The Impact of COVID-19 pandemic on Islamic and conventional financial markets: International empirical evidence

Abstract

The current global COVID-19 pandemic is adversely affecting financial markets, including commodities, conventional stocks, and Islamic stocks. This paper empirically investigates the extent to which COVID-19 effects may drive interdependence in markets. We fit copulas to pairs of returns before and during the ongoing epidemic shock, analyze the observed changes in the dependence structure, and discuss asymmetries on the propagation of crisis. We also use the findings to construct portfolios possessing desirable expected behavior. We find that the dependence structure changes significantly during the global pandemic providing valuable information on how the COVID-19 crisis affects inter-dependencies. The selected portfolio, including gold and Islamic return indices, has the best performance outside the COVID-19 crisis, and slightly more performing during the bear markets validating gold’s intrinsic characteristic to be a safe haven. However, the portfolio performances, when combining the Brent with Islamic or conventional indices, have the same trend for the whole period. Our findings contribute to help investors better adjust their investment strategies.

1. Introduction

Compared to conventional finance, “Islamic finance" is relatively recent. Indeed, it is only since the end of the Second World War and the beginning of the independence movements of the countries of Islamic tradition that the first reflections on what will later be qualified as Islamic finance emerge. For several decades, Islamic finance remained a mere intellectual exercise. However, from the early 1960 s onwards, we have witnessed the first attempts to make it a reality, whether under the aegis of governments or private individuals. For example, in 1962, the Malaysian government sponsored a pilgrimage fund (Hachicha & Ben Amar, 2015), and in 1963 the Egyptian economist Ahmad Al-Najjar established a small savings bank in the agricultural town of Mit Ghamr (See Appendix 1). However, the beginning of the independence movements was neither the only nor the main factor that explains the emergence of Islamic finance: instead, the enormous revenues realized by the Gulf Cooperation Council countries following the oil shocks allowed its realization and fueled its growth. On the economic level, the oil embargo declared in October 1973 by the Gulf region’s oil-producing countries led to a quadrupling of the oil price between October and December 1973. This shock led to a transfer of revenues from Western countries to oil-exporting countries. GCC member countries rapidly accumulated wealth through which Islamic finance, hitherto an embryonic concept, became a feasible project (See Appendix 2). The link between the development of Islamic finance and the level of oil revenues was verified following the fall in hydrocarbon prices in the first half of the 1980 s. This shock significantly contributed to the slowdown in the development of Islamic financial institutions in the Gulf countries (Mohieldin, 1997).

With assets amounting, according to ICD Thomson Reuters, to approximately $2523.5 billion in 2019, Islamic finance has become an essential part of the international financial system and, at a double-digit average annual growth rate over the last decade (Mollah, Hassan, Al-Farooque, & Mobarek, 2017), has undoubtedly been one of its most dynamic components. This dynamics is mainly explained by the revenues accumulated by the Gulf countries following the rise in oil prices (Ben Amar, 2019, Kaur Grewal, 2013). According to ICD Thomson Reuters’ estimates, the Islamic finance industry in 2019 includes 2601 financial institutions, including 301 banks, 335 takaful, and retakaful companies, 1701 funds, and 592 other financial institutions. 1 Interest in Islamic finance has grown considerably since the early 1990 s and was reinforced after the financial crisis in 2007 2 which, according to some empirical studies, showed that Islamic banks were more resilient than conventional banks (e.g. Fakhfekh et al., 2016, Trad et al., 2017) and that countries that integrated Islamic finance into their financial systems were relatively less affected (Hasan & Dridi, 2010 3 ). According to Rambaud (2014), this can be explained by “the requirements of Islamic law [which, according to him, reduce] the exposure [of banks] to the risk that could be generated by toxic credits.” Indeed, and since investment with an ex-ante fixed income (riba) is prohibited, as well as hoarding, the saver is, ipso facto, an investor, which implies the absence of a gap between investment and savings in Islamic finance. Moreover, backing any financial transaction with a tangible asset is one of the fundamental principles of Islamic finance. Allais (1993) recognizes the danger of the absence of a real counterpart on the stability of the world economy: “if it is possible to buy without paying or sell without holding, the trade-off between the present and the future can create macroeconomic imbalances,” he says. In Islamic finance, any financial transaction must be systematically backed by a real and not imaginary or notional asset (Hasan, 2014, Zarrokh, 2010). Thus, Islamic finance is supposed to ensure a close link between the real economy and finance. Hasan (2014) adds that the link is obvious not only because Islamic financial institutions cannot sell what they do not own, but also, and above all, because they consider their clients as partners.Nevertheless, the superiority of Islamic finance in terms of resilience remains ambiguous at both the empirical and the theoretical levels. In the theoretical literature, there is a debate between the supporters of Islamic finance and its detractors who think that it is nothing but religious hypocrisy. Proponents of Islamic finance argue that its business model differs significantly from that of conventional finance and that this difference enhances its efficiency and stability (Khan and Mirakhor, 1987, Khan, 1986). However, those who regard Islamic finance as religious hypocrisy (e.g., Kuran, 1995) point out that Islamic and conventional finance are different in form but similar in substance and add that Islamic banks have no particular advantages in terms of efficiency and stability. This debate can also be found in the empirical literature. Indeed, recent studies have shown that Islamic banks are not necessarily more resilient than conventional ones (Ben Amar, 2019, Ben Amar et al., 2017, Ouerghi, 2014). Islamic finance’s relative resilience is still relevant today, especially with the recent COVID-19 pandemic shock, which changed the outlook unexpectedly and involved heavy human, economic and financial consequences (Elliot, 2020). As a result of the high level of uncertainty that the pandemic shock involved, in one week only, from the 24th to the 28th of February, global stock markets lost about US$6 trillion in terms of capitalization (Ozili & Arun, 2020).

While the effects of the COVID-19 outbreak on economic growth and financial markets are currently receiving increasing attention of economists and politicians (e.g., Aloui et al., 2020, Belaid et al., 2021, Ben Amar et al., 2020, Ben Amar et al., 2020, Gourinchas, 2020, Mzoughi et al., 2020, Ben Amar, Hachicha, & Halouani, 2020), the existing literature has not yet examined the extent to which Islamic finance was more or less resilient than conventional finance to the pandemic shock. Thus, this paper aims to fill this gap and address these issues by focusing on the impact of the COVID-19 pandemic shock on Islamic and conventional stock markets for different geographical regions in relation with energy commodity markets. This paper contributes to the existing Islamic empirical literature by providing investors with important information about the relationship between the selected markets, especially during crisis periods. Specifically, it uses copula theory to analyze the impact of the COVID-19 crisis on conventional and Islamic stock markets, when combined with crude oil and gold markets. To this end, it focuses on the dependence structure between two strategic commodities, crude oil and gold, and a set of conventional and Islamic indices from different regions - Europe, Asia, Pacific, Latin America, North America, and GCC countries -, before and during the COVID-19 outbreak. Second, it estimates the tail dependencies between each of the two strategic commodities and each of the regional stock markets considered in order to identify the most performing pairs. The results reveal that all asset pairs exhibit a significant shift in dependence structure between the two sub-periods considered (i.e., before and during the COVID-19 crisis), which provides information on how the COVID-19 pandemic affected interdependencies.

The rest of this paper is organized as follows: Section 2 describes the data and the method. Section 3 displays and discusses the empirical results. Section 4 concludes.

2. Methodology

2.1. Modeling marginal distributions

Marginal densities need to be specified before fitting copulas. The conditional mean of the univariate margins μ _t has a dynamic behavior described by an autoregressive (AR) moving average (MA) process given by

$${\mu }_t={\theta }_0+\sum _{j=0}^p{\theta }_j{r}_{t-j}+\sum _{k=0}^q{\varphi }_k{\xi }_{t-k}$$ (1)

where ξ _t−k = σ _t−k ν _t−k and θ ₀, θ _j and ϕ _k are constant parameters and the AR and MA parameters, respectively, and p and q are positive integers. Conditional volatility processes ${\sigma }_t^2$, exhibit a temporal dynamics described by generalized autoregressive conditional heteroskedasticity (GARCH)(Bollerslev, 1986):

$${\sigma }_t^2={\omega }_0+\sum _{k=1}^r{\omega }_k{\sigma }_{t-k}^2+\sum _{h=1}^s{\alpha }_h{ϵ}_{t-h}^2$$ (2)

where ω ₀ is a constant, while ω _k, α _h are the parameters of ARCH and GARCH respectively.

2.2. Modeling dependence

Copula theory has gained huge interest among researchers and practitioners when modeling the dependence structure between conventional financial assets (e.g., Bouri et al., 2018, Bouri et al., 2019, Mokni and Youssef, 2019, Reboredo et al., 2016, Tiwari et al., 2019) and Islamic financial assets (e.g., Shahzad et al., 2018, Shahzad et al., 2018 and Usman, Jibran, Amir-ud Din, & Akhter, 2018), due to their ability to capture not only the degree of the dependence but also its structure as well as potential asymmetry in the tail dependence, which are highly relevant and essential in risk analysis and portfolio management. These functions may overcome the shortcomings of the classical correlation coefficient, which is linear and incapable of providing information about the tail dependence. Hence, copulas are a flexible alternative to correlation, whose detection can be enhanced for any distribution. Making use of results of Sklar (1959), the dependence structure between crude oil and gold regarding their relationship with Islamic and conventional indices is statistically fully characterized by their joint distribution expressed using copula functions. Consider a sample of random iid vectors X _i ∈ R ^p as X _i = (X _1;i; : : : ; X _p;i) ~ Xiid. Their margins are denoted by F ₁, . . . , F _p. Let F be the distribution function of X and C the copula of X, thus:

$$F(x_1,...,x_p)=C(F_1(x_1),...,F_p(x_p))$$ (3)

Also, copula functions allow assessing the probability that two variables experience joint extreme upward or downward movements through upper (right) λ _U and lower (left) tail dependence λ _L, computed by:

$${\lambda }_L=\underset{u\to 0^{+}}{\lim }Pr[U_2<u∕U_1<u]$$ (4)

Thus, a positive value of λ _L indicates that the two financial market distributions simultaneously exhibit extreme downward movements.

$${\lambda }_U=\underset{u\to 1^{-}}{\lim }Pr[U_2>u∕U_1>u]$$ (5)

When introducing copulas, these two coefficients are

$${\lambda }_U=\underset{u\to 1}{\lim }\frac{C(u,u)}{1-u}and{\lambda }_L=\underset{u\to 0}{\lim }\frac{C(u,u)}{u}$$ (6)

A lower (upper) tail dependence means that we have a non-zero probability of observing minimal (substantial) values for one variable together with minimal (large) values for another variable.

