Risk spillovers between crude oil and agricultural commodities in India: Insights from two major global disruptions

Abstract

This study examines the dynamics of the integration of and risk transmission between India’s crude oil price index and nine agricultural commodity spot prices. The study uses daily price volatility data from June 2015 to December 2023 and a time-varying parameter vector autoregressive framework to investigate short-term and long-term connectedness dynamics. The results show a strong relationship between crude oil and agricultural commodity prices in India. Soybean oil is a significant transmitter within the network, and crude oil emerges as a primary net recipient. The average total connectedness index (TCI) value was about 36 per cent, with 27 per cent in the short term and 8.9 per cent in the long term. The average TCI value during the Coronavirus disease (COVID-19) period was 32.6 per cent, and that during the Russia–Ukraine war period was 24.3 per cent, with volatility spillovers in the network occurring primarily in the short term. Dynamic variations in connectedness over time highlight the impact of specific events, such as the COVID-19 pandemic and the Russia–Ukraine war. This study identifies the net transmitters and net receivers of volatility shocks within a network, providing insights that can help with portfolio diversification and strategic decision-making.

1. Introduction

Commodities play a crucial role in a country’s economy, mainly due to their geopolitical and financial significance. Oil is the primary energy source that influences the price movement of other commodities. Studies indicate that if crude oil prices remain elevated for an extended period, the current surge in food commodity prices will likely persist for a significantly longer duration compared with previous periods of boom (Baffes, 2007). Oil prices tend to affect the prices of agricultural commodities because farmers use significant amounts of oil in the form of fertilizers and pesticides to improve crop yields and, ultimately, income. Oil is also essential in transportation, and when oil prices increase, prices for the consumer also increase. Oil and agricultural commodity prices increased sharply from 2006 to mid-2008. Studies show that the food crisis during this period was due to rising oil prices (Du, Yu and Hayes, 2011). When oil prices increase, people switch to biofuels made from maize (or corn) and soybeans, leading to rising prices of other agricultural commodities (Chang and Su, 2010). Another school of thought suggests that increased global economic activity accounts for higher agricultural commodity prices (Radetzki, 2006; Gilbert, 2010). For instance, during 2003–2008, the booming Asian giants India and China were responsible for the third commodity boom (Radetzki, 2006). India and China used more raw materials than developed economies, increasing demand and leading to higher crude oil prices.

The increase in the financialization of energy and commodity markets offers a variety of trade-offs for investors and consumers. The connectedness of different markets plays a vital role in managing risk and investment strategies. Investors and other stakeholders always look to protect their investments, and a common way is to diversify portfolios across different markets. However, the analysis of connectedness between different markets shows that shocks from one market can spill over to the other (Guhathakurta, Dash and Maitra, 2020; Guo and Tanaka, 2022; Farid et al., 2022; Mishra and Ghate, 2022; Mishra, Agrawal and Patwa, 2022; Mishra, Renganathan and Gupta, 2024). The interest in crude oil and commodity prices is not new, but it has increased in recent years due to the potential for diversification and other benefits. In addition, the spillover between crude oil and agricultural commodities has gained attention in the context of risk management (Ji et al., 2018; Tiwari et al., 2022). The relationship between crude oil and agricultural commodities is complex and significantly impacts the global economy. The relationship is more complicated and unpredictable in times of crisis, such as the global financial crisis of 2008 or the Coronavirus disease (COVID-19) pandemic. The COVID-19 pandemic restricted domestic and international demand, leading to a significant decrease in the market for crude oil from the transportation sector (Ji and Chu, 2020).

Regarding the financialization of agricultural commodities in India, the central government has taken many initiatives, including launching the National Agriculture Market (eNAM) in 2016. It is a pan-India electronic trading portal that connects the existing Agriculture Produce Market Committee (APMC) mandis¹ to form a single national market for agricultural commodities. The e-NAM platform provides farmers with better marketing opportunities to sell their produce through an online competitive and transparent price discovery system and an online payment facility. The e-NAM portal provides access to all APMC-related information and services. This includes commodity arrivals, quality and price information, buy and sell offers, and e-payment settlement directly into farmers’ accounts, among other services.

Geopolitical risks affect agricultural commodity prices by influencing supply and demand in the spot market. These events can also reshape market participants’ expectations, creating behavioural biases that influence hedging and speculation decisions (Goyal, Mensah and Steinbach, 2024). Two events in the past couple of years have significantly affected crude oil markets: the COVID-19 pandemic and the Russia–Ukraine war. The COVID-19 pandemic caused significant uncertainty and financial stress in industries, including travel, tourism, and hospitality, as well as in supply chains, consumption, production, and prices of products (Chang, McAleer and Wong, 2020). Several studies have examined the connectedness between crude oil and commodities during the COVID-19 pandemic (Sun et al., 2021). The decline in demand for crude oil negatively impacts commodity prices due to the COVID-19 pandemic and continues to be low compared with historical levels and financial assets.² In addition, commodity markets have become more interconnected with other financial markets because of globalization, financial liberalization, trade integration, and the financialization of commodities. The financialization of commodities has made them more liquid and accessible to trade, attracted many investors, and made them vulnerable to speculation.³ A high degree of connectedness between crude oil prices and commodity prices in India and increasing volatility and spillovers between commodity markets pose significant challenges for investors and policymakers (Pal and Mitra, 2019; Mishra, Agrawal and Patwa, 2022; Shahani and Taneja, 2022; Mishra et al., 2023; Anand and Mishra, 2023; Mishra, Anand and Venkatasai Kappagantula, 2025). Recently, Goyal, Mensah and Steinbach (2024) used the time-varying parameter vector autoregressive model to assess the impact of geopolitical risk and shocks on agricultural commodity markets, accounting for demand (exports), supply (input prices), inventory, speculation, and economic fluctuations. The results show that geopolitical risks significantly affect corn and soybean futures prices and market behaviours in the short to medium term. Oil and agricultural commodity prices rose during the Russia–Ukraine war due to geopolitical risk.

