The COVID-19 pandemic uncertainty, investor sentiment, and global equity markets: Evidence from the time-frequency co-movements

Abstract

We use daily data of the Google search engine volume index (GSVI) to capture the pandemic uncertainty and examine its effect on stock market activity (return, volatility, and illiquidity) of major world economies while controlling the effect of the Financial and Economic Attitudes Revealed by Search (FEARS) sentiment index. We use a time–frequency based wavelet approach comprising wavelet coherence and phase difference for our empirical assessment. During the early spread of the COVID-19, our results suggest that pandemic uncertainty, and FEARS sentiment strongly co-move, and increased pandemic uncertainty leads to pessimistic investor sentiment. Furthermore, our partial wavelet analysis results indicate a synchronization relationship between pandemic uncertainty and stock market activities across G7 countries and the world market. Our results are robust to the inclusion of alternative pandemic fear measure in the form of equity market volatility infectious disease tracker. The pandemic uncertainty and associated sentiment implications could be one plausible reason for increased volatility and illiquidity in the market, and hence, policymakers should look upon this issue for the financial market stability perspective.

1. Introduction

This paper examines the COVID-19 pandemic information internet-search intensity1 (Pandemic, hereafter) impact on the stock market returns, volatility, and (il)liquidity of major world economies after controlling the effect of investor sentiment. The coronavirus (COVID-19) pandemic has been acknowledged as a global catastrophe with unprecedented socio-economic consequences. The gravity of the situation was explained by the WHO Director-General as “a unique virus, with unique features. We are in uncharted territory.”2 Baker et al., 2020, Huynh et al., 2021, Mishkin and White (2020), Salisu and Akanni (2020), among others, also emphasized that this pandemic spread is indeed different. Amid the COVID-19 spread, analysts have opined that the fallout of the coronavirus pandemic will be a threat to the stability of the world economy. During the spread of the pandemic, the combined effects of three crucial transmission channels related to demand shock, supply shock, and financial shock have been unprecedented (OECD, 2020, World Bank, 2020a, World Bank, 2020b). As COVID-19 continues to disseminate every day, policymakers and market participants worldwide are concerned about its profound economic consequences.

Stock market activities have received much attention in the popular financial press and various policy domain discussions. A fast-growing body of recent research suggests that due to the increasing spread of COVID-19, global stock markets have experienced high volatility and lower stock returns (Al-Awadhi et al., 2020, Baker et al., 2020, Ramelli and Wagner, 2020, Shaikh and Huynh, 2021, Sharif et al., 2020, Zaremba et al., 2020, Zhang et al., 2020). A common consensus that underscores the pandemic’s adverse effects on stock market activity is the economic fallout, heightened policy uncertainty, liquidity concerns, gloomy market sentiment, and looming global financial crisis (Baker et al., 2020, OECD, 2020). There was increasing concern that the stock market looks increasingly dissociated from economic reality (Capelle-Blancard and Desroziers, 2020, Krugman, 2020, Shiller and Malkiel, 2020). For instance, Capelle-Blancard and Desroziers (2020) noted that “since the beginning of the crisis, stock prices seem to be running wild. They first ignored the pandemic, then panicked when Europe became its epicenter. Now, they are behaving as if the containment of half the world’s population will have no economic impact after all.” In other words, the relationship between stock performance has been primarily driven by the oscillation between greed and fear (Krugman, 2020). If speculative bubbles (positive or negative) last long, the objective relation between fundamentals and stock valuations may become suboptimal (Shiller, 2003). The COVID-19 pandemic represents an excellent example of how investor sentiment and difficulty in predicting the severity of the resulting economic disruption can affect stock market activity (Shiller & Malkiel, 2020). One crucial aspect that has been largely ignored is the interrelationship between pandemic information search intensity, investor sentiment, and stock market activity. Our empirical analysis focuses on two research questions. First, whether a time–frequency co-movement relationship exists between the COVID-19 pandemic uncertainty and investor sentiment measured by the Financial and Economic Attitudes Revealed by Search (FEARS) index. Second, whether the COVID-19 pandemic uncertainty influences global stock market behavior after controlling the effects of investor sentiment (FEARS index). The existing recent literature highlights several potential arguments to motivate us for our research questions.

Our first research question identifies the potential relationship between pandemic fear (measured through the internet search volume index information for the COVID-19 and Coronavirus search terms) and investor sentiment. It is essential to understand the relationship between pandemic uncertainty and investor sentiment. The COVID-19-induced investors' fear appears to be higher in the equity segment for the first time since the market crash of 1987 and the global financial crisis of 2008–2009 (Shaikh & Huynh, 2021). Moreover, excessive fear could significantly affect investment sentiments and decisions (Costola et al., 2021, Huynh et al., 2021, Salisu and Akanni, 2020). Fear increases investor pessimism, leading to investors' overreaction to bad information or news (Su, Liu, & Fang, 2021). Because the full effect of any new information related to the pandemic spread is not immediately apparent to market participants, some participants may vastly underreact (underestimate) or overreact (overestimate) to the arrival of such information in the news (Shiller & Malkiel, 2020). Since there is no fundamental psychological principle that people tend to overreact or underreact (Shiller, 2003), the short-term implications of sentiment-induced mispricing on the market during COVID-19 cannot be ruled out completely. Behavioral finance suggests that investors’ psychological biases could be more pronounced when there is more significant uncertainty (Donadelli et al., 2017, Garcia, 2013). Accordingly, we hypothesized that the increased pandemic-induced uncertainty (Baker et al., 2020) could have led to pessimistic investor sentiment. Recent studies by Salisu and Akanni, 2020, Vasileiou, 2021, and Smales (2021) suggest that investors are perhaps paying more attention to search information to resolve uncertainty concerning the COVID-19 crisis rather than fundamental information. Given the incomparable economic consequences of COVID-19, it has received large-scale media attention worldwide. Recent empirical evidence also suggests that increased volatility in the equity markets results from overwhelming panic generated by coronavirus-related news (Biktimirov et al., 2021, Haroon and Rizvi, 2020, Zaremba et al., 2020). A related research strand on information flow and demand suggests that high media pessimism predicts financial market behavior with increasing volatility and lower return (Garcia, 2013, Tetlock, 2007). Thus, it is reasonable to argue that the COVID-19 pandemic uncertainty could have crucial implications for investor sentiment, and these two are independent information variables. However, limited empirical evidence exists that examines the time-varying relationship between the information transmission channel of pandemic information demand (Pandemic uncertainty or pandemic attention) and investor sentiment. The second important aspect is the lack of sufficient empirical evidence on the COVID-19 impact on aggregate stock market behavior, i.e., stock return, volatility, and market il(liquidity). Existing research majorly focuses on the stock return, and volatility aspects (Anastasiou, Ballis, & Drakos, 2022; Costola et al., 2020; Su et al., 2021, Tripathi and Pandey, 2021; Vasileiou, 2021; Wang et al., 2021, Zaremba et al., 2020 among others) and the implication of market liquidity has largely been ignored. Moreover, available literature fails to control the effect of prevailing exogeneous market sentiment while analyzing the impact of COVID-19 pandemic uncertainty on stock market activity. Our approach first considers the co-movement between FEARS sentiment and pandemic uncertainty. Subsequently, it assesses the independent effects of pandemic uncertainty on stock returns, volatility, and (il)liquidity after removing sentiment impact.