Various types of bivariate copula functions have been employed, allowing systemic and asymmetric dependence structures. The bivariate symmetric copulas include the Normal copula, the Student’s-t copula, the Plackett copula and the symmetrized Joe Clayton copula (SJC). The asymmetric copulas are the Gumbel (rotated Gumbel) copula (which is designed to asymmetrically account for upper (lower) tail dependence) and the Clayton (rotated Clayton) copulas (with a lower (upper) tail independence). The selection of the fitted copula function is based on minimizing the negative log likelihood the AIC and BIC information criteria.4

3. Data and primary results

Our empirical investigation chose the spot price of Brent crude oil, the price of gold, and Islamic and conventional stock market indices of six different geographical regions, including Europe, North America, Pacific, Asia, Latin America, and the Gulf Cooperation Council (GCC). The selected Islamic indices are matched with conventional indices to ensure valid comparisons. Our data are expressed in US dollars and collected from Bloomberg covering the period from October 2, 2018, (around the time without any crisis) to January 11, 2021 (during the ongoing COVID-19 pandemic). Figs. 1 and 2 display the dynamics of all price indices and their returns, the latter computed as the first differences of the natural logs. A visual inspection of these figures reveals that all price indices show a common trend, except gold, especially in that most hit their peak around mid-February 2020, when most indices, except gold, experienced unprecedented levels. As for gold, the plot of its returns shows that it experienced variable oscillations in various time periods in 2020–2021. Tables 1 and 2 present the descriptive statistics, preliminary tests, and correlation matrix before and during the ongoing health crisis. Large differences between maximum and minimum price returns are observed for crude oil (Brent), going from − 0.074–0.136 before the COVID-19 crisis and from − 0.279–0.19 during the ongoing pandemic. Notably, during the current COVID-19 pandemic, the minimum returns for all the indices are negative, but the Islamic indices are less negative than the conventional ones. This observation is similar to that of Mansor and Bhatti (2011) where both groups of indices experienced negative returns during the crisis, but Islamic indices performed slightly better than their conventional peers. Average returns were close to zero, whereas differences in standard deviation indicate dispersion in volatility across crude oil, gold, and financial indices. Indeed, crude oil is more volatile before the COVID-19 crisis and less volatile during the ongoing crisis. This finding is supported by the earlier graphical evidence (Figs. 1 and 2). Likewise, the skewness is negative before and during the current pandemic (except for gold and Brent before the pandemic), showing a shift to the left. High kurtosis values were common to crude oil and gold during the ongoing corona-virus crisis, even if they were low for gold before the current turbulent period, pointing out the impact of the medical shock and implying fat tails in the return distributions. Regarding conventional and Islamic indices, Kurtosis values remain high before and during the COVID-19 pandemic, specifically for MSCI North America conventional index (MXNA), MSCI Latin America Islamic index (MILA), and the two indices of GCC countries (MIGC and MXGCC). For the rest of the selected indices, kurtosis values become higher during the ongoing crisis, showing thick tail distributions. Furthermore, the Jarque-Bera (JB) test strongly rejects the normality of the unconditional distribution. These findings motivate the use of copula theory in our empirical investigation.

Fig. 1.

Time series plots of daily indices.

Fig. 2.

Time series plots of daily returns.

Table 1.

Descriptive statistics of return series and unconditional correlations before COVID-19 pandemic.

	MIEU	MXEU	MINA	MXNA	MIPC	MXPC	MIMS	MXMS	MILA	MXLA	MIGC	MXGCC	Brent	Gold
Panel A: Descriptive statistic and primary tests
Min	-0.0323	-0.0314	-0.031	-0.0329	-0.0374	-0.0325	-0.0341	-0.0397	-0.0928	-0.044	-0.05	-0.049	-0.0743	-0.0218
Max	0.0248	0.0278	0.043	0.0459	0.033	0.0297	0.032	0.0316	0.037	0.0427	0.0399	0.0336	0.136	0.0246
Mean	0.00021	0.00024	7.9E-05	0.00029	1.4E-05	5.3E-05	0.00022	0.00022	-6.6E-05	0.00032	5.6E-05	0.00015	-0.0007	0.0007
Std.Dev.	0.008	0.0077	0.0091	0.0094	0.0084	0.0078	0.0095	0.009	0.0134	0.0128	0.0086	0.0076	0.0211	0.0069
Skewness	-0.47 *	-0.419 *	-0.351 *	-0.374 *	-0.39 *	-0.409 *	-0.314 *	-0.35 *	-1.17 *	-0.342 *	-0.24 *	-0.86 *	0.245 *	0.185 *
Kurtossis	1.49 *	1.883 *	2.75 *	3.24 *	2.21 *	2.18 *	1.19 *	1.97 *	6.46 *	0.97 *	6.97 *	8.26 *	6.44 *	1.29 *
JB	42.232 *	57.517 *	109.44 *	149.7 *	74.42 *	73.37 *	24.44 *	59.43 *	640.3 *	19.18 *	661.5 *	964.*	564.87 *	24.5 *
ADF	-10.641 *	-10.547 *	-9.92 *	-10.174 *	-10.17 *	-9.716 *	-9.27 *	-9.68 *	-10.856 *	-10.56 *	-9.896 *	-9.98 *	-10.81 *	-10.2 *
KPSS	0.17	0.188	0.208	0.219	0.347	0.273	0.212	0.159	0.142	0.047	0.095	0.108	0.246	0.06
Q(20)	19.55	21.9	19.3	14.87	15.58	17.53	21.5	15.59	19.73	16.56	12.94	10.94	21.87	50.91
ARCH(10)	1.069	1.155	5.49	4.74	7.44	7.53	1.37	1.92	0.513	2.25	2.03	1.39	0.788	1.24
Panel B: Correlation matrix
	MIEU	MXEU	MINA	MXNA	MIPC	MXPC	MIMS	MXMS	MILA	MXLA	MIGC	MXGCC	Brent	Gold
MIEU	1
MXEU	0.923	1
MINA	0.609	0.603	1
MXNA	0.593	0.596	0.969	1
MIPC	0.181	0.174	0.166	0.147	1
MXPC	0.198	0.206	0.165	0.149	0.979	1
MIMS	0.54	0.54	0.519	0.513	0.546	0.552	1
MXMS	0.522	0.523	0.427	0.418	0.583	0.598	0.952	1
MILA	0.44	0.452	0.497	0.47	0.127	0.124	0.392	0.339	1
MXLA	0.41	0.39	0.484	0.447	0.114	0.112	0.329	0.271	0.85	1
MIGC	0.258	0.238	0.186	0.21	0.235	0.247	0.289	0.307	0.142	0.181	1
MXGCC	0.29	0.269	0.212	0.229	0.259	0.274	0.345	0.368	0.197	0.233	0.96	1
Brent	0.35	0.306	0.391	0.335	0.136	0.134	0.24	0.21	0.279	0.281	0.112	0.152	1
Gold	-0.232	-0.233	-0.227	-0.24	-0.02	-0.019	-0.16	-0.137	-0.033	-0.045	-0.14	-0.165	-0.164	1

Notes: In the above table, MIEU (MSCI Europe Islamic local index), MXEU (MSCI Europe index), MINA (MSCI North America Islamic index), MXNA (MSCI North America index), MIPC (MSCI Pacific Islamic index), MXPC (MSCI Pacific index), MIMS (MSCI Emerging Market Asia Islamic index), MXMS (MSCI Emerging Market Asia index), MILA (MSCI Emerging Market Latin America Islamic index), MXLA (MSCI Emerging Market Latin America index), MIGC (MSCI GCC countries combined Islamic index) and MXGCC (MSCI GCC countries combined index) designate the return series of the Islamic and conventional indices. All return series are computed as the first difference of the natural log price for daily data. JB indicates the Jarque-Bera normality test with the null hypothesis that the variable from which the sample was taken follows a normal distribution. ADF and KPSS denote the Augmented Dickey-Fuller (1979) and unit root tests. Q(20) denotes the Ljung Box test statistics for serial correlation, whereas ARCH(10) indicates autoregressive heteroscedasticity tested through Engle (1982) at 20 lags. *, ** and **** denote significances levels at 1,5% and 10%, respectively, and rejection of the null hypothesis for normality, unit root, stationarity, no autocorrelation, and conditional heteroscedasticity.

Table 2.

Descriptive statistics of return series and unconditional correlations during COVID-19 pandemic.

