Global pandemic crisis and risk contagion in GCC stock markets

Abstract

This study investigates how the COVID-19 outbreak has shaped the volatility spillover between oil and Gulf Cooperation Council (GCC) stock markets. Contagion analysis is conducted by implementing a vector error correction (VECM) asymmetric BEKK model, wherein both cointegration and asymmetric features are considered. Financial market uncertainty caused by the recent health crisis is captured using Baker et al.’s (2020) newly developed infectious disease tracker. Our results indicate a significant discrepancy in the GCC group, as shock and volatility linkages between oil and equities are more apparent for some countries but not for others. The estimated VECM-asymmetric BEKK model reveals cross-market asymmetric spillover effects only in Kuwait, Qatar, and Saudi Arabia. We report that the global pandemic has strongly affected crude oil market volatility, while the GCC region seems to be less affected by the emergence of the new infectious disease. Our findings underscore the diversification opportunities offered by Gulf equity markets to international investors.

1. Introduction

This study aims to investigate how the volatility connectedness between oil and stock markets has changed after the onset of the coronavirus disease 2019 (COVID-19) crisis. We particularly focus on net oil-exporting countries—namely, the Gulf Cooperation Council (GCC) countries—because of the relevance of the research question for such countries depending completely on oil exports. The historic drop of the West Texas Intermediate (WTI) crude price—trading at approximately negative US$ 37 per barrel on April 20, 2020—has raised serious concerns for the Gulf region. Heavy reliance on hydrocarbon revenues leads to weaker fiscal and external positions, thereby exacerbating the economic cost of the current recession. As global financial markets have plummeted to unprecedentedly low levels since the Great Depression, equity markets in oil-exporting countries could face further challenges owing to their dependence on oil.

In fact, several recent studies have examined the connectedness between stock and oil markets in the COVID-19 era, thus showing the dramatic shutdowns of several indexes and demonstrating the importance of volatility spillovers between financial, commodity, and oil markets (Akhtaruzzaman et al., 2021; Ashraf, 2020; Corbet et al., 2020; Goodell, 2020; Jawadi and Sellami, 2021; Yousfi et al., 2021; Zhang et al., 2020). Nevertheless, studies on GCC countries as energy-dependent economies are relatively scarce. Most COVID-19 financial literature focuses on shocks and risk spillovers in the US and Chinese financial markets. Studying the volatility contagion in GCC countries during the unprecedented health crisis will serve as a decision support tool for investors. From an empirical viewpoint, the extant literature has used the daily data for total COVID-19 cases and related deaths that are publicly available at World Health Organization (WHO) platforms (see Al-Awadhi et al., 2020; Baig et al., 2021; Sharif et al., 2020; Xu, 2021).1 Additionally, some studies have used a dummy variable corresponding to key dates of the COVID-19 outbreak in China (see Corbet et al., 2020; Mirza et al., 2020).

To consider the role of pandemic-related development and its impact on financial market volatility, we consider the daily infectious disease equity market volatility (ID-EMV) index of Baker et al. (2020), who used a newspaper-based method to measure the impact of news about infectious disease outbreaks on equity market behavior. The daily changes in this variable can be attributed to public attention to infectious diseases and its effect on stock markets. Additionally, countries in the GCC group are heavily reliant on oil revenues. Their stock markets are expected to co-move with oil prices, as some common forces are likely to co-drive them over time. Thus, a cointegration analysis is conducted to test for the presence of a long-term relationship between stock market price indices and oil prices in the Gulf region. Furthermore, the overall average volatility in financial markets has increased tremendously following the outbreak of the global health crisis. Therefore, asymmetric volatility spillovers are expected to occur, as negative oil shocks would be more detrimental to the stock market than an equivalent shock before the crisis. We propose conducting a volatility contagion analysis using a vector error correction model (VECM) asymmetric BEKK model, wherein both cointegration and asymmetric features are considered. This framework allows us to examine how the dynamics of volatility transmission among the oil and equity prices have changed after the onset of the global pandemic. Our study covers the group of GCC countries (Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the UAE) from January 4, 2019, to January 29, 2021.

The rest of the paper is structured as follows: Section 2 reviews the existing literature on the dynamic interdependence between oil and GCC stock prices. Section 3 presents the data and discusses their properties. Section 4 outlines the empirical strategy for using the multivariate asymmetric BEKK model. Section 5 discusses the main empirical results. In Section 6, a robustness analysis is provided. Finally, conclusion and some policy implications are given in Section 7.

2. Empirical literature

A myriad of empirical studies has examined the dynamics of risk trnamission between different markets, particularly oil and equity markets (e.g., Agren, 2006; Hammoudeh et al., 2009; Singh et al., 2010; Awartani and Maghyereh, 2013; Tortorice, 2018; and Huisman et al., 2021). For the past two years, there has been a growing interest in the risk transmission phenomenon between oil and stock markets in the COVID-19 era (e.g., Akhtaruzzaman et al., 2021; Corbet et al., 2020; Goodell, 2020; Mensi et al., 2020; Salisu et al., 2020; Sharif et al., 2020; Yousfi et al., 2021; Zhang et al., 2020). The global economy has been significantly affected by the health emergency due to the shutdown of several oil and stock markets; for instance, the S&P500 index, DJI average, and NASDAQ index all fell to unprecedented levels.

Nevertheless, studies on GCC countries, such as net oil-exporting economies, remain sparse. In a recent study, Abuzayed and Al-Fayoumi (2021) investigated the change in oil price volatility transmission to the Gulf states during the recent health crisis. The authors implemented a dynamic conditional correlation (DCC) GARCH model and found that oil volatility spillover was more pronounced during the recent pandemic. Similarly, McMillan et al. (2021) examined the role of the crude oil market as a key determinant of US–GCC stock market connectedness. Using the asymmetric DCC-GARCH model, their results confirmed that oil shocks and volatility significantly explain changes in the US–GCC dependence; however, the role of the COVID-19 pandemic was neglected. Using the Kalman filter to generate oil price shocks, Refai et al. (2021) found that the equity market reaction was greater in Kuwait, Saudi Arabia, and the UAE on March 11, 2020, the day that the WHO declared COVID-19 as a pandemic.

Notwithstanding the apparent similarity, the empirical literature has highlighted the heterogeneous reactions of Gulf equity markets with respect to oil price volatility. According to Abuzayed and Al-Fayoumi (2021), vulnerability to extreme oil volatility seems to be more apparent in Saudi Arabia and the UAE than in other countries in the region. Fayyad and Daly (2014) compared the interdependence between the oil and Gulf stock markets to that of the UK and US markets within a vector autoregressive framework. The authors indicated that Qatar and the UAE had the highest equity market responsiveness among the GCC economies. Similarly, Alqahtani et al. (2019) examined the dependence of GCC equities to crude oil market uncertainty using the ARMA-DCC-EGARCH and time-varying Student-t copula models. Their main results show that the impact of oil price uncertainty differs across Gulf countries, with Oman and Bahrain exhibiting less sensitivity to stock returns. These insights are particularly useful for investors and portfolio managers seeking low-risk opportunities. Moreover, the analysis of causal links among the oil and equity prices remains mixed and inconclusive with respect to the causality direction in the region. Employing a panel Granger causality test, Arouri and Rault (2010) reported a two-way Granger causality for Saudi Arabia but found no causal relationship for the other Gulf countries.

Moreover, the extant empirical literature underlines the asymmetric impact of oil on stock markets as a feature of the GCC group. Using Shin et al.’s (2014) nonlinear ARDL approach, Fasanya et al. (2021) observed that most equity markets in the Gulf region responded asymmetrically to oil prices for the period from 1992–2016. According to Mohanty et al. (2011), oil price changes have an asymmetric impact on equity prices at both country and industry levels. However, some studies confirm our empirical results, which reveal that Arab Gulf nations do not react to oil price changes in the same way. Ben Cheikh et al. (2021) documented an asymmetric reaction of equity prices to oil shocks in some GCC nations but not in others. Our study seeks to further elucidate the possible asymmetry in the volatility contagion by using a multivariate asymmetric BEKK model.

