Resources Policy 82 (2023) 103490

Contents lists available at ScienceDirect

Resources Policy
journal homepage: [Link]/locate/resourpol

Dynamic connectedness in commodity futures markets during Covid-19 in

India: New evidence from a TVP-VAR extended joint
connectedness approach
Aswini Kumar Mishra a, *, Vairam Arunachalam b, Dennis Olson c, Debasis Patnaik a
a
Department of Economics and Finance, BITS Pilani KK Birla Goa Campus, India
b
School of Accountancy & PwC/Joseph A. Silvoso Distinguished Professor, University of Missouri, Columbia, USA
c
Department of Business Administration, Oregon State University-Cascades, 1500 SW Chandler Avenue, Bend, OR, 97702, USA

A R T I C L E I N F O A B S T R A C T

JEL classification: This paper presents a unique time-varying parameter vector autoregression (TVP-VAR) based extended joint
JEL Code connectedness approach to quantify the connectedness and transmission mechanism of shocks of nine com­
C32 modities futures returns (namely; Gold and Silver from the category of precious metals; Copper, Lead, Zinc,
C5
Nickel and Aluminium from the category of base or industry metals; Natural Gas and Brent Crude Oil from energy
F3
G15
sector) obtained from Multi Commodity Exchange of India Limited (MCX) from January 1, 2018 to December 31,
2021. This paper employs Balcilar et al. (2021)’s TVP-VAR extended joint connectedness approach, which
Keywords:
Commodity futures market
combines the TVP-VAR connectedness approach of Antonakakis et al. (2020) with the joint spillover approach of
COVID-19 pandemic Lastrapes and Wiesen (2021), to investigate the dynamic connectedness among the select commodity futures of
TVP-VAR interest. Our findings show that system-wide dynamic connectedness varies over time and is driven by economic
Dynamic connectedness events. The pandemic shocks appear to have an impact on system-wide dynamic connectedness, which peaks
Joint connectedness during the COVID-19 pandemic. Crude oil and zinc are the primary net shock transmitters, whereas gold and
silver are the primary net shock receivers. We also discovered that the role of aluminum in shock transmitters
and shock receivers changed during the course of the investigation. Pairwise connectivity, on the other hand,
shows that Zinc, Copper, Nickel, and Crude oil are the key drivers of gold price changes, explaining the network’s
high degree of interconnectivity. During the study period, it was also discovered that silver has a significant
influence on gold. Furthermore, in comparison to natural gas, gold’s spillover activity is still relatively modest
(on a scale), indicating that gold is less sensitive to market innovations.

1. Introduction Büyüksahin and Harris, 2011; Silvennoinen and Thorp, 2016; Delatte
and Lopez, 2013; Naeem et al., 2022a). Investors and portfolio managers
An understanding of the interdependencies and spillovers between have found it increasingly challenging to achieve portfolio diversifica­
different asset types is necessary. The importance of this knowledge is tion. Understanding how returns spill over from one market to another is
becoming increasingly critical to the success of portfolio management becoming increasingly important.
because of the increasing regulatory and fundamental convergence Liquidity in commodity markets has increased dramatically over the
across financial markets, which is driving out the diversification po­ past couple of decades. Sharp increases in non-agricultural commodity
tential for investors and increasing financial market volatility. prices in recent years have drawn the attention of academics in studying
Commodity markets’ greater financialization has lowered the nega­ the drivers of non-agricultural commodity prices. Global financial and
tive impact of their physical characteristics, such as decreased liquidity economic crises have made commodities particularly vulnerable to price
and a higher cost of capital reallocation, on investment and hedging fluctuations that can go either way, including ups and downs. Energy
demands (Hong and Yogo, 2012). Literature on the subject of com­ commodity prices, for example, are vulnerable to enormous swings and
modities markets’ diversification value is varied (Greer, 2000; severe volatility because of the high macroeconomic uncertainty,

* Corresponding author.
E-mail addresses: aswini@[Link], aswinimishra1@[Link] (A.K. Mishra), arunachalam@[Link] (V. Arunachalam), dolson_prof@yahoo.
com (D. Olson), marikesh@[Link] (D. Patnaik).

[Link]
Received 10 August 2022; Received in revised form 27 December 2022; Accepted 13 March 2023
Available online 20 March 2023
0301-4207/© 2023 Elsevier Ltd. All rights reserved.
A.K. Mishra et al. Resources Policy 82 (2023) 103490

economic and energy crises, political conflicts, and legislation aimed at events such as the recent COVID-19 pandemic crisis. In our findings, we
mitigating dangerous climate change (Mensi et al., 2014; Rahman et al., demonstrate that there are influences of a shock in a typical market on
2022). There has been a rise in demand for precious metals like gold as a the entire network, implying that in the process of managing the in­
result of inflation fears (Bilgin et al., 2018; O’Connor et al., 2015) and vestment portfolio, investors and managers would be more careful, and
financial market downturns (Baur and Lucey, 2010; Baur and McDer­ he/she would realize that there are contagions of uncertain and risk as
mott, 2010). Even when macroeconomic shocks push commodities and an early warning signal to consider the investment strategy.
investors’ portfolio returns in different ways, commodity futures have The remaining sections are organized as follows. Section 2 provides a
divergent returns and volatility with low linkage to financial markets comprehensive overview of the pertinent literature. Section 3 provides a
(Kang et al., 2017). Commodities have taken on a greater role in port­ full discussion of the data and technique utilised. The empirical findings
folio allocation since the stock market collapse of 2002 (Agieri and are discussed in Section 4, while the study’s conclusion and scope for
Leccadito D, 2017; Creti et al., 2013). An increase in institutional in­ further research are presented in Section 5.
vestors’ purchases of commodity index-related instruments This begs
the question of whether or not financialization-induced over-speculation 2. Review of literature
drives commodity prices up or down (Basak and Pavlova, 2016). As a
result of the volatility in the commodities markets, policymakers pay This section examines the pertinent literature on the interconnec­
special attention to inflationary pressures. All economies benefit greatly tedness of a variety of financial markets as well as the interconnected­
from a tight connection between the pricing of goods in the future. ness of non-agricultural commodities. In light of the interconnectedness
Commodities are now seen as a way to diversify a portfolio’s risk, thanks of a vast array of financial markets, there is a multitude of research
to their financialization (Choi and Hammoudeh, 2010; Tang and Xiong, leveraging interconnectedness to measure contemporary financial risk
2012; Vivian and Wohar, 2012). and its management.
Needless to mention, global financial markets have suffered enor­ Using spillovers as a measure of the market’s interconnectedness has
mous losses and dramatic upheavals since the 2008 financial crisis been increasingly popular in recent years. Systemic and systematic
(Cembalest, 2020) and recently COVID-19 has affected financial markets dangers may be affected by a person’s ability to stay connected (Andries
worldwide. Numerous empirical studies show that COVID-19 has and Galasan, 2020). Some examples of systemic/deterministic risk
influenced financial market interconnectivity. include economic cycles and wars as well as natural calamities and
As the COVID-19 pandemic has caused disruptions to global financial regime shifts (also known as market risk). Commodity futures market
markets, several researchers have documented the evolution of the shocks, credit markets, asset markets, repurchase agreements and
agricultural commodity markets during this period. This research con­ money market funds have all been subjected to systemic risk in recent
tributes to the current literature on the interconnectedness of com­ years. Systemic risk, on the other hand, is primarily caused by macro­
modity futures markets in India by analyzing the effects of the COVID-19 economic factors and is more concerned with the danger of systemic
pandemic on the dynamic interconnectedness and transmission mech­ contagion and system collapse. During a financial crisis, “systemic risk”
anism of shocks of three types of commodity futures returns namely; refers to the spread of bankruptcy, illiquidity, and losses throughout the
energy (Brent crude oil and natural gas), precious metals (gold and sil­ entire financial system.
ver) and industrial metals (aluminum, copper, zinc, lead, and nickel), It is the degree of interdependence and interconnectedness between
based on daily dataset obtained from Multi-Commodity Exchange of system components that is referred to as connectedness (Maggi et al.,
India Limited (MCX) between January 1, 2018 and December 31, 2021. 2020). According to Torrente and Uberti (2021), a portfolio’s assets
These three types of commodity futures are important investing tools grow more interdependent as the complexity of their relationships in­
and alternatives for commodity investors and financial investors alike creases. During the Great Recession of 2007–2008, systemic risk dras­
(Daskalaki and Skiadopoulos, 2011). tically rose. Therefore, the connections between variables in a network
To examine the interconnectedness of commodity futures markets in are an effective method for determining whether or not a network has
India, this study employs Balcilar et al. (2021)’s TVP-VAR extended joint systemic risk. Financial interconnectedness is a potential source of sys­
connectedness technique, which combines Antonakakis et al. (2020)’s temic risk, according to Maggi et al. (2020). Systemic risk could be
TVP-VAR connectedness approach with the joint spillover approach of identified using a phrase such as “connectedness.” As a broad term, it
Lastrapes and Wiesen (2021). This combination has numerous advan­ encompasses a vast array of financial risk metrics. Consider market risk
tages: (i) there is no need to choose an arbitrary window size, (ii) no research (such as Belsley et al., 2005; Maggi et al., 2020) and systemic
observations are lost, (iii) it is less susceptible to outliers, (iv) it adapts risk in general, which incorporates a variety of connectivity metrics
better to parameter changes, and (v) the generalized forecast error from empirical and theoretical literature in quantitative and qualitative
variance decomposition (GFEVD) is calculated more precisely. The studies (system wide connectedness, see e.g., Acemoglu et al., 2015;
network approach allows us to capture not only the spillovers between Acharya et al., 2012; Billio et al., 2012; Figini et al., 2020; Mishra et al.,
each commodity’s returns but also the cross-market spillovers between 2022a; 2022b). According to these studies, financial risk is influenced by
returns. the speculative nature of assets such as commodity futures, which are
Our findings show that system-wide dynamic connectedness is regarded as profitable but highly volatile. In this paradigm, it is neces­
affected by economic events. The pandemic shocks appear to alter sary to investigate these networks for potential systemic risk sources and
system-wide dynamic connectedness, which peaks during the COVID-19 alter investment portfolios accordingly. Reducing asset and portfolio
pandemic. Crude Oil and Zinc are the principal net transmitters of allocation volatility can reduce an investor’s portfolio’s downside risk
shocks, while Gold and Silver are the principal net receivers of shocks. by as much as one-third (Silvennoinen and Thorp, 2016; Mensi et al.,
Over time, aluminum’s role in shock transmitters and receivers changed. 2014; Antonakakis et al., 2018). Due to the connection between com­
Pairwise connectedness shows that Zinc, Copper, Nickel, and Crude oil modity market financialization and the degree of interconnection be­
influence gold price swings, which explains the network’s high degree of tween commodity markets, governments could benefit from investing in
interconnectivity. Silver has a large impact on gold during the study certain commodity markets (Kang et al., 2017).A recent study in the
period. In comparison to natural gas, gold’s spillover activity is still Indian context, investigates the causal relationships among spot and
relatively low, indicating that gold is less sensitive to advances in this futures prices of crude oil and gold for the Indian market (Mishra et al.,
market. Thus, our findings have important implications for investors and 2022c). However, this study doesn’t examine the dynamic connected­
policymakers, as well as practices from contagions across varied markets ness, which is quite relevant from risk management perspective.
and their policy interconnections. Our findings also highlight the Jana et al. (2022) use a series of econometric and machine learning
growing interconnections within unexpected and highly uncertain tools to analyze the influence of COVID-19 on the US equity market