	MIEU	MXEU	MINA	MXNA	MIPC	MXPC	MIMS	MXMS	MILA	MXLA	MIGC	MXGCC	Brent	Gold
Panel A: Descriptive statistic and primary tests
Min	-0.113	-0.123	-0.115	-0.128	-0.058	-0.06	-0.058	-0.058	-0.15	-0.161	-0.163	-0.17	-0.28	-0.0586
Max	0.084	0.082	0.089	0.091	0.0656	0.0663	0.063	0.056	0.121	0.114	0.068	0.054	0.19	0.0496
Mean	3.6E-05	-9E-05	0.0003	0.0007	0.0005	0.0004	0.0011	0.001	0.0004	-0.0005	0.0002	-4.8E-05	-0.0006	0.0007
Std.Dev.	0.0167	0.0174	0.02	0.021	0.013	0.0129	0.015	0.014	0.0286	0.0288	0.015	0.015	0.0436	0.012
Skewness	-1.35 *	-1.54 *	-0.85 *	-0.98 *	-0.0013	-0.006	-0.039 *	-0.675 *	-1.15 *	-1.395 *	-4.89 *	-5.71 *	-1.48 *	-0.765 *
Kurtossis	10.87 *	11.66 *	8.4 *	9.42 *	3.88 *	4.34 *	3.2 *	3.72 *	7.6 *	8.38 *	53.39 *	61.16 *	13.22 *	4 *
JB	1400.9 *	1624.2 *	822.22 *	1035 *	168.47 *	210.24 *	121.21 *	175.2 *	7.705 *	870.4 *	32,899 *	43,234 *	2049.2 *	205.08 *
ADF	-8.27 *	-8.07 *	-7.7 *	-7.71 *	-7.88 *	-7.84 *	-7.7 *	-7.5 *	-7.63 *	-7.69 *	-8.55 *	-8.32 *	-8.73 *	-9.16 *
KPSS	0.2193	0.249	0.189	0.179	0.439	0.386	0.326	0.386	0.548	0.53	0.36	0.431	0.324	0.112
Q(20)	51.1 * *	61.12 * *	196.734 * *	193.26 * *	51.83 * *	47.41 * *	25.48	28.76	43 * *	56.09 * *	86.32 * *	87.96 * *	35.16 * **	29.45
ARCH(10)	9.72 * *	6.77 * *	17.56 * *	17.86 * *	11.196 * *	11.78 * *	16.47 * *	15.54 * *	21.67 * *	16.841 * *	1.28	1.04	2.32 * **	1.77
Panel B: Correlation matrix
	MIEU	MXEU	MINA	MXNA	MIPC	MXPC	MIMS	MXMS	MILA	MXLA	MIGC	MXGCC	Gold	Brent
MIEU	1
MXEU	0.984	1
MINA	0.723	0.738	1
MXNA	0.698	0.718	0.984	1
MIPC	0.509	0.528	0.383	0.381	1
MXPC	0.536	0.556	0.409	0.404	0.985	1
MIMS	0.59	0.589	0.504	0.5	0.608	0.63	1
MXMS	0.609	0.619	0.495	0.488	0.63	0.655	0.967	1
MILA	0.684	0.694	0.767	0.77	0.45	0.481	0.623	0.63	1
MXLA	0.707	0.734	0.78	0.786	0.446	0.48	0.592	0.607	0.967	1
MIGC	0.512	0.509	0.448	0.433	0.358	0.37	0.438	0.467	0.496	0.509	1
MXGCC	0.528	0.532	0.443	0.432	0.395	0.406	0.453	0.489	0.512	0.533	0.98	1
Gold	0.445	0.434	0.423	0.401	0.211	0.216	0.26	0.285	0.41	0.427	0.494	0.507	1
Brent	0.131	0.113	0.152	0.142	0.183	0.186	0.126	0.141	0.138	0.08	-0.034	-0.028	0.05	1

Notes: In the above table, MIEU (MSCI Europe Islamic local index), MXEU (MSCI Europe index), MINA (MSCI North America Islamic index), MXNA (MSCI North America index), MIPC (MSCI Pacific Islamic index), MXPC (MSCI Pacific index), MIMS (MSCI Emerging Market Asia Islamic index), MXMS (MSCI Emerging Market Asia index), MILA (MSCI Emerging Market Latin America Islamic index), MXLA (MSCI Emerging Market Latin America index), MIGC (MSCI GCC countries combined Islamic index) and MXGCC (MSCI GCC countries combined index) designate the return series of the Islamic and conventional indices. All return series are computed as the first difference of the natural log of the price for daily data. JB represents the Jarque-Bera normality test with the null hypothesis that the variable from which the sample was taken follows a normal distribution. ADF and KPSS denote the Augmented Dickey Fuller(1979) and Kwiatkowski et al.(1992) unit root tests. Q(20) denotes the Ljung Box test statistics for serial correlation whereas ARCH(10) indicates autoregressive conditional heteroscedasticity tested through Engle (1982) at 20 lags.* ** , * * and * denote significance levels at 1,5, and 10%, respectively, and rejection of the null hypothesis for normality, unit root, stationarity, no autocorrelation and conditional heteroscedasticity.

Furthermore, the correlation matrices (Tables 1 and 2) show that there is a strong association between conventional and Islamic indices before the occurrence of the current medical shock, which is amplified during the medical crisis period: the maximum value for the correlation coefficient before the COVID-19 pandemic is around 0.979 (0.985 during the COVID-19 pandemic), and a minimum value around − 0.24 ( − 0.035 during the COVID-19 pandemic). This evidence can be explained by the fact that stock market indices have indeed a strong tendency to move together due to their similar financial situation and even more during periods of high volatility. This is previously proved by Abdul Karim, Lee, Abdul Karim, and Jais (2012), who finds that both the Islamic banking financing and stock market variables are cointegrated with several macroeconomic variables (inflation, real exchange rate, interest rates, and economic activity as represented by the industrial production index) both before and during the subprime mortgage crisis.

4. Empirical analysis

Our investigation begins by analyzing the daily returns of crude oil, gold, six conventional indices, and six Islamic indices, highlighting their stylized facts and then their dependence structure over two time periods: from October 2, 2018, to December 31, 2019,5 (before the COVID-19 pandemic) and from January 1, 2020, to January 11, 2021 (during the COVID-19 pandemic). Lastly, an analysis of the evolution through time of the accumulated gains of two equally weighted portfolios is carried out.

Tables 3 and 4 show the specifications of the marginal distributions before and during the COVID-19 crisis. The change in models and especially the significance of the excess of volatility for all the variables presents the impact of the COVID-19 crisis. Next, examining the dependence structure between the daily returns needs a transformation into the uniform space. This will be obtained using the empirical cumulative distribution, and the eight selected copula families are estimated via inference functions for margins (IFM).6 Selection of the best fitting copula follows by comparing the the values of the negative log-likelihood value and the Akaike Information Criterion (AIC). The nonparametric copula densities, computed according to Deheuvels (1979), are reported in Figure A.1 (see the Appendix). A visual inspection shows that the nonparametric densities point to major evidence of weakly or nonsymmetric comovement before and during the pandemic period between financial markets and commodity markets, as indicated by the fact that the probability mass is unevenly located along the quadrant formed by the points (0,0) and (1,1). This result is consistent with the results of the dependence structure estimations. Tables 5 6 7 8 9 10 11 12 report, each, the estimation results for the eight considered copula functions. Fundamentally, the analysis will be divided into two strands highlighting similarities and/or differences between the crude oil market and the gold market with respect to regional MSCI conventional and Islamic indices. Specifically, it is about answering two questions: What characterizes the dependence structure between the crude oil market and conventional and Islamic MSCI indices before and during the ongoing pandemic? and what characterizes the dependence structure between the gold market and conventional and Islamic MSCI indices?

Table 3.

Estimation marginal results before COVID-19 pandemic.

	MIEU	MXEU	MINA	MXNA	MIPC	MXPC	MIMS	MXMS	MILA	MXLA	MIGC	MXGCC	Gold	Brent
Mean equation
C(M)	0.0007	0.0008	0.0007	0.0009
	(0.0324)	(0.0125)	(0.021)	(0.002)
AR(1)	-0.854			0.95					-0.765
	(0.0012)			(0.00)					(0.00)
MA(1)	0.851			-0.98					0.88
	(0.0012)			(0.00)					(0.00)
Variance equation
C(V)	5.46E06		3.14						0.3E04
	(0.0357)		(0.041)						(0.04)
ARCH(ϕ1)	0.147	0.164	0.188	0.2	0.068	0.086	0.024	0.03	0.057	0.029	0.002	0.006	0.05	0.1
	(0.0217)	(0.02)	(0.001)	(0.02)	(0.012)	(0.02)	(0.04)	0.037)	(0.017)	(0.04)	(0.009)	(0.008)	(0.001)	(0.01)
GARCH(β1)	0.781	0.77	0.79	0.84	0.93	0.88	0.96	0.96	0.76	0.96	0.97	0.965	0.95	0.92
	(0.000)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Deg.Freedom	4.872	4.42	4.65	4.18	7.66	6.48	7.78	5.27	5.28	8.14	2.47	2.55	4.98	3.13
	(0.0003)	(0.00)	(0.00)	(0.00)	(0.00)	(0.003)	(0.007)	(0.0002)	(0.005)	(0.007)	(0.00)	(0.00)	(0.0001)	(0.00)
Diagnostic tests
Log-Likelihood	1127.927	1143.224	1108.26	1108.157	1116.104	1141.03	1062.23	1085.89	964.92	962.97	1140.96	1180.43	1169.57	834.7
AIC	-6.898	-7.004	-6.79	-6.78	-6.85	-6.99	-6.5	-6.66	-5.9	-5.9	-7.002	-7.2	-7.17	-5.11
Q(20)	9.48	13.55	13.88	15.4	13.97	16.36	26.47	20.72	7.6	35.4	40.5	34.2	16.2	5.39
ARCH(10)	0.258	0.258	0.99	0.86	0.7	0.87	1.18	0.93	0.17	2.55	2.5	1.5	0.6	0.25

