International Agri Commodities Investment Interdependence in a

Fractionally Cointegrated VAR Framework

Authors – Arnav Goel, Anshuman Bhartiya, Smit Sanghvi, Aakash Satayavada, Pushkar Devasthali, A
Rameswar Patra, Reetish Rath, Vaibhav Kolse, Vandan Doshi

Highlights

● Examines the long-term equilibrium relationships between energy and agricultural commodities using
the FCVAR model.
● Demonstrates significant spillover effects from energy prices (Brent Oil, Natural Gas, Gasoline) to key
agricultural commodities (Wheat, Soybean, Corn, Cotton, Sugar, Rice, Coffee, and Cocoa.).
● Provides insights into the implications for investment strategies and risk management in commodity
markets.

Abstract

This study investigates the long-term interdependence between international energy and agricultural
commodities using a Fractionally Cointegrated Vector Autoregressive (FCVAR) model. By analyzing daily
price data from January 2004 to June 2024, this research identifies significant cointegrating relationships and
persistent co-movements driven by shared economic factors. The findings reveal strong spillover effects from
energy prices to agricultural markets, suggesting that fluctuations in energy costs play a crucial role in shaping
agricultural commodity prices. These insights have important implications for investors seeking cross-hedging
opportunities and for policymakers aiming to stabilize food prices and ensure market efficiency.

Keywords

Fractional Cointegration, Commodity Markets, Energy Prices, Agricultural Commodities, Long-Memory

Process, Investment Interdependence

1. Introduction

The global commodity markets are interconnected and heavily influenced by price movements in the energy
sector. Rising energy costs, particularly crude oil prices, affect the production and transportation costs of
agricultural commodities, leading to volatility and co-movement across these markets. Given the growing
demand for biofuels and the integration of global supply chains, understanding the relationship between energy
and agricultural commodities is critical for investors, policymakers, and market participants.

This research employs the FCVAR model, a robust econometric framework that accounts for fractional
integration, to explore the long-term equilibrium relationships between selected energy and agricultural
commodities. The study aims to uncover the degree of market integration and the persistence of price shocks,
providing valuable insights into the dynamics of commodity price transmission and its implications for
investment strategies.
Research Objectives

● To examine the long-term equilibrium relationships between energy and agricultural commodities.
● To assess the persistence of price shocks across these interconnected markets.
● To analyze the implications of these interdependencies for investment strategies and policy decisions.

2. Related Literature Review

2.1 Market Integration and Price Transmission

The literature on commodity markets highlights the significant impact of energy prices on agricultural
commodities. Studies by Baffes (2007) and Nazlioglu (2011) demonstrate that rising oil prices often lead to
higher agricultural prices through increased costs of inputs like fertilizers and transportation. Biofuel production
further amplifies these linkages, particularly for crops like corn and soybean (Serra & Zilberman, 2013).

2.2 Long-Memory Processes in Commodity Markets

Research by Baillie and Bollerslev (1994) introduced the concept of long-memory processes in financial time
series, suggesting that shocks have persistent but decaying effects over time. Johansen and Nielsen (2012)
extended this concept to cointegration analysis, proposing the FCVAR model to account for fractional
integration. This model allows for a more flexible analysis of long-term dependencies, making it suitable for
studying the interconnectedness of commodity markets.

2.3 Empirical Evidence of Volatility Spillovers

The increasing financialization of commodity markets has led to greater volatility and spillover effects between
energy and agricultural sectors (Tang & Xiong, 2012). Studies using VAR, VECM, and GARCH models have
found strong evidence of price transmission, particularly during periods of economic uncertainty (Nazlioglu et
al., 2020). However, few studies have employed the FCVAR model to capture the persistent co-movements
observed in these markets.

2.4 Identified Research Gaps

Despite extensive research, several gaps remain:

 Limited exploration of long-term dependencies using fractional integration techniques like FCVAR.
 Insufficient analysis of the impact of geopolitical events on the interdependence between energy and
agricultural markets.
 A need for a comprehensive approach that integrates both energy and agricultural sectors to assess
investment interdependence holistically.

3. Data

3.1 Descriptive Statistics Analysis

The dataset used for this analysis includes daily price data for key energy and agricultural commodities
spanning from January 2004 to June 2024. This extensive time frame captures multiple economic cycles,
offering valuable insights into both short-term fluctuations and long-term trends. The commodities analyzed
include:

● Energy Commodities: Brent Oil, Natural Gas, and Gasoline.

● Agricultural Commodities: Wheat, Soybean, Corn, Cotton, Sugar, Rice, Coffee, and Cocoa.

The descriptive statistics (Figure 1) provide a comprehensive summary of the key features of these
commodities' price behaviors over the analysis period.

Na
Br S
Sta tur Gas Soy Co Co
en Wh Co u Ri Coc
tis al olin bea tto ffe
t eat rn g ce oa
tic Ga e n n e
Oil ar
s
4 4
Co 48 48 489 48 489 48 48 8 8 48 48
unt 90 90 0 90 0 90 90 9 9 90 90
0 0
1
59 46 77 1 14 26
Me 77 4.2 214 111 6.
8.4 1. .8 3. 8. 67.
an .4 4 .34 8.2 9
9 36 9 3 63 83
7
19 18 39 8. 7. 86
Mi 1.4 41. 29 527 13
.3 6. .1 4 0 .6
n 8 18 3 .25 21
3 25 4 5 6 5
1 1
59 48 36 62 11
0.2 2.7 163 2. 1. 22
.3 4.2 911 0. .8 5.
5 2 .1 7 0 27
5 5 75 7 9
1 4
1
74 56 39 73 1 13 25
3.5 209 104 3.
0.5 .3 2.2 9. .7 6. 4. 25.
6 .9 5 0
2 5 75 6 3 43 5
6
1 1
136 58 85 17
0.7 97 4.8 264 69 9. 5. 29
1.1 0. .2 5.
5 .6 4 .64 4 7 3 29
9 25 7 94
3 9
3 2
14 14 83 21 30 11
Ma 15. 427 177 5. 4.
6. 25. 1. 5. 4. 87
x 38 .62 1 3 4
08 25 25 15 9 8
1 6
23 15 14 24 5. 2. 43 10
2.2 62. 279
Std .9 9.7 5. .7 1 8 .9 51.
7 51 .49
8 6 34 1 2 2 9 37

Figure 3.1 – Descriptive Statistics

Key Observations from Descriptive Statistics

A. Analysis of Mean Prices

Brent Oil has a mean price of 77.40, highlighting its role as a critical global benchmark for energy pricing. It
influences production costs across industries, especially in transportation and manufacturing. The high mean
reflects sustained global demand and its sensitivity to supply chain disruptions.

Soybean follows with a mean price of 1118.20, making it the most valuable agricultural commodity in the
dataset. Its high price is driven by its dual role in food production and biofuels, coupled with strong global
demand for animal feed.

Cocoa, with a mean price of 2667.83, is the highest among all agricultural commodities analyzed. The
elevated price level reflects its limited supply, concentrated production regions (e.g., West Africa), and
vulnerability to climate factors affecting yields.

In contrast, Natural Gas shows a significantly lower mean price of 4.24. This can be attributed to abundant
supply in key markets like the United States, advancements in extraction technologies, and seasonally
influenced demand patterns.

Sugar has the lowest mean price (16.97), suggesting a well-regulated market with relatively stable production
levels. Government subsidies and trade regulations often play a role in keeping sugar prices consistent.

B. Volatility Insights

Cocoa exhibits the highest standard deviation (1051.37), indicating extreme price volatility. This suggests that
Cocoa is highly sensitive to supply shocks, weather disruptions, and geopolitical factors in its main
production regions. The volatility also reflects speculative trading, as Cocoa is a key commodity in financial
markets.

Among energy commodities, Gasoline (std of 62.51) shows substantial price variability, reflecting its direct link
to crude oil prices and consumer demand, especially during peak travel seasons.

Brent Oil (std of 23.98) also demonstrates significant volatility due to its exposure to global economic
conditions, geopolitical events, and changes in OPEC’s production policies.

Natural Gas, with a lower standard deviation of 2.27, indicates relatively less volatility compared to other
energy commodities. This suggests that despite seasonal demand fluctuations, the availability of ample
reserves and consistent supply from major producers helps stabilize its price.

C. Price Range and Identification of Outliers

The wide range between minimum and maximum prices for several commodities highlights the potential for
extreme price swings. For example, Brent Oil prices varied from a low of 19.33 to a high of 146.08, showing
significant spikes during major economic or geopolitical crises, such as the 2008 financial meltdown
and recent conflicts affecting oil supply.

Cocoa’s maximum price of 11878, which is far above its mean of 2667.83, suggests the presence of severe
outliers. These extreme values can be attributed to unexpected supply disruptions and sudden surges in
demand from the global chocolate industry.

Natural Gas also has a notable range (1.48 to 15.38), reflecting sharp price fluctuations during severe weather
events like cold winters, which drive up demand for heating.

D. Interquartile Range (IQR) Analysis

The IQR offers insights into the central tendency and stability of prices. Wheat and Corn have moderate IQRs
(e.g., Wheat ranges from 484.25 to 694.00), indicating relatively predictable price behavior. This stability is
likely supported by consistent agricultural output and strategic stockpiling practices by major producers.

Cocoa and Soybean exhibit wider IQRs, suggesting greater price variability even within the central 50% of
observations. This reflects the influence of external factors such as adverse weather conditions, global demand
shifts, and export restrictions that affect supply availability.

Rice (IQR from 11.04 to 15.39) and Sugar (IQR from 12.71 to 19.73) show narrower spreads, indicating more
stable price distributions. These commodities benefit from government interventions, including price controls
and import/export regulations, which help maintain consistent pricing.

Insights for Investment and Policy Implications:

The observed high volatility in Cocoa and Gasoline suggests that these commodities carry substantial risk
for investors, making them attractive for speculative trading but also requiring robust risk management
strategies. Derivative instruments like futures contracts can be used to hedge against unexpected price
swings.

Natural Gas’s lower volatility and tighter IQR indicate a more stable market environment, making it a
potentially safer option for conservative investors seeking exposure to the energy sector without the
heightened risk seen in oil markets.

Policymakers need to be aware of the extreme price volatility in key commodities like Cocoa and

Soybean, as these fluctuations can impact food security and global trade. Strategic reserves and intervention
measures (e.g., subsidies, export controls) may be necessary to stabilize these markets during periods of
crisis.

The post-2021 price surges seen across multiple commodities reflect a new phase of heightened uncertainty,
driven by factors such as climate change, ongoing geopolitical tensions, and disruptions in global supply
chains. This underscores the need for adaptive policy frameworks and agile risk management approaches to
handle future market volatility.

3.2 Time Series Analysis of Price Trends

The time series depiction of commodity prices (Figure 2) provides a comprehensive view of the price
movements for selected energy and agricultural commodities over the 20-year period from January 2004 to
June 2024. The chart illustrates the dynamics and co-movements of different commodities, highlighting periods
of stability, volatility, and significant price shocks.
 Figure 3.2 – Time Series Depiction of Price Trends

Overview of Commodity Price Trends

The time series graph displays a diverse range of price behaviors across the commodities analyzed,
indicating varying degrees of sensitivity to global economic conditions, supply chain disruptions, and market-
specific factors.

