Economic policy uncertainty and commodity market volatility: implications for economic recovery.

As a consequence of the COVID-19 pandemic outbreak, most commodities experienced significant price drops, which were expected to continue well into 2020. As a result, the Markov switching model is used to study the influence of policy uncertainty and the COVID-19 pandemic on commodity prices in the USA. Commodity markets are stimulated by economic policy uncertainty, according to results from a two-state Markov switching model. In both high and low regimes, economic policy uncertainty (EPU) influences the commodity market, according to the study's findings. However, in the high regime, EPU has a greater influence on the energy and metal sectors. EPU has different influences on commodity markets in high- and low-volatility regimes, according to this study. There is a wide range of correlations between COVID-19 outcomes and EPU and how the prices of natural gas, oil, corn, silver, soybean, copper, gold, and steel respond to these tremors, in both high- and low-volatility tenure. Oil and natural gas, on the other hand, are unaffected by shifts in COVID-19 death rates under either regime. Results show that in both high- and low-volatility regimes, the demand and supply for most commodities are responsive to historical prices.

Introduction

Chinese commodities are more dependent on global-related sectors as global financial integration and the expansion of industrialization in emerging economies continue (Yan and Wang 2021; Zhu et al. 2021). Commodity markets have been a key source of worldwide concern for the past few years because of the frequent and dramatic changes in commodity prices and the increased demand for commodities from investors (Rajput et al. 2021). There is a huge deal of attention on the fluctuation of commodity prices in this context. Commodity markets have been the subject of numerous studies, both theoretical and empirical, from a variety of perspectives.

Academics are concerned about the consequences of economic policy uncertainty (EPU). Recent events have led to a widening financial crisis that continues to spread throughout the international economy (Ellis and Liu 2021). The risk transmission across multiple financial markets has been strengthened as an outcome of these events (Wei and Han 2021). A wide range of economic policies is routinely used by countries to ensure the smooth development of their internal economy. It is apparent that policy is a factor in commodity markets. As a result, examining the link between China’s commodity markets and economic policy uncertainties is of enormous practical importance.

Since its discovery in Wuhan, China, the unique coronavirus disorder 2019 (COVID-19) has caused unprecedented public health problems and severe social and economic consequences around the globe. There have been above 8 million cases of the disorder, and roughly 0.46 million deaths, globally (Kumar and Nayar 2021; Zanke et al. 2021). More than twice as much as at the time of the global economic crisis of 2008–2009 is predicted to be lost in 2020 as a result of a worldwide loss approaching 3%. From the time when the COVID-19 pandemic started, several variables have changed dramatically. Since the outbreak of the COVID-19 epidemic, oil prices have largely reduced. Global oil requirement was predicted to shrink by roughly 10% in 2020, which is greater than twice as much as the next greatest reduction in 1980 (Aktar et al. 2021), which was around 7%. COVID-19 epidemic has a detrimental impact on all countries, but the fall in oil prices has a different effect on countries that are either oil importers or exporters.

As the COVID-19 epidemic has spread, so has the level of uncertainty. It is possible to see an increase in global economic policy uncertainty from 234 in January 2020 to 298 in September 2020, based on the news-based global economic policy uncertainty index (PPP-adjusted GDP) (Bai 2021a). Pandemic uncertainty stems from a variety of factors, including the disease’s infectiousness, widespread distribution, and short incubation period. The length of time that the effects of the pandemic-induced (and unexpected) shifts in consumer expenditure would last also adds to the uncertainty. Uncertainty about the future of economic policy has been shown to have a negative impact on the economy’s ability to grow and create jobs, particularly in policy-sensitive industries such as healthcare and the defense sector.

In general, financial markets, and stock markets in particular, have been negatively influenced by the COVID-19 epidemic (Yumei et al. 2021). During the week ending February 28, 2020, global stock markets saw their largest weekly losses since the 2008 global financial crisis. The value of the global equities market plummeted by 30% in March 2020 (Mohsin et al. 2021b; Notteboom et al. 2021). The financial market’s reaction to the COVID-19 pandemic has been more negative than that of previous infectious illness pandemics, such as the Spanish influenza pandemic of 1918, which murdered a projected 2.0% population of the world (Asghar et al. 2021; Rao et al. 2022). The 2nd and 3rd waves of the virus have produced volatility in various nations, despite the global stock market recovery in the latter part of the year.