Notes:This table reports the optimal fitted margin models based on the information criteria and the significances of the estimated parameters at the 5% level. MIEU (MSCI Europe Islamic local index), MXEU (MSCI Europe index), MINA (MSCI North America Islamic index), MXNA (MSCI North America index), MIPC (MSCI Pacific Islamic index), MXPC (MSCI Pacific index), MIMS (MSCI Emerging Market Asia Islamic index), MXMS (MSCI Emerging Market Asia index), MILA (MSCI Emerging Market Latin America Islamic index), MXLA (MSCI Emerging Market Latin America index), MIGC (MSCI GCC countries combined Islamic index) and MXGCC (MSCI GCC countries combined index) designate the return series of Islamic and conventional indices. All return series are computed as the first difference of the natural log price for daily data. C(M) and C(V) are constants for the mean and variance equation, respectively. Q(20) denotes the Ljung Box test statistics for serial correlation whereas ARCH(20) indicates autoregressive conditional heteroscedasticity, tested through Engle (1982) at lags 20.

Table 4.

Estimation marginal results during COVID-19 pandemic.

	MIEU	MXEU	MINA	MXNA	MIPC	MXPC	MIMS	MXMS	MILA	MXLA	MIGC	MXGCC	Gold	Brent
Mean equation
C(M)			0.001	0.001			0.001	0.001					0.001
			(0.005)	(0.003)			(0.006)	(0.04)					(0.014)
AR(1)		-0.46	-0.53	-0.52								0.64	0.69	-0.76
		(0.0005)	(0.00)	(0.00)								(0.00)	(0.00)	(0.00)
MA(1)		0.36	0.35	0.34								-0.6	-0.74	0.82
		(0.005)	(0.0001)	(0.00)								(0.00)	(0.00)	(0.00)
Variance equation
C(V)	0.12E04	0.14E04	0.1E04	0.1E04	0.08E04	0.07E04	0.25E04	0.16E04	06E04	0.48E04			0.1E04
	(0.05)	(0.03)	(0.08)	(0.028)	(0.04)	(0.03)	0.008)	(0.01)	(0.015)	(0.03)			(0.05)
ARCH(ϕ1)	0.18	0.22	0.35	0.39	0.15	0.15	0.16	0.14	0.23	0.26	0.36	0.54	0.13	0.19
	(0.01)	(0.018	(0.012)	(0.004)	(0.006)	(0.003)	(0.019)	(0.028)	(0.002)	(0.002)	(0.04)	(0.04)	(0.008)	(0.01)
GARCH(β1)	0.77	0.73	0.7	0.64	0.79	0.79	0.71	0.76	0.67	0.67	0.9	0.9	0.83	0.86
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Deg.Freedom			3.65						6.37		2.09	2.09	3.12	3.38
			(0.006)						(0.008)		(0.00)	(0.00)	(0.00)	(0.00)
Diagnostic tests
Log-Likelihood	770.3	765.7	787.9	771.6	814.06	822.19	771.7	794.7	644.18	653.54	919.7	925.75	836.95	567.02
AIC	-5.72	-5.67	-5.82	-5.71	-6.05	-6.11	-5.73	-5.9	-4.77	-4.84	-6.84	-6.87	-6.2	-4.19
Q(20)	33.6	31.09	9.94	5.82	39.12	32.87	21.9	18.3	7.04	13.99	11.46	11.35	6.6	19.09
ARCH(10)	1.8	1.54	0.45	0.25	2.24	1.82	0.89	0.68	0.52	1.11	0.91	0.85	0.42	1.25

Notes: This table reports the optimal fitted margin models based on the information criteria and the significance of the estimated parameters at the 5% level. MIEU (MSCI Europe Islamic local index), MXEU (MSCI Europe index), MINA (MSCI North America Islamic index), MXNA (MSCI North America index), MIPC (MSCI Pacific Islamic index), MXPC (MSCI Pacific index), MIMS (MSCI Emerging Market Asia Islamic index), MXMS (MSCI Emerging Market Asia index), MILA (MSCI Emerging Market Latin America Islamic index), MXLA (MSCI Emerging Market Latin America index), MIGC (MSCI GCC countries combined Islamic index) and MXGCC (MSCI GCC countries combined index) designate the return series of the Islamic and conventional indices. All return series are computed as the first difference of the natural log of the price for daily data. C(M) and C(V) denote constants for the mean and variance equation, respectively. Q(20) denotes the Ljung Box test statistics for serial correlation, whereas ARCH(20) indicates autoregressive conditional heteroscedasticity, tested through Engle (1982) at lags 20.

Table 5.

Estimated dependence structure before COVID-19 pandemic between Brent and conventional indices.

Copula	MXEU	MXNA	MXPC	MXMS	MXLA	MXGCC
Gaussian
ρ	0.3463	0.3421	0.1187	0.2029	0.3283	0.1789
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
LL	-20.7573	-20.2294	-2.304	-6.8337	-18.5361	-5.2842
AIC	-41.5084	-40.4527	-4.6018	-13.6613	-37.0661	-10.5623
Clayton
ϕ	0.4757	0.415	0.1886	0.2606	0.4019	0.2411
λL	0.2329	0.1882	0.1001	0.07	0.178	0.0564
λU	0	0	0	0	0	0
LL	-17.7847	-14.9229	-4.15	-7.3601	-13.9046	-6.2484
AIC	-35.5633	-29.8396	-8.2938	-14.714	-27.803	-12.4905
Rotated Clayton
γ	0.4461	0.4664	0.0956	0.2156	0.4029	0.14
λL	0	0	0	0	0	0
λU	0.2115	0.2263	0.0616	0.0402	0.179	0.0071
LL	-16.2805	-18.152	-1.0891	-4.6533	-13.4142	-2.0366
AIC	-32.5549	-36.2979	-2.172	-9.3	-26.8223	-4.067
Plackett
ω	3.3637	2.883	1.3864	1.8	2.8612	1.6453
λL	0	0	0	0	0	0
λU	0	0	0	0	0
0	0	0
LL	-23.5938	-19.2606	-1.6749	-5.6329	-20.0368	-4.5795
AIC	-47.1814	-38.5751	-3.3436	-11.2597	-40.0675	-9.1529
Gumbel
φ	1.2843	1.2749	1.1	1.1422	1.2465	1.1
λL	0	0	0	0	0	0
λU	0.2845	0.2776	0.1792	0.1654	0.2562	0.1221
LL	-19.7693	-20.9033	-2.2332	-7.6931	-15.5828	-2.9296
AIC	-39.5325	-41.8004	-4.4603	-15.38	-31.1594	-5.853
Rotated Gumbel
ε	1.2937	1.261	1.1017	1.1388	1.2451	1.1214
λL	0.2912	0.2673	0.1838	0.1621	0.2551	0.1445
λU	0	0	0	0	0	0
LL	-21.6929	-18.0086	-4.6327	-6.76	-15.1447	-5.3372
AIC	-43.3797	-36.041	-9.2593	-13.5151	-30.2833	-10.6682
Student
	0.3686	0.3435	0.1098	0.197	0.3388	0.1838
DF	6.5972	8.4964	5.6757	7.4652	70.413	99.8714
λL	0.1001	0.0581	0.0556	0.0427	0	0
λU	0.1001	0.0581	0.0556	0.0427	0	0
LL	-23.4904	-21.8526	-5.0638	-8.4	-18.5793	-5.268
AIC	-46.9684	-43.6929	-10.1154	-16.7892	-37.1462	-10.5237
SJC
β1	0.1626	0.213	9E-11	0.06	0.1461	1.9E-04
β2	0.2147	0.1391	0.096	0.0833	0.1414	0.1068
λL	0.2147	0.1391	0.0962	0.0833	0.1414	0.1068
λU	0.1627	0.213	0	0.0606	0.1461	0.0002
LL	-22.1893	-21.8251	-4.5596	-8.4	-16.3062	-5.8389
AIC	-44.3664	-43.6378	-9.1073	-16.7868	-32.6	-11.6656

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

Table 6.

Estimated dependence structure during COVID-19 pandemic between Brent conventional indices.