The Russia–Ukraine war added further stress to the markets. Russia is the world’s leading natural gas exporter, the second-largest oil exporter, and the top wheat exporter. Two primary ways the war has affected commodity markets are through the physical destruction of productive capacity and the sanction-induced disruption of trade and production. Energy and food prices have risen significantly, raising concerns about energy and food security. Food prices soared to levels comparable with or even higher than those during the 2007–8 period due to the financialization of derivative markets (Tang and Xiong, 2012; Huchet and Fam, 2016). Concerns regarding the immediate and long-term effects of the Russia–Ukraine war on the production and trade of commodities, particularly those in which Russia and Ukraine are significant participants, have been the primary cause of the increased volatility in commodity prices since February 2022. This war shows how geopolitical shocks affect global commodity markets. Russia invaded Ukraine, a major agricultural producer and exporter, disrupting its commodity trade (Ahn, Kim and Steinbach, 2023; Steinbach, 2023). Such conflicts demonstrate how unexpected geopolitical events can disrupt commodity supply chains, causing price fluctuations across nations (Zhang et al., 2023; Steinbach and Yildirim, 2024). Thus, the importance of agricultural commodities increased significantly during the Russia–Ukraine war (Khalfaoui et al., 2023). According to Just and Echaust (2022), wheat, rice, and maize account for more than half of the world’s total plant-based food energy. The aforementioned commodities contribute to global food security.

Although numerous studies have examined crude oil and commodity markets in the global market and the western Arab region, little is known about the return connectedness or volatility spillover between India’s agricultural commodities and crude oil markets. Many studies investigate the connectedness or volatility spillover between India’s agricultural commodities and crude oil markets in global events like the COVID-19 pandemic and the Russia–Ukraine war.

India is a net importer of crude oil and relies heavily on imports to meet its energy needs. Russia has emerged as a significant supplier of discounted crude oil to India, especially after western sanctions were imposed on Russia following the Ukraine invasion. This has been a notable shift in India’s crude oil import patterns, as the Middle East was traditionally the primary source of oil imports. The availability of discounted crude oil from Russia has allowed India to mitigate some of the economic impact of global oil price surges. India’s agricultural sector is mainly domestic, with the nine commodities mentioned (bajra, barley, cotton, turmeric, coriander, jeera, soybean, soy oil, and mustard) being produced domestically and exported. However, oil prices significantly influence agricultural prices indirectly through transportation, fertilizer, and irrigation costs, which are energy-intensive.

Lower crude oil costs could reduce input costs for farmers, such as diesel for irrigation and transportation. This might alleviate upward pressure on agricultural prices. Given India’s reliance on energy-intensive fertilizer imports, cheaper crude oil could indirectly lower fertilizer production costs. This would positively affect agricultural economics. With reduced energy costs, India might be able to price its agricultural exports more competitively, which could be advantageous during global supply disruptions caused by the Russia–Ukraine war. The Russia–Ukraine crisis has caused volatility in global energy and commodity markets. India’s ability to secure discounted oil during this period has likely cushioned some inflationary pressures. However, the transmission of these effects to agricultural prices depends on factors like logistics efficiency, government subsidies, and global market conditions for agricultural exports.

Thus, the objective of this study is two-fold. The first is to investigate short-term and long-term spillovers of risks between crude oil prices and agricultural commodity prices. Specifically, we investigate the average, net total directional, and net pairwise directional connectivity to capture the subtle dynamics of the relationship between crude oil prices and agricultural commodity prices in India. The second is to closely examine the connectedness between commodities and crude oil during COVID-19 (11 March 2020 to 23 February 2022) and the effect of the Russia–Ukraine war (24 February 2022 to 30 December 2023). The study uses a time-varying parameter vector autoregressions (TVP-VAR) approach⁴ based on the Chatziantoniou, Gabauer and Gupta (2023) framework. Data come from India, including the Indian crude oil index⁵ (the crude oil price index) and nine highly traded agricultural commodities from June 2015 to December 2023. The nine agricultural commodities can be classified into four categories: cereals and pulses (barley and bajra), fibres (cotton), spices (turmeric, coriander, and cumin seeds or jeera), and oil and oil seeds (soybean, refined soybean oil, and mustard seed).

This study contributes to the existing literature by enhancing understanding of the dynamic connectedness between crude oil and agricultural commodities, particularly during significant global disruptions such as the COVID-19 pandemic and the Russia–Ukraine war. By employing the time-varying parameter vector autoregression (TVP-VAR) model and analysing both short- and long-term connectedness, the research builds on prior studies like those by Antonakakis, Chatziantoniou and Gabauer (2020) and Baruník and Křehlík (2018), adding depth to the exploration of risk contagion and integration within commodity markets. Unlike earlier studies that primarily rely on static averages, this study employs dynamic connectivity measures, uncovering nuanced, event-specific impacts of crude oil price fluctuations. Furthermore, introducing portfolio analysis based on connectedness indices offers a novel perspective for assessing hedging effectiveness (HE) and optimizing diversification. The comparison of minimum variance, minimum correlation, and minimum connectedness portfolios (MCoPs) advances portfolio management strategies, demonstrating the practical applications of connectedness measures. By showing that crude oil is often a net receiver of shocks and soybean oil, a significant transmitter, the research aligns with and extends previous findings, providing a more comprehensive framework for understanding commodity market behaviour and informing risk management and policy decisions.

The study reveals several key findings about the dynamic interactions between crude oil and agricultural commodities. The dynamic total connectedness measure shows consistent trends regardless of rolling window sizes, indicating robustness in the methodology. This implies that spillover effects between crude oil and commodities are systemic and not highly sensitive to temporal segmentation. Crude oil consistently emerges as a net receiver of shocks, particularly during crises such as the Russia–Ukraine war and COVID-19. In contrast, soybean oil is identified as a significant net transmitter. This suggests that crude oil acts as a buffer or absorber of market disruptions, while soybean oil is pivotal in propagating volatility across commodities. The minimum variance portfolio (MVP) outperforms other strategies in terms of HE, as indicated by Sharpe ratios (SRs). However, the MCoP demonstrates greater stability in diversification. This suggests that portfolio strategies informed by connectedness metrics can better enhance risk mitigation compared with traditional correlation-based approaches. Long-run connectedness exceeds short-run connectedness, indicating that commodities and crude oil do not react immediately to market shocks but show sustained integration over time. This finding reflects the slow diffusion of information and structural adjustments in these markets.

