Analysing Market Volatility and Economic Policy Uncertainty of South Africa with BRIC and the USA During COVID-19

Abstract

The contagious COVID-19 disease not only brought about a global health crisis but also a disruption in the global economy. The uncertainty levels regarding the impact of the disease increased volatility. This study analyses stock market volatility and Economic Policy Uncertainty (EPU) of South Africa (SA) with that of the United States of America (USA) and Brazil, Russia, India, and China (BRIC) during the COVID-19 pandemic. The study aims to analyse volatility spillovers from a developed market (USA) to emerging markets (BRIC countries) and also to examine the causality between EPU and stock returns during the COVID-19 pandemic. By employing the GARCH-in-Mean model from a sample of daily returns of national equity market indices from 1 January 2020 to 31 March 2022, SA and China are shown to be the most volatile during the pandemic. By using the diagonal Baba, Engle, Kraft, and Kroner (BEKK) model to analyse spillover effects, evidence of spillover effects from the US to the emerging countries is small but statistically significant, with SA showing the strongest impact from US market shocks. From the Granger causality test, Brazil’s and India’s equity markets are shown to be highly sensitive to changes in EPU relative to the other countries.

1. Introduction

The previously unheard-of levels of stock volatility emanating from the COVID-19 pandemic shook global financial markets, putting the resilience and stability of developed and emerging countries to the test. During this period, investors had serious concerns about market performance and the ability of different economies to cope with the crisis. The crash of various major indices in financial markets gave an indication that the pandemic had adversely impacted the global market and that a market crisis was brewing. For example, in the USA, financial markets saw one of the most significant crashes in history during March 2020, during the outbreak of the COVID-19 pandemic. There was a 7% decline in the S&P 500 index in just four minutes on 9 March 2020 amid fears of the pandemic (Samitas et al., 2022). The same index also declined by 9.51% and 11.98% on 12 March 2020 and 16 March 2020, respectively, marking the biggest daily declines since Black Monday on 19 October 1987, when it declined by 20.4% (Imbert, 2020). Rakshit and Neog (2022) assert that markets were more volatile than they were during the Global Financial Crisis (GFC) in 2008.

Of all the stock market events, e.g., the GFC, DotCom Bubble, COVID-19 pandemic, stock return volatility—the fluctuation of returns over a given time period—receives a lot of attention and consideration when it comes to investing activities because of the effects it has on portfolio risks and expected returns for investors (Adrian & Rosenberg, 2008). Due to its effect on financial market liquidity, high return volatility may potentially lead to lower returns, particularly for risk-averse investors, because it is often associated with increased uncertainty and risk. Owing to this, there is a conventional view in financial markets, built on the principle of ambiguity aversion, that markets are averse to uncertainty and volatility. As a result, a decrease in stock return may occur because investors avoid highly volatile markets in favour of more stable ones, a trend known as the flight-to-safety (Talwar et al., 2021).

Scholars and industry professionals have shown great interest in stock returns and volatility for both theoretical and empirical reasons, as they play important roles in risk management and the creation of optimal portfolios (Bhowmik & Wang, 2020). The primary objective of this study is to analyse South Africa (SA)’s stock volatility together with that of the BRIC countries and the USA during the pandemic. Moreover, this research study aims to analyse volatility spillovers from the developed markets (USA) to emerging markets (BRIC countries). Lastly, the study aims to examine the causality between EPU and stock returns during the COVID-19 pandemic.

Prior studies mostly focus only on the relationship between volatility spillover effects within emerging or developed countries during the pandemic, often without conducting cross-country comparative analyses, nor considering the impact of EPU thereof (Rakshit & Neog, 2022; Khan, 2023; Sahoo & Kumar, 2023). Singh et al. (2024) conducted a study in which the authors explored the relationship between risk–return and volatility spillover on stock returns of BRIC countries; this study will extend the analysis by including a developed country, the USA, and conduct a cross-country analysis of spillover effects. This study will contribute to the body of knowledge by providing a comparative analysis of how policy uncertainty in developed and developing markets affects stock returns during extreme market events such as the COVID-19 pandemic. The case of SA is particularly important since it has a dual economic structure, and it is often seen as a gateway to African markets. Furthermore, SA has been one of the leading investors in renewable energy and sustainability amid the ongoing energy crises, and it heavily relies on sectors such as hospitality and tourism, which were adversely impacted by the pandemic as a result of lockdowns. The country also experienced digital transformation, particularly in digital public policy, and fintech transformation, as a result of the COVID-19 pandemic. Thus, the case of SA will offer a distinct understanding of global shock transmission. From the insights provided by this comparative analysis, policymakers can formulate more effective economic policies to stabilise markets during crisis periods. In addition to helping contextualise SA’s market behaviour during the pandemic, the comparison of SA’s market volatility with that of the USA and the BRIC countries will offer a global perspective and lessons that may be applied to other emerging economies. By studying the volatility transmissions between developed and developing economies, the study will enhance the understanding of how stock return volatility is spilled over across different regions.

The remaining sections of the study will provide a review of the literature, the methodology and results of the study, discussion, and the conclusion.

2. Literature Review

2.1. EPU and Stock Return Volatility

There is a link between EPU and stock volatility shown by existing literature (Dai et al., 2021; Binotto & Marschner, 2024; Batabyal & Killins, 2021). In a study aimed at investigating the link between stock prices and policy uncertainty from 2001 to 2013 in seven OECD countries, Chang et al. (2015) used a bootstrap panel for a causality test and found that high policy uncertainty decreases stock prices the United Kingdom (UK) and USA. The existence of this relationship between EPU and stock market volatility is further ascertained by Yuan et al. (2022), where the authors examined the effect of EPU on the risk of stock price crash on commercial banks. The study revealed that EPU increases the crash risk of the stock prices of banks.