Copula	MXEU	MXNA	MXPC	MXMS	MXLA	MXGCC
Gaussian
ρ	0.3615	0.3617	0.181	0.2203	0.3765	0.3473
LL	-18.764	-18.787	-4.4636	-6.6657	-21.4884	-17.2188
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-37.52	-37.5665	-8.1996	-13.324	-40.9693	-34.4302
Clayton
ϕ	0.6715	0.7286	0.395	0.4012	0.6363	0.5191
LL	-25.8833	-29.4325	-10.0673	-12.3877	-23.916	-16.9202
λL	0.3562	0.3862	0.1451	0.1777	0.3364	0.2631
λU	0	0	0	0	0	0
AIC	-51.7592	-58.8575	-20.1272	-24.768	-47.8246	-33.833
Rotated Clayton
γ	0.4546	0.4258	0.1429	0.1763	0.4831	0.5068
LL	-12.7777	-10.8395	-1.4842	-2.2839	-14.1372	-17.0279
λL	0	0	0	0	0	0
λU	0.2177	0.1963	0.0078	0.0196	0.2381	0.2547
AIC − 25.55	-21.6716	-2.9608	-4.5603	-2.284	-34.0483	-17.028
Plackett
	3.8645	4.3167	1.9713	2.0228	3.9602	3.2991
LL	-21.5128	-25.1052	-5.5479	-6.5672	-24.0595	-16.8227
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-43.018	-50.203	-11.0883	-13.127	-48.1115	-33.6379
Gumbel
φ	1.3225	1.3292	1.1274	1.1408	1.3389	1.3095
LL	-20.0251	-19.2148	-3.6226	-4.4745	-20.7274	-24.4586
λL	0	0	0	0	0	0
λU	0.311	0.3155	0.1507	0.164	0.3218	0.3021
AIC	-40.043	-38.4222	-7.2378	-8.9416	-41.4474	-42.9098
Rotated Gumbel
ε	1.375	1.3998	1.1837	1.1984	1.3688	1.3171
LL	-29.1683	-31.3577	-11.0288	-12.5607	-27.2875	-22.025
λL	0.3445	0.3592	0.204	0.2168	0.3407	0.3074
λU	0	0	0	0	0	0
AIC	-58.33	-62.7079	-22.0502	-25.1139	-54.5676	-44.0425
Student
ρ	0.3645	0.3912	0.1947	0.216	0.3994	0.3248
DF	2.1	22.1	2.9472	3.9938	2.7445	2.2101
LL	-37.6154	-40.1287	-14.7897	-12.4821	-30.8869	-31.4977
λL	0.3132	0.3259	0.1791	0.1328	0.278	0.2855
λU	0.3132	0.3259	0.1791	0.1328	0.278	0.2855
AIC	-75.2158	-80.2426	-29.5645	-24.9492	-61.7589	-62.9805
SJC
β1	0.1193	0.032	8.8E-07	1.8E-09	0.1399	0.224
β2	0.3591	0.4068	0.2197	0.2407	0.3275	0.2398
LL	-29.4658	-31.192	-11.1211	-13.0758	-26.8342	-25.2389
λL	0.3591	0.4068	0.2197	0.2407	0.3275	0.2398
λU	0.1193	0.032	0	0	0.1399	0.224
AIC	-58.9167	-62.369	-22.2272	-26.1366	-53.6534	-50.4629

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

Table 7.

Estimated Copula before COVID-19 pandemic between Brent and Islamic indices.

Copula	MIEU	MINA	MIPC	MIMS	MILA	MIGC
Gaussian
ρ	0.4037	0.4124	0.1223	0.2387	0.3116	0.121
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
LL	-28.9016	-30.2944	-2.4476	-9.535	-16.6017	-2.3973
AIC	-57.7971	-60.5827	-4.889	-19.0639	-33.1973	-4.7885
Clayton
ϕ	0.5688	0.5429	0.1865	0.3012	0.3929	0.1421
λL	0.2956	0.2789	0.0243	0.0253	0.1713	0.0076
λU	0	0	0	0	0	0
LL	-24.7679	-23.214	-4.09	-9.3342	-13.8291	-2.3843
AIC	-49.5297	-46.4218	-8.1825	-19.6623	-27.6521	-4.7624
Rotated Clayton
γ	0.5099	0.5933	0.0974	0.2488	0.3671	0.0982
λL	0	0	0	0	0	0
λU	0.2568	0.3109	0.0008	0.0007	0.1514	0.0009
LL	-19.8137	-26.5815	-1.136	-6.088	-11.5021	-1.1278
AIC	-39.6213	-53.1568	-2.2658	-12.1699	-22.998	-2.2494
Plackett
ω	3.8481	3.8069	1.3845	1.9549	2.7275	1.3594
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
LL	-31.8214	-31.3337	-1.7093	-7.7736	-17.2978	-1.7599
AIC	-63.6367	-62.6612	-3.4125	-15.541	-34.5895	-3.5136
-9.1529
Gumbel
φ	1.3332	1.3629	1.1	1.1566	1.23	1.1
λL	0	0	0	0	0	0
λU	0.3181	0.3371	0.1221	0.1221	0.2428	0.1221
LL	-49.5185	-61.9132	-3.6514	-17.5041	-28.5277	-1.9983
AIC	-49.5185	-61.9132	-3.6514	-17.5041	-28.5277	-1.9983
Rotated Gumbel
ε	1.3493	1.3504	1.1	1.1615	1.2348	1.1
λL	0.3285	0.3292	0.1221	0.124	0.247	0.1221
λU	0	0	0	0	0	0
LL	-27.9438	-28.1772	-4.3512	-8.6212	-14.7976	-1.4259
AIC	-55.8815	-56.3482	-8.6963	-17.2362	-29.589	-2.8456
Student
ρ	0.4217	0.4247	0.1116	0.2399	0.3218	0.1237
DF	14.3245	6.4445	7.0751	18.5986	26.1475	98.7955
λL	0.0244	0.124	0.0344	0.0025	0.0009	0
λU	0.0244	0.124	0.0344	0.0025	0.0009	0
LL	-29.6307	-33.7642	-4.2746	-9.7997	-16.77	-2.3068
AIC	-59.249	-67.5162	-5.5369	-19.587	-33.5285	-4.6012
SJC
β1	0.1687	0.2713	2.78E-08	0.0634	0.1188	0.0017
β2	0.2678	0.2085	0.0864	0.1099	0.1559	0.0218
λL	0.2678	0.2085	0.864	0.1099	0.1559	0.0218
λU	0.1687	0.2713	0	0.0634	0.1559	0.0017
LL	-27.6877	-32.0808	-4.3583	-10.3844	-15.8463	-2.315
AIC	-55.363	-64.1493	-8.7046	-20.6366	-31.6809	-4.6176

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

Table 8.

Estimated Copula during COVID-19 pandemic between Brent and Islamic indices.

Copula	MIEU	MINA	MIPC	MIMS	MILA	MIGC
Gaussian
ρ	0.3772	0.3963	0.1779	0.2081	0.3394	0.3296
LL	-20.5669	-22.8884	-4.3081	-5.9327	-16.3982	-15.4131
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-41.1263	-45.7693	-8.6088	-11.8579	-32.7888	-30.8186
Clayton
ϕ	0.7084	0.8092	0.3393	0.3682	0.6011	0.473
LL	-28.3419	-33.9519	-9.2292	-10.7283	-22.4801	-14.5809
λL	0.3759	0.4246	0.1297	0.1522	0.3156	0.231
λU	0	0	0	0	0	0
AIC	-56.6764	-67.8964	-18.451	-21.449	-44.9528	-29.1544
Rotated Clayton
γ	0.4606	0.5264	0.1513	0.1632	0.366	0.4638
LL	-13.3104	-14.8284	-1.7066	-2.0135	-8.6105	-15.0058
λL	0	0	0	0	0	0
λU	0.222	0.268	0.0102	0.0143	0.1505	0.2243
AIC	-26.6134	-29.6493	-3.4057	-4.0195	-17.2134	-30.0042
Plackett
ω	3.8982	5.4111	1.9224	1.8733	3.3842	3.0453
LL	-22.3948	-32.8375	-5.141	-5.7323	-18.5439	-14.8223
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-44.7822	-65.6676	-10.2745	-10.7371	-37.0803	-29.6372
Gumbel
φ	1.3293	1.3986	1.129	1.1261	1.2738	1.2803
LL	-20.7025	-25.5724	-3.8742	-3.6122	-14.5	-18.5074
λL	0	0	0	0	0	0
λU	0.3156	0.3585	0.1523	0.1493	0.2769	0.2816
AIC	-41.3975	-51.1372	-7.741	-7.2169	-28.9934	-37.0073
Rotated Gumbel
ε	1.3899	1.4581	1.1755	1.1823	1.3278	1.2912
LL	-31.109	-36.7685	-10.0867	-11.1903	-23.8455	-19.1666
λL	0.3534	0.3914	0.1966	0.2028	0.3145	0.2895
λU	0	0	0	0	0	0
AIC	-62.2105	-73.5295	-20.1678	-22.3731	-47.6835	-38.3257
Student
ρ	0.3663	0.4609	0.1829	0.2022	0.3588	0.3135
DF	2.1	2.1	2.9558	4.6424	3.0985	2.6268
LL	-37.3642	-48.8506	-14.4828	-10.1665	-24.0764	-25.8912
λL	0.3141	0.3609	0.1745	0.1042	0.2351	0.2474
λU	0.3141	0.3609	0.1745	0.1042	0.2351	0.2474
AIC	-74.7135	-97.6862	-28.9507	-20.3181	-48.1378	-51.7674
SJC
β1	0.1096	0.115	7.1E-09	3.4E-09	0.0219	0.1999
β2	0.3783	0.4248	0.2085	0.2225	0.3439	0.2148
LL	-31.6306	-36.4509	-10.2154	-11.5144	-23.5551	-21.8
λL	0.3783	0.4248	0.2085	0.2225	0.3439	0.2148
λU	0.1096	0.115	0	0	0.0219	0.1999
AIC	-63.2463	-72.8868	-20.4159	-23.0138	-47.0952	-43.5862

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

Table 9.

Estimated dependence structure before COVID-19 pandemic between Gold and conventional indices.