2
A.K. Mishra et al. Resources Policy 82 (2023) 103490

during the first wave of Coronavirus. They have demonstrated that early with an extended joint connectedness technique is used by the author.
in the outbreak, worldwide COVID-19s worries impacted the US equity This study shows that pandemic shocks affect system-wide dynamic
market differently across sectors. Naeem et al. (2022b) conducted a new connectivity, which peaks during the COVID-19 pandemic. Net total
study that examined the return and volatility connection of the rising directional connectivity shows that cryptocurrencies and stocks play
green asset and the well-established US industry stock and commodity different roles depending on their internal qualities and external shocks.
markets from September 2010 to July 2021. They discover that the Bitcoin and Binance Coin receive shocks while Ethereum transmits. The
time-varying return and volatility connectivity have shown significant US stock market transmits shocks, while Hong Kong and Shanghai are
crisis leaps. net receivers. Pairwise connectivity shows that cryptocurrencies caused
Using multivariate GARCH models, the literature on spillovers in­ stock market volatility at the start of 2020 during the COVID-19
vestigates the transmission of returns and volatility among equity assets pandemic shock.
(Baele, 2005), oil and financial markets (Creti et al., 2013), agricultural
markets (Wright, 2011), gold and oil assets (Ewing and Malik, 2013), 3. Data and methodology
and energy and cereal prices (Mensi et al., 2014). There is little empir­
ical study on energy, metals, agriculture, and animals. Nazlioglu (2011) 3.1. Data description
investigated oil and agricultural commodities prices’ nonlinear causal­
ities. Vivian and Wohar (2012) explored the transmission of volatility Our data covers period from January 1, 2018 to December 31, 2021.
between energy, animal products, cereals, industrial metals, Additionally, we divide the entire sample into two subsamples: those
manufacturing inputs, and precious metals. Creti et al. (2013) applied collected before and during the COVID-19 epidemic. This division is
DCC-GARCH to analyze volatility transmission among commodity based on the declaration of the World Health Organization, which for
groupings (energy, precious metals, agricultural, non-ferrous metals, the first time officially revealed the existence of the 2019 coronavirus
food, oleaginous, exotic, cattle) (energy, precious metals, agricultural, pandemic (so-called COVID-19) on December 31, 2019. Consequently,
non-ferrous metals, food, oleaginous, exotic, livestock). Mensi et al. we classify the two time periods as pre Covid-19 (from January 1, 2018
(2014) evaluated the influence of OPEC press releases on the dynamic to December 31, 2019) and post Covid-19 (from January 1, 2020 to
spillovers of returns and volatility between the energy and grain com­ December 31, 2021). The data includes daily futures prices for Gold and
modity markets. DCC-GARCH is unable to detect sources of spillover. Silver from the category of precious metals; Copper, Lead, Zinc, Nickel
Using volatility and equity market uncertainty indices, Gorzgor et al. and Aluminium from the category of base metals; Natural Gas and Brent
(2016) evaluated commodity market volatility. Barbaglia et al. (2020) Crude Oil from energy sector which are obtained from Multi Commodity
investigated the spillovers of energy, agriculture, and biofuel volatility. Exchange of India Limited (MCX). We calculate the first log-differenced
Dahl et al. (2020) examined energy, grain, and cattle futures using series: yit = log (xit ) − log (xit− 1 ) which can be interpreted as the return
copula approach. Evaluated the spillover effects between oil and agri­ to select commodity futures. In terms of whole observations, all series in
cultural commodity return volatility. This technique disregards nonlin­ Table 1 are reported with a positive average return.
earity, asymmetry, and the time horizon. Tiwari et al. (2020) explored Natural gas and crude oil markets have the highest variance, making
multiscale causality and interactions among energy, food, agriculture, the two markets the riskiest assets throughout the sample periods, as
industry, and metals. Kang et al. (2017) applied the Diebold and Yilmaz indicated in Panel 1 A and Fig. 1 demonstrates the volatility of these
(2012) spillover index and bivariate dynamic equicorrelation GARCH series.
(DECO-GARCH) model to analyze the time-varying spillover between Furthermore, this research finds Natural gas market has the highest
the WTI crude oil, precious metal (gold and silver), and agricultural variance during pre-Covid period, whereas crude oil has witnessed the
(corn, rice, wheat) commodity futures markets. The authors discover highest variance during the Covid-19 study period. Additionally, all of
links between market performance and volatility spillovers. The global the series are leptokurtic, which means the distributions have fatter tails
financial crisis and the European debt crisis boost market spillovers and than a normal distribution. This confirms the Jarque and Bera (1980)
the costs of hedging. Agieri and Leccadito D (2017) investigated the normality test, which found that all assets are considerably non-
contagion of energy, food, and metals using delta Conditional normally distributed. All returns are steady at the 1% significance
Value-at-Risk (CoVaR). The authors identify market contagion risks level according to the ERS unit root test of Silvennoinen and Thorp
caused by financial variables in the energy and metal markets, as well as (2016). Finally, a weighted portmanteau test by Figini et al. (2020)
the financial and economic fundamentals of the food markets. validates our decision to describe the interconnectivity of series by
Yahya et al. (2022) conducted a study that provided insight on the adopting a TVP-VAR technique with a time-varying var­
energy-biodiesel/food argument by evaluating the interconnectedness iance–covariance structure to characterize the interconnectedness of the
of the crude oil market and the markets for biodiesel and rapeseed oil in series. Since the research aims to find the linkages between crypto­
Europe. They applied the DCC-Student-t copula to daily data from July currencies and the stock market, we examine the interconnectedness of
17, 2008 to April 17, 2020, and found that during bearish periods, the these two markets before and during the COVID-19 pandemic.
conditional connectedness between crude oil prices and biodiesel Panels 1 B and 1C highlight the primary statistics of two subsamples
(rapeseed oil) prices is stronger than during bullish periods, indicating a with the same statistics as Panel A. However, Table 1 emphasizes sta­
tendency for positive comovement to persist with a decline in crude oil tistically significant differences between the two periods for these
prices. Akyildirim et al. (2022) investigate the connection and direc­ included series. Surprisingly, a positive average return is reported for all
tional spillovers for agricultural commodities futures markets and commodity futures during the post-COVID-19 period. On the other
sentiment indices. Similarly, Mishra and Ghate (2022) examined return hand, a negative average return is reported for base metals like copper,
and volatility spillover dynamics amongst a network of basic metals in zinc, lead and aluminium and natural gas from the energy commodity
India from January 2011 to March 2020. Their analysis shows that base futures during the pre-COVID-19 period. In addition, with the exception
metal returns and volatility are interrelated and imply market risk. of lead and aluminum, other markets became more volatile after COVID-
A very recent study by Ha (2022) investigates the interconnections 19, since all variances increased. The results of the ERS unit root test and
between cryptocurrency and the stock market, as well as the effects of the weighted portmanteau test on these variables during these two pe­
uncertain events such as the COVID-19 health crisis on the dynamic riods are more likely to be identical to those obtained from tests on the
interconnections between these two markets. The TVP-VAR combined entire sample, leading us to conclude that modeling the

