Degree and structure of return dependence among commodities, energy stocks and international equity markets during the post-COVID-19 period.

This paper examines the safe-haven role of copper, iron, gold, silver, and energy stocks for international equity markets during the COVID-19 pandemic. Specifically, the degree and structure of return dependence at different points of conditional return distributions are examined for the pre-COVID and post-COVID periods. The results show that copper is a weak safe-haven for the US equity market at the upper-tail of conditional distribution of cooper returns during the post-COVID period. Gold loses its hedge status during the post-COVID period while silver is a strong safe-haven against international equity markets at the upper-tail of conditional return distribution of silver. Further, iron pose weak safe-haven properties against international equity markets when iron returns are extremely positive. However, neither conventional nor green energy stocks act as safe-haven against international equity markets. Current results may provide guidance for risk management, portfolio management and policy decisions during the post-COVID-19 period.

Introduction

The coronavirus (COVID-19) pandemic is one of the most impactful shocks to the financial markets since the beginning of the modern economic era. Following its declaration as a pandemic by World Health Organization (WHO), return volatility of equity markets increased dramatically (e.g. Debakshi and Basistha, 2021; Gao et al., 2021; Hong et al., 2021; Uddin et al., 2021) accompanied with severe daily losses (e.g. Baker et al., 2020; Zhang et al., 2020), shock spillovers (e.g. Elgammal et al., 2021) and financial instability (e.g. Shehzad et al., 2021b). According to the empirical findings of Baker et al. (2020), no previous pandemic, including the Spanish Flu, hit financial markets as hard as COVID-19. On 12th of March, MSCI global, MSCI developed markets, MSCI emerging markets, MSCI Europe and Standards & Poor 500 (S&P500) equity market indices dropped by 12%, on average. Similar daily losses experienced on 11th and 16th of March as well, albeit with a slightly lower magnitude. Another notable loss experienced on March 9, 2020, MSCI equity index of global energy producers dropped by more than 20%. These patterns represent a great amount of wealth loss for equity investors, worldwide.

The COVID-19 emerged as a great challenge for equity market investors. During such times, investors look for safer assets because their risk aversion is more sensitive to acute losses than gains (see Tversky and Kahneman, 1991), indicating increased demand towards assets whose returns are uncorrelated or negatively correlated with equity market returns. Such assets are labeled as safe-haven assets in a pioneering work by Baur and Lucey (2010). Safe-haven assets exhibit return behavior similar to those of hedge assets during market turbulence. Specifically, return on a hedge asset, on average, is uncorrelated or negatively correlated with equity market indices during tranquil times while a safe-haven asset exhibits same relation with equity markets during crisis. Although there is no theory to guide us on choosing assets with safe-haven properties, empirical evidence shows that investors engage in flight-to-safety behavior amid to significant shocks to financial markets (e.g. Baur and Lucey, 2009; Forbes and Rigobon, 2002; Hartmann et al., 2004; Sarwar, 2017 are among others). Accordingly, there is value in searching for assets which may improve risk-return characteristics of portfolios during financial turbulence.

Diversification is made possible when there is less than perfect correlation among the returns of financial assets (Bai and Green, 2010; Chandar et al., 2009). However, shocks triggered by crises increase co-movements among financial asset returns, particularly equity returns tend to become highly correlated (see for instance Gherghina and Simionescu, 2021; Hasman and Samartin, 2008; So et al., 2021; Youssef et al., 2021). This would limit benefits of international diversification of equity portfolios (Ibragimov and Walden, 2007), hence motivates investors for including other asset classes into their portfolios (Baur and McDermott, 2010). Over the last two decades natural resources have gained popularity as financial commodities (Corbet et al., 2020) and relative protection against shocks. Empirical evidence showed that inclusion of precious metals into investment portfolios offers diversification benefits (see also Belousova and Dorfleitner, 2012; Canover et al., 2009; Hillier et al., 2006; Jensen et al., 2002). More importantly, examination of relationship among equity markets and different asset classes reveals important information about rescue measures for policy makers during turbulence and thus improves stability of financial markets (Shahzad et al., 2018).

Earlier papers revealed the ability of gold and several other metals to provide inflation hedging (e.g. Bampinas and Panagiotidis, 2015; Beckmann and Czudaj, 2013; Erb and Harvey, 2013; McCown and Zimmerman, 2006), exchange rate risk hedging (e.g. Capie et al., 2005; Hammoudeh et al., 2009; Reboredo, 2013; Reboredo and Rivera-Castro, 2014) and oil price risk hedging (e.g. Yaya et al., 2016). Ciner et al. (2013) reported that gold can act as a safe-haven for exchange rate risk in the UK and US. While, results from Bampinas and Panagiotidis (2015) showed that silver can hedge inflation in the UK but not in the US. According to the results of Ciner (2001), stable relationship among gold and silver returns disappeared following 1990s, indicating that these assets may provide different hedging implications in managing risky portfolios. Moreover, Umar et al. (2019) reported that copper can hedge credit risk of US metal and mining industry better than other metals. According to the finding of Agyei-Ampomah et al. (2014), copper offers hedging and safe-haven properties better than gold against losses in sovereign bonds. In addition, Roache and Rossi (2010) reported that copper (gold and silver) price responds positively (negatively) to announcements about an improvement in economic measures. Because, copper is an important metal used as an input in manufacturing and tend to be more demanded during economic expansion (see also Elder et al., 2012). Finally, Salisu and Adediran (2019) reported potential of iron to hedge inflation in Australia. Accordingly, return behavior of these metals and their connection to other assets exhibit different patterns in different economic settings as well as during different economic cycles. Therefore, it is important to examine their safe-haven properties during different economic episodes.

Baur and Lucey (2010) and Baur and McDermott (2010) are among the frontiers to examine the safe-haven properties of financial commodities against equity portfolios. Baur and Lucey (2010) documented that gold can act as a safe-haven for the US, UK and German equity markets amid to financial crisis of 2008. And, Baur and McDermott (2010) reported that gold has safe-haven properties for European and the US markets. But, Shahzad et al. (2019) revealed that gold is a weak safe-haven commodity for equity indices of developed and developing markets in face of crises experienced during January 2010 to February 2018. And, results of Bredin et al. (2015) showed that gold can act as a safe-haven for only up to a year (see also Baur and Glover, 2012; Baur and McDermott, 2016; Klein, 2017). However, hedging and safe-haven properties of metals, both precious and industrial, against equity portfolios have been underexamined for the post-COVID period. The recent and ongoing pandemic and relatively better return performance of commodities compared to international equity markets during the post-COVID-19 period (please see Table 1 ) present a strong motivation to examine the ability of financial commodities to act as safe-haven assets for international equity investors. In addition to that, findings of Zhang et al. (2021) revealed that the ability of oil price movements to predict stock returns has declined by almost 90% during the post-COVID period. Indicating decoupling of energy related assets and stock market returns during the post-COVID-19 period.

Table 1

Descriptive statistics for daily returns.

	Mean	Median	Std. D.	Max	Min	Skew.	Kurt.	p(J-B)	Obs.
Panel A: Descriptive statistics for pre-COVID-19 period; January 03, 2018 – March 10, 2020
Commodities and Precious Metals
Copper	−0.009	−0.018	1.078	3.722	−4.370	−0.045	4.176	0.000	549
Gold	0.038	0.052	0.716	3.095	−3.718	−0.070	5.611	0.000	549
Iron	0.034	0.022	1.570	13.398	−16.563	−1.648	40.629	0.000	549
Silver	1.410	0.347	3.103	11.567	−7.437	1.331	4.115	0.000	549
Global energy equity indices
Energy	−0.108	0.047	1.551	4.434	−20.444	−4.234	55.976	0.000	549
Clean energy	0.037	0.000	1.223	5.450	−11.994	−1.683	20.381	0.000	549
Equity Market Indices
Global	−0.010	0.100	1.016	4.386	−8.024	−1.284	11.735	0.000	549
Emerging	−0.071	0.071	1.248	4.957	−7.072	−0.552	5.259	0.000	549
Developed	−0.032	0.055	0.943	3.855	−8.729	−1.821	17.450	0.000	549
Asia	−0.027	0.042	1.199	3.786	−4.968	−0.354	3.863	0.000	549
BRIC	−0.023	0.072	1.284	4.654	−6.564	−0.469	4.490	0.000	549
EU	−0.034	0.083	1.005	3.569	−9.090	−1.838	16.278	0.000	549
North America	0.016	0.100	1.057	3.444	−5.849	−0.952	6.560	0.000	530
US	−0.012	0.083	1.093	4.840	−7.901	−0.963	11.131	0.000	549
Panel B: Descriptive statistics for post-COVID-19 period; March 11, 2020 – July 08, 2021
Commodities and Precious Metals
Copper	0.173	0.233	1.515	4.251	−7.005	−0.602	5.155	0.000	349
Gold	0.026	0.093	1.135	4.739	−5.519	−0.589	6.678	0.000	349
Iron	0.248	0.100	1.372	6.909	−9.130	−0.783	16.152	0.000	349
Silver	0.287	0.264	2.799	8.222	−12.671	−0.688	6.575	0.000	349
Global energy equity indices
Energy	0.086	0.057	2.592	12.731	−13.155	−0.251	8.827	0.000	349
Clean energy	0.195	0.366	2.688	10.800	−13.709	−0.757	7.575	0.000	349
Equity Market Indices
Global	0.114	0.220	1.679	8.383	−11.633	−1.377	17.265	0.000	349
Emerging	0.079	0.220	1.884	7.245	−13.329	−1.579	15.064	0.000	349
Developed	0.091	0.197	1.650	7.962	−12.184	−1.835	18.935	0.000	349
Asia	0.077	0.154	1.774	6.720	−12.073	−1.295	12.371	0.000	349
BRIC	0.060	0.162	1.864	6.638	−11.609	−1.106	10.023	0.000	349
EU	0.093	0.199	1.776	8.714	−12.391	−1.817	18.310	0.000	349
North America	0.113	0.149	1.284	5.980	−7.255	−0.378	8.095	0.000	339
US	0.121	0.179	1.774	8.968	−12.765	−1.033	17.584	0.000	349

Notes: Table reports descriptive statistics for daily returns on copper futures, gold, iron futures, silver futures, MSCI global energy producer equity index (Energy), MSCI global clean energy producer equity index (Clean), MSCI global equity index (Global), MSCI emerging markets equity index (Emerging), MSCI developed markets equity index (Developed), MSCI Asian markets equity index (Asia), MSCI BRIC markets equity index (BRIC), MSCI European markets equity index (EU), MSCI North American markets equity index (North A.), and S&P500 (US) for the pre-COVID period in Panel A and post-COVID period in Panel B. The time spans from January 3, 2018 to July 28, 2021. J-B indexes Jarque-Bera test of normality.

Against this backdrop, current paper examines diversification, hedging and safe-haven properties for two of the most liquid precious metals (i.e. gold and silver), two important metals used in production (i.e. iron, copper) and energy equity indices (i.e. global energy equity index and global clean energy equity index) for international equity market investors. The motivation behind the choice of these asset classes derived from their heterogenous relationship with equity market indices documented in the prior literature. While, energy related stocks represent another asset class with potential hedging or safe-haven properties given simultaneous green energy recovery stimulus plans initiated by a number of developed economies during the post-COVID-19 period (please see Chen et al., 2020; Helm, 2020). According to Wan et al. (2021), these developments should improve the performance of greener stocks.

Ability of financial commodities to protect investors’ portfolio during post-COVID-19 period is examined in a limited number of empirical papers. Corbet et al. (2020) reported that correlation coefficient among gold and equity returns has changed, from negative to positive, during the initial stage of COVID-19 in China, before the declaration of COVID-19 as a pandemic. Omane-Adjepong and Alagidede (2021) also failed to document any safe-haven property of gold for equity market returns in Africa. Results from Akhtaruzzaman et al. (2020) revealed that safe-haven properties of gold for international equity investors diminished after March 16, 2020. While, Ji et al. (2020) reported strong safe-haven property of gold for equity investors during December 2019 to March 2020. Dutta et al. (2020) examined safe-haven properties of gold for oil price crush amid to COVID-19 and revealed that gold can act as a safe-haven for oil price shocks during the initial phase of the pandemic. Although gold established itself in empirical works as a potential safe-haven asset, ability of different types of commodities to act as a safe-haven has been under researched for the post-COVID-19 period. Further, existing studies examine the safe-haven properties of gold during the initial stages of the COVID-19, focusing on specific region or market. Papers of Elgammal et al. (2021) and Shehzad et al. (2021a) are two exceptions which examined the interaction among global equity markets and gold during the post-COVID-19 period. Using a GARCH (1,1) model, Elgammal et al. (2021) examined mean and volatility spillovers among gold, energy and global equity market index for the period from January 13, 2015 to March 15, 2020, covering both pre-COVID and post-COVID periods. Their results revealed higher bidirectional spillovers among gold, and both equity and energy markets and higher cross-shock spillovers from energy to gold market during the post-COVID period, indicating lower benefits to diversification among these assets. While Shehzad et al. (2021b) compared the performance of gold and bitcoin to act as a safe-haven against the equity market indices of Asia, Europe and the US from January 06, 2020 to August 24, 2020, documenting that gold is a better safe-haven against equity market indices compared to bitcoin.

Different from the existing literature, current paper investigates the ability of four metals (i.e. copper, gold, iron and silver) and two energy indices (i.e. clean energy producer index and conventional energy producer index) to act as a safe-haven against eight different equity indices (i.e. global, emerging, developed, Asia, Europe, BRIC, North America and the US). Accordingly, results of this paper provide information for both regional and international equity investors for optimizing decisions during the post-COVID period via considering a wider range of potential safe-haven assets. Current paper also investigates a longer time span for the post-COVID period compared to existing papers, the sample ends in July 28, 2021. Therefore, results may provide guidance for investors with longer investment horizon. Further, a different methodology to investigate the dependence relationship among financial commodities and equity market indices is adopted. The novel Quantile Regression (QR) based model of Baur (2013) is used to test the impact of COVID-19 on the degree and structure of return relationship (i.e. return dependence) at different points of conditional return distributions of commodities and energy indices. The novelty of QR based model to capture the impact of shocks to financial asset prices derives from its ability to provide rich set of information about the dependence structure of financial asset returns for the pre-event and post-event periods at different points of conditional return distribution. Accordingly, it also accounts for dynamic changes in the dependence among financial asset returns. Specifically, it enables the decomposition of the impact of a shock on the degree of dependence and structure of dependence, allowing one to test contagion effect on and around median as well as lower and upper tails of return distributions. Thus, current results provide specific information for portfolio optimization and policy making. Further, according to Baur (2013), methods which neglect dependence in tails may incorrectly conclude a contagion effect among asset classes and thus fail to provide correct information for portfolio management and policy making. Results from author reveled that contagion of 2008 crisis occurred only at the upper tail of the conditional return distribution of equity indices, indicating that international equity market returns, in fact, became more dependent during good market state. Similar findings documented by Azimli (2020) for the impact of COVID-19 on industry returns and systematic risk connectedness in the US.

The QR has econometric advantages as well to model return dependence among random variables amid to black swan events like COVID-19. First, due to its dependence on median function rather than the mean, it is not required to make distributional assumptions related to the residuals since the model is robust to outliers (Baur and Schulze, 2005), which is a common feature of financial time series during financial turbulence. According to Bae et al. (2003), shocks lead to non-linear shifts and thus should not be modelled relying on the conditional mean function. Second, Hu (2006) showed that crises significantly change the linear dependence among random variables (see also Koenker, 2005). The QR approach can model any type of dependence structure among variables due to its ability to model the entire conditional distribution of the dependent variable overall 99 quantiles. It is not required to determine the shape and/or nature of the dependence structure among random variables (Yu et al., 2003). Because the shape and nature of structure is revealed by parameter estimates overall 99 quantiles. Paper of Bilgili et al. (2021) offers a nice discussion about the advantages of QR.

The results of this paper can be listed as follows. The return dependence among copper and international equity markets have significantly reduced at the upper-tail of conditional return distribution of copper during the post-COVID period. Degree of dependence among iron and international equity market returns reduced at the lower tail of conditional iron returns. Gold lost its weak (strong) hedge status against equity market (the US) returns given significant increase in return dependence around median and upper-tail of conditional gold returns during the post-COVID period. While, silver was a weak hedge during tranquil times and become a strong safe-haven against international equity markets during the post-COVID period at the upper tail of silver returns. And, dependence among energy stocks and equity markets significantly increased during the post-COVID period, eliminating any benefit to include them into broad equity market portfolios.

This paper has several contributions to the existing literature examining the impact of shocks on the dependence relationship among financial assets. First, none of the existing literature explicitly examines the safe-haven property of broad range of financial commodities for the post-COVID period for a long investment horizon. Earlier papers examined if gold is a safe-haven for Chinese equity markets (Corbet et al., 2020), African equity markets (Omane-Adjepong and Alagidede, 2021) and top 50 stocks in Euro area, top 225 in Japan, top 50 in China and the US (Akhtaruzzaman et al., 2020) before the declaration of COVID-19 as a pandemic by World Health Organization (WHO). Ji et al. (2020) reported that gold was a safe-haven against equity markets during December 2019 to March 2020. Although Elgammal et al. (2021) offers investigation among global equity, gold and energy markets during the post-COVID period, current paper differs in several aspects. For instance, this paper extends the analysis period for more than one year for the post-COVID period and analyzes safe haven ability of four different commodities and two energy equity indices (i.e. green and conventional energy) against eight different equity market indices. The dependence among commodities and energy indices with equity markets investigated at different points of conditional return distribution. Therefore, current paper documents degree of dependence at lower-tail, upper-tail and intermediate quantiles of commodity and energy index returns. Thus, current paper provides richer information for portfolio management and hedging strategies during post-COVID period for both regional and international equity investors as well as policy makers.

Second, by using the QR based approach of Baur (2013) current paper documents different stages of hedge and safe-haven properties of four metals and two energy indices against international and regional equity market indices. Specifically, if returns are uncorrelated (negatively correlated) at specific points of conditional return distribution during tranquil times the asset is said to be a weak (strong) hedge. While, if returns are uncorrelated (negatively correlated) at specific points of conditional return distribution during crisis times the asset is said to be a weak (strong) safe-haven (please refer to Baur and McDermott (2010) for further explanation). It is documented that copper is a weak safe-haven for the US equity markets during the post-COVID period at the upper-tail of conditional distribution of cooper returns due to the fact that copper returns are uncorrelated with the returns of the US equity market returns. While, silver is a strong safe-haven at the upper-tail of conditional return distribution of silver because silver returns further detached from the returns of the returns of equity market indices, coefficient estimates on the marginal change during the post-COVID period are negative and significant at the upper-tail of conditional silver returns. But, iron is a weak safe-haven only when iron returns are extremely positive due to negligible dependence of iron returns with equity markets. However, in contrast to early empirical evidence that gold can acts as safe-haven during early stages of COVID-19 (e.g. Ji et al., 2020), current results indicate that it loses its safe-haven status in the long-run. Corbet et al. (2020), Akhtaruzzaman et al. (2020) and Omane-Adjepong and Alagidede (2021) failed to document any safe-haven property of gold during the short-run, initial stage of COVID-19. Accordingly, investors and portfolio managers seeking alternative assets to protect their portfolios during turbulence should start looking for commodities other than gold. Current results reveal that copper, iron and silver are better candidates than gold for the post-COVID-19 period to protect portfolio risk.

Third contribution is related to trading behavior of investors amid to significant shocks. The results show that both degree of dependence and structure of dependence have significantly changed for most of the asset classes. And, majority of marginal change took place at tails of conditional return distributions. Accordingly, current findings highlight the importance of tail events during shocks to financial markets. Further, different regions reveal different dependence structure with financial commodities. Overall, emerging markets and BRIC equity indices show lower dependence with commodities during the post-COVID period. Accordingly, current results extend the literature investigating the ability of financial commodities for hedging and portfolio optimization with respect to regional and international equity markets during post-COVID-19 period (e.g. Akhtaruzzaman et al., 2020; Corbet et al., 2020; Elgammal et al., 2021; Shehzad et al., 2021b; Omane-Adjepong and Alagidede, 2021 are among others).

Documented findings are important for policy makers, portfolio managers, investors and researchers. Policy makers may use the results to initiate rescue measures and attain market stability during the post-COVID-19 period. According to the findings different markets have different dependence structure with commodities and energy indices. Therefore, the results may guide policy makers in different regions to craft strategies that address regional heterogeneity and different structure of return relationship of equity markets with commodity and energy index returns. Current results are also important for portfolio managers and investors in attaining better portfolio optimization during the post-COVID period. QR Method adopted can reveal the silent features of the dependence among different asset groups. The information related to dependence at tails is important for portfolio managers and investors wishing to hedge their risk during turbulence. Such analysis may provide precise information to engage in hedging strategies and understand the similarities and differences among the behavior of commodity and energy index returns under different market states. Therefore, information whether the hedging and safe-haven properties of commodities and energy stocks change amid to uncertainty would help investors and portfolio managers to optimize for mean and variance. Finally, current results are also relevant for researchers. A close attention should be paid towards structural changes in tail dependence among financial assets in face of significant shocks to economies in the future research. Because, the dependence structure among financial commodities and equity market indices significantly changed at tails while remained relatively constant on and around median during the post-COVID period.

Following the introduction, Section 2 presents the data and methodological framework employed. Section 3 reports empirical findings and discusses the implications of the degree and structure of dependence among financial assets during both pre-COVID period and post-COVID period. Section 4 concludes the paper with a summary of results and their implications.

Data and methods

Data

This paper uses daily prices for four different commodities, namely copper future prices, gold spot prices,1 iron future prices and silver future prices, two energy equity indices, namely MSCI global energy producer equity index and global clean energy producer equity index, and eight regional equity market indices, namely MSCI global equity market index, MSCI emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index and S&P500 equity index. Since the impact of COVID-19 can be characterized as a sudden shift in financial asset prices and investors react extreme negative shocks relatively fast daily prices are used. This choice is also in line with those of Baur and Lucey (2010) and Salmi et al. (2018) among others thus enables a direct comparison. Sampling period is from January 2, 2018 to July 29, 2021. Daily prices for commodities and indices are from investing.com. Returns used in empirical tests are obtained using the following equation:where Ri,t is return on commodity or energy index i on day t. Evidence of sharp increase/decrease in daily returns for commodities and equity indices can be seen in Fig. 1 . Prior literature mainly focused on the safe-haven properties of financial assets for equity investors for the period prior to the declaration of COVID-19 as a pandemic. However, there is a clear indication that the structure of returns has significantly changed towards the end of the first quarter of 2020 and extreme movements in asset returns happened around the same period. While most of the series have different structure during the sampling period, all of them experience high volatility starting from the end of the first quarter of 2020, March. Another notable finding is that the structure of most of the series never returned to their post-pandemic structure, justifying the need for a longer horizon investigation of the return association among these assets.

Fig. 1

Daily return plots for the variable between January 3, 2018 to July 28, 2021.

The descriptive statistics for returns are given in Table 1 for the pre-COVID period, January 03, 2018 to March 10, 2020, in Panel A and post-COVID period, March 11, 2020 to July 28, 2021, in Panel B. The table reports mean, median, standard deviation, maximum and minimum values, skewness, kurtosis, probability value from Jarque-Bera normality test and number of daily observations for two precious metals, two industrial metals, two energy indices and eight equity market indices. All the returns are calculated from US Dollar prices. In Panel A, highest mean value for daily returns is observed for silver, this is also evident in Fig. 1. Among commodities, only copper losses during the tranquil period. Silver has the highest mean daily return and also highest risk (standard deviation). Only positive skewness value is observed for silver indicating that investors in silver would have increased their wealth due to right-skewed distribution of silver returns. For instance, during the pre-COVID period the maximum daily gain (loss) for gold is 3.1% (−3.7%) and 11.6% (−7.4%) for silver. Lucey and Li (2015) revealed that silver offers safe-haven properties when gold does not. Mean daily returns for all the equity indices are negative, except for global clean energy producer equity index (Clean), before the pandemic. The highest daily lost experienced in global energy producer equity index (Energy) with −20.4% followed by iron, −16.6%.

In panel B, mean daily returns for all the financial assets examined are positive, silver has the highest mean. The lowest daily loss experienced in gold with −5.5% which accounts for only half of a single day loss experienced in other assets, except for copper. Despite positive mean daily returns on financial commodities and equity indices during the post-COVID period, daily distributions of all the returns are negatively skewed with high kurtosis values. Accordingly, return distributions are leptokurtic and thus showing high daily losses. Probability values of Jarque-Bera (p(J-B)) statistics reported in Column 8 reject the hypothesis that returns are normally distributed.

