillovers: evidence from the Chinese COVID-19 epicentre

Dynamic Conditional Correlation

Tail risk contagion between international financial markets during COVID-19 pandemic

Sleep disturbances, anxiety, and generalized anxiety during the COVID-19 shut down phase in Germany: relation to infection rates, deaths, and German stock index DAX Sleep Med

Stock market returns, volatility, correlation and liquidity during the COVID-19 crisis: evidence from the Markov switching approach

The COVID-19 lockdown in India: impacts on the economy and the power sector

The impact of Covid-19, associated behaviours and policies on the UK economy: a computable general equilibrium model

A study of systemic risk of global stock markets under COVID-19 based on complex financial networks

A mathematical model to guide the re-opening of economies during the COVID-19 pandemic

Mental health outcomes of the COVID-19 pandemic and a collapsing economy: perspectives from a developing country

Constructing a Global Fear Index for the COVID-19 Pandemic

Impacts of the COVID-19 pandemic on financial market connectedness

The impact of COVID-19 on the Chinese stock market: sentimental or substantial?

The impact of COVID-19 on stock market performance in Africa: a Bayesian structural time series approach

The impact of COVID-19 on emerging stock markets

Circular economy as a COVID-19 cure? Resour

Stock Return and the COVID-19 pandemic: evidence from Canada and the US

Stock market reaction to COVID-19: evidence from U.S. Firms' International exposure

Financial markets under the global pandemic of COVID-19

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.