Covid-19 and oil and gold price volatilities: Evidence from China market.

Gold and crude oil are the influential commodities of the stock markets and real economy of the world in financial crises as well as in COVID-19 periods. However literature mainly focused on the effects of these commodities' prices only, and the volatilities in the prices of these commodities altogether with the prices got little attention. To fill in a major research gap, our study intends to estimate the dynamic relationship between oil prices, gold prices, oil prices volatilities and gold prices volatilities on the stock market of China. Using daily data over the period from 2009 to 2021, the study applied Autoregressive Distributed Lag (ARDL) bound test approach for the purpose of empirical estimation. Moreover, Non linear ARDL and asymmetric Causality analysis has also been applied for more comprehensive asymmetric estimation. The findings of our study indicated that gold prices and oil prices negatively affect stock market of China in the long run. In terms of implied volatility index of these commodities, study finds negative impact of price volatility of oil but positive impact of the price volatility of gold on the country's stock market in the long run. However, in the short run, only oil price and gold prices have significant effect on the China's stock market. On the basis of our findings, we recommend the investors to make rational decisions in response to the uncertainties in these markets and should consider gold as a safe haven to hedge themselves in times of uncertainty. Policymakers should take appropriate actions and adopt proper mechanisms for dealing with the quick uncertainty flow of information from the oil to the stock market.

Introduction

High uncertainty decreases investments and enhances investor concerns making resource allocation and portfolio crisis management a big challenge for hedgers, traders, portfolio managers and equity investments Bernanke (1983). In the period of peak uncertainty, stock market integration and interdependence increase, that reduce the opportunities for diversification and force the investors to search other investment opportunities to control investment risks. Gold and oil are the important commodities which are necessary for the worldwide economy and frequently added in the investors’ equities portfolios Singleton (2014) Büyükşahin and Robe (2014). Oil is a very volatile product, as well as its price fluctuations provide useful information for forecasting commodities and prices of financial asset Regnier (2007). Gold, on the other side, is often seen as a safe-haven asset during times of crisis Baur and Lucey (2010). To diversify existing equities portfolios, investors frequently switch between gold and oil or simply combine these two Soytas et al. (2009). However, in recent time, the notion that gold and oil have been defined by increased volatility and turbulence, is a source of concern for investment decision making(e.g., due to rise in oil price in summer 2008, 2014 to 2016 crash in oil price and 2006 food crisis). As a result, precise understanding of the way information is spilled over and conveyed between the oil, gold and stock markets should help investors to make risk management and better portfolio decisions(Aliedan, 2021; Alosaimi, 2021; Alshboul et al., 2021; Cifuentes-Faura, 2021; Lipińska, 2021; Mensi et al., 2021; Pinmongkhonkul et al., 2021).

In 2019, the fast spread of the novel corona virus illness (COVID-19) has heightened commodity and financial market uncertainty Mhalla (2020), Zhang et al. (2020). Governments have been forced to apply strict restrictions (e.g., countries or towns' lockdown, prevent migration, enforce social isolation, and cease corporate operations) as the confirmed cases and number of deaths have increased resulting in economic consequences. For example, China's economy shrank by 6.8 percent in the start of 2020, the largest contraction since the beginning of GDP quarterly records. As the COVID-19 cases expanded dramatically, stock markets began to fall, with commodity and equity investors increasingly becoming risk-averse as a result of the rapid rise in deaths and cases (2020); Corbet et al. (2020), Harjoto et al. (2021) He et al. (2020). Salisu and Vo (2020)discovered that news related to health improves the anticipated stock returns accuracy, confirming the link between the public health and stock market situation. Azimli (2020)has determined that COVID-19 spread influenced the risk-return nexus dependence structure. The world's current health crisis, when combined with the recent damaging impacts of re-occurring financial crises, political conflicts, trade wars between China and the United States has exacerbated economic uncertainty and spillovers Sharif et al. (2020), and made management of portfolios more difficult than before. The COVID-19 outbreak, according to Bakas and Triantafyllou (2020), had a substantial influence on the financial sectors and real economy contributing to increased uncertainty. In reaction to a dramatic decline in global demand, the crude oil value (US reference) (WTI) plummeted 37 $ per barrel in April 2020 when economic activity slowed significantly during lockdowns in practically all developed countries. For the very first quarter of 2020, the US International Energy Agency predicts a 30 percent increase. While price volatility of oil has a negative effect on global industrial manufacturing, gold prices have demonstrated a strong upward tendency during the pandemic. Gold, Oil, stock markets theoretically move in lockstep together, with commodity information influencing stock performance. Volatility in Crude oil prices has an impact on discount rates and corporate cash flows. The oil prices increase raises the costs of production and reduces flows of cash, lowering share values. Different factors, e.g. corporate future cash flows, foreign exchange markets, economic cycles, discount factors, might theoretically explain the potential links between stock markets and crude oil. The gold demand rises as a source of hedging against falling equities prices, and its price rises as a consequence Khalfaoui et al. (2019); Sarwar et al. (2019); Waheed et al. (2018) Mensi et al. (2021). Despite the fact that gold is important for currency trading and hedging, gold price volatility could have a severe effect on financial markets because a rise in price volatility of gold is associated with a risky investment situation, while a decrease in gold price volatility is associated with a safe investment situation Baur (2012). As a result, understanding that gold price volatility is critical for hedging decisions, derivative valuation, financial markets, and for the economy as a whole is indeed necessary Brüntrup (2021); (Fok et al.,; Gakpo et al., 2021); Gokmenoglu and Fazlollahi (2015); (Haaskjold et al., 2021; Zhao, 2021).