Over the years, the stock market has seen its share of ups and downs. Many studies have been undertaken to examine the macro- and microeconomic causes, oil price movements, inflation, recession, and interest rate movements, among other things, that have contributed to these changes in the economy (Anser et al. 2020; Khokhar et al. 2020; Mohsin et al. 2021a). Since the factors that cause the stock market’s ups and downs might be explained by economic policy uncertainty or volatility, this study focuses on these two issues in particular. When the COVID-19 epidemic broke out in countries and communities around the world, it had an impact on the global economy in 2020 that was unprecedented in the past century (Fu et al. 2021; Hou et al. 2019; Iqbal et al. 2021). Businesses are struggling with lost revenue and supply chain network interruptions as facilities close and lockdown measures are extended around the world. Unemployment is also at historic highs. There is still a lot of uncertainty, despite the fact that governments around the world are scrambling to put fiscal and monetary policies in place in order to lessen the recession’s consequences. Volatility is expected to rise in lockstep with rising uncertainty as a result of all of this.

In a number of areas, this work adds to the body of knowledge. Markov swapping dynamic model, newly created by Balcilar et al. (2016), has never been employed to analyze the monitoring and control of commodities’ volatility and returns via EPU and investor attitudes, to the best of our knowledge. Causation in returns and volatility can be found using the Markov swapping dynamic model, which employs a more generic procedure. The regime-swapping model was applied to define the time series shifts between distinct commodity price regimes—particularly crude oil prices—due to structural breaks. In the Markov regime-swapping model, we applied a lag-dependent variable to correct omitted-variable bias, contributing to model parameter estimate bias. At high speeds, the Markov regime-swapping method can represent provisional volatilities. The technique has been utilized to apprehend significant abrupt changes in oil price fluctuations (Hamilton 1989). Additional research also shows that the Markov swapping model can accurately examine the volatility in oil-based product forthcoming value chain (Fong and See 2002) as well as to forecast the transition possibilities among low- and high-growth regimes and examine the mean change in the US GDP with oil price. Misspecification errors, structural breaks, and frequent outliers in financial time series are not a problem for the swapping dynamic model.

As a result of our research, policymakers will be able to better understand the impact of COVID-19 on commodities under various regimes. It may be advantageous for portfolio managers and investors, especially during uncertain times like pandemics, to hedge efficient short-term dangers in their assets and portfolio. Using our findings, investors and regulators may better assess and forecast commodities return transitions in volatile locations.

Literature review

Commodity markets and effect of COVID-19

Most recent studies on commodity markets have focused on the financialization of commodities. Its phenomenon has begun in 2004, when fund influxes into the market rose from $15 billion to over $450 billion in April 2011 according to Gao et al. (2016) and Huang et al. (2021a). In the mid-2000s, commodity derivatives trading expanded dramatically, according to Zhang et al. (2021c), whereas Abbas et al. (2022) point out that signs of commodity financialization soared during the 2008–2009 global financial crisis. Contrary to popular belief, Zhang et al. (2021b) report that the financialization of the metals and agricultural markets has been cyclical, with a de-financialization occurring between 2014 and 2017. Zhu et al. (2021) further show that commodities in Canada have new diversification options during the period of greatest financialization. The returns and volatility of commodities are mostly driven by financial variables, as demonstrated by Rajput et al. (2021). It has been shown that global macroeconomic conditions have an impact on the stages of the commodity price cycle. The relationship between the Chinese stock market, commodity markets, and the world crude oil price are dynamic, according to Yu et al. (2021). In addition, commodities have been shown by Kim and Yasuda (2021) to be advantageous as a place of refuge, hedge, or portfolio diversification. Because commodities play a critical role in the economy, Zhu et al. (2020) show that both traditional and non-traditional financial policies can influence commodity prices. Commodity price variations have a significant influence on economic growth, according to Scarcioffolo and Etienne (2021). The volatility of commodity markets is asymmetric, which means the volatility is larger following a positive price tremor than a negative price tremor.

Commodity prices have a projecting material for exchange rates, according to Huang et al. (2021b). There is evidence that commodity prices can be forecast using information from other markets, according to Yang et al. (2021b). It has also been shown that commodity prices may be used to predict inflation (Long et al. 2021). Commodity prices can be predicted by global trade uncertainty, according to Ellis and Liu (2021) and Wu et al. (2022a). It has been shown that commodities play a key part in the active consideration of climate, sickness, economic, or geopolitical dangers, or “hazard fear” by Chakraborty and Thomas (2020). Commodity prices can forecast GDP growth, according to Wang et al. (2022). It has been shown that commodity spot and future prices are linked by Tran (2021). A key driver of commodity prices, according to Sha et al. (2020), is speculation. Shafiullah et al. (2021) demonstrate intraday return predictability for commodity ETFs using high-frequency data. Furthermore, Dai et al. (2020) show that Chinese commodities futures markets have intraday momentum.