Given the observed heterogeneity across the Gulf equity markets, the study of asymmetric volatility spillovers and the global pandemic context will serve as a decision support tool for investors. Alqahtani et al. (2019) examined the dependence of GCC stock markets on oil price uncertainty using a copula-based methodology. The authors underscored the presence of different levels of dependencies rather than homogenous profiles in the region. The Gulf countries can be classified into three groups, with Kuwait and Qatar markets having the most negative dependence vis-à-vis oil uncertainty. This grouping could offer some opportunities for international investors to benefit from the consequent diversification advantages. The ability to identify the heterogeneous reaction across GCC equities as collectively the largest global exporter of oil may present potential diversification opportunities. Furthermore, GCC linkage patterns in the crude oil market are expected to be different compared to net oil-importing countries, resulting in potential market segmentation and portfolio diversification opportunities. Insofar as Gulf economies exhibit a limited reaction to global shocks, such as the unprecedented COVID-19 health emergency, international investors and portfolio managers would be provided different investment and portfolio diversification possibilities.2

3. Empirical strategy

To examine the dynamic relationships among the oil and Gulf equity markets, we implement a VECM as follows:

$$\Delta o_t=c_1+{\alpha }_1{z}_{t-1}+\sum _{j=1}^p{a}_{1j}\Delta o_{t-j}+\sum _{j=1}^p{b}_{1j}\Delta s_{t-j}+d_{1}IDEM{V}_t+{\epsilon }_{1t}$$ (1)

$$\Delta s_t=c_2+{\alpha }_2{z}_{t-1}+\sum _{j=1}^p{a}_{2j}\Delta o_{t-j}+\sum _{j=1}^p{b}_{2j}\Delta s_{t-j}+d_{2}IDEM{V}_t+{\epsilon }_{2t}$$

where $\Delta {\mathrm{o}}_t$ and $\Delta s_t$ indicate the daily closing prices for equity and crude oil, respectively. $z_{t-1}$ is the lagged error correction term. $IDEM{V}_t$ is the daily ID-EMV index as developped by Baker et al. (2020). We introduce the ID-EMV index as an exogenous variable in the system. $c_i$, ${\alpha }_i$, $a_{ij}$, $b_{ij}$, and $d_i$ are the parameters to be estimated for $i=1,2$ and $j=1,\dots ,p$, where $p$ is the lag order in the system. The residuals ${\epsilon }_{1,t}$ and ${\epsilon }_{2,t}$, obtained from Equation (1), are then used as inputs in the next phase. A multivariate asymmetric GARCH model is then estimated, which allows for spillovers in volatility between oil and stock prices in the Gulf countries. Within this framework, the cross relationships enable GCC equities to be sensitive to oil price volatility asymmetry, even if no asymmetries exist in stock volatility. We introduce asymmetries into conditional volatility following the approach of Kroner and Ng (1998), who adapt Glosten et al.’s (1993) threshold GJR-GARCH model to a multivariate setting. Our empirical specification utilizes the VECM-asymmetric BEKK model, which has the following two-dimensional compacted form:

$$H_t=\left(C+D{x}_t\right){\left(C+D{x}_t\right)}^{\prime }+A^{\prime }{\epsilon }_{t-1}{\epsilon }^{\prime }{}_{t-1}A+B^{\prime }{H}_{t-1}B+G^{\prime }{\eta }_{t-1}{\eta }^{\prime }{}_{t-1}G$$ (2)

$\left(C+D{x}_t\right)$ is a $\left(2\times 2\right)$ lower triangular matrix, where $x_t$ is an exogenous variable representing the daily ID-EMV index. $A$ and $B$ are $\left(2\times 2\right)$ matrices that indicate the effects of past shocks and past volatility on current conditional variance. $H_t$ is a $\left(2\times 2\right)$ conditional covariance matrix. ${\epsilon }_{1t}$ and ${\epsilon }_{2t}$ are the unexpected shocks. ${\eta }_{1t}=\text{max}\left[0,-{\epsilon }_{1t}\right]$ and ${\eta }_{2t}=\text{max}\left[0,-{\epsilon }_{2t}\right]$ are the threshold terms that allow distinguishing between negative shocks. The unfolded covariance model is as follows:

$$\begin{array}{cc}\hfill \left[\begin{array}{cc}{h}_{11t}& h_{12t}\\ .& h_{22t}\end{array}\right]& =\left[\begin{array}{cc}{c}_{11}+d_{11}{x}_t& 0\\ c_{21}+d_{21}{x}_t& c_{22}+d_{22}{x}_t\end{array}\right]\left[\begin{array}{cc}{c}_{11}+d_{11}{x}_t& 0\\ c_{21}+d_{21}{x}_t& c_{22}+d_{22}{x}_t\end{array}\right]\hfill \\ & +{\left[\begin{array}{cc}{a}_{11}& a_{12}\\ a_{21}& a_{22}\end{array}\right]}^{\prime }\left[\begin{array}{cc}{\epsilon }_{1t-1}^2& {\epsilon }_{1t-1}{\epsilon }_{2t-1}\\ .& {\epsilon }_{2t-1}^2\end{array}\right]\left[\begin{array}{cc}{a}_{11}& a_{12}\\ a_{21}& a_{22}\end{array}\right]\hfill \\ & +{\left[\begin{array}{cc}{b}_{11}& b_{12}\\ b_{21}& b_{22}\end{array}\right]}^{\prime }\left[\begin{array}{cc}{h}_{11t-1}& h_{12t-1}\\ .& h_{22t-1}\end{array}\right]\left[\begin{array}{cc}{b}_{11}& b_{12}\\ b_{21}& b_{22}\end{array}\right]\hfill \\ & +{\left[\begin{array}{cc}{g}_{11}& g_{12}\\ g_{21}& g_{22}\end{array}\right]}^{\prime }\left[\begin{array}{cc}{\eta }_{1t-1}^2& {\eta }_{1t-1}{\eta }_{2t-1}\\ .& {\eta }_{2t-1}^2\end{array}\right]\left[\begin{array}{cc}{g}_{11}& g_{12}\\ g_{21}& g_{22}\end{array}\right]\hfill \end{array}$$ (3)

$G$ is a $\left(2\times 2\right)$ matrix, wherein the diagonal elements capture the asymmetric effect of previous shocks on the current conditional variance. The off-diagonal elements of $G$ measure the asymmetries across the markets. If parameter $g_{ij}$ is statistically significant, there is evidence of asymmetric risk contagion between markets $i$ and $j$. We use dots [.] to indicate that some matrices are symmetric. Within the abovementioned VECM-asymmetric BEKK model, the risk contagion is also examined by depicting the time-varying conditional correlations over the estimation period.3

4. Data and their properties

The dataset used here comprises daily GCC stock market indices (Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and UAE) and Brent spot prices in US dollars from January 4, 2019, to January 29, 2021. As GCC stock markets are closed on Fridays and Saturdays, the data used here correspond to the four common trading days for these markets: Monday through Thursday. The daily closing prices for equity and crude oil are extracted from Thomson Reuters DataStream. We compute the daily returns for the first logarithmic differences in daily prices. We report the summary statistics of the daily returns for the entire period in Table 1 .

Table 1.

Main statistics for daily data

	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE	Brent
Mean	0.063	0.059	0.001	0.026	0.027	0.052	0.004
Median	0.054	0.104	0.000	0.027	0.117	0.079	0.291
Min.	-8.001	-19.035	-9.369	-12.533	-16.755	-15.757	-37.493
Max.	2.413	5.082	4.406	3.497	6.831	7.508	19.077
Std. Dev.	0.726	1.552	0.848	1.187	1.541	1.679	4.001
Skewness	-3.531	-5.397	-2.944	-3.099	-3.674	-2.630	-2.584
Kurtosis	39.173	61.408	38.725	31.680	37.489	26.970	26.782
Jarque–Bera	26736***	65600***	25890***	17584***	24627***	12741***	12554***
$Q\left(20\right)$	33.332***	58.013***	22.082**	13.694***	65.011**	31.965***	22.735**
ARCH(20)	18.248**	45.876***	16.344**	22.851**	25.208***	25.357***	20.388***
Correlation with Brent	0.30	0.43	0.29	0.41	0.43	0.39	1.00
Zivot–Andrews test
Level	-3.719	-4.386	-4.296	-3.576	-3.843	-4.478	-3.743
First difference	-7.419***	-7.713***	-6.242***	-9.261***	-8.585***	-7.210***	-8.546***
Lee–Strazicich test
Level	-3.318	-4.050	-3.955	-3.209	-3.562	-4.396	-3.565
First difference	-8.574***	-9.821***	-17.557***	-20.778***	-9.630***	-9.533***	-9.594***

Notes: *** $p<0.01$; ** $p<0.05$; * $p<0.1$.