We aim to estimate the dynamic effects of gold prices and oil prices as well as the volatilities in these prices on the Chinese stock market in the present study. There are a number of reasons to consider the China market. First, China has surpassed the USA as the largest oil consumer and importer in the world as cited in Tian et al. (2021). Meanwhile, as the China has become the second largest economy of the world, in the 18th People Congress held in 2012 November, in order to attract a group of international investors, the government of China increased the credit line available to certified investors internationally (QFII) Tian et al. (2021). International investors have used the Shenzhen and Shanghai indexes as benchmarks. The United States and China are the top oil consumers of the world, China is a key partner of trade of the OPEC (Organization of Petroleum Exporting) economies Lin et al. (2021). The Shanghai index, in 2019, is the second largest stock exchange in the world in the Asia Pacific area, with a market capitalization of $6.19 trillion (after the Japan's Nikkei 225 index). The Shenzhen index, on the other hand, is rated fourth (following the stock exchange market of Hong Kong). Furthermore, because they are available to global investors, the stock markets of China are particularly exposed to external shocks. Oil is the world's traded commodity, with fluctuations in prices affecting not only other valuable goods, but also the prices of stock markets. Oil price variations are critical for the economic growth. Variations in the stock prices could be observed by looking at patterns in gold and oil prices, as commodities and stock market spillovers can happen at any time, but especially during times of crisis(Mensi et al., 2021).

As stated that China is currently a major player in the international oil industry (asit is the second-biggest oil consumer and the world's greatest net importer of oil), and with its rapid economic growth and growing importance in the world economy, the stock market of China is flourishing, attracting a significant group of investors from all over the world in the same time. As a result, in recent years, the relationship between stock market of China and oil prices has gotten a lot of attention Bouri et al., 2017a, Bouri et al., 2017b, Broadstock et al. (2012), Chen and Lv (2015), Luo and Qin (2017) Zhu et al. (2016), Caporale et al. (2015). The study on the Chinese oil-stock nexus has also been extended from a nonlinear approach in a few studies that focus on the asymmetric effects of the prices of oil on the stock market, ignoring the nonlinear associations between these two variables in the short and long run Wen et al. (2019). Meanwhile, according to Lai and Tseng (2010), the stock market of China is a safe haven for foreign investors, and also a hedge for their portfolios, particularly in times of monetary turbulence. In addition, we have noticed a steady increase in gold demand in China in recent time. Soaring global market prices, high property and stock market uncertainty and higher expectations of inflation, are all the determinants that have contributed to considerable growth in the investment of domestic gold market of China. Gold, just like any other asset, can be viewed as the best inflation hedger and viable alternative investment. It is worth mentioning that COMEX's three month gold futures prices are the worldwide gold trading benchmark. Such prices will exert a big effect on the equity markets of China, as gold has become an important part of asset allocation, and China has the world's fastest growth in commodities including gold. There will be a significant empirical and clear correlation between the variable, as we expect that China's strong marginal gold demand is related to the global benchmark at gold prices (Arouri et al., 2015).

There are two main contributions of our study to the literature; first, our study is the first research in context of China that examines dynamic linkage of gold prices, oil prices as well as their volatilities with its stock market. As far as the authors know, none of the studies have investigated these linkages together in a model taking China as a case study. Moreover, the impact of gold prices did not get much attention particularly in China market. Therefore, our study accompanies those few studies that tried to investigate this relationship, and goes a step further by estimating this relationship by taking gold prices as well as their volatilities into consideration. Secondly, we applied ARDL bound testing approach, which is far better that Johansen Co integration approach because of its numerous advantages to estimate the long term associations among oil prices, gold prices, volatilities in gold prices and oil prices and stock markets Le and Chang (2016). ARDL bound test can easily applied, no matter whether the studied variables are integrated of order 0, order 1, or is a combination of both, unlike the other methods of assessing co integration. However, the variables should not be integrated of order 2; else, ARDL bound approach would not work. We can evaluate the cointegration through bound test process of the ordinary least squares, once the model's lag order has been determined. The lag values of variables are being used in the ARDL technique, which assists to avoid the endogeniety problem. To estimate this cointegration relationship with tiny samples, model works well and it avoids difficulties of weak estimation power(Pesaran, 1997).

The remainder of the study is laid out as follows: Review of the relevant literature is included in Section 2. The estimation techniques that were employed are presented in Section 3. Section 4 summarizes the findings, while Section 5 concludes and discusses policy implications.