3
A.K. Mishra et al. Resources Policy 82 (2023) 103490

Table 1
Summary statistics.
Panel 1 A: Whole sample (1027 Observations)

Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil

Mean 0.049* 0.046 0.045 0.015 0.029 0.067 0.043 0.037 0.037
(0.071) (0.387) (0.237) (0.726) (0.536) (0.175) (0.300) (0.730) (0.727)
Variance 0.749*** 2.843*** 1.504*** 1.951*** 2.229*** 2.520*** 1.734*** 11.581*** 11.475***
Skewness − 0.680*** − 0.882*** − 0.122 1.663*** 0.087 − 0.105 0.368*** 0.407*** − 0.932***
(0.00) (0.00) (0.108) (0.00) (0.250) (0.169) (0.00) (0.00) (0.00)
[Link] 4.636*** 8.870*** 3.418*** 25.818*** 3.863*** 1.832*** 8.993*** 4.592*** 30.451***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
JB 998.721*** 3499.723*** 502.551*** 28997.706*** 639.722*** 145.453*** 3483.615*** 930.521*** 39828.367***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
ERS − 14.007*** − 15.229*** − 12.625*** − 10.707*** − 14.517*** − 3.002*** − 9.883*** − 15.914*** − 11.927***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.003) (0.00) (0.00) (0.00)
Q(20) 19.341** 27.323*** 10.652 12.531 7.749 9.391 18.945** 22.366*** 72.058***
(0.023) (0.001) (0.436) (0.266) (0.751) (0.572) (0.027) (0.006) (0.00)
Q2(20) 85.863*** 190.580*** 13.245 1.702 57.217*** 37.511*** 42.505*** 152.165*** 567.847***
(0.00) (0.00) (0.215) (1.000) (0.00) (0.00) (0.00) (0.00) (0.00)

Pearson Correlation Coeff. Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil

Gold 1.000*** 0.713*** 0.136*** 0.149*** 0.106*** 0.128*** 0.092*** − 0.023 0.028
Silver 0.713*** 1.000*** 0.222*** 0.104*** 0.168*** 0.209*** 0.126*** 0.017 0.146***
Copper 0.136*** 0.222*** 1.000*** 0.294*** 0.510*** 0.499*** 0.394*** 0.008 0.227***
Lead 0.149*** 0.104*** 0.294*** 1.000*** 0.384*** 0.269*** 0.212*** 0.027 0.091***
Zinc 0.106*** 0.168*** 0.510*** 0.384*** 1.000*** 0.412*** 0.291*** 0.028 0.157***
Nickel 0.128*** 0.209*** 0.499*** 0.269*** 0.412*** 1.000*** 0.352*** 0.059 0.166***
Aluminium 0.092*** 0.126*** 0.394*** 0.212*** 0.291*** 0.352*** 1.000*** 0.047 0.122***
Natural Gas − 0.023 0.017 0.008 0.027 0.028 0.059 0.047 1.000*** 0.083***
Brent Crude Oil 0.028 0.146*** 0.227*** 0.091*** 0.157*** 0.166*** 0.122*** 0.083*** 1.000***

Panel 1B: Pre Covid-19 pandemic (511 Observations)

Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil

Mean 0.057* 0.034 − 0.012 − 0.011 − 0.032 0.063 − 0.01 − 0.039 0.026
(0.053) (0.454) (0.811) (0.868) (0.600) (0.361) (0.875) (0.754) (0.780)
Variance 0.450*** 1.058*** 1.220*** 2.397*** 1.904*** 2.463*** 1.946*** 7.770*** 4.238***
Skewness 0.267** 0.105 0.558*** 2.809*** 0.505*** 0.379*** 1.066*** 0.188* 0.220**
(0.014) (0.325) 0.000 0.000 0.000 (0.001) 0.000 (0.080) (0.041)
[Link] 1.540*** 1.832*** 4.963*** 32.555*** 3.436*** 1.071*** 12.727*** 7.408*** 5.234***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
JB 56.556*** 72.413*** 551.000*** 23237.576*** 273.041*** 36.687*** 3545.491*** 1171.354*** 587.420***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
ERS − 9.142*** − 9.379*** − 9.578*** − 7.692*** − 8.993*** − 2.160** − 7.016*** − 9.608*** − 7.271***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.031) (0.00) (0.00) (0.00)
Q(20) 22.542*** 9.874 14.136 7.553 10.035 6.613 16.216* 21.977*** 18.370**
(0.006) (0.519) (0.162) (0.771) (0.501) (0.857) (0.079) (0.007) (0.034)
Q2(20) 54.792*** 46.511*** 3.813 0.951 7.023 8.218 25.605*** 166.290*** 8.368
(0.00) (0.00) (0.989) (1.000) (0.821) (0.702) (0.001) (0.00) (0.686)

Pearson Correlation Coeff. Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil

Gold 1.000*** 0.757*** 0.166*** 0.217*** 0.111** 0.213*** 0.129*** 0.027 0.124***
Silver 0.757*** 1.000*** 0.254*** 0.179*** 0.157*** 0.273*** 0.152*** − 0.014 0.149***
Copper 0.166*** 0.254*** 1.000*** 0.242*** 0.418*** 0.404*** 0.316*** − 0.039 0.281***
Lead 0.217*** 0.179*** 0.242*** 1.000*** 0.311*** 0.283*** 0.175*** − 0.011 0.093**
Zinc 0.111** 0.157*** 0.418*** 0.311*** 1.000*** 0.338*** 0.203*** − 0.011 0.254***
Nickel 0.213*** 0.273*** 0.404*** 0.283*** 0.338*** 1.000*** 0.341*** − 0.012 0.155***
Aluminium 0.129*** 0.152*** 0.316*** 0.175*** 0.203*** 0.341*** 1.000*** 0.043 0.108**
Natural Gas 0.027 − 0.014 − 0.039 − 0.011 − 0.011 − 0.012 0.043 1.000*** 0.106**
Brent Crude Oil 0.124*** 0.149*** 0.281*** 0.093** 0.254*** 0.155*** 0.108** 0.106** 1.000***

Panel 1C: Post Covid-19 pandemic (516 Observations)

Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil

Mean 0.04 0.057 0.102* 0.042 0.089 0.071 0.094* 0.111 0.048
(0.374) (0.548) (0.084) (0.442) (0.206) (0.315) (0.083) (0.519) (0.800)
Variance 1.047*** 4.616*** 1.781*** 1.513*** 2.548*** 2.582*** 1.522*** 15.365*** 18.664***
Skewness − 0.881*** − 0.873*** − 0.542*** − 0.653*** − 0.209* − 0.552*** − 0.600*** 0.436*** − 0.925***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
[Link] 3.936*** 5.745*** 2.550*** 4.791*** 3.981*** 2.520*** 2.914*** 2.932*** 21.805***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
JB 399.811*** 775.297*** 165.071*** 530.127*** 344.461*** 162.732*** 213.572*** 201.212*** 10295.804***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
ERS − 10.533*** − 11.220*** − 8.677*** − 4.371*** − 11.356*** − 1.419 − 11.069*** − 10.896*** − 9.171***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.157) (0.00) (0.00) (0.00)
Q(20) 13.4 19.994** 9.332 13.871 12.495 13.044 12.948 15.629* 48.328***
(continued on next page)

4
A.K. Mishra et al. Resources Policy 82 (2023) 103490

Table 1 (continued )

Panel 1C: Post Covid-19 pandemic (516 Observations)

Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil

(0.205) (0.017) (0.579) (0.177) (0.269) (0.229) (0.236) (0.097) (0.00)

Q2(20) 26.832*** 59.094*** 14.237 19.975** 52.208*** 45.019*** 23.185*** 38.298*** 272.639***
(0.001) (0.00) (0.157) (0.017) (0.00) (0.00) (0.004) (0.00) (0.00)

Pearson Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil
Correlation Coeff.

Gold 1.000*** 0.708*** 0.122*** 0.108** 0.105** 0.078 0.072 − 0.046 − 0.002
Silver 0.708*** 1.000*** 0.218*** 0.077 0.181*** 0.197*** 0.129*** 0.027 0.145***
Copper 0.122*** 0.218*** 1.000*** 0.359*** 0.574*** 0.579*** 0.473*** 0.034 0.217***
Lead 0.108** 0.077 0.359*** 1.000*** 0.475*** 0.257*** 0.265*** 0.062 0.107**
Zinc 0.105** 0.181*** 0.574*** 0.475*** 1.000*** 0.476*** 0.379*** 0.051 0.125***
Nickel 0.078 0.197*** 0.579*** 0.257*** 0.476*** 1.000*** 0.367*** 0.109** 0.185***
Aluminium 0.072 0.129*** 0.473*** 0.265*** 0.379*** 0.367*** 1.000*** 0.051 0.145***
Natural Gas − 0.046 0.027 0.034 0.062 0.051 0.109** 0.051 1.000*** 0.076
Brent Crude Oil − 0.002 0.145*** 0.217*** 0.107** 0.125*** 0.185*** 0.145*** 0.076 1.000***

Notes: (i) *Denote significance at 10% significance level; **Denote significance at 5% significance level and ***Denote significance at 1% significance level.
(ii) Skewness: D’Agostino (1970) test; Kurtosis: Anscombe and Glynn (1983) test; JB: Jarque and Bera (1980) normality test; ERS: Silvennoinen and Thorp (2016)
unit-root test; Q10) and Q2 (10): Figini et al. (2020) weighted portmanteau tests.

Fig. 1. Return series plot for commodities under study.

interconnectedness of the series using a TVP-VAR approach with a time- 3.2.1. Vector autoregression with time-varying parameters
varying variance-covariance structure is well supported. This section begins by describing the TVP-VAR connectivity
approach in conjunction with Diebold and Yilmaz’s (2012) original
3.2. Empirical methodology technique. Using the Bayesian information criterion (BIC), we estimate a
TVP-VAR model with a lag duration of order one:
Diebold and Yilmaz first established the most prominent economet­
yt = Bt yt− 1 + εt εt ∼ N(0, Σt ) (1)
ric technique for examining interlinkage (2012). Barunk and Kehlk
(2018) extended the literature on connectedness with their concept of
vec(Bt ) = vec(Bt− 1 ) + vt vt ∼ N(0, Rt ) (2)
frequency connectedness, which combines the Diebold and Yilmaz
(2012) framework with Stiassny’s spectral decomposition (1996). Uti­ where yt , yt− 1 and εt are K × 1 dimensional vector and Bt and Σt are K ×
lizing this technology, the researchers monitor contagions in a preset K dimensional matrices. vec (Bt ) and vt are K2 × 1 dimensional vectors
network in order to mitigate negative consequences caused by a specific
whereas, Rt is a K2 × K2 dimensional matrix. This model allows all pa­
economic shock. A shortcoming of the original method is its reliance on
rameters (Bt ) and hence the relationship across series to vary over time.
an arbitrarily selected rolling-window size for time-variable interlink­
Additionally, the variance-covariance matrices, Σt and Rt are also
ing. Consequently, a number of proposals have been made to address
varying over time Several prior research have demonstrated that the
this issue, including the use of mean squared prediction error to estab­
variances and covariances of financial markets change over time,
lish the appropriate window size (Antonakakis et al., 2020) and the
resulting in fluctuating market and investment risk.
application of the joint spillover index (Lastrapes and Wiesen, 2021). In
Subsequently, the TVP-VIMA model is written as follows: yt =
this research, we apply TVP-VAR in conjunction with an expanded joint ∑
h=0 Nh,t εt− 1 where N0 = IZ and εt denotes a symmetric white noise
connectedness technique, proposed by Balcilar et al. (2021).
shocks that Z × Z time-varying covariance matrix E(εt εt ) = Σt varies
′

with time. Therefore, the L-step forecast error is as follows:

5
A.K. Mishra et al. Resources Policy 82 (2023) 103490

∑
L− 1
φt (L) = yt+L − E(yt+L |yt , yt− 1 , …) = Nl,t εt+L− l − − − − − − − − − − − − − − (3)
l=0

where network spillovers with a higher degree have a greater value.

Lastly, the net pairwise directional spillovers can de represented as:
A matrix of forecast error covariance can be written as follows: gen,net
Xi,t = gSTij,t
gen,to
− gSTij,t
gen,from gen,net
. If Xij,t > 0, suggesting that series i

( )
(4)
′ ′
E φt (L)φt (L) = Nl,t Σt Nh,t , − − − − − − − − − − − − − − − − − − − − − − − − −

has a more considerable influence on series j.

The proposed framework relies on Pal and Mitra (2019) L-step ahead 3.2.2. Technique with an extended joint connectedness
generalized forecast error variance decomposition (GFEVD). The The gSTij,t and jSTij,t are assumed:
GFEVD, gSTij,t , represents an impact of a shock stemming from variable j
∑
Z
on variable i and can be written as follows: [Link]
Xi←•,t = jSTij,t , (11)
( ) [ ( )⃒ ]2 j=1,i∕
=j
E φ2i,t (L) − E φi,t (L) − E φi,t (L) ⃒εj,t+1 , …, εj,t+1
gen gen
φij,t (L) = φij,t (L) = ∑
Z
E(φ2it (L)) jnt,to
X•←i,t = jSTji,t , (12)
(5) j=1,i∕
=j

− 1(
L∑ )2
′
ei Nlt Σt ej 1∑ Z
[Link] 1∑ Z
jSIi = Xi←•,t = X jnt,to .
= l=0
( ) (6) z i=1 z i=1 i→•
( ′ ) L∑
− 1
′ ∑ ′
e Σt e j ⋅ ei Nlt Nlt ei
l=0 t We follow Lastrapes and Wiesen (2021) to generalize the scaling
approach, which the scaling factor η differs by each row as follows:
φgen
ij,t (L) [Link]
gSTij,t = ,− − − − − − − − − − − − − − − − − − − − − − − Xi←•,t
∑
L
ηi = − − − − − − − − − − − − − − − − − − − − − − − − − −
φgen
ij,t (L) gen,from
Xi←•,t
j=1