Copula	MXEU	MXNA	MXPC	MXMS	MXLA	MXGCC
Gaussian
ρ	-0.2329	-0.2084	0.0055	-0.1245	-0.0208	-0.0968
LL	-9.0655	-7.2124	-0.0049	-2.5401	-0.0705	-1.5283
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-18.1248	-14.4186	-0.0037	-5.0741	-0.1349	-3.0505
Clayton
ϕ	1E-04	1E-04	1E-04	1E-04	1E-04	1E-04
LL	0.0059	0.0052	0.0008	0.0036	0.0005	0.0025
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	0.0179	0.0166	0.0078	0.0134	0.0072	0.0112
Rotated Clayton
γ	1E-04	1E-04	0.0584	1E-04	1E-04	1E-04
LL	0.0053	0.0046	-0.4772	0.0014	0.0008	0.0024
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	0.0167	0.0154	-0.9482	0.0091	0.0078	0.011
Plackett
ω	0.4817	0.5485	1.0436	0.6498	1.052	0.855
LL	-9.308	-5.8776	-0.0296	-3.0014	-0.0435	-0.4409
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-18.6097	-11.749	-0.053	-5.9966	-0.0809	-0.8757
Gumbel
φ	1.1	1.1	1.1	1.1	1.1	1.1
LL	13.4	11.1682	2.4805	6.9418	4.0022	7.1644
λL	0	0	0	0	0	0
λU	0.1221	0.1221	0.1221	0.1221	0.1221	0.1221
AIC	26.8066	22.3426	4.9672	13.8897	8.0106	14.3349
Rotated Gumbel
ε	1.1	1.1	1.1	1.1	1.1	1.1
LL	13.5743	11.1151	3.8242	8.6147	4.0713	7.2682
λL	0.1221	0.1221	0.1221	0.1221	0.1221	0.1221
λU	0	0	0	0	0	0
AIC	27.1547	22.2364	7.6546	17.2356	8.1487	14.5425
Student
ρ	-0.241	-0.2074	0.0071	-0.1323	-0.0073	-0.088
DF	44.2788	15.2727	8.8053	7.1774	18.2113	22.1353
LL	-9.1195	-7.5204	-1.3351	-4.6857	-0.359	-1.7419
λL	0	0.0001	0.0113	0.0111	0.0003	0
λU	0	0.0001	0.0113	0.0111	0.0003	0
AIC	-18.2266	-15.0284	-2.6579	-9.359	-0.7057	-3.4714
SJC
β1	9.5E-07	9E-07	4.4E-04	9.5E-07	9.5E-07	9.5E-07
β2	9.536	9.536	1.862	9.536	9.536	9.536
LL	2.9457	2.595	-0.1247	1.476	0.5265	1.4077
λL	0	0	0	0	0	0
λU	0	0	0.0004	0	0	0
AIC	5.9037	5.2022	-0.2371	2.9643	1.0653	2.8277

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

Table 10.

Estimated dependence structure before COVID-19 pandemic between Gold and Islamic indices.

Copula	MIEU	MINA	MIPC	MIMS	MILA	MIGC
Gaussian
ρ	-0.2331	-0.1972	0.0033	-0.1539	-0.0181	-0.087
LL	-9.082	-6.4448	-0.0017	-3.8936	-0.0534	-1.2347
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-18.1578	-12.8834	0.0027	-7.7811	-0.1006	-2.4632
Clayton
ϕ	1E-04	1E-04	1E-04	1E-04	1E-04	1E-04
LL	0.0056	0.005	0.0007	0.0043	0.0005	0.0023
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	0.0174	0.0162	0.0076	0.0148	0.0071	0.0108
Rotated Clayton
γ	1E-04	1E-04	0.0471	1E-04	1E-04	1E-04
LL	0.0057	0.0039	-0.3033	0.0022	0.0004	0.0023
λL	0	0	0	0	0	0
λU 0	0	0	0	0	0
AIC	0.0175	0.014	-0.6005	0.0106	0.0069	0.0108
Plackett
ω	0.4873	0.5642	1.0534	0.5929	1.0041	0.8534
LL	-9.3455	-5.2164	-0.0453	-4.4512	-0.0003	-0.4606
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-18.684	-10.4266	-0.0845	-8.8963	0.0056	-0.9151
Gumbel
φ	1.1	1.1	1.1	1.1	1.1	1.1
LL	13.5888	9.7163	2.6002	7.816	3.9659	7.5082
λL	0	0	0	0	0	0
λU	0.1221	0.1221	0.1221	0.1221	0.1221	0.1221
AIC	27.1838	19.4387	5.2065	15.6381	7.9379	15.0225
Rotated Gumbel
ε	1.1	1.1	1.1	1.1	1.1	1.1
LL	13.5401	10.4725	3.7527	10.1085	3.9925	6.8071
λL	0.1221	0.1221	0.1221	0.1221	0.1221	0.1221
λU	0	0	0	0	0	0
AIC	27.0864	20.9511	7.5116	20.2232	7.9913	13.6204
Student
ρ	-0.2406	-0.1868	0.0095	-0.162	-0.0086	-0.0871
DF	100	8.3218	9.9987	7.3272	15.0219	88.5308
LL	-9.1208	-7.5957	-1.0989	-5.9699	-0.4757	-1.2481
λL	0	0.0047	0.0073	0.008	0.001	0
λU	0	0.0047	0.0073	0.008	0.001	0
AIC	-18.2294	-15.179	-2.1855	-11.9275	-0.9391	-2.4838
SJC
β1	9.5E-04	9.5E-07	1.5E-06	9.5E-07	9.5E-07	9.5E-07
β2	9.536	9.536	9.536	9.536	9.536	9.536
LL	2.9723	2.3827	0.1047	1.8238	0.438	1.3394
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	5.9569	4.7778	0.2217	3.66	0.8883	2.6911

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

Table 11.

Estimated dependence structure during COVID-19 pandemic between Gold and Islamic indices.

Copula	MIEU	MINA	MIPC	MIMS	MILA	MIGC
Gaussian
	0.0139	0.0845	0.1452	0.0792	0.0904	-0.0136
LL	-0.0258	-0.9626	-2.8543	-0.8432	-1.1004	-0.0248
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-0.0441	-1.9177	-5.7011	-1.6789	-2.1932	-0.0421
Clayton
ϕ	0.0699	0.1382	0.2235	0.0913	0.1301	1E-04
LL	-0.6152	-2.0176	-4.3183	-0.8405	-1.8335	0.0009
λL	0	0.0066	0.045	0.0005	0.0048	0
λU	0	0	0	0	0	0
AIC	-1.2229	-4.0277	-8.6292	-1.6736	-3.6596	0.0093
Rotated Clayton
γ	0.0563	0.1145	0.1576	0.1382	0.1407	0.063
LL	-0.4513	-1.4772	-2.2529	-2.0383	-2.063	-0.5061
λL	0	0	0	0	0	0
λU	0	0.0023	0.0123	0.0066	0.0072	0
AIC	-0.8952	-2.9469	-4.4984	-4.0692	-4.1186	-1.0047
Plackett
ω	0.9481	1.2456	1.5937	1.2793	1.2241	0.9809
LL	-0.0325	-0.5469	-2.7659	-0.7553	-0.4807	-0.0052
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-0.0576	-1.0864	-5.5243	-1.5031	-0.9539	-0.0029
Gumbel
φ	1.1	1.1	1.106	1.1	1.1	1.1
LL	-0.5692	-2.922	-4.3841	-2.7707	-3.9766	-0.0299
λL	0	0	0	0	0	0
λU	0.1221	0.1221	0.1288	0.1221	0.1221	0.1221
AIC	-1.131	-5.8366	-8.7608	-5.5339	-7.9457	-0.0524
Rotated Gumbel
ε	1.1	1.1	1.1263	1.1	1.1	1.1
LLL	-0.7408	-2.2655	-5.9186	-1.2249	-2.2693	3.2151
λL	0.1221	0.1221	0.1495	0.1221	0.1221	0.1221
λU	0	0	0	0	0	0
AIC	-1.4741	-4.5236	-11.2298	-2.4424	-4.531	6.4376
Student
ρ	-0.0146	0.0412	0.1574	0.064	0.0507	-0.0024
DF	2.8634	2.9217	4.1385	3.5893	2.9884	5.2408
LL	-10.0711	-10.022	-8.0012	-7.3482	-10.4073	-3.5279
λL	0.1193	0.1316	0.1094	0.1058	0.1307	0.0448
λU	0.1193	0.1316	0.1094	0.1058	0.1307	0.0448
AIC	-20.1272	-20.0291	-15.9875	-14.6816	-20.7996	-7.0408
SJC
β1	2.3E-06	0.0233	0.0222	0.0548	0.0617	0.0169
β2	0.0161	0.0208	0.0954	1.2E-07	0.004	2.4E-10
LL	-1.0076	-2.7938	-5.6008	-2.5628	-3.3904	-0.5345
λL	0.0161	0.0208	0.0954	0	0.004	0
λU	0	0.0233	0.0222	0.0548	0.0617	0.0169
AIC	-2.0002	-5.5726	-11.1867	-5.1107	-6.7658	-1.054

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

Table 12.

Estimated dependence structure during COVID-19 pandemic between Gold and Conventional indices.