Literature review

The present literature is full of the mixed evidences about the role of gold and oil commodities in the stock market performance Boldanov et al. (2016); Büyüksahin et al. (2009); Gorton and Rouwenhorst (2006); Hammoudeh et al. (2014)because commodities give diversification advantages to the investors when added to their portfolios. Salant and Henderson (1978)evaluated the impact of government sales policy expectations on the real gold prices. It was concluded that price dropped sharply after announcements that a government auction was more likely. Attempts by the government to fix the price or maintain a price ceiling by selling its stockpile would eventually result in a speculative attack. Hatemi-J and Irandoust (2015) considered GCC countries to estimate the impact of oil price shocks and growth in gold prices to stock returns and concluded no significant impact of oil and gold prices on stock returns. Starr and Tran (2008) scrutinized the panel data for 21 countries to analyze the important factors that affect physical and portfolio demand for gold. According to study findings, there was significant difference between physical and portfolio demands. Moreover, physical and portfolio demands also differed among developing and developed countries. Similarly, Solt and Swanson (1981)analyzed the effect of changes in gold and silver prices on investor's perceptions and concluded that there was significant dependence between these prices that investors failed to distinguish. Similarly Rubio (1989) also strongly rejected the safe haven properties of government bonds and gold while analyzing the Spanish capital market data.

Among recent studies, Mensi et al. (2021) investigated the price switching spill over impacts between the Chinese and US stock, gold markets and oil markets both in pre and during COVID 19 era. In their study, Markov-Switching Vector Autoregressive Model indicated that own shocks in stock markets were the primary determinants of these markets, whereas Connectedness analysis showed that stock and gold markets were the receivers or contributors of spillovers in the regimes of high and low volatility, whereas oil was spillover contributor and receiver in the high and low regimes of volatility respectively. Gokmenoglu and Fazlollahi (2015) investigated the impact of oil prices, gold prices, oil volatility and gold volatility indices on stock market price of the USA. Applying ARDL co-integration estimation approach, the study found that a long term association existed between the variables, and stock market index revealed to be reverted towards its long term equilibrium by a slow daily adjustment speed. Applying VAR-GARCH model, Arouri et al. (2015) examined how gold price volatility affected the stock market prices over the 2004–2011 period in China. Their findings revealed that volatility in gold price exerted a major influence on the stock market performance of China. Singhal, Choudhary, and Biswal (2019) estimated the dynamic linkage among oil prices, stock market index, gold prices and exchange rate in Mexico. The findings from ARDL estimation revealed that gold prices positively affected stock markets but did not affect exchange rate significantly, whereas oil prices affected stock market prices positively but affected exchange rate negatively over the studied period (Das et al., 2018). applied Non Parametric Causality-in-Quantile estimation to analyze the dependence of gold, crude oil and stocks with financial stress and concluded that gold and oil had bilateral causality with financial stress both at means and variance. Applying DCC-GARCH connectedness approach (Benlagha and El Omari, 2022) estimated how COVID-19 pandemic affected the dynamic connectedness among oil, gold, and five leading stock markets namely Germany, China, USA, UK and Japan. According to the author's conclusions strong connectedness was found to be present among gold, oil and stock markets in COVID 19 period as compared to pre-COVID period. The study of Khalfaoui et al. (2021)was an attempt to estimate the dependence structure between non-energy and energy commodity pairs at various quantiles and frequencies. Findings revealed a low meaningful dependency between energy and non-energy commodities markets at various quantiles and frequencies and some non-energy commodity markets had a neutral relationship with global energy commodity prices. Sarwar, Tiwari, and Tingqiu (2020)selected Bombay, Karachi and Shangai stock markets to study the volatility spillover between oil and stock markets returns before and after crisis period through bivariate BEKK-GARCH model. The findings revealed bidirectional volatility spillover between oil and stock markets in Karachi stock market, unidirectional in Shanghai stock market and mixed (unidirectional or bidirectional) in Bombay stock market. Moreover, the results did not reveal any significant difference in spill over effects between and after crisis. Khalfaoui et al. (2019)analyzed the volatility spillovers between stock market and oil market and portfolio and hedging implications of oil exporting and oil importing countries. The findings of corrected conditional correlation (cDCC) and dynamic conditional correlation (DCC) GARCH models indicated that the lagged volatility in the stock and oil market impacted volatility significantly in the respective markets. Negative shocks were higher in magnitude than positive shocks in asymmetric analysis. Portfolio optimization findings indicated that oil-exporting countries investors should hold more oil assets for risk hedging purposes and no interdependencies were found between markets both in oil exporting and oil importing countries. Sarwar et al. (2019) analyzed the shock transmission and volatility spillovers and shock transmission between firm oil assets and stocks in Pakistani firms by using the BEKK-GARCH model. Volatility spillover had been found from stocks to oil and from oil to stocks in analysis. The portfolio optimization results demonstrated the importance of oil and gas assets in the creation of an optimal portfolio.