The impact of EPU on stock returns of the Canadian stock market, one of the leading countries in the OECD inter-governmental organisation in terms of economic growth, was studied by Batabyal and Killins (2021). The authors used monthly returns over a period of 1985 to 2015 and employed linear and nonlinear ARDL models to identify the potential asymmetric impacts that EPU may have on the equity returns in both long-run and short-run. The study revealed that stock markets in Canada are significantly impacted by changes in EPU. The results showed negative coefficients in both nonlinear and linear modelling, which imply that when there is an increase in EPU, stock returns tend to decrease.

To further support the relationship between EPU and stock return volatility, Ghani and Ghani (2024) used a GARCH-MIDAS model to research how EPU affect the Pakistan stock market volatility (emerging market) and that of the USA, UK and China (developed). Monthly EPU indices and daily stock data was used. The results of the study revealed that EPU index of the USA was a more powerful predictor of the volatility of the Pakistan stock market. Furthermore, the EPU index of the UK was also a predictor of the Pakistan equity market volatility, although not as significant as that of the USA. In contrast, China showed no evidence of equity market volatility prediction during the sample period from August 2010 to December 2020. The authors also extended their analysis to during the COVID-19 period, identical results further showed that USA’s EPU index was a more potent predictor for the stock market volatility in Pakistan, followed by the UK. China was again found to not have predictive nature of the Pakistan’s equity market volatility even during the COVID-19 period.

Rani et al. (2025) did a study where they researched how the EPU impacted the Indian stock market and exchange rate volatility from 2000 to 2021 using the GARCH-MIDAS model. The authors used the EPU of the US as the proxy for global EPU to show the effect of uncertainty at both the global and domestic level. COVID-19 pandemic and the Global Financial Crises were incorporated as dummy variables to show the effect of major extreme market events. The study found that when EPU increases, Indian stock market volatility also increased during extreme market events. Furthermore, the study also revealed that during periods of high EPU, the India’s exchange rate volatility intensified. The results were consistent with the findings of Batabyal and Killins (2021).

2.2. Volatility Spillovers Between Developed and Emerging Markets

The concept of market spillovers implies that a shock or instability to one stock market increases affects the stock returns of other markets (Hasan et al., 2019). Various studies proved that volatility in one market has the potential to spillover into other markets, particularly in periods of higher market turbulence (N. Prasad et al., 2018; Gamba-Santamaria et al., 2019). N. Prasad et al. (2018) found that, in line with previous research and predictions, the USA stock market is the primary source of volatility transmission worldwide, ascertaining that risk spillovers are more prevalent, particularly in the event of extreme market crises. E. S. Prasad et al. (2009) assert that emerging markets, such as BRIC countries, are becoming more vulnerable to external shocks from developed markets due to factors such as the illiquidity of their markets, the slow progress in economic development, and the stability in financial institutions and strict regulatory frameworks.

Global stock market integration has simplified the movement of capital in both emerging and developed markets. To ascertain the existence of stock market integration, Mensi et al. (2023) investigated the asymmetric returns between various international indices by using the spillover index developed by Diebold and Yılmaz (2014). The study revealed that price variations in equity indices are influenced by major international market events such as Brexit, GFC, and the COVID-19 pandemic. The results of the study also revealed that there exists time-varying volatility spillovers between international stock markets. Thus, high market integration is also associated with increased volatility spillovers between financial markets.

Hanif et al. (2025) did a study on how the Gulf Cooperation Council (GCC) global bonds and equity markets are interconnected, focusing on the US bonds and European Monetary Union by employing quantile coherency and quantile connectedness methodologies. Using a sample period from July 2007 to September 2023, the study found that Saudi Arabia and United Arab Emirates had the most impact on the returns of other GCC countries. Additionally, the study also found that during periods of extreme markets and global crises, the interconnectedness of GCC markets with bonds responds differently that when the global economy is stable.

Bhar and Nikolova (2009) investigated the degree of integration and dynamic interactions between the regional, world, and the four emerging economies being BRIC markets—Brazil, Russia, India, and China. India was found to have the highest level of market integration, with Brazil, Russia, and China following, respectively. The authors also discovered a negative relationship between the conditional volatility of India and the Asia-Pacific region, and of China and the world. These negative correlations provide a diversification opportunity for investors.

Notwithstanding the extensive research investigating the relationship between market volatility and EPU in both developing and developed markets, the current studies predominantly focus on emerged and larger developing economies, being Pakistan and BRIC countries. There still exists an understudy on the South African equity market during the pandemic, albeit being highly exposed to global shocks and considered the gateway to the African market. Moreover, there were major policy shifts in SA during the COVID-19 pandemic which may have further exacerbated policy uncertainty. This study will close that gap and extend empirical evidence by analysing SA’s stock volatility with that of BRIC countries and the USA during the pandemic. Additionally, volatility transmissions from the USA to BRIC countries, and the causality between the EPU and stock returns between January 2020 and March 2022 will be analysed to bridge this gap.