According to existing research, it is also critical to make a distinction between different commodity groups, such as energy and agriculture, and livestock and precious metals. Considering 21 different commodities, Dai et al. (2021) conclude, for example, that valuable and manufacturing metals are a superior hedge and safe-like haven than other commodities. In contrast, Yuan et al. (2020) show that the volatility of crude oil prices is more adversely affected by pandemic uncertainty than gold prices. As far as commodities are associated, only crude oil has an opposite leverage impact, according to Dai and Yin (2020); Lee et al. (2019); Shen et al. (2021); and Xiang and Qu (2020). According to Zhang et al. (2021a), the long-term price equilibrium association between industrial metal and crude oil markets exists, but not between the agricultural and the gold market. According to Yuan et al. (2022), the gold and crude oil markets are more responsive to market dimensions, but soybeans are not. According to Wu et al. (2022b), there is a bigger time-varying impact on agricultural commodity prices than there is on metal and energy prices. We can see from Shang et al. (2021) the short-term and medium-long-term transmission intensity of metals, while energy is highest in both time frames. Gold futures, according to Zhou et al. (2021), can be used to protect against stock market losses, although the vast variety of commodity futures appear to be viewed as a distinct asset category because of the increasing financialization of commodities. The energy futures market also actively takes part in the coordination of stock and commodity markets, as shown by Bai (2021b). In order to see if the price overreaction behavior varies between various commodities, we looked at 20 distinct commodity futures because of this heterogeneity.

Furthermore, several past studies examine the connection between commodities and other asset types. For instance, Chen et al. (2020) explore the volatility connectivity between credit defaults swaps (CDS) and commodities and find that commodities can transfer volatility to CDS. Volatility transmission varies in strength according to the commodity type, with energy commodities and precious metals having the strongest impact. There is a correlation between commodities and stocks between BRICS nations and the USA, according to Dash and Maitra (2021). Researchers have discovered an ever-shifting network structure among these assets, with effects on the network that can be felt both locally and internationally. Li (2021) expresses that the volatility spillover between energy and agricultural commodities is asymmetric. With regard to agricultural commodities, there is also a major risk of spillover from energy sources. Crude oil is a particular commodity that provides greater modification profits than other commodities, according to Aloui et al. (2016), who differentiate between non-energy and energy commodities. Herding behavior may play a substantial role in explaining the movement of commodity prices, according to some researchers. As an instance, Fasanya et al. (2021) discover that 24 Chinese commodities exhibit positive response business, noisy business, and a steer mentality; on the other hand, Aslam et al. (2022) show that steering behavior varies between markets and it is asymmetric. The bottom line is that, in earlier studies, commodities were shown to interact significantly with other asset classes and to be a fence and safe-like haven for other asset categories under certain conditions. Each commodity category has its unique characteristics that must be taken into account (energy, agriculture, non-energy, industrial metals, precious metals). The herding behavior of investors may also affect commodity prices. These observations have prompted us to investigate the futures price overreaction behavior of a broad sample of 20 commodities.

These studies show that commodity futures have a significant impact on financial and economic systems. Various pricing patterns have been studied, including cycles, hedge, predictability, and safe-like haven. However, given the impact of the COVID-19 outbreak, we have not seen any studies examining the sensitivity of commodity futures prices based on hourly data. As Deev and Plíhal (2022) demonstrate, uncertainty shocks can have a considerable impact on commodity prices. The COVID-19 epidemic’s impact on the commodity market was the subject of scholarly investigations at the time of this study. None of them has examined their pricing overreactions. For instance, Wang and Sun (2017) found that the volatility of commodity prices is affected by the number of deaths and definite cases caused by the COVID-19. According to Yang et al. (2021a), a worldwide fear index for the COVID-19 epidemic has the ability to anticipate commodity prices, with commodity yields being positively connected with an increase in COVID-19-related dread.

We can infer from the aforementioned studies that it is critical to comprehend the commodity market’s response to the COVID-19 pandemic. No current studies have examined the price extreme reaction behavior of commodities at the time of the COVID-19 epidemic, to our knowledge. By studying the price movements of commodities at the time of the COVID-19 epidemic, we add to what is already known about commodity futures pricing and can therefore assist policymakers and investors in better appraising commodity investment risks and possibilities in the future.