Brent Oil exhibits the most dramatic price fluctuations among all the commodities. It serves as a global
benchmark for crude oil, and its price movements reflect changes in global demand, geopolitical risks, and
supply shocks. The sharp spikes and subsequent drops are evident in several key periods –

- A significant dip during the 2008 financial crisis, followed by a strong recovery that peaked around
2014 due to robust global demand and OPEC’s production control.

The most pronounced spike occurs post-2021, influenced by the COVID-19 pandemic and subsequent
recovery, supply chain issues, and geopolitical events like the Russia-Ukraine conflict, which severely
disrupted global oil supplies.

A. Energy Commodities Analysis

Natural Gas prices show noticeable seasonal fluctuations, reflecting the impact of weather patterns and
heating demand, especially in colder months. The time series indicates spikes during periods of extreme
weather events (e.g., harsh winters) and disruptions in gas supply, as seen in the mid-2000s and early 2020s.

Gasoline prices closely track the movements of Brent Oil but exhibit less volatility in some periods. This
correlation is due to the direct relationship between crude oil prices and refined gasoline products. Gasoline
prices show peaks during peak travel seasons and economic recoveries, while demand shocks (e.g., during
the pandemic lockdowns) caused temporary declines.
 B. Agricultural Commodities Analysis

Soybean, Corn, and Wheat display moderate price volatility compared to energy commodities. These
agricultural products are sensitive to changes in weather conditions, crop yields, and shifts in biofuel demand:

- Soybean prices show significant fluctuations around 2012, likely influenced by drought conditions in the
U.S. Midwest, which impacted crop yields. Additionally, increased demand for soy-based biofuels
contributed to price increases.

- Wheat and Corn prices exhibit spikes during the 2008 food crisis, driven by global supply shortages,
export restrictions, and rising biofuel demand.

- Another noticeable peak occurs in 2021, reflecting disruptions from the COVID-19 pandemic and
adverse weather events affecting production.

Cocoa displays notable peaks and high volatility, particularly around 2010 and again post-2021. Cocoa
production is concentrated in a few regions (e.g., West Africa), making it vulnerable to supply shocks caused
by political instability, adverse weather conditions, and changes in global demand from the chocolate industry.

Stable Agricultural Commodities

Rice, Sugar, and Cotton exhibit more stable trends with relatively lower volatility compared to other
commodities. These staples tend to have predictable production cycles and consistent demand:

- Rice prices show minimal fluctuations, indicating stability in global rice production and strong
government interventions, including subsidies and export controls, aimed at ensuring food security.

- Sugar prices, while relatively stable, display occasional spikes related to changes in global trade
policies, production issues, and shifts in demand for sugar-based ethanol.

- Cotton prices reflect changes in global textile demand, weather conditions affecting crop yields, and
trade policies. Notable price spikes around 2011 can be attributed to supply shortages and strong
demand from the textile industry.

C. Notable Market Events and External Shocks

Several external shocks are evident in the time series data, reflecting significant global economic events -

The 2008 financial crisis had a widespread impact on commodity prices, leading to sharp declines due to
reduced global demand and financial market instability.

The COVID-19 pandemic in 2020 caused a temporary collapse in demand for both energy and agricultural
commodities, followed by a rapid recovery as economies reopened, highlighting the sensitivity of these
markets to global demand shocks.

The Russia-Ukraine conflict in 2022 significantly impacted energy markets, leading to a surge in Brent Oil
and Gasoline prices. The conflict disrupted key agricultural exports, causing price spikes in Wheat and Corn
due to supply chain disruptions.

D. Co-Movement and Correlation Analysis:

The time series graph suggests a degree of co-movement between energy and certain agricultural
commodities, particularly during periods of economic crisis or geopolitical events. The strong correlation
between Brent Oil and Gasoline prices is evident, reflecting their direct relationship.
Agricultural commodities like Soybean and Corn also show co-movement during certain periods, driven by
common factors such as biofuel demand and adverse weather conditions affecting crop yields.

Conversely, Rice and Sugar appear to be less correlated with the broader market trends, highlighting their
relative stability and the influence of regulatory controls that buffer against extreme price swings.

Implications for Investors and Policymakers:

The analysis of time series trends reveals significant volatility in energy commodities, underscoring the
need for effective hedging strategies, especially for investors exposed to oil markets. The strong co-movement
between Brent Oil and Gasoline suggests opportunities for cross-hedging.

For agricultural commodities, understanding seasonal patterns and the impact of external shocks is crucial for
risk management. Price stability in staples like Rice and Sugar indicates lower risk exposure, but sudden
disruptions (e.g., trade restrictions, weather events) can still impact these markets.

Policymakers must consider the broader implications of commodity price volatility, particularly in light of recent
events like the pandemic and geopolitical conflicts. Stabilizing prices through strategic reserves and market
interventions can help mitigate the adverse effects of extreme price fluctuations on consumers and industries

4. Methodology

4.1 Introduction to the FCVAR Model

The Fractionally Cointegrated Vector Autoregressive (FCVAR) model is employed in this study to examine
the long-term equilibrium relationships between energy and agricultural commodities. Unlike traditional
Cointegrated Vector Autoregressive (CVAR) models, which assume integer integration (I(1) processes), the
FCVAR model accommodates fractional integration, allowing for more flexible analysis of time series data
that exhibit long-memory properties and slow mean reversion.

In real-world time series, such as commodity prices, deviations from equilibrium often persist for extended
periods, indicating fractional integration rather than standard integration. The FCVAR model captures these
dynamics by allowing both the variables and their deviations from equilibrium to be fractionally integrated,
offering a nuanced approach to identifying long-run dependencies that would be missed by conventional
cointegration model. The key equation of the FCVAR model can be expressed as follows:
 This model formulation allows for the estimation of both the fractional differencing parameters and the
cointegrating vectors, making it a robust tool for analysing persistent, slow-reverting behaviours in the
commodity markets.

4.2 Stationarity Testing - Augmented Dickey-Fuller (ADF) Test

The Augmented Dickey-Fuller (ADF) test was conducted to assess the stationarity of the commodity price
series, a crucial step in determining the appropriate model specifications. Stationarity testing helps identify
whether the time series exhibits a unit root, indicating non-stationarity, or if it is stationary, implying stable
statistical properties over time.
Brent Natural Soybea
Statistic Gasoline Wheat Corn Cotton Sugar Rice Coffee Cocoa
Oil Gas n
Test Statistic -2.249 -3.8965 -3.4553 -3.2062 -2.6277 -2.7264 -2.8638 -2.8256 -2.8074 -1.8381 -0.8076
p-value 0.189 0.0021 0.0093 0.0874 0.0874 0.0696 0.0497 0.0547 0.0572 0.3617 0.8169
Used Lags 5 32 23 23 30 18 32 19 31 22 32
Observations 4884 4857 4866 4866 4859 4871 4857 4857 4858 4867 4857
Critical Value
-3.4317 -3.4317 -3.4317 -3.4317 -3.4317 -3.4317 -3.4317 -3.4317 -3.4317 -3.4317 -3.4317
(1%)
Critical Value
-2.8621 -2.8621 -2.8621 -2.8621 -2.8621 -2.8621 -2.8621 -2.8621 -2.8621 -2.8621 -2.8621
(5%)
Critical Value
-2.5671 -2.5671 -2.5671 -2.5671 -2.5671 -2.5671 -2.5671 -2.5671 -2.5671 -2.5671 -2.5671
(10%)
Figure 4.1 – ADF Test Results

The ADF test results (refer to Figure 3) showed mixed outcomes across the commodities:

Stationary Series: Natural Gas, Gasoline, Wheat, and Cotton exhibited p-values below the 5%
significance level, indicating that these series are stationary and do not require further differencing
before model estimation.

Non-Stationary Series: Brent Oil, Soybean, Corn, Sugar, Rice, Coffee, and Cocoa had p-values above
0.05, suggesting non-stationarity. These series likely exhibit long-memory properties and are well-
suited for analysis using the FCVAR model, which accommodates fractional integration.

The identification of non-stationary series necessitates testing for fractional cointegration, as the FCVAR model
can capture both short-term dynamics and long-term equilibrium relationships for these variables.

4.3 Correlation Analysis

The correlation matrix (refer to Figure 4) provides an overview of the linear relationships between the
selected commodities. Key observations include -

High Correlation between Brent Oil and Gasoline (0.95): This strong positive correlation reflects the
direct relationship between crude oil and its refined product, gasoline. Price movements in Brent Oil are
closely linked to gasoline prices, driven by changes in crude oil supply, refining costs, and market
demand.
 Strong Correlation between Soybean and Corn (0.89): The high correlation between these
agricultural commodities suggests shared influences, such as biofuel demand and weather-related
supply shocks. Both crops are used extensively in animal feed and biofuel production, leading to co-
movements in their prices.

Weak Correlations for Natural Gas: Natural Gas shows weak or negative correlations with most other
commodities, indicating its relative independence from the broader agricultural and oil markets. The
price of Natural Gas is influenced more by seasonal factors (e.g., heating demand) and specific supply
dynamics, making it a potentially useful diversification asset.

The correlation analysis highlights the interconnected nature of some commodities while also pointing out
distinct, independent price movements for others, providing valuable insights for risk management and
investment strategies.

Figure 4.2 – Correlation Matrix

4.4 Estimation Process in FCVAR

The estimation of the FCVAR model involves several critical steps aimed at identifying and analyzing long-term
equilibrium relationships between fractionally integrated time series –

Lag Selection - The optimal lag length for the FCVAR model was chosen using information criteria such as
the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). These criteria help
balance model complexity and goodness-of-fit, ensuring the inclusion of relevant past observations in the
analysis.

Cointegration Rank Testing - Likelihood ratio (LR) tests were conducted to determine the number of
cointegrating relationships, denoted as r. The cointegration rank indicates the number of distinct long-term
equilibria present among the selected commodities.

Maximum Likelihood Estimation - The FCVAR model parameters, including the fractional differencing orders
(d and b), the cointegration matrix (β), and the adjustment coefficients (α), were estimated using maximum
likelihood methods. This estimation captures the long-run equilibrium relationships and the speed of
adjustment towards equilibrium.
Hypothesis Testing - Following the estimation, hypothesis tests were performed to impose restrictions on the
model parameters. For example, tests for long-run exogeneity examined whether certain variables do not
respond to deviations from equilibrium. These tests provide insights into the relative importance and influence
of each commodity in the long-term relationships.

4.5 Model Validation and Diagnostics

The residuals of the estimated FCVAR model were analysed for serial correlation using multivariate white
noise tests. The absence of significant serial correlation in the residuals indicates a well-specified model,
validating the adequacy of the FCVAR framework for capturing the dynamics of the selected commodities.

The validated FCVAR model was then used for forecasting and scenario analysis, offering insights into
potential future trends based on the detected long-term relationships.

5. Empirical Results and Discussion

5.1 Data Analysis: 2 Energy vs 2 Agricultural Commodities

This analysis focuses on Brent Oil (C01), Natural Gas (NG1), Wheat (W-1), and Soybean (S1), representing
critical sectors of the global economy. These commodities were selected due to their significant roles in energy
markets, agriculture, and broader global supply chains.