3. Data and Methodology

3.1. Data Collection

The secondary financial data, being the closing prices of the indices, was sourced from IRESS BFA and www.investing.com (historical data archive). The closing prices of the indices were collected from 1 January 2020 to 31 March 2022. This sample period was chosen because it captures the outbreak, peak and the immediate impact of the pandemic. Furthermore, this period highlights initial investor and market reaction and how market sentiment changed during the COVID-19 pandemic. Additionally, significant fiscal and monetary policies were introduced by governments and central banks worldwide during this time, which is important information to have when studying policy uncertainty (Mihailova-Borisova, 2021). The study used the daily index returns because they capture the price swings that occur within a trading day, which is important in assessing volatility, and monthly data was used to be consistent with EPU indices, which are published monthly. Additionally, monthly data is less sensitive to high-frequency fluctuations and irregularities that often affect daily data. Thus, the study had a total number of 3353 daily and 324 monthly return observations, including EPU index returns.

The equity indices from each country used in this study were a representation of the stock markets in their respective countries. Specifically, the daily and monthly (for correlation analysis with the policy uncertainty indices) price data to be used for the following markets: Brazil (IBOVESPA), India (NIFTY 50), China (Shanghai Composite Index), Russia (MOEX Russia Index), South Africa (JSE ALSI), and the US (S&P 500). These indices are suitable for comparative analysis since they are widely used and accurately reflect the performance of the respective national markets (Muguto & Muzindutsi, 2022).

The study used the Baker et al. (2016) monthly country-level EPU indices for the five countries (excluding SA as it is not published). This indicator is determined by counting how frequently terms like “uncertainty,” “tax,” “spending,” “uncertainties,” “uncertain,” “deficit,” “Federal Reserve,” “legislation,” and “regulation” appear in newspaper articles. As a proxy for uncertainty, the other component gauges the degree of disagreement among economic forecasters (Baker et al., 2016; Hung, 2021; Davis, 2016). The SA EPU index was sourced from NWU, developed by Prof. Waldo Krugell and Prof. Raymond Parsons. The index includes three components: newspaper coverage on uncertainty, views on uncertainty by a leading group of economists, and input from Stellenbosch University’s Bureau for Economic Research regarding constraints to manufacturers doing business among themselves.

The developers of the index set a threshold of 50; a value above 50 would indicate high policy uncertainty and anything lower than 49 indicates less uncertainty. The SA policy uncertainty index is published on a quarterly basis, as such, the data had to be interpolated using linear, constant and quadratic to enhance the data robustness before being fitted in any model. Linear interpolation was chosen over quadratic and constant interpolation because of its simplicity and asymptotic optimality (Blu et al., 2004).

3.2. Methodology

The methodology used in this study includes various statistical techniques such as Descriptive Statistics, Stationarity Test, ARCH effect test, GARCH-M, BEKK-GARCH and the Granger causality Test.

The study used log returns, which have a more symmetrical and accurate normal distribution relative to simple returns. Statistical modelling and analysis benefit from this characteristic because many financial models assume regularly distributed returns (Gundersen et al., 2023). The returns of each index were calculated as follows:

$$R_t=ln({\displaystyle \frac{Pt}{Pt-1}}) \times 100$$

(1)

where $R_t$ represents the index daily return on day t, $P_t$ is the index closing price at day t, and $Pt-1$ is the previous day’s closing price of the index.

3.2.1. Stationarity Test

Unit root testing was used to verify if the series is stationary since a non-stationary time series of financial returns could restrict the applicability of reliable forecasting models in informal testing (Timmermann & Granger, 2004). A time series is said to be stationary if a change in time does not result in a change in the distribution’s shape (Chaudhary et al., 2020). As a result, the study employed the Augmented Dickey–Fuller (ADF) and the Phillips-Perron (PP) and the Kwiatkowski Phillips Schmidt-Shin (KPSS) tests to test if the financial series is stationary or not. The two tests being the ADF and PP are based on the assumption that the time series has a unit root, which suggests non-stationarity, in contrast to the alternative hypothesis that the time series does not have a unit root, which suggests stationarity (Dickey & Fuller, 1981). The KPSS test assumes stationarity and that the null hypothesis states that the series does not contain a unit root.

H₀. 

The null hypothesis in the ADF and PP test states that the series contains a unit root (the data is non-stationary).

H₁. 

The alternative hypothesis in the ADF and PP test states that the series does not contain a unit root (the data is stationary).

H₀. 

The null hypothesis in the KPSS test states that the series does not contain a unit root (the data is stationary).

H₁. 

The alternative hypothesis in the KPSS test states that the series contains a unit root (the data is not-stationary).

3.2.2. ARCH Effect Test

Subsequent to the confirmation that the series is stationary, the study proceeded to test the presence of heteroskedasticity by employing the Autoregressive Conditional Heteroskedasticity-Lagrange Multiplier (ARCH-LM) Test, which is the acknowledged highest standard test for detecting ARCH and is a preliminary step to GARCH modelling (Sjölander, 2011). Engle and Ng (1982) proposed the ARCH model to represent the volatility’s time-varying nature.

The following auxiliary regression model was used to test for ARCH of order p:

$$u_t^2={\gamma }_0+{{\gamma }_{1}u}_{t-1}^2+{{\gamma }_{2}u}_{t-2}^2+\cdots +{{\gamma }_{p}u}_{t-p}^2+v_t$$

(2)

where t and u denote the time period and residual square (which will be measured by the basic regression model), respectively. The constant term is represented by γ, while $v_t$ represents the error term. Lastly, p denotes the number of lagged terms in the model.

The alternative hypothesis, which suggests that there are ARCH effects in the returns series, varies from the null hypothesis of the ARCH-LM test, which implies the existence of ARCH effects in the returns series. The null hypothesis in this test can be represented by the following:

$$H_0={\gamma }_0={\gamma }_1={\gamma }_2={\gamma }_p=0$$

(3)

The conditional mean equation that was performed in order to test for ARCH effects is shown in Equation (4) below:

Yt = µ + ε_t

(4)

where $Y_t$ = conditional mean

$\mu$ = a constant

${\epsilon }_{\mathrm{t}}$ = error term that is normally distributed with zero mean.