Macroeconomic volatility and commodity markets

The empirical evidence supports the hypothesis that the volatility of commodities responds to macroeconomic factors. For example, according to Bianchi (2021), monetary policy and inflation are responsible for gold price volatility. US financial policy news has a calming influence on commodity volatility, according to Ayadi et al. (2020). Economic activity and volatility are linked in several studies (e.g., Aharon and Qadan 2018; Bahloul and Gupta 2018), and Rehman and Vo (2021) show that economic activity news has a rapid and large impact on metal futures’ volatility.

The erratic nature of Chinese commodity markets is the subject of another body of research (Ji et al. 2018). Economic news from both China and the USA affects Chinese commodity volatilities, as An et al. (2020) demonstrate. According to Chen et al. (2021), macroeconomic factors such as GDP growth, industrial production, and money supply affect commodities futures’ volatility. They also show that the Chinese commodity markets are influenced by the economies of both China and the USA. Liang et al. (2021) explore the lead-lag association between macroeconomic futures returns and forecasts, are the most relevant to our research. By focusing on volatility, we stray from their research, and we look at a much broader range of commodities.

Research on the influence of macroeconomic information on commodity prices is relatively restricted and has been unsuccessful to come to an agreement. Batista Soares and Borocco (2021) found no “compelling evidence” that energy prices respond to US macroeconomic news, but Shi and Shen (2021) observed more volatility in crude oil futures prices on the days of events. According to Suleman et al. (2021), macroeconomic news has little effect on metal futures prices. There is evidence of “a quick and large response” to macro-news, as Umar et al. (2021) write, and Christie-David and Cai show that futures prices of gold and silver react strongly to economic data. Similarly, Bakas and Triantafyllou (2018) find a considerable increase in gold futures market volatility in the wake of positive news, and they demonstrate that this increase is correlated with greater belief dispersion.

Recession and financial crises have been shown to have a greater impact on asset prices’ reaction to macroeconomic news than other times (e.g., Sobti 2020; Sun et al. 2021). As Ma et al. (2021) explain, one reason for this is that the announcements may be interpreted as signs of future economic development. Investors’ moods fluctuate wildly during recession and crisis situations; therefore, this could also be the cause (Hu et al. 2020). A further possibility is that commodity markets have become more and more financialized. According to Marfatia et al. (2021), enlarged co-movements between the commodity and stock markets have led to heightened volatility. According to Ahmed and Huo (2021), financialization has led to a previously unseen shift toward a more volatile risk appetite. In conjunction with the indication of increased dissemination of information, this explains the higher volatility in the 2007–2009 financial crisis than earlier.

Research methods and data

Data

The study rely on daily explanations of COVID-19 cases (found by the number of tainted US citizens with an entirely different strain of the virus), oil prices (determined by the WTI benchmark crude oil price), and the US-EPU (news-based index) to calculate their results. On the CDC’s website, COVID-19 data is retrieved. Furthermore, the statistics on the oil market are received from DataStream, while the information on EPU is taken from the EPU website. The data for commodity prices (gas, silver, gold, steel, copper, corn, and soybean) is obtained from Thomson Reuters DataStream.

Methodology

An MS technique is used to examine the influence of economic policy uncertainty and COVID-19 cases on commodity market prices. The developments in US financial policy in the late 1980s prompted the use of the MS approach. We use Fallahi’s (2011) two-state MS model (MS(2)) to analyze the link between economic policy uncertainty, COVID-19, and commodity market prices. There are three regime-switching variables: mean (st) and variance (st) and economic policy uncertainty and COVID-19 cases.

Markov regime-switching approach

The latent process drives the time series utilized in the MS model, which is considered to be stationary (Alizadeh and Nomikos 2004). Consequently, it is impossible to observe the states around which the time series evolves. EPUst, where t = 0, 1 is assumed to be a regime-dependent coefficient of economic policy uncertainty together with the mean, variance, and stddev (Alizadeh et al. 2008). As a result, these characteristics change throughout time in relation to the regimes. High prices of commodities are tied to the 0th regime; a low price of commodities is linked to the 1st regime. Commodities’ prices are predicted to be higher and volatility to be lower when the market is expanding as a result of economic policy uncertainty. As a result, a high-growth and low-volatility regime is indicated by 0 > 1 and 0 > 1, respectively. MS(2) can be written as follows: where Eq. 1 uses state-dependent intercepts (Zt) and state-dependent switching variables EPU and COVID-19 cases to represent a change in commodities prices and state-dependent switching variables (Zt) to represent a change in state-dependent switching variables .