Our sample consisted of G7 countries and an overall measure of the world market. Existing research highlights that the empirical analysis of information in Google’s relative search volume index can help measure public attention on the global epidemic and its spread (Effenberger et al., 2020, Ginsberg et al., 2009) and household sentiment (Da, Engelberg, & Gao, 2015). A rise in the number of internet searches during the COVID-19 crisis induced a faster rate of information flow into financial markets (Costola et al., 2021, Smales, 2021, Tripathi and Pandey, 2021), enhancing the fear of the pandemic (Su et al., 2021) and therefore, a higher search volume index of pandemic information can be interpreted as a measure of coronavirus-related uncertainty and perceived risk (Chundakkadan and Nedumparambil, 2021, Szczygielski et al., 2022, Wang et al., 2021). We use daily data of the GSVI to capture COVID-19 pandemic information demand and investor FEARS sentiment index. Our empirical design considers the wavelet-based time–frequency analysis. We utilized a time–frequency-based approach comprising wavelet coherence and partial wavelet analysis for our empirical assessment. This empirical approach enabled us to study the interdependence of pandemic information demand, sentiment, and stock market behavior to better understand the possible interrelationship (Rubbaniy et al., 2021a, Rubbaniy et al., 2021b). We examine the co-movements and the lead-lag relationship between the Pandemic uncertainty and FEARS sentiment index using wavelet coherence and phase difference approach. While investigating the effects of pandemic uncertainty on stock market behavior, we resorted to controlling FEARS sentiment effects using partial wavelet coherency analysis. The use of partial wavelet coherency allows us to examine the scale-specific and localized relationship between the pandemic uncertainty and stock market activity (return, volatility, and liquidity) after eliminating the influence of common dependence between exogenous market sentiment and pandemic uncertainty. Our robustness test uses the newly constructed Equity Market Volatility Infectious Disease Tracker or EMV (Baker et al., 2019, Baker et al., 2020) as an alternative proxy for the pandemic fear sentiment. The application of wavelet analysis combines the advantage of drawing inference from the time series data by focusing on the time-varying relationship between both time and frequency domains (Al-Yahyaee et al., 2019, Dash et al., 2019, Rubbaniy et al., 2021a, Rubbaniy et al., 2021b, Sharif et al., 2020). The time–frequency-based wavelet analysis helps to accommodate these two domains within a unified framework (Crowley, 2007, Dash and Maitra, 2019, Mensi et al., 2019, Ramsey, 2002, Sharif et al., 2020).

Our time–frequency co-movement results suggest that the FEARS sentiment and Pandemic indexes are positively correlated. Results indicate that pandemic spread leads to pessimistic investor sentiment. On 1 March 2020, this strong coherence intensified after COVID-19 was declared a global pandemic. Our partial wavelet analysis results show signs of a synchronization relationship across G7 countries and the world market between pandemic uncertainty and stock market activities (after controlling for FEARS sentiment). Our results are robust to the use of alternative measure of pandemic fear sentiment.

This study extends the rapidly growing body of research on the impact of COVID-19 on the stock market behavior. Our study also provides some initial evidence on the pandemic induced uncertainty and stock market liquidity. Existing literature majorly focused on the return and volatility dimensions of the stock market behavior, and the trading or liquidity aspect of the market has been largely ignored. To the best of our knowledge, no known study focuses on the stock market's return, volatility, and (il)liquidity effect amid the pandemic spread. Our findings contribute to the ongoing discussion between the co-movements of pandemic induced uncertainty and investor sentiment. Existing literature majorly focuses on the COVID-19 investor attention or pandemic uncertainty measures derived from the search volume information on the stock market without accounting for the exogeneous sentiment effect in the market. There has been a limited examination of the impact of pandemic-information search on investor sentiment and its effect on stock markets activity. Our empirical approach also brings novelty with the use of wavelet-based time–frequency analysis. The wavelet-based approach helps to accommodate different investment horizons and understand the relationship between pandemic attention, sentiment, and stock market behavior in various time–frequency domains. Markets are composed of multiple agents operating in each moment at different time scales (short- and long-term). It is essential to accommodate both the time and frequency domains within a unified framework to simultaneously assess how our variables were related at different frequencies and how such a relationship has evolved. This approach provides a scope to explore the relationship across different frequencies of sentiment variations, which is essential for investors considering different investment opportunities (e.g., daily -movements for short-term investors) and for policymakers looking at the broader aspect of the financial market stability. This paper also extends the related strand of literature on the implications of internet search-based market participants’ expectation measures of financial market behavior.

The remainder of this paper is structured as follows. Section 2 provides a brief overview of related literature. Section 3 presents data and variables. Section 4 discusses our empirical approach. Section 5 outlines our main empirical results and robustness tests. Lastly, section 6 concludes the paper.

2. Review of related literature

The coronavirus (COVID-19) outbreak in early 2020 has posed adverse economic consequences on the global economy and financial markets worldwide (Contessi and De Pace, 2021, OECD, 2020, World Bank, 2020a, World Bank, 2020b). As Shaikh and Huynh (2021) cogently describe the unique pandemic effect, “the COVID19 is an uncontained epidemic; just wait and watch!” The disruptive effect of the pandemic shock and the international health crisis on financial markets is well documented in the recent strand of literature. Amid the pandemic spread, the stock market behavior worldwide has drawn considerable attention. The January-May 2020 market crash does not account for the bursting of the asset price bubble. Instead, the market crash occurs for the first time in economic history when fundamentals are strong and there is a panic selling. Baker et al. (2020) highlight that the unprecedented U.S. stock market volatility during the first quarter of 2020 surpasses the Great Depression, the Great Financial Crisis (GFC), and the Spanish Flu pandemic.

This section presents a brief review of recent literature that examines the implication of COVID-19 associated uncertainty and fear on the global equity market. The inimitable effect of the COVID-19 pandemic outbreak on the global financial system can be cauterized as a decline in liquidity, increase in volatility, lower return, and cross-market or cross-asset economic shock spillover (Al Guindy, 2021, Baker et al., 2020, Contessi and De Pace, 2021, Huynh et al., 2021, Paule-Vianez et al., 2021, Rubbaniy et al., 2021b, Shaikh and Huynh, 2021, Xu et al., 2021, Zaremba et al., 2020, Zhang et al., 2021, among others). Although the empirical research question remains focused on COVID-19 and stock market behavior, the approach to measuring COVID-19′s impact varies considerably. For example, existing literature documents operationalization of Global COVID-19 fear index (Al-Awadhi et al., 2020, Mazumder and Saha, 2021, Rubbaniy et al., 2021b, Salisu and Akanni, 2020, Salisu et al., 2020a, Zhang et al., 2021) measured from the daily death and confirmed cases, Feverish sentiment (Huynh et al., 2021), pandemic anxiety indexes (Yu, Xiao, & Liu, 2021), the COVID19+ positive sentiment index (Anastasiou et al., 2022), Equity Market Volatility Infectious Disease Tracker (EMV) (Al Rababa'a et al., 2021, Bai et al., 2021, Baker et al., 2020, Bouri et al., 2021, Salisu et al., 2020b), Pandemic intensity information search (Goel & Dash, 2021), Coronavirus-related news (Biktimirov et al., 2021, Sun et al., 2021), COVID-19 Twitter intensity (Al Guindy, 2021), investor attention or pandemic attention through GSVI (Costola et al., 2021, Smales, 2020, Smales, 2021, Tripathi and Pandey, 2021, Xu et al., 2021), and pandemic-induced fear sentiment or pandemic uncertainty (Chen et al., 2020, Chundakkadan and Nedumparambil, 2021, Liu et al., 2021, Lyócsa et al., 2020, Paule-Vianez et al., 2021, Su et al., 2021, Subramaniam and Chakraborty, 2021, Szczygielski et al., 2022; Vasileiou, 2021). Table 1 presents a brief overview of recent literature and its findings. In the contemporary strand of literature, pandemic-induced fear sentiment, pandemic uncertainty, and pandemic attention have been used interchangeably, although measured through a similar internet search volume intensity approach. Taken together, a large body of recent literature (Al-Awadhi et al., 2020, Baker et al., 2020, Costola et al., 2021, Chundakkadan and Nedumparambil, 2021, Shaikh and Huynh, 2021, Smales, 2021, Szczygielski et al., 2022, Yu et al., 2021; Vasileiou, 2021; Zhang et al., 2021 among others) has documented a significant negative (positive) effect of the COVID-19 pandemic on stock returns (market volatility).