For robustness of the results, the study also applied the White Test to test the presence of heteroskedasticity

3.2.3. Multivariate GARCH Models

The study made use of multivariate GARCH models because of the models’ ability to capture the risk premium (GARCH-M) and volatility spillovers (BEKK-GARCH), and the time-varying volatility across different stock markets. Other models like the GARCH-MIDAS were not employed due to its complexity and the study’s sample size, as the model often requires larger data sets.

GARCH-M Model

Time-varying relationships between time series can be captured using standard estimate approaches such as multivariate GARCH-type models. Additionally, in studies of contagion, multivariate GARCH models have been employed to examine volatility and correlation transmission as well as spillover effects (He et al., 2008).

$$\mathrm{Mean} \mathrm{equation}: Y_t= \mu + \mathsf{\lambda }{\sigma }_t^2+{\epsilon }_t$$

(5)

$$\mathrm{Variance} \mathrm{equation}: {\sigma }_t^2=\omega +{\alpha }_1{\epsilon }_{t-1}^2+{\beta }_1{\sigma }_{t-1}^2.$$

(6)

where $Y_t$ is the asset’s return at time t, μ is the average return, ω is the constant term, α is the ARCH term’s coefficient, β is the GARCH term’s coefficient, and ${\epsilon }_{\mathrm{t}}$ is the residual return. Additionally, $\omega >0$, ${\alpha }_1$ ≥ 0, and ${\beta }_1$ ≥ 0. The risk premium is denoted by the parameter λ in the mean equation. A high coefficient (λ) indicates a positive correlation between the return and volatility. Moreover, ${\sigma }_t^2$ denotes the conditional variance at time t. A security’s return could be influenced by its volatility (risk), especially in economic crisis periods. The GARCH-M (1,1) model includes a heteroskedasticity factor into the mean equation in order to explain such events.

BEKK-GARCH Model

$$\mathrm{Mean} \mathrm{equation}: Y_t= \mu +{\mathsf{\epsilon }}_{\mathrm{t}}$$

(7)

$$\mathrm{Variance} \mathrm{equation}: H_t=C^{\prime }C+B^{\prime }{H}_{t-1}B+A^{\prime }{\epsilon }^{\prime }\epsilon A$$

(8)

where $H_t$ is the conditional covariance matrix at time t. A is the ARCH effect coefficient matrix, B is the GARCH effect coefficient matrix, C is the constant matrix, and ε represents residuals. Each coefficient is of a 2 × 2 order, as modelling all of the six indices at once would have led to a large number of parameters and subsequently making the interpretation of the results ambiguous. The ARCH impacts from previous returns to current conditional variances are examined in Matrix A. The GARCH influence from previous conditional variance to current conditional variances is examined in Matrix B (Engle & Kroner, 1995).

To analyse how covariance equations describe volatility and cross-country volatility of the six emerging and developed countries included in the study, the diagonal BEKK model was applied to the USA and the emerging markets (Brazil, India, China, Russia, and SA).

3.2.4. EPU Granger Causality Test

The following regressions were used to test if EPU Granger-causes the returns of a particular country’s stock index:

For Brazil:

$${Brazil}_t=\alpha +{\sum }_{i=1}^p{\beta }_1{Brazil}_{t-1}+{\sum }_{i=1}^p{\gamma }_i{EPU}_{t-i}+{\mathsf{\epsilon }}_t$$

(9)

The hypothesis to be tested is:

$$H_0:{\gamma }_1={\gamma }_2={\gamma }_3\dots {\gamma }_p=0$$

(10)

EPU will be implied to Granger-cause the stock market index returns if the coefficient ${\gamma }_i$ are significant jointly. Therefore, the null hypothesis would be rejected.

The study employed nonlinear models instead of simultaneously applying both linear and linear because the intention was to capture the time varying volatility and asymmetry to economic policy uncertainty and the COVID-19 pandemic. While both approaches would have enhanced the robustness of the results, nonlinear models are more suitable in modelling stock market behaviour, especially during periods of heightened volatility. Linear models tend to fall short on capturing volatility clustering and detecting volatility spillovers.

4. Results and Discussions

Table 1. Descriptive statistics of the log returns.

National Stock Indices
	Brazil	SA	Russia	India	USA	China
Mean	0.00002	0.00044	−0.00019	0.00078	0.00088	0.00011
Median	0.00092	0.00075	0.00163	0.00196	0.00134	0.00067
Maximum	0.13022	0.07262	0.18262	0.08400	0.08968	0.05554
Minimum	−0.15993	−0.10227	−0.40467	−0.13904	−0.12765	−0.08039
Std. Dev.	0.02134	0.01490	0.02529	0.01574	0.01587	0.01147
Skewness	−1.56751	−1.06205	−7.35896	−1.77245	−0.94880	−0.90530
Kurtosis	19.08597	12.05834	132.21820	19.55472	18.51147	9.65846
Jarque–Bera	6233.45	1951.33	370697.90	6401.29	5667.63	1029.64
Probability	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000

displays the descriptive statistics of the log returns;

Table 2. Descriptive statistics of the country-level EPU.