Brent Oil, the most widely used global benchmark for crude oil pricing, is integral to the energy market,
influencing industrial operations and transportation costs worldwide. Natural Gas, another key energy
commodity, plays a dual role in industrial processes and household energy consumption. On the agricultural
side, Wheat and Soybean are essential staples that drive global food supply, pricing structures, and trade.

Together, these commodities enable an analysis that spans individual price dynamics and explores the long-
term equilibrium relationships between the energy and agricultural sectors.

The choice of these commodities aligns with the research objective of examining fractional cointegration, a
statistical property signifying that non-stationary time series are fractionally integrated but exhibit co-movement
over the long run. This property is pivotal in understanding price dependencies in commodity markets, where
long-term influences are often masked by short-term volatility.

Economic Influences on Price Dynamics

The price dynamics of the selected commodities are shaped by distinct economic and structural factors.
Energy commodities like Brent Oil and Natural Gas are particularly sensitive to geopolitical events,
supply-demand imbalances, and advancements in energy technologies. These factors contribute to short-
term price volatility while also establishing trends that influence long-term price behavior.
Conversely, agricultural commodities such as Wheat and Soybean are predominantly affected by climate
conditions, crop yields, and international trade policies. These factors result in prolonged price trends
influenced by systemic disruptions and seasonal cycles. By capturing these long-term and short-term
influences, fractional cointegration becomes a valuable analytical framework for exploring the intricate price
relationships in these markets.

Fractional Differencing and Persistence Analysis

The fractional differencing parameters (d-values) provide insights into the persistence or "memory" in the price
series of each commodity. These values quantify the degree of differencing required to make the time series
stationary while retaining essential long-term dependencies.

Brent Oil 0.377071

Natural Gas 0.454756

Wheat 0.901566

Soyabean 0.881723

Table 5.1 - Fractional Differencing Parameters

 Brent Oil (d = 0.377): The moderate d-value indicates that Brent Oil prices exhibit some long-term
memory, where past price changes moderately affect future trends. This behavior aligns with the dual
nature of the oil market, which experiences short-term shocks such as changes in production quotas or
geopolitical tensions alongside sustained trends driven by global energy demand and technological
innovations in energy production.
 Natural Gas (d = 0.454): Similar to Brent Oil, the d-value for Natural Gas reflects moderate
persistence. However, it exhibits slightly higher memory effects, influenced by factors like weather
changes, infrastructure development, and global policy shifts toward cleaner energy sources. These
factors suggest that while Natural Gas prices are subject to short-term volatility, they also reflect
enduring structural changes.
 Wheat (d = 0.901) and Soybean (d = 0.882): Agricultural commodities show significantly higher
persistence compared to energy commodities, with d-values nearing 1. This indicates that price
changes in Wheat and Soybean are less susceptible to short-term shocks and more influenced by
long-term factors like crop cycles, climate change, and global trade policies. Their high persistence
underscores the necessity of applying substantial fractional differencing to achieve stationarity while
preserving crucial long-term dependencies.

Implications for Analysis

The persistence levels indicated by the d-values affirm the importance of fractional differencing in
preparing time series data for advanced modelling techniques like the FCVAR framework. These
methods enable the simultaneous examination of short-term fluctuations and long-term equilibrium
relationships, offering a comprehensive understanding of price dynamics across energy and agricultural
markets.

Significance of Fractional Differencing in FCVAR

The estimated d-values for each commodity emphasize the critical role of fractional differencing in the FCVAR
model. Fractional differencing serves as a method to transform non-stationary time series into a stationary
form while retaining the inherent long-term dependencies or memory present in the data.

By applying fractional differencing, the time series is appropriately pre-processed, allowing the subsequent
FCVAR model to effectively capture both short-term dynamics (e.g., market shocks or volatility) and long-
term equilibrium relationships (e.g., co-movement between energy and agricultural prices). This dual
focus ensures a comprehensive analysis of commodity markets.

5.2 Johansen Cointegration Test Results

The Johansen cointegration test was conducted to determine the presence and number of long-term
equilibrium relationships among the selected commodities - Brent Oil, Natural Gas, Wheat, and Soybean. This
test is particularly effective for systems involving multiple variables, as it identifies the number of
cointegrating vectors and provides critical insights into the interdependencies of the series.

Trace Statistics and Critical Values

Table 5.2(a) summarizes the results of the Johansen test, which includes the trace statistics for hypotheses of
0, 1, 2, and 3 cointegrating vectors along with the corresponding critical values at 90%, 95%, and 99%
confidence levels.

Johansen Test Trace Statistics [4175.22729406, 1861.98702326, 183.64181788, 32.91590947]

Critical Values (90%, 95%, 99%) [44.4929, 47.8545, 54.6815]

[27.0669, 29.7961, 35.4628]
[13.4294, 15.4943, 19.9349]
[2.7055, 3.8415, 6.6349]

Eigenvalues [0.30423785, 0.23140205, 0.02335873, 0.00514836]

Table 5.2 (a) - Johansen Test Results

Rank (Null Trace Critical Value Conclusion

Hypothesis) Statistic (99%)

H0:r=0 4175.23 54.68 Reject H0: At least one cointegrating

relationship exists

H0:r≤1 1861.99 35.46 Reject H0: At least two cointegrating

relationships exist

H0:r≤2 183.64 19.93 Reject H0: At least three cointegrating

relationships exist

H0:r≤3 32.92 6.63 Reject H0: Four cointegrating

relationships exist

Table 5.2 (b) - Johansen Hypothesis

The results demonstrate that all trace statistics are significantly greater than their respective critical values,
consistently rejecting the null hypotheses.
The trace statistic for H0: r=0(4175.23) exceeds the 99% critical value (54.68) by a large margin, indicating
strong evidence of at least one cointegrating relationship. Similarly, the subsequent trace statistics (1861.99,
183.64, and 32.92) surpass their respective critical values, confirming the existence of four cointegrating
relationships among the commodities.

5.3 Eigenvalues and Strength of Relationships

Cointegrating Vector Eigenvalue Strength of Relationship

1 0.304 Strong

2 0.231 Moderate-Strong

3 0.023 Weak

4 0.005 Very Weak

Table 5.3 (a) - Eigenvalue Relationship

The first eigenvalue (0.304) highlights the strongest long-term relationship, reflecting significant co-movement
among the selected commodities.

The second eigenvalue (0.231) also signifies a relatively strong linkage, whereas the third and fourth
eigenvalues (0.023 and 0.005) represent weaker connections.

The ranking of eigenvalues underscores the dominance of the first two cointegrating vectors in capturing the
long-term dynamics within the system.

Interpretation and Implications

The presence of four cointegrating relationships indicates that the selected commodities exhibit significant
long-term co-movement, driven by shared economic factors and interdependencies. While individual
prices may fluctuate in the short term due to market-specific shocks, the results confirm their tendency to
adjust toward a common equilibrium in the long run.

The strength of the first two cointegrating vectors, as evidenced by the larger eigenvalues, suggests that these
relationships play a critical role in shaping the overall dynamics of the system. These findings are particularly
relevant for econometric modeling, such as Vector Error Correction Models (VECM), which can capture both
short-term adjustments and long-run equilibrium relationships.

Conclusion

The Johansen cointegration test results confirm that Brent Oil, Natural Gas, Wheat, and Soybean are
cointegrated, with four significant long-term relationships detected. The dominance of the first two vectors
highlights robust interconnections, particularly between energy and agricultural markets, which are likely
influenced by global economic trends and market forces. These results provide a solid foundation for further
modeling and analysis, emphasizing the importance of considering both short- and long-term dynamics in
commodity market studies.

5.4 Vector Error Correction Models (VECM)

The Vector Error Correction Model (VECM) reveals the intricate dynamics between Brent Oil, Natural Gas,
Wheat, and Soybean, uncovering their short-term interactions and long-term equilibrium relationships.

coef std err z P>¿ z∨¿ [0.025 0.975 ¿

L1.Brent Oil - C01 -0.2802 0.012 -23.050 0.000 -0.304 -0.256

L1.Natural Gas - NG1 0.0921 0.091 1.017 0.309 -0.085 0.269

L1.Wheat - W-1 -0.0027 0.002 -1.733 0.083 -0.006 0.000

L1.Soyabean - S1 -0.0054 0.001 −4.966 0.000 −0.008 -0.003

coef std err z P>¿ z∨¿ [0.025 0.975 ]

L1.Brent Oil - C01 -0.0009 0.002 -0.572 0.568 -0.004 0.002

L1.Natural Gas - NG1 -0.3370 0.012 -28.148 0.000 -0.361 -0.314

L1. Wheat - W-1 -0.0004 0.000 -1.743 0.081 -0.001 4.4e-05

L1.Soyabean - S1 2.345 e−05 0.000 0.162 0.871 −0.000 0.00

Table 5.4 (a) - VECM Results – 1

coef std err z P>¿ z∨¿ [0.025 0.975 ¿

L1.Brent Oil - C01 -0.2645 0.107 -2.469 0.014 -0.475 -0.055

L1.Natural Gas - NG1 -0.7832 0.798 -0.982 0.326 -2.347 0.780

L1. Wheat - W-1 -0.0576 0.014 -4.264 0.000 -0.084 -0.031

L1.Soyabean - S1 −0.0300 0.010 −3.113 0.002 −0.049 −0.011

coef std err z P>¿ z∨¿ ¿ 0.975 ¿

L1.Brent Oil - C01 -0.3443 0.128 -2.690 0.007 -0.595 -0.093

L1. Natural Gas - NG1 1.7454 0.953 1.832 0.067 -0.122 3.613

L1. Wheat - W-1 0.0881 0.016 5.456 0.000 0.056 0.120

L1.Soyabean - S1 −0.1082 0.012 −9.400 0.000 −0.131 −0.086

Table 5.4 (b) - VECM Results – 2

Loading coefficients (alpha) for Brent Oil

coef std err z P>¿ z∨¿ [0.025 0.975 ¿
ec1 0.4565 0.019 24.255 0.000 0.420 0.493

ec2 1.3412 0.029 46.425 0.000 1.285 1.398

 Loading coefficients (alpha) for Natural Gas
coef std err z P>¿ z∨¿ [0.025 0.975 ¿
ec1 -0.0004 0.000 -1.545 0.122 -0.001 0.000

ec2 −0.0005 0.000 −1.167 0.243 −0.001 0.000

Table 5.4 (c) - VECM Results - 3

Loading coefficients (alpha) for Wheat

coef td err z P>∣ z | [0.025 0.975 ¿
ec1 0.4565 0.019 24.255 0.000 0.420 0.493

ec2 1.3412 0.029 46.425 0.000 1.285 1.398

Loading coefficients (alpha) for Soyabean
coef std err z P>¿ z∨¿ [0.025 0.975 ¿
ec1 0.7293 0.022 32.441 0.000 0.685 0.773

ec2 −0.4698 0.035 −13.614 0.000 −0.537 −0.402

Table 5.4 (d) - VECM Results – 4

Cointegration Relations for loading-coefficients – Column 1

coef std err z P>¿ z∨¿ [0.025 0.975 ¿
beta. 1 1.0000 0 0 0.000 1.000 1.000

beta. 2 5.286e-17 0 0 0.000 5.29e-17 5.29e-17

beta. 3 0.1567 0.022 7.195 0.000 0.114 0.199

beta. 4 −0.7430 0.012 −62.779 0.000 −0.766 −0.720

Cointegration Relations for loading-coefficients – Column 2
coef std err z P>¿ z∨¿ [0.025 0.975 ¿
beta. 1 -1.041e-17 0 0 0.000 -1.04e-17 -1.04e-17

beta. 2 1.0000 0 0 0.000 1.000 1.000

beta. 3 -0.6269 0.012 -50.786 0.000 -0.651 -0.603

beta. 4 0.2999 0.007 44.704 0.000 0.287 0.313

Table 5.4 (E) - VECM Results - 5

In the short run, significant linkages emerge, with Brent Oil prices showing mean-reverting behavior and being
influenced by agricultural commodities like Wheat (-0.0027) and Soybean (-0.0054). This underscores the
impact of supply chains and shared production costs. Natural Gas exhibits strong mean-reverting tendencies (-
0.337) and largely independent short-term behavior, while Soybean maintains significant connections to both
energy and agricultural markets.