Here, the probabilities of remaining in regime 0 and 1 are p 00 and p 11, while p01 and p10 indicate the movement of probabilities between the two corresponding regimes; thus,

The mean and variance are supposed to behave in MS (2) model.

A low-average-growth regime (St = 1) and a high-average-growth regime (St = 0). In order to achieve worldwide optimization of parameters, we started with more than 1000 estimated specification beginning values. The LR test, residual analysis, and the regime classification measure (RCM) were also used to identify the best model.

Results and discussion

Table 1 presents the descriptive statistics of the studied variables. Table 1 also includes the the Ljung-Box first [Q(1)], Jarque–Bera normality test (JB), the first [ARCH(1)] and fourth [Q(4)] autocorrelation tests, and fourth [ARCH(4)]-order Lagrange multiplier (LM) tests for autoregressive conditional heteroskedasticity (ARCH). First- and fourth-order autocorrelation and autoregressive conditional heteroskedasticities are found for both logarithmic levels and logarithmic differences. The WTI series is more volatile than the gas, silver, gold, steel, copper, corn, and soybean in both logarithmic levels and logarithmic differences measured by the coefficient of variation.

Table 1

Descriptive statistics

	WTI	Gas	Silver	Gold	Steel	Copper	Corn	Soybean
Mean	3.337	1.406	3.003	1.265	2.703	1.139	2.433	1.025
SD	1.910	1.317	1.719	1.185	1.547	1.067	1.392	0.960
Min	0.352	2.309	0.317	2.078	0.285	1.870	0.257	1.683
Max	7.499	4.897	6.749	4.407	6.074	3.967	5.467	3.570
Skewness	0.742	0.714	0.668	0.643	0.601	0.578	0.541	0.521
Kurtosis	0.696	0.287	0.626	0.258	0.564	0.232	0.507	0.209
JB	207.398	160.993	186.658	177.092	167.992	194.802	151.193	214.282
Q(1)	1846.705	1839.056	1809.771	1802.275	1773.575	1766.229	1738.104	1730.905
Q(4)	7341.834	7219.099	7194.997	7074.717	7051.097	6933.223	6910.075	6794.558
ARCH(1)	1847.788	1822.354	1810.832	1785.907	1774.616	1750.189	1739.123	1715.185
ARCH(4)	1844.866	1824.999	1807.969	1788.499	1771.809	1752.729	1736.373	1717.674

***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively

Unit root tests

A linear trend and a constant are both included in the test equation in Table 2, Panel A, which shows results from unit root tests on the log levels of the series. Panel B presents the results of experiments using only a constant as a unit root for the first variations in the log series. There are many different types of unit root tests, including the augmented Dickey-Fuller test (Dickey and Fuller 1979); the Phillips-Perron Z unit root test (Phillips and Perron 1988); MZ and MZt, the modified Phillips-Perron tests of Perron and Ng (1996); and Z, the Phillips-Perron Z unit root test of Phillips and Perron (Phillips and Perron 1988). GLS detrending is required for the Z, MZ, and MZt tests. Lag order is determined for the ADF unit root statistic by testing the significance of each successive lag at the level of 10% significance. We use the modified Bayesian information criterion (BIC)–based data-dependent technique of Ng and Perron (2001) to pick the bandwidth or lag order for the MZt, MZt, DF-GLS, and KPSS tests. Table 2 shows that the KPSS test rejects the null hypothesis that the series is stationary. The null hypothesis of nonstationary series cannot be rejected by any other test. There is a KPSS test in Panel B that does not reject the null hypothesis of stationary series, while other tests do. First-difference stationarity is seen in all the series. To summarize, we find that the logarithmic discrepancies between the selected variables series are not steady.