	Brazil EPU	China EPU	India EPU	Russia EPU	SA EPU (Linear)	USA EPU
Mean	219.4989	619.4874	80.12948	423.7374	56.36067	246.3727
Median	204.1461	592.8005	75.14697	392.8705	56.72646	201.4337
Maximum	368.8998	934.717	167.7502	793.6345	59.86547	503.9633
Minimum	62.591	356.0635	23.35276	185.3315	50.3	136.9202
Std. Dev.	79.52119	161.0367	33.9958	177.0099	2.502065	105.4151
Skewness	0.05001	0.334922	0.95975	0.696472	−0.699927	0.938931
Kurtosis	2.194942	2.107329	3.983364	2.44754	2.751587	2.794674
Jarque–Bera	0.685544	1.297452	4.845296	2.339066	2.105522	3.717211
Probability	0.7098	0.522711	0.088686	0.310512	0.348973	0.15589

displays the unit root properties of the stock market indices as captured by the ADF and PP techniques; and

Table 3. Unit root results from the log returns.

	Brazil	Russia	India	China	SA	USA
ADF test	−7.38666	−39.82584	−23.95359	−22.52699	−23.40412	−7.09691
	0.0000	0.0001	0.0000	0.0000	0.0000	0.0000
PP test	−28.09865	−37.41825	−24.03939	−22.54641	−23.39304	−30.32080
	0.00000	0.00000	0.0000	0.0000	0.0000	0.0000
Result	Stationary	Stationary	Stationary	Stationary	Stationary	Stationary
Test critical values at level: 1% = −3.44, 5% = −2.86, 10% = −2.56
	Brazil	Russia	India	China	SA	USA
KPSS test	0.12508	0.19749	0.12099	0.12094	0.08055	0.07099
Result	Stationary	Stationary	Stationary	Stationary	Stationary	Stationary
Test critical values at level: 1% = 0.739%, 5%= 0.463, 10% = 0.347

For Russia, alpha + beta = 1.04, indicating nonstationarity of GARCH process, authors noted otherwise, they wrotestationarity is met in their response file from the previous round.

and

Table 4. ARCH-LM results.

	F-Stat	Prob. F	Obs. R-Squared	Prob. Chi-Square
Brazil	58.35912	0.0000	52.16920 ***	0.0000
Russia	13.29714	0.0003	12.98696 ***	0.0003
India	13.21546	0.0003	12.90617 ***	0.0003
China	13.12979	0.0003	12.86512 ***	0.0003
SA	10.76566	0.0011	10.57808 **	0.0011
USA	53.31980	0.0000	48.68441 ***	0.0000
Conclusion	Present	Present	Present	Present

Significant at 1% level ***, significant at 5% level **.

exhibit the ARCH-LM test results and conditional correlation coefficients, respectively. Figure 1 presents the return volatility patterns of the US and the BRIC.

4.1. Descriptive Statistics

It is essential to analyse a number of descriptive statistical properties before developing several volatility models for the study. These include the residuals’ standard deviation, skewness, kurtosis, and Jarque–Bera normality test. The primary goal of analysing these descriptive statistics is to determine the distributional properties of the return series before performing the required tests on the econometric models.

Table 1 presents the summary of the descriptive statistics for the daily log returns of the series from 1 January 2020 to 31 March 2022. The findings show that there is a considerable amount of variation in the equity market returns among the six nations. Russia has the biggest negative mean return, which indicates a marginal loss on average, whereas the USA and India have the highest average returns. The high average returns show a slight gain over time. The largest median return (0.00196) comes from India, indicating a skewed distribution of returns. The two countries that exhibited the highest standard deviations were Russia (0.02529) and Brazil (0.02134), which suggests that the two countries had the highest risk comparatively, since standard deviation is a measure of risk. Although the COVID-19 outbreak was in China, Table 1 shows that the country had the lowest risk. China’s State-Owned Entities (SOEs) play a crucial role in the economy, and they contribute the most revenue to the country (Huang & Nicolas, 2021). Moreover, the majority of these SOEs are not listed on public exchanges, in contrast with the rest of the world, where most of the big companies are publicly traded. Thus, the country was less vulnerable to the financial shock brought by the pandemic.

Kurtosis shows the existence of extreme outliers in terms of both significant gains and losses in its return distribution (Lai, 2012). Since the kurtosis statistics are greater than 3, all of the return series are leptokurtic, meaning they have noticeably fatter tails and higher peaks. Additionally, Table 1 shows that with a kurtosis of 132.21820, Russia exhibits extreme outliers, or “fat tails.” This might have been due to the fact that exports of gas and oil are vital to Russia’s economy (Yang et al., 2021). Fluctuations in the price of commodities around the world during pandemic might have led to abrupt market surges or collapses, which could have resulted in extremely high or low profits. Lastly, a significant Jarque–Bera statistic with a probability of 0 is shown for every country, indicating that the returns are not normally distributed.

It can be deduced from Table 2 that China had the highest mean EPU of 619.4874, which suggests higher overall policy uncertainty. From the measure of dispersion, Russia and China exhibits the highest standard deviations, indicating constant changes to policy uncertainty during the pandemic. From the Kurtosis, India is the market that shows signs of leptokurtosis (with a coefficient of 3.983364) from the results of the country-level EPU, the null hypothesis of normality is rejected for all markets since the p-values are more than 0.05.

4.2. Stationarity Tests

Table 3 exhibits the ADF, PP and KPSS test results of the series. The results of the ADF test show that each country’s test statistic is extremely negative (ranging from −7.09691 to −37.82548), and their corresponding p-values are also very low (all 0.0001 or less). This shows that the time series for each country is stationary and strongly rejects the null hypothesis of the presence of a unit root. Similarly, the results of the Phillips-Perron test indicate significantly negative statistics and p-values of 0.0000, proving that each country’s series is stationary. The test statistic from the KPSS results of all indices are below the critical values, confirming stationarity. Therefore, all the tests confirming the absence of a unit root, it can be concluded that the equity market returns in all the countries analysed are stationary. The stationarity test results for the monthly stock returns and country-level EPU indices are exhibited an Appendix A.