In continuation, Smyth and Narayan (2018) made a review of existing literature about oil prices and stock prices relationships and concluded that higher prices of oil caused interest rates to rise. Xiao, Hu, Ouyang, and Wen (2019)applied the quantile regression approach in an attempt to find that how fluctuations in oil price volatility impacted changes in volatility of China stock markets. Their findings revealed that the effect was positive and relatively larger in bearish market conditions but existed in the short run mainly. Moreover, Luo and Qin (2017)analyzed how price volatility of oil and shocks in oil prices affected the stock market of China and returns of five sectors, and found that there was a positive impact of oil price shocks on Chinese stock prices. Bouri, Jain, Biswal, and Roubaud (2017) examined the nonlinear causality and cointegration between crude oil, international gold, and Indian stock market. The findings indicated that cointegration relationships existed between these variables and there existed a positive and non linear effect of the gold and oil volatilities on the Indian stock market volatility. Recently, Yousaf et al. (2021) tried to study that whether gold acted as a safe haven and a hedger in the era of COVID-19 pandemic by analyzing 13 Asian stock markets and found that gold acted as a strong safe haven for Indonesia, China, Vietnam and Singapore, but a weak safe haven in Thailand and Pakistan. Similarly, Ming, Zhang, Liu, and Yang (2020) found the safe-haven properties of the gold for the Chinese stock market during period of two severe financial crises (the global financial crises of 2007–2008, equity market crash of 2015) and negative movements of the financial markets(Rao, 2017).

The review of above literature revealed us that although many studies have researched the effects of oil prices and oil price volatilities in various countries or group of countries on stock market performances Raza et al. (2016) Gokmenoglu and Fazlollahi (2015) Luo and Qin (2017), however very little attention has been given to the effect of gold prices or gold price volatility index on the stock markets. This is particularly scarce in terms of China Arouri et al. (2015). To our best knowledge, although investigations have been done in terms of the relationship of implied gold prices volatilities and oil prices volatilities with stock markets, however, we cannot find any study that has investigated the impact of the prices of oil and gold as well as their implied volatilities together in the context of any country. This study is novel in this aspect that it estimates the impact of the prices of oil and gold and the volatilities in these prices on the stock market of China, because not only international commodity prices, but uncertainty of their prices has significant implications for stock market investments (Pitts, 2017). The rising gold and oil volatility serves as a warning to producers and investors, exposing them to danger. As a result, understanding price volatilities of these two commodities improves our understanding of financial and stock markets and allows us to get a clearer picture of the economy as a whole. Furthermore, previous studies mostly tried to estimate the spillover effects of these variables on stock markets performances and ignored the understanding of the dynamic relationship between these variables in the long run(Miller and Wager, 2017). So our study is a major contribution in the existing literature because it is going to estimate the dynamic relationship between oil gold prices as well as the volatilities in these commodities prices and the stock market of China by applying the ARDL bound testing approach in addition to NARDL and asymmetric causality analysis, which has various advantages over conventional granger causality and OLS estimations as stated in the introduction part.

Empirical methodology

Data and model specification

The econometric model in this regard, according to the theories and literature review, considers oil prices, gold prices, volatility in gold prices, volatility in oil prices as explanatory variables, while stock market price as dependent variable. Our study model in its functional form is specified as:

The econometric form of the model can be written as

Where, Sp = stock price index, oil = oil prices, gold = gold price, goldvolt = gold price volatility index and oilvolt = oil price volatility index, and t = time period. We represented Chinese stock market with CSI300 index. It is a capitalization weighted index that covers the 300 most liquid and largest traded stocks in the Shenzhen Stock Exchanges and Shanghai Stock Exchange, and about 60% of the market capitalization is represented by this. We measured oil prices by crude oil prices (dollars per barrel) and data is assessed through economic data by Federal Reserves. The data of gold prices is taken from World Gold Council, measured in terms of USD/per troy ounce. In addition, along with using typical price series, GVZ index is employed as a measurement for gold price volatility, and OVX taken as the measurement for the volatility of oil price. GVZ and OVX are considered indices since they indicate the uncertainty of market, on the bases of both past data of volatility, and investors' expectations for market situations in future Gokmenoglu and Fazlollahi (2015); Ji and Fan (2012). The data of these two indices has been collected from CBOE official website (CBOE, 2021) spanning over Sep 2009 to Dec 2021. To improve the economic interpretation, variables are used in logarithmic form.

ARDL bound Co integration test

The goal of the research is to examine and understand the relationship of international gold prices and oil prices, volatility in oil price and volatility in gold prices with China stock market. ARDL bound technique is applied in the study to estimate the long run association between gold price, gold price volatility, oil price, oil price volatility, and stock prices Pesaran et al. (1996). Co integration establishes the long-term systemic movement of two or more than two macroeconomic variables. The cointegration equation is approximated using long run error term in the error correction model (known as error correction term), when cointegration among variables has been recognized. It reflects the speed at which the variables adjust over time, which gives an indication of the relationship's stability. The ARDL method provides a number of advantages. ARDL approach can be applied in all situations; no matter the studied variables have zero order of integration or of 1, or a combination of both, unlike the other techniques of assessing co integration. However, there should not be integration of order 2 in the series; else, the ARDL approach would not be used. Second, once the lag order of model has been determined, cointegation is tested using the usual least squares bound test approach (OLS). Third, we use the lagged variables in the ARDL technique, which helps to overcome the endogeneity problem. Finally, when we model the co-integrating relationship with tiny samples, the model works better and avoids concerns with weak power Pesaran (1997).

Initially, Phillip Person KPSS, ADF unit root tests are used in the levels and in the first difference both in our investigation. These are applied to ensure the stationarity of the variables and that neither of them has second order of integration. On the basis of the specifications, the ARDL model is constructed, based on the AIC (Akaike information criterion) in the second stage. After that, bounds testing is used to see if the dependent and explanatory variables have a co-integrating connection. Due to the scarcity of relevant literature, there is no agreement on the directions of long term associations, thus we estimated the UECM (unrestricted error correction model) regressions as follows: where, oil represents international oil price, gold represents gold prices, goldvolt represents volatility in gold prices, oilvolt denotes volatility in oil prices, and last, sp is China stock market. The H0 of all of the above equations is as follows:

oil equation:  =  =  =  = 0

gold equation:  =  =  =  = 0

oilvolt equation:  =  =  =  = 0

goldvolt equation:  =  =  =  = 0

sp equation:

The H0 in the above four equations assumes the absence of any long term link between the series. Wald F-test is applied to test this hypothesis. It calculates the F statistic as well as lower and upper critical values by detecting the combined significance of lag terms of these variables in the equation. If F statistics exceeds the upper limit critical value, it provides the evidence for cointegration, or vice versa. The outcome does not provide any conclusion if the F statistics lie between the upper and lower bound critical values. The cointegrating equation is approximated using error term once the cointegration between variables has been established, and is known as error correction term. It reflects the rate at which the variables adjust over time, which gives an indication of the relationship's stability.where, the ECTt-1 phrase denotes the long term equilibrium adjustment speed following a short-term shock. In post estimation diagnostics, we applied the stability tests such as CUSUM (Cumulative sum) and CUSUMSQ (cumulative sum of squares) and the statistical results are within the critical bound at 5 percent, showing the stability of regression equation.

Non-linear ARDL estimation

For robustness, we apply Non linear ARDL and Asymmetric Causality Analysis also in estimation procedure. The NARDL model is conceptually more attractive than the other nonlinear approaches to cointegration because of the reasons listed below. NARDL outperforms the threshold cointegration model because it compensates for both short- and long-run asymmetries, whereas the threshold cointegration model takes long-run asymmetry only into account. Because it accounts for both long- and short asymmetry, NARDL beats the threshold cointegration model. Third, it separates short- and long-run impacts from explanatory variables to the dependent variable. Even if all 3 of the preceding facts may be evaluated using a nonlinear threshold VECM or smooth transition model, these methods may have a convergence issue due to the large number of parameters and that is not the case with the NARDL model. Fourth, unlike other ECM models that require all time series to be integrated in the same order, the NARDL model eliminates this requirement and permits data series with varied integration orders to be combined. This flexibility is very important for macroeconomic series. Through negative and positive partial sum decompositions of independent variates, the NARDL model combines long-run and short run nonlinearities. The partial sum approach is used in asymmetric decomposition to integrate the decrease and increase in the variables Khalfaoui et al. (2021); Sarwar et al. (2021); Sarwar et al. (2021). Furthermore, Hatemi-J (2012) causality test is used to determine the asymmetric causality direction. The test is unique in that it distinguishes between the causal effects of negative and positive shocks. The unit root test is often used as a preliminary step in the technique to determine the maximum order of integration of variables (dmax). The next step is to use the model information criteria to determine the vector autoregression optimal lag length (k). After that, the (k + dmax)th order vector auto regression framework is estimated. Finally, the usual Wald test is employed to determine whether the variables have a causal relationship.

Results and discussion

Before moving toward the ARDL Bounds testing, the tests for unit root are performed to evaluate the order of integration in the series. The series in the ARDL approach should not be integrated of order 2; else, the erroneous outcomes would be produced. Hence, three different tests for unit root, like PP, KPSS, and ADF are applied. We used these tests at the level, with time trend variables and with constant, and at first difference, only at constant. Table 1 below presents these results. The H0 in the PP and ADF tests assumes the non stationarity of the series, whereas it is opposite in KPSS test. The test results show that neither of the series has second order of integration and therefore permits us to follow the ARDL bound estimation.

Table 1

Unit root Findings.

Unit root (level)	ADF	PP	KPSS
Oil	−2.132	−2.334	0.795a
	0.421	0.410
Gold	−1.561	−1.564	0.348a
	0.673	0.670
Oilvolt	- 4.0571
	0.336
Goldvolt	−1.245	−1.675
	0.679	0.671
Sp	−0.048	0.094	0.254a
Unit root (first difference)
D(oil)	−57.567	−57.60	0.067
	0.000a	0.000a
D(gold)	53.81	53.81	0.012
	0.000a	0.000a
D(oilvolt)	−56.75	−55.01	0.07
	0.000a	0.000a
D(goldvolt)	−50.95	−50.94	0.06
	0.000a	0.000a
D(sp)	51.86	−52.28	0.062
	0.000a	0.000a

Indicates the rejection of null hypothesis at 5 percent significance level. In PP and ADF, null hypothesis assumes the non stationarity or unit root in the series. But reverse is true in the case of KPSS.