Table 2

Unit root test

Variable	ADF	Zα	MZα	MZt	DF-GLS	KPSS	Zivot-Andrews
Panel A: unit root tests in levels
LnOIL	1.422	15.546	5.573	1.345	1.466	5071.387***	1.355
LnGAS	1.547	15.253	4.234	1.354	1.334	816.345***	1.354
LnSilver	1.479	16.168	5.796	1.399	1.524	5274.242***	1.409
LnGold	1.608	15.863	4.404	1.408	1.387	848.999***	1.408
LnCopper	1.538	16.815	6.028	1.455	1.585	5485.212***	1.465
LnSteel	1.544	15.229	4.228	1.352	1.332	815.039***	1.352
LnCORN	1.6	17.487	6.269	1.513	1.649	5704.621***	1.524
LnSoybean	1.482	14.619	4.058	1.298	1.279	782.438***	1.298
LnEPU	1.664	18.187	6.52	1.574	1.715	5932.805***	1.585
LnCOVID	1.423	14.035	3.896	1.246	1.227	751.140***	1.246
Panel B: first differences unit root test
LnOIL	10.456***	534.347***	43.346***	4.465***	3.345***	0.234
LnGAS	12.345***	700.342***	56.234***	6.234***	7.343***	0.323
LnSilver	10.874***	555.72***	45.08***	4.643***	3.479***	0.244
LnGold	12.839***	728.356***	58.484***	6.484***	7.637***	0.336
LnCopper	11.309***	577.949***	46.884***	4.829***	3.618***	0.253
LnSteel	12.326***	699.221***	56.144***	6.224***	7.331***	0.323
LnCORN	11.762***	601.067***	48.759***	5.022***	3.763***	0.264
LnSoybean	11.833***	671.253***	53.899***	5.975***	7.038***	0.318
LnEPU	12.232***	625.11***	50.709***	5.223***	3.914***	0.274
LnCOVID	11.359***	644.402***	51.743***	5.736***	6.756***	0.297

Regime switching model

The Markov switching approach was used to study the influence of economic policy uncertainty and COVID-19 on commodity prices. A modest sample size and increased understanding of the link between economic policy uncertainty, COVID-19, and commodity prices lead us to focus on only two regimes for further investigation in this study. MS(2) specification was shown to be more closely aligned with macroeconomic correlations. As a result of the residuals from the linear and MS(2) model estimations, Table 1 presents descriptive statistics and diagnostic tests. Both models have good residual qualities, according to the statistics. The insignificance of the Jarque–Bera test shows that the residuals have a normal distribution. Heteroscedasticity and autocorrelation are not present in the computed residuals as confirmed by ARCH, Q(12) of the Ljung-Box model, and Q2(12). MS(2) model is a true data generator because RCM was 1.7887, which is closer to 0 than any other value (DGP). There are lower AIC, SC, and HQ values in MS(2) than a linear model, according to the information criteria. MS(2) best fits the data for the period 1980–2017 when it comes to the association between economic policy uncertainty, COVID-19, and the commodity prices in the USA, according to information criterion, diagnostic tests, and RCM.

Oil and gas market

There are substantial correlations between 0, 1, and the estimated coefficients (e.g., 0, 1, 0, and 1). It is said that economic activity expands significantly in a high-growth regime like regime 0. Regime 0, for example, has the highest intercept coefficient (0 = 0.086) and the lowest volatility (0 = 0.0014) of any of the possible regimes. The regime’s high mean value and low volatility indicate that the economy is expanding at the time. At an estimated value of 1 = 0.049, regime 1 represents a time of minimal growth, whereas at an estimated 1 = 0.012, regime 1 represents a very significant variance. According to the results, 0 > 1 is true. Mean and variance values show that regime 0 is associated with high growth and moderate volatility, while regime 1 is associated with low growth and high volatility. Low-growth regimes are more volatile than high-growth ones; hence, this conclusion can be drawn.

Economic policy uncertainty and COVID-19 have a favorable and considerable impact on oil and gas returns under both regimes, as shown by regression parameters (see Table 3). However, under a high-growth environment, economic policy uncertainty has a greater impact than in a low-growth regime. In a high-growth regime, a 1% rise in the economic policy uncertainty index results in a 0.033% and 0.842% increase in oil and gas returns, respectively. However, in a low-regime, a 1% rise in economic policy uncertainty would lead to a 0.056% and 0.049% increase in oil and gas returns. In spite of these findings, the commodity market in the USA is influenced by the economic policy uncertainty and COVID-19 in a different way under each of these regimes. This finding supports the idea that in developed nations like the USA, the link between economic policy uncertainty and commodity market is not one-to-one and is dependent on the regime in place. There is a discrepancy between the findings of Albulescu et al. (2019) and Badshah et al. (2019).