As shown in Table 1, the daily returns of Gulf equity markets are close to zero on average over the entire sample period as a result of the health crisis. On average, Oman exhibits the worst performance (0.001%) in the region, which is very close to the weak returns of the crude oil market (0.004%). The highest average returns are recorded in Bahrain, Kuwait, and the UAE. Among the Gulf nations, the highest risk—indicated by the standard deviation measure—is found in the UAE, whereas the Bahraini market is the least volatile. The Jarque–Bera test statistics reject the null hypothesis of normality for all series at the 1% significance level. This is confirmed by the negative skewness and high kurtosis for most return series. There is evidence of residual autocorrelation using Ljung and Box's (1978) test. Similarly, ARCH effects are found using Engle's (1982) test for conditional heteroskedasticity. Finally, both the Zivot and Andrews (1992) and Lee and Strazicich (2003) tests are conducted to check for unit roots in oil and stock prices in levels and first differences. Table 1 indicates that price indices are integrated of order 1 and confirms the stationarity for all return (first differences) series. Finally, Table 1 displays the unconditional correlations between oil price changes and equity returns. Although cross-market co-movements are positive, they are not high and do not exceed 0.50 of the correlation coefficient across the Gulf countries. The highest correlation is found in Kuwait and Saudi Arabia, which are the most oil-reliant in the region (Ben Cheikh et al., 2018).

Our empirical specification considers the daily ID-EMV index developed by Baker et al. (2020) to explore the explanatory ability of infectious disease pandemics, such as that of COVID-19, on daily volatility. More specifically, we use the ID-EMV index to gauge the impact of the COVID-19 crisis on the connectedness between GCC equity prices and the crude oil market. The ID-EMV is a newspaper-based index of financial market uncertainty caused by infectious diseases. Instead of focusing on the shocks of isolated public health incidents, this index will allow us to quantify and gain an overall view of the global infectious disease pandemic over time.4 Data on the daily ID-EMV index are obtained from the following website: http://www.policyuncertainty.com/infectious_EMV.html. As shown in Fig. 1 , the outbreak of the COVID-19 pandemic corresponds to higher values of the ID-EMV index as an indication of an increasing number of news reports on this unprecedented public health shock to financial markets. The increased volatility is more apparent since the WHO declared COVID-19 as a public health emergency on February 20, 2020, and then as a global pandemic on March 11, 2020. Simultaneously, the GCC stock markets exhibited a sharp collapse following the eruption of the infectious disease pandemic. Fig. 2 shows that the time evolutions of the oil and Gulf stock prices are quite similar, which can indicate more integrated markets during the health crisis episode.

Fig. 1.

Daily ID-EMV tracker.

Notes: The ID-EMV tracker is a newspaper-based index constructed by Baker et al. (2020), which is used to measure the impact of news about infectious disease outbreaks on equity market behavior. The data cover the period from January 04, 2019, to January 29, 2021.

Fig. 2.

Time series plots of Brent oil and GCC stock prices.

Notes: The plots displayed above comprise daily GCC stock market indices and Brent spot prices from January 4, 2019, to January 29, 2021. The daily closing prices for equity and crude oil are extracted from Thomson Reuters DataStream.

5. Empirical results

5.1. Cointegration analysis

We first conducted Johansen's 1988, 1991, 1995) cointegration tests before estimating the mean model specified in Equation ((1). The number of cointegrating vectors ($r$) in the system is determined by employing the widely used trace, ${\lambda }_{trace}\left(r\right)$, and the maximum eigenvalue, ${\lambda }_{max}\left(r\right)$, tests. To ensure the robustness of the presence of a possible cointegrating relationship, we used Johansen et al.’s (2000) test, wherein the standard trace tests are extended to allow for trend and level breaks at several known points. Johansen et al. (2000) provided two versions of the trace test. $H_l\left(r\right)$ and $H_c\left(r\right)$ for when there are ($q--1$) breaks in a linear trend and constant trend, respectively, where $q$ represents subsamples. As a break point, we considered March 11, 2020, the day that the WHO declared COVID-19 as a pandemic.5

We used different information criteria to determine the optimal lag length to be introduced in the VECM. However, the information criteria, such as Akaike information criterion (AIC) or Bayesian information criterion (BIC), are not designed for heteroscedastic data and hence can be sensitive to outliers.6 For comparison purposes between Gulf economies, a VECM with four lags is estimated for each country in the region (see Sadorsky, 2012). The results of the cointegration tests are reported in Table 2 . The trace and maximal eigenvalue tests revealed the presence of one cointegrating vector, as the null hypothesis of no cointegration was rejected for all the GCC members. As expected, oil prices and Gulf equities were co-driven by common stochastic forces.

Table 2.

Cointegration test results.

	$H_0$: rank = $r$	${\lambda }_{trace}\left(r\right)$	${\lambda }_{max}\left(r\right)$	$H_l\left(r\right)$	$H_c\left(r\right)$	Cointegration vector
						${\beta }^{\prime }=\left(1,{\beta }_1,{\beta }_2\right)$
Bahrain	$r=0$	16.314**	15.521**	17.115**	15.721**	1	0.496***	8.560***
							(0.013)	(0.055)
	$r=1$	3.307	3.307	5.137	5.047
Kuwait	$r=0$	17.919**	15.663**	18.543***	16.437**	1	0.046***	2.369***
							(0.008)	(0.038)
	$r=1$	2.256	2.256	3.154	2.977
Oman	$r=0$	17.410**	16.127**	17.193**	15.261**	1	0.059***	8.560***
							(0.005)	(0.022)
	$r=1$	3.681	3.681	3.988	3.181
Qatar	$r=0$	18.126***	17.245***	19.235***	17.271***	1	0.321***	5.909***
							(0.012)	(0.054)
	$r=1$	2.880	2.880	3.445	3.180
Saudi Arabia	$r=0$	19.386***	18.361***	21.325***	17.481***	1	0.059***	8.688***
							0.008	0.037
	$r=1$	3.0243	3.024	3.198	2.902
UAE	$r=0$	19.086**	18.102***	25.375**	17.279**	1	0.241***	3.960***
							0.007	0.031
	$r=1$	1.390	1.390	2.569	2.249

Notes: *** $p<0.01$; ** $p<0.05$; * $p<0.1$.

Next, we examined the estimates from the VECM, which represents the conditional mean equations in the system, as shown in Table 3 . The error correction model enabled us to identify the return spillovers among the oil and equity markets and conduct causality analysis. Table 3 confirms the significant impact of oil price changes on Gulf equities, except for the UAE, at the 5% significance level: The lagged oil price changes significantly impact the current equity prices at the 1% significance level for Kuwait, Oman, Qatar, and Saudi Arabia. This outcome is unsurprising, as Bahrain and the UAE are the least reliant on oil in the Gulf region.

Table 3.

Estimates of the VECM.