The above findings highlight the complex, time-sensitive nature of risk contagion in commodity markets. The study underscores the importance of dynamic modelling in capturing the evolving interdependencies among commodities and crude oil. Additionally, the results provide practical insights into constructing more resilient portfolios, suggesting that connectedness-based approaches can outperform traditional methods during economic instability. This has implications for policymakers and investors, emphasizing the need to consider short- and long-term dynamics in risk management strategies.

The rest of the paper is organized as follows. Section 2 describes the data used. Section 3 explains the methodology. Section 4 discusses the empirical findings, and Section 5 concludes the paper.

2. Data and descriptive statistics

This study investigates the connectedness between the Indian crude oil price index and the spot prices of nine agricultural commodities from 1 June 2015 to 31 December 2023. Further analyses of two specific subperiods also are made. These subperiods include COVID-19 (11 March 2020 to 23 February 2022) and the effect of the Russia–Ukraine war (24 February 2022 to 30 December 2023). The data for all crude oil indices and agricultural commodities are taken from the National Commodity and Derivatives Exchange Limited. In this study, we analyse nine agricultural commodities of importance in India’s markets, which are subcategorized as cereals and pulses (barley and bajra), fibres (cotton), spices (turmeric, coriander, cumin seeds [jeera]), and oil and oil seeds (soybean, refined soybean oil, and mustard seeds). Figure 1 shows the standardized prices of crude oil and the nine traded commodities in India from June 2015 to December 2023.

Table 1 displays the summary statistical information of the commodities. Among all the commodities, bajra and cumin seeds (jeera) exhibit the highest mean returns (0.015 per cent each), whereas coriander, soybean, and mustard have negative mean returns. Due to political risk and global volatility around price shocks, more significant fluctuations in the price of crude oil are expected.

Soybean exhibited a significantly more significant variance (1.93 per cent) than the other commodities, with mustard and bajra having variances of 1.673 per cent and 1.43 per cent, respectively, followed by coriander. Bajra not only has a substantial chance of earning a profit but also comes with considerable risk. The skewness and kurtosis values indicate that all the return series are asymmetrical. Among the ten commodities, three—cotton, soybean, and crude oil—exhibit negative skewness; the others have right-skewed distributions. Barley, jeera (cumin seeds), soybean, and crude oil have large kurtosis values. Furthermore, all the return series have a leptokurtic distribution characterized by a heavy and long tail, including extreme outlier values. The Jarque–Bera test indicates that the return series does not exhibit normality. The Pearson correlation test measures the association between the commodities presented in the heat map (Figure 2). The results indicate a mostly positive correlation among the commodities.

3. Methodology

This study’s risk spillover network model relies on the research of Chatziantoniou, Gabauer and Gupta (2023). Antonakakis, Chatziantoniou and Gabauer (2020) extend the variance decomposition of the time-varying parameter vector autoregression (TVP-VAR) model over the dynamic vector authoregression model (Diebold and Yılmaz, 2012, 2014). As a result, a framework is developed to analyse risk spillover networks over time by incorporating a Kalman filter and forgetting factors into estimating variance-covariance matrices. Thus, the extension overcomes the following shortcomings: (1) it can adeptly overcome the difficulty of selecting a random window size, (2) it can avert the loss of critical observations, (3) the outliers exert no influence, and (4) the parameters can be calculated with greater precision during substantial alterations. The TVP-VAR model with order P is represented as

In the above equation, |$y,{y_{t - 1}}$| and |${\varepsilon _t}$| are |$N \times 1$| dimensional vectors. |$\sum\nolimits_t {} $| and |${{\Phi}_{it}}$|⁠, |$i = 1, \ldots ,p$| are |$N \times N$|-dimensional matrices, the former representing the time-varying variance-covariance matrix and the latter representing the time-varying VAR coefficients. Koop, Pesaran and Potter (1996) and Pesaran and Shin (1998) proposed the concept of generalized forecast error variance decomposition (GFEVD), which is based on the Wold representation theorem. The GFEVD |$\left( {{{\tilde C}_{ijt}}\left( H \right)} \right)$| is based on the seminal work of Diebold and Yılmaz (2012, 2014) and is defined as the effect of a shock in variable ‘j’ on a variable ‘I’. It can be expressed as follows:

The equations mentioned above illustrate how variable ‘j’ contributes to the variance in variable ‘i’ prediction error for horizon H. As the sum of the rows of |${\tilde C_{ijt}}\left( H \right)$| is not 1, it needs to be normalized to derive |${\tilde \theta _{ijt}}$| for better analysis of risk spillover effects. Through normalization, the following identities are obtained in this paper: |$\mathop \sum \nolimits _{i = 1}^N{\tilde C_{ijt}}\left( H \right) = 1$| and |$\,\sum\nolimits_{j = 1}^N {\sum\nolimits_{i = 1}^N {{{\tilde C}_{ijt}}\left( H \right) = N} } $|⁠.

Next, we calculated all the connection indicators to create risk spillover indices. First, total directional connectedness (TO) measures how much a shock in the variable ‘i’ affects all other variables ‘j’:

On the other hand, the total directional connectedness (FROM) others indicates the extent to which a variable ‘i’ receives shocks from other variables ‘j’, which is measured as follows:

Based on the above two equations, the net total directional connectedness (NET) of variable ‘i’ is defined as

If the value of |${\rm{NE}}{{\rm{T}}_{it}}\left( H \right)$| > 0, variable ‘i’ is the net transmitter and can drive other variables’ risk levels. Conversely, if the value of |${\rm{NE}}{{\rm{T}}_{it}}\left( H \right)$| < 0, variable ‘i’ acts as a net receiver, and other variables determine risk spillover. The risk total spillover index was developed to measure the spillover level of the complete financial risk spillover network using the following equation:

In other words, |${\rm{TC}}{{\rm{I}}_t}\left( H \right)$| measures the average level of the network of connectedness, that is, how a change in one variable usually affects all the other variables. The higher the value, the greater the market risk, and vice versa.