Copula	MXEU	MXNA	MXPC	MXMS	MXLA	MXGCC
Gaussian
ρ	-0.0068	0.0692	0.1412	0.1163	0.0273	-0.0109
LL	-0.0062	-0.6441	-2.6975	-1.8252	-0.0996	-0.0159
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-0.0049	-1.2807	-5.3876	-3.6428	-0.1918	-0.0244
Clayton
ϕ	0.0517	0.1176	0.2075	0.1171	0.0675	1E-04
LL	-0.3554	-1.5492	-3.8813	-1.2948	-0.5708	0.0008
λL	0	0.0028	0.0354	0.0027	0	0
λU	0	0	0	0	0	0
AIC	-0.7034	-3.0909	-7.7551	-2.5821	-1.1342	0.009
Rotated Clayton
γ	0.0479	0.1063	0.1698	0.1972	0.0817	0.0695
LL	-0.3499	-1.3573	-2.6901	-3.6979	-0.8371	-0.6159
λL	0	0	0	0	0	0
λU	0	0.0015	0.0169	0.0298	0.0002	0
AIC	-0.6923	-2.7072	-5.3727	-7.3884	-1.6667	-1.2243
Plackett
ω	0.8747	1.1367	1.5171	1.4835	1.0118	0.975
LL	-0.2028	-0.1849	-2.1617	-1.8862	-0.0016	-0.0085
λL	0	0	0	0	0	0
λU	0	0	0	0	0	0
AIC	-0.3982	-0.3623	-4.316	-3.765	0.0043	-0.0096
Gumbel
φ	1.1	1.1	1.1105	1.1059	1.1	1.1
LL	-0.0921	-2.7586	-5.0188	-4.6142	-1.9463	-0.1439
λL	0	0	0	0	0	0
λU	0.1221	0.1221	0.1333	0.1285	0.1221	0.1221
AIC	-0.1766	-5.5098	-10.0302	-9.221	-3.8851	-0.2803
Rotated Gumbel
ε	1.1	1.1	1.1211	1.1	1.1	1.1
LL	-0.0039	-1.5992	-5.3459	-2.4546	-0.1602	3.0093
λL	0.1221	0.1221	0.14443	0.1221	0.1221	0.1221
λU	0	0	0	0	0	0
AIC	-0.0003	-3.1909	-10.6844	-4.9018	-0.3129	6.0261
Student
ρ	-0.0423	0.0114	0.1404	0.0897	-0.0022	-0.0055
DF	2.6572	2.8156	3.5759	3.4635	2.8099	4.8142
LL	-11.9711	-10.4453	-9.1821	-8.6701	-10.5522	-4.1672
λL	0.1234	0.1291	0.1278	0.1183	0.1256	0.0528
λU	0.1234	0.1291	0.1278	0.1183	0.1256	0.0528
AIC	-23.9273	-20.8758	-18.3493	-17.3253	-21.0894	-8.3195
SJC
β1	0.0072	0.0298	0.0418	0.0947	0.0354	0.0177
β2	3.6E-06	0.0049	0.0773	1.2E-07	2.15E-10	3.7E-09
LL	-0.6711	-2.4451	-5.6676	-4.3562	-1.5304	-0.6309
λL	0	0.0049	0.0773	0	0	0
λU	0.0072	0.0298	0.0418	0.0947	0.0354	0.0177
AIC	-1.3273	-4.8753	-11.3203	-8.6974	-3.0459	-1.2468

Note: In this table, we present the estimated bivariate copula for each pair of returns through minimum negative log-likelihood, and the AIC.

The first strand focuses on the relationship between the crude oil market and conventional and Islamic stock markets before and during the COVID-19 crisis. Indeed, ever since the seminal work of Hamilton (1983) 7 , a large body of the academic literature has been devoted to the study of the impact of oil prices on macroeconomic variables (Filis and Chatziantoniou, 2014, Lippi and Nobili, 2012, Mork, 1989) and stock market returns.8 A wide strand of the literature has examined the impact of oil prices on stock market returns. As the stock price is the present value of expected future cash-flows, the literature has identified two channels through which the oil price can affect stock prices: direct channels by altering future cash flows and indirect channels by affecting the equity risk premium and, consequently, the discount rate. While these direct channels are based on the idea that oil is necessary to produce many goods, and that fluctuations in oil prices may alter the demand (i.e., consumption, investment and public expenditure decisions) and, consequently, income and cash flows (Guo & Kliesen, 2005),9 the indirect channel is based on the hypothesis that changes in oil prices could have an impact on the equity risk premium which, together with the nominal interest rate, is the main determinant of the rate at which future cash flows are discounted (Abul Basher, Haug, & Sadorsky, 2018). Indeed, oil price fluctuations can affect macroeconomic variables including GDP growth rates, inflation, and exchange rates (Hamilton & Herrera, 2004) and thus indirectly drive equity risk premiums’ which in turn affect the discount rates applied to the cash-flows in stock valuation models.

The estimated dependence structure between the crude oil market and regional MSCI conventional indices have mainly three types of copulas: (i) the Student copula (with Pacific, Asian, and North America indices), (ii) the Plackett copula (with European and Latin America indices) and (iii) the Clayton copula (with GCC countries index). As for the estimated dependence structure between the crude oil market and regional MSCI Islamic indices, the estimated copulas are: (i) the Student copula (with North America index), (ii) the normal copula (with the GCC countries index), (iii) the Plackett copula (with Latin America and European indices), and (iv) the SJC copula (with Asian and Pacific indices). Notably, the fitted copulas between the crude oil market and the conventional and Islamic indices corresponding to each region have the same tail dependence characteristics before the occurrence of the ongoing medical shock, except for GCC countries. This is due, at least in part, to the dependence of the GCC countries on oil to achieve their economic growth and to the fact that the GCC countries are resource-based economies, reflecting, to some extent at least, their low level of diversification, which is a mechanical consequence of their heavy dependence on oil revenues (the ’Dutch disease’).10 This result is consistent with those of Hammoudeh, Dibooglu, and Aleisa (2004) 11 .

Regarding the relationship between the crude oil market and the conventional and the Islamic indices, we find that the Student copula is the best function for capturing the dependence structure between each pair, except for the Asian stock market (modeled using the SJC copula). Notably, each of the two employed types of copula functions has significant tail dependence values, indicating an extreme dependence that is more intense during the critical time period, even if the copula before the turbulent period does not allow dependence during extreme movements. These findings argue that both conventional and Islamic indices were affected by the COVID-19 crisis. It is also interesting to note that the estimates (of the copula parameters and the upper and lower tail values) are more drastically higher when using the conventional indices than the Islamic indices. This finding is in accordance with the findings of Soke Fun, Abd Rahman, Muhamad Yusuf, and Zamzamin (2014) that Islamic indices outperformed their conventional counterparts during crisis periods. This may due to one of the characteristics of Islamic institutions, viz., that they support their investments.

The second strand focuses on the relationship between the gold market and conventional and Islamic stock markets before and during the current pandemic. The dependence structure between gold and MSCI conventional indices is captured using the Student copula allowing symmetric tail dependence, except for the European index (captured using the Plackett copula without tail dependence). Before the COVID-19 crisis, the dependence between the gold market and MSCI conventional stock, was more intense than that between the gold market and MSCI Islamic stocks, except for North America, where we see only a slight difference. The same results are to be seen when estimating the relationship between the gold market and the MSCI Islamic stocks after January 1, 2020. Indeed, the Student copula seems to be the best symmetric copula for modeling the dependence structure between the gold market and the selected financial stock markets, taking into account the dependence during extreme events. The sign of the Student copula remains positive during the ongoing pandemic for the majority of the series, affirming the impact of the crisis. This finding is consistent with Kuran (1995) and Ben Amar (2019), indicating that, in practice, there are few differences between the modus operandi of Islamic finance and the methods used in conventional finance, and with Hammoudeh, Mensi, Reboredo, and Nguyen (2014), suggesting that the Sharia-compliance rules are not strong enough to make the Islamic stock universe very different from the conventional stock markets. The intensity of the relationship is higher during the corona-virus crisis than before, and is also higher with the MSCI conventional stocks than the MSCI Islamic stocks (except in Europe, North America, and Latin America, where the differences are only slight). This is also supported by Sukmana and Kholid (2010), suggesting that Islamic indices are more resilient to crises than conventional indices.

Lastly, we consider the evolution over time of the accumulated gains of equally-weighted portfolios. Based on estimates of the tail dependence, we selected pairs showing the most positive upper tail dependence before the current pandemic and the lowest lower tail dependence during the critical period of the COVID-19 pandemic. The results show that the pairs that show the lowest extreme dependence during bull markets are gold and the MSCI GCC countries conventional and Islamic indices (gold-MXGCC with λ _L = 0.0528 and gold-MIGC with λ _L = 0.0448) and the crude oil and MSCI Pacific conventional and Islamic indices (Brent-MXPC with λ _L = 0.1791 and Brent-MIPC with λ _L = 0.1745). On the other hand, the pairs that show the most substantial positive dependence at extreme levels before COVID-19 pandemic period are gold and MSCI Pacific conventional index and MSCI Asian Islamic index (gold-MXPC with λ _U = 0.0113 and gold-MIMS with λ _U = 0.0088) and two MSCI North American conventional and Islamic indices (Brent-MXNA with λ _U = 0.0581 and Brent-MINA with λ _U = 0.124). The dependence between gold and MIMS (Gold-MIMS) during the pre-COVID-19 crisis period is, at least in part, due to the fact that Asia is a region where Islamic financial engineering, and hence the Islamic finance industry, is quite developed.

We then construct equally weighted portfolios and compare their cumulative returns in order to detect the best performing portfolio before and during the COVID-19 period. When comparing the cumulative returns between gold and both Asian Islamic (Gold-MIMS) and conventional Pacific (Gold-MXPC) indices, we observe that the Gold-MIMS portfolio performs significantly better, i.e., it records relatively higher cumulative returns (see Fig. 3). These findings are in line with Hammoudeh et al. (2014), who affirming the usefulness of the information to investors respecting Sharia and diversification with commodities such as gold. Indeed, backing any financial transaction with a tangible asset is one of the fundamental principles of Islamic finance. Thus, in Islamic finance, any financial transaction must be systematically backed by a real asset and not an imaginary or notional one (Hasan, 2014, Zarrokh, 2010). As a result, Islamic finance is supposed to ensure a close link between the performance of real assets and that of purely financial assets.