Explanatory Variable	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE
	Dependent Variable: Oil prices
Constant	6.306***	4.592***	5.615***	3.832***	7.170***	-0.121**
	(1.447)	(1.529)	(2.242)	(0.672)	(2.090)	(0.059)
$\Delta {\mathrm{o}}_{t-1}$	-0.038*	-0.039	-0.037	-0.041**	-0.051**	-0.000
	(0.023)	(0.024)	(0.024)	(0.021)	(0.023)	(0.030)
$\Delta {\mathrm{o}}_{t-2}$	-0.019	-0.021	-0.027	-0.038*	-0.040*	-0.007
	(0.024)	(0.024)	(0.024)	(0.022)	(0.022)	(0.030)
$\Delta {\mathrm{o}}_{t-3}$	0.027	0.026	0.018	0.021	0.003	0.044
	(0.024)	(0.022)	(0.022)	(0.022)	(0.022)	(0.030)
$\Delta {\mathrm{o}}_{t-4}$	0.046**	0.037*	0.036*	0.041**	0.039*	0.035
	(0.022)	(0.022)	(0.020)	(0.019)	(0.021)	0.030)
$\Delta {\mathrm{s}}_{t-1}$	0.082	-0.435***	-0.040	6.121***	-0.010	0.026
	(0.053)	(0.149)	(0.057)	(0.042)	(0.040)	(0.075)
$\Delta {\mathrm{s}}_{t-2}$	0.041	0.458***	0.007	-1.858***	0.023	0.010
	(0.058)	(0.145)	(0.054)	(0.039)	(0.039)	(0.075)
$\Delta {\mathrm{s}}_{t-3}$	-0.147***	-0.239*	-0.006	-7.390***	0.036	-0.005
	(0.051)	(0.144)	(0.052)	(0.039)	(0.040)	(0.074)
$\Delta {\mathrm{s}}_{t-4}$	-0.020	0.038	-0.006)	3.567***	0.042	0.027
	(0.052)	(0.043)	(0.056)	(0.039)	(0.035)	(0.074)
$\text{IDEM}{\mathrm{V}}_{t-1}$	-0.039***	-0.019**	-0.085***	-0.195***	-0.018**	-0.012**
	(0.014)	(0.009)	(0.034)	(0.015)	(0.009)	(0.001)
$z_{t-1}$	-0.164***	-0.126***	-0.027**	-0.038***	-0.120***	0.002**
	(0.038)	(0.042)	(0.012)	(0.013)	(0.035)	(0.000)
	Dependent Variable: Stock prices
Constant	-0.046	0.888	1.055***	0.009	-0.434	0.054**
	(0.285)	(0.627)	(0.007)	(0.001)	(1.151)	(0.024)
$\Delta {\mathrm{o}}_{t-1}$	0.001**	0.022***	0.031***	0.025***	0.037***	0.009
	(0.001)	(0.009)	(0.005)	(0.003)	(0.010)	(0.012)
$\Delta {\mathrm{o}}_{t-2}$	0.012***	-0.003	0.002	0.001***	-0.018*	0.020*
	(0.003)	(0.008)	(0.005)	(0.000)	(0.010)	(0.012)
$\Delta {\mathrm{o}}_{t-3}$	-0.002	0.007	0.005	0.000	0.004	0.009
	(0.005)	(0.008)	(0.005)	(0.000)	(0.010)	(0.012)
$\Delta {\mathrm{o}}_{t-4}$	0.009**	0.009	0.007*	0.000	-0.000	0.021*
	(0.004)	(0.008)	(0.004)	(0.000)	(0.010)	(0.012)
$\Delta {\mathrm{s}}_{t-1}$	-0.040*	0.061***	0.113***	1.041	0.095***	0.049
	(0.021)	(0.022)	(0.023)	(0.000)	(0.021)	(0.030)
$\Delta {\mathrm{s}}_{t-2}$	0.040	0.006	0.084***	-0.025	0.034	-0.013
	(0.025)	(0.023)	(0.022)	(0.000)	(0.023)	(0.030)
$\Delta {\mathrm{s}}_{t-3}$	0.028	0.024	-0.065***	-0.011	0.002	0.014
	(0.020)	(0.021)	(0.021)	(0.000)	(0.022)	(0.030)
$\Delta {\mathrm{s}}_{t-4}$	0.022	0.053**	0.013	-0.008	0.007	0.083***
	(0.020)	(0.022)	(0.016)	(0.000)	(0.021)	(0.030)
$\text{IDEM}{\mathrm{V}}_{t-1}$	0.002	-0.024	-0.016***	0.034***	0.008	-0.001
	(0.008)	(0.017)	(0.000)	(0.001)	(0.019)	(0.000)
$z_{t-1}$	-0.004	-0.007	-0.006***	-0.000	-0.000	0.000
	(0.003)	(0.005)	(0.001)	(0.000)	(0.004)	(0.000)
Residual analysis
$Q\left(20\right)$	5.260	7.572	9.885	9.384	21.977	24.362
	[0.999]	[0.994]	[0.970]	[0.978]	[0.341]	[0.220]
$Q^2\left(20\right)$	74.661***	72.655***	92.456***	76.864***	99.015***	87.548***
	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]
ARCH(20)	68.239***	71.348***	89.070***	71.996***	95.068***	84.024***
	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]

Granger causality tests
$H_0$: ${\gamma }_{1,1}={\gamma }_{1,2}={\gamma }_{1,3}={\gamma }_{1,4}=0$	4.647***	4.315***	3.326**	7.664***	4.243***	9.761***
	[0.000]	[0.004]	[0.018]	[0.000]	[0.001]	[0.000]
$H_0$: ${\beta }_{2,1}={\beta }_{2,2}={\beta }_{2,3}={\beta }_{2,4}=0$	1.455	1.694	1.284	1.563	0.753	0.654
	[0.223]	[0.165]	[0.277]	[0.177]	[0.555]	[0.611]

Note: Results are obtained from the VECM specification as in Equation (1) over the period from January 4, 2019 to January 29, 2021. The Fig.s in parentheses are standard errors, and those in square brackets are $p$-values. *** $p<0.01$; ** $p<0.05$; * $p<0.1$.

Additionally, the impact of the ID-EMV index on the crude oil market is significant and robust across the different specifications. Financial market uncertainty due to infectious diseases significantly negatively impacts oil price changes. However, there is weak evidence that stock returns react significantly to unprecedented public health emergencies. Only Oman and Qatar seem to exhibit a negative response to the COVID-19 pandemic. As is well known, GCC financial markets are more segmented and less integrated with global markets than are those in other developed or emerging economies. International investors still have restricted access to Gulf stock markets, and liquidity remains low; thus, a lesser effect of the global health crisis is expected for the region.

The residual analysis presented in Table 3 indicates that heteroskedasticity remains, whereas serial autocorrelation disappears. Moreover, the Granger causality tests show some evidence of unidirectional causality from oil to equity prices in some countries in the region. Our results contrast with those of Hammoudeh and Choi (2006), who find no causality between oil price and any GCC stock market using the VECM framework. Notably, a few previous studies have provided evidence of causal relationship from the Saudi stock markets to oil prices (see Arouri and Rault, 2010; Basher, 2006; Hammoudeh and Aleisa, 2004). Given its major role in worldwide energy markets, it is not surprising that changes in Saudi Arabia's economic conditions would cause changes in oil prices.

5.2. Volatility contagion analysis

Next, we examined the shock and volatility transmission among oil and GCC equity prices using the VECM-asymmetric BEKK model. The estimated conditional volatility equations and standardized residual analysis results are displayed in Tables 4 and 5, respectively. The parameters of our multivariate GARCH models were estimated via the quasi-maximum likelihood estimation using the Broyden–Fletcher–Goldfarb–Shanno algorithm. The diagnostic tests in Table 5 show standardized residuals from the multivariate asymmetric GARCH, which indicate that autocorrelation and heteroskedasticity issues have disappeared.

Table 4.

Estimation results from the VECM-asymmetric BEKK model.