Table 3

Effect of policy uncertainty and COVID Oil and gas commodities

Variable	ΔOIL	ΔGAS
Mean (μ0)	0.029***	0.082***
	(0.849)	(0.666)
Mean (μ1)	0.023***	0.032***
	(0.350)	(1.046)
Variance (σ0)	0.026***	0.036***
Variance (σ1)	0.056***	0.049***
ΔEPU0	0.033***	0.842***
	(1.031)	(2.936)
ΔEPU1	0.062***	0.342***
	(1.404)	(3.312)
Mean (μ0)	0.054***	0.116***
	(0.714)	(0.673)
Mean (μ1	0.054***	0.066***
	(0.706)	(0.230)
Variance (σ0	0.081***	0.019***
Variance (σ1	0.098***	0.015***
ΔCOVID	0.058***	0.876***
	(1.248)	(2.970)
ΔCOVID1	0.087***	0.098***
	(1.696)	(3.951)

Effect on gold and silver market

COVID-19 and the economic policy uncertainty index have a statistical impact on gold returns in a low-volatility regime, according to the Markov switching model results from Table 4. Gold returns show a statistically significant reaction under the high-volatility regime. Gold returns rose by 0.659% and 0.183% in a high-volatility environment when COVID-19 cases and economic policy uncertainty increased by 1%. Due to gold strong tie to the US and global economy, it is unlikely to function as a safe-haven commodity for investors, even though gold returns showed a less substantial positive response to all exogenous variables (Bhar and Hammoudeh 2011). In periods of low and high volatility, the lagged return on gold has a substantial negative and positive relationship with the latter’s return. This shows that gold returns are affected by historical events, such as the COVID-19 epidemic. Contrary to previous findings, results from Markov switching suggest that it is more likely to remain in a lower volatility regime than a higher one. Silver returns, unlike gold, show no substantial reaction to external variables in the low-volatility regime. Silver prices rose by 0.036%, the biggest change since 2019. This could be a result of this. Imports and demand from China surged due to the relaxation of COVID-19 restrictions and the implementation of the stimulus package, driving this increase.

Table 4

Effect of policy uncertainty and COVID gold and silver commodities

Variable	ΔSilver	ΔGold
Mean (μ0)	0.039***	0.096***
	(0.688)	(0.430)
Mean (μ1)	0.039***	0.059***
	(0.770)	(0.190)
Variance (σ0)	0.036***	0.011***
Variance (σ1	0.043***	0.021***
ΔEPU0	0.079*	0.659***
	(0.371)	(0.139)
ΔEPU1	0.163***	-0.183***
	(0.544)	(0.312)
Mean (μ0)	0.059***	0.096***
	(0.874)	(0.309)
Mean (μ1	0.059***	0.089***
	(0.255)	(0.477)
Variance (σ0	0.026***	0.011***
Variance (σ1	0.021***	0.021***
ΔCOVID	0.098***	0.659***
	(0.791)	(2.300)
ΔCOVID1	0.183***	0.183***
	(0.964)	(2.463)

Effect on steel and copper market

Steel Markov switching model results in Table 5 show a 1% change in COVID-19 cases, and economic policy uncertainty will stimulate steel returns by 0.020% and 0.012% in a low-volatility environment. There is no correlation between steel returns and confirmed COVID-19 cases, recovery, or economic policy uncertainty in the high-volatility regime. Strong demand for steel in China, which led to a 25% price increase in the third quarter of 2020 when COVID-19 limits were removed due to low reported cases, maybe the cause of the substantial positive coefficient found under low-volatility regimes. In both regimes, steel returns have a strong correlation with its lagging returns. There is a strong correlation between steel prices and historical demand and supply. In the first quarter of 2019, the collapse of the Brumadinho dam in Brazil caused production at Vale to be disrupted due to a scarcity of transportation and workers caused by the COVID-19 outbreak. These steel Markov switching model equations show that regimes with low volatility have a higher likelihood of remaining than regimes with the high volatility of shifting. According to previous studies, gold’s cumulative impulse response during the height of the COVID-19 epidemic was more stable than that of other metal commodities, such as copper, silver, and aluminum (Apergis et al. 2021). However, a correlation between COVID-19 and copper returns has been reported.

Table 5

Effect of policy uncertainty and COVID steel and copper commodities

Variable	Copper	Steel
Mean (μ0)	0.028***	0.065***
	(6.619)	(1.979)
Mean (μ1	0.014***	0.028***
	(0.195)	(1.079)
Variance (σ0	0.005***	0.020***
Variance (σ1	0.010***	0.012***
ΔEPU0	0.068***	0.0228***
	(0.760)	(0.179)
ΔEPU1	0.152***	0.152***
	(0.933)	(2.352)
Mean (μ0)	0.035***	0.072***
	(6.266)	(1.986)
Mean (μ1	0.035***	0.035***
	(0.186)	(1.034)
Variance (σ0	0.002***	0.013***
Variance (σ1	0.033***	0.018***
ΔCOVID	0.034***	(0.635)
	(0.767)	(2.106)
ΔCOVID1	0.015***	0.059***
	(0.940)	(2.359)