Fig. 3.

The evolution of accumulated gains for both portfolios throughout the span of the data.

Fig. 4 shows that the cumulative return curves have a common trend. However, we can see that the portfolio including the Islamic index (Gold-MIGC) slightly outperforms the second portfolio (Gold-MXGCC), especially during the crisis period. This resilience may be due to the Sharia-based rules that characterize Islamic finance. To summarize, these results suggest that, before and during the COVID-19 period, portfolios combining gold and Islamic assets (from Asian [MIMS] and GCC [MIGC] regions) outperform portfolios consisting of gold and conventional assets (from Pacific [MXPC] and GCC regions [MXGCC]).

Fig. 4.

The evolution of accumulated gains for both portfolios throughout the span of the data.

The incorporation of crude oil in two portfolios, the first, -including the North American conventional index (MXNA) and the second the North American Islamic index (MINA), allowed us to detect that, over the entire period, that the Brent-MINA combination is slightly better performing, in terms of cumulative portfolio return, than the Brent-MXNA combination, even if it is negative (see Fig. 5). In particular, unprecedented negative returns appear when crude oil prices become very volatile, mainly due to the price war between Russia and OPEC and the collapse in demand. As for the conventional and Islamic MSCI Pacific indices, Fig. 6 shows a similar pattern for the two portfolios including crude oil, while the portfolio incorporating the conventional index (Brent-MXPC) performs slightly better than the one incorporating the Islamic index (Brent-MIPC). In addition, the results also suggest that portfolios combining Brent with either the conventional or Islamic Pacific or North American indices allow investors to have the same performance. Furthermore, our results reveal that gold, when combined with stock market indices (whether Islamic or conventional), provides investors with a better performance. Figs. 7, 8, 9, 10.

Fig. 5.

The evolution of accumulated gains for both portfolios throughout the span of the data.

Fig. 6.

The evolution of accumulated gains for both portfolios throughout the span of the data.

Fig. 7.

The evolution of accumulated gains for both portfolios throughout the span of the data.

Fig. 8.

The evolution of accumulated gains for both portfolios throughout the span of the data.

Fig. 9.

The evolution of accumulated gains for both portfolios throughout the span of the data.

Fig. 10.

The evolution of accumulated gains for both portfolios throughout the span of the data.

5. Conclusions

The increasing integration of financial and energy commodity markets provides international investors new investment opportunities and the benefits of diversification. However, under extreme market conditions, it becomes difficult for investors to obtain the benefits of diversification and hedge their positions.

This article has examined the effects of COVID-19 on six geographical regional conventional and Islamic markets, including Europe, North America, Pacific, Asia, Latin America, and the Gulf Cooperation Council. Specifically, the selected Islamic indices are matched with conventional indices to ensure valid comparisons covering the period from October 2, 2018, (around the time without any crisis) to January 11, 2021 (during the ongoing COVID-19 pandemic). Our research contributes to the Islamic empirical literature by giving important information for investors in their investment decision on the relationship between the selected markets, especially during a bear period. The method used is based on the theory of copulas in order to model the dependence structure before and during the ongoing pandemic and assess the probability that two series in two different markets experience join extreme upward or downward movements. Evidently, before fitting bivariate copulas, the marginal densities have to be specified. The dynamic behavior of the conditional mean has been described by an AutoRegressive Moving Average (ARMA) process and the dynamic behavior of the conditional volatility described by a Generalized AutoRegressive Conditional Heteroskedasticity (GARCH) process. The results highlight that all markets show a significant persistence in their volatility process, as benefits a crisis. A notable change in the dependence structure between crude oil and Islamic and conventional indices before and during the unprecedented event provides valuable information on how the COVID-19 pandemic affects the inter-dependencies. It is also interesting to note that the estimates (of the copula parameters and the upper and lower tail values) are more drastically higher when using the conventional indices than the Islamic indices. This is at least partly due to the principles of Islamic finance, particularly the backing any financial transaction with a tangible asset.

Regarding the dependence structure between gold and Islamic and conventional indices, the estimates show an increase of the parameters of the Student copula, highlighting the impacts of the crisis, except the European indices, for which the copula function changes. The results are in line with previous studies (e.g., Kuran, 1995; Ben Amar, 2019), indicating that, in practice, there are few differences between the modus operandi of Islamic finance and the methods used in conventional finance. Moreover, it seems that the Sharia-compliance rules are not strong enough to make the Islamic stock universe very different from the conventional stock markets (Hammoudeh et al., 2014). To sum up, our findings may support the conclusions of Sukmana and Kholid (2010) suggesting that Islamic indices are more resilient towards crisis than conventional indices.

Furthermore, we investigated portfolio risk management issues based on previous estimates. Indeed, selecting the pairs showing the lowest left tail dependence (λ _L) outside the turbulence period and the pairs showing the highest right tail dependence (λ _U) during the crisis period, we considered the evolution through time of the accumulated gains of pairs of equally-weighted portfolios, including gold, crude oil, and financial indices. The results reveal that the selected portfolio, including gold and Islamic indices, has the best performance during both periods, validating the role of gold as a safe haven. Moreover, the results show little difference between the portfolios including Brent with Islamic or conventional indices.

To sum up, this study contributes to the growing literature on the economic impacts of the COVID-19 pandemic, by focusing on the Islamic and conventional stock markets. The results have several practical implications for international investors and policymakers,- who wish to invest in Islamic and/or conventional stock markets in order to better manage their portfolios, especially during market downturns. Indeed, during bear market conditions, the presence of a dependence between two markets suggest that the investment in these two asset classes needs careful placement in a portfolio. Practically, Islamic equity markets are usually considered as significant parts for optimal diversification benefits. Based on the distributions of cumulative returns, our results suggest that combining gold and Islamic indices in the same portfolios seems to be able to mitigate the effects of the crisis.

For future research, it may be attractive to extend our empirical analysis in many directions. For instance, investors will be interested in further information from the investigation of the relationship between energy commodity markets and conventional and Islamic indices across different investment horizons and market conditions (e.g., combining the wavelet and Cross-Quantilogram techniques). It will also be useful to characterize the level of spillovers among these markets under different investment horizons when computing, for example, the systemic risk. Moreover, given that all indices recorded negative returns during the COVID-19 period, driven by the high level of global uncertainty, it seems that Islamic stock indices were negatively affected by investor sentiment (fear, anxiety and stress). This opens the door also for another future in-depth study of the violation of the principles of Sharia, by examining the possibility of developing new hedging instruments for Islamic markets.

Appendices

Appendix 1 In 1963, Ahmad al-Najjar created a small savings-investment bank in the agricultural town of Mit Ghamr (Egypt), hence the name of this entity. This bank invested the deposits in its possession in the real estate sector. In 1967, when the bank’s network spread throughout the country through nine branches, increasing the number of depositors by 14 times and the volume of deposits by 45 times between 1963 and 1967, the authorities decided to close the bank and liquidate all its assets. The government justified this decision with a number of reasons, the main ones being the failure to comply with registration and licensing procedures and the incompetence of the bank’s managers ensuring effective management of the funds collected. Appendix 2 The idea of creating an Islamic bank was first discussed at the supranational level during the summit of the Organization of the Islamic Conference (OIC) in Lahore, Pakistan, from 22 to 24 February 1974, four months after the first oil shock. Indeed, following a suggestion made by Saudi Arabia, Algeria and Somalia, the OIC member countries decided by mutual agreement, to create a supranational financial institution. This initiative took shape in 1974 with the creation of the Islamic Development Bank (IDB) based in Jeddah, the first financial institution to insert the adjective -Islamic’ in its corporate name. The emergence of this bank, currently owned by a consortium of 56 countries, responded to the need to establish an institution capable of financing government projects and promoting economic and social development in countries with a Muslim tradition. At the time of its creation, no one knew what an Islamic bank could be or how it should operate to comply with Shari’a rules. It took no less than two years for the IDB to complete its first murabaha transaction with Algeria. Public support for the establishment of an “Islamic bank” has stimulated private initiative in GCC countries. The first the private Islamic bank, the Dubai Islamic Bank, was established in 1975. Since then, the number of Islamic banks has continued to grow, first in Arab countries, then in Asia and finally in the West: It was not until 22 years after the creation of the first private Islamic bank in Dubai that the first Western conventional bank, Citibank, decided to open its first Islamic banking subsidiary in Bahrain in 1997. The adoption of Islamic finance by conventional banks can be explained by the high savings potential of these Gulf-sponsored institutions.

Figs. 11, 12, 13, 14, 15, 16, 17, 18.

Fig. 11.

Copulas densities between crude oil (Brent) and conventional indices before COVID-19 pandemic.

Fig. 12.

Copulas densities between crude oil (Brent) and Islamic indices before COVID-19 pandemic.

Fig. 13.

Copulas densities between crude oil (Brent) and conventional indices during COVID-19 pandemic.

Fig. 14.

Copulas densities between crude oil (Brent) and Islamic indices during COVID-19 pandemic.

Fig. 15.

Copulas densities between Gold and conventional indices before COVID-19 pandemic.

Fig. 16.

Copulas densities between Gold and Islamic indices before COVID-19 pandemic.

Fig. 17.

Copulas densities between Gold and conventional indices during COVID-19 pandemic.

Fig. 18.

Copulas densities between Gold and Islamic indices during COVID-19 pandemic.

Figure A.1. Empirical copula densities.