	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE
$c_{11}$	0.049	-0.104	0.131*	1.028***	1.006***	0.205***
	(1.021)	(1.006)	(0.080)	(0.038)	(0.043)	(0.053)
$c_{12}$	-0.028	-0.169**	0.089	0.000	-1.081***	-0.001**
	(0.028)	(0.087)	(0.056)	(0.001)	(0.041)	(0.001)
$c_{22}$	0.198***	1.074***	1.151***	0.003***	1.005***	0.004***
	(0.022)	(0.053)	(0.026)	(0.000)	(0.030)	(0.000)
$d_{11}$	0.034***	0.050***	0.037***	0.033***	0.019***	0.030***
	(0.004)	(0.006)	(0.005)	(0.004)	(0.005)	(0.004)
$d_{21}$	0.010***	0.034	0.012***	0.000***	0.014***	0.000***
	(0.002)	(0.005)	(0.003)	(0.000)	(0.004)	(0.000)
$d_{22}$	-0.015***	-0.004	-0.017***	-0.000***	0.001	0.000
	(0.002)	(0.004)	(0.003)	(0.000)	(0.003)	(0.000)
$a_{11}$	-0.021	0.055**	0.091**	0.137***	0.081**	0.083***
	(1.040)	(0.028)	(0.042)	(0.028)	(0.043)	(0.034)
$a_{12}$	0.011	-0.020	-0.008	-0.001***	-0.104***	0.001
	(1.016)	(1.021)	(0.022)	(0.000)	(0.014)	(0.001)
$a_{21}$	0.081*	0.220***	0.084	0.507***	-0.273***	-1.715
	(0.045)	(0.044)	(1.069)	(0.131)	(0.035)	(3.059)
$a_{22}$	0.355***	0.349***	-0.217***	0.239***	0.104***	-0.245***
	(0.025)	(0.027)	(0.051)	(0.028)	(0.036)	(0.037)
$b_{11}$	0.921***	1.013***	0.967***	1.009***	0.949***	0.964***
	(0.005)	(0.004)	(0.006)	(0.006)	(0.007)	(0.006)
$b_{12}$	0.000	0.007	0.004	0.000	0.003	0.001
	(0.002)	(0.006)	(0.003)	(0.000)	(0.005)	(0.001)
$b_{21}$	-0.016	0.117***	0.116***	0.433**	0.096***	-2.174
	(1.019)	(0.019)	(0.030)	(0.251)	(0.027)	(1.814)
$b_{22}$	0.902***	0.865***	0.839***	0.888***	0.867***	0.863***
	(0.014)	(0.018)	(0.016)	(0.012)	(0.013)	(0.016)
$g_{11}$	0.340***	0.305***	0.306***	0.316***	0.274***	0.332***
	(0.022	(0.025)	(0.024)	(0.027)	(0.029)	(0.025)
$g_{12}$	0.014	0.030	0.005	0.002	0.048**	0.000
	(0.009	(0.028)	(0.011)	(0.003)	(0.022)	(0.001)
$g_{21}$	0.100	0.097***	0.207	0.358***	0.084***	0.487
	(0.068	(0.003)	(0.166)	(0.085)	(0.020)	(0.759)
$g_{22}$	0.036	0.149**	0.547***	0.407***	0.511***	0.412***
	(0.056)	(0.076)	(0.039)	(0.036)	(0.036)	(0.044)
Testing cross-variance effects:
$H_0$: $a_{12}=a_{21}=b_{12}=b_{21}=0$	64.126	36.155	54.326	71.234	47.383	26.841
	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]
Testing asymmetric variance:
$H_0$: $g_{11}=g_{12}=g_{21}=g_{22}=0$	38.987	34.454	46.244	69.632	30.823	29.542
	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]	[0.000]
Log-likelihood	-6493.368	-6927.122	-6455.751	-6849.143	-7274.694	-6665.083
AIC	5.474	5.089	6.608	4.783	7.441	7.253
BIC	5.534	5.180	6.722	4.842	7.555	7.346

Note: The results are obtained from the BEKK-GARCH model as in Equation (2). The Fig.s in parentheses are standard errors, and those in square brackets are $p$-values. *** $p<0.01$; ** $p<0.05$; * $p<0.1$.

Table 5.

Standardized residuals diagnostic tests.

	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE
$Q\left(20\right)$	25.548	17.040	23.664	24.575	23.090	24.002
	(0.181)	(0.626)	(0.257)	(0.218)	(0.284)	(0.242)
$Q^2\left(20\right)$	6.421	8.185	14.115	15.292	10.988	13.996
	(0.998)	(0.990)	(0.824)	(0.759)	(0.946)	(0.830)
ARCH(20)	0.348	0.402	0.705	0.753	0.540	0.696
	(0.996)	(0.991)	(0.820)	(0.765)	(0.950)	(0.833)

Note: $Q\left(20\right)$ and $Q^2\left(20\right)$ are Ljung and Box (1978) autocorrelation tests of order 20. ARCH(20) is the Engle (1982) test for conditional heteroskedasticity of order 20.

The estimated coefficients for the ARCH and GARCH terms from the VECM-asymmetric BEKK are referred to by the coefficients $a_{ij}$ and $b_{ij}$, respectively. In most cases, own conditional ARCH effects are statistically significant. Similarly, own conditional GARCH effects are crucial in explaining conditional volatility and are statistically significant at the 1% level in each multivariate asymmetric GARCH model. Notably, the conditional volatilities of these stock markets are more affected by their own lagged conditional volatility than by their lagged shocks. For the GCC group, the past values of the conditional variance in a given Gulf financial market are crucial for forecasting future volatility.

Thereafter, we analyzed the nature of the risk spillover between the oil and GCC stock markets, namely, how shocks and volatility are transmitted. Table 4 shows that when testing cross-variance effects, that is, $H_0$: $a_{12}=a_{21}=b_{12}=b_{21}=0$, restrictions are clearly rejected. This confirms that the cross-relationships across all conditional moments and their shocks cannot be ignored. Specifically, our results indicate significant shock spillovers from oil prices to the GCC stock markets, except for Oman and the UAE. Table 4 shows the bidirectional shock transmission for Qatar and Saudi Arabia, with negative spillovers between Brent oil prices and the Saudi market. For volatility connectedness, we find unidirectional connectedness from oil to most Gulf countries, except for Bahrain and the UAE.

Our findings are consistent with Arouri et al. (2011), who reported that the risk spillover runs from oil to GCC equity prices. The current conditional volatility of stock prices is positively impacted by the past volatility of the crude oil market in the region because of its heavy reliance on hydrocarbons for export earnings and fiscal revenues. However, Bahrain and the UAE are the relatively less oil-dependent GCC countries in the region (see Ben Cheikh et al., 2021). Using the VAR-BEKK-GARCH model, Jouini and Harrathi (2014) revealed that the oil and stock market volatility connectedness is very weak in Bahrain and the UAE.

Additionally, our VECM-asymmetric BEKK model enables us to examine the asymmetric reaction of conditional volatility to negative shocks. First, when testing $H_0$: $g_{11}=g_{12}=g_{21}=g_{22}=0$, the restrictions on asymmetric cross-variance effects are strongly rejected, indicating the presence of significant asymmetries. Specifically, asymmetric responses to negative own shocks are highly significant in the crude oil market. This is not surprising, as the COVID-19 pandemic has dramatically impacted oil prices, which plunges into the negative territory.7 Asymmetries in the negative own shocks are also apparent for most GCC countries, except for Bahrain. For the cross-market asymmetric responses between oil and stock prices, positive unidirectional spillovers running from oil to equity prices are found only in Kuwait and Qatar. For the Gulf region, Awartani and Maghyereh (2013) analyze the return and volatility connectedness among stock markets and oil prices using the Diebold and Yilmaz (2009, 2012) approach. The authors find asymmetric shock and volatility transmission from oil to equities in the GCC group, while the flow of information in the opposite direction is marginal. In the case of our study, among GCC stock indices, there is evidence of bidirectional asymmetric interdependencies between oil and Saudi equity prices. Again, the result is consistent with some previous studies on the Gulf countries where evidence of two-way Granger causality among oil and Saudi stock indices is found (see, e.g., Arouri and Rault, 2010).

Furthermore, within our multivariate GARCH setting, we intend to gauge the role of financial market uncertainty related to the recent COVID-19 health crisis. As Table 4 shows, the impact of the ID-EMV index is highly significant and positive on the permanent volatility of the crude oil market at the 1% significance level. This confirms the role of the recent pandemic in exacerbating the decline in the global demand for oil. However, the GCC region seems to be less affected by the infectious diseases-based financial market uncertainty, as stock price fluctuations are increasing only in Qatar and Saudi Arabia. For the remaining Gulf nations, namely, Bahrain, Kuwait, Oman, and the UAE, their stock markets appear to be more stable during public health emergencies. These insights are particularly relevant for investors and portfolio managers seeking low-risk opportunities during turbulent periods.

The properties of the conditional second moment of the oil and stock markets are also depicted in Fig. 3 , where the estimated conditional variance is plotted over a longer time period from January 4, 2010, to January 29, 2021. In fact, the crude oil market has experienced massive fluctuations and boom–bust cycles in recent years, namely, the great plunge of 2014–2015, which we intend to consider in the volatility contagion analysis. Although the oil and GCC stock markets appear to exhibit similar patterns, the Brent oil series exhibits volatility that is almost always above the other indices, especially since the beginning of the global pandemic. The time-varying conditional volatility of the crude oil market is highly unstable and fluctuates substantially during the global health emergency.

Fig. 3.

Time-varying conditional volatility in the oil and GCC stock markets.