− − − − − − − −
(7) (13)

where ei denotes a Z × 1 zero selection vector that have a unity on its i th 1∑ Z

gen η= η − − − − − − − − − − − − − − − − − − − − − − − − − −
position and φij,t (L), (L), which represents a proportional reduction in z i=1 i
the variance of the prediction error of variable i as a result of condi­ − − − − − − − −
tioning on the future shocks of variable j. (14)
∑ gen
The Zj=1 φij,t (L) ∕
= 1 is normalized to unity, leading to the value of
Our η scaling and the one derived from the joint connectivity method
gSTij,t . We write this metric as follows:
are the same; the only difference is that our method offers greater
∑
Z flexibility because each row has its own scaling factor. Finally, the
gen,from
Xi←•,t = gSTij,t, (8) following actions must be coded:
Lastly, we can obtain:
j=1,i∕
=j

∑
Z
gen,to
Xi→•,t = gSTij,t. . (9) (1) jSTij,t − ηi gSTij,t ,
j jnt,from
j=1,i∕
=j (2) jSTii,t = 1 − XZ ,
gen,net jnt,to ∑Z
The net total directional connectedness is presented as: Xi,t = (3) Xi→•,t = j=1,j∕=i jSTij,t .
gen,to gen,net gen,net
Xi→•,t− Xi←•,t
gen. from
. If Xi,t < 0(Xi,t > 0), variable i implies a net
receiver (transmitter) of shocks. In other words, variable i is driven by (is Finally, allowing the scaling parameter to vary by row allows to
driving) other variables in the network. compute the net total and pairwise directional connectedness measures
The total connectedness index (TCI) demonstrates the interconnec­ as follows:
tedness within the network. We define the TCI as:

1∑ Z
1∑ Z
gSTt = [Link]
Xi←•,t = X gen,to − − − − − − − − − − − − − − − − − − − (10)
z i=1 z i=1 i→•,t

6
A.K. Mishra et al. Resources Policy 82 (2023) 103490

Xi,tjnt,net = Xi→•,t
jnt,to jnt,from
− Xi←•,t , (15) 4. Empirical results

jnt,net
Xij,t = gSTji,t − gSTij,t . (16) This section starts by providing the average TCI (total connectedness
index) values for a full set of data before displaying the pattern of the TCI
Even if the interpretations are same to the original connectivity across the studied time. By studying changes in the TCI’s pattern before
methodology, the outcomes are more accurate since the row sum and after the COVID-19 health crisis, we also examine the effects of this
normalization method flaws are solved (Caloia et al., 2019). unpredictable occurrence on the interlinkages among the considered
This technique is less sensitive to outliers, does not require an arbi­ markets. In order to gain a better understanding of not only the role of
trary window size, and detects parameter changes more precisely each commodity market (i.e., net transmitter or net receiver of shocks),
(Antonakakis et al., 2020a). In addition, Balcilar et al. (2021)’s work but also the bilateral interaction between commodities, we have pro­
have not only employed the joint connectedness approach proposed by vided results for net total connectedness and net pairwise connected­
Lastrapes and Wiesen (2021), which yields more accurate results ness. A similar strategy is also applied to two subsamples to demonstrate
because the normalization technique is theoretically derived as opposed the COVID-19 pandemic’s effects on the network. Both the explanation
to the standard normalization technique used by Diebold and Yilmaz and the implications are based on the study’s dynamic analysis, as it is
(2009, 2012, 2014), but they have also extended this framework. This the dynamic analysis that allows connectedness to be linked to specific
proposed paradigm allows for increased flexibility and even enables the events during the sample period and so makes it simpler to appreciate
calculation of net pairwise connection metrics, which was previously the evolution of connectedness of this network over time.
unachievable.

Table 2
Averaged joint connectedness.
Panel A: Whole Sample

Commodities Precious Metal Base Metal Energy

Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil FROM

Gold 42.84 41.62 3.81 2.56 2.23 2.83 1.94 0.54 1.61 57.16
Silver 39.16 41.15 5.16 1.51 3.31 4.52 1.86 0.63 2.71 58.85
Copper 2.22 3.66 54.62 4.99 12.28 10.87 7.25 0.40 3.71 45.38
Lead 2.18 1.46 5.34 73.72 8.44 4.10 2.78 0.67 1.30 26.28
Zinc 1.30 2.04 12.37 7.76 61.38 7.86 4.04 0.60 2.65 38.62
Nickel 1.72 3.09 11.31 3.84 7.92 63.48 5.42 0.97 2.25 36.52
Aluminium 1.07 1.34 7.61 2.79 4.32 5.91 74.87 0.84 1.25 25.13
Natural Gas 0.52 1.15 0.64 1.05 0.97 1.01 0.83 92.09 1.74 7.91
Brent crude 0.68 1.73 3.82 1.01 2.52 2.22 1.32 0.89 85.82 14.18
TO 48.85 56.09 50.06 25.52 42.00 39.32 25.44 5.53 17.22 310.03
Own 91.70 97.24 104.68 99.24 103.38 102.80 100.31 97.62 103.04 TCI
NET − 8.30 − 2.76 4.68 − 0.76 3.38 2.80 0.31 − 2.38 3.04 34.45
NPT 0.00 2.00 8.00 4.00 6.00 5.00 3.00 2.00 6.00

Panel B: Pre-COVID-19 pandemic

Commodities Precious Metal Base Metal Energy

Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil FROM

Gold 40.04 37.93 5.59 4.22 2.50 4.59 2.13 0.41 2.58 59.96
Silver 36.08 38.07 7.62 2.76 3.08 6.62 2.41 0.65 2.70 61.93
Copper 3.35 5.58 59.14 4.64 8.33 9.17 4.77 0.51 4.50 40.86
Lead 3.11 2.19 4.68 75.07 5.79 4.42 2.21 0.91 1.62 24.93
Zinc 1.63 2.29 8.52 5.74 68.77 6.76 2.30 0.28 3.71 31.23
Nickel 3.22 4.97 9.81 4.49 6.81 60.73 6.12 0.93 2.91 39.27
Aluminium 1.64 1.90 5.27 2.64 2.91 7.29 76.38 0.88 1.09 23.62
Natural Gas 0.34 0.64 0.70 0.89 0.86 0.52 0.72 93.48 1.85 6.52
Brent crude 1.83 1.93 4.53 1.44 3.72 2.63 1.19 0.86 81.87 18.13
TO 51.20 57.43 46.73 26.83 34.01 42.01 21.85 5.43 20.96 306.44
Own 91.23 95.50 105.87 101.90 102.78 102.74 98.22 98.91 102.83 TCI
NET − 8.77 − 4.50 5.87 1.90 2.78 2.74 − 1.78 − 1.09 2.83 34.05
NPT 0.00 1.00 8.00 4.00 7.00 3.00 3.00 5.00 5.00

Panel C: Post-COVID-19 pandemic

Commodities Precious Metal Base Metal Energy

Gold Silver Copper Lead Zinc Nickel Aluminium Natural Gas Brent Crude Oil FROM

Gold 43.41 41.94 2.54 1.40 2.33 2.38 2.14 1.91 1.94 56.59
Silver 39.92 41.72 3.61 0.80 4.43 3.82 1.53 0.97 3.19 58.28
Copper 1.34 2.69 46.68 6.21 15.73 14.88 9.40 0.37 2.70 53.32
Lead 1.74 1.34 6.96 68.79 11.48 4.21 3.78 0.55 1.14 31.21
Zinc 1.16 2.08 15.77 10.20 52.37 10.23 5.96 0.83 1.40 47.63
Nickel 0.84 2.39 15.39 3.72 10.52 58.23 5.23 1.64 2.05 41.77
Aluminium 0.66 0.95 9.63 3.78 5.92 5.24 71.68 0.86 1.28 28.32
Natural Gas 0.82 1.29 0.43 1.39 1.03 1.82 0.71 91.26 1.25 8.74
Brent crude 0.21 1.68 2.68 1.14 1.34 2.11 1.50 0.85 88.50 11.50
TO 46.69 54.36 57.02 28.65 52.79 44.68 30.25 7.99 14.94 337.37
Own 90.09 96.09 103.70 97.44 105.16 102.91 101.92 99.25 103.44 TCI
NET − 9.91 − 3.91 3.70 − 2.56 5.16 2.91 1.92 − 0.75 3.44 37.49
NPT 1.00 3.00 7.00 2.00 5.00 6.00 5.00 2.00 5.00

7
A.K. Mishra et al. Resources Policy 82 (2023) 103490

4.1. Averaged dynamic connectedness results transmitters effecting other markets over this sample time.
With the onset of the COVID-19, however, some changes are noticed.
Using the complete set of observations and subsets of observations With the exception of gold, silver, natural gas which remained net re­
based on the day the COVID-19 pandemic was first proclaimed, Table 2 ceivers while copper, zinc, nickel and Brent Crude oil remained the net
presents the average results for the interconnections of different markets transmitters, the responsibilities of two other markets changed. After the
within the network of varied markets. In this table, the diagonal element pandemic outbreak, aluminium transformed into net transmitters, while
indicates that the volatility of a given market is accounted for by its own lead became net receivers. With the results demonstrating the post-
shocks, while the off-diagonal elements describe the contribution of this COVID-19 situation, it is evident that three of base metal commodity
market to the volatility of others (FROM) and the contribution of others futures and Brent crude oil from the energy sector investigated assumed
to this market’s volatility (TO). Particularly, in Table 2, the rows indi­ the role as net transmitters during uncertain times. This revealed that
cate each market’s contribution to a given market’s forecast error investing in these commodities can be viewed as a safe move, implying
variance, while the columns represent the effect that a specific com­ that the volatility observed on the commodity futures market can be
modity futurehas on all other commodity futures individually. explained by the state of these specific Indian base metal and energy
Overall, the principal diagonal elements represent one’s personal futures market.
contribution, whereas off-diagonal elements represent contributions
“from” or “to” others. The rows of Table 2 indicate the contribution of 4.2. Dynamic total connectedness results
each individual commodity asset to the forecast error variance of a
specific commodity asset in the system, while the columns represent the It is important to note that the average results are typically used to
effect a specific commodity asset has on all other commodity assets summarize the underlying relationships. These conclusions cannot be
independently. applied to the investigation of a single occurrence or major shock, such
With an average TCI of 34.5 percent, the network of all markets is as the COVID-19 pandemic. Consequently, it is essential to apply the
able to account for 34.5 percent of the observed network developments. dynamic or time variant of total connectedness to evaluate market dy­
Consequently, the idiosyncratic effects can account for about 65 per cent namics and role changes over time. The necessity to explore the swap­
of the variance of the system’s forecast error. The average results in ping of the roles of net transmitter and net receiver is one of the cases
Table 2 also indicated that co-movement tends to occur among the proving the usefulness of this paradigm. Fig. 2 depicts the intertemporal
markets listed. In this system, both the precious metals (gold and silver) growth of the TCI.
are deemed to be the net recipient of shocks. On the other hand, lead The TCI fluctuates during the sample period. From 2018 through the
from the base metals and natural gas from energy futures are found to be first quarter of 2019, it was observed that the TCI values witnessed an
the net recipient of shocks. On average, gold has a greater tendency to be upward trend and reached the peak at approximately 45 per cent. The
influenced than to impact other markets. Four base metals namely; TCI values thereafter started decreasing till the end of the year 2019 to
copper, zinc, nickel and aluminium and Brent crude oil from the energy approximately 25 percent. However, when confronted with the COVID-
sector are the net shock transmitters due to their inclination to affect 19 pandemic, in the first quarter of 2020, the TCI peaked at approxi­
rather than be influenced by other markets. mately 40 per cent and thereafter, a declining trend observed later for
Taking into account the two subgroups (pre and post COVID-19 short period before we witness an upward trend till the end of 2021. The
pandemic) of observations, this research illuminates how a market increased value of TCI observed at the start of 2020 are in sync with
might play distinct roles within the specified time periods. Prior to the previous studies (Balcilar et al., 2021; Zhang and Broadstock, 2020).
occurrence of COVID-19, the network of all markets may still be held Notably, the increasing value of the TCI can be seen as spillovers be­
accountable for their advances and evolutions (TCI = 34.05 per cent). tween different markets. Therefore, regarding previous studies in which
This percentage climbed to 37.5 per cent when the COVID-19 pandemic holding other assets as safe havens were mentioned and scrutinized (Ji
began, indicating that the idiosyncratic impact may have been respon­ et al., 2020) and commodities peak due to the occurrence of the
sible for 62.5 per cent during the epidemic. These findings provide COVID-19 pandemic (Balcilar et al., 2021), the paper’s findings are
credence to the notion that comovement occurred frequently, particu­ supported, suggesting that the dynamic evolution of the TCI reported
larly during uncertain periods such as the COVID-19 pandemic. Prior to sensitivity when coping with uncertain situations or in encountering the
COVID-19, the average joint connectedness values of gold, silver, COVID-19 pandemic (Ha, 2022).
aluminium and natural gas indicated that they were shock receivers.
Consequently, copper, lead, zinc, nickel and Brent crude oil are the net

Fig. 2. Dynamic total connectedness

Note: Implementation of Balcilar et al. (2021)
methodology to set up the lead (20 leads) and lag
length (1 lag) order of forecast error variance
decomposition of in our TVP-VAR system. The
robustness tests were also performed by modifying
these values. Both the joint interlinkages (the black
shaded area) and the original interlinkages are dis­
played (the red line).

8
A.K. Mishra et al. Resources Policy 82 (2023) 103490

Fig. 3. Dynamic net total directional connectedness

Note: Same as Fig. 2.

Fig. 4. Time-variant of net total directional connectedness during the COVID-19 Pandemic
Note: Same as Fig. 2.

4.3. Net total and pairwise directional connectedness results which we examine pairs of commodities to see how their interaction has
changed over time in regard to these two potential roles. Fig. 3 depicts
The next step is to look at the network connectivity results. This level these observations. For each commodity under discussion, positive
of analysis will make it easier to classify diverse commodities as net numbers represent the net transmitting role and negative values repre­
transmitting and net receiving. This allows for the detection of potential sent the net receiving role.
role transfers between the two entities. In other words, depending on the Brent crude oil and copper tend to be the most persistent net trans­
time interval, any commodity in our network can act as a net transmitter mitters of spillover shocks in terms of net overall connection. With the
or net receiver of system shocks over time. We begin with net total exception of a few extremely brief time periods, zinc and nickel are the
connectivity, which determines if a commodity’s position in relation to same. Aluminium and lead, on the other hand, played both functions
all other commodities remains consistent across the examination period. throughout time, however their net transmitting role is clearly more
Furthermore, we offer our findings on pairwise net connectivity, in prominent, while times in which these assume a net receiving role do not

9
A.K. Mishra et al. Resources Policy 82 (2023) 103490

Fig. 5. Dynamic net pairwise directional connectedness

Note: Same as Fig. 2.

Fig. 6. Dynamic net pairwise directional connectedness during the COVID-19 Pandemic
Note: Same as Fig. 2.

appear to be significant. Furthermore, Crude Oil looks to have been a broadcasters, whereas gold and silver have continued to be long-term
significant net shock transmitter for the most of the 2018–2020 period. net receivers.
We’ve recently noticed (i.e., at the end of our sample) that Crude Oil is Next, we examine the net pairwise connectivity results shown in
once again becoming a large net transmitter. The network’s constant net Fig. 5. In an effort to determine the significance of gold within our
receivers, with the exception of brief intervals, are gold, silver, and network of commodities, we focus on the spillover effects associated
natural gas. with gold. The first thing to observe is that gold and all other com­
Fig. 4 depicts the connection during the period of uncertainty caused modities may eventually assume both roles as transmitters and
by the COVID-19 pandemic outbreak. In the early stages of 2020, natural recipients.
gas was the net transmitter, before reverting to becoming the net The above figure demonstrates that zinc, copper, nickel, and crude
receiver of shocks. The opposite is true for nickel and copper. Most oil are the key shock drivers for gold throughout the whole study period.
crucially, crude oil and zinc have continued to be long-term net This conclusively illustrates that these commodities markets are linked