The lag structure of the variables examined is still important in the ARDL model. As a result, the lag length maximum of 4 lags, is determined using the Akaike Information Criterion (AIC), Schwarz Information Criterion (SIC) and Hannan-Quinan Information Criterion (HQ) as shown in Table 2 below.

Table 2

Results of optimal lag order selection.

Lag	LogL	LR	FPE	AIC	SC	HQ
0	50.758	55.889	1.35e-04	−2.93540	−4.4364	−2.5369
1	132.744	18.131a	8.23e-12	−4.4567	−2.5356	−4.2445
2	138.748	22.2354	2.53e-04	−3.1245	−4.6958	−3.3751
3	222.242	23.8415	2.41e-05	−5.9589	−1.76114	−4.4437
4	235.244a	27.8441a	2.56e-04	−5.6576a	−1.2561a	−3.3549a

Represents optimal lag order. LR, sequential modified LR test statistic, FPE = final prediction error, AIC = Akaike information criterion, SIC Schwarz information criterion, HQ = Hannan–Quinan information criterion.

The cointegration test is applied to see if the movements of the variables are related in the long run (See Table 3 ). Co-integration establishes the long-term structural co movements of two or more than two macroeconomic variables. Cointegration eliminates the risk of erroneous correlation between variables. Keeping stock price as the dependent variable, cointegration is observed among the variables. This means that anytime there is any change in the system, the stock market will move first, followed by the gold price and its volatility, oil price and its volatility. When gold price volatility is considered as explained variable, we observed co-integration to be present with 10% significance level. This is because the F- statistics for gold price volatility and stock price, at 5% and 10%, respectively, are higher than the critical values of upper bound limits. When oil volatility and gold volatility are considered as explained variables, cointegration is found to be insignificant. This means that, any changes in gold and oil prices in the long run, are unrelated to changes in stock price, oil price and gold price volatility.

Table 3

ARDL-bound cointegration test findings.

Dependent variable	Cointegration (H0)	F-stat structure of optimal lag	F-statistics	Outcome
Oil	α_1 = α_2 = α_3 = α_4 = 0	(2,1,2,2)	2.308	No cointegration
Gold	β_1 = β_2 = β_3 = β_4 = 0	(1,4,4,1)	1.293	No cointegration
Oilvolt	γ_1 = γ_2 = γ_3 = γ_4 = 0	(4,2,1,4)	1.234	No cointegration
Goldvolt	μ_1 = μ_2 = μ_3 = μ_4 = 0	(4,2,1,1)	3.554	Cointegration**
Sp	δ_1=δ_2=δ_3=δ_4=0	(4,1,2,4)	5.545*	Cointegration*

AIC criterion provides the basis for optimal lag length. 4 is found to be the optimal lag length. * and ** denotes 5 percent and 10 percent statistical significance levels respectively.

After unit root and cointegration estimations, long run coefficient estimates of ARDL bound testing approach are presented in Table 4 below. All of the variables are observed to impact the Chinese stock market significantly either in the positive or the negative way.

Table 4

Long run coefficients estimated through the ARDL model.

Dependent variable (Stock price)	Coefficients	Prob-value
Oil	−4.603	0.000
Gold	−2.896	0.007
Oilvolt	−0.134	0.000
Goldvolt	0.062	0.028
Cons	14.436	0.000

The findings imply that gold prices exert a non-favourable effect on stock prices (see Table 5 ). One percent increase in gold prices, sp declines by 2.8 percent. It shows that gold assets act as hedger and have safe haven properties for China stock markets. From the earlier studies, Arouri et al. (2015); Low et al. (2016); Yousaf et al. (2021) strongly support our findings that gold possesses hedging qualities and safe haven qualities for the China stock markets in financial crisis period and period of COVID 19 pandemic. However findings presented by Jain and Biswal (2016) and Singhal et al. (2019) for a study of India and Mexico respectively are totally in contrast with us. They discovered that stock markets' downward movements are triggered by negative gold price fluctuations. An increase in the international pricesof gold is anticipated to contribute to the nation's economic development, and stock markets show this by trending upward.

Table 5

Error correction ARDL model in short run.

All variables	d-SP	t-stat
Error correction term	−0.0152***	−1.842
dSP(-1)	−1.0507	−1.236
doil(-1)	0.0574***	−1.063
dgold(-1)	−1.0211	−0.005
dgold(-2)	0.0134	1.734
dgold(-3)	0.0530***	3.415
dgoldvolt(-1)	−0.013	−1.543
doilvolt(-1)	0.0032	1.437
Cons	1.000	0.000

The rise in oil prices exerts a significant and negative influence on the Chinese stock market. For each percent increase in oil prices, SP decreases by 4.6 percents. This finding is plausible because China is an oil importing, rather than a producing country. An increase in global oil output would entice investors to put their money into oil-producing economies, rather than in oil consumption economies. Bai and Koong (2018) and Singhal et al. (2019) are in support of our finding as these studies also observed the negative influence of these prices on stock markets of China and Mexico respectively. However the finding of Luo and Qin (2017) is in sharp contrast with us who indicated that oil prices influenced stock markets of China positively.