Notes: The plots displayed above comprise time-varying conditional volatilities of oil and GCC stock returns estimated from Equation (2) over the period from January 4, 2010, to January 29, 2021. The lightly shaded regions in the graphs cover the COVID-19 pandemic episode.

Furthermore, using the VECM-asymmetric BEKK model, we plotted the time-varying conditional correlations between oil and stock prices in the region over the same time period. In Fig. 4 , it is evident that the dynamic conditional correlations vary significantly. Even if the correlations are still in the positive range most of the time, they alternate in sign. The eruption of the COVID-19 pandemic marked greater levels of correlation compared to other episodes, exceeding the value of 0.5, in all the Gulf countries. The different plots revealed that dynamic conditional correlations hit their maximum in March 2020, as the impact of the unprecedented health crisis was felt. It is interesting that there is a decreasing trend in the dynamic correlation in the post-COVID-19 era in Bahrain, Kuwait, and Saudi Arabia. Conditional correlations reached low levels during the summer of 2020 and even entered the negative range for the case of Bahrain. For only a few periods after the COVID-19 outbreak have these GCC stock markets offered an opportunity for meaningful portfolio diversification. Moreover, the dramatic oil crash of 2014–2015 seems to entail a dynamic correlation pattern that is comparable to that of the COVID-19 period, particularly for Kuwait and Oman.

Fig. 4.

Time-varying conditional correlations from the VECM-asymmetric BEKK model.

Notes: The plots displayed above consist of time-varying conditional correlations among oil and GCC stock prices obtained from Equation (2) over the period from January 4, 2010, to January 29, 2021. The lightly shaded regions in the graphs cover the COVID-19 pandemic episode.

5.3. Robustness analysis

In this section, the reliabilities of our results from the VECM-asymmetric BEKK model are compared and contrasted using different specifications. As empirical results achieved in the previous section may be specific to the setup of the model or the selected time period, we investigated the robustness of the results in three different ways. First, we experimented with the two restricted correlation models—namely, constant conditional correlation (CCC) and DCC specifications—to test whether the risk spillover of the dynamic interdependence between oil and equity prices is still robust. Second, we introduced the daily ID-EMV index in the multivariate GARCH setting as an endogenous variable. Third, we conducted a comparative analysis between the 2008 Global Financial Crisis (GFC) and the recent global health emergency by estimating our VECM-asymmetric BEKK model over the period from January 1, 2007 to December 31, 2009.

Tables A1 and A2 in the Appendix show the results from the CCC- and DCC-GARCH models, which reveal that shock and volatility transmissions from oil to GCC equity prices are still robust. Similarly, the presence of asymmetric volatility reactions of conditional volatility to negative shocks is apparent in the oil—and most GCC—stock markets. Moreover, the CCC and DCC specifications confirm that the recent global pandemic has affected the GCC region less. Finally, based on the information criteria, namely, AIC and BIC, the BEKK model appears to be a closer fit to the DCC specification. Similarly, comparing the log-likelihood functions reveals that both BEKK and DCC models provide a much better fit than the CCC model. Chang et al. (2011) and Arouri et al. (2011) reached the same conclusion, revealing the superior ability of BEKK-GARCH model over the restricted correlation models. Next, we examined the robustness of our results by considering the possible endogeneity of the daily ID-EMV index. In our benchmark model, we introduced financial market uncertainty due to infectious diseases as an exogenous variable in the system. To check the reliability of our results, we estimated a trivariate asymmetric BEKK-GARCH model; herein, as displayed in Table A3 in the Appendix, the ID-EMV index is considered to be endogenous. Again, the dynamics of return and volatility spillovers remain consistent with the evidence obtained from our benchmark bivariate model.

Finally, we investigated the risk spillover effects during another period of turmoil, the 2008 GFC, and whether volatility transmission is as prominent as that during the COVID-19 crisis. Table 6 reports the results from the re-estimation of our VECM-asymmetric BEKK model from January 1, 2007, to December 31, 2009.8 Although there is evidence of risk spillover between the oil and Gulf equities during the 2008 financial crisis, the unprecedented COVID-19 global health crisis seems to trigger higher risk transmission in the GCC region. As expected, the 2020 lockdown entailed higher uncertainty levels and stronger risk spillovers. Moreover, noteworthily, the presence of asymmetries in volatility transmission is less apparent during the 2008 GFC compared to the COVID-19 outbreak, where bad volatility is transmitted more than good volatility. Our results are in line with those of Jebabli et al. (2021), who analyzed the volatility contagion between stock markets and oil during different crisis periods. Using Diebold and Yilmaz's (2012) approach, the authors pointed out that risk transmission during the recent health crisis exceeded that recorded throughout the 2008 financial turmoil.

Table 6.

Estimation of the VECM-asymmetric BEKK model during the GFC.

	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE
$c_{11}$	-0.055	1.005	1.056	0.104***	0.151***	0.018
	(1.042)	(1.051)	(1.046)	(0.036)	(0.046)	(1.038)
$c_{12}$	0.224***	-0.163**	-0.249***	-0.161***	-0.342***	-0.465***
	(0.030)	(0.077)	(0.060)	(0.041)	(0.040)	(0.029)
$c_{22}$	0.001	-0.072	0.084	0.054	0.051	0.237***
	(1.172)	(0.200)	(1.118)	(0.075)	(1.109)	(0.046)
$a_{11}$	0.014	0.106***	0.071**	0.115***	0.064***	0.134***
	(0.044)	(0.039)	(0.034)	(0.029)	(0.030)	(0.041)
$a_{12}$	0.000	0.110	-0.018	-0.028	-0.100***	0.032
	(0.009)	(1.153)	(0.017)	(0.025)	(0.031)	(0.026)
$a_{21}$	0.057	0.110**	0.166	-0.034	0.142***	-0.012
	(1.050)	(0.053)	(0.145)	(0.032)	(0.028)	(0.031)
$a_{22}$	0.340***	0.249***	-0.167**	0.322***	0.090*	-0.092*
	(0.038)	(0.035)	(0.071)	(0.037)	(0.054)	(0.053)
$b_{11}$	0.949***	0.978***	1.009***	0.973***	1.012***	0.978***
	(0.003)	(0.006)	(0.005)	(0.005)	(0.007)	(0.005)
$b_{12}$	0.001	0.004	0.009	0.003	0.032	0.016
	(0.002)	(0.004)	(1.014)	(0.004)	(0.029)	(0.026)
$b_{21}$	-0.007	0.009	-0.026	0.033***	0.021***	-0.017
	(1.021)	(0.026)	(1.027)	(0.012)	(0.009)	(0.015)
$b_{22}$	0.907***	1.007***	0.837***	0.900***	0.830***	0.845***
	(0.018)	(0.009)	(0.015)	(0.014)	(0.020)	(0.022)
$g_{11}$	0.245***	0.223***	0.128***	0.224***	0.254***	0.214***
	(0.026)	(0.037)	(0.031)	(0.034)	(0.035)	(0.037)
$g_{12}$	0.019	0.005	-0.002	0.014	0.007	0.047
	(0.014)	(0.038)	(1.018)	(1.020)	(0.032)	(0.038)
$g_{21}$	0.013	0.254	0.076	-0.029	0.142	0.007
	(1.0152)	(0.169)	(0.077)	(1.048)	(0.250)	(0.028)
$g_{22}$	-0.067	0.262***	0.349***	0.429***	0.607***	0.540***
	(0.063)	(0.051)	(0.062)	(0.051)	(0.056)	(0.048)
Log-Likelihood	-3173.506	-3289.159	-3123.972	-3402.367	-3459.908	-3651.695
AIC	6.545	6.780	6.444	7.011	7.128	7.519
BIC	6.744	6.980	6.643	7.210	7.327	7.718

Note: The results are obtained from the asymmetric BEKK model (ID-EMV index excluded) from January 1, 2007, to December 31, 2009. The numbers in parentheses are standard errors. *** $p<0.01$; ** $p<0.05$; * $p<0.1$.

