COVID-19’s disasters are perilous than Global Financial Crisis: A rumor or fact?

Highlights

• •Financial markets of the world confronted an extreme collision in their market values due to COVID-19.
• •COVID-19 exposed countless sway on financial volatility of the US and European markets than GFC.
• •During the period of COVID-19, Asian markets are making available better opportunities to diversify the financial risk.
• •COVID-19 revealed a negative and significant association with returns of S&P 500 index.

Abstract

This investigation employed the Asymmetric Power GARCH model and found that COVID-19 substantially harms the US and Japan's market returns. Moreover, COVID-19 has influenced the variance of the US, Germany, and Italy's stock markets more than the Global Financial Crises (GFC). However, GFC indicated a more significant impact on the financial volatility of the Nikkei 225 index and SSEC than COVID-19. The study confirmed the leverage effect for the S&P 500, Nasdaq Composite Index, DAX 30, Nikkei 225, FTSE MIB, and SSEC. The analysis authenticated that the health crisis that befell due to COVID-19 have imperatively originated the financial crisis globally; however, the Asian markets still make available better prospects for portfolio optimization.

1. Introduction

In the past, World health organization (WHO) revealed various epidemics that influenced the vast number of people and economies around the world, e.g., Influenza, Ebola, and Severe Acute Respiratory Syndrome (SARS). These epidemics greatly affected the economy of the world; particularly, 5% gross domestic product (GDP) of the US was reduced owing to Influenza (Palese, 2004); while, the US encountered the economic deficit of $53 billion as result of EBOLA and it caused more than 11,300 deaths in the world (Hai et al., 2004). The SARS transpired from China in 2002 and affected more than 8000 people. It depreciated the 1% GDP of China and instigated the economic damage of $54 billion worldwide (Peiris et al., 2003). Recently, a new pandemic reported from Wuhan, China, known as a novel coronavirus (COVID-19), with a specific ability to transmit from one person to another, has spread nearly all over the world, and prominently distracted the lives of the people universally. On April 17, 2020, more than 1,865,413 people have been infected, and 110,008 deaths have been documented due to COVID-19 in the world (Financial Times, 2020). In order to stop the transmission of COVID-19, each nation has locked its business, markets, and people are forced to live in their houses. Subsequently, unemployment has been enriched; the supply side has been damaged, economic growth, tourism level, and sale of the traveling sector have been reduced (Leduc and Liu, 2020). Also, the stock markets of the world confronted an extreme collision in their market values. The market value of the Standard & Poor (S&P) 500 index dropped to 30% since the outbreak of COVID-19. Fig. 1 exhibited the diminution in the market value of the world's reputed stock index due to COVID-19. The verdicts reported that, during March, stock markets of Germany, France, and Italy encountered extreme deterioration in their market values.

Fig. 1.

Decline in stock market values due to COVID-19

Data source: Statista.com.

Moreover, the US, UK, Spain, Hong Kong, and China's stock markets run into a decline of 14.9%, 21.4%, 25.1%, 14.7%, and 12.1% in their prices since March 08 – March 18, respectively. Notably, the market value of the MSCI World Index faced 17.5% dropped from March 06 - March 18. Adam (2020) reported that financial markets are near to collapse as they were during the financial crises of 2008. Georgieva (2020) argued that COVID-19 brought the earth near to financial crises more perilous than Global Financial Crises (GFC) 2007-2009. Additionally, the financial volatility index (VIX), also known as "Fear gauge," has moved to the uppermost level, higher than the GFC era, while the US 10 year treasury yield index moved down to record low level (Leduc and Liu, 2020). The Asian Development Bank evaluated that the global cost of COVID-19 can be $4.1 trillion (ADB, 2020). Hence, this information intimated the importance of exploring the impacts of COVID-19 on reputed stock markets of the globe and resembled its impact on disasters of GFC. This investigation sought the impact of COVID-19 and GFC on the returns and variance of stock markets of the US, Germany, Italy, China, and Japan; and compared the parameters of both periods.

Also, the study analyzed the role of China and the US trade war on their stock market returns and variance. The investigation also scrutinized the influence of news and volatility persistence by reason of COVID-19 in these markets. This project answers the superior queries of the policymakers, investors, financial analysts, portfolio managers, and academicians. For instance, first, does the COVID-19 generated a more significant crash in stock markets than GFC? Second, does COVID-19 has a higher impact on stock returns than GFC? Third, does the volatility that ensued in stocks as a result of COVID-19 are higher than GFC? Fourth, do markets have a significant leverage effect? Fifth, do shocks befell in these markets continued for the long term period? Sixth, what tactics can be used to stabilize the market volatility? The author deems that answers to these queries will clear the role of COVID-19 and GFC for financial markets. As per the author's best information, no study has been conducted to analyze and compare the imperious role of COVID-19 and GFC on stock market returns and variance.

First, this study engaged the Augmented Dickey-Fuller test, and Phillip Perron test to examine the unit root level of study variables and then Autoregressive Conditional Heteroskedasticity-Lagrange Multiplier (ARCH-LM) test is employed to ascertain the ARCH effect in the data. The findings disclosed that all the variables are stationary at I(0), and a significant ARCH effect exists. Moreover, the Jarque-Bera test revealed that these market's returns are not normally distributed. Consequently, Asymmetric Power Generalized Autoregressive Conditional Heteroskedasticity (APGARCH) model is appropriate for this study. The rest of this project is outlined as follows. Chapter 2 reviews the methodological section, chapter 3 debates on the findings of this project, chapter 4 epitomizes the conclusion, and chapter 5 provides the references applied in this project.

2. Methodological path

2.1. Data

This investigation utilized the day-wise data from the period of June 30, 2007, to April 07, 2020. The day-wise prices of the US (S&P 500, Nasdaq Composite index), Germany (DAX 30), Italy (FTSE MIB), Japan (Nikkei 225), and Chinese (SSEC) stock markets are salvaged from the database of Yahoo Finance. Further, this investigation used the dummy variables (0, 1) to ascertain the impact of GFC and COVID-19. Likewise, Choudhry, 2010 and Kiymaz and Berument (2003) utilized the dummy variables to analyze the impact of daily and yearly anomalies on stock returns.

2.2. Methodology

The study has calculated the daily returns of S&P 500 (SPR), Nasdaq Composite Index (NSR), DAX 30 (DXR), FTSE MIB (FMBR), Nikkei 225 Index (NKR), and SSEC (SCR) by following the method of Shehzad and Sohail (2018) as follows,

$$R_t=\left(ln\left(C{S}_t/C{S}_{t-1}\right)\right)*100$$ (1)

here R_t, CS_t, and CS_t-1 signify the projected day-wise returns, the closing price of a stock at time t, and closing value of a stock on the previous day, while ln symbolizes the natural log. In order to capture the non-linear influence of COVID-19 and GFC in the daily returns and variance of financial markets, this investigation employed Asymmetric Power GARCH model introduced by Ding et al. (1993). Moreover, this analysis utilized the student-t distribution, as it can handle excess kurtosis (Bollerslev, 1987). The mean and variance equation of APGARCH model can be defined as (Ding et al., 1993),

$${\lambda }_{ti}={\xi }_0+{\xi }_{1}GF{C}_{ti}+{\xi }_{2}COVI{D}_{ti}+{\epsilon }_{ti}$$ (2)

$${\Psi }_{ti}^v={\vartheta }_0+{\varphi }_1\text{GF}{\mathrm{C}}_{ti}+{\varphi }_2\text{COVI}{\mathrm{D}}_{ti}+\sum _{i=1}^p{({\alpha }_i\left|z_{t-1}\right|+{\chi }_i{z}_{t-1})}^v+\sum _{i=1}^q{\beta }_i{\Psi }_{ti-1}^v$$ (3)

where,

$$Z=\frac{{\epsilon }_{ti}}{{\varpi }_{ti-1}}\sim t.d.\left(0,{\Psi }_{ti},\Gamma \right)$$ (4)

here in the mean equation Eq. (2), λ_ti and ε_ti denotes the returns and error term of stock market i, respectively. The error term is founded on historical information (ϖ_ti − 1) and presumed that it follows the student-t density (t.d.), with Γ degree of freedom. Moreover, ξ₀, ξ₁, and ξ₂ indicate the constant term, dummy variable of GFC, and COVID-19. Likewise, in the variance equation Eq. (3), β_i directs the variations in its own variance series, i.e., GARCH effect and v denote the power term coefficient. Moreover, α_i and χ_i evaluate the effects of return changes on its own series, i.e., ARCH effect and asymmetric impact of series i, i.e., leverage effect, respectively. A positive χ_i indicates that negative news has a greater impact than positive news on market volatility (Ding, 2011). Besides, φ₁ and φ₂ determine the effect of GFC and COVID-19 on variance series i, respectively. By considering the methodology of (Ding et al., 1993; Hentschel, 1995), various standard ARCH and GARCH models can be nested in APGARCH model by stipulating the permitted values for α_i, χ_i, β_i, and v (Teräsvirta, 2012). Table 1 defines the inherent limitations to produce the various ARCH and GARH models nested in APGARCH model. This study assigned the number 1 to the days of GFC, i.e., June 30, 2007- December 30, 2009, and 0 otherwise. While to evaluate the sway of COVID-19, the number 1 is assigned from January 01, 2020 – April 07, 2020, otherwise 0. This examination does not consider the dummy variable of the normal period to avoid the dummy trap. Further, the investigation include the dummy variable of the trade war between China and the US for the stock markets of China and the US as follows,

$${\lambda }_{ti}={\xi }_0+{\xi }_{1}GF{C}_{ti}+{\xi }_{2}COVI{D}_{ti}+{\xi }_{3}CWA{R}_{ti}+{\epsilon }_{ti}$$ (5)

$${\Psi }_{ti}^v={\vartheta }_0+{\varphi }_1\text{GF}{\mathrm{C}}_{ti}+{\varphi }_2\text{COVI}{\mathrm{D}}_{ti}+{\varphi }_3\text{CWA}{\mathrm{R}}_{ti}+\sum _{i=1}^p{({\alpha }_i\left|z_{t-1}\right|+{\chi }_i{z}_{t-1})}^v+\sum _{i=1}^q{\beta }_i{\Psi }_{ti-1}^v$$ (6)

where ξ₃ and φ₃ denote the impact of China and the US trade war in mean (Eq. (5)) and variance (Eq. (6)) equation, respectively.

Table 1.

Nested ARCH and GARCH model limitations.

Approach	Αi	χi	βi	v
ARCH	independent	0	0	2
GARCH	independent	0	independent	2
GJR GARCH	αi(1+χi)2	4αiχi	0	2
GJR ARCH	αi(1+χi)2	4αiχi	independent	2
TARCH	independent	|χ|≤1	0	1
Taylor ARCH	independent	0	0	1
Taylor GARCH	independent	0	independent	1
NARCH	independent	0	0	independent
Generalized TARCH	independent	|χ|≤1	independent	1
Asymmetric ARCH	independent	|χ|≤1	0	2
Asymmetric GARCH	independent	|χ|≤1	independent	2
PGARCH	independent	0	independent	independent
APGARCH	independent	|χ|≤1	independent	independent

Source: author's calculation.

3. Results and discussions

The summary statistics of study variables given in Table 2 informed that all the stock markets have negative mean returns during the COVID-19 era. In addition, all the markets exposed negative skewness with high kurtosis values, inferring that chances of loss are in height. Table 3 stated the results of the Augmented Dickey-Fuller (ADF) test Dickey et al., 1979, Phillip Perron (PP) test (Perron, 1988), and Autoregressive Conditional Heteroskedasticity (ARCH) test (Engle, 1982). The findings described that all the return series are stationary at I(0), and the significant ARCH effect is present. Hence, the APGARCH model perfectly fits the study.

Table 2.

Summary statistics of stock returns.

	NKR	FMBR	SCR	NSR	SPR	DXR
Mean	-0.313724	-0.402455	-0.147617	-0.215405	-0.147623	-0.51546
Std. Dev.	2.445463	3.453028	1.769321	3.626117	2.977544	2.925632
Skewness	0.505023	-2.31595	-1.558152	-0.510258	-0.556915	-0.875228
Kurtosis	4.942365	14.04434	8.030458	5.449286	7.903214	9.651646
Jarque-Bera	12.78127	424.3192	90.4602	19.36125	106.3956	130.0984
Probability	0.001677	0	0	0.000062	0	0

Source: author's calculation.

Table 3.

Unit root and Heteroskedasticity test

	ADF		PP		ARCH-LM
Variable	Level	1st Diff.	Level	1st Diff.	F-statistics
SPR	-66.03021***	-22.63017***	-66.25453***	-828.5502***	426.7328***
NSR	-64.45909***	-64.63904***	-65.08732***	-816.2465***	466.3889***
DXR	-56.48601***	-22.39084***	-56.48463***	-785.2905***	48.2608***
NKR	44.8633***	21.9552***	-53.6732***	-721.3891***	329.7114***
FMBR	-59.11521***	-26.05068***	-59.1209***	-810.3178***	80.99362***
SCR	-54.81653***	-23.65203***	-54.84421***	-521.7985***	109.9907***

Note: *, **, and *** denotes the 10%, 5%, and 1% level of significance, respectively. Source: author's calculation.

3.1. Mean model reckoning

The aftermaths of the mean equation for each returns series, specified in Table 4 , discovered that GFC (ξ₁) and COVID (ξ₂) have a negative sway on SPR. However, the coefficient of GFC (ξ₁) remains inconsequential. Moreover, CWAR indicated a positive but insignificant impact on SPR. The mean equation of NSR publicized that GFC (ξ₁) has a negative and substantial relationship with NSR. Nonetheless, COVID (ξ₂) and CWAR (ξ₃) demonstrated a positive but insignificant influence on NSR. The outcomes of DXR remain extraneous, but NKR quantified that GFC (ξ₁) and COVID (ξ₂) have negative control on NKR, but the effect of COVID (ξ₂) is insignificant. Further, GFC (ξ₁) disclosed a destructive and momentous effect on FMBR, whereas the impact of COVID (ξ₂) on FMBR is noted to be positive and irrelevant. The return series of SCR designated that GFC (ξ₁) has a positive, but CWAR (ξ₃) has an adverse reaction on it; however, COVID-19 divulged positive and minor control.

Table 4.

Results of mean equation.

Variable	Coefficient	Std. Err	Z-Statistics	Prob.
SPR
ξ1	-0.074166	0.051206	-1.448404	0.1475
ξ2	-0.033553	0.03892	8.625601	0.0000
ξ3	0.024108	0.031249	0.771508	0.4404
ξ0	0.046116	0.013268	3.475718	0.0005
NSR
ξ1	-0.07019	0.057106	-2.91165	0.0359
ξ2	0.068919	0.049588	1.389834	0.1646
ξ3	0.029659	0.059823	0.49578	0.62
ξ0	0.071684	0.016054	4.465119	00000
DXR
ξ1	-0.074194	0.053305	-1.391876	0.164
ξ2	0.003962	0.05155	0.076855	0.9387
ξ0	0.043952	0.017596	2.497806	0.0125
NKR
ξ1	-0.105667	0.059912	-1.763687	0.0778
ξ2	-0.046816	0.054285	-0.862421	0.3885
ξ0	0.052843	0.019775	2.672166	0.0075
FMBR
ξ1	-0.114119	0.057484	-1.985253	0.0471
ξ2	0.017097	0.129362	0.132167	0.8949
ξ0	0.017786	0.022873	0.777598	0.4368
SCR
ξ1	0.159696	0.081796	1.952361	0.0509
ξ2	0.088564	0.093508	9.471250	0.0000
ξ3	-0.031935	0.049069	-0.650818	0.5152
ξ0	0.034007	0.02037	1.669487	0.095

Source: author's calculation.

3.2. Variance equation fallouts

The inferences of the variance equation for each market are unveiled in Table 5 . It defines that all constant values (ϑ₀) are significant, implying that mean values are divergent from zero. The fluctuations in returns due to GFC (φ₁) and COVID (φ₂) imperatively affect the instability of SPR. However, COVID (φ₂) revealed a trifling effect on the stability of NSR. Also, the instability caused by CWAR (φ₃) indicated an inconsequential effect. The findings of DXR and NKR variance equation stated that uncertainty occurred due to GFC (φ₁) at time t, expressively upsurge the volatility of these markets at time t+1. Further, variability happened owing to COVID (φ₂) dramatically influence the variance of DXR, but it showed inconsequential sway on the variance of NKR. Likewise, COVID (φ₂) documented the feeble effect as compared to GFC (φ₁) for FMBR. The variance reckoning of SCR particularized that the coefficient of CWAR (φ₃) and GFC (φ₁) exposed a positive and insignificant, although the coefficient of COVID (φ₂) nominated positive and significant impact on the market variance. Fig. 2 confirmed that stock markets of the US, Germany, and Italy have high conditional variance because of COVID-19 as compared to GFC. On the other hand, the conditional variance of Nikkei 225 and SSEC is high during the GFC period.

Table 5.

Results of variance equation.

Variable	Coefficient	Std. Err	Z-Statistics	Prob.
SPR
ϑ0	0.037938	0.004409	8.604057	0
αi	0.138593	0.01163	11.91649	0
χi	0.994209	0.069184	14.3704	0
βi	0.861692	0.011788	73.09983	0
Φ1	0.036573	0.009111	4.014814	0.0001
Φ2	0.000766	0.006152	1.448201	0.0737
Φ3	-0.000543	0.005079	-0.10683	0.9149
v	0.869052	0.098541	8.819156	0
Γ	6.44113	0.707081	9.109459	0
NSR
ϑ0	0.049766	0.005443	9.142806	0
αi	0.124476	0.008233	15.11968	0
χi	0.999998	0000000	4800000	0
βi	0.861615	0.010638	80.99348	0
Φ1	0.03738	0.009882	3.782666	0.0002
Φ2	0.003984	0.007429	0.53633	0.5917
Φ3	0.006086	0.00959	0.634627	0.5257
v	0.897292	0.095359	9.409617	0
Γ	7.359537	0.910916	8.079272	0
DXR
ϑ0	0.034172	0.005377	6.355084	0
αi	0.073512	0.008658	8.490796	0
χi	0.997048	0.084556	11.79154	0
βi	0.922762	0.009217	100.1114	0
Φ1	0.008138	0.004861	1.674201	0.0941
Φ2	0.008571	0.010051	0.852748	0.3938
v	0.879833	0.134396	6.546564	0
Γ	8.037546	0.937191	8.576207	0
NKR
ϑ0	0.05136	0.006832	7.518056	0
αi	0.104771	0.010673	9.816201	0
χi	0.883281	0.081964	10.77651	0
βi	0.883054	0.011243	78.54085	0
Φ1	0.023895	0.007921	3.016814	0.0026
Φ2	-0.003074	0.006092	-0.504632	0.6138
v	0.864164	0.111283	7.765457	0
Γ	6.600645	0.658709	10.02058	0
FMBR
ϑ0	0.0346	0.004874	7.098794	0
αi	0.090903	0.007194	12.63516	0
χi	0.999987	0.000001	4800000	0
βi	0.898316	0.009178	97.87958	0
Φ1	0.01981	0.00834	2.37527	0.0175
Φ2	0.001447	0.006227	-2.32424	0.0100
v	1.064429	0.118983	8.946031	0
Γ	6.246663	0.689603	9.058344	0
SCR
ϑ0	0.011564	0.003894	2.969629	0.003
αi	0.067482	0.009789	6.893807	0
χi	0.13211	0.069902	1.889922	0.0588
βi	0.938788	0.007753	121.0913	0
Φ1	0.024274	0.016066	1.510895	0.1308
Φ2	0.000971	0.008832	10.99064	0.0000
Φ3	0.003074	0.005022	0.612173	0.5404
v	1.382652	0.265928	5.199348	0
Γ	4.475004	0.405198	11.04399	0

Source: author's calculation.

Fig. 2.

Conditional variance plots of each stock market.

The investigation specified that negative news has a more significant impact on these markets as compared to positive news of the same magnitude, as χ_i showed positive and significant values for these markets. Moreover, the study originates that shocks measured through standardized residuals, and one-period lag variance extensively distress its own impulsiveness, inferring that resilient ARCH (α_i) and GARCH (β_i) effect exist in these markets, respectively. The sum of the ARCH (α_i) and GARCH (β_i) coefficients represents the volatility persistence in a market (Dedi and Yavas, 2016). This evaluation argued that shocks generated due to COVID-19 would persist for an extended period in these markets, as the sum of ARCH (α_i) and GARCH (β_i) parameters of these markets is near to one. In addition, the power term identified the value of 0.86, 0.89, 0.87, 0.86, 1.06, and 1.38 for SPR, NSR, DXR, NKR, FMBR, and SCR, respectively. It indicates the utilization of the model, which permits the power factor to be analyzed. The degree of freedom parameter (Γ) of these markets is significant, and earned fair values range from 4.4 to 8.03. It recognized that the investigation has efficaciously captured the actual fat-tailed returns distribution of these stocks.

3.3. Diagnostic measures

In order to ascertain the serial correlation and Heteroskedasticity in the square of standardized residuals of the model, this examination employed Ljung-Box Q and ARCH LM test. The consequences accessible from Table 6 , mentioned that there was no serial correlation and Heteroskedasticity in the square of standardized residuals for each model, concluding that all the models performed are correct.

Table 6.

Diagnostic parameters.

	Serial Correlation		Heteroskedasticity
Variables	Q statistics	P-value	F-statistics	P-value
SPR	33.071	0.609	2.127126	0.1448
NSR	42.096	0.224	2.238769	0.1067
DXR	46.167	0.119	2.275491	0.1315
NKR	28.143	0.822	0.531143	0.4662
FMBR	37.113	0.417	0.117112	0.7322
SCR	44.462	0.157	1.353077	0.2448

Source: author's calculation.

4. Conclusion

This study applied the APGARCH model to analyze the non-linear behavior of financial markets of the US, Germany, Italy, Japan, and China during the COVID-19 and GFC period. The findings specified that COVID-19 has a substantial and harmful impact on stock returns of the S&P 500; however, it showed an inconsequential impact on the Nasdaq Composite index. As well, the conditional variance of European and the US markets, during the era of COVID-19, is high as compared to the GFC epoch; but conditional variance during the GFC period is high in Asian markets. Hence, the European and the US markets are more affected by COVID-19 as compared to Asian markets. Thus, Asian markets still provide better opportunities to diversify financial risk. Additionally, this investigation confirmed the leverage effect in these markets. The investigation argued that COVID-19 had stopped the economic circle throughout the world, and it can cause more dangerous shocks in these markets. The study confirmed that the health crisis of COVID-19 has successfully produced the financial crisis. Consequently, a significant portion of the budget should be spent to mitigate this kind of pandemic in the future. Fallouts of this study have the same implications for the other markets of the US, Europe, and Asia. These findings are vital for policymakers, investors, academicians, portfolio managers, and researchers.

Funding

The study is supported by National Natural Science Foundation of China (No.71673043).

CRediT authorship contribution statement

Khurram Shehzad: Methodology, Software, Writing - original draft. Liu Xiaoxing: Supervision. Hayfa Kazouz: Writing - review & editing, Data curation.

Declaration of Competing Interest

“The authors reported no potential conflict of interest”