4.3. ARCH-LM Test Results

ARCH and GARCH models are used to study the persistence of volatility in various national stock markets. The ARCH-LM test of the mean equations of the stock markets under study is shown in Table 4, wherein the observed R-square with its probability value and the estimated F-statistics with its probability value are displayed. According to the ARCH-LM tests of the mean equation, the equity markets of all the countries appear to have an ARCH effect, confirmed by a very low probability value, which is closer to zero for both F-statistics and observed R-square. Although still statistically significant, South Africa exhibits the lowest degree of ARCH effects, when compared to the other countries. Brazil and the USA demonstrate very strong ARCH effects with very high F-statistics and Chi-square values. Thus, the null hypothesis that there is no presence of an ARCH effect is rejected for all countries, permitting the study to estimate the extensions of GARCH models.

4.4. GARCH-M Results

Table 5. GARCH-M results.

Coefficient	Brazil	Russia	India	China	SA	USA
Mean equation
µ (constant)	0.000325	0.000714	0.000614	−0.000744	0.000006	0.000704
λ (Risk Premium)	−0.124954 **	−0.046123	0.003488	0.017777	0.570590	−0.086469 *
Variance equation
$\omega$ (Constant)	0.000014 ***	0.000006 *	0.000005 ***	0.000017 ***	0.000014 ***	0.000008 ***
${\alpha }_1$ (ARCH effect)	0.129258 ***	0.286396 ***	0.134174 ***	0.226032 ***	0.156687 ***	0.244489 ***
${\beta }_1$ (GARCH effect)	0.823029 ***	0.760305 ***	0.844575 ***	0.655250 ***	0.761081 ***	0.720585 ***
${\alpha }_1$$+{\beta }_1$	0.952287	1.046701	0.978749	0.881282	0.917768	0.965074

Significant at 1% level ***, significant at 5% level **, significant at 10% level *.

presents the estimates of the GARCH-M (1,1) model, which is used to find the relationship between risk and return among the equity market indices of various countries in investigation. The constants in the mean equation all the countries are positive and statistically insignificant, except for China, indicating that the returns of the countries are driven by volatility. Thus, high returns are associated with high volatility but the trend is not consistent and that the risk-return trade cannot be relied upon. China’s constant is not only statistically insignificant but negative, indicating negative stock returns, which can be ascribed to China being where the COVID-19 outbreak began. After China, South Africa’s equity market offers the lowest returns; the close-to-zero return (0.000006) during times of low volatility. Although not statistically significant, this may be an indication of fundamental problems in South Africa’s equity market, such as low investor confidence or less potential for growth relative to other countries.

The parameter λ denotes the risk premium. The λ parameter shows how investors perceive risk, with the assertion that higher-risk assets would have a higher return. The results in Table 5. exhibit the USA and Brazil to have a statistically significant but negative risk-return trade-off, meaning the higher the risk, the lower the returns will be.

The parameter $\omega$ in the variance equation indicates the long-term volatility experienced by each equity market during the COVID-19 pandemic on average. From Table 5, it can be deduced that all the $\omega$ parameters are statistically significant at 1% level, implying that the equity markets in all countries always have level volatility. China, followed by Brazil and SA, had the highest long-term volatility, with the former having a constant of 0.000017 and both Brazil and SA exhibiting constants of 0.00014. With a constant value of 0.000005, India had the lowest baseline volatility relative to other countries. In light of that, India was less volatile in the long term.

From Table 5, the parameter (α₁) denotes the ARCH effects, which capture how short-term volatility reacts to market shocks. The ARCH effects in the variance equation explain how the current volatility is impacted by recent volatility or the arrival of new information. The statistical significance across all countries indicates that current volatility was significantly impacted by past volatility. Russia and the USA have the highest ARCH effects of 0.286396 and 0.244489, respectively, suggesting that new information has a significant impact on the current volatility. Thus, markets quickly respond to shocks, like the COVID-19 pandemic, by increasing volatility. In contrast to Russia and the USA, Brazil and India exhibit the lowest ARCH effects. Signalling that current volatility responds to new information or recent market volatility with less aggression compared to the USA and Russia.

The GARCH coefficient (β₁) denotes persistent volatility over time. Volatility is said to remain in the market for a sustained period post-shock when there is a high (β₁) value. There is statistical significance across all stock markets, implying that market volatility is persistent over time. With Brazil (0.823029) and India (0.844575) having the highest (β₁) value, this shows that volatility typically lasts for a long period of time following a shock in those two markets. This implies that subsequent to a heightened volatility, these markets take longer to settle. Contrary to Brazil and India, China showed the lowest (β₁) value, proposing that China’s stock market stabilises quickly post a shock. This can be because of the government’s increased market interventions to reduce volatility.

Persistence of variance is represented by the sum of (α₁) and (β₁) coefficients. The interaction between the long-term volatility persistence (GARCH) and short-term shocks (ARCH) is reflected in this sum. Strong persistence is indicated by a sum that is either near to or more than 1, indicating that volatility may continue to increase for an extended period of time following a shock. Russia showed persistence of variance of 1.046701, suggesting extremely high persistence in volatility. Since the (α₁) + (β₁) sum is greater than 1, it implies that volatility increases with time. This is indicative of an unstable stock market where volatility is difficult to decrease and may need the government’s intervention. The Chinese stock market shows lower persistence of volatility, with a (α₁) + (β₁) sum of 0.881282. This implies that periods of high volatility decay with time relatively quicker.