Correlation analysis among variables is provided in Table 2 for two sub-periods, pre-COVID period in Panel A and post-COVID period in Panel B. Results indicate higher correlation coefficients among returns during the post-COVID period compared to the tranquil period. Notably, gold and silver were negatively correlated with equity market indices during the tranquil period. But, the correlation coefficients have changed to positive during the post-COVID period. Corbet et al. (2020) also reported sign change among correlation coefficients of gold and equity market returns in China amid to COVID-19.

Table 2

Correlation results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
Panel A: Correlations for pre-COVID-19 period; January 03, 2018 – March 10, 2020
(1)Copper	1
(2)Gold	0.126	1
(3)Iron	0.070	−0.050	1
(4)Silver	0.068	0.233	−0.018	1
(5)Energy	0.343	−0.087	0.080	0.055	1
(6)Clean	0.307	0.007	0.038	0.045	0.667	1
(7)Global	0.353	−0.113	0.041	−0.017	0.752	0.746	1
(8)Emerg.	0.426	−0.009	0.034	−0.019	0.635	0.725	0.819	1
(9)Devel.	0.369	−0.039	0.046	−0.001	0.775	0.760	0.905	0.834	1
(10)Asia	0.430	−0.034	0.035	−0.026	0.580	0.684	0.791	0.976	0.798	1
(11)BRIC	0.439	−0.019	0.062	−0.029	0.611	0.714	0.789	0.968	0.794	0.957	1
(12)EU	0.382	−0.023	0.043	−0.004	0.751	0.742	0.879	0.806	0.973	0.769	0.764	1
(13)NA	0.244	−0.123	0.048	0.005	0.526	0.437	0.557	0.376	0.459	0.371	0.381	0.426	1
(14)US	0.278	−0.167	0.020	−0.032	0.718	0.727	0.933	0.783	0.872	0.755	0.747	0.841	0.543	1
Panel B: Correlations for post-COVID-19 period; March 11, 2020 – July 28, 2021
(1)Copper	1
(2)Gold	0.177	1
(3)Iron	0.191	−0.012	1
(4)Silver	0.314	0.718	0.104	1
(5)Energy	0.441	0.104	0.131	0.225	1
(6)Clean	0.289	0.252	0.023	0.305	0.541	1
(7)Global	0.381	0.261	0.102	0.297	0.734	0.764	1
(8)Emerg.	0.401	0.269	0.109	0.293	0.692	0.762	0.885	1
(9)Devel.	0.403	0.272	0.127	0.321	0.781	0.742	0.953	0.877	1
(10)Asia	0.387	0.264	0.109	0.281	0.645	0.746	0.856	0.988	0.841	1
(11)BRIC	0.371	0.249	0.095	0.270	0.616	0.737	0.818	0.969	0.799	0.972	1
(12)EU	0.402	0.267	0.121	0.336	0.782	0.737	0.941	0.855	0.989	0.816	0.778	1
(13)NA	0.248	0.066	0.122	0.153	0.437	0.502	0.583	0.487	0.580	0.478	0.461	0.563	1
(14)US	0.321	0.243	0.083	0.261	0.682	0.736	0.973	0.847	0.909	0.815	0.780	0.898	0.553	1

Model and methods

Empirical model employed in this paper is a manifestation of novel QR approach introduced by Baur (2013) to examine the impact of COVID-19 on the degree of dependence and structure of dependence on various asset classes. This approach is based on the base QR model of Koenker and Bassett (1978) which can be specified as follows:where Qy(τ) indexes the τth conditional quantile of dependent variable (y) under the assumption of linear dependence on vector of independent variables (x). The dependence on each quantile is given by the coefficient estimates, (β(τ)). Specifically, equation (2) models the dependence among the specific quantiles of y and unspecific quantiles of x and β estimates the impact of x on different points of the conditional distribution of y. According to Baur (2013), the QR framework reveals the structure of dependence based on the vector of β estimates at different quantiles depending on the degree of dependence of y and x at each quantile. The average degree of dependence reveled by averaging β estimates over all quantiles.

The empirical model which allows to test these patterns by decomposing dependence into two factors, (i) the degree of dependence and (ii) the structure of dependence, takes the following form:where indexes either daily returns on commodities (i.e. copper, gold, iron and silver) or energy equity indices (i.e. MSCI global energy producer index and MSCI global clean energy producer index); x indexes one of the eight equity market index mentioned in Section 2.1 (i.e. MSCI global equity market index, MSCI emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index, and S&P500); and D is dummy taking a value of 1 for the period March 11, 2020 to July 28, 2021, and otherwise zero. Accordingly, Equation (2) specified above reveals the degree and structure of dependence for the full sample as well as for the post-COVID period. More specifically, Equation (2) estimates the impact of COVID-19 on the degree and structure of dependence on individual assets conditional on the quantile of the return distribution on asset i (i.e. commodity returns and energy index returns). Lower quantiles reveal the dependence during bad state and higher quantiles reveal the dependence during good state of asset returns. By this way, dynamic relationship for tranquil times and crisis times is estimated with the help of the D .

According to Baur (2013), change in structure and degree of dependence induced by a shock is given by averaging the coefficient estimate of the interaction term, . If is larger (smaller) than 0 there is a positive (negative) change in the degree of dependence induced by the shock relative to average dependence given by . Specifically, provides the marginal change and the full effect can be obtained by adding up the coefficients of for a specific quantile. While a change in the structure of dependence happens if coefficient estimates of varies across different quantiles. By this way, coefficient estimates provide an estimation of dependence at different points of conditional distribution before and after the shock while accounting for any change in the structure of dependence induced by the shock. This framework enables us to examine the degree and nature of relationship among returns during tranquil and crisis times, revealing if tested assets become more coupled or decoupled with broad equity indices at different points of conditional return distribution. By this way, information related to hedging and/or safe-haven properties of examined assets against equity portfolios revealed.

Empirical results

This section reports empirical results from applying the QR framework explained under Section 2.2 to a set of commodities and equity market indices as well as energy producer indices and equity market indices. There is no need to make a priori about the structure of association among returns because the QR framework can assume any type of structure given the estimates across different quantiles. Dependence could have a constant structure if estimates are constant across all quantiles or could take other forms such as symmetric if estimates are lower (higher) in tails with equal coefficients and asymmetric with a left-tail dependence if estimates are higher in the lower-tail or asymmetric with a right-tail dependence if estimates are higher in the upper-tail. In addition, the structure of dependence could be a combination of any of these structures. To avoid the burden of an ad hoc choice, the estimations are made for all 99 quantiles but coefficients are reported for a selection of quantiles to improve readability.2

Dependence among copper and equity market returns

Table 3 reports estimates for the dependence among copper and international equity market returns during tranquil times, given by β coefficients in Panel A, and marginal change in return dependence during post-COVID-19 period, given by coefficients in Panel B. Except for North America and the US, β estimates in Panel A reveal an asymmetric u-shaped dependence with higher coefficients at tails compared to intermediate quantiles. The dependences among copper and global, emerging, developed, Asia, Europe and BRIC markets are positive and significant almost in every quantile of conditional return distribution of copper. But, coefficient estimates are considerably lower than 1.00 indicating that including copper into equity portfolios of these markets may provide diversification benefit during normal times. The dependence for North American and the US market with copper have different structure, their returns are uncorrelated at the upper-tail of copper returns showing that during good state of copper returns it may act as a weak hedge against North American and the US equity market investors. Copper is a better hedge for North American equity investors than the US given by insignificant β estimates of North American equity index for 90th quantile above. It is important to note that, as can be seen in Table 1, mean daily returns on copper during January 03, 2018 to March 10, 2020 is negative thus its inclusion into equity portfolios may increase downside risk (see also Hillier et al., 2006). Finally, the last raw of Panel A reports average degree of dependence across all quantiles. Developed markets equity index has the highest dependence degree with copper while North American equity index has the lowest. Accordingly, hedging role of copper may differ for different equity markets.

Table 3

Estimation results for copper.

Quantile	Global	Emerging	Developed	Asia	Europe	BRIC	North America	US
Panel A: Pre-COVID period (β); 03/01/2018–March 10, 2021
0.01	0.621 (1.137)	0.404* (1.836)	0.394 (1.151)	0.455* (1.872)	0.481 (1.330)	0.388* (1.936)	0.069 (0.741)	0.501 (0.771)
0.02	0.718*** (5.649)	0.580*** (8.891)	0.459 (1.004)	0.588*** (5.653)	0.855*** (9.607)	0.595*** (0.069)	0.229 (1.149)	0.657*** (7.614)
0.03	0.569*** (4.015)	0.452*** (3.356)	0.607*** (4.937)	0.537*** (3.231)	0.725*** (8.072)	0.577*** (7.375)	0.359** (2.141)	0.586*** (6.407)
0.04	0.494*** (3.868)	0.392*** (6.598)	0.574*** (5.560)	0.371*** (5.176)	0.645*** (8.374)	0.450*** (6.407)	0.471*** (2.924)	0.495*** (4.326)
0.05	0.485*** (4.335)	0.401*** (6.650)	0.552*** (4.540)	0.413*** (4.609)	0.606*** (6.699)	0.430*** (7.648)	0.333 (0.742)	0.341 (1.374)
0.10	0.378*** (3.649)	0.338*** (5.801)	0.437*** (4.685)	0.405*** (5.819)	0.457*** (6.050)	0.359*** (6.095)	0.169 (1.392)	0.248** (1.988)
0.25	0.273*** (6.020)	0.284*** (4.907)	0.343*** (5.229)	0.307*** (7.114)	0.360*** (6.069)	0.290*** (5.750)	0.221*** (5.101)	0.214*** (6.372)
0.50	0.360*** (6.834)	0.314*** (8.707)	0.384*** (5.718)	0.338*** (8.472)	0.362*** (5.153)	0.299*** (8.302)	0.233*** (4.698)	0.268*** (6.065)
0.75	0.435*** (5.828)	0.393*** (10.239)	0.461*** (5.454)	0.409*** (8.481)	0.425*** (6.279)	0.397*** (11.284)	0.295*** (6.639)	0.319*** (8.126)
0.90	0.406*** (13.946)	0.414*** (5.574)	0.424*** (8.545)	0.449*** (6.657)	0.414*** (6.787)	0.451*** (7.590)	0.374*** (3.760)	0.320*** (5.477)
0.95	0.390*** (3.163)	0.389*** (2.922)	0.386*** (2.581)	0.409*** (4.146)	0.375** (1.986)	0.349*** (3.535)	0.156 (0.441)	0.245*** (3.843)
0.96	0.350*** (3.748)	0.333*** (3.930)	0.363*** (2.954)	0.346*** (3.788)	0.334*** (3.452)	0.335*** (3.635)	0.058 (0.409)	0.274*** (3.242)
0.97	0.347*** (3.481)	0.331*** (3.996)	0.313*** (4.049)	0.356*** (3.969)	0.310*** (3.602)	0.337*** (3.730)	0.150 (0.488)	0.253*** (3.033)
0.98	0.417*** (2.841)	0.412*** (2.964)	0.361*** (3.533)	0.505*** (2.592)	0.356*** (3.684)	0.459*** (3.118)	0.103 (0.262)	0.098 (1.100)
0.99	0.625*** (11.861)	0.542*** (3.081)	0.524*** (8.188)	0.687*** (2.963)	0.490*** (8.277)	0.596*** (3.759)	−0.222* (−1.891)	−0.071 (−0.664)
Average overall τ	0.378	0.354	0.417	0.381	0.401	0.357	0.241	0.277
Panel B: Change in COVID period (θ);11/03/2020–July 28, 2021
0.01	−0.165 (−0.304)	0.012 (0.054)	0.129 (0.379)	0.016 (0.065)	0.029 (0.078)	0.001 (0.007)	0.466 (0.569)	−0.139 (−0.221)
0.02	−0.382*** (−3.167)	−0.239*** (−3.188)	−0.069 (−0.139)	−0.199* (−1.717)	−0.489*** (−5.539)	−0.261*** (−3.188)	0.034 (0.126)	−0.383*** (−4.670)
0.03	−0.283** (0.047)	−0.119 (−0.841)	−0.283** (−2.100)	−0.162 (−0.949)	−0.386*** (−3.040)	−0.260*** (−2.757)	0.141 (0.542)	−0.350*** (−4.040)
0.04	−0.270** (−0.036)	−0.003 (−0.010)	−0.166 (−0.711)	−0.072 (−0.662)	−0.212 (−0.846)	−0.105 (−0.408)	0.036 (0.166)	−0.296*** (−2.827)
0.05	−0.161 (−0.662)	−0.042 (−0.197)	−0.105 (−0.429)	−0.065 (−0.373)	−0.182 (−0.909)	−0.094 (−0.469)	0.222 (0.515)	−0.170 (−0.693)
0.10	−0.087 (−0.649)	−0.014 (−0.119)	0.031 (0.242)	−0.104 (−1.019)	−0.092 (−0.564)	−0.088 (−1.048)	0.262* (1.960)	−0.031 (−0.221)
0.25	0.070 (0.984)	0.083 (1.000)	0.011 (0.134)	0.065 (0.977)	−0.025 (−0.324)	0.080 (1.151)	0.055 (0.764)	0.046 (0.649)
0.50	0.023 (0.796)	0.004 (0.042)	0.017 (0.086)	−0.037 (−0.409)	−0.018 (−0.171)	−0.027 (−0.411)	−0.036 (−0.445)	0.010 (0.109)
0.75	−0.070 (−0.737)	−0.037 (−0.360)	−0.076 (−0.704)	−0.049 (−0.459)	−0.072 (−0.834)	−0.100* (−1.950)	−0.048 (−0.732)	0.014 (0.240)
0.90	−0.052 (0.203)	−0.145*** (−2.960)	−0.056 (−0.945)	−0.109 (−1.543)	−0.111 (−1.362)	−0.183*** (−2.677)	−0.058 (−0.510)	−0.062 (−0.971)
0.95	−0.101 (−.805)	−0.102 (−0.761)	−0.132 (−0.884)	−0.091 (−0.896)	−0.118 (−0.618)	−0.057 (−0.566)	0.136 (0.390)	0.065 (0.882)
0.96	−0.053 (−0.549)	−0.086 (−0.923)	−0.104 (−0.841)	−0.065 (−0.654)	−0.070 (−0.687)	−0.048 (−0.477)	0.211 (1.463)	−0.004 (−0.018)
0.97	−0.038 (−0.366)	−0.031 (−0.352)	−0.037 (−0.455)	−0.030 (−0.317)	−0.034 (−0.379)	−0.070 (−0.702)	0.091 (0.292)	−0.005 (−0.026)
0.98	−0.139 (−0.412)	−0.181 (−1.140)	−0.068 (−0.661)	−0.272 (−1.235)	−0.065 (−0.654)	−0.243 (−1.503)	0.100 (0.253)	0.037 (0.456)
0.99	−0.550*** (−5.366)	−0.352 (−1.612)	−0.178*** (−3.376)	−0.549* (−1.943)	−0.154*** (−2.905)	−0.448** (−2.440)	0.353*** (3.242)	0.133 (1.596)
Average overall τ	−0.031	−0.019	−0.033	−0.047	−0.057	−0.057	0.040	−0.011

Notes: This table presents the coefficient estimations for the relationship between copper future returns (dependent variable) and equity market returns including MSCI global equity market index, MSCI emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index, and S&P500 (independent variables). Panel A reports average degree of dependence for the pre-COVID period given by β estimates and Panel B reports marginal change induced by COVID-19 shock given by ϴ estimates. Estimation period is from January 03, 2018 to July 28, 2021. Numbers in parentheses are t-statistics. ***, ** and * indicate statistical significance 1%, 5% and 10% level, respectively. Model.

Panel B reports estimates for to assess the marginal effect on the degree and structure of dependence among copper and equity market returns during the post-COVID-19 period. The structure of marginal change during post-COVID-19 period is an inverted u-shape for global, developed, Europe and BRIC equity market indices with higher negative dependence coefficients at tails of copper returns. Accordingly, the dependence structure among copper returns and equity market indices reduced at both tails of conditional copper returns. For example, full effect during the post-COVID period on the degree of dependence among copper and global equity index at τ = 99 is given by βglobal(0.99) + global(τ = 0.99) = 0.625–0.550 = 0.075. Showing that, when copper experiences extreme positive returns, the dependence becomes negligible thus copper may act as a weak safe-haven asset for global equity market investors. The degree of dependence on left-tail has also reduced but not as large as the reduction observed for right-tail. Similar pattern is also observable for BRIC market index. But for developed and European equity market indices the reduction in dependence at τ = 99 is considerably lower, βdeveloped(0.99) + developed(τ = 0.99) = 0.524–0.178 = 0.346 and βEurope(0.99) + Europe(τ = 0.99) = 0.490–0.154 = 0.336. Therefore, benefits to diversification have increased for developed and Europe equity indices during post-COVID period but copper cannot act as a safe-haven against these markets. Marginal effect of COVID on the degree of dependence among copper and Asian market returns has the same structure, an inverted u-shape with higher negative values at tails, albeit coefficient estimates of Asia are weakly significant at both tails. For emerging markets equity index, marginal change during COVID-19 is only significant at the lower quantiles with negative estimates of . But, for North America return dependence at τ = 99 has increased and sign of dependence changed from negative to positive, eliminating ability of copper to act as a safe-haven for North American equity portfolio. And, the degree of dependence among copper and US market returns have significantly reduced at the lower-tail of copper returns while the estimates of US are all insignificant at upper-tail. Accordingly, copper and US market returns are uncorrelated during good states of copper returns showing that copper may act as a weak safe-haven for US equity index during good episodes of copper returns.

Overall, except for North America and emerging market portfolios, copper and equity market returns have decoupled during the post-COVID-19 period at both tails of conditional copper returns. Specifically, returns of copper and the US equity index is unrelated, indicating that copper may act as a weak safe-haven asset for the US equity market investors during post-COVID period. Diversification benefits of copper, except for North America, have also increased due to lower dependence structure at different return states of cooper. It is important to note that, daily returns on copper futures have averaged to 0.173% (please see Panel B of Table 1) during the post-COVID period, which is higher than average daily returns of all the equity market indices. Therefore, copper may help to improve risk-return characteristics of equity portfolios during the post-COVID period. This finding contradicts with those of Kopyl and Lee (2016) who examined if copper can provide safe-haven properties during market downturns for a number of developed market equity indices and reported that it cannot for the period from 1964 to 2014. Further, Sadorsky (2014) reported increased dynamic correlation among stock and copper returns for the post-2008 period, indicating lower benefits to include copper in the stock portfolios.

Fig. 2 plots pre-COVID, post-COVID and marginal change during post-COVID period in the degree and the structure of dependence among copper and global equity market returns.3 Plot indicates change from a u-shaped dependence structure during tranquil period to an asymmetric dependence with lower values at both extremes during the post-COVID period.

Fig. 2

Estimation Results: Dependence among copper and MSCI global equity index before (pre-10/03/2020) and after the declaration of COVID-19 as a global pandemic (post-11/03/2021).

Dependence among gold and equity market returns

This section formally tests if gold is a safe-haven asset against equity market indices for the post-COVID period. Evidence during the early stage of the pandemic showed that gold cannot act as a safe-haven against Chinese (Corbet et al., 2020) and African equity markets (Omane-Adjepong and Alagidede, 2021). Results of Akhtaruzzaman et al. (2020) further suggest that safe-haven property of gold for international equity investors has diminished following the mid of March 2020.

Panel A of Table 4 reports the coefficient estimates for the dependence among gold and international equity market returns considering different points of conditional distribution of gold returns. Overall, during the pre-pandemic period gold can act as a strong hedge for North American and US equity index portfolios given that the coefficient loadings are significant and negative for the intermediate (for both) and upper quantiles (for the US only) while insignificant for lower (for the US). Accordingly, equity market portfolios tend to lose when gold returns are positive but they become uncorrelated when gold returns are in a bad state, indicating strong hedging potential of gold. For the other markets, gold may act as a weak hedge given the insignificant coefficient estimates of βs over different quantiles.

Table 4

Estimation results for gold.

Quantile	Global	Emerging	Developed	Asia	Europe	BRIC	North America	US
Panel A: Pre-COVID period (β); 03/01/2018–March 10, 2021
0.01	0.091 (0.649)	0.077 (0.602)	0.089 (0.587)	−0.300 (−1.290)	0.098 (0.605)	0.221 (1.604)	0.306*** (2.841)	0.084 (0.710)
0.02	−0.056 (−0.959)	−0.091 (−1.246)	−0.088 (−1.008)	−0.111 (−1.162)	−0.069 (−0.876)	−0.083 (1.248)	0.409*** (7.330)	−0.048 (−0.919)
0.03	−0.100 (−1.356)	−0.078 (−1.600)	−0.103 (−1.438)	−0.101 (−1.576)	−0.034 (−0.234)	−0.081 (−1.406)	0.332 (1.517)	−0.092 (−1.405)
0.04	−0.118 (−1.594)	−0.050 (−0.475)	−0.125 (−1.619)	−0.062 (−0.438)	0.042 (0.223)	−0.114** (−2.001)	0.179 (0.655)	−0.113* (−1.707)
0.05	−0.116 (−1.222)	−0.034 (−0.278)	−0.055 (−0.247)	−0.036 (−0.319)	0.084 (0.551)	−0.044 (−0.335)	0.150 (0.847)	−0.088 (−0.799)
0.10	−0.007 (−0.056)	0.031 (0.463)	0.037 (0.362)	0.032 (0.478)	0.037 (0.361)	0.018 (0.301)	0.027 (0.364)	−0.089 (−0.683)
0.25	−0.040 (−0.747)	0.029 (0.766)	−0.002 (−0.029)	0.019 (0.488)	0.004 (0.060)	0.022 (0.642)	−0.096*** (−3.618)	−0.070** (−2.217)
0.50	−0.022 (−0.842)	0.001 (0.041)	−0.011 (−0.447)	−0.018 (−0.683)	0.014 (0.527)	−0.014 (−0.579)	−0.075* (−1.843)	−0.053* (−1.670)
0.75	−0.102** (−2.402)	0.004 (0.128)	−0.043 (−0.830)	−0.033 (−0.914)	−0.043 (−0.965)	−0.007 (−0.192)	−0.143*** (−3.208)	−0.144*** (−3.726)
0.90	−0.215*** (−3.745)	−0.107** (−2.316)	−0.150* (0.053)	−0.106*** (−2.418)	−0.163** (−2.220)	−0.095* (−1.665)	−0.203*** (−3.872)	−0.189*** (−4.158)
0.95	−0.278*** (−4.401)	0.016 (0.258)	0.005 (0.040)	−0.005 (−0.094)	0.016 (0.127)	0.002 (0.035)	−0.224** (−2.375)	−0.223*** (−4.262)
0.96	−0.246** (−2.291)	0.021 (0.331)	−0.024 (−0.162)	0.021 (0.333)	−0.077 (−0.326)	0.013 (0.175)	−0.268*** (2.605)	−0.205*** (−4.203)
0.97	−0.194 (−1.480)	−0.070 (−0.573)	−0.173 (−1.398)	−0.095 (−0.770)	−0.157 (−1.618)	−0.086 (−0.776)	−0.299*** (−3.375)	−0.229*** (−3.722)
0.98	−0.343 (−1.480)	−0.087 (−0.982)	−0.130 (−1.225)	−0.135 (−1.555)	−0.102 (−1.169)	−0.067 (−0.690)	−0.223 (−1.266)	−0.290*** (−4.124)
0.99	−0.222 (−1.540)	−0.155*** (−2.255)	−0.289 (−1.038)	−0.122 (−1.098)	−0.227 (−0.890)	−0.113 (−1.146)	−0.123 (−0.749)	−0.336*** (−7.138)
Average overall τ	−0.083	−0.004	−0.033	−0.024	−0.017	−0.012	−0.087	−0.104
Panel B: Change in COVID period (θ);11/03/2020–July 28, 2021
0.01	0.053 (0.328)	0.070 (0.478)	0.042 (0.247)	0.470** (2.129)	0.032 (0.183)	−0.175 (−1.636)	−0.095 (−0.902)	0.069 (0.514)
0.02	0.105 (1.344)	0.125 (1.608)	0.131 (1.352)	0.117 (1.199)	0.102 (1.209)	0.105 (1.401)	−0.383*** (−4.038)	0.115 (1.306)
0.03	0.192** (2.060)	0.145 (1.574)	0.206** (2.294)	0.187* (1.643)	0.127 (0.807)	0.139** (1.974)	−0.281 (−1.244)	0.183* (1.870)
0.04	0.225** (2.356)	0.115 (1.110)	0.204** (2.404)	0.147 (1.063)	0.021 (0.113)	0.195*** (2.805)	−0.101 (−0.593)	0.240** (2.180)
0.05	0.216** (1.975)	0.127 (1.050)	0.156 (0.725)	0.183 (1.229)	0.003 (0.020)	0.203 (1.401)	−0.111−0.593()	0.210* (1.640)
0.10	0.213 (1.492)	0.091 (1.183)	0.163 (1.365)	0.124 (1.439)	0.132 (1.140)	0.127* (1.714)	−0.093 (−1.104)	0.293** (2.012)
0.25	0.225** (2.044)	0.150*** (2.868)	0.245** (2.022)	0.175*** (2.744)	0.186 (1.523)	0.118* (1.952)	0.075 (0.889)	0.195*** (2.739)
0.50	0.201*** (3.034)	0.172*** (3.658)	0.201*** (2.809)	0.175*** (3.639)	0.164*** (2.389)	0.142*** (3.932)	0.076 (0.973)	0.206*** (3.547)
0.75	0.275*** (4.839)	0.174*** (3.229)	0.252*** (3.724)	0.185*** (4.120)	0.222*** (3.529)	0.150*** (3.558)	0.170** (2.181)	0.310*** (5.647)
0.90	0.419*** (3.916)	0.306*** (2.838)	0.367*** (3.449)	0.304*** (3.672)	0.356*** (3.024)	0.279*** (4.488)	0.308*** (3.038)	0.347*** (6.825)
0.95	0.505*** (7.688)	0.195** (2.482)	0.291** (2.242)	0.221*** (3.738)	0.249* (1.913)	0.217*** (3.205)	0.386** (2.344)	0.419*** (7.818)
0.96	0.470*** (4.358)	0.175*** (2.692)	0.286* (1.926)	0.196*** (3.051)	0.326 (1.383)	0.215*** (0.739)	0.486*** (2.940)	0.411*** (8.158)
0.97	0.430*** (3.577)	0.278** (2.278)	0.426*** (3.422)	0.324*** (2.612)	0.382*** (3.895)	0.324*** (2.869)	0.551*** (5.281)	0.442*** (6.929)
0.98	0.599** (2.577)	0.309*** (3.411)	0.400*** (3.709)	0.365*** (4.144)	0.340*** (3.859)	0.287*** (2.713)	0.422** (2.389)	0.493*** (4.784)
0.99	0.479*** (3.290)	0.397*** (5.299)	0.595** (2.123)	0.382*** (3.437)	0.492* (1.915)	0.274* (2.435)	0.383** (2.398)	0.593*** (12.091)
Average overall τ	0.269	0.177	0.237	0.199	0.196	0.161	0.113	0.264

Notes: This table presents the coefficient estimations for the relationship between gold returns (dependent variable) and equity market returns including MSCI global equity market index, MSCI emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index, and S&P500 (independent variables). Panel A reports average degree of dependence for the pre-COVID period given by β estimates and Panel B reports marginal change induced by COVID-19 shock given by ϴ estimates. Estimation period is from January 03, 2018 to July 28, 2021. Numbers in parentheses are t-statistics. ***, ** and * indicate statistical significance 1%, 5% and 10% level, respectively. Model.

Panel B reports the marginal change in the degree of dependence during the post-COVID-19 period. The results suggest a significant change in the structure of dependence, from left-tailed dependence into a right-tailed dependence for all the equity market indices. Almost all the estimates are significant and positive on and above 25th percent quantile of conditional return distribution of gold. For example, degree of dependence among gold and global markets equity index returns increased from −0.222 during tranquil period to 0.258 during post-COVID period at 99th quantile, given by β(0.99) + (τ = 0.99) = −0.222 + 0.479 = 0.258. Similar shifts are observable for all the other equity indices. Highest marginal change experienced for developed markets equity index and US equity index. This finding is in line with those of Corbet et al. (2020) that reported increased in correlation coefficient among gold and Chinese market equity index amid to COVID-19. Akhtaruzzaman et al. (2020) also reported that gold has lost its safe-haven property following March 16, 2020 for international equity investors. And, Elgammal et al. (2021) documented increased mean and volatility spillovers among gold and global equity market portfolio for the post-COVID-19 period.

Fig. 3 illustrates the dependence among gold and global equity market index for the full sample and sub-sample along with marginal change. The thin black line shows dependence for the tranquil period, bold grey line shows the marginal change in the dependence for the post-COVID and dashed black line shows the total dependence for the post-COVID period indicating an overall increase in the average dependence compared to tranquil period. This structural change in dependence has converted gold from being a hedge asset to an asset that only offers diversification benefits to international equity investors. Albeit, ability of gold to improve risk-return characteristics of equity portfolios is limited due to its low daily returns during the post-COVID period (0.026%) as reported in Panel B of Table 1. Salisu et al. (2021) reported that gold can provide hedge against US sectoral equity returns but the hedging performance of gold is lower during the initial stage of the COVID-19 period. The current findings further reveal that gold lost its status as a hedge asset during the post-COVID period for investors with relatively longer investment horizon. Dimitriou et al. (2020) reported that gold can act as a safe-haven asset for MSCI global equity market index amid to 2008 financial crisis and European Sovereign Debt Crisis. Results obtained here suggest that gold does not act as a safe-haven asset for international equity investors, including the MSCI global portfolio, during the post-COVID period. Another notable result related to safe-haven property of gold is reported by Baur and Lucey (2010) indicating that gold can act as a safe-haven only during the initial stage of global financial crisis of 2008. Further analysis conducted to test their finding for the post-COVID period. According to unreported results gold cannot act as a safe-haven against international equity market returns for the first fifteen days following the declaration of COVID as pandemic.

Fig. 3

Estimation Results: Dependence among gold and MSCI global equity index before (pre-10/03/2020) and after the declaration of COVID-19 as a global pandemic (post-11/03/2021).

Dependence among iron and equity market returns

Estimates for dependence among iron and international equity market returns reported in Panel A of Table 5 show an asymmetric left-tailed dependence during the tranquil times. Except for North American equity market index, there are positive and significant dependence estimates among iron and equity market returns at quantile 2 with stronger degree of dependence for global (slightly lower than 1.00), developed (slightly lower higher 1.00), and European (slightly higher than 1.00) equity markets. Accordingly, returns of iron and equity market indices move together only at a specific bad state of iron returns (i.e. at τ = 0.02). Except for North America and the US, iron and equity market returns are positively and significantly dependent at τ = 0.98. While, when iron experiences extreme positive returns (i.e. at τ = 0.99), iron and international equity market returns are uncorrelated. Accordingly, iron is a weak hedge against international equity markets when iron returns are at τ = 0.99. It is also important to note that, daily returns on iron have averaged to 0.034% but average returns on equity indices are all negative during the pre-COVID-19 period. For North American equity index, dependence coefficient is negative, significant and high (−0.724) at the extreme left-tail of conditional return distribution of iron (τ = 0.01). However, coefficient estimates of βNorth America are negative on and above 90th percent quantile. Accordingly, iron provides strong (weak) hedging against North American equity market returns when iron returns are at a bed (good) state.

Table 5

Estimation results for iron.

Quantile	Global	Emerging	Developed	Asia	Europe	BRIC	North America	US
Panel A: Pre-COVID period (β); 03/01/2018–March 10, 2021
0.01	0.346 (1.014)	0.301 (0.341)	0.308 (0.769)	0.287 (0.434)	0.321 (0.503)	0.326 (0.525)	−0.724*** (−3.979)	0.3324 (0.965)
0.02	0.914*** (12.756)	0.741*** (5.801)	1.078*** (7.801)	0.734*** (6.756)	1.082*** (3.749)	0.811*** (6.032)	0.079 (0.230)	0.644** (2.313)
0.03	0.130 (0.211)	0.166 (0.245)	−0.072 (−1.091)	0.449 (0.814)	−0.090*** (−4.024)	0.498 (1.537)	0.2380.904	−0.103*** (−3.950)
0.04	0.0396 (0.012)	0.237 (0.523)	−0.071** (−2.331)	0.224 (0.717)	−0.065** (−2.059)	0.406 (1.612)	0.254 (0.535)	−0.079** (−2.318)
0.05	−0.029 (−0.791)	0.047 (0.113)	−0.031 (−1.131)	0.227 (0.711)	−0.031 (−1.159)	0.269 (1.335)	0.234* (1.705)	−0.035 (−0.934)
0.10	0.056** (1.969)	0.068** (1.970)	0.053** (2.254)	0.094* (1.665)	0.050** (2.032)	0.133 (1.450)	0.116** (2.289)	0.043 (1.206)
0.25	0.083 (1.603)	0.084*** (2.651)	0.088** (1.974)	0.059 (1.585)	0.067 (1.555)	0.101*** (3.257)	0.125*** (2.606)	0.054 (1.162)
0.50	0.039 (1.168)	0.027 (1.209)	0.041 (1.023)	0.030 (1.286)	0.036 (1.058)	0.034 (1.573)	0.046 (1.515)	0.013 (0.420)
0.75	0.040 (0.623)	−0.006 (−0.132)	0.030 (0.409)	−0.010 (−0.226)	0.033 (0.524)	0.010 (0.216)	0.088** (2.153)	−0.003 (−0.068)
0.90	0.171*** (2.538)	0.100 (1.328)	0.232*** (4.058)	0.089 (1.302)	0.217*** (3.148)	0.102 (1.364)	0.152 (1.078)	0.078 (0.651)
0.95	0.215** (2.014)	0.169 (1.122)	0.150 (0.377)	0.153 (1.321)	0.143 (0.760)	0.219** (2.023)	0.261 (0.703)	0.080 (1.504)
0.96	0.134** (2.213)	0.191 (0.905)	0.164** (2.098)	0.210 (1.429)	0.142** (2.438)	0.336*** (2.662)	0.388 (0.996)	0.119*** (2.674)
0.97	0.250* (1.845)	0.239 (1.333)	0.212** (2.100)	0.297 (1.366)	0.191* (1.800)	0.316* (1.825)	0.423 (0.606)	0.193** (1.977)
0.98	0.332** (2.411)	0.333*** (2.962)	0.277*** (2.736)	0.365*** (3.159)	0.280*** (2.698)	0.365*** (3.212)	0.449 (0.140)	0.154 (0.077)
0.99	0.138 (0.464)	0.329 (0.162)	0.273 (0.643)	0.391 (0.175)	0.111 (0.428)	0.351 (0.263)	0.281 (0.554)	0.151 (0.382)
Average overall τ	0.082	0.068	0.083	0.075	0.078	0.090	0.103	0.044
Panel B: Change in COVID period (θ);11/03/2020–July 28, 2021
0.01	0.234 (0.563)	0.215 (0.059)	0.330 (0.669)	0.223 (0.063)	0.276 (0.371)	0.071 (0.046)	1.489*** (13.013)	0.220 (0.540)
0.02	−0.546*** (−7.268)	−0.968*** (−6.969)	−0.689*** (−5.112)	−0.971*** (−8.073)	−0.698** (−2.493)	−1.018 (−0.250)	0.361 (0.373)	−0.547 (−0.544)
0.03	−0.060 (−0.072)	−0.317 (−0.456)	0.080 (0.102)	−0.631 (−1.074)	0.067 (0.112)	−0.689* (1.933)	0.116 (0.427)	−0.047 (−1.598)
0.04	−0.125 (−0.037)	−0.310 (−0.615)	−0.039 (−0.172)	−0.297 (−0.848)	−0.042 (−0.180)	−0.487 (−1.570)	0.027 (0.059)	−0.014 (−0.064)
0.05	0.060 (0.218)	0.008 (0.016)	0.068 (0.200)	−0.201 (−0.495)	−0.080*** (−3.607)	−0.250 (--0.886)	−0.020 (−0.130)	0.063 (0.213)
0.10	−0.038 (−0.183)	0.025 (0.492)	−0.085 (−1.391)	0.008 (0.105)	−0.079 (−1.541)	−0.035 (−0.333)	−0.028 (−0.345)	−0.007 (−0.030)
0.25	−0.066 (−1.181)	−0.063 (−1.620)	−0.070 (−1.367)	−0.023 (−0.500)	−0.051 (−1.048)	−0.065* (−1.654)	−0.098* (−1.676)	−0.032 (−0.642)
0.50	−0.021 (−0.589)	−0.004 (−0.167)	−0.012 (−0.299)	−0.001 (−0.031)	−0.005 (−0.144)	−0.013 (−0.493)	0.005 (0.130)	0.004 (0.127)
0.75	0.012 (0.199)	0.060 (1.357)	0.017 (0.233)	0.067 (1.508)	0.014 (0.412)	0.043 (0.906)	−0.051 (−0.668)	0.050 (1.023)
0.90	−0.043 (−0.666)	0.012 (0.156)	−0.114** (−2.015)	0.030 (0.449)	−0.095 (−1.450)	0.004 (0.042)	0.047 (0.329)	0.038 (0.323)
0.95	−0.033 (−0.284)	−0.014 (−0.090)	0.028 (0.066)	0.026 (0.211)	0.033 (0.160)	−0.037 (−0.358)	0.037 (0.099)	0.089 (1.391)
0.96	0.081 (1.345)	−0.056 (−0.026)	0.037 (0.451)	−0.003 (−0.022)	0.059 (0.984)	−0.125 (−1.140)	−0.053 (−0.137)	0.073 (1.631)
0.97	0.001 (0.005)	−0.021 (−0.119)	0.024 (0.250)	−0.054 (−0.261)	0.042 (0.423)	−0.051 (−0.298)	−0.039 (−0.056)	0.031 (0.356)
0.98	−0.057 (−0.436)	−0.089 (−0.801)	−0.019 (−0.196)	−0.098 (−0.834)	−0.021 (−0.220)	−0.087 (−0.745)	−0.020 (−0.006)	0.097 (0.049)
0.99	0.209 (0.701)	−0.019 (−0.010)	0.063 (0.150)	−0.039 (−0.018)	0.220 (0.855)	0.005 (0.004)	0.043 (0.050)	0.167 (0.427)
Average overall τ	−0.022	−0.013	−0.022	−0.014	−0.025	−0.031	−0.003	0.009

Notes: This table presents the coefficient estimations for the relationship between iron future returns (dependent variable) and equity market returns including MSCI global equity market index, MSCI emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index, and S&P500 (independent variables). Panel A reports average degree of dependence for the pre-COVID period given by β estimates and Panel B reports marginal change induced by COVID-19 shock given by ϴ estimates. Estimation period is from January 03, 2018 to July 28, 2021. Numbers in parentheses are t-statistics. ***, ** and * indicate statistical significance 1%, 5% and 10% level, respectively. Model.

Panel B reports the marginal change during COVID-19. Overall, the degree of dependence in quantile 2 has significantly reduced for equity markets other than BRICs, North America and the US. For instance, degree of dependence for Asia, the most effected index, has reduced from 0.734 during tranquil period to −0.237 during post-COVID period (βAsia(0.02) + Asia(τ = 0.02) = 0.734–0.971 = −0.237). Emerging markets experienced a similar decline, from 0.741 during tranquil period to −0.227 during post-COVID period (βemerging(0.02) + emerging(τ = 0.02) = 0.741–0.968 = −0.237). Accordingly, downside risk of combining iron with these equity indices has significantly reduced during the post-COVID period. Daily returns of iron during the post(pre)-COVID-19 period has averaged to 0.248%(0.034%). For the other points of conditional return distribution of iron, the iron and equity market return dependence has not changed significantly. Accordingly, iron may act as a weak safe-haven against international equity market returns due to detached returns at lower tail and insignificant returns at upper extreme of iron return distribution. Compared with the existing literature, Salisu and Adediran (2019) reported that iron can hedge against inflation in Australia during January 2009 to December 2018 that is conditional on macroeconomic condition. In addition, Adekoya and Oliyide (2020) revealed great hedging potential of iron against lowered oil demand amid to economic activity shocks. Current findings indicate that iron returns have decoupled form equity market returns at the lower quantiles of iron returns and iron also acts as a weak safe-haven asset for international equity investors when iron experiences extreme positive returns during the post-COVID-19 period.

Three scatter plots representing dependence among iron and global equity market index for the pre-COVID (black line), post-COVID (bold grey line) and total effect for the post-COVID (dashed black line) presented in Fig. 4 . As can be seen dependence has slightly reduced at the left-tail during the post-COVID period while the structure of dependence remained similar, an asymmetric u-shaped dependence structure with higher left-tail dependence.

Fig. 4

Estimation Results: Dependence among iron and MSCI global equity index before (pre-10/03/2020) and after the declaration of COVID-19 as a global pandemic (post-11/03/2021).

Dependence among silver and equity market returns

Table 6 reports regression results concerning the degree of dependence among silver and equity market returns. Estimates for β coefficients in Panel A reveal an asymmetric left-tailed u-shaped dependence with higher positive coefficients during extreme bad state of silver returns. At the other points of conditional return distribution of sliver, equity markets-silver dependence is mainly insignificant. Although estimates show high β values at the lowest extreme of silver returns, daily returns of silver have averaged to 1.410%, indicating weak (strong when τ = 90) hedging benefits when silver returns are on and over median. However, returns of silver and North American equity index is positive and significant at both tails, eliminating the ability of sliver to act as a hedge for the equity market index of North America.

Table 6

Estimation results for silver.

Quantile	Global	Emerging	Developed	Asia	Europe	BRIC	North America	US
Panel A: Pre-COVID period (β); 03/01/2018–March 10, 2021
0.01	1.195*** (3.449)	1.075*** (0.474)	1.292*** (3.057)	−0.605 (−0.622)	1.355*** (2.737)	1.205*** (3.194)	1.519** (2.276)	1.122*** (3.378)
0.02	0.259 (0.244)	0.180 (0.474)	0.610 (1.133)	0.177 (0.385)	0.635 (0.657)	0.137 (0.412)	0.705*** (4.016)	0.393 (0.268)
0.03	0.099 (0.218)	0.324 (0.712)	0.197 (0.219)	0.323 (0.627)	0.598* (1.767)	0.260 (0.692)	0.483 (1.299)	0.124 (0.322)
0.04	0.245 (0.456)	0.359 (1.295)	0.442 (1.190)	−0.184 (−0.725)	0.305 (0.472)	−0.061 (−0.098)	0.296 (1.309)	0.189 (0.433)
0.05	0.340 (0.987)	−0.029 (−0.089)	0.492** (1.978)	−0.053 (−0.175)	0.486* (1.913)	−0.022 (−0.061)	0.401 (1.554)	0.359** (2.218)
0.10	0.145 (0.155)	0.173* (1.854)	0.210 (1.288)	0.160 (1.479)	0.210 (1.133)	0.183* (1.766)	0.027 (0.132)	0.163 (1.039)
0.25	0.113* (1.914)	0.135*** (2.769)	0.141* (1.854)	0.118** (2.344)	0.142* (1.940)	0.131*** (2.702)	0.024 (0.456)	0.059 (1.233)
0.50	−0.002 (−0.025)	0.103 (1.531)	0.127* (1.769)	0.095 (1.257)	0.115* (1.702)	0.112* (1.723)	0.036 (0.457)	−0.049 (−0.601)
0.75	−0.255 (−1.316)	−0.191 (−1.131)	−0.260 (−0.948)	−0.230 (−1.283)	−0.277 (−1.345)	−0.213 (−1.183)	−0.170 (−1.206)	−0.307** (−2.016)
0.90	−1.247*** (−4.720)	−1.036** (−2.469)	−1.246*** (−4.620)	−1.013** (−2.346)	−0.937 (−1.218)	−0.835* (−1.784)	−0.180 (−0.148)	−0.477 (−1.319)
0.95	−0.184 (−0.522)	−0.427 (−1.370)	0.029 (0.072)	−0.483 (−1.455)	−0.098 (−0.292)	−0.420* (−1.93)	−0.106 (−0.354)	−0.101 (−0.345)
0.96	−0.056 (−0.179)	−0.192 (−0.525)	0.162 (0.582)	−0.174 (−0.527)	−0.029 (0.912)	−0.169 (−0.517)	−0.118 (−0.362)	−0.060 (−0.220)
0.97	0.170 (1.539)	0.128 (0.692)	0.222 (1.412)	0.155 (0.886)	0.184 (1.071)	0.053 (0.305)	0.152 (0.572)	0.136 (1.349)
0.98	0.075 (0.661)	0.083 (0.479)	0.217 (1.326)	0.112 (0.703)	0.152 (0.944)	0.046 (0.261)	0.315 (1.517)	0.031 (0.269)
0.99	−0.142 (−1.241)	0.192 (1.113)	−0.123 (−0.993)	0.174 (1.212)	0.305 (1.579)	−0.185 (−1.209)	0.450*** (2.585)	−0.112 (−1.401)
Average overall τ	−0.076	−0.065	−0.022	−0.090	−0.032	−0.066	0.015	−0.071
Panel B: Change in COVID period (θ);11/03/2020–July 28, 2021
0.01	−0.609 (−1.507)	−0.570 (−1.278)	−0.642 (−1.282)	1.164 (1.269)	−0.767 (−1.259)	−0.649 (−1.423)	−1.824** (−2.287)	−0.601 (−1.576)
0.02	0.305 (0.293)	0.289 (0.722)	−0.039 (−0.071)	0.333 (0.693)	−0.085 (−0.089)	0.276 (0.788)	−0.511 (1.643)	−0.043 (−0.029)
0.03	0.355 (0.777)	0.056 (0.118)	0.352 (0.390)	0.086 (0.166)	−0.145 (−0.422)	0.155 (0.392)	−0.061 (−0.155)	0.318 (0.819)
0.04	0.301 (0.553)	−0.032 (−0.112)	0.189 (0.468)	0.616** (2.400)	0.234 (0.358)	0.456 (0.726)	0.031 (0.127)	0.345 (0.781)
0.05	0.252 (0.702)	0.406 (1.256)	0.106 (0.376)	0.532* (1.745)	0.160 (0.468)	0.484 (1.317)	−0.119 (−0.446)	0.023 (0.094)
0.10	0.339 (1.149)	0.315*** (2.749)	0.290 (1.089)	0.377* (1.902)	0.314 (1.130)	0.246 (1.525)	0.156 (0.686)	0.165 (0.954)
0.25	0.383* (1.741)	0.341* (1.921)	0.504*** (2.836)	0.331*** (3.685)	0.426** (2.371)	0.271*** (2.703)	0.344* (1.718)	0.360*** (3.686)
0.50	0.481*** (3.728)	0.372*** (3.577)	0.423** (1.965)	0.411*** (3.487)	0.369** (2.055)	0.285** (2.290)	0.346** (2.370)	0.544*** (4.163)
0.75	0.840*** (4.156)	0.789 (4.231)	0.892*** (3.105)	0.813*** (3.781)	0.871*** (3.918)	0.730*** (3.428)	0.598*** (2.878)	0.889*** (5.542)
0.90	1.652 (1.000)	1.231** (2.458)	1.875*** (2.728)	1.212** (2.449)	1.647** (2.005)	1.061* (1.746)	0.464 (0.376)	0.603* (1.743)
0.95	0.055 (0.143)	0.313 (0.859)	−0.170 (−0.377)	0.377 (0.965)	−0.020 (−0.053)	0.269 (0.816)	−0.022 (−0.065)	−0.015 (−0.047)
0.96	−0.103 (−0.298)	0.119 (0.237)	−0.206 (−0.116)	0.113 (0.125)	−0.122 (−0.395)	0.007 (0.015)	−0.090 (−0.253)	−0.087 (−0.290)
0.97	−0.384*** (−2.640)	−0.384 (−1.538)	−0.447** (−2.373)	−0.426* (−1.731)	−0.394** (−1.963)	−0.354 (−1.182)	−0.439 (−1.360)	−0.330** (−2.447)
0.98	−0.337** (−2.457)	−0.391* (−1.749)	−0.489*** (−2.613)	−0.441** (−1.970)	−0.408** (−2.239)	−0.438 (−1.623)	−0.659** (−2.532)	−0.282** (−2.099)
0.99	−0.218* (−1.830)	−0.596*** (−2.890)	−0.254* (−1.911)	−0.616*** (−3.260)	−0.637*** (−3.115)	−0.315 (−1.351)	−0.923*** (−4.246)	−0.228** (−2.390)
Average overall τ	0.523	0.483	0.518	0.518	0.509	0.442	0.312	0.463

Notes: This table presents the coefficient estimations for the relationship between silver returns (dependent variable) and equity market returns including MSCI global equity market index, MSCI emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index, and S&P500 (independent variables). Panel A reports average degree of dependence for the pre-COVID period given by β estimates and Panel B reports marginal change induced by COVID-19 shock given by ϴ estimates. Estimation period is from January 03, 2018 to July 28, 2021. Numbers in parentheses are t-statistics. ***, ** and * indicate statistical significance 1%, 5% and 10% level, respectively. Model.

Panel B reports marginal change in the degree of dependence. Reported coefficients indicate significant increase in degree of dependence on and around median and decrease at the upper-tail of the conditional return distribution of silver. Except for BRIC equity index, COVID-19 period has led to the decoupling of silver and equity market returns at the right-tail of conditional distribution of silver returns. Accordingly, silver has become a safe-haven asset against these equity indices during good state of silver returns. For example, total post-COVID degree of dependence among European markets equity index and silver at the extreme right-tail is βEurope(0.99) + Europe(τ = 0.99) = 0.305 + 0.637 = −332. Same patterns are observable also for other equity market indices except for BRICs due to insignificant estimates of BRIC at upper quantiles of silver returns. Accordingly, marginal changes in the structure of dependences among silver and equity market returns indicate that during the post-COVID-19 period equity markets tend to decrease when silver experience extreme positive returns. Given that mean daily returns of silver is the highest among all assets investigated during the post-COVID period (0.287%), it can be regarded as a safe-haven asset against equity markets.

Fig. 5 is a scatter plot for dependence among silver and global markets equity returns. Dashed line plots the total effect and shows that the degree of dependence among silver and global index returns for the quantiles between τ = 1 to τ = 90 is around 0.50. At the upper extreme of conditional silver returns, the dependence becomes negative. Accordingly, silver may act as a diversifier for global equity portfolio when silver returns are at bad state or around median. But silver acts as a safe-haven when silver experiences extreme positive returns. According to the last raw of Panel B, average degree of dependences among silver and equity market returns have increased considerably. Accordingly, results also suggest that examination of average change would be misleading without examining tail dependence.

Fig. 5

Estimation Results: Dependence among silver and MSCI global equity index before (pre-10/03/2020) and after the declaration of COVID-19 as a global pandemic (post-11/03/2021).

Dependence among global energy producer index returns and equity market returns

This section examines the dynamic dependence structure among MSCI global energy producer index and international equity markets to test if energy stocks pose any safe-haven properties during post-COVID-19 period. Panel A of Table 7 reports coefficient estimates examining the degree of dependence at different points of conditional return distribution of global energy producer index during tranquil period. Results indicate high degree of left-tailed dependence (positive) with coefficient estimates of above 1.00 level from τ = 0.01 to τ = 0.25. The coefficient estimates of β fall steadily form τ = 0.25 to τ = 98 but still maintain the positive sign and significance. Accordingly, global energy producer index offers very limited diversification benefits only when its returns are at a good state. However, due to large downside risk given by high β estimates on and below the median and negative average returns of energy index as reported in Table 1 Panel A, its inclusion into equity portfolios is highly risky. Different from other markets, energy index offers hedging benefits to North American equity index investors only when conditional energy returns are at extreme positive state, when τ = 0.99.

Table 7

Estimation results for global energy producers.

Quantile	Global	Emerging	Developed	Asia	Europe	BRIC	North America	US
Panel A: Pre-COVID period (θ); 03/01/2018–March 10, 2021
0.01	1.849*** (10.001)	1.187*** (5.405)	1.961*** (92.707)	1.101 (1.119)	1.876*** (75.371)	1.337*** (4.398)	1.810*** (2.597)	1.466*** (9.753)
0.02	1.637*** (12.507)	1.119*** (4.290)	1.848*** (13.609)	0.986** (2.134)	1.660*** (6.356)	0.934** (2.440)	0.996** (2.052)	1.269*** (11.511)
0.03	1.534*** (9.794)	0.965*** (6.330)	1.623*** (11.638)	0.884** (2.388)	1.516*** (9.808)	0.897*** (2.867)	0.555*** (3.552)	1.202*** (11.276)
0.04	1.314*** (8.395)	0.902*** (7.790)	1.639*** (12.916)	0.983*** (3.556)	1.493*** (1.493)	0.904*** (2.640)	0.517*** (3.487)	1.123*** (11.637)
0.05	1.410*** (12.124)	0.858*** (7.533)	1.566*** (12.808)	0.879*** (2.972)	1.403*** (7.232)	0.952*** (4.620)	0.662*** (3.670)	1.117*** (6.829)
0.10	1.288*** (13.201)	0.972*** (7.451)	1.306*** (11.192)	0.895*** (7.210)	1.100*** (11.934)	0.968*** (8.853)	0.706*** (5.206)	1.189*** (14.924)
0.25	1.097*** (20.172)	0.749*** (11.910)	1.152*** (16.139)	0.762*** (11.920)	1.029*** (14.561)	0.677*** (13.448)	0.741*** (14.713)	0.982*** (18.969)
0.50	0.956 (18.968)	0.644*** (12.532)	1.070*** (17.992)	0.606*** (12.373)	0.931***15.501	0.591*** (11.953)	0.643*** (8.462)	0.879*** (19.99)
0.75	0.909*** (13.782)	0.624*** (11.017)	1.139*** (12.178)	0.576*** (10.089)	1.025*** (13.267)	0.581*** (9.291)	0.629*** (8.238)	0.840 (15.083)
0.90	0.937*** (3.705)	0.618*** (7.657)	1.054*** (15.451)	0.596*** (6.026)	0.901*** (9.617)	0.571*** (7.544)	0.332*** (3.038)	0.809*** (4.162)
0.95	0.838 (1.105)	0.642** (1.995)	1.103*** (5.935)	0.729* (1.762)	0.834*** (4.659)	0.476*** (4.045)	0.082 (0.395)	0.716* (1.903)
0.96	0.692 (1.149)	0.834*** (4.259)	0.969 (0.983)	0.819*** (2.700)	0.906*** (5.562)	0.624* (1.775)	0.211 (0.303)	0.481 (0.507)
0.97	0.479*** (3.530)	0.580 (1.149)	0.807*** (2.804)	0.628 (1.589)	0.830*** (2.983)	0.453 (1.302)	−0.012 (−0.007)	0.318*** (4.148)
0.98	0.421*** (3.756)	0.603 (1.284)	0.565* (1.882)	0.370 (0.506)	0.608 (1.189)	0.343 (0.432)	−0.220 (−0.825)	0.249*** (3.204)
0.99	0.123 (1.089)	0.303 (0.260)	0.208 (0.962)	0.454 (0.203)	0.342 (1.531)	0.642 (0.151)	−0.820*** (−2.721)	0.065 (0.599)
Average overall τ	1.021	0.694	1.129	0.677	1.022	0.655	0.611	0.888
Panel B: Change in COVID period (δ);11/03/2020–July 28, 2021
0.01	−0.963*** (−4.604)	−0.211 (−0.920)	−0.873*** (−8.037)	−0.049 (−0.065)	−0.747*** (−2.82)	−0.349 (−1.255)	−0.536 (−0.932)	−0.740*** (−4.285)
0.02	−0.613*** (−4.010)	−0.118 (−0.445)	−0.621*** (−4.306)	−0.057 (−0.123)	−0.453* (−1.923)	−0.036 (−0.094)	−0.108 (−0.228)	−0.331*** (−2.681)
0.03	−0.416** (−2.111)	−0.022 (−0.135)	−0.461*** (−3.079)	0.072 (0.193)	−0.410** (−2.524)	−0.013 (−0.041)	0.337** (2.019)	−0.305** (−2.520)
0.04	−0.223 (−1.025)	0.103 (0.724)	−0.457*** (−2.931)	−0.018 (−0.064)	−0.383** (−2.245)	−0.006 (−0.016)	0.386** (2.408)	−0.274*** (−2.672)
0.05	−0.242 (−1.637)	0.115 (0.815)	−0.330** (−2.214)	0.121 (0.397)	−0.274 (−1.329)	−0.037 (−0.180)	0.296 (1.566)	−0.210 (−1.146)
0.10	−0.132 (−1.200)	−0.018 (−0.132)	−0.083 (−0.496)	0.048 (0.375)	0.055 (0.395)	−0.096 (−0.756)	0.199 (1.079)	−0.226*** (−2.586)
0.25	0.033 (0.333)	0.177** (2.279)	0.076 (0.732)	0.165* (1.872)	0.078 (0.707)	0.168 (1.113)	−0.035 (−0.176)	−0.011 (−0.176)
0.50	0.166** (2.214)	0.285*** (3.094)	0.236* (1.785)	0.260 (1.276)	0.121 (1.224)	0.204 (1.325)	0.022 (0.146)	0.143** (2.057)
0.75	0.269*** (3.326)	0.321*** (4.232)	0.119 (1.132)	0.320** (2.122)	0.083 (0.957)	0.174 (1.159)	−0.004 (−0.031)	0.199** (1.989)
0.90	0.312 (1.222)	0.407*** (3.806)	0.291*** (3.649)	0.311*** (2.602)	0.315** (2.174)	0.247*** (2.638)	0.417** (2.545)	0.245 (1.428)
0.95	0.431 (0.582)	0.461 (1.440)	0.260 (1.382)	0.188 (0.442)	0.386** (2.151)	0.332* (1.819)	0.984*** (3.092)	0.459 (1.219)
0.96	0.558 (0.948)	0.290 (1.520)	0.369 (0.377)	0.194 (0.563)	0.303* (1.861)	0.138 (0.360)	0.926 (1.311)	0.665 (0.708)
0.97	0.739*** (5.948)	0.579 (1.146)	0.512* (1.808)	0.521 (1.273)	0.383 (1.367)	0.319 (0.801)	0.997 (0.587)	0.801*** (10.279)
0.98	0.773*** (7.366)	0.497 (1.056)	0.699** (2.508)	0.701 (1.171)	0.525 (1.075)	0.389 (0.444)	1.230*** (4.388)	0.859*** (11.816)
0.99	0.944*** (10.032)	0.774 (0.659)	0.933*** (7.429)	0.581 (0.264)	0.723*** (3.285)	0.122 (0.029)	2.082*** (6.839)	0.918*** (13.435)
Average overall τ	0.129	0.263	0.137	0.231	0.111	0.150	0.155	0.126

Notes: This table presents the coefficient estimations for the relationship between MSCI global energy producer index returns (dependent variable) and equity market returns including MSCI global equity market index, MSCI emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index, and S&P500 (independent variables). Panel A reports average degree of dependence for the pre-COVID period given by β estimates and Panel B reports marginal change induced by COVID-19 shock given by ϴ estimates. Estimation period is from January 03, 2018 to July 28, 2021. Numbers in parentheses are t-statistics. ***, ** and * indicate statistical significance 1%, 5% and 10% level, respectively. Model.

Marginal change during post-COVID-19 period is given by estimates in Panel B. Results indicate that dependence in the lower(upper)-tail has significantly reduced(increased). Structure of dependence changed from s-shaped asymmetric dependence to a flatter dependence structure. Given the total effect during the post-COVID period, energy equity index cannot provide any diversification, hedging or safe-haven benefits to international equity index investors.

Dependence structures among energy and global equity market returns for the pre-COVID and post-COVID periods are given by a black and bold grey scatter plot in Fig. 6 , respectively. Dashed line indicates the total effect during the post-COVID period. Overall, during the post-COVID-19 period, asymmetric dependence structure with higher positive dependence at left-tail and lower positive dependence at right-tail converted into a flatter dependence structure with average dependence of above 1.00.

Fig. 6

Estimation Results: Dependence among MSCI energy producer index and MSCI global equity index before (pre-10/03/2020) and after the declaration of COVID-19 as a global pandemic (post-11/03/2021).

Dependence among global clean energy producer index returns and equity market returns

This section examines the dependence structure among clean energy and equity index returns. The coefficient estimates in Panel A of Table 8 reveal a right-tailed u-shaped asymmetric dependence structure with higher positive coefficients at tails relative to intermediate quantiles. Coefficient estimates at the lower tail of the distribution of clean energy returns is higher than 1.00 showing that combining international equity market portfolios with global clean energy producer portfolio cannot provide any diversification benefit before the COVID-19 pandemic but increases the downside risk.

Table 8

Estimation results for global clean energy equity index.

Quantile	Global	Emerging	Developed	Asia	Europe	BRIC	North America	US
Panel A: Pre-COVID period (β); 03/01/2018–March 10, 2021
0.01	1.044*** (20.766)	1.198*** (31.499)	0.970*** (25.779)	1.635*** (19.377)	0.941*** (28.580)	1.272*** (24.939)	1.346*** (9.725)	1.100*** (12.361)
0.02	1.148*** (40.192)	1.193*** (4.140)	1.055*** (33.488)	1.357*** (8.237)	0.978*** (26.293)	1.356*** (20.142)	1.164*** (3.966)	1.108*** (33.264)
0.03	1.197*** (30.677)	1.027*** (6.024)	1.071*** (30.463)	1.289*** (7.719)	1.036*** (30.662)	1.081*** (6.986)	0.790*** (3.205)	1.018*** (7.755)
0.04	1.217*** (30.862)	0.895*** (7.727)	1.104*** (34.023)	1.078*** (4.772)	1.054*** (29.122)	1.035*** (6.772)	0.635*** (3.899)	0.953*** (6.622)
0.05	1.134*** (3.596)	0.851*** (9.611)	1.128*** (40.074)	1.036*** (5.835)	1.083*** (30.826)	0.929*** (5.562)	0.593*** (3.524)	0.886*** (6.172)
0.10	0.844*** (11.151)	0.677*** (12.847)	0.961*** (11.439)	0.702*** (7.094)	0.994*** (8.492)	0.680*** (8.677)	0.604*** (6.144)	0.751*** (10.062)
0.25	0.839*** (16.347)	0.610*** (16.361)	0.955*** (10.960)	0.652*** (14.878)	0.819*** (11.216)	0.600*** (16.922)	0.494*** (8.807)	0.751*** (16.964)
0.50	0.845*** (14.924)	0.671*** (18.906)	0.964*** (19.225)	0.664*** (17.466)	0.857*** (17.777)	0.623*** (16.971)	0.425*** (5.999)	0.801*** (17.783)
0.75	0.859 (14.674)	0.688*** (13.549)	0.936*** (14.997)	0.689*** (15.389)	0.844*** (13.368)	0.604*** (13.915)	0.361*** (4.613)	0.819*** (10.557)
0.90	0.812*** (6.294)	0.716*** (8.446)	0.960*** (6.877)	0.622*** (8.946)	0.810*** (6.407)	0.631*** (8.868)	0.246 (1.480)	0.625*** (7.359)
0.95	0.756*** (3.748)	0.797*** (3.954)	0.908*** (3.629)	0.728*** (3.782)	0.835*** (5.291)	0.808*** (4.739)	0.170 (0.322)	0.665 (0.963)
0.96	0.804*** (4.046)	0.848*** (4.694)	0.944*** (2.735)	0.861*** (5.654)	0.740*** (2.793)	0.883*** (4.290)	0.202 (0.458)	0.522*** (2.696)
0.97	0.888** (2.528)	0.997*** (4.018)	1.064 (1.581)	0.940*** (4.844)	1.020*** (5.046)	0.885** (2.446)	0.219 (0.415)	0.597** (2.289)
0.98	1.094*** (2.620)	1.085*** (4.238)	1.233*** (2.873)	1.052*** (4.707)	1.195*** (4.104)	1.099*** (4.523)	0.107 (0.165)	0.706** (2.214)
0.99	1.618*** (7.920)	1.439*** (4.329)	1.751*** (17.800)	1.264*** (4.212)	1.471*** (5.909)	1.352*** (10.752)	0.680 (0.772)	1.217*** (3.132)
Average overall τ	0.874	0.703	0.967	0.705	0.883	0.681	0.442	0.785
Panel B: Change in COVID period (θ);11/03/2020–July 28, 2021
0.01	0.306*** (4.356)	−0.077 (−0.690)	0.341*** (6.897)	−0.501*** (−3.260)	0.343*** (4.585)	−0.183*** (−2.614)	−0.139 (−0.798)	0.031 (0.168)
0.02	0.048 (0.645)	−0.045 (−0.172)	0.368*** (6.462)	−0.147 (−0.861)	0.271*** (4.553)	−0.123*** (−2.923)	0.153 (0.452)	−0.044 (−1.093)
0.03	0.000 (0.004)	0.042 (0.248)	0.282** (2.402)	−0.106 (−0.607)	0.238*** (4.195)	0.016 (0.107)	0.531*** (2.851)	0.043 (0.305)
0.04	−0.039 (−0.473)	0.156 (1.391)	0.186** (2.101)	0.142 (0.634)	0.137* (1.825)	0.088 (0.585)	0.523*** (3.006)	0.187 (1.135)
0.05	0.037 (0.118)	0.225** (0.374)	0.213** (2.046)	0.167 (1.005)	0.068 (0.957)	0.155 (1.078)	0.557*** (3.006)	0.281* (1.793)
0.10	0.329*** (3.897)	0.414*** (5.815)	0.106 (1.184)	0.411*** (3.582)	0.024 (0.201)	0.367*** (2.107)	0.415*** (3.484)	0.333** (2.228)
0.25	0.302*** (3.781)	0.460*** (5.607)	0.156* (1.712)	0.442*** (4.934)	0.208* (1.956)	0.442*** (9.408)	0.261*** (2.392)	0.297*** (4.064)
0.50	0.339*** (3.156)	0.410*** (4.656)	0.198*** (3.045)	0.385*** (6.555)	0.213* (1.681)	0.338*** (3.601)	0.492** (2.481)	0.342*** (2.838)
0.75	0.343*** (3.716)	0.344*** (3.985)	0.243*** (3.355)	0.392*** (5.820)	0.234*** (3.291)	0.358*** (3.364)	0.649*** (4.715)	0.308*** (3.018)
0.90	0.392 (2.982)	0.329*** (3.718)	0.210 (1.443)	0.473*** (5.487)	0.319** (2.471)	0.446*** (4.901)	0.830*** (0.182)	0.414*** (5.359)
0.95	0.440** (2.001)	0.299 (1.560)	0.353 (1.346)	0.415** (2.063)	0.343** (2.148)	0.233 (1.100)	0.974* (1.740)	0.487 (0.702)
0.96	0.422* (1.947)	0.268 (1.540)	0.346 (1.102)	0.374** (2.514)	0.451* (1.665)	0.153 (0.756)	0.991** (2.067)	0.596*** (2.656)
0.97	0.376 (1.082)	0.130 (0.279)	0.222 (0.332)	0.273 (1.330)	0.144 (0.461)	0.157 (0.455)	1.078** (2.031)	0.523* (1.857)
0.98	0.110 (0.259)	0.069 (0.279)	0.111 (0.269)	0.166 (0.705)	0.056 (0.224)	0.116 (0.553)	1.091* (1.672)	0.487 (1.557)
0.99	−0.208 (−1.318)	−0.189 (−0.617)	−0.304*** (−4.247)	0.045 (0.173)	−0.147 (−0.624)	−0.065 (−0.508)	0.637 (0.664)	0.104 (0.284)
Average overall τ	0.323	0.373	0.199	0.391	0.190	0.345	0.512	0.329

Notes: This table presents the coefficient estimations for the relationship between MSCI global clean energy producer index returns (dependent variable) and equity market returns including MSCI global equity market index, MSCI clean emerging markets equity index, MSCI developed markets equity index, MSCI Asian markets equity index, MSCI BRIC markets equity index, MSCI European markets equity index, MSCI North American markets equity index, and S&P500 (independent variables). Panel A reports average degree of dependence for the pre-COVID period given by β estimates and Panel B reports marginal change induced by COVID-19 shock given by ϴ estimates. Estimation period is from January 03, 2018 to July 28, 2021. Numbers in parentheses are t-statistics. ***, ** and * indicate statistical significance 1%, 5% and 10% level, respectively. Model.

Panel B reports coefficient estimates for ϴ that measures the marginal change in the degree of dependence at different quantiles. The degree of dependences for most of the market portfolios with clean energy returns have increased at lower quantiles and on and around the median. For example, degree of dependence among the returns of clean energy producer index and global equity market index has increased from 1.044 during tranquil times to 1.350 during the post-COVID period at τ = 0.01 (βglobal(0.01) + global(τ = 0.01) = 1.044 + 0.306 = 1.350). On median, degree of dependence has increased from 0.845 to 1.184 (βglobal(0.50) + global(τ = 0.50) = 0.845 + 0.339 = 1.184. Other equity market indices experienced similar affect. Overall, clean energy stocks cannot provide any diversification or hedging benefit to international equity investors during tranquil times or any safe-haven properties during post-COVID period.

Fig. 7 plots degree and structure of dependence among clean energy and equity returns and dynamic change during the post-COVID-19 period on different points of conditional returns distribution of clean energy index. It is clearly observable that right-tailed asymmetric u-shaped dependence structure converted into a flatter dependence structure.

Fig. 7

Estimation Results: Dependence among MSCI clean energy producer index and MSCI global equity index before (pre-10/03/2020) and after the declaration of COVID-19 as a global pandemic (post-11/03/2021).

Conclusion

This paper analyzes the role of copper, gold, iron, silver and energy stocks, both conventional and green, in the global financial markets during the pre-COVID-19 and post-COVID-19 periods. Using the QR based approach of Baur (2013) dependence structure among commodities and international equity market indices have decomposed into two components, (i) the degree of return dependence and (ii) the structure of return dependence. This method enables the examination of marginal change in the degree and structure of return dependence among two random variables during different regimes. Accordingly, the degree and the structure of dependence among commodities and index returns estimated for the pre-COVID period while accounting for the marginal changes during the post-COVID period. Due to its focus on the dynamic change in the degree of dependence among financial asset classes during different market states, current paper is related to the prospect theory of Koenker and Bassett (1978), such as: investors are loss averse (Duxbury and Summers, 2004) and dramatically switch among financial assets in face of financial shocks (Ang et al., 2005). Current results provide valuable information related to the trading behavior under extreme market conditions with a specific focus to commodities, energy stocks and international equity markets during the post-COVID period.

The empirical results show that both the degree and the structure of dependence among returns of commodities and equity markets changed significantly during the post-COVID-19 period. For instance, during the pre-COVID period copper was a weak hedge for North America and the US equity markets when copper returns are at positive state. During the post-COVID period, degree of dependence among copper and global and developed equity indices reduced and become negligible at the upper-tail of copper returns, showing that copper became a weak safe-haven during the post-COVID period. While the structure of dependence among gold and equity returns is considerably different. Before the pandemic gold was a weak hedge for global, emerging, developed, Asia, Europe, BRIC and North American equity markets and a strong hedge for the US specifically during good states of gold returns. But, during the post-COVID-19 period dependence among gold and equity markets have significantly increased, indicating that gold cannot act as a safe-haven for international equity markets for the post-COVID-19 period. While, silver was a weak hedge during the pre-COVID period and became a strong safe-haven during the post-COVID period at the upper-tail of conditional silver distribution. Given high daily returns reported for silver during both pre-COVID and post-COVID periods, it offers great potential to improve risk-return characteristics of equity portfolios. Dependence of iron returns with international equity indices unchanged, it was a weak hedge for international markets during the pre-COVID period and a weak safe-haven during the post-COVID period when iron returns are extremely positive. Finally, related to energy stocks, neither conventional energy producer equity index nor green energy producer equity index act as hedge or safe-haven against international equity markets. The degree of average dependence among energy indices and equity markets increased significantly during the post-COVID period eliminating any benefits of including energy stocks into broad equity market portfolios.

Several policy implications can be drawn from the empirical results. First, international equity market investors should consider silver, and to some extent copper and iron instead of gold and energy stocks to optimize their portfolio risk during the post-COVID period. Specifically, silver found to act as a strong safe-haven asset against international equity market portfolios when gold does not. These findings would be useful for policy makers in attaining financial stability and portfolio managers and investors in optimizing risky portfolios during the post-COVID period. Second, researchers should pay close attention towards structural changes in tail dependence among financial assets in face of significant shocks to economy. It has been showed that, dependence structure among financial commodities and equity market indices significantly changed at tails while relatively remained constant on and around median. Therefore, models that neglect tail dependences would offer misleading guidance.

Author statement

Asil Azimli: Conceptualization, Methodology, Formal Analysis, Writing and Visualization.