Moving towards the impacts of volatilities in commodities prices, we observed that increase in oil price volatilities (OVX) has highly significant and adverse effects on the markets. For each percent increase in volatility index reduces stock prices by 0.13%. It means that higher oil price volatility has a negative effect on the Chinese stock market, making them more vulnerable to the OVX. It's possible that the stock market of China is catching the effects of price volatility of oil on economic activity, and that increasing volatility in oil price has an influence on expected returns. Furthermore, OVX, the indicated volatility obtained by USO options, includes both historical oil spot price volatility and investors' predictions for future uncertainty. As a result, it is more useful and has a greater impact than actual variance, which only includes historical information on price volatility of oil. From empirical literature, Oberndorfer (2009), Joo and Park (2021)and Luo and Qin (2017) and Xiao et al. (2018) specifically provide support for our results in the context of China. However, Bastianin, Conti, and Manera (2016) concluded that oil price uncertainity did not affect stock market in G-7 economies. And finally, gold prices volatility (GVZ) has indicated to be positively impacting the stock market prices of China. In terms of coefficient, one percent rise in gold market price volatility increases SP by 0.06% in the long term. In contradiction with the finding of (Ali et al., 2020) and (Raza et al., 2016) who found negative impact of price volatility of gold on the stock markets, our finding is plausible and in line with the finding of (Gokmenoglu and Fazlollahi, 2015)because when the gold market becomes more uncertain, some investors may decide to shift to the stock market.

After the co integrated interaction is formed, the short run dynamics are studied using an error correction model. The error correction term is statistically significant at the 5% significance level. This means that following a systemic shock, the stock prices will revert to the long run equilibrium. A relatively small error term coefficient, on the other hand, indicates the slow adjustment speed to equilibrium.Where,*, **, *** denotes 1, 5 and 10 percent significance level, respectively.

The table above shows various independent variable lags, having significant impact on stock market in the short run. It is clearly indicated from the table that (dgold (−3)) indicates that if the first difference of 3 days earlier oil price increases by one percent, SP today increases by 5.7%, and (doil (−1)) demonstrates that a 1% rise of the first difference of the preceding day's gold price will raise the sp by 5.3% today. The remaining of the variables do not have any short run effect on the Chinese stock price.

Post ARDL estimation diagnostics

Last, we used (CUSUM) and (CUSUMSQ) tests proposed by (Brown et al., 1975) to check stability in the short run and long run coefficients as the structural changes in variables may exist due to multiple or single structure break. The CUSUM and CUSUMSQ lines shown in Fig. 1 and Fig. 2 fall under critical boundaries at 5%, supporting the stability and fitness of the model (see Fig. 4) (see Fig. 5) (see Fig. 6) (see Fig. 3).

Fig. 1

CUSUM plotfor coefficient stability in ARDL model (2009–2021).

Fig. 2

CUSUMSQ plot for coefficient stability in ARDL model (2009–2021).

Fig. 4

Short run and long run multipliers for oil prices.

Fig. 5

CUSUM plot for coefficient stability in NARDL model.

Fig. 6

CUSUMSQ plot for coefficient stability in NARDL model.

Fig. 3

Short run and long run multipliers for oil prices.

NARDL estimation results

For robustness, we also applied NARDL approach and the corresponding short run and long run results are provided in Table 6 .

Table 6

NARDL long run and Short run Results.

Long Run EstimationVariables Coefficient P-value
Oil	−3.033	0.009
Gold	−1.223	0.090
Oilvolt	−1.246	0.010
Goldvolt	1.602	0.018
Short Run Estimation
dSP(-1)	−0.027	0.000
doil(-1)+	−0.298	0.009
doil(-1)-	−0.089	0.023
dgold(-1)-	−0.010	0.008
Dgold(-1)+	−1.223	0.006
dgold(-2)-	−0.099	0.998
dgold(-2)+	−0.786	0.334
dgold(-3)-	−0.667	0.076
dgold(-3)+	−1.432	0.088
dgoldvolt(-1)-	0.114	0.678
dgoldvolt(-1)+	1.550	0.124
doilvolt(-1)-	−0.967	0.835
doilvolt(-1)+	−0.889	0.776
Cons	1.000	0.000
Diagnostics Statistics p-value
LM test	0.9733	0.608
Heterosckedasticity test	0.7778	0.885
Normality test	0.332	0.455

The results of long run NARDL analysis are similar to those of ARDL bound test. The results indicate that gold prices and oil prices have significant negative impact on stock prices i.e., rise in gold prices reduce the stock prices but decline in gold prices increase stock prices. Similar to ARDL results, the findings of NARDL reveal that oil price volatility has adverse influence on stock prices but gold prices have positive impact on stock prices in the long run. Moving towards long run dynamics of NARDL estimation, it is found that the results are similar to those in the short run. i.e., gold price, oil price and oil price volatility adversely affect stock prices whereas gold price volatility has positive effect on stock prices.