Table 1.

Overview of recent literature: COVID-19 pandemic and financial markets.

Articles	COVID-19 Measure	Market	Approach	Findings
Szczygielski et al. (2022)	Pandemic uncertainty: Google search query	Equity market: 68 industries	Panel regression	The COVID-19 related uncertainty is associated with higher level of volatility
Anastasiou et al. (2022)	Positive sentiment index (COVID19+)	G20 stock markets	Panel-GARCH	COVID19+ affects investor sentiment by increasing stock return
Bai et al. (2021)	Infectious Disease Equity Market Volatility Tracker (EMV-ID)	The US, China, UK, and Japan	GARCH-MIDAS	Infectious disease pandemic (EMV-ID) impacts market volatility
Huynh et al. (2021)	Feverish sentiment: RavenPack database	G-20 countries equity market	TVP-VAR model	Investor sentiment positively (negatively) predicts the stock volatility (return)
Al Rababa'a et al. (2021)	Infectious diseases equity market volatility (EMV) index	Gold, oil, and US dollar	Quantile regression	The infectious diseases EMV index affects gold, oil, and US dollar returns.
Liu et al. (2021)	Fear sentiment: Baidu search volume Index	China stock market	Time series model	Fear sentiment exacerbates crash risk
Smales (2021)	Investor attention: Google search volume	G20 countries	EGARCH	Investor attention negatively influences global stock returns
Costola et al. (2021)	Public concern: Google search query	Equity market	Regression	Search query volumes predicts equity market returns.
Rubbaniy et al. (2021a)	Global COVID-19 fear index	Commodity markets	Wavelet analysis	Soft commodities show safe-haven behavior
Chundakkadan and Nedumparambil (2021)	Investor attention: Google search query	59 countries equity market	Panel data model	Pandemic and the resultant sentiment-driven trade increases volatility in the market.
Subramaniam and Chakraborty (2021)	COVID-19 fear index: Google search query	Equity market	Regression	Negative association between COVID-19 fear and stock returns
Wang et al. (2021)	Expected and unexpected investor attention	U.S. stock market	Regression	Unexpected attention to COVID-19 is noisy
Goel and Dash (2021)	Pandemic intensity: Google search query	53 countries equity market	Panel data model	Pandemic search intensity has a negative effect on equity market return
Tripathi and Pandey (2021)	Investor attention: Google search query	25 equity market	Panel data model	Information from non-systematic sources contributed to the price decline
Su et al. (2021)	Pandemic-induced fear (PIF) index: Google search query	China stock market	Regression	Investors' internet search behaviors heighten the fear of the pandemic
Mazumder and Saha (2021)	Fear index: reported cases and deaths	US Stock Market	Regression	IPO returns are sensitive to the fear of the pandemic
Sun et al. (2021)	Coronavirus-related news (CRNs)	Equity market	Event study	Covid-19 is negatively associated with investor sentiment
Yu et al. (2021)	Pandemic anxiety indexes	BRICS and G7 countries	Regression	Time varying relationship: stock returns and the epidemic anxiety indexes
Rubbaniy et al. (2021b)	Global COVID-19 fear index	Cryptocurrency	Wavelet analysis	Safe-haven properties of cryptocurrencies is not plausible
Zhang et al. (2021)	COVID-19 pandemic: death rate	36 countries equity market	Panel regression	Significant negative impact on economic sentiment
Paule-Vianez et al. (2021)	COVID-induced fear: Google search query	G7 countries bond market	Regression	COVID-induced fear was associated with an increase in country risk perception
Xu et al. (2021)	Pandemic attention: Baidu search Index	China stock market	Panel regression	The stock market response to firm-specific information is decelerated
Al Guindy (2021)	COVID-19 Twitter intensity	US Stock Market	Regression	Increase in Twitter COVID-19 intensity corresponds to negative market returns
Harjoto et al. (2021)	Pandemic transmission speed	78 countries equity market	Panel regression	Stock markets react negatively to the COVID-19 infection and death cases
Vasileiou (2021)	Coronavirus Fear Index (CFI): Google	US stock market	Regression	Market inefficiency exists during the pandemic spread.
Smales (2020)	Investor attention: Google search query	S&P500 composite index	Regression	Heightened investor attention negatively influenced US stock returns
Chen et al. (2020)	Fear sentiment: Google search query	Cryptocurrency	VAR models	Market volatility has been exacerbated by fear sentiment
Biktimirov et al. (2020)	Media sentiment/coverage intensity	Equity market	Regression	The sentiment scores are generally negative
Salisu et al. (2020a)	Global fear index (GFI)	Commodity markets	Panel data model	Positive relationship between commodity price and GFI.
Salisu et al. (2020b)	Equity Market Volatility Infectious Disease Tracker (EMV)	24 emerging market stocks	Panel regression	Emerging stock markets are more vulnerable than developed market stocks
Salisu and Akanni (2020)	Global COVID-19 fear index (GFI)	Equity markets	Panel regression	Inverse relationship between stock prices and the fear index
Lyócsa et al. (2020)	Fear of the coronavirus: search query	10 Global equity markets	Regression	Excess search volume predicts price variation
Bouri et al. (2021)	Equity Market Volatility Infectious Disease Tracker (EMV)	Energy market	EGARCH	EMV helps to predict the forecast accuracy.