6. Concluding remarks

In this study, we examined how the volatility transmission between the oil and GCC stock markets changed after the onset of the global health crisis. We captured the COVID-19 outbreak's impact using Baker et al.’s (2020) newspaper-based infectious disease index. This indicator captures the increase in the financial uncertainty related to infectious diseases. We conducted volatility contagion analysis by implementing a VECM-asymmetric BEKK model, in which both cointegration and asymmetric features were considered. Our results indicate a significant discrepancy in the GCC group, as return and volatility linkages between the oil and equity markets are more apparent for some countries but not for others. The estimated VECM-asymmetric BEKK model revealed the presence of asymmetries in the conditional variances with respect to negative own shocks. However, for the transmission across the oil and stock markets, we found asymmetric spillover effects only in Kuwait, Qatar, and Saudi Arabia. Clearly, the global COVID-19 pandemic has significantly affected crude oil market volatility, while the GCC region seems to be less affected by the onset of the new infectious disease.

Our findings underscore the potential diversification opportunities in GCC equity markets for global investors. Although GCC economies depend heavily on oil revenues, their equity markets would provide a promising possibility for portfolio diversification. This is especially true for less oil-reliant countries, such as Bahrain and the UAE, where the volatility spillover effects from oil to equities are marginal. Notably, GCC financial markets are more segmented and less integrated with global markets than those in other developed or emerging economies. International investors still have restricted access to Gulf stock markets, and liquidity remains low; thus, a lesser effect of the global health crisis is expected for the region. Although equity markets in the region exhibit heterogeneous reactions to oil price volatility and the COVID-19 pandemic, their behavior differs from that of international stock markets, thereby presenting significant portfolio diversification possibilities for investors. Only a few periods after the COVID-19 outbreak show a decreasing trend in the dynamic conditional correlation with the oil market in Bahrain, Kuwait, and Saudi Arabia. Conditional correlations reached low levels by July 2020, and even entered the negative range. During an unprecedented global public health emergency, international investors would benefit from portfolios that are diversified across the GCC group. Investment decisions are also influenced by the region's enhanced financial liberalization and ongoing structural reforms. The implementation of adjustment policies that reduce reliance on the hydrocarbon sector and enhance the private sector's role is key toward tempering stock market sensitivity to oil shocks over time.

Declaration of Competing Interest

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

Appendix

 

Table A1.

Estimation results from the constant conditional correlation (CCC) model.

	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE
$c_{11}$	1.002***	0.076***	0.021	1.023***	-0.015	0.068*
	(0.038)	(0.024)	(1.025)	(0.014)	(1.016)	(0.029)
$c_{22}$	0.039***	0.063***	0.056***	0.072***	0.118***	0.165***
	(0.006)	(0.011)	(0.008)	(0.013)	(0.015)	(0.027)
$d_{11}$	0.037***	0.028***	0.022**	0.015**	0.019***	0.028***
	(0.011)	(0.011)	(0.010)	(0.007)	(0.007)	(0.009)
$d_{22}$	0.002***	0.005***	0.004***	0.002*	-0.002**	0.008**
	(0.001)	(0.002)	(0.001)	(0.001)	(0.001)	(0.004)
$a_{11}$	0.001	0.004	0.001	-0.020***	-0.016**	-0.004
	(0.010)	(0.010)	(0.010)	(0.007)	(0.007)	(0.008)
$a_{12}$	0.046	0.009	-0.014	-0.018	0.015	-0.013
	(0.042)	(0.029)	(0.034)	(0.014)	(1.018)	(1.018)
$a_{21}$	0.019***	-0.027***	-0.015	-0.028***	-0.029***	-0.025*
	(0.006)	(0.009)	(1.011)	(0.007)	(0.009)	(0.016)
$a_{22}$	0.135***	0.075***	0.028*	0.044***	-0.004	0.033**
	(0.017)	(0.018)	(0.018)	(0.013)	(0.011)	0.017)
$b_{11}$	0.911***	1.001***	0.909***	1.013***	0.872***	0.837***
	(0.011)	(0.013)	(0.014)	(0.008)	(0.009)	(0.012)
$b_{12}$	0.045***	1.053**	0.048	0.238***	0.753***	0.060
	(1.324)	(0.427)	(0.054)	(0.053)	(0.044)	(0.097)
$b_{21}$	-0.157***	0.119	0.721***	-0.023	0.295***	0.020
	(0.058)	(1.078)	(0.031)	(0.029)	(0.060)	(0.066)
$b_{22}$	0.808***	1.017***	0.721***	1.008***	0.644***	0.744***
	(0.018)	(0.022)	(0.031)	(0.022)	(0.035)	(0.032)
$g_{11}$	0.131***	0.128***	0.121***	0.116***	0.103***	0.132***
	(0.018)	(0.020)	(0.018)	(0.012)	(0.014)	(0.016)
$g_{12}$	0.027	0.045	0.003	0.009	0.057**	0.001
	(0.043)	(0.027)	(0.064)	(0.026)	(0.017)	(0.002)
$g_{21}$	0.087	0.078***	0.243	0.267***	0.094***	0.379
	(0.064)	(0.004)	(0.183)	(0.065)	(0.033)	(0.539)
$g_{22}$	-0.014	0.108***	0.302***	0.205***	0.365***	0.227***
	(0.022)	(0.026)	(0.040)	(0.031)	(0.042)	(0.040)
${\rho }_{12}$	0.078***	0.122***	0.146***	0.219***	0.229***	0.170***
	(0.017)	(0.021)	(0.023))	(0.011)	(0.018)	(0.022)
Log-Likelihood	-6536.540	-6991.228	-6490.476	-7211.417	-7298.847	-7427.234
AIC	6.683	7.146	6.640	7.374	7.463	7.593
BIC	6.777	7.239	6.745	7.479	7.568	7.698

Note: The results are obtained from the constant conditional correlation (CCC) model. ${\rho }_{12}$ is the conditional correlation coefficient between oil and equity prices for the CCC model. The numbers in parentheses are standard errors. *** $p<0.01$; ** $p<0.05$; * $p<0.1$.

Table A2.

Estimation results from the dynamic conditional correlation (DCC) model.

	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE
$c_{11}$	1.001***	0.187***	1.040***	1.006***	-0.023	0.090***
	(0.039)	(0.011)	(0.014)	(0.007)	(1.024)	(0.026)
$c_{22}$	0.034***	0.095***	0.053***	0.079***	0.112***	0.178***
	(0.006)	(0.003)	(0.001)	(0.011)	(0.015)	(0.025)
$d_{11}$	0.032***	0.082***	0.025***	0.013**	0.021***	0.016*
	(0.012)	(0.004)	(0.004)	(0.005)	(0.007)	(0.009)
$d_{22}$	0.002***	0.009***	0.004***	0.001	-0.002	0.003
	(0.001)	(0.001)	(0.001)	(1.002)	(1.001)	(1.003)
$a_{11}$	0.002	0.078***	0.006	-0.015**	-0.001	0.001
	(1.010)	(0.003)	(1.005)	(0.006)	(1.010)	(1.011)
$a_{12}$	0.073	0.043*	-0.049	-0.067***	-0.064***	-0.003
	(1.048)	(0.024)	(1.034)	(0.014)	(0.024)	(1.025)
$a_{21}$	0.024***	-0.016**	-0.027***	-0.024***	-0.025***	-0.008
	(0.006)	(0.008)	(0.001)	(0.008)	(0.007)	(1.002)
$a_{22}$	0.103***	0.204***	0.032***	0.039***	0.006	0.034**
	(0.018	(0.007)	(0.000)	(0.010)	(1.011)	(0.017)
$b_{11}$	0.921***	0.842***	1.003***	1.017***	0.870***	0.903***
	(0.011	(0.002)	(0.005)	(0.006)	(0.016)	(0.015)
$b_{12}$	0.707**	0.125***	0.411***	0.453***	0.606***	0.059
	(0.276)	(0.023)	(0.066)	(0.007)	(0.112)	(0.059)
$b_{21}$	0.091***	0.109***	0.089***	-0.010	0.104***	0.055***
	(0.036)	(0.006)	(0.008)	(0.015)	(0.030)	(0.021)
$b_{22}$	0.874***	0.651***	0.701***	1.082***	0.733***	0.713***
	(0.019)	(0.004)	(0.006)	(0.019)	(0.028)	(0.030)
$g_{11}$	0.128***	0.054***	0.129***	0.113***	0.096***	0.142***
	(0.019)	(0.006)	(0.008)	(0.011)	(0.015)	(0.019)
$g_{12}$	0.035	0.023	0.012	0.007	0.039**	0.001
	(1.026)	(1.036)	(0.015)	(1.013)	(0.019)	(1.003)
$g_{21}$	0.083	0.086***	0.262	0.347***	0.092***	0.518
	(1.077)	(0.003)	(0.206)	(0.057)	(0.037)	(1.519)
$g_{22}$	-0.015***	0.032***	0.309***	0.206***	0.303	0.254***
	(0.023)	(0.010)	(0.028)	(0.027)	(0.035)	(0.039)
${\theta }_1$	0.023	0.051**	0.050***	0.033***	0.020**	0.040***
	(0.026)	(0.021)	(0.012)	(0.009)	(0.008)	(0.009)
${\theta }_2$	0.295	0.801***	0.786***	0.833***	0.870***	0.948***
	(0.384)	(0.054)	(0.040)	(0.049)	(0.040)	(0.014)
Log-Likelihood	-6520.783	-6998.345	-6491.482	-7208.360	-7300.318	-7419.404
AIC	6.672	7.188	6.642	7.372	7.465	7.586
BIC	6.780	7.287	6.750	7.480	7.573	7.694