The model stability, as well as the short- and long-run multipliers of rises and drops in the price of oil, gold, oil price volatility, and gold price volatility, have been assessed. The findings demonstrated that the study's model is stable, and that the impact of both rises and falls in the explanatory variables takes several years to fully felt.

CUSUM and CUSUM Square Plots

Asymmetric causality analysis

Last, the findings of Hatemi-J asymmetric causality test are shown in Table 7 . It is clearly indicated from the table that the null hypothesis that the positive shocks ingold, oil, goldvolt and oilvolt do not cause positive shocks in Sp, and negative shocks in gold, oil, goldvolt and oilvolt do not cause negative shocks in SP are rejected. Hence it is concluded that asymmetric Causal association is present between SP and all of the variable series.

Table 7

Asymmetric causality analysis.

Null hypothesis	Test value	Critical Value (1%)	Critical Value (5%)
Gold+ does not cause SP+	180.66	18.09	11.50
SP+does not cause Gold+	7.65	5.55	3.09
Gold− does not cause SP−	126.39	22.77	11.99
SP- does not cause Gold-	14.56	13.32	4.09
Oil+ does not cause SP+	5.87	11.50	4.89
SP+ does not cause Oil+	176.62	17.09	11.09
Oil− does not cause SP−	.78	21.33	4.76
SP- does not cause Oil-	7.33	13.54	4.09
Oilvolt+ does not cause SP+	12.87	11.40	11.60
SP+ does not cause Oilvolt+	8.36	15.65	4.99
Oilvolt− does not cause SP−	6.77	14.99	3.97
SP- does not cause Oilvolt-	5.89	7.20	4.67
Goldvolt+ does not cause SP+	8.38	7.67	3.77
SP+ does not cause Goldvolt+	7.01	6.66	4.95
Goldvolt− does not cause SP−	7.34	18.01	4.30
SP- does not cause Goldvolt-	16.36	15.65	4.89

Conclusion and policy implications

Gold and oil markets, in addition to their impact on the economy, have also been found to have interactions with stock markets. The relevance of oil to economic activity as well as gold for safe haven and hedging purposes, is well documented in the literature. Given the importance of these commodities to an economy, the stock market of China has also been observed to have interactions with the markets of gold and oil. However, most previous research on the volatility associations between stock and commodities markets primarily focused on historical volatility measurements. Our study is different and novel in this aspect that in addition to gold and oil prices, from sep 2009 to Dec 2021, we examined the correlations between gold, oil, as well as their volatility indexes and stock market of China. Implied volatility indexes are produced from option prices and hence indicate investors' expectations of market circumstances in future, as opposed to historical price volatility measures, only showing historical volatility information (Liu et al., 2013). In this way, they serve as a useful indicator of market risk in the future. In terms of methodology, we used the ARDL Bounds testing, non linear ARDL and asymmetric causality analysis techniques. Findings of the study imply that international gold prices, oil prices negatively impact the stock market of China in the long run. This shows that as global gold and oil prices increase, stock market declines. In terms of implied indices of volatility, we found that oil price volatility has negative, but gold price volatility has positive impact on the Chinese stock market in the long run. These results are useful because a thorough understanding of commodity volatility patterns might aid in the valuation of hedging strategies and derivatives. These volatility indices also may aid in improved stock market trend forecasting, particularly in emerging markets like China. In short run dynamics, our study found the significant effect of gold and oil prices only on the stock market prices. All other variables did not found to have any significant impact in the short term in ARDL analysis only. Moreover, the findings imply that index of stock market converges towards its long term equilibrium by 0.1% of the speed of adjustment (daily) by the contribution of gold price, oil prices as well as volatilities in these prices.

On the basis of our study findings, we recommend that investors and commodities hedgers might use the findings to reduce their portfolio risks. Market participants can employ implied volatility options to hedge volatility risk in the oil, gold and Chinese stock markets, as well as to manage their equity commodity portfolios, as implied volatility contracts have been accepted as risk management instruments. Not only in energy and financial crises periods, but also in health crises scenarios, gold remains a great safe haven commodity as a hedge against risks in investment. As a result, policymakers and investors should implement gold reserve rules that maximize equity portfolios while also ensuring financial system stability. Investors must reap maximum benefits of the information provided by the implied volatility indices and should make quick decisions in response to these volatilities in the oil market produce. In case when the oil market becomes more volatile, they must improve their stock risk management and should invest less in stocks. Policymakers should strengthen the Chinese stock market's basic mechanism and build quick reaction mechanisms for dealing with the quick uncertainty flow of information from the oil to the stock market. Investors should also consider the volatility links between these markets when designing the system. Furthermore, officials should proactively embrace some tactics to assist investors to invest in the stock market of China and to lessen investor concern of rising oil market volatility.

Author statement

We have revised manuscript titled “Covid-19 and oil and gold price volatilities: Evidence from China Market”. Now we are submitting revised draft. There is no author's conflict among authors in this research. The plagiarism of this manuscript is not greater than 13%. So, this will be a good edition in existing literature.