However, three plausible issues constrain our understanding of the COVID-19 related uncertainty or fear impact on the stock market. First, the existing literature has considered pandemic search intensity or fear index to measure investor sentiment and ignored the presence of prevailing sentiment effect in the market. For instance, Sun et al., 2021, Zhang et al., 2021 document that pandemic-induced fear harms economic and investor sentiment. Moreover, the expected and unexpected segments of investor attention to COVID-19 have an asymmetric effect on the equity market, and incredible attention to COVID-19 is noisy (Wang et al., 2021). Drawing on the argument of Sun et al., 2021, Zhang et al., 2021, one may argue that pandemic-induced fear may have exuberated the pessimistic sentiment in the market. On the other hand, due to the prevailing pessimistic sentiment attributable to the market crash, the deterrent effect of pandemic fear or uncertainty is substantially negative. Second, the theoretical underpinning of the pandemic information search as a measure of pandemic fear or uncertainty (or equivalent risk perception) has a similar connotation to the investor attention-based sentiment measures (e.g., FEARS index, Da et al., 2015). Da et al., 2015, Burggraf et al., 2021, and Goel and Dash (2021) findings suggest that investor sentiment can be directly measured through the internet search volume information, i.e., Financial and Economic Attitudes Revealed by Search (FEARS) index. Smales (2021) suggests that rather than searching for information on potential stocks to buy, retail investors are searching for information to resolve uncertainty about household FEARS during the COVID-19 crisis. Behavioral finance literature provides two competing theories for investor attention pricing in the financial markets, i.e., limited attention bias (Kahneman, 1973) and the price pressure hypothesis (Barber and Odean, 2008). Given that the potential investment universe is vast, with a binding constraint of cognitive capability, investors selectively pay attention to a limited set of information (Smales, 2021). This argument postulates that when investor investment decisions focus on selective attention-grabbing information, it results in temporary price pressure (Da et al., 2015). Existing literature have highlighted that investors' emotion explains the variations in financial markets and determines the asset allocation strategy of investors (Al Guindy, 2021, Garcia, 2013, Sun et al., 2021, Tetlock, 2007, Wang et al., 2021).

Since prices react to new information only when investors pay attention to it, high levels of attention to pessimistic news may cause selling pressure and sudden price reactions due to investors' response to such news. The third is the time-varying relationship between pandemic search volume information (pandemic fear or uncertainty measure) and investor sentiment. Given the severity of the COVID-19 effect, one may assume that pessimistic sentiment is more prevalent in the market during the early spread of the pandemic. Still, attention to pandemic information is not a comprehensive view of the market sentiment. In the COVID-19 pandemic, fear, anxiety, and worries have significant psychological consequences combined with sociocultural, cognitive, and behavioral aspects (Arora, Jha, Alat, & Das, 2020). The persistent flow of COVID-19 spread information with increasing infection and death cases, perceived vulnerability to disease, widespread media coverage, and the resulting economic uncertainty affected many people’s mental well-being with increased intensity of anxiety (worry), stress, fear, and sadness (Arora et al., 2020, Aslam et al., 2020). All taken together have resulted in the psychosocial development process of “coronaphobia” (Asmundson & Taylor, 2020). Consequently, the fear associated with “coronaphobia” may generate a negative sentiment, reduce the performance of stock markets, and therefore, investors may resort to rebalancing their portfolio asset allocation (Aslam et al., 2020, Capelle-Blancard and Desroziers, 2020, Chundakkadan and Nedumparambil, 2021, Liu et al., 2021, Smales, 2020, Su et al., 2021; Vasileiou, 2021; Wang et al., 2021).

The abovementioned discussion highlights certain essential limitations in the existing literature, which fails to control the effect of investor sentiment while examining the implication of pandemic uncertainty in the stock market. The Covid-19 pandemic fear or uncertainty or pandemic attention measured through the internet search intensity may not be a complete source of investor sentiment characterization. Investor attention-related sentiment measures (like FEARS) and pandemic fear or uncertainty measures are no different. It is equally likely that both may exist in the market simultaneously and feed on each other during market turbulence. Therefore, ignoring the effect of FEARS sentiment in the model prediction of pandemic uncertainty on stock prices may not reveal an accurate picture. We offer answers to the following two research questions based on the preceding arguments. First, whether there exists a co-movements and lead-lag relationship between the FEARS sentiment and Pandemic uncertainty across the major world economies. Second, whether the Pandemic uncertainty impacts the four broad aspects of stock market behavior, i.e., return, volatility, liquidity, and illiquidity, after controlling the effect of FEARS sentiment. Fig. 1 provides a theoretical framework on the relationship between Pandemic uncertainty, investor FEARS sentiment, and stock market behavior concerning our two research questions.

Fig. 1.

Framework of pandemic uncertainty, investor sentiment, and stock market behaviour.

Our findings help to shed more insights on the COVID-19 impact on the financial market after controlling for the effect of investor sentiment. Furthermore, the available literature emphasizes either the stock market return or volatility. To the best of our knowledge, no known study focuses on the stock market's return, volatility, and (il)liquidity effect amid the pandemic spread. Our paper also brings novelty in the empirical design by using wavelet-based time–frequency analysis. The benefit of the wavelet methods (WTC) is their ability to unite both the time domain (e.g., consideration of changes in time-period of a variable at a specified frequency) and the frequency domain (e.g., change in frequency of a variable but not periods) aspect of time-series data (Al-Yahyaee et al., 2019, Dash and Maitra, 2019, Dash et al., 2019). This approach effectively examines the co-movements between financial variables over time and under different frequencies (high, medium, and low) (Al-Yahyaee et al., 2019, Crowley, 2007, Mensi et al., 2019, Rubbaniy et al., 2021a, Rubbaniy et al., 2021b; Sharif et al., 2021).

3. Data and variables

This study uses the daily data over a sample period of January 2020 till May 2020 (138 daily observations), when several European countries and the USA became epicenters of the COVID-19 pandemic. Similar sample selection criteria were also employed by Contessi and De Pace (2021), Salisu et al. (2020), and Smales (2021), among others.

3.1. Stock market data

Our sample focuses on a worldwide measure and G7 country specific variables (Canada, France, Germany, Italy, Japan, the UK, and the USA). Our sample countries were affected by the COVID-19 at different stages of pandemic spread. The rationale behind G7 country selection was based on their significance in global economic activities and financial markets. We measured G7 country and world market daily stock market activities using four proxies: stock return, time-varying volatility using generalized autoregressive conditional heteroskedastic (GARCH) processes, the Amihud (2002) illiquidity measure, and turnover as a liquidity measure. We considered the S&P TSX composite, CAC 40, DAX 30, FTSE Italia All-Share, NIKKEI, FTSE 100, Nasdaq 100, and S&P Global 100 benchmarking market index for Canada, France, Germany, Italy, Japan, the UK, the USA, and the world market, respectively. We collected our stock market data from the Thomson Reuters database.

3.2. The COVID-19 uncertainty

Following Da et al., 2015, Effenberger et al., 2020, we utilized the GSVI as a proxy for attention search3 . In order to create a Pandemic search intensity variable, we used two search terms, i.e., “Coronavirus” and “COVID-19”, by restricting our search geography to different countries around the world. The final Pandemic index is a simple average of the standardized search intensity of two search terms (Coronavirus and COVID-19). However, we do not make any country specific language translation for the words “Coronavirus” and “COVID-19” as they are considered as synonym of a global pandemic (Chundakkadan and Nedumparambil, 2021, Costola et al., 2021, Liu et al., 2021, Smales, 2021, Subramaniam and Chakraborty, 2021). Searching for information about a particular subject clearly shows that one is paying attention to that subject (Smales, 2021). Moreover, the World Health Organization (WHO) official announcement of disease outbreak news (30 Jan 2020) considers the novel coronavirus (COVID-19) outbreak as a public health emergency of international concern (PHEIC), irrespective of any global, regional, and country-level differentiation.

Existing literature highlight that stock market volatility amid the pandemic spread is subject to good or bad news and even responses to fake news and policy changes (Shaikh & Huynh, 2021). However, our measure of pandemic uncertainty is different from the news-based uncertainty measure (Sun et al., 2021) and follows the search volume-based measure of pandemic uncertainty (Chundakkadan and Nedumparambil, 2021, Costola et al., 2021, Liu et al., 2021, Su et al., 2021, Smales, 2021; Vasileiou, 2021).

3.3. Sentiment indicator

Our analysis employs sentiment indicator measured by the Financial and Economic Attitudes Revealed by Search (FEARS). To construct FEARS sentiment index, our approach closely follows Da et al. (2015). For better consistency, our search keywords remained unchanged irrespective of the country selection. Consistent with Gao, Ren, and Zhang (2020), our search terms closely follow the same primitive word list suggested by Da et al. (2015), however, with reference to a specific country our search terms have been translated into the country specific language (e.g., French, Italian, Japanese language for France, Italy, and Japan respectively). For our sample countries like the world market, Canada, the UK, and the USA we use the English language for the exact primitive word list suggested by Da et al., 2015, Goel and Dash, 2021, and Burggraf et al. (2021).

Along with internet search intensity-based FEARS sentiment measure, we also use the newly constructed Equity Market Volatility Infectious Disease Tracker (EMV) developed by Baker et al., 2019, Baker et al., 2020 as an alternative pandemic sentiment indicator. The EMV is a newspaper text analysis-based equity market volatility tracker that moves with the VIX and with the realized volatility of returns on the S&P 5004 . Similar approach has been used by Al Rababa'a et al., 2021, Bai et al., 2021, Bouri et al., 2021, Salisu et al., 2020b as the measure of pandemics fear.

Fig. 2 displays a consistent co-movement between the FEARS and Pandemic time series across all the panels, suggesting the impacts of the COVID-19 pandemic on the sentiment towards the FEARS sentiment index is pervasive. The time series co-movement of the FEARS and Pandemic search volume supports the notion that the pandemic has potential deterrent effects on economic activities. Besides, we also examine the pair-wise rolling window correlation (R_Corr.) for a 20-day window between FEARS sentiment and Pandemic indexes of each country in the sample. We observed a high degree of correlation between coronavirus-information searches and the FEARS sentiment indexes. The strength of the relationship is higher for the USA, UK, Canada, and Japan. We notice that the correlation has intensified to 50% at the end of March. In our sample, Italy shows a sharp jump in the correlation at the end of April. The results exhibit the contagion effects of a Pandemic on the FEARS sentiment.

Fig. 2.

Time series comovement between FEAR and pandemic uncertainty. Notes: The above figure shows time series comovement between the sentiment (FEARS), pandemic search intensity (Pandemic Uncertainty), and the rolling correlations (R_Corr.) between two series, i.e., FEARS sentiment and Pandemic. The Y-axis (secondary, right side) captures R_Corr. The Y-axis (primary, left side) captures the FEARS and Pandemic variable. We have taken 20 days as the window to estimate the correlations. Given the short sample of the data, we cannot afford to take a longer window; however, we find that at 20 days windows, the correlations estimates are stable and consistent. The sample period is from 02 January 2020 till 18 May 2020 (138 daily observations).

Table 2 presents the descriptive statistics of all the variables. We use Augmented Dickey–Fuller unit root tests (Panel B of Table 2) to check the stationarity of the data. Panel (A) of Table 2 suggests a negative mean return for all the G-7 markets during our sample period. This is consistent with the extreme market downfall in the global equity markets during the early spread of COVID-19. There is heighted market uncertainty measured by return volatility for markets like Canada, Germany, Italy, and the USA. The global stock markets continue to exhibit a high degree of volatility, with a cumulative loss of 12.35% between January and May 2020 and a more than $9 trillion loss since the outbreak of COVID-19 (Salisu et al., 2020). Our Pandemic sentiment measure found to be consistently negative across the G-7 markets. The prevailing high level of pessimism in the global market during the COVID-19 pandemic spread is consistent with the notion that due to heightened economic uncertainty amid pandemic spread (Baker et al., 2020), investors may prefer to pay more attention to search information concerning the COVID-19 crisis to resolve uncertainty (Salisu and Akanni, 2020, Smales, 2021, Vasileiou, 2021).

Table 2.

Descriptive statistics.

Panel (A) Descriptive Statistics
Countries	FEARS		Pandemic		Return		Volatility		Liquidity		Illiquidity
	Mean	Stdev	Mean	Stdev	Mean	Stdev	Mean	Stdev	Mean	Stdev	Mean	Stdev
World	0.001	0.497	−0.007	2.906	−0.049	2.630	10.15	16.49	17.26	0.40	0.001	0.018
Canada	0.005	0.302	−0.001	0.986	−0.113	3.215	22.65	56.75	24.15	0.37	0.486	0.756
France	0.001	0.279	−0.001	0.989	−0.274	2.789	13.18	20.10	23.15	0.46	0.196	0.391
Germany	−0.001	0.256	−0.001	0.989	−0.164	2.821	13.26	19.43	23.20	0.44	0.153	0.212
Italy	−0.001	0.278	−0.001	0.986	−0.286	2.952	15.47	32.84	25.05	0.37	0.231	0.393
Japan	−0.001	0.278	−0.003	0.983	−0.140	2.193	5.08	4.72	25.17	0.37	0.900	1.976
UK	0.001	0.328	−0.001	0.987	−0.211	2.536	9.54	14.66	25.31	0.44	1.656	3.580
USA	−0.001	0.387	−0.001	0.989	0.062	3.246	15.06	25.17	23.84	0.34	1.112	2.053
Panel (B) Augmented Dickey–Fuller test (ADF) Unit Root Tests
Countries	ADF	P-value	ADF	P-value	ADF	P-value	ADF	P-value	ADF	P-value	ADF	P-value
World	−12.53	{0.00}	−2.64	{0.08}	−13.31	{0.00}	−2.57	{0.09}	−4.30	{0.00}	−11.10	{0.00}
Canada	−13.69	{0.00}	−3.87	{0.02}	−13.07	{0.00}	−5.52	{0.00}	−3.47	{0.00}	−4.73	{0.00}
France	−18.58	{0.00}	−3.87	{0.02}	−9.51	{0.00}	−3.92	{0.00}	−2.72	{0.00}	−4.57	{0.00}
Germany	−17.35	{0.00}	−3.87	{0.02}	−9.26	{0.00}	−4.21	{0.01}	−2.84	{0.00}	−4.78	{0.00}
Italy	−19.06	{0.00}	−3.80	{0.00}	−10.69	{0.00}	−3.84	{0.01}	−3.45	{0.00}	−5.39	{0.00}
Japan	−14.81	{0.00}	−3.86	{0.00}	−8.12	{0.00}	−1.98	{0.09}	−2.44	{0.00}	−5.26	{0.00}
UK	−14.18	{0.00}	−3.87	{0.00}	−9.64	{0.00}	−4.30	{0.01}	−3.11	{0.00}	−8.68	{0.00}
USA	−13.78	{0.00}	−3.87	{0.00}	−15.97	{0.00}	−2.45	{0.09}	−2.47	{0.00}	−7.18	{0.00}

Notes: This table reports the descriptive statistics and unit root test results of all variables for the World stock market and the G7 countries. Pandemic is the Google search intensity index of COVID-19 pandemic. FEARS is the sentiment index constructed using the Google search volume intensity index (Da et al., 2015). We consider SandP TSX composite, CAC 40, DAX 30, FTSE Italia All Share, NIKKEI, FTSE 100, Nasdaq 100, and SandP Global 100 as benchmarking market index for the Canada, France, Germany, Italy, Japan, UK, USA, and the World market. Volatility is measured through the GARCH estimation, Liquidity is measured as turnover, and Illiquidity is measured through the Amihud (2002) price impact measure. The sample period is from 02 January 2020 till 18 May 2020 (138 daily observations).

4. Empirical approach

In this paper, we use an empirical framework based on a wavelet approach. We investigate co-movement between the FEARS sentiment and Pandemic sentiment measures using the Wavelet Coherence (WTC) Analysis. To conduct our empirical analysis, we considered three wavelet methods, namely Wavelet Coherence Analysis (WCA), The Wavelet Phase Angle (WPA), and Partial Wavelet Coherence Analysis (PWCA). The WTC and the wavelet phase angle offer the benefits of testing the multi-time horizon co-movement of two indexes. The wavelet approach allows us to study the co-movements between sentiment and stock market behavior over time and under high, medium, and low frequencies (Crowley, 2007, Ramsey, 2002) and could be helpful for both short and long-term investors (Aguiar-Conraria et al., 2008, Mensi et al., 2020 Dash and Maitra, 2019, Mensi et al., 2019). Since, managing portfolio risk requires a strategic approach encompassing the time and frequency domain relationship between economic or financial indicators (e.g., sentiment and volatility of the market) it is useful for designing appropriate investment strategies. In this section, we briefly discuss the wavelet approach adopted to address our research questions.

4.1. Wavelet coherence analysis (WCA)

Wavelet coherence measures the strength of association between two time series. It gives a local correlation over time as well as across frequencies. In order to estimate wavelet coherence, each time series has to be transformed into a continuous wavelet. The continuous wavelet transform is defined as the convolution of the $p_t$ is expressed as:

$$W_p(\tau ,s)={\int }_{-\infty }^{+\infty }p(t)\frac{1}{\sqrt{s}}\psi \overline{\frac{(t-\tau )}{s}}dt$$ (1)

where τ and $s$ is the time position controlling its location and scale controlling the width of the wavelet, respectively. $1/\sqrt{s}$ normalizes the wavelet transforms to make them comparable across scales and time series. A measurement of the amplitude of a specific time series or variance of the time series at each time and each scale can be achieved by following the wavelet power spectrum (WPS)$\mathit{WP}{S}_p(\tau ,s)={\left|\mathit{Wp}(\tau ,s)\right|}^2$.

Similar to time series $p_t$, $q_t$ can also be transformed into a continuous wavelet.

In the next step, wavelet coherence is calculated by taking the ratio of the cross-spectrum to the product of the spectrum of two signals or time series. The wavelet coherence of two time series p_t and q_t is given as:

$$R_{\mathit{pq}}^2(\tau ,s)=\frac{{\left|S(s^{-1}{W}_{\mathit{pq}}(\tau ,s)\right|}^2}{S(s^{-1}{\left|W_p(\tau ,s)\right|}^2)S(s^{-1}{\left|W_q(\tau ,s)\right|}^2)}$$ (2)

where S represents smoothing operator both over time and scale, with 0 ≤ R²(τ,s) ≤ 1. A high (low) R² value suggests higher (lower) co-movement between two series. The co-movement between two time series over time and frequencies or scales is denoted by the contour plot. The level of co-movement is determined by the contour plot color. A yellow (blue) color denotes strong (weak) co-movement. The wavelet in time is not completely localized, hence contains edge effects. The edge effects are taken care of by a cone of influence (COI). The COI demarcates an area within which the relationship is distorted. As shown by Torrence and Compo (1998), the wavelet power significance testing can be done by judging against the null hypothesis stating that the data-generating process is given by an AR(0) or AR(1) stationary process. However, more general processes demand Monte Carlo simulations.

4.2. The wavelet phase angle (WPA)

The phase difference indicates delays of the oscillations of two time series as a function of time and scale. The wavelet phase difference can be estimated as:

$${\phi }_{\mathit{pq}}(\tau ,s)={\tan }^{-1}\left(\frac{\mathfrak{I}(W_{\mathit{pq}}(\tau ,s)}{\mathfrak{R}(W_{\mathit{pq}}(\tau ,s)}\right)$$ (3)

where $\mathfrak{I}$ and $\mathfrak{R}$ is an imaginary and real operator, respectively. The phase relationship between two time series is given by the wavelet phase angle. The circular phase is calculated to quantify the phase relationship, with statistical significance greater than 5% plus within the COI. Black arrows refer to phase differences. A phase difference of zero, denoted by arrows pointed to the right, suggests that time-series co-move at the specified frequency. In contrast, arrows directed to the left signify that the two series are negatively correlated and anti-phase. If arrows are directed towards [0,п/2], then series are in phase, and y leads x (if x and y are the first and second time series in order); if arrows are pointed towards [-п/2,0], then x leads y negatively. If arrows show the direction towards [п/2, п], then x leads y positively; arrows towards the direction of [-п,- п/2] show that y leads x negatively. Monte Carlo simulations are used to test the statistical significance of coherence.

4.3. Partial wavelet coherence analysis (PWCA)

The PWCA enables the user to consider three time-series. Let us denote the three time-series as: $X_n$, ${Y1}_n$ and ${Y2}_n$, and let their wavelet transformation be: ${\mathit{WL}}_n^X$(Š), ${\mathit{WL}}_n^{Y1}$(Š) and ${\mathit{WL}}_n^{Y2}$(Š) respectively. Thus, the squared PWC may be defined as:

$$\begin{array}{c}{({\mathit{RcP}}_n^{XY1\_Y2})}^2=\frac{{\left|{\mathit{Rc}}_n^{XY1}-{\mathit{Rc}}_n^{XY2}.R_n^{Y2Y1}\right|}^2}{(1-{({\mathit{Rc}}_n^{YX2})}^2)(1-{({\mathit{Rc}}_n^{X2X1})}^2)};\\ {({\mathit{RcP}}_n^{XY2\_Y1})}^2=\frac{{\left|{\mathit{Rc}}_n^{XY2}-{\mathit{Rc}}_n^{XY1}.R_n^{Y2Y1}\right|}^2}{(1-{({\mathit{Rc}}_n^{YX1})}^2)(1-{({\mathit{Rc}}_n^{X2X1})}^2)}.\end{array}$$ (3)

5. Results discussion

5.1. Time-frequency co-movement between FEARS sentiment and pandemic uncertainty

Fig. 3 examines the co-movements between the FEARS sentiment and Pandemic uncertainty by employing wavelet coherence and phase difference approaches. The strength of coherence ranges from yellow (strong) to blue (weak). The horizontal (vertical) axis shows time (frequency). The frequency numbers denote the number of days5 . The black contours in the figure suggest significant coherent regions at a 5 percent level of significance. The outer solid line of the cone indicates the cone of influence. We notice that the COVID-19 pandemic has a high impact on the FEARS sentiment starting from 21 January until 30 April across all frequency bands at the world level. Our evidence also coincides with two significant events around that time and of important implications for the world economy. First, on 21 January, the first travel-related case in the USA was confirmed in Washington state. Second, the WHO experts conducted a brief field visit to Wuhan, China, on 21 January 2020; on 22 January 2020, preliminary evidence of human-to-human transmission in Wuhan, China, was confirmed by the WHO. Finally, on 30 January 2020, WHO Director-General recommended that the COVID-19 outbreak constituted an International Public Health Emergency.

Fig. 3.

Wavelet coherency: pandemic uncertainty and FEARS sentiment. Notes: This figure presents the wavelet coherency test results between Pandemic Uncertainty and FEARS sentiment for G7 countries and the world market. The horizontal and vertical axis represents time in months and the period, respectively. The colors range from deep blue (low coherence/no correlation) to deep yellow (high coherence/perfect correlation). The area within the black contours indicates coherency significance at 5% level. If arrows pointed toward the right (left) indicates both Pandemic and FEARS are in-phase (out-of-phase).

Countries such as the US, UK, Canada, and Japan show higher Pandemic impacts on FEAR sentiment. Effenberger et al., 2020, Sun et al., 2021 confirmed that the worldwide public interest COVID-19 reached its first peak at the end of January when the numbers of newly infected patients started to increase exponentially in China. The strong coherence across all bands indicates the contagion effects between pandemic and market sentiment; however, more meaningful coherence is present at the scale between 16 and 32 days, i.e., the medium term. The magnitude of co-movements increased after 01 March, which overlaps with the periods: i) beginning of health crisis in most countries worldwide; ii) implementation of strict rules regarding social distancing and quarantine; and iii) moderate to intense economic lockdown across countries. Besides, France, Germany, and Italy also show coherence between the pandemic and FEAR sentiment. These three countries reveal the significant contagion effects of the pandemic from 21 March to 10 April at the frequency band of 4–8 days.

Furthermore, we examined the lead-lag relationship between the Pandemic uncertainty and FEARS sentiment indexes. In Fig. 3, the arrows pointing to the right ($\to$) and left ($\leftarrow )$ indicate that the first and second series are in-phase and out-of-phase, respectively. The in-phase or out-of-phase relationship shows whether the two are positively or negatively co-moving. The upwards (downwards) [$\uparrow$($\downarrow$)] arrows indicate the FEARS is leading (lagging). The right up ($\nearrow )$ and left down ($\swarrow )$ arrows suggest the FEARS sentiment is lagging. We notice that both the FEARS sentiment and Pandemic uncertainty are positively correlated as the arrows are directed to the right. The FEARS sentiment is led by pandemic conditions of the country and world. In a nutshell, we observe a strong positive relationship between the FEARS sentiment and Pandemic due to COVID-19. The world majors like the US, UK, Canada, and Japan show stronger co-movements between the two series. The strong coherence intensified after 01 March, when the spread of COVID-19 has substantially increased across the world.6

Effenberger et al. (2020) also supports the view that worldwide interest peaked in mid-March 2020, shortly after COVID-19 was declared a pandemic. Consistent with our observation, Effenberger et al. (2020) note that the highest search volume index for the pandemic peaked during mid of March, which is in line with rigorous policies by the government regarding the rapid spread. The results corroborate the findings of Ramelli and Wagner (2020) that the extreme event like COVID-19 is likely to have a significant effect on economic uncertainty, and thus on the investor sentiment. Therefore, the impacts stemming from economic uncertainty, sentiment, and financial markets are difficult to ascertain (Rush, 2020, Stirling et al., 2020). The increase in economic uncertainty may be due to a fall in global trade, stock price reactions to COVID-19, the decline of credit lines, and bond-floating activities during the pandemic crisis (World Bank, 2020b, Baker et al., 2020, Huynh et al., 2021, OECD, 2020, Shaikh and Huynh, 2021, World Bank, 2020a). Our results also support the arguments of Bloom et al., 2018, Fan et al., 2018 toward the deterrent impact of pandemic events on increasing economic and policy uncertainty. Overall, our results reveal that the COVID-19 pandemic represents a menacing risk that is stirring up feverish behavior in investors worldwide (Capelle-Blancard and Desroziers, 2020, Ramelli and Wagner, 2020, Subramaniam and Chakraborty, 2021, Wang et al., 2021, Zhang et al., 2021). The potential impact of pandemic search intensity on investor sentiment could be due to the availability cascade effect observed by Kahneman (2011). An availability cascade effect (Kahneman, 2011) is a self-sustaining chain of events that could be due to the emotional reaction by investors due to the excessive media coverage of pandemic spread (Capelle-Blancard and Desroziers, 2020, Huynh et al., 2021, Krugman, 2020, Shiller and Malkiel, 2020, Sun et al., 2021, Tripathi and Pandey, 2021). Similarly, analyses of sentiments and emotions evoked by news headlines regarding the coronavirus disease outbreak revealed that the news headlines had high emotional scores with negative polarity (Aslam et al., 2020).

5.2. Partial wavelet coherency analysis

Fig. 4 reveals the PWCA between stock market variables (return, volatility, liquidity, and illiquidity) of G7 countries and Pandemic uncertainty after controlling the FEARS sentiment. Since Pandemic uncertainty and FEARS sentiment can have significant correlations due to heightened concern of economic uncertainty, any effects of the FEARS sentiment must be controlled to examine the co-movements between stock market variables and Pandemic intensity. In a more illustrative way, phycologists described that “when it comes to making decisions that involve risks, we humans can be irrational in quite systematic ways. When we see actual death tolls climbing daily, as we do with the coronavirus-another factor besides our sensitivity to losses comes into play: fear.”7 As previously mentioned, PWCA squared helps to detect the correlation between two time series, y(t) and x₁(t), after controlling for the effect of x₂(t).

Fig. 4.

Partial wavelet: pandemic uncertainty, FEARS sentiment, and stock market behaviour. Notes: The yellow area within the patches surrounded by black line indicates the high power significant at 5% level of significance. Mone-Carlo simulation is employed to find whether the patches are significant. We present the color bar on the right side of each figure. The color blue signifies low power and blue, reddish yellow, and yellow indicates high, higher, and highest power, respectively. The greater the density of the colors is, the higher the power of the wavelet. The color bar also indicates the degree of correlations, which range between 0.5 (50%) and 0.9 (90%). Wavelet squared coherency detects the different investment horizons, i.e. short, medium, and long-term across the sample period. The X-axis presents the timeline for each stock market variable and pandemic pair whereas the Y-axis measures scale or frequency.

For example, in Fig. 4, columns 1, 2, 3, and 4 represent the co-movements of Pandemic uncertainty with stock returns, volatility, liquidity, and illiquidity, respectively, after controlling the effect of FEAR sentiment. Fig. 4 reveals that the pandemic co-movements with the stock market returns of the world and all G7 markets. Except for the UK, we notice a strong co-movement for all other stock markets. We documented that the relationship between pandemic and stock returns (after controlling for FERAS sentiment) shows signs of synchronization across countries. Moreover, our results indicate a strong co-movement at the lower frequency scale of 8–16 days. However, the partial coherence peaked between February to March 2020 in Canada, France, Italy, and Japan. During peak periods, the coherence extends to higher frequencies (4–8 days) for Canada, Italy, Japan, and the world.

Our results for the Pandemic intensity and stock return behavior are consistent with the findings of Al-Awadhi et al., 2020, Ramelli and Wagner, 2020, and Zhang et al. (2020). A similar co-movement pattern can be observed between pandemic and stock return volatility for France, Italy, and Japan. Interestingly, unlike stock returns, the volatility co-movement with Pandemic intensity is not strongly apparent. Our results are consistent with the notion that the prolongation of the coronavirus pandemic and associated stringent policy responses are important sources of financial market volatility (Zaremba et al., 2020). In Fig. 4, the third and fourth columns show that liquidity and illiquidity were affected due to the world pandemic. Countries that strongly display the effects of pandemic fear on (il)liquidity include the UK, Canada, France, Italy, and Japan. Very short-term (il)liquidity, up to a week, was influenced by Pandemic intensities in the UK, France, Germany, Italy, and Japan. However, in Canada, France, and Japan, the (il)liquidity shows strong concern due to the pandemic for the frequency of 8–16 days, equivalent to a period between a week to a fortnight. Our results are consistent with Dash et al. (2019) findings, which suggest that increasing concern over policy uncertainty could deteriorate market liquidity for countries such as Canada, France, Germany, and the USA.

5.3. Robustness tests

In this section we carry out additional robustness test using the Equity Market Volatility Infectious Disease Tracker (EMV) (Baker et al., 2019, Baker et al., 2020) as an additional pandemic fear sentiment indicator. Since VIX or volatility index is popularly used in the behavioral asset pricing tests as a proxy for the pessimistic sentiment, consistent with the argument of Al Rababa'a et al., 2021, Bai et al., 2021, Bouri et al., 2021, and Salisu et al. (2020b) we resort to the operationalization of EMV as an alternative measure of pandemic fear sentiment. The robustness test focusses on the partial wavelet coherence test between pandemic uncertainty and stock market activity of our sample countries after controlling the effects of EMV pandemic fear sentiment index. The reason for the choice of alternative pandemic fear sentiment measure follows the argument that in times of crisis, FEARS sentiment index might not be very effective in catheterizing investor sentiment as pandemic effect will be more dominant8 . Since Pandemic uncertainty and FEARS sentiment may feed each other during the market turmoil and heightened economic concern amid the pandemic spread, it is reasonable to reexamine our previous findings in the presence of alternative pandemic fear sentiment.

Fig. 5 presents the PWCA between G7 countries stock market return, volatility, and (il)liquidity, and Pandemic uncertainty after controlling the effect of EMV or pandemic fear sentiment. Fig. 5 shows that the pandemic co-movements with the stock market returns of the world and all G7 markets after controlling the effects equity market volatility due to infectious disease (EMV). Our results indicate a strong co-movement at the lower frequency scale of 8–16 days. Similar to Fig. 4, the partial coherence peaked between February to March 2020 in Canada, France, Italy, and UK. We do not notice a significant co-movement between stock market volatility and pandemic uncertainty; however, illiquidity shows relatively higher coherency with pandemic uncertainty compared to liquidity. Taken together, our results are robust to the inclusion of alternative pandemic sentiment measure as EMV and co-movements between Pandemic uncertainty and stock market variables (return, volatility, and (il)liquidity) is consistently visible.

Fig. 5.

Pandemic uncertainty and stock market behaviour: controlling the effect of Pandemic fear sentiment (EMV). Notes: The yellow area within the patches surrounded by black line indicates the high power significant at 5% level of significance. Mone-Carlo simulation is employed to find whether the patches are significant. We present the color bar on the right side of each figure. The color blue signifies low power and blue, reddish yellow, and yellow indicates high, higher, and highest power, respectively. The greater the density of the colors is, the higher the power of the wavelet. The color bar also indicates the degree of correlations, which range between 0.5 (50%) and 0.9 (90%). Wavelet squared coherency detects the different investment horizons, i.e. short, medium, and long-term across the sample period. The X-axis presents the timeline for each stock market variable and pandemic pair whereas the Y-axis measures scale or frequency.

6. Conclusion

This paper uses daily data from the G7 countries and world equity markets to examine the pandemic uncertainty (measured by search volume index) implications for the stock market behavior. The stock market behavior has been analyzed from the return, volatility, and liquidity aspects. Using wavelet-based time–frequency analysis, this paper focusses on two important research questions. First, whether there exists a co-movements and lead-lag relationship between the FEARS sentiment and Pandemic uncertainty across the major world economies. Second, whether the Pandemic uncertainty impacts the four broad aspects of stock market behavior, i.e., return, volatility, liquidity, and illiquidity, after controlling the effect of FEARS sentiment. Our findings suggest that the rapid spread of the COVID-19 pandemic increased the pessimistic sentiment environment within the world’s major economies. We notice that the COVID-19 pandemic has a high impact on the FEARS sentiment during the early spread of the pandemic (January- March 2020). Results indicate that pandemic spread leads to pessimistic investor sentiment. The macroeconomic expectations hypothesis suggests that stock markets are forward‐looking in nature, and hence, market participants’ trading behaviors are directly related to expectations about future economic uncertainty. Our results reveal that, due to heightening economic uncertainty amid COVID-19’s spread (Baker et al., 2020, Huynh et al., 2021, OECD, 2020, Shaikh and Huynh, 2021, Sharif et al., 2020, Wang et al., 2021, World Bank, 2020a, World Bank, 2020b, Zhang et al., 2020), market participants may resort to the pessimistic considerations regarding future expected cash flow and discount rates, which affects stock market activity. Furthermore, the relationship between the pandemic uncertainty and stock market activities (after controlling for FERAS sentiment) shows signs of synchronization across G7 countries and the world market. The results are robust to alternative measures of pandemic fear sentiment.

Overall, this study highlights that there is a time-varying lead-lag relationship between pandemic uncertainty and the impact of pandemic uncertainty on stock market is persistent even after controlling the effect of investor’s FEARS sentiment. Our findings are having practical implications for the financial market participants and policy makers. Our findings highlight the importance of investor sentiment measures for the financial modeling and asset pricing purpose, and even in the short-horizon trading strategies. Markets are composed of investors operating in each moment at different time scales (short- and long-term). Our findings supported by time–frequency co-movement accommodate both the time and frequency domains within a unified framework to simultaneously assess how sensitive the stock market activity (return, volatility, and liquidity) to pandemic uncertainty at different time horizons. Moreover, pandemic uncertainty matters for the equity markets return, volatility, and trading activities. The pandemic uncertainty and associated sentiment implications could be one plausible reason for increased volatility and illiquidity in the market, and hence, policymakers should look upon this issue for the financial market stability perspective. After the initial spread of the pandemic, there are several policy interventions undertaken by the Government authorities. Given that our sample only accounts for the initial spread of the pandemic, future research could bring more insights on the pandemic uncertainty and stock market activity after the introduction of the policy measures and vaccination drive around the world.