Effect on Agriculture Commodity market

Even though the COVID-19 pandemic had a limited impact on agricultural commodities, the global and domestic supply chain disruption and limits on exports or stockpile commodities raise worries about food security issues. Table 6 shows the results of the Markov switching regression on corn and soybean commodities. Death and recovery situations exhibit negative coefficients for low-volatility regimes with no statistical inference, but a substantial positive coefficient at high volatility. In a low-volatility regime, a 1% increase in confirmed cases results in a 0.04% decrease in corn returns. An increase of 1% in confirmed cases will statistically improve corn returns by approximately 7.7% correspondingly under the high-volatility regime. As a result of the fall in oil and natural gas production as a result of low market prices, COVID-19 cases and the economic policy uncertainty index had a positive association with corn returns. This could have an impact on the pricing of biofuel crops like corn and soybeans. If the results are insignificant, it may be because the poor sensitivity of crops like maize to external shocks that are not fundamental may be the cause. This may imply that most agricultural products are essential to global food security. Even in moderate volatility, the impact of economic policy uncertainty on maize returns is significant—and this uncertainty can have a detrimental influence on the economy in high volatility. There is a 3.5% chance of low volatility and a 1.8% chance of a high volatility transition.

Table 6

Effect of policy uncertainty and COVID on agriculture commodities

Variable	ΔCORN	ΔSoybean
Mean (μ0)	0.068***	0.093***
	(0.659)	(0.427)
Mean (μ1	0.068***	0.056***
	(0.189)	(0.807)
Variance (σ0	0.035***	0.008***
Variance (σ1	0.030***	0.018***
ΔEPU0	0.018***	0.056***
	(0.862)	(2.207)
ΔEPU1	0.052***	0.180***
	(0.833)	(2.383)
Mean (μ0)	0.037***	0.071***
	(0.628)	(0.985)
Mean (μ1	0.017***	0.034***
	(0.037)	(0.071)
Variance (σ0	-0.004***	0.016***
Variance (σ1	-0.023***	0.026***
ΔCOVID	-0.077***	0.064***
	(0.769)	(2.185)
ΔCOVID1	0.161***	0.015***
	(0.942)	(2.358)

Conclusion and policy implication

This research analysis describes the link between COVID-19 number of cases and economic policy uncertainty on commodity prices. The study is grounded on a two-state Markov switching technique. The findings encourage the existence of a non-linear link between COVID-19 number of cases and economic policy uncertainty on commodity prices in the USA. The conclusion demonstrates that economic policy uncertainty exerts a large optimistic influence on commodity prices in low- and high-growth regimes. However, the influence of economic policy uncertainty on commodity prices was rather considerable in the elevated growth phase. This indicates that commodity prices react differentially to economic policy uncertainty in low- and high-growth regimes in the USA. This in addition shows that the relationship between economic policy uncertainty and commodity prices is non-linear. In our COVID-based Markov technique, extreme COVID-19 definite cases are troublesome for prices of oil commodities due to COVID-19 mitigation actions that severely restricted transport and travel which accounts for approximately 67% of oil requirement in a low-volatility regime. Rising COVID-19 cases disturb the price of natural gas requirement, although the influence is significantly slighter given the predominant usage of natural gas for the generation of electricity and domestic cooling and heating because of COVID-19 laws on travel limitations. On the other hand, elevated COVID-19 revival situations will diminish natural gas yields due to loose lockup regulations. The association between soybean profits, corn profits to the COVID-19 casualty, confirm, recovery cases, and economic policy uncertainty index is optimistic in elevated volatility regimes. In a less volatile environment, corn earnings reveal minimal association because of the low susceptibility of agricultural products to outside tremors. Indication from the research reveals soybean earnings are reactive to past growth in supply and demand of soybeans in both regimes.

These studies can offer awareness for the prevarication possibility of silver and gold in the days of pandemics. Silver and gold prevarication possibility varies with time and is tenure dependent, suggesting that they change among Markov regimes. It is possible for short-term investors to properly hedge against systematic risks in their portfolios. Our findings can serve as a reference for future investors looking to invest in similar pandemics. Regulators can use the findings to assess and estimate the likelihood that the market will remain in a certain regime and begin the process of transitioning to a new normal. The results of natural gas and oil earnings can help oil-exporting nations formalize measures against future worldwide pandemics on the world market for energy commodities. OPEC, for example, has the power to limit supply in order to increase demand because most commodities are influenced by previous market trends. Using dynamic autoregressive distributed lag models, future studies can examine the impact of hypothetical shocks on commodity markets.