Note: The results are obtained using the dynamic conditional correlation (DCC) model. ${\theta }_1$ and ${\theta }_2$ are the estimated coefficients from Engel's (2002) recursion used to generate the correlation matrix within the DCC model. The numbers in parentheses are standard errors. *** $p<0.01$; ** $p<0.05$; * $p<0.1$.

Table A3.

BEKK-GARCH estimation results with the endogenous ID-EMV index.

	Bahrain	Kuwait	Oman	Qatar	Saudi Arabia	UAE
$c_{11}$	0.005	1.048	-0.118***	1.023***	0.290***	-0.204
	(1.038)	(1.066)	(0.043)	(0.035)	(0.041)	(1.038)
$c_{21}$	-0.052**	0.038	-0.064	-0.069	-0.209***	-0.105*
	(0.025)	(1.077)	(1.075)	(1.064)	(0.040)	(0.064)
$c_{22}$	0.165***	0.233***	0.241***	0.266***	0.250***	0.373***
	(0.020)	(0.024)	(0.025)	(0.031)	(0.030)	(0.031)
$c_{31}$	0.052*	0.063	-0.121***	0.050	-0.004***	0.035
	(0.030)	(0.070)	(0.047)	(0.033)	(0.033)	(0.040)
$c_{32}$	0.015	0.008	-0.135***	0.021	0.040*	-0.011
	(0.023)	(0.034)	(0.042)	(0.034)	(0.024)	(0.024)
$c_{33}$	-0.150***	0.167***	0.000	-0.150***	0.164***	0.167***
	(0.019)	(0.029)	(0.137)	(0.022)	(0.015)	(0.019)
$a_{11}$	0.040	0.049**	-0.027	-0.019	0.100***	0.058**
	(0.028)	(0.025)	(0.025)	(0.030)	(0.025)	(0.023)
$a_{12}$	0.012***	0.004	0.033***	-0.017	-0.088***	0.041*
	(0.005)	(0.008)	(0.007)	(0.012)	(0.011)	(0.016)
$a_{13}$	-0.027***	-0.037***	-0.043***	-0.023***	-0.004	-0.038***
	(0.008)	(0.008)	(0.008)	(0.008))	(0.010	(0.008)
$a_{21}$	0.063	0.060	-0.121***	-0.046	-0.183***	0.023
	(0.043)	(0.058)	(0.046	(0.033))	(0.033)	(0.022)
$a_{22}$	0.319***	0.217***	0.120***	0.186***	0.112***	-0.111***
	(0.024)	(0.026)	(0.041)	(0.030)	(0.030)	(0.041)
$a_{23}$	-0.021	-0.066***	-0.105***	-0.083***	-0.091***	-0.065***
	(0.018)	(0.020)	(0.020)	(0.012)	(0.014)	(0.011)
$a_{31}$	-0.080***	-0.077***	-0.093***	-0.096***	-0.054***	-0.053***
	(0.016)	(0.015)	(0.018)	(0.016)	(0.013)	(0.011)
$a_{32}$	-0.006*	-0.026***	-0.005	-0.012**	-0.010	-0.020**
	(0.004)	(0.007)	(0.005)	(0.006)	(0.007)	(0.008)
$a_{33}$	0.351***	0.402***	0.448***	0.375***	0.374***	0.396***
	(0.021)	(0.025)	(0.026)	(0.021)	(0.020)	(0.026)
$b_{11}$	0.962***	0.971***	0.971***	0.964***	0.949	0.966***
	(0.004)	(0.005)	(0.003)	(0.004)	(0.007)	(0.004)
$b_{12}$	0.000	0.004	0.005***	0.000	0.007*	0.009**
	(0.002)	(0.003)	(0.002)	(0.003)	(0.005)	(0.004)
$b_{13}$	-0.008***	-0.006*	-0.009***	-0.009***	-0.004	-0.003
	(0.003)	(0.003)	(0.003)	(0.003)	(0.005)	(0.003)
$b_{21}$	1.015	-0.062	1.102	0.040*	0.102***	-0.013
	(1.016)	(1.024)	(0.023)	(0.021)	(0.023)	(1.014)
$b_{22}$	0.919***	1.014***	0.854***	1.008***	0.863***	0.877***
	(0.011)	(0.011)	(0.014)	(0.011)	(0.013)	(0.013)
$b_{23}$	0.008	-0.001	0.095***	0.010	-0.011	0.007
	(0.008)	(1.008)	(0.017)	(1.010)	(1.012)	(0.008)
$b_{31}$	0.019***	0.020***	0.022***	0.025***	0.015***	0.013***
	(0.005)	(0.005)	(0.006)	(0.005)	(0.004)	(0.004)
$b_{32}$	0.002	0.008***	-0.001	0.002	0.002	0.004
	(0.001)	(0.002)	(0.002)	(0.002)	(0.002)	(0.003)
$b_{33}$	0.930***	0.905***	0.886	0.918***	0.917***	0.908***
	(0.007)	(0.010)	(0.011)	(0.008)	(0.008)	(0.010)
$d_{11}$	0.331***	0.297***	0.333***	0.337***	0.302***	0.334***
	(0.021)	(0.025)	(0.021)	(0.022)	(0.028)	(0.020)
$d_{12}$	0.013*	-0.023	0.018*	0.018	0.028	-0.021
	(0.008)	(0.014)	(0.011)	(0.015)	(0.022)	(0.018)
$d_{13}$	0.042***	0.035***	-0.002	0.027***	0.006	0.020**
	(0.011)	(0.011)	(0.012)	(0.010)	(0.011)	(0.010)
$d_{21}$	0.138**	0.190***	0.157***	-0.161***	-0.122**	0.004
	(0.074)	(0.067)	(0.061)	(0.057)	(0.051)	(0.034)
$d_{22}$	0.136***	0.331***	0.564***	0.427***	0.547***	0.509***
	(0.048)	(0.034)	(0.036)	(0.037)	(0.037)	(0.037)
$d_{23}$	0.018	0.057**	-0.059**	0.023	0.025	-0.027
	(0.029)	(0.024)	(0.028)	(0.023)	(0.021)	(0.018)
$d_{31}$	-0.025	-0.056	-0.078	-0.003	-0.005	0.004
	(0.035)	(0.047)	(0.057)	(0.046)	(0.056)	(0.054)
$d_{32}$	-0.016*	-0.052*	-0.101***	-0.017	-0.006	-0.094**
	(0.010)	(0.029)	(0.032)	(0.023)	(0.034)	(0.039)
$d_{33}$	-0.115	-0.143**	0.089	-0.078	0.011	-0.001
	(0.073)	(0.060)	(0.062)	(0.055)	(0.055)	(0.060)
Log-Likelihood	-9260.128	-9728.839	-9227.800	-9980.536	-10036.164	-10158.720
AIC	9.525	9.973	9.464	10.229	10.286	10.411
BIC	9.789	10.186	9.677	10.442	10.499	10.624

Note: The results are obtained from a trivariate asymmetric BEKK model, where variables have the following order in the system: $\Delta {\mathrm{o}}_t$, $\Delta s_t$, and $IDEM{V}_t$. The numbers in parentheses are standard errors. *** $p<0.01$; ** $p<0.05$; * $p<0.1$.