10
A.K. Mishra et al. Resources Policy 82 (2023) 103490

and that dynamic portfolio management should be emphasized. During and other commodities, as proven by our study for certain time periods,
the majority of the study period, silver had the greatest impact (in terms suggests that policymakers should prioritize the decoupling of infor­
of scale and amplitude) on gold. However, net spillover activity for gold mation between these two markets when formulating regulations.
continues at relatively low levels (i.e., in terms of size) relative to natural Consequently, ensuring that policy actions do not produce unintended
gas, indicating that gold is typically less responsive to advances in this consequences. In order to prevent systemic repercussions, it is essential
market. to adopt policies that prevent negative contagion or overflow from one
Since the outbreak of Covid-19, zinc, natural gas, copper, nickel, commodity market to another. Our findings have significant implica­
crude oil and aluminium have continued to communicate shocks for the tions for investors constructing portfolio and asset allocation strategies,
gold futures market. Because of this, gold market volatility can be mitigating downside risk, and hedging.
explained by these commodities, with the most severe impact occurring Nonetheless, the study is limited to examining the case of com­
in early 2020. (Fig. 6). modities futures market interconnectedness in India. As a part of further
study, one may examine how did the COVID-19 pandemic affect the
5. Conclusions and scope for further research interconnectedness of commodities and financial markets. Similarly,
researchers may use Adekoya et al. (2022)’s asymmetric time-varying
Our study adds to the body of knowledge by estimating the time- vector autoregressive (TVP-VAR) connected approach to examine the
varying network connectedness across commodities futures markets in interconnectedness between these commodities as this technique out­
India, with a focus on the COVID-19 epidemic period. We can capture performs the traditional TVP-VAR model due to its capacity to identify
not just the spillovers between commodity returns, but also the cross- three types of spillovers, namely normal, positive, and negative
market spillovers between returns using the network technique. Using spillovers.
Balcilar et al. (2021) ‘s TVP-VAR extended joint connectedness
approach, we use a network connectedness approach to evaluate how CRediT author statement
closely commodity futures market in India are intertwined. This paper
uses daily returns data from January 1, 2018 to December 31, 2021, for Aswini Kumar Mishra: Supervision, Conceptualization, Methodol­
nine commodities futures (namely; Gold and Silver from the category of ogy, Investigation, Writing-review-editing; Vairam Arunachalam:
precious metals; Copper, Lead, Zinc, Nickel and Aluminium from the Formal-analysis, Investigation, Writing-original-draft; Dennis Olson:
category of base metals; Natural Gas and Brent Crude Oil from energy Formal-analysis, Investigation, Writing-original-draft, Visualization;
sector) traded at the Multi Commodity Exchange of India Limited and Debasis Patnaik: Investigation, Writing-original-draft.
(MCX).
From an investment and portfolio selection perspective, this di­
minishes the significance of precious and base metals and energy com­ Declaration of competing interest
modities during times of crisis. Despite the fact that system-wide
connection data provide some broader insight, it is vital to distinguish The authors declare that they have no known competing financial
between net-transmitter and net-receiver commodities in terms of their interests or personal relationships that could have appeared to influence
contribution to the system. The information identifying the net receiver the work reported in this paper.
and net recipient can be exploited by a number of investors to develop
optimal investment and risk management strategies. Data availability
Our findings indicate that these commodities markets are fairly
interrelated. However, while focused on the COVID-19 health issue, the The authors do not have permission to share data.
degree of interlinkage increases, as evidenced by the relatively high TCI
value (approximately 34.5 per cent in the full sample and 37.5 per cent References
in the post-COVID-19 period). We might conclude, based on these sta­
Acharya, V., Engle, R., Richardson, M., 2012. Capital shortfall: a new approach to
tistics, that our developed network is exposed to a significant level of ranking and regulating systemic risks. Am. Econ. Rev. 102 (3), 59–64.
market risk. Specifically, the COVID-19 epidemic inspired this investi­ Acemoglu, D., Ozdaglar, A., Tahbaz-Salehi, A., 2015. Systemic risk and stability in
gation to identify time-varying system-wide links. We discovered the financial networks. Am. Econ. Rev. 105 (2), 564–608.
Adekoya, O.B., Akinseye, A.B., Antonakakis, N., Chatziantoniou, I., Gabauer, D.,
strongest link between system-wide dynamic connectivity and the
Oliyide, J., 2022. Crude oil and Islamic sectoral stocks: Asymmetric TVP-VAR
COVID-19 outbreak. According to net total directional connectivity, the connectedness and investment strategies. Resour. Pol. 78 [Link]
role of each commodity is affected by both its internal and exterior resourpol.2022.102877.
characteristics. The location of aluminum changes from receiver to Agieri, B., Leccadito D, A., 2017. Assessing contagion risk from energy and non-energy
commodity markets. Energy Econ. 62, 312–322.
transmitter. Gold and silver continue to be regarded as the two net shock Akyildirim, E., Cepni, O., Pham, L., Uddin, G.S., 2022. How connected is the agricultural
recipients, while crude oil and zinc remain the two net shock trans­ commodity market to the news-based investor sentiment? Energy Econ. 113 https://
mitters. In addition, our work contributes to and strengthens the present [Link]/10.1016/[Link].2022.106174.
Andries, A.M., Galasan, E., 2020. Measuring financial contagion and spillover effects
body of knowledge (Liu et al., 2019; Pal and Mitra, 2019; Roman et al., with a state-dependent sensitivity value-at-risk model. Risks 8 (1), 5.
2020; Silvennoinen and Thorp, 2016; Umar et al., 2021; Balcilar et al., Antonakakis, N., Gabauer, D., Gupta, R., Plakandaras, V., 2018. Dynamic connectedness
2021) about interconnectedness in commodity markets and the impact of uncertainty across developed economies: A time-varying approach. Econ. Lett.
166, 63–75. [Link]
of shocks upon the economy. Anscombe, F.J., Glynn, W.J., 1983. Distribution of the kurtosis statistic b 2 for normal
While the overall system-wide connectivity analysis may reveal some samples. Biometrika 70 (1), 227–234.
patterns, the pairwise analysis enables us to identify patterns that may Antonakakis, N., Chatziantoniou, I., Gabauer, D., 2020. Refined measures of dynamic
connectedness based on time-varying parameter vector autoregressions. J. Risk
not be immediately apparent. Pairwise connectedness demonstrates that Financ. Manag. 13 (4), 84. [Link]
zinc, copper, nickel, and crude oil are the primary drivers of gold price Baele, L., 2005. Volatility spillover effects in European stock markets. J. Financ. Quant.
fluctuations. Given this, it is evident that various commodities markets Anal. 40, 373–401.
Balcilar, M., Gabauer, D., Umar, Z., 2021. Crude Oil futures contracts and commodity
are highly intertwined and that dynamic portfolio management is the
markets: new evidence from a TVP-VAR extended joint connectedness approach.
optimal strategy. Throughout the duration of the investigation, silver Resour. Pol. 73, 102219. [Link]
exerts a substantial influence on gold, according to another report. In Barbaglia, L., Croux, C., Wilms, I., 2020. Volatility spillovers in commodity markets: a
relation to natural gas, gold’s spillover activity is still relatively low (in large t-vector autoregressive approach. Energy Econ. 85.
Baruník, J., Křehlík, T., 2018. Measuring the frequency dynamics of financial
terms of scale), indicating that gold is less susceptible to innovations in connectedness and systemic risk. J. Financ. Econom. 16 (2), 271–296. [Link]
this market. The time-varying nature of the relationship between gold org/10.1093/jjfinec/nby001.

11
A.K. Mishra et al. Resources Policy 82 (2023) 103490

Basak, S., Pavlova, A., 2016. A model of financialization of commodities. J. Finance 71, Lastrapes, W.D., Wiesen, T.F., 2021. The joint spillover index. Econ. Modell. 94,
1511–1556. 681–691.
Baur, D.G., McDermott, T.K., 2010. Is GOLD a safe haven? International evidence. Liu, X., Pan, F., Yuan, L., Chen, Y., 2019. The dependence structure between crude oil
J. Bank. Finance 34 (8), 1886–1898. futures prices and Chinese agricultural commodity futures prices: measurement
Baur, D.G., Lucey, B.M., 2010. Is Gold a Hedge or a Safe Haven? An Analysis of Stocks, based on Markov-switching GRG copula. Energy 182, 999–1012.
Bonds and Gold. Finan. Rev. 45 (2), 217–229. [Link] Maggi, M., Torrente, M.-L., Uberti, P., 2020. Proper measures of connectedness. Ann.
6288.2010.00244.x. Finance 16 (4), 547–571.
Belsley, D.A., Kuh, E., Welsch, R.E., 2005. Regression Diagnostics: Identifying Influential Mensi, W., Hammoudeh, S., Nguyen, D.K., Yoan, S.M., 2014. Dynamic spillovers among
Data and Sources of Collinearity, 571. John Wiley & Sons. major energy and cereal commodity prices. Energy Econ. 43, 225–243.
Bilgin, M.H., Gogolin, F., Lau, M.C.K., Vigne, S.A., 2018. Time-variation in the Mishra, A.K., Agrawal, S., Patwa, J.A., 2022a. Return and volatility spillover between
relationship between white precious metals and inflation: A cross-country analysis. India and leading Asian and global equity markets: an empirical analysis. J. Econ.
J. Int. Finan. Mark. Inst. Money 56, 55–70. [Link] Finance Adm. Sci. 27 (54), 294–312.
intfin.2018.03.001. Mishra, A.K., Theertha, A., Amoncar, I.M. and R L, M. (2022b), "Equity market
Billio, M., Getmansky, M., Lo, A.W., Pelizzon, L., 2012. Econometric measures of integration in emerging economies: a network visualization approach", J. Econ. Stud.,
connectedness and systemic risk in the finance and insurance sectors. J. Financ. Vol. ahead-of-print No. ahead-of-print. [Link]
Econ. 104 (3), 535–559. Mishra, A.K., Ghate, K., 2022. Dynamic connectedness in non-ferrous commodity
Büyüksahin, B., Harris, J.H., 2011. Do speculators drive crude Oil futures prices. Energy markets: evidence from India using TVP-VAR and DCC-GARCH approaches. Resour.
J. 32 (2), 167–202. Pol. 76 [Link]
Caloia, F.G., Cipollini, A., Muzzioli, S., 2019. How do normalization schemes affect net Naeem, M.A., Karim, S., Uddin, G.S., Junttila, J., 2022a. Small fish in big ponds:
spillovers? A replication of the Diebold and Yilmaz (2012) study. Energy Econ. 84, connections of green finance assets to commodity and sectoral stock markets. Int.
104536. Rev. Financ. Anal. 83, 102283.
Cembalest, M., 2020. Eye on the Market. J.P. Morgan Asset Management, 20 March. Mishra, A.K., Ghate, K., Renganathan, J., Kennet, J.J., Rajderkar, N.P., 2022c. Rolling,
Choi, K., Hammoudeh, S., 2010. Volatility behavior of oil, industrial commodity, and recursive evolving and asymmetric causality between crude oil and gold prices:
stock markets in a regime-switching environment. Energy Pol. 38, 4388. evidence from an emerging market. Resources Policy 75, 102474.
Creti, A., Joëts, M., Mignon, V., 2013. On the Links between Stock and Commodity Naeem, M.A., Karim, S., Rabbani, M.R., Nepal, R., Uddin, G.S., 2022b. Market
Markets’ Volatility. Energy Economics. integration in the Australian National Electricity Market: fresh evidence from
D’Agostino, R.B., 1970. Transformation to normality of the null distribution of g1. asymmetric time-frequency connectedness. Energy Econ. 112, 106144.
Biometrika 679–681. Nazlioglu, S., 2011. World oil and agricultural commodity prices: evidence from
Dahl, R.E., Oglend, A., Yahya, M., 2020. Dynamics of volatility spillover in commodity nonlinear causality. Energy Pol. 39, 2935–2943.
markets: linking crude oil to agriculture. J. Commod. Market. 20, 100111. O’Connor, F.A., Lacey, B.M., Batten, J.A., Baur, D.G., 2015. The financial economics of
Daskalaki, C., Skiadopoulos, G., 2011. Should investors include commodities in their gold-a survey. Int. Rev. Financ. Anal. 41, 186–205.
portfolios after all? New evidence. J. Bank. Finance 35 (10), 2606–2626. Pal, D., Mitra, S.K., 2019. Correlation dynamics of crude oil with agricultural
Delatte, A.L., Lopez, C., 2013. Commodity and stock markets: some stylized facts from a commodities: a comparison between energy and food crops. Econ. Modell. 82,
copula approach. J. Bank. Finance 37 (12), 5346–5356. 453–466.
Diebold, F.X., Yilmaz, K., 2009. Measuring financial asset return and volatility spillovers, Rahman, M.L., Shahzad, S.J.H., Uddin, G.S., Dutta, A., 2022. Comparing the risk
with application to global equity markets. Econ. J. 119 (534), 158–171. spillover from oil and gas to investment grade and high-yield bonds through optimal
Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive directional copulas. Energy J. 43 (1).
measurement of volatility spillovers. Int. J. Forecast. 28 (1), 57–66. Roman, M., Górecka, A., Domagała, J., 2020. The linkages between crude oil and food
Diebold, F.X., Yilmaz, K., 2014. On the network topology of variance decompositions: prices. Energies 13 (24), 6545. [Link]
Measuring the connectedness of financial firms. In: Journal of Econometrics, Vol. Silvennoinen, A., Thorp, S., 2016. Crude oil and agricultural futures: an analysis of
182. Elsevier Ltd., pp. 119–134. [Link] correlation dynamics. J. Futures Mark. 36 (6), 522–544.
Ewing, B.T., Malik, F., 2013. Volatility transmission between gold and oil futures under Stiassny, A., 1996. A spectral decomposition for structural VAR models. Empir. Econ. 21
structural breaks. Int. Rev. Econ. Finance 25, 113–121. (4), 535–555. [Link]
Figini, S., Maggi, M., Uberti, P., 2020. The market rank indicator to detect financial Tang, K., Xiong, W., 2012. Index investment and the financialization of commodities.
distress. Econom. Stat. 14, 63–73. Financ. Anal. J. 68, 54–74.
Gorzgor, G., Lau, C.K.M., Biglin, M., 2016. Commodity markets volatility transmission: Torrente, M.-L., Uberti, P., 2021. Connectedness versus diversification: two sides of the
roles of risk perceptions and uncertainty in financial markets. J. Int. Financ. Mark. same coin. Math. Financ. Econ 1–17.
Inst. Money 44, 35–45. Tiwari, A.K., Nasreen, S., Shhbaz, M., Hammoudeh, S., 2020. Time-frequency Causality
Greer, R.J., 2000. The Nature of Commodity Index Returns. The Journal of Alternative and Connectedness between International Prices of Energy, Food, Industry,
Investments, pp. 45–52, 3(1). Agriculture and Metals. Energy Economics, 104529.
Ha, L.T., 2022. Interlinkages of cryptocurrency and stock markets during COVID-19 Umar, Z., Gubareva, M., Teplova, T., 2021. The impact of Covid-19 on commodity
pandemic by applying a TVP-VAR extended joint connected approach. J. Econ. Stud. markets volatility: analyzing time-frequency relations between commodity prices
[Link] ahead-of-print, No. ahead-of-print. and coronavirus panic levels. Resour. Pol. 73 (C), 102692 [Link]
Hong, H., Yogo, M., 2012. What does futures market interest tell us about the resourpol.2022.102692.
macroeconomy and asset prices? J. Financ. Econ. 105 (3), 473–490. Vivian, A., Wohar, M.E., 2012. Commodity volatility breaks. Journal of International
Jana, R.K., Ghosh, I., Jawadi, F., Uddin, G.S., Sousa, R.M., 2022. COVID-19 news and the Financial Institutions & Money 22, 395–422.
US equity market interactions: an inspection through econometric and machine Wright, B.D., 2011. The economics of grain price volatility. Appl. Econ. Perspect. Pol. 33,
learning lens. Ann. Oper. Res. 1–22. 32–58.
Jarque, C.M., Bera, A.K., 1980. Efficient tests for normality, homoscedasticity and serial Yahya, M., Dutta, A., Bouri, E., Wadström, C., Uddin, G.S., 2022. Dependence Structure
independence of regression residuals. Econ. Lett. 6 (3), 255–259. between the International Crude Oil Market and the European Markets of Biodiesel
Ji, Q., Zhang, D., Zhao, Y., 2020. Searching for safe-haven assets during the COVID-19 and Rapeseed Oil, 197. Renewable Energy, pp. 594–605.
pandemic. Int. Rev. Financ. Anal. 71, 101526. [Link] 10.1016/[Link].2020. Zhang, D., Broadstock, D.C., 2020. Global financial crisis and rising connectedness in the
101526. international commodity markets. Int. Rev. Finan. Anal. 68, 101239.
Kang, S.H., McIver, R., Yoon, S.-M., 2017. Dynamic spillover effects among crude oil,
precious metal, and agricultural commodity futures markets. Energy Econ. 62,
19–32.

12