4.5. BEKK-GARCH Results

Table 6. Diagonal BEKK-GARCH model results.

Parameter	M(1,1)	M(1,2)	M(2,2)	A1(1,1)	A1(2,2)	B1(1,1)	B1(2,2)
Brazil
Coefficient	0.00001 ***	0.00000 ***	0.00002 ***	0.47094 ***	0.30863 ***	0.85461 ***	0.92356 ***
Test statistic	4.12035	1.75721	3.75086	9.61699	16.23000	29.65024	71.70501
Russia
Coefficient	0.00001 ***	0.00000 ***	−0.00000 ***	0.64444 ***	0.48528 ***	0.70175 ***	0.94441 ***
Test statistic	5.39898	2.64852	−3.43135	10.60173	11.35717	17.17922	95.69021
India
Coefficient	0.00001 ***	0.00000 ***	0.00000 ***	0.44567 ***	0.36981 ***	0.86352 ***	0.91568 ***
Test statistic	5.20821	2.30721	3.30427	11.91624	15.53375	42.00448	68.51800
China
Coefficient	0.00001 ***	0.00000 *	0.00000 **	0.48271 ***	0.26965 ***	0.83874 ***	0.92470 ***
Test statistic	4.54043	1.67795	2.25833	10.26079	6.71171	28.78107	34.03373
SA
Coefficient	0.00001 ***	0.00000 ***	0.00001 ***	0.40846 ***	0.31865 ***	0.89229 ***	0.91043 ***
Test statistic	4.06435	4.43190	3.05558	9.59824	12.87369	48.96879	45.68100

Significant at 1% level ***, significant at 5% level **, significant at 10% level *.

exhibits the coefficients of spillover effects. The diagonal BEKK-GARCH model was used to determine how the BRIC emerging markets may have been affected by a developed country, being the USA. Consistent with previous studies (Bundoo & Ramlukun, 2022; Malik et al., 2022; Kaura et al., 2022; Mensi et al., 2016), the study employed the diagonal BEKK-GARCH model to account for shocks and volatility spillover effects. The model was applied to the US and the emerging markets to investigate how the conditional expectation and covariance equations represent the individual volatility and cross-country volatility of the five emerging and developed countries included in the study. The ARCH terms (which measure short-term persistence of volatility) are represented as A1(1,1) and A1(2,2), and GARCH (which accounts for both short-term volatility persistence (ARCH effect) and long-term volatility persistence) are denoted as B1(1,1) and B1(2,2), respectively. Lastly, constant terms are denoted by M(1,1), M(1,2) and M(2,2).

The constant variables in the BEKK-GARCH model indicate the variance and covariance between the USA and the specific underlying emerging country. The baseline variance for the USA is represented by M(1,1). This positive number for the USA denotes a consistent volatility level that persists despite external shocks or previous volatility impacts. The baseline variance for the paired nation (such as Brazil, Russia, etc.) is captured by M(2,2). Most emerging nations have positive values, which suggests that they are inherently volatile, as with the USA. Russia’s negative value, however, points to unique baseline volatility behaviour that might be a result of market-specific variables such as geopolitical tensions. The constant M(1,2) represents the baseline covariance between each emerging country and the USA. All the countries exhibited positive constant values, except for Russia, which has a slightly negative baseline return. The negative return can be attributed to Russia’s geopolitical tensions and weak investor sentiment in the country. The small M(1,2) values point to little but significant inherent co-movements in the markets of the USA and the BRIC. Notably, SA exhibits the strongest spillover effect from the US, with a test statistic of 4.43190, suggesting that SA is mostly affect by the US market shocks relative to the other emerging countries. Even in the absence of market shocks, these covariances demonstrate a certain degree of inherent relationship between each country and the USA.

The ARCH effects of the BEKK-GARCH model show how previous shocks affect current volatility, thus capturing the short-term persistence of volatility. The ARCH effects become much more significant considering that the analysis includes the COVID-19 pandemic period (1 January 2020 to 31 March 2022). This period was characterised by high uncertainty, unprecedented global economic shocks, and increased market volatility, which makes the short-term volatility persistence described by ARCH effects particularly relevant. A1(1,1) and A1(2,2) show how market volatility is impacted by shocks to the USA and the paired BRIC nations, respectively.

A1(1,1) and A1(2,2) have been shown to be significant, indicating that the influence of news on one index further influences the conditional covariance in other indices as well. Given its higher vulnerability to shocks, Russia exhibits the largest ARCH effect on its own and when paired with the US market, showing significant short-term volatility persistence. Russia’s stock market is the most reactive, with an A1(1,1) coefficient of 0.64444, implying that current market volatility is significantly impacted by recent events. Additionally, with an A (2,2) coefficient of 0.48528, Russia’s market is significantly impacted by the USA’s market returns.

The BEKK-GARCH results on Table 6 reveal a weak ARCH effect and a stronger GARCH effect. The coefficients of the GARCH effects being represented by B1 show that every country had a strong persistence of volatility during the pandemic, with the B1 coefficients varying from 0.70175 to 0.94411. This suggests that previous volatility shocks have a persistent impact on future volatility. Relative to the other BRIC countries, Russia has the highest volatility persistence, with B1(2,2) = 0.94441, indicating that volatility shocks in the Russian market have a longer-lasting effect. Thus, reflecting market sensitivity to worldwide events during the pandemic and geopolitical concerns. Moreover, with B1(1,1) = 0.89229, South Africa likewise shows large volatility persistence, suggesting that shocks to the volatility have a significant impact on its future levels. The considerable volatility persistence between the USA and South Africa is also indicated by the high B1(2,2) value.

4.6. Log Returns of the Stock Market Indices

Figure 1 depicts the log returns of various stock market indices in different countries during the COVID-19 pandemic. From Figure 1, it is noticeable that there was a volatility spike early in 2020, which coincides with the global onset of the pandemic. All six indices show a strong decline in returns during March 2020. Although the returns seem to stabilise post-March 2020, they cluster around the 0% mark. There is another volatility spike in early 2022, which is led by Russia, which can largely be attributed to the geopolitical tensions between Russia and Ukraine, which we later saw Russia invading Ukraine in February 2022 (Lin & Wang, 2024).

4.7. Granger Causality Test

The Granger causality test was used to investigate the causal relationship in the short run among the country-level EPU and stock markets in the analysis. The test was chosen because it examines the joint significance of each variable and its lags. With a p-value of 0.0067 and 0.0066, Brazil and India, respectively, were the only two countries that showed a predictive relationship with the EPU. In contrast, the null hypothesis of the Granger causality test for the other countries was accepted. The results in

Table 7. Granger Causality results for all countries.

Null Hypothesis:	F-Statistic	p-Value
SA EPU does not Granger Cause SA returns	1.31867	0.2921
Brazil EPU does not Granger Cause Brazil returns	6.58177	0.0067
Russia EPU does not Granger Cause Russia returns	0.61590	0.5506
India EPU does not Granger Cause India returns	6.61650	0.0066
China EPU does not Granger Cause China returns	1.96270	0.1679
USA EPU does not Granger Cause USA returns	0.82397	0.4538

. also show that Brazil and India’s equity markets are highly sensitive to changes in the EPU relative to the other countries. The other countries might be shielded from the immediate effects of changes in economic policy uncertainty by stronger domestic factors, bigger economies, or more robust financial systems.

5. Conclusions and Policy Implications

Broadly, the study aimed to analyse the effect of COVID-19 on the stock market volatility of five emerging countries and one developed country using GARCH models to draw comparisons. To achieve this objective, the study made use of daily log returns from 1 January 2020 to 31 March 2022 as a sampling period. Two multivariate GARCH models, viz., GARCH-M and diagonal BEKK-GARCH models, were employed to model volatility and to check the conditional variance in the variables. The GARCH-M model presented key insights into the individual countries’ ARCH and GARCH effects and the risk premium using both the mean and variance equations. From the mean equation of the GARCH-M model, the results showed that SA and China were perceived as the riskiest markets during the pandemic. To invest in these two markets in the midst of the pandemic, investors required more compensation to cater for the additional risk taken, which is consistent with the findings of a study conducted by Singh et al. (2024). The results from the variance equation highlighted each market’s volatility persistence and how each individual country responded to past shocks during the COVID-19 pandemic. While the USA and India showed the strongest volatility persistence, China emerged as the quickest country to dissipate volatility.

To analyse how volatility spills over from developed to emerging markets, the study employed the BEKK-GARCH model, pairing the USA with each country. The key findings from the results showed that all the stock markets had high persistence of volatility, particularly for Brazil and Russia, indicating that past volatility has a significant impact on future volatility. In addition, all the countries showed a high sensitivity to volatility shocks, with China and Russia being the most reactive. The evidence of spillover effects from the US to the emerging countries was small but statistically significant, with SA showing the strongest impact from US market shocks, which is in conformity with the studies conducted by Samitas et al. (2022) and N. Prasad et al. (2018).

From the Granger causality test, EPU appeared to have more significant impact in Brazil and India than in any other country, contrary to the results found from estimating the VAR model. The results underscore the need for the governments of these two nations to formulate policies that will mitigate the control that external risks like EPU have on their domestic economies, such as supporting more state-owned companies in their economies.

Overall, this study’s outcomes have great influence on policy implications. The evidence of spillover effects from the US to the BRIC countries essentially means that SA’s market regulators (SA being the most vulnerable) should monitor potential global risk channels (by potentially including global and domestic uncertainty indicators in their risk assessments), with China and the US being its biggest trade partners. Furthermore, the significance of the impact of EPU on both emerging and developing stock markets highlights the importance of policymakers having proper communication channels to mitigate the exacerbation of investor uncertainty and to stabilise the perception of investors, particularly during extreme market events.

From the study’s findings, it can be recommended that countries diversify their economies instead of having to rely on traditional sectors to buffer their vulnerability when there are unfavourable economic conditions, as such sectors tend to perform badly in global economic downturns. Moreover, strict and robust regulatory frameworks are imperative to ensure the stability of financial markets. Constant monitoring of global economic trends is also imperative for the early detection of potential economic and policy changes.

6. Limitations and Scope for Future Research

Since the study mainly focused on the BRIC countries and the USA, other studies can expand the analysis beyond these and include countries such as Turkey and the United Arab Emirates in order to assess the impact of COVID-19 on more diverse economic structures. This study’s sample size was only from 1 January 2020 to 31 March 2022; future research can widen this period to monitor the trends before, during, and after an extreme market event. Additionally, by making use of economic variables such as fiscal and monetary policy changes, the explanatory power of the models would be augmented when analysing the impact of economic policy uncertainty. Furthermore, other models like the GARCH-MIDAS and DCC-GARCH models could be used in modelling volatility to enhance the robustness of the results.