In the long run, the cointegration analysis confirms a shared equilibrium among the commodities, highlighting
persistent interdependencies driven by global economic forces and production linkages. The error correction
terms (ECTs) reveal varying speeds of adjustment to deviations from equilibrium: Soybean, with a rapid
adjustment rate of 72.93%, demonstrates resilience to shocks, while Brent Oil, correcting at just 1.87%, reflects
slower stabilization influenced by geopolitical and macroeconomic factors.

These findings emphasize the interconnectedness of commodity markets and their varying responses to
shocks, offering critical insights for forecasting, risk management, and understanding global market dynamics.
The VECM framework thus serves as a robust tool for analyzing the interplay of energy and agricultural
commodities.

5.5 FCVAR MODEL RESULTS - BRENT OIL VS WHEAT, SOYBEAN, CORN

This report summarizes the application of the Fractionally Cointegrated Vector Autoregressive (FCVAR) model
to analyze the long-term and short-term relationships between Brent oil and three agricultural commodities
wheat, soybean, and corn. The analysis is structured into stages, including lag selection, cointegration testing,
model estimation, residual analysis, and forecasting.

Lag Selection and Model Order Determination

The initial phase involved determining the optimal lag order (denoted as k) to capture the short-term
dependencies between Brent oil and the selected agricultural commodities. The Akaike Information Criterion
(AIC) and Bayesian Information Criterion (BIC) were utilized to evaluate various lag orders.

 AIC Minimization: A lag order of k=3 minimized the AIC, highlighting its ability to better capture the
dynamic relationships between variables.
 BIC Minimization: A lag order of k=0 minimized the BIC, indicating a simpler model with fewer
parameters.
Final Model Decision
To balance the trade-off between model complexity and explanatory power:
 A lag order of k=2 was selected for the analysis.

 Serial correlation tests on the residuals demonstrated that k=2 provides an adequate explanation of the
short-term interactions while maintaining model efficiency.
The selected lag order strikes a balance between capturing dynamic relationships and minimizing residual
dependencies, ensuring robust short-term relationship modeling.
Lag Selection Results

Dimension of Number of observations in

4 6390
system: sample:

Order for WN Number of observations for

12 6390
tests: estimation:

Restricted
No Initial values: 0
constant:
 Unrestricted
No Leve1 parameter: Yes
constant:

k r d b LogL LR pv AIC BIC

3 4 0.984 0.984 -84016.15 69.71 0.000 168170.30* 168636.91

2 4 0.976 0.976 -84051.01 21.90 0.146 168208.02 168566.43

1 4 0.973 0.973 -84061.96 105.39 0.000 168197.92 168448.13

0 4 1.005 1.005 −84114.66 0.00 0.000 168271.31 168413.32*

Table 5.5 (a) – Parameter Estimates and Results

Additionally, tests for serial correlation in the residuals showed acceptable results for k = 2, indicating that this
lag choice helps explain the short-term interactions adequately.

k pmvQ pQ1 pLM1 pQ2 pLM2 pQ3 pLM3 pQ4 pLM4

3 0.00 0.00 0.21 0.04 0.87 0.00 0.42 0.00 0.51

2 0.00 0.00 0.19 0.00 0.50 0.00 0.39 0.00 0.37

1 0.00 0.00 0.17 0.00 0.43 0.00 0.38 0.00 0.35

0 0.00 0.00 0.14 0.00 0.34 0.00 0.36 0.00 0.45

Table 5.5 (b) - Serial Correlation of Residuals

This section lays the foundation for subsequent steps in the analysis, including the evaluation of cointegration
and estimation of the FCVAR model. Further insights into long-term relationships and forecasting accuracy are
presented in the following sections.

5.5.1 Cointegration Rank Testing

The next stage of the analysis focused on cointegration rank testing to determine the number of stable, long-
term relationships between Brent oil and the selected agricultural commodities - wheat, soybean, and corn.
This step is critical to identifying whether these commodities exhibit co-movement over the long term, even
amidst short-term fluctuations.

Methodology

(unable to get table using mathpix)

Table 5.5.1 (a) - Likelihood Ratio Tests

The test results indicated that the most statistically significant rank is r = 1, meaning that there is one
cointegrating relationship among Brent oil, wheat, soybean, and corn.
 This finding suggests that while the prices of these commodities may vary in the short term, they are bound by
a stable, long-term equilibrium relationship. For instance, any deviation in wheat, soybean, or corn prices will
likely see an eventual correction toward equilibrium with Brent oil over time.

Fractionally Cointegrated VAR: Estimation Results

Number of observations in
Dimension of system: 4 6390
sample

Number of observations for

Number of lags: 2 6390
estimation

Restricted constant: No Initial values: 0

Unrestricted constant: No Level parameter: Yes

starting value for d: 0.800 Parameter space for d: (0.010 , 2.000)

Starting value for b : 0.800 Parameter space for b: (0.010 , 2.000)

Cointegrating rank: 1 AIC: 168223.067

Log-1ikelihood: -84067.533 BIC: 168520.616

log(det(Omega_hat)): 14.961 Free parameters: 44

Fractional parameters:

Coefficient Estimate Standard error

d 0.962 0.011
Table 5.5.1 (b) - Fractionally Cointegrated VAR: Estimation Results
The model’s estimated fractional differencing parameter d was approximately 0.962, close to 1, indicating
high persistence in the time series data. A d value near 1 suggests that price shocks to these commodities
have long-lasting effects, with the impacts fading slowly over time. This level of persistence is characteristic of
commodities, where factors like supply chain disruptions, seasonal fluctuations, and geopolitical events can
have prolonged influences on prices. The model's ability to account for fractional integration, rather than
treating the data as purely stationary or purely integrated, enhances the accuracy of long-term forecasting for
these commodities.

5.5.2 Adjustment Dynamics and the Alpha Matrix

Adjustment matrix (alpha):

Var 1(Brent -0.001

Oil) The Alpha matrix, which shows the adjustment speeds of each commodity
towards the long-term equilibrium, indicates that wheat and soybean adjust
SE 1 ( 0.000 ) more rapidly to deviations from equilibrium, while Brent oil and corn
exhibit slower adjustments. In practical terms, this means that wheat and
Var2(Wheat) 0.009
soybean prices are more reactive to imbalances, likely due to their higher
SE 2 ( 0.004 ) demand elasticity and shorter production cycles compared to Brent oil and corn.
The estimated cointegrating equation, also known as the Beta matrix, revealed
Var3(Soybean) -0.020 the specific long-term relationships between the commodities. Brent oil
(represented as Var1 in the output) has a positive relationship with wheat
prices and a negative relationship with soybean and corn prices.

This means that, in the long term, Brent oil and wheat tend to move in the same
 SE 3 ( 0.004 )

Var 4(Corn) -0.007

SE 4 ( 0.002 )
Table 5.5.2 (a) Alpha Matrix

Cointegrating equations (beta):

Variable CI equation 1

Var1(Brent Oil) 1.000

Var2(Wheat) -0.529

Var3(Soybean) 0.218

var4(Corn) -0.053

Table 5.5.2 (b) – Beta Matrix

5.5.3 Long-Run and Short-Run Effects

The Long-Run matrix (Pi) summarizes the overall long-term relationships between Brent oil and the agricultural
commodities, combining the effects of both the Beta (cointegration) and Alpha (adjustment) matrices. This
matrix shows relatively small values, indicating that while there is a stable equilibrium relationship, each
commodity's long-term impact on the others is modest.

This finding is typical in markets where commodities are influenced by global supply and demand but may not
have strong direct dependencies.

Long-run matrix (Pi):

Variable Var 1 Var 2 Var 3 Var 4

Var 1(Brent Oil) -0.001 0.000 -0.000 0.000

Var 2(Wheat) 0.009 -0.005 0.002 -0.001

Var 3(Soybean) -0.020 0.010 -0.004 0.001

Var 4(Corn) -0.007 0.004 -0.001 0.000

Level parameter (mu):

Var 1(Brent Oil) 25.092

SE 1 ( 1.404 )
 Var 2(Wheat) 247.499

SE 2 ( 9.517 )

Var 3(Soybean) 456.501

SE 3 ( 13.990 )

Var 4(Corn) 200.749

SE 4 ( 7.484 )
Table 5.5.3 (a) - Long – Run Matrix

The short-term dynamics, captured by the Lag Matrices (Gamma_1 and Gamma_2), reflect how lagged values
of each commodity affect others over shorter periods. The relatively low values in these matrices indicate that
short-term interactions are less significant than the long-term relationship captured in the Beta and Alpha
matrices. This suggests that while there may be minor short-term reactions among these commodities, they
are primarily driven by individual supply and demand factors in the short run.

Lag Matrix 1

Var 1(Brent var Var Var 4(Corn)

Variable
oil) 2(Wheat) 3(Soybean)

Var 1(Brent -
-0.001 0.006 0.004
Oil) 0.008

SE 1 ( 0.018 ) ( 0.002 ) ( 0.001 ) ( 0.003 )

-
Var 2(Wheat) -0.023 0.118 0.012
0.093

SE 2 ( 0.123 ) ( 0.020 ) ( 0.012 ) ( 0.026 )

Var -
-0.279 -0.016 0.078
3(Soybean) 0.030

SE 3 ( 0.156 ) ( 0.020 ) ( 0.020 ) ( 0.033 )

Var 4(Corn) -0.079 -0.020 0.019 0.048

SE 4 ( 0.080 ) ( 0.010 ) ( 0.008 ) ( 0.021 )

Lag Matrix 2
Variable Var 1(Brent Oil) Var 2(Wheat) Var 3(Soybean) Var 4(Corn)

Var 1(Brent Oil) -0.012 0.001 -0.001 0.003

SE 1 ( 0.016 ) ( 0.002 ) ( 0.001 ) ( 0.003 )

Var 2(Wheat) -0.196 0.012 -0.008 0.052

SE 2 ( 0.128 ) ( 0.016 ) ( 0.013 ) ( 0.028 )

Var 3(Soybean) 0.055 -0.030 0.016 0.026

 SE 3 ( 0.162 ) ( 0.020 ) ( 0.017 ) ( 0.035 )

Var 4(Corn) 0.069 -0.008 0.000 0.004

SE 4 ( 0.084 ) ( 0.011 ) ( 0.008 ) ( 0.019 )

Table 5.5.3 (b) - Lag Matrix 1,2

Table 5.5.3 (c) - Roots of the Characteristic Polynomial

5.5.4 Residual Analysis and Model Fit

To assess the model's fit, we conducted a White Noise Test on the residuals. Ideally, residuals should exhibit
white noise characteristics, meaning they should have no significant patterns or autocorrelations, indicating
that the model has captured the main dependencies.

The results of the test showed that while individual variables (commodities) had residuals approximating white
noise, the multivariate residuals were not perfectly white noise, suggesting some remaining patterns. This
outcome indicates that while the FCVAR model effectively captures the primary long-term relationships, there
may still be minor dependencies left unexplained.

Table 5.5.4 (a) - White Noise Tests (Var1-4-Brent Oil, Wheat, Soyabean, Corn)

5.5.5 Forecasting Commodity Prices

The final part of the analysis involved forecasting the prices of Brent oil, wheat, soybean, and corn for the next
12 periods.

The forecast plot reveals that wheat and soybean are projected to continue exhibiting high levels of volatility,
consistent with their historical behaviour, while corn prices are expected to remain relatively stable.

This forecast provides valuable insights for stakeholders in these markets, including farmers, investors, and
policymakers, allowing them to anticipate price trends and plan accordingly. The dashed line in the plot
demarcates the start of the forecast period, highlighting the model’s predicted trajectories based on historical
data.
 Figure 5.5.5 Commodity Prices Forecast

5.5.6 Conclusion

This FCVAR model analysis reveals a stable, long-term relationship between Brent oil and the
agricultural commodities of wheat, soybean, and corn. Despite their individual volatilities, these
commodities exhibit a common trend influenced by Brent oil prices, especially in the long term.

 Wheat prices are positively correlated with Brent oil, while soybean and corn show weaker or opposite
relationships.
 The high fractional differencing parameter indicates that price shocks have long-lasting effects,
especially for wheat and soybean, which adjust quickly back to equilibrium.

In terms of practical application, this analysis is valuable for anyone involved in commodity markets. The long-
term dependencies identified here allow for better planning and risk management strategies, while the forecast
provides a data-driven projection of future price trends.

5.6 MODEL RESULTS - BRENT OIL VS COTTON, SUGAR AND RICE (More optimized than previous
models)

In this study, we applied the FCVAR model to investigate the cointegration between Brent oil, representing
energy commodities, and three agricultural commodities: cotton, sugar, and rice. The primary aim is to
determine whether these commodities share a common stochastic trend, which could imply a long-term
equilibrium relationship. Such analysis is essential for understanding the dynamic linkages between energy
and agricultural markets, potentially influenced by factors such as biofuel policies, cost-push inflation, and
global economic shifts.

5.6.1 Lag Selection

The lag selection results from the FCVAR model indicate an optimal lag order of two for the analysis of Brent
Oil versus Cotton, Sugar, and Rice, with the dimension of the system set at four (reflecting the four
commodities under study). By comparing the AIC and BIC values, we find that the model with k=1 minimizes
both criteria (AIC = 48044.83 and BIC = 48295.05), suggesting it strikes the best balance between model
complexity and fit among the alternatives.

Lag Selection Results

Dimension of system: 4 Number of observations in sample: 6390

Order for wN tests: 8 Number of observations for estimation: 6390

Restricted constant: No Initial values: 0

Unrestricted constant: No Level parameter: Yes

 Parameter Estimates and Information Criteria
k r d b LogL LR pv AIC BIC

0.99 0.15
2 4 0.999 -23974.51 21.81 48055.03 48413.44
9 0

0.98 0.00
1 4 0.985 -23985.42 208.30 48044.83∗¿ 48295.05∗¿
5 0

1.03 0.00
0 4 1.033 -24089.57 0.00 48221.13 48363.15
3 0
Tests for Serial Relations of Residuals
k pmVQ pQ1 pLM1 pQ2 pLM2 pQ3 pLM3 pQ4 pLM4
2 0.00 0.00 0.36 0.00 0.08 0.00 0.25 0.02 0.44
1 0.00 0.00 0.30 0.00 0.07 0.00 0.17 0.02 0.43
0 0.00 0.00 0.02 0.00 0.01 0.00 0.07 0.00 0.00

Table 5.6.1 Lag Selection Results

The likelihood ratio (LR) test for k=1 yields a significant LR statistic of 208.30 with a p-value of 0.000,
reinforcing the suitability of this model specification. Lower AIC and BIC scores generally imply a more efficient
model that better captures the underlying data structure without overfitting.

The serial correlation tests on residuals reveal p-values that are largely above conventional significance levels,
suggesting that the residuals of the model are approximately white noise. For example, the p-values for the
multivariate Q-statistic (pmvQ) and the various lags of the LM test are all above 0.05, indicating no significant
autocorrelation in the residuals across multiple lags. This finding implies that the model does a reasonable job
of accounting for temporal dependencies in the data, with little systematic pattern remaining in the residuals.

Such results are favourable, as they suggest that the FCVAR model specification with k=1 provides a good fit
to the data without leaving notable patterns unexplained. This model configuration, therefore, allows us to
explore cointegration and long-term relationships among these commodities with a high degree of confidence
in its adequacy.

5.6.2 Cointegration Rank Testing

The cointegration rank testing results shown below provide crucial insights into the long-term relationships
between Brent Oil, Cotton, Sugar, and Rice prices. The table reports the Log-likelihood, LR (likelihood ratio)
statistic, and the p-values for different assumed ranks (0 to 4). The LR statistic tests the hypothesis of no
cointegrating relationships (or fewer cointegrating vectors) against the alternative of at least one cointegrating
relationship. A lower p-value indicates stronger evidence for cointegration.

Likelihood Ratio Tests for Cointegrating Rank

Dimension of system: 4 Number of observations in sample: 6390

Number of lags: 1 Number of observations for estimation: 6390

Restricted constant: No Initial values: 0

Unestricted constant: No Level parameter: Yes

 Log-
Rank d b LR statistic P-value
likelihood

0 0.975 0.975 -24014.884 58.934 0.014

1 0.978 0.978 -24003.248 35.662 0.040

2 0.982 0.982 -23993.969 17.105 0.122

3 0.985 0.985 -23985.442 0.050 1.000

4 0.985 0.985 -23985.417 ---- -

Table 5.6.2 Likelihood Ratio Tests For Cointergrating Rank

a. Rank 0 - the LR statistic is 58.934 with a p-value of 0.014, indicating significance at the 5% level,
meaning we reject the null hypothesis of no cointegration. This suggests that there is at least one
cointegrating vector, pointing to a long-term equilibrium relationship among the variables.
b. Rank 1 - the LR statistic is 35.662 with a p-value of 0.040, still significant but weaker than Rank 0.
c. Rank 2 - the LR statistic drops further to 17.105 with a p-value of 0.122, which is not significant at the
5% level, suggesting limited support for additional cointegrating vectors beyond the first one.

Finally, ranks 3 and 4 show LR statistics that are statistically insignificant (with p-values of 1.000), confirming
that these ranks do not add further explanatory power.

These results imply that a model with a cointegration rank of 1 is appropriate, as it captures the essential
long-term relationship among the commodities without unnecessary complexity. The presence of one
cointegrating vector suggests a stable relationship over time among Brent Oil, Cotton, Sugar, and Rice, which
could be influenced by shared economic factors such as global supply-demand dynamics, currency
fluctuations, and market shocks. This finding has practical implications for investors and policy-makers, as it
highlights interdependencies that might be exploited for hedging, forecasting, and decision-making in
commodity markets.

Model Analysis

5.6.3 Fractional Differencing Parameter and Model Fit

The fractional differencing parameter d=0.978 (with a small standard error of 0.009) indicates that the series
are close to being fully integrated but exhibit fractional behavior, suggesting a slow reversion to equilibrium.
The close-to-one value of d implies that shocks to the system have lasting effects but eventually revert over
time. This slow mean-reversion process is characteristic of markets that are interconnected, where
adjustments happen gradually.

Fractionally Cointegrated VAR : Estimation Results

Dimension of system: 4 Number of observations in sample: 6390

Number of lags: 1 Number of observations for estimation: 6390

Restricted constant: No Initial values: 0

Unrestricted constant: No Leve1 parameter: Yes

starting value for d: 0.750 Parameter space for d: (0.010 , 2.000)

 starting value for b : 0.750 Parameter space for b: (0.010 , 2.000)

Cointegrating rank: 1 AIC: 48062.496

Log-1ikelihood: -24003.248 BIC: 48251.846

log ⁡¿ Free
-3.839 28
det(Omega_hat)): parameters:
Fractional Parameters
Coefficient Estimate Standard Error

d 0.978 0.009
Table 5.6.3 FCVAR - Estimated Results

5.6.4 Cointegrating Equations (Beta) and Adjustment Matrix (Alpha) -

The cointegrating equation highlights the equilibrium relationship among Brent Oil, Cotton, Sugar, and Rice.
The equation shows Brent Oil (set as 1 in the vector) having a direct link with other commodities, with Cotton (-
2.997), Sugar (6.365), and Rice (-5.135) having varying impacts. These coefficients signify the equilibrium
relationship; for instance, a 1-unit change in Brent Oil is associated with roughly a 3-unit change in Cotton in
the opposite direction, a 6.4-unit change in the same direction in Sugar, and a 5.1-unit change in Rice in the
opposite direction.

Cointegrating Equations(beta)
Variable CI equation 1

Var1(Brent Oil) 1.000

Var2(Cotton) -2.997

Var3(Sugar) 6.365

Var4(Rice) -5.135

Adjustment Matrix(alpha)

Variable CI equation 1

Var 1(Brent Oil) -0.001

SE 1 ( 0.000 )

Var 2(Cotton) 0.001

SE 2 ( 0.000 )

Var 3(Sugar) 0.000

SE 3 ( 0.000 )

Var 4(Rice) 0.000

SE 4 ( 0.000 )
 Table 5.6.4 Alpha and Beta Matrix

The adjustment matrix (Alpha) shows how quickly each commodity responds to deviations from the
equilibrium. Brent Oil has a near-zero adjustment coefficient (-0.001), indicating it minimally adjusts to changes
in the equilibrium, possibly acting as a leading or influential commodity. In contrast, other commodities adjust
more significantly, with Cotton (0.001) and Rice (0.002) showing higher responsiveness to maintain
equilibrium.

5.6.5 Long-Run Matrix (Pi), Lag Matrix (Gamma), and Roots of the Characteristic Polynomial -

The long-run matrix (Pi) shows the interaction among the variables in the cointegrated system. For instance,
Cotton's effect on Brent Oil is -0.003, suggesting a small negative influence.

The lag matrix (Gamma_1) illustrates short-term adjustments. Notable values include Cotton's influence on
Brent Oil (-0.012) and Sugar’s adjustment based on Cotton (0.003), signifying short-term dependencies that
differ from the long-term equilibrium.

The roots of the characteristic polynomial, with moduli close to 1, further indicate the model's stability and the
persistence of shocks. Roots near the unit circle imply that the system is stable but responsive, capturing long-
term dependencies effectively.

Long-run matrix (Pi):

Variab1e Var 1(Brent oil) Var 2(Cotton) Var 3(Sugar) Var 4(Rice)

Var 1(Brent Oil) -0.001 0.003 -0.006 0.005

Var 2(Cotton) 0.001 -0.004 0.008 -0.007

Var 3(Sugar) 0.000 -0.000 0.000 -0.000

Var 4(Rice) 0.000 -0.000 0.000 -0.000

Level parameter (mu):

Var 1(Brent Oil) 25.078

SE 1 ( 1.440 )

Var 2(Cotton) 51.034

SE 2 ( 1.452 )

Var 3(Sugar) 6.126

SE 3 ( 0.334 )

Var 4(Rice) 5.218

SE 4 ( 0.203 )
 Lag Matrix 1(Gamma_1)
Var 1(Brent Var Var Var
Variable
Oil) 2(Cotton) 3(Sugar) 4(Rice)

Var 1(Brent
-0.012 -0.015 0.142 0.159
Oil)

SE 1 ( 0.016 ) ( 0.012 ) ( 0.057 ) ( 0.090 )

Var 2(Cotton) 0.006 0.148 -0.010 0.095

SE 2 ( 0.014 ) ( 0.017 ) ( 0.061 ) ( 0.097 )

Var 3(Sugar) -0.006 -0.003 0.024 0.017

SE 3 ( 0.003 ) ( 0.003 ) ( 0.016 ) ( 0.021 )

Var 4(Rice) -0.002 0.002 0.010 0.159

SE 4 ( 0.002 ) ( 0.002 ) ( 0.008 ) ( 0.016 )

Roots of the characteristic polynomial

Number Real part Imaginary part Modu1us

1 11.816 -39.421 41.154

2 11.816 39.421 41.154

3 7.218 0.000 7.218

4 5.988 0.000 5.988

5 1.005 0.000 1.005

6 1.000 0.000 1.000

7 1.000 0.000 1.000

8 1.000 0.000 1.000

Table 5.6.5 Long Run - matrix and Lag Matrix

5.6.6 Multivariate White Noise Test on Residuals

The Multivariate White Noise (MVWN) test on residuals is used to determine whether the residuals (errors)
from the FCVAR model are white noise, meaning they lack autocorrelation and exhibit purely random behavior.
If the residuals are white noise, it indicates that the model has effectively captured the structure of the
data, leaving no predictable pattern in the residuals.

 The overall Multivariate (Multivar) Q-statistic is 278.289, with a p-value of 0.000. This low p-value (<
0.05) suggests that, collectively, the residuals for all variables exhibit some level of autocorrelation,
indicating that the model may not have completely captured all the dynamics in the data.

White Noise Test Results(lag = 8)

 Variable Q P-val LM P-val

Multivar 278.289 0.000

Var1(BrentOIl) 27.921 0.000 9.438 0.307

Var2(Cotton) 53.266 0.000 14.701 0.065

Var 3(Sugar) 38.200 0.000 11.730 0.164

Var4(Rice) 18.000 0.021 7.850 0.448

Table 5.6.6 White Noise Test Results

5.6.7 Individual Variables -

Var1 (Brent Oil): The Q-statistic for Var1 is 27.921, with a p-value of 0.000, which indicates significant
autocorrelation in Var1's residuals. The LM statistic is 9.438, with a higher p-value of 0.307, indicating no
significant autocorrelation based on this test.

Var2 (Cotton): The Q-statistic of 53.266 and a p-value of 0.000, which shows significant autocorrelation in the
residuals. The LM statistic is 14.701, with a marginal p-value of 0.065, suggesting slight evidence of
autocorrelation.

Var3 (Sugar): The Q-statistic is 38.200 with a p-value of 0.000, also indicating significant autocorrelation. The
LM test statistic is 11.730 with a p-value of 0.164, suggesting no significant autocorrelation based on the LM
test.

Var4 (Rice): The Q-statistic for Var4 is 18.000, with a p-value of 0.021, which is marginally significant,
indicating slight autocorrelation. The LM test statistic for Var4 is 7.850 with a p-value of 0.448, showing no
significant autocorrelation in the residuals.

In summary, while the LM tests do not show significant autocorrelation for any variable, the Q-statistics for
Var1, Var2, and Var3 indicate that some autocorrelation remains in the residuals for these variables,
suggesting that the model may have missed certain patterns.

This could imply that further refinement of the model, such as adjusting lag structure or exploring alternative
model specifications, might be necessary to fully capture the dynamics in these time series.

Graphical Interpretation

Figure 5.6.7 Commodity Prices Forecast

Co-Movement Patterns - Throughout the timeline, there are notable instances where the commodities exhibit
parallel movements, suggesting a level of interdependence or shared influence. For example, a significant
spike is observed for both Brent oil and cotton around the mid-section of the graph, likely due to shared
economic or market events that impact both energy and agricultural commodities. These patterns of
simultaneous rise and fall indicate that external macroeconomic factors, such as global demand shifts, financial
crises, or policy changes, may be impacting these commodities together.

Differences in Volatility - The graph demonstrates that Brent oil experiences considerably higher volatility
compared to the agricultural commodities, as evidenced by the sharp peaks and troughs. Cotton shows
moderate volatility, reflecting sensitivity to seasonal changes and supply-demand dynamics in the agricultural
market. In contrast, sugar and rice display relatively stable, less volatile price trends, which could be attributed
to their nature as staple goods with more inelastic demand compared to energy commodities.

Distinct Seasonal and Long-Term Trends - Brent oil's sharp price fluctuations align with known global
events impacting the oil market, such as geopolitical tensions or production adjustments by major oil-producing
countries. Cotton also shows seasonal fluctuations, likely due to cyclical agricultural production and demand
patterns. Meanwhile, the trends in sugar and rice are more subdued, underscoring their relative stability and
the lower influence of external market shocks compared to energy commodities.

This graph provides a visual confirmation of the hypothesis that energy prices can indirectly influence
agricultural prices over the long term. Specifically, spikes in Brent oil prices may affect agricultural production
costs (e.g., fuel, transportation, and fertilizer costs), which in turn may impact agricultural commodity prices
over time. Thus, the observed co-movement suggests that oil price trends could serve as an indicator of cost
pressures in agricultural markets.

5.6.8 Conclusion

The FCVAR model, combined with visual analysis of the graph, reveals a meaningful fractional cointegrating
relationship between Brent oil and selected agricultural commodities (cotton, sugar, and rice). This model
outcome and the observed patterns in the graph indicate that these commodities do not operate in isolation
over the long run; rather, they are influenced by shared economic factors, with Brent oil potentially acting as a
major driver given its role in agricultural production and distribution costs.

The fractional cointegration results highlight the persistence of price shocks from Brent oil to agricultural
commodities. The high fractional differencing parameter (𝑑 = 0.978) suggests that short-term fluctuations in oil
prices can have lasting effects on agricultural commodities. For stakeholders including investors, policymakers,
and supply chain managers, understanding energy price movements can provide valuable insight into future
trends in agricultural prices, essential for budgeting, pricing strategies, and planning in both sectors.

5.7 FCVAR - Brent Oil vs All Agricultural Commodities

The Fractionally Cointegrated Vector Autoregression (FCVAR) model is a robust econometric tool designed to analyze
long-term relationships among time-series variables that exhibit fractional integration. Unlike traditional cointegration
models, which assume integer-order integration, FCVAR allows for fractional orders, capturing persistent memory and
long-range dependencies often observed in economic and financial data. This flexibility makes it particularly suitable for
examining commodity markets, where prices are influenced by complex dynamics such as global supply-demand shifts
and macroeconomic factors.

In this analysis, the FCVAR model is employed to investigate the interdependencies between Brent Oil and eight
agricultural commodities, using a dataset of 6,390 observations. By identifying cointegrating relationships and long-run
equilibria, the model provides a deeper understanding of price transmission mechanisms and the persistent linkages across
these critical markets.
 5.7.1 Lag Selection Results

The lag selection results provide the foundational structure for modelling the relationship between Brent Oil and
agricultural commodities. With a system dimension of 9 variables, the analysis encapsulates Brent Oil and 8 agricultural
commodities, allowing a holistic exploration of interdependencies within these markets. The dataset consists of 6,390
observations, ensuring sufficient data to enhance the robustness and reliability of the estimations. A lag order of 12 was
selected for the white noise tests, ensuring that the residuals of the model meet stationarity assumptions—a critical
requirement for reliable inference in time-series models. Importantly, the model does not incorporate restricted or
unrestricted constants, simplifying the structure while maintaining its ability to capture the underlying dynamics
effectively.

Lag Selection Results

Dimension of system: 9 Number of observations in sample: 6390

Order for WN tests: 12 Number of observations for estimation: 6390

Restricted constant: No Initial values: 0

Unrestricted constant: No Level parameter: Yes

Table 5.7.1 Lag Selection Results

5.7.2 Parameter Estimates and Information Criteria

The comparison of models with varying numbers of cointegrating relations (k) provides insights into the optimal model
structure. The analysis evaluates models based on log-likelihood, likelihood ratio (LR) statistics, and information criteria,
such as Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC).

k r d b LogL LR pv AIC BIC

3 9 0.975 0.975 -147841.41 355.86 0.000 296350.82* 298609.49

2 9 1.013 1.013 -148019.34 199.65 0.000 296544.67 298255.58

1 9 0.978 0.978 -148119.16 428.03 0.000 296582.32 297745.47

0 9 1.018 1.018 -148333.18 0.00 0.000 296848.35 297463.74*

Table 5.7.2 Parameter Results

The model with k=3 achieves the highest log-likelihood value (−147841.41), indicating the best fit to the data. The LR
statistic confirms that the inclusion of additional cointegrating relations significantly improves the model fit, with p-values
consistently below 0. Among the evaluated models, the k=3 has the lowest AIC (296350.82), highlighting it as the best-fit
model when prioritizing accuracy. However, the model with k=0 achieves the lowest BIC (297463.74), favouring
parsimony. This trade-off between complexity and simplicity underlines the importance of considering both criteria
depending on the research objective.

5.7.3 Integration and Cointegration Parameters

The fractional integration parameters (d, b) across the models further substantiate the presence of long-memory processes,
a hallmark of commodity markets. Values close to or slightly above 1 (e.g., d=0.975 and b=0.975 for k=3) suggest
fractional cointegration, capturing the persistent interdependence and price linkages between Brent Oil and agricultural
commodities. This nuanced approach extends beyond traditional cointegration methods, accommodating the long-run
equilibrium relationships with greater flexibility.

5.7.4 Serial Correlation of Residuals

The tests for serial correlation in the residuals reveal varied results. The overall multivariate Q-statistic indicates the
presence of significant autocorrelation (p=0.000), suggesting residual dependencies across the system. At the individual
level, some commodities, such as Rice (Var8), exhibit minimal autocorrelation (p>0.05 for Q-statistics), indicating well-
specified residual structures for these variables. In contrast, commodities like Brent Oil (Var1) and Cocoa (Var9) display
significant autocorrelation (p<0.05), pointing to potential issues with the model's lag structure or specification for these
variables. These findings suggest the need for further refinement in the model to fully address residual dependencies and
improve the accuracy of the FCVAR framework.

(Unable to get editable table from mathpix) Table 5.7.4 Tests For Serial Correlation Of Residuals

5.7.5 Evaluation of Cointegration Ranks in the FCVAR Model

The cointegration rank testing results offer a comprehensive view of the long-term relationships among Brent Oil and
agricultural commodities. Table __examining ranks highlights how the model improves in fit as the rank increases. At
rank 0, the log-likelihood value is −148150.56, with an LR statistic of 262.452 and a highly significant p-value of 0.000,
indicating that the model fit improves significantly when moving to a higher rank. However, the improvements diminish
as the rank increases, with p-values becoming nonsignificant (p>0.05) from rank 3 onwards. This indicates that a
cointegration rank of 2 or 3 is optimal, balancing model complexity and statistical significance without overfitting.

Rank d b Log-Likelihood LR statistic P -value

0 0.995 0.995 -148150.56 262.452 0
1 0.996 0.996 -148110.9 183.133 0.007
2 1.001 1.001 -148086.04 133.4 0.06
3 1.002 1.002 -148063.17 87.672 0.36
4 1.006 1.006 -148045.72 52.759 0.777
5 1.008 1.008 -148033.97 29.261 0.935
6 1.011 1.011 -148026.06 13.449 0.983
7 1.012 1.012 -148021.21 3.751 0.998
8 1.013 1.013 -148019.36 0.045 1
9 1.013 1.013 -148019.34 ---- ----
Table 5.7.5 (a) - Cointegration Rank By FCVAR Results
Table details the estimation results after selecting a cointegration rank of 1. The fractional integration
parameter (d=0.996) is estimated with high precision (standard error = 0.007), highlighting the strong
persistence and long-memory behaviour inherent in the time series data. The log-likelihood of the model at this
rank is −148110.904, and the model achieves an AIC of 296599.808 and a BIC of 297877.918, indicating a
good balance between fit and parsimony. While the higher ranks slightly improve the log-likelihood, they fail to
produce statistically significant improvements, reinforcing the decision to choose rank 1.
Fractionally Cointegrated VAR: Estimation Results
 Dimension of system: 9 Number of observat vations in sample: 6390

Number of lags: 2 Number of observ vations for estimation: 6390

Restricted constant: No Initial values: 0

Unrestricted constant: No Level parameter: Yes

Starting value for d: 0.800 Parameter space for d: ( 0.010 , 2.000)

Starting value for b: 0.800 Parameter space for b: (0.010 , 2.000)

Cointegrating rank: 1 AIC: 296599.808

Log-1ikelihood: -148110.904 BIC: 297877.918

log(det(Omega_hat)): 20.816 Free parameters: : 189

Fractional parameters:

Coefficient Estimate Standard error

d 0.996 0.007
Table 5.7.5 (b) - FCVAR - Estimated Results

The results obtained from the cointegration rank testing and subsequent FCVAR model estimation carry
significant implications for understanding the dynamics between Brent Oil and agricultural commodities. The
optimal rank of 1 suggests that a single long-term equilibrium relationship governs the interplay between these
markets. This implies that, despite their individual volatilities, the commodities are linked through a common
long-term factor, likely driven by global economic and market forces such as energy-agriculture
interdependencies or shared responses to macroeconomic shocks. The high fractional integration parameter
(d=0.996) indicates persistent, slow-decaying dependencies, emphasizing that shocks to these markets are not
quickly absorbed and can have lasting impacts on price movements.

For policymakers and market participants, these results highlight the need to consider the interconnectedness
between energy and agricultural commodities when formulating strategies. For instance, changes in oil prices
may have spill over effects on agricultural markets, particularly those tied to biofuels such as corn and soybean.
Similarly, understanding these relationships can assist investors in diversifying portfolios or hedging against
commodity price risks. Overall, the results demonstrate the FCVAR model’s utility in capturing and interpreting
the intricate dynamics of interconnected global markets.

5.7.6 Cointegrating Equations (Beta)

The cointegrating equation represents the long-term equilibrium relationship among Brent Oil and the agricultural
commodities. Brent Oil (Var1) is normalized with a coefficient of 1, serving as the reference variable. Among the
agricultural commodities, Cocoa (Var7) exhibits the strongest positive relationship with a coefficient of 10.724, indicating
that it has a significant influence in driving the long-term equilibrium with Brent Oil. In contrast, Sugar (Var6) shows the
strongest negative relationship with a coefficient of −5.739, suggesting an inverse effect on the equilibrium relationship.
Other commodities, such as Wheat (Var2, −0.544) and Corn (Var4, 0.253), exhibit moderate contributions, highlighting
their roles in maintaining the equilibrium state but with less influence compared to Cocoa and Sugar. The coefficients
reflect the strength and direction of the relationships between Brent Oil and each commodity. Positive coefficients, such
as those of Cocoa and Corn, suggest that increases in these commodities' prices are associated with upward adjustments in
the equilibrium price of Brent Oil. Conversely, negative coefficients, such as for Sugar and Wheat, indicate that increases
in their prices contribute to downward adjustments in the equilibrium.
 Cointegrating equations (beta):
Variable CI equation 1
Var1(Brent Oil) 1
Var2(Wheat) -0.544
Var3(Soybean) 0.05
Var4(Corn) 0.253
Var5(Cotton) 0.242
Var6(Sugar) -5.739
Var7(Rice) 10.724
Var8(Coffee) -0.056
Var9(Cocoa) 0.003

Table 5.7.6 Beta Matrix

5.7.7 Adjustment Matrix (Alpha)

The adjustment matrix captures the short-term dynamics of how Brent Oil and agricultural commodities respond to
deviations from the long-term equilibrium. The adjustment coefficient for Brent Oil is −0.001, indicating a very slow rate
of correction back to equilibrium when shocks occur. Cocoa and Sugar, despite their strong contributions to the
equilibrium state, exhibit near-zero adjustment coefficients (−0.000), suggesting that these commodities do not actively
participate in short-term corrections.

Interestingly, some agricultural commodities, such as Wheat (0.008) and Soybean (−0.015), exhibit relatively larger
adjustment coefficients, indicating that they play a more active role in stabilizing the system in response to short-term
shocks. These dynamic highlights the diversity in the responsiveness of different commodities, where some actively
correct disequilibrium while others remain relatively inert.

Adjustment matrix (alpha):

Variable CI Equation 1

Var 1(Brent Oil) -0.001

SE 1 ( 0.000 )

Var 2(Wheat) 0.008

SE 2 ( 0.004 )

Var 3(Soybean) -0.015

SE 3 ( 0.005 )

Var 4(Corn) -0.010

SE 4 ( 0.002 )

Var 5(Cotton) -0.002

SE 5 ( 0.000 )

Var 6(Sugar) -0.000

SE 6 ( 0.000 )
 Var 7(Rice) -0.000

SE 7 ( 0.000 )

Var 8(Coffee) -0.001

SE 8 ( 0.001 )

Var 9(Cocoa) -0.005

SE 9 ( 0.020 )
Table 5.7.7 Alpha Matrix

The Beta and Alpha matrices collectively provide valuable insights into the market dynamics between Brent Oil and
agricultural commodities. The Beta coefficients underscore the dominant role of Cocoa and Sugar in shaping long-term
price relationships, with Cocoa exerting a strong positive influence and Sugar a significant negative impact. In contrast,
the Alpha coefficients reveal that short-term adjustments are primarily driven by Wheat and Soybean, while Brent Oil
and most other commodities exhibit minimal short-term responsiveness.

These results suggest that persistent shocks in Cocoa and Sugar prices can have long-lasting effects on Brent Oil,
emphasizing their strategic importance in energy-agriculture market interdependencies. Simultaneously, the active short-
term adjustments by Wheat and Soybean highlight their stabilizing role in mitigating market imbalances. These insights
are crucial for policymakers and market participants in managing commodity price risks and understanding the broader
implications of energy-agriculture linkages.

5.7.8 Long-Run Matrix (Pi)

The long-run matrix reflects how deviations from the long-term equilibrium relationship influence each variable in the
system. The values in the matrix are generally small, suggesting that adjustments in individual variables due to
disequilibrium are relatively minor. For instance, Brent Oil (Var1) shows near-zero coefficients with other variables,
including Cocoa (Var7, −0.014) and Sugar (Var6, 0.007). This implies that while Brent Oil is part of the long-term
equilibrium, it is less affected by deviations in these relationships in the short run. Similarly, the minor adjustments for
Rice (Var8) and Coffee (Var9) (−0.001 and 0.003, respectively) reinforce the idea that short-term deviations in these
commodities are not strongly driven by the system's disequilibrium.

Variable Var 1 Var 2 Var 3 Var 4 Var 5 Var 6 Var 7 Var 8 Var 9

Var 1(Brent Oil) -0.001 0.001 -0.000 -0.000 -0.000 0.007 -0.014 0.000 -0.000

Var 2(Wheat) 0.008 -0.004 0.000 0.002 0.002 -0.044 0.082 −0.000 0.000

Var 3(Soybean) -0.015 0.008 -0.001 -0.004 -0.004 0.087 -0.163 0.001 -0.000

Var 4(Corn) -0.010 0.005 -0.000 -0.002 -0.002 0.055 -0.102 0.001 -0.000

Var 5(Cotton) -0.002 0.001 -0.000 -0.000 -0.000 0.009 -0.018 0.000 -0.000

Var 6(Sugar) -0.000 0.000 -0.000 -0.000 -0.000 0.000 -0.000 0.000 -0.000

Var 7(Rice) -0.000 0.000 -0.000 -0.000 -0.000 0.002 -0.003 0.000 -0.000
Var 8(Coffee) -0.001 0.000 -0.000 -0.000 -0.000 0.004 -0.007 0.000 -0.000

Var 9(Cocoa) −0.005 0.003 −0.000 −0.001 −0.001 0.031 −0.057 0.000 −0.000
Table 5.7.8 Long Run Matrix

The significant contributions of Cocoa and Sugar in the cointegrating relationships (as seen earlier) are contrasted by
their small influence in the adjustment process, further emphasizing that these variables are more impactful in the long-
term equilibrium rather than short-term corrections. The long-run matrix highlights how the interplay among variables
primarily maintains the broader equilibrium without significant short-term disruptions.

5.7.9 Level Parameters

The level parameters provide a snapshot of the baseline values of the variables in the system. Coffee (Var9) exhibits the
largest baseline value (830.000), significantly outpacing the others, reflecting its high relative pricing or market impact
within the system. In contrast, Brent Oil (Var1, 25.062) and Cocoa (Var7, 5.182) have much lower baseline values,
indicating a relative divergence in their pricing structures compared to agricultural commodities such as Soybean (Var3,
456.500) or Wheat (Var2, 247.509).

Var 1(Brent Oil) 25.062

SE 1 ( 1.365

Var 2(Wheat) 247.509

SE 2 ( 8.487

Var 3(Soybean) 456.500

SE 3 ( 13.678

Var 4(Corn) 200.744

SE 4 ( 7.103

Var 5(Cotton) 51.083

SE 5 ( 1.514

Var 6(Sugar) 6.175

SE 6 ( 0.322

Var 7(Rice) 5.182

SE 7 ( 0.194

Var 8(Coffee) 116.488

SE 8 ( 2.865

Var 9(Cocoa) 830.000

SE 9 ( 23.470

Table 5.7.9 Level Parameter Results

These baseline values are essential for understanding the relative magnitudes of the variables and their potential
sensitivity to systemic changes. High values, such as Coffee, suggest that its pricing behavior might dominate the
equilibrium or require greater adjustments to maintain systemic balance.

5.7.10 Lag Matrices (Gamma_1 and Gamma_2)

The lag matrices capture the short-term dynamics and dependencies across variables. Brent Oil (Var1) shows weak but
noticeable short-term interactions with Sugar (Var6, 0.102 in Gamma_1) and Cocoa (Var7, 0.095 in Gamma_1),
indicating a limited but measurable immediate influence. However, the coefficients for most variables remain small,
reinforcing the subdued short-term effects typical of commodity systems dominated by long-term relationships.

Table 5.7.10 (a) - Lag Matrix 1 (Gamma_1) Results

[Var1- Brent Oil, Var2-Wheat, Var3- Soybean, Var4-Corn, Var5-Cotton, Var6-Sugar,

Var7-Rice, Var8-Coffee, Var9-Cocoa]

Across both lag matrices, the interactions remain consistent, with small coefficients indicating weak short-term
dependencies. This suggests that while the variables are connected through long-term cointegration, immediate price
shocks do not propagate strongly across the system. However, slight deviations, such as Coffee (Var9) influencing Wheat
(Var2, 0.075 in Gamma_2), highlight specific dynamics that may warrant closer scrutiny.

Table 5.7.10 (b) - Lag Matrix 2 (Gamma_2) Results

[Var1- Brent Oil, Var2-Wheat, Var3- Soybean, Var4-Corn, Var5-Cotton, Var6-Sugar,

Var7-Rice, Var8-Coffee, Var9-Cocoa]
 5.7.11 Roots of the Characteristic Polynomial

The characteristic roots provide insights into the system's stability. Most roots are close to 1, indicating a stable
cointegration system with persistent relationships. A few larger roots (e.g., 7.250) suggest that some variables, such as
Coffee or Soybean, might have stronger persistent effects on the system. These roots confirm the long-memory
characteristics seen earlier and validate the appropriateness of the FCVAR model for analyzing these interdependencies.

Table 5.7.11 Roots of the Characteristic Polynomial

Implications

 Long-Term Stability: The small coefficients in the long-run and lag matrices reinforce the system's long-term
stability, driven by persistent but stable cointegrated relationships.
 Short-Term Dynamics: Weak short-term dependencies suggest limited immediate price transmissions among
variables, indicating that price adjustments are primarily long-term phenomena.
 Variable-Specific Roles: Commodities like Coffee and Cocoa play prominent roles in the equilibrium, with
Coffee's high baseline reflecting its market dominance and Cocoa's influence primarily driving long-term
relationships.

These findings underscore the importance of focusing on long-term policies and strategies for managing energy-
agriculture market interactions. Short-term shocks may not significantly disrupt the equilibrium, but persistent dynamics
demand careful monitoring and response.

5.7.12 Multivariate Results

The multivariate Q-statistic for the system is 1881.991 with a highly significant p-value of 0.000, indicating the presence
of serial correlation across the residuals of the system. This result suggests that, at the multivariate level, the model does
not fully eliminate residual autocorrelation, implying that some dependencies among the variables remain unaccounted for
by the selected lag structure or model specification. This finding underscores the need for potential adjustments in the
model's dynamics to address autocorrelation and improve its fit.

5.7.13 Univariate Results

At the individual variable level, the Q-statistics for most variables, such as Brent Oil (Var1, Q=39.399, p=0.000) and
Cocoa (Var7, Q=53.040, p=0.000), indicate significant residual autocorrelation. This pattern is consistent across most
variables, highlighting the persistence of autocorrelation even when considered individually. However, Rice (Var8,
Q=18.087, p=0.113) is an exception, as it shows no significant autocorrelation at the univariate level. This suggests that
the residual structure for Rice is better captured by the model compared to other variables.

Variable Q P-val LM P-val

Multivar 1881.991 0.000

Var1(Brent Oil) 39.399 0.000 14.266 0.284

Var2(Wheat) 50.022 0.000 11.879 0.455

Var3(Soybean) 40.453 0.000 13.224 0.353

Var4(Corn) 34.193 0.001 11.207 0.511

Var5(Cotton) 55.082 0.000 15.738 0.204

Var6(Sugar) 51.678 0.000 17.637 0.127

Var7(Rice) 53.040 0.000 16.724 0.160

Var8(Coffee) 18.087 0.113 9.688 0.643

Var9(Cocoa) 269.472 0.000 11.000 0.529

Table 5.7.13 White Noise Test Results (lag = 12)

5.7.14 Lagrange Multiplier (LM) Test

The LM test results provide further insights into the serial correlation of residuals. Unlike the Q-statistics, the LM test for
all variables shows nonsignificant p-values (p>0.05), indicating no strong evidence of serial correlation in the residuals for
individual variables. For example, Brent Oil (Var1, LM=14.266, p=0.284p = 0.284p=0.284) and Sugar (Var6,
LM=17.637, p=0.127) exhibit no significant serial correlation under the LM test. This discrepancy between the Q-
statistics and LM test suggests that while overall autocorrelation exists in the system, it is not pronounced enough to
dominate the individual variables.

The results highlight a mixed picture of residual autocorrelation in the FCVAR model. The significant multivariate Q-
statistic suggests that systemic autocorrelation persists across the variables, warranting further refinements in the lag
structure or other model specifications. However, the nonsignificant LM test results at the univariate level suggest that
individual variables are less affected by serial correlation, implying that the overall autocorrelation is driven by the
interactions among variables rather than their standalone dynamics.

To address these findings, model adjustments such as exploring additional lags or alternative specifications might be
necessary. Additionally, the results emphasize the importance of considering multivariate dependencies in interpreting
residual dynamics, particularly for commodities like Brent Oil and Cocoa, which exhibit strong residual correlations at the
univariate level. These insights are critical for ensuring the robustness of the model and its reliability in capturing the
dynamics of commodity markets.

5.7.15 Forecasting
The forecast graph provides a visualization of the price trajectories of Brent Oil and key agricultural commodities,
including Wheat, Soybean, Corn, Cotton, Sugar, Rice, Coffee, and Cocoa. The forecast highlights distinct dynamics
across commodities, with varying levels of price volatility and growth trends.

Figure 5.7.15 Commodity Prices Forecast

Key Observations

1. Cocoa’s Price Surge: Among the commodities, Cocoa exhibits the most pronounced price trajectory, with a
sharp surge in the forecast period, especially after the 6000th time point. This could indicate significant market
shocks, supply constraints, or heightened demand, emphasizing its unique volatility compared to other
commodities.
2. Brent Oil and Agricultural Commodities: Brent Oil, Soybean, Corn, and Wheat show more gradual and stable
price growth, reflecting their strong linkages to macroeconomic conditions and steady supply-demand dynamics.
These commodities exhibit lower levels of forecasted volatility compared to Cocoa, suggesting relatively stable
long-term behavior.
3. Relatively Stable Commodities: Commodities such as Sugar, Rice, and Coffee display much flatter trajectories
with minimal fluctuations. This indicates that these commodities may have more predictable pricing behaviors
due to steady production and demand conditions or lower sensitivity to global economic fluctuations.

The forecast underscores the diverse dynamics within the commodity markets. Cocoa's sharp price increase highlights the
need for careful monitoring and risk management strategies for commodities prone to extreme volatility. In contrast, the
steady trends in Brent Oil and key agricultural commodities reflect their resilience and importance as stable economic
indicators. These insights can guide market participants in tailoring their strategies for portfolio diversification, risk
management, and identifying potential investment opportunities across volatile and stable commodities.

5.7.16 Overall Conclusion

The application of the Fractionally Cointegrated Vector Autoregression (FCVAR) model to Brent Oil and agricultural
commodities provides deep insights into the intricate interdependencies and dynamics that govern these markets. The
FCVAR model stands out by effectively capturing long-term relationships while accounting for the fractional integration
of variables, which represents the persistent and slow-decaying nature of shocks. With a cointegration rank of 1, the
results emphasize the presence of a singular dominant equilibrium relationship, likely influenced by shared
macroeconomic factors such as inflation, global trade dynamics, and energy-agriculture linkages. The high fractional
parameter (d≈0.996d) highlights that commodity prices exhibit long-term dependencies that traditional
econometric models often fail to capture.
  Comparison with Other Models

Compared to traditional econometric approaches like Vector Autoregression (VAR) and Cointegrated VAR (CVAR), the
FCVAR model demonstrates significant advantages. VAR models rely on the assumption of stationarity and are incapable
of addressing nonstationary variables that are common in commodity markets. CVAR models, while addressing
nonstationary, assume integer-order integration, thereby overlooking the fractional integration that characterizes long-
memory processes in commodity price dynamics. Studies such as Johansen (1995) and Granger (1986) emphasize the
importance of incorporating fractional integration for commodities, as persistent shocks often span multiple
economic cycles. By incorporating fractional cointegration, the FCVAR model offers superior precision in modelling
these long-term dependencies, leading to more accurate forecasts and actionable insights.

For instance, traditional models may underestimate the prolonged impact of oil price shocks on agricultural commodities,
whereas the FCVAR model explicitly accounts for the persistence and gradual adjustments of prices. Furthermore, as seen
in previous research (e.g., Baumeister & Kilian, 2016), traditional models often oversimplify the relationships between
energy and agriculture, treating them as independent, while the FCVAR framework acknowledges their intricate
interdependencies. The results of this analysis align with findings from empirical studies, which demonstrate that
incorporating fractional dynamics leads to better risk management and forecasting, particularly in volatile markets.

 Investor Perspective

For investors, the results of this study provide actionable insights into commodity market dynamics. The strong positive
linkages between Brent Oil and agricultural commodities like Soybean and Corn suggest that energy price fluctuations
can serve as leading indicators for agricultural commodity markets, particularly those connected to biofuels. This
interconnectedness underscores the need for diversification strategies that consider both energy and agriculture sectors to
hedge against systemic risks.

The forecasted price trajectories further highlight opportunities and risks. Commodities like Cocoa, with high forecasted
volatility, present opportunities for speculative investments but carry substantial risk, necessitating robust hedging
strategies. Conversely, stable commodities like Sugar and Rice are more suited for conservative investors seeking
predictable returns with lower exposure to market shocks. Additionally, the persistent nature of price shocks revealed by
the FCVAR model underscores the importance of long-term investment horizons, as short-term corrections may not fully
capture the recovery or growth potential of these