Thus far, the abovementioned measures can assess only the overall risk spillover effect, not the specific risk spillover effect, in both the short and long term. Therefore, referring to Chatziantoniou, Gabauer and Gupta (2023), the following frequency response function is introduced: |${\Psi }\left( {{{\rm{e}}^{ - i\omega }}} \right) = \mathop \sum \limits _{h = 0}^\infty {{\rm{e}}^{ - i\omega h}}{{\Psi }_h}$|⁠. In the previous equation, |$i = \sqrt { - 1} $| and |$\omega $| denote the frequencies at which the spectral density of |${y_t}$| at frequency |$\omega $| is considered. The spectral density of |${y_t}$| at frequency |$\omega $| can be defined as the Fourier transform of |${\rm{TVP - VMA}}\left( \infty \right)$|⁠:

The frequency of GFEVD is calculated using the approach by Baruník and Křehlík (2018) and normalized using the equation below:

|${\tilde C_{ijt}}\left( \omega \right)$| denotes the portion of the spectrum of the |$i$| variable at a given frequency |$\omega $| attributable to shocks in the |$j$| variable.

The study summarizes all frequencies within a specific range to analyse the short- and long-term risk spillover effects, with |$d = \left( {a,b} \right):a,b \in \left( { - {\pi },{\pi }} \right),a \lt b$|⁠:

Finally, we compute the total directional connectedness to others (TO), from others (FROM), net total directional connectedness (NET), and total connectedness index (TCI) to assess risk spillover for a specific frequency range d,

3.1. Portfolio construction and back testing

Three methodologies are used to evaluate the impact of asymmetry dynamics on the effectiveness of asset class hedging in portfolio construction.

3.1.1. MVP.

This approach aims to construct a financial asset portfolio with minimum volatility across multiple variables (Markovitz, 1959).

where |${{\bf{\it{w}}}_{{{\Sigma }_t}}}$| is the k × 1 portfolio weight vector, I is the unit vector, and Σ is the TVP-VAR model’s conditional variance-covariance matrix.

3.1.2. Minimum correlation portfolio

The minimum correlation portfolio (MCP), developed by (Christoffersen et al., 2014), is a method similar to the MVP. Portfolio weights are derived by minimizing conditional correlations. The formula for computing the MCP is as follows:

where |${R_t}$| is the k × k dimensional conditional correlation matrix.

3.1.3. MCoP

The MCoP uses the TVP-VAR method, as described in the study by Broadstock, Chatziantoniou and Gabauer (2022). Assets that have no impact on other assets and are not affected by them receive more significant allocations in the portfolio. The following are the specifics:

where PCI is the pairwise connectedness index matrix.

The appendix section delineates the mathematical principles underlying the MCP approach, which prioritizes the minimization of asset correlations.

3.1.4. Portfolio back testing

We use the SR (Sharpe, 1994), Sortino ratio (StR) (Sortino and Price, 1994), and HE (Ederington, 1979) to evaluate the portfolio methods. The SR and StR are given as follows:

where |${r_{\rm{p}}}$| represents the portfolio returns on the assumption that there is no risk-free rate. The SR enables us to evaluate and compare portfolio strategies by considering their returns in relation to their level of risk. As a result, a higher SR

) leads to a greater portfolio return. In the StR, |${R_{\rm{p}}}$| and |${R_{\rm{f}}}$| represents portfolio and risk-free rate, while DR represents the downside risk, which quantifies the standard deviation of negative returns. It is an enhancement of the SR that assesses risk-adjusted performance by concentrating solely on downside risk, which better corresponds to investor apprehensions regarding losses rather than overall volatility. Similar to the SR, a higher StR will provide high risk-adjusted portfolio performance.

HE measures portfolio risk reduction. Using the HE test statistic (Antonakakis, Chatziantoniou and Gabauer, 2020), the metric information is evaluated. The study employs the one-dimensional HE to assess the degree of risk reduction among the assets in the portfolio. Edrington’s (1979) HE is the prevalent instrument used to demonstrate the decrease in hedged variance compared with unhedged portfolio variance. The study employs three hedging techniques: MVP, MCP, and MCoP. As the objective is not to assess the possible impacts of these strategies, the downward bias of the HE can be disregarded since we are focused solely on risk reduction, as indicated by the metrics. The following is the calculation for the metric:

The portfolio variance is denoted as |${\mathop{\rm var}} \left( {{r_{\rm{p}}}} \right)$|⁠, the variance of asset i is represented as |${\mathop{\rm var}} \left( {{r_i}} \right)$|⁠, and HE refers to the variance reduction in the unhedged position of asset i. The risk decreases as the HE indices increase and vice versa.

4. Results and discussion

4.1. Averaged dynamic connectedness

Table 2 shows the average results for the entire sample period without considering the dynamic impact of events at specific time points. The results in the table include values from the whole sample of observations, high-frequency or short-run values (in parentheses), and low-frequency or long-run values (in curly brackets). Throughout the sample period, the low-frequency results enable us to discern between the short- and long-term outcomes. To avoid doubt, the components in main diagonal in Table 2 represent shocks associated with specific variables, also known as idiosyncratic shocks. On the other hand, the off-diagonal components address how variables interact within that network. For example, suppose we concentrate on the diagonal entry under the column labelled ‘bajra.’ It shows that shocks originating from this commodity account for 81.69 per cent of the connectedness, with 67.02 per cent in the short term and 14.67 per cent in the long term. The remaining 18.31 per cent is explained by interactions among all the variables in the network.

The model indicates that the average TCI is 35.91 per cent, meaning that the market’s observed network accounts for this percentage; consequently, the impact of one commodity on the others in the network is 35.91 per cent; similarly, the idiosyncratic effects can be 73.77 per cent, i.e. the impact the commodities have on themselves. The average value of the TCI index is 35.91 per cent, with the short run accounting for 27.04 per cent of the value and the long run accounting for 8.87 per cent. According to the results, the network of variables is responsible for 35.91 per cent of the forecast error variance in this network of variables. Our finding is in line with the previous studies of Anand and Mishra (2024) and Goyal, Mensah and Steinbach (2024). The remaining 64.09 per cent is explained by factors represented by each variable’s unique component. Moreover, short-term developments—that is, shock transmissions—are the main drivers of connectedness (27.04 per cent).

Analysing specific agricultural commodity types, we find that soybean oil, on average, is the primary source of development transmission in this network of variables, contributing 9.2 per cent of the total. Soybean (4.62 per cent) and cumin seeds (jeera) (1.68 per cent) followed. The network processes data rapidly regarding frequency bands when considering soybean and soybean oil. Short-term values of 8.09 per cent and 2.3 per cent are the main drivers of connectedness for soybean and soybean oil, respectively. Table 2 shows three values for each entry in the format‒‒total, (short term), {long term}, or frequency bands. Notably, the network processes information faster for soybean oil because soybean oil’s connectedness to the network is predominantly influenced by short-term dynamics, with a value of 8.09 per cent. Similar trends are observed for soybean and cumin seeds (jeera). With respective shares of 2.3 and 0.9 per cent, soybean and cumin seeds (jeera) tend to transmit volatility spillovers in the network, mainly in the short term.

Shifting our attention to the net recipients, Table 2 makes it clear that, on average, the primary recipient in this network is crude oil, with an average of −5.06 per cent, followed by bajra, which averages −4.28 per cent. Regarding the frequency bands, crude oil is notably more affected in the short term, at −4.27 per cent, than in the long term. Similar trends are seen in bajra, which also primarily impacts the short term, with a value of −2.93 per cent. When considering the oil basket, we observe that soybean and soybean oil are the primary net transmitters. Our results support the previous findings of Dahl, Oglend and Yahya (2020). Mustard is observed to be the net receiver due to its short-term dynamics. In the long run, mustard is also the net transmitter of shocks. In the spice basket, cumin seeds (jeera) are the net transmitter of shocks; the rest are the net recipients. Coriander is more susceptible to short-term shocks (−1.73 per cent) than long-term shocks (−0.97 per cent). In the cereal basket, both barley and bajra are the net recipients of the shocks. Bajra is the second-highest receiver of shocks in the network. We also see that crude oil is the primary recipient of shocks among the network of commodities. This is explained in subsequent sections.

Table 3 shows the average results from 11 March 2020 to 23 February 2022 and analyses the dynamic impact of COVID-19. The average TCI value is 32.55 per cent, with the short run accounting for 25.84 per cent of the total value and the long run accounting for 6.71 per cent. Soybean oil is again the net transmitter among the network of variables, and crude oil is the net receiver of shocks. The short-term impact of the shock is more evident in the case of both the transmitter soybean oil and the receiver crude oil. Further observations revealed that barley from cereals and the pulse basket were also major transmitters of shocks in the short term. Considering the oil and oil seed basket, we see that soybean oil and soybean are the major transmitters of shocks in this network of variables. This is consistent with the results we found earlier in the section. In the period of tranquillity, crude oil becomes the net receiver of the shocks. The spice basket was second in line to become the net receiver of the shocks. Coriander and jeera were also the net receivers of shocks, with 5.89 and 2.74 per cent, respectively.

Table 4 covers the period from 24 February 2022 to 30 December 2023. This period is used to analyse the effect of the Russia–Ukraine war. Both countries are significant participants in the international commodities market. The war brought increasing volatility to the commodities market. When the TCI is analysed, the average value is 24.27 per cent, with the short run accounting for 17.26 per cent of the value and the long run accounting for 7.01 per cent. During this period, we observed that the oil and oil seed baskets were the major transmitters of shocks among the network variables. Soybean oil was the main net transmitter at 6.8 per cent, with 5.22 per cent representing short-term impacts. A plausible reason is that Russia is one of the largest exporters of crude oil. Cereals and pulse baskets were the main shock receivers. Bajra was the main net receiver of shocks at 3.45 per cent. The spices basket followed the cereal and pulses basket, with the turmeric market being the next net recipient of shocks at 2.43 per cent.

Although Tables 2, 3, and 4 contain information about the average connectedness behaviour, our empirical framework enables a more dynamic analysis, which can be extremely useful in understanding the underlying relationships. Simply put, focusing solely on average connectedness behaviour provides a limited understanding of the risk-contagion dynamics that influence how network variables interact. This strategy conceals the consequences of specific events during the study period. The average network plots presented in Figure 3 illustrate the various contagion effects, which aim to enhance comprehension of the overall levels of interconnectedness.

Figure 3 shows the Network Plot for each contagion event. The graph contains plots (a), (b), (c) and (d), which represent the case for the overall network, the pre-COVID period, the COVID-19 period and the Russian–Ukraine period, respectively. The individual network figure in each plot shows the overall period; the second one shows the short-term and the third shows the long-term. The lines’ thickness indicates the connectivity level between the two variables. At the same time, the size of the nodes represents the degree of interaction of the variable in the network. Concerning overall interconnectedness, jeera, soybeans, and soy oil function as transmitters, and the remaining items serve as receivers. The node size of soybean oil indicates that it is the primary transmitter in the network over the long term. The connectedness is relatively low in the short term, with turmeric having no role, whereas cotton acts as the transmitter in the network. In the medium term, mustard acts as the transmitter, soybean serves as the primary transmitter, and crude oil is a receiver more in the medium than in the short term. During the COVID-19 pandemic, significant interaction among the variables was lacking. However, the dynamics are entirely different in the context of the Russia–Ukraine war. In this scenario, jeera, soybean, soybean oil, and mustard act as transmitters, and cotton plays a significant role as the primary receiver in the network. However, in the short term, cotton acts as the transmitter and crude oil acts as the receiver. Similarly, in the medium term, cotton becomes the primary receiver. The partial interaction or relatively lower interaction between the variables during the pandemic is one of the essential features of the commodity market.

4.2. Total dynamic connectedness

Network connectedness, or the average effect of a shock in one variable’s price on the other, is measured by the TCI. Conversely, a higher value indicates greater market risk. Figure 4 displays the overall TCI evolution (black-shaded region), the short-run TCI evolution (1-5 days), and the long-run evolution (5 and more days).

Connectivity reflects the mutual movement of these commodities. These co-movement patterns show the dependence and risk spillover between commodities. As a result, the relationship between the crude oil and agricultural commodities markets has received much attention recently. Millions of people were thrust into poverty, resulting from the abrupt increase in commodity prices caused by the COVID-19 and 2008 financial crises (De Hoyos and Medvedev, 2011). One significant development that has had an enormous impact on the market is policy. Biodiesel (mainly derived from soybeans) and bioethanol (mainly from corn/maize) are considered technological alternatives to traditional fuels such as gasoline and diesel. A close market integration between the energy and agricultural markets is anticipated because the production of biofuels is highly dependent on the supply of agricultural commodities. Market integration will likely be the most significant factor in agricultural development (Tyner and Taheripour, 2008). Food prices will rise due to the anticipated increase in demand for these goods, brought about by the continuous rise in biofuel production. According to Avalos (2014), these two markets were historically independent until 2006, when ethanol use increased to the point where it began to affect global energy prices. Fuels have been blended with ethanol for environmental purposes. In 2003, India implemented fuel blending policies, introducing 5 per cent ethanol-blended gasoline. However, in the scenario of crude oil prices versus nine agricultural commodities over the sampled period, we observe values of overall TCI (i.e. black-shaded region) ranging from 20 per cent to more than 50 per cent. There were some spikes in late 2015 and early 2016. In comparison, the most significant increases occurred during this time and, more recently, during the COVID-19 era.

During 2019 and early 2020, both overall and short-term connectedness increased significantly. As a result, we can see that the energy and agricultural commodities markets are highly integrated (for a short time) during the shock period. Furthermore, the periods in our sample with the highest connectedness are most likely associated with world events that significantly impacted the global economy and the oil market. The global economy experienced a decrease in oil prices in 2014 due to an oversupply of oil in the international market and a slowdown in growth in China and other emerging economies. There is enough evidence to conclude that the COVID-19 pandemic significantly impacted the crude oil market in the early months of 2020 when the crisis peaked (Zhang and Hamori, 2021).

4.3. Net total directional connectedness

Net total directional connectedness (NTDC) measures the overall strength of the directional spillovers from one asset or market to another. It is calculated by taking the sum of all the positive directional spillovers from Asset A to Asset B and subtracting the sum of all the negative directional spillovers from Asset A to Asset B. A shaded area with positive values indicates that the corresponding crude oil type is a net transmitter of price shocks to the rest of the network. Negative values represent net recipients. NTDC can explain why commodity prices affect crude oil prices because it shows a solid directional connection between crude oil and other commodities. For example, a National Bureau of Economic Research study revealed that crude oil has a solid net directional connection to other commodities, such as soybean and zinc (Diebold, Liu and Yılmaz, 2017). NTDC helps understand the complex relationships between different asset markets. NTDC can be used to identify the markets that are most interconnected and therefore most vulnerable to shocks. Investors can use this information to make more informed investment decisions. The results are illustrated in Figure 5.

We found that crude oil was the net receiver of the shocks until late 2022 and, after that, was a net transmitter due to post-COVID-19 market recovery. As explained in the previous section, COVID-19 caused the market to become more interconnected. Our result is consistent with the findings of Jebabli, Kouaissah and Arouri (2022). The findings show that crude oil is a net recipient of volatility shocks in the system of volatility spillovers between stock and energy markets during the COVID-19 crisis. Another significant development that took place during that time was Russia’s attack on Ukraine. Russia is one of the major oil suppliers to Europe and Asia. The sanctions that the USA and Europe imposed on Russia sent shocks to the market. The positive transmission effect is due to the increase in the crude oil price. Early in 2015, we saw that crude oil was transmitting shocks, the impact of the surge in the price of crude oil after Russia invaded Crimea, China’s financial crisis, and Organization of the Petroleum Exporting Countries production cut. Since 2015, crude oil prices have consistently experienced shocks from commodity prices. In the post-COVID-19 era of 2020, prices were slightly above 0, resulting in shocks to other commodities, but when the markets settled, prices started to increase.

Among all the commodities, soybean and soybean oil were the net transmitters in the network both in the short and long run. Soybeans are a major feedstock for biodiesel production; hence, crude oil prices are one of the influencing factors in this case. An increase in soybean-based biodiesel production following an increase in oil prices has significantly affected global agricultural grain production and prices (Avalos, 2014). In cereal and pulse baskets (bajra and barley), bajra is the net transmitter. We observe that barley is the net receiver of volatility shocks during economic distress (Figure 5—late 2015 and 2019). Our results are consistent with the findings of Just and Echaust (2022). Barley showed significant price volatility during the period, which was generally unrelated to the basket shocks.

Similar trends were observed in cotton. Cotton has been a net receiver for a long time, but our finding of short-run connectedness suggests that cotton is a net transmitter of shocks in the post-COVID era. With respect to the spice basket, we see that in periods of economic uncertainty, such as in 2015, spices were the net transmitter. Coriander was a net transmitter in 2015, while cumin seeds (jeera) were a net transmitter until mid-2016. Turmeric maintained its almost neutral position. Turmeric can hedge the risk or altogether avoid it during tranquil periods. The oil and oil seed basket is the most significant because governments worldwide are trying to use carbon-neutral fuels and migrate to fuels other than fossil fuels. Soybean and soybean oil are the net transmitters of shocks in the basket, as they are among the main components in biofuel production. Net transmission increases during periods of shock and uncertainty. Finally, mustard is the net receiver of the shocks in the basket.

4.4. Net pairwise directional connectedness

Net pairwise directional connectedness (NPDC) measures the direction and magnitude of the spillover effects between two variables. The results are shown in Figure 6. The figure provides a more detailed picture of the evolution of connectedness over time and practically validates the aforementioned analysis. To be more precise, we use the variable that occurs first in the title to interpret the findings. A positive NPDC indicates that the first variable is a net transmitter of shocks to the second variable. A negative NPDC value suggests that the first variable is a net recipient of shocks from the second variable.

We observed that soybean oil was the net transmitter of shocks to crude oil for most of the study period. The result is consistent with the aforementioned findings, especially in periods of economic uncertainty. Soybean oil is strongly transmitted to crude oil in the short term (Jebabli, Kouaissah and Arouri, 2022). Among the commodities in cereal baskets, barley is the shock transmitter to bajra. Furthermore, the observation tells us that bajra receives shocks from mustard most of the time, but bajra is the short-run shock transmitter during a period of tranquillity—the effect of oil and oil seeds on cereals. Cereals are the net shock recipients for most of the period. The interplay of connectedness and the spillover of shock between spices and cereals tells us that spices are generally the transmitters of shocks to cereals. Barley has shown volatility for spices. Sometimes, it is the net recipient for the spices; other times, it is the net transmitter to the spices. Spices show minimal connectedness among themselves. Turmeric, cumin seeds (jeera), and coriander transferred shocks to each other throughout the study, with no specific substantial resemblance. We found that cotton was the net transmitter of shocks to crude oil in the short term from 2019 to 2020. The year of the pandemic saw a rise in crude oil prices and brought uncertainty to the market. Our previous discussion showed that soybean oil and soybeans also transmit shocks to cotton. Overall, cotton was the net transmitter of shocks to the cereals and pulses basket. Among cereals, bajra was the major recipient of shocks from cotton. Oil and oil seeds were the net transmitters of the shocks to the cotton plants.

Throughout the sample period, oil and oil seeds were the net transmitters of shocks, especially during uncertain periods. Soybean and soybean oil were the main transmitters in the commodity network. A plausible reason is that as crude oil prices increase, people begin to look for alternatives, and biofuels are the best available. Soybeans are the major component that helps produce biofuel, so their importance increases during periods of economic uncertainty. Among the oil and oil seed baskets, soybean oil is the net transmitter of shocks to mustard and soybeans. Crude oil has also received shocks from oil and oil seed baskets. The major transmitters are soybean oil and soybean. Crude oil is essential because most recent agricultural production requires fertilizers and pesticides, which require crude oil and its derivatives to operate. In addition, cotton continues to transmit moderate and severe shocks to crude oil throughout the period, such as toward the end of 2021–3. Finally, soybeans show a similar shock trend to cotton with crude oil. For most of the sample period, soybeans were the major shock transmitter, with some substantial and moderate shocks.

The direction and magnitude of pairwise connectedness are consistent with previous findings (Just and Echaust, 2022; Sun et al., 2021). As we wrap up this section, it is important to highlight that even though the crude oil market is highly interconnected, we also observe instances in which the network variables act as either transmitters or recipients of shocks. Portfolio diversification is potentially considered. Discoveries regarding long-term connectedness, particularly frequency bands, could be valuable to long-term investors. By thoroughly evaluating market events, these investors can construct portfolios based on entities that transmit longer-term shocks. On the other hand, information regarding short-term relationships could be more valuable for speculative purposes. This study can provide useful insights for investors seeking to enhance risk management and policymakers aiming to comprehensively understand trends in the commodities market.

4.5. Robustness check

We tested the robustness of the methodology using two other rolling window sizes (150 and 250 observations) as can be seen in Figure 7a and b). The results in these figures show that the dynamic total connectedness smoothing increases as the window length increases. The system’s dynamic spillover effect is resistant to window sizes since the fundamental trend remains.

We also performed the 5-year rolling window with a 1-year loading period as can be seen in Figure 8.

4.6. Portfolio implication analysis

This section offers a comparative analysis of portfolio diversification using three techniques: the MVP, MCP, and MCoP. The effectiveness of the approaches is evaluated using the SR to determine the best hedging technique for the given assets. Therefore, we will demonstrate whether the MCoP generated from the pairwise connectedness index in our study offers effective portfolio diversification. We evaluate the hedge effectiveness of each strategy. Table 5 displays the most valuable players (MVPs).

Table 5 can be interpreted as, for example, by investing in a one-dollar portfolio of 1 per cent or one cent in oil, one can reduce the risk or volatility by 98 per cent. The optimal portfolio allocation for the MVP consists of 9 per cent or 9 cents investment in bajra, 17 per cent or 17 cents in barley, 20 per cent or 20 cents in cotton, 14 per cent or 14 cents in turmeric, 3 per cent or 3 cents in coriander, 13 per cent or 13 cents in jeera, 2 per cent or 2 cents in soybean, 19 per cent or 19 cents in soy oil, 3 per cent or 3 cents in mustard, and 1 per cent or 1 cent in crude oil for a one-dollar portfolio. This allocation results in a significant reduction in volatility, with bajra experiencing an 87 per cent decrease, barley 73 per cent, cotton 71 per cent, turmeric 77 per cent, coriander 86 per cent, jeera 83 per cent, soybean 90 per cent, soy oil 78 per cent, mustard 89 per cent, and crude oil 98 per cent. It is important to emphasize that most of the investment should be allocated to soybean oil and barley. Notably, soy oil is the leading net transmitter, and barley is the leading net receiver, predominantly during periods of economic distress. Investing 1 per cent in crude oil can significantly lessen the persistence of volatility by 98 per cent. Crude oil is known to consistently receive spillover during crises. We contrast this with the metric that is more reliant on correlation in the subsequent analysis.

Table 6 shows the MCP for the assets. Constructing this portfolio involves allocating investments as follows: 16 per cent or 16 cents in bajra, 9 per cent or 9 cents in barley, 8 per cent or 8 cents in cotton, 11 per cent or 11 cents in turmeric, 9 per cent or 9 cents in coriander, 6 per cent or 6 cents in jeera, 7 per cent or 7 cents in soybean, 3 per cent or 3 cents in soy oil, 1 per cent or 1 cents in mustard, and 19 per cent or 19 cents in crude oil in a one-dollar portfolio. This allocation results in a reduction in volatility of 48 per cent for bajra, −6 per cent for barley, −16 per cent for cotton, 9 per cent for turmeric, 46 per cent for coriander, 31 per cent for jeera, 61 per cent for soybean, 11 per cent for soy oil, 55 per cent for mustard, and 92 per cent for crude oil. A negative HE indicates a rise in portfolio volatility. It is important to observe that the allocation of investments is significantly distinct from that of the MVP. Here, most of the investment is allocated to crude oil and bajra, with crude oil being the recipient of funds and bajra being the sender. Investing 19 per cent in crude oil can lessen the persistence of volatility by 92 per cent.

Table 7 shows the MCoP for the assets. This portfolio can be constructed by allocating the following percentages to different commodities: 14 per cent or 14 cents to bajra, 11 per cent or 11 cents to barley, 10 per cent or 10 cents to cotton, 11 per cent or 11 cents to turmeric, 10 per cent or 10 cents to coriander, 8 per cent or 8 cents to jeera, 8 per cent or 8 cents to soybean, 4 per cent or 4 cents to soy oil, 9 per cent or 9 cents to mustard, and 15 per cent or 15 cents to crude oil in the one-dollar portfolio. This allocation will result in a reduction in volatility of 64 per cent for bajra, 27 per cent for barley, 20 per cent for cotton, 37 per cent for turmeric, 63 per cent for coriander, 53 per cent for jeera, 73 per cent for soybean, 39 per cent for soy oil, 69 per cent for mustard, and 95 per cent for crude oil. There appears to be a slight variation in the investment levels between the minimum correlation and MCoPs. However, the latter offers a higher level of HE. Like the MCP, a significant investment is allocated to crude oil and bajra. Crude oil is a net receiver, and bajra is a net transmitter.

Table 8 presents the portfolio analysis outcomes for each technique. The average of the returns is not equal. Therefore, each portfolio strategy will yield varying levels of profit. Based on the SR and StR, it may be inferred that the MVP performs better than the minimum connectedness and minimum correlation-based portfolios. Based on the provided mean and standard deviation, we can conclude that the MCoP generally achieves greater optimality than the MCP and the MCP indicates negative HE for many commodities.

5. Conclusion

This study offers ground for further exploration of changes in the crude oil market and crude oil’s effects on commodity markets following the current Russia–Ukraine war (Salisu, Pierdzioch and Gupta, 2021). This study considered the common pool concept for commodities and crude oil types. The study also examined the extent of integration and the potential for risk contagion within a network of variables. Following Antonakakis, Chatziantoniou and Gabauer (2020), we applied the dynamic connectivity technique based on a TVP-VAR model. Simultaneously, we examined connectedness in the short and long runs to explore events that had a rather extended impact on the global crude oil market. In other words, following (Baruník and Křehlík, 2018), we assumed connectivity inside both a high-frequency band (i.e. 5 days) and a low-frequency band (i.e. from 5 to 100 days).

The findings from this study revealed that soybean oil proved to be a significant transmitter of shocks and crude oil emerged as the primary net recipient. However, relying solely on averages overlooks nuanced risk-contagion dynamics. Our dynamic analysis revealed the impact of specific events, providing a more comprehensive understanding of the above. Using net total directional connectedness, we could gauge the effects of crude oil fluctuations on other agricultural commodities. We were also able to mark specific historical impacts of crude oil price fluctuations, which resulted in crude oil causing shock pulses to other commodities. However, when widely examined, crude oil was considered a shock receiver for most periods (extended range).

Next, the study considered net directional and pairwise connected measures to determine the real interaction between agricultural commodities and crude oil. Every variety of commodity can function in the network as a net transmitting or net receiving entity. During the study’s sample period, singular incidents showed how various commodity types transitioned between roles. Conversely, greater ambiguity about how COVID-19 lockdowns affected the economy may be able to shed light on the prevalence of long-term connectivity. In this case, the study revealed strong connectivity through net directional and net pairwise analyses. This study showed that the market is not as highly integrated as predicted. However, we found a high long-run connectedness compared with short-run connectedness, suggesting that the variables do not respond swiftly to market developments. Sometimes, bands show high connectedness, which can be attributed to events (such as war and COVID-19) strongly influencing the network between oil and agricultural commodities.

Acknowledgements

We would like to take this opportunity to thank the editor, guest editor, and anonymous referees for useful comments and suggestions on the earlier version of this paper. The usual disclaimer applies.

Funding

We acknowledge the financial support from Indian Council of Social Science Research, New Delhi, India (F.No. 02/70/2021-22/MJ/RP).

Footnotes

References

Appendix

Equation A1 delineates the mathematical principles underlying the MCP approach, which prioritizes the minimization of asset correlations rather than variances, as seen in the MVP. The conditional correlation matrix R_t is derived from the conditional variance-covariance matrix |${h_t}$|⁠, which encompasses the conditional variances and covariances of asset returns at time t.

1. Conditional variance-covariance matrix:

|${H_t}$|

|${H_t}$| is a N × N matrix (for N assets) in which the diagonal elements |${h_{ii,t}}$| denote the conditional variances of the asset returns. The off-diagonal elements |${h_{ii,t}}\,$| denote the conditional covariances between asset i and asset j at time t.

2. Diagonal matrix of conditional standard deviations

|${\rm{diag}}{\left( {{H_t}} \right)^{ - 0.5}}$| diag(⁠|${H_t}$|⁠) is a diagonal matrix that includes the conditional variances h_ii,_t along its diagonal,

|${\rm{diag}}{\left( {{H_t}} \right)^{ - 0.5}}$| represents the inverse square root of the diagonal matrix, which standardizes the covariances by their respective standard deviations:

3. Normalization to obtain

|${{\bf{\it{R}}}_{\bf{\it{t}}}}$|

This standardizes the covariances in |${H_t}$| to yield correlations, using equation A1 ensuring that

The diagonal elements of |${R_t}$| are 1, representing the correlation of an asset with itself.

The off-diagonal elements denote the pairwise correlations between assets |${\rho _{ij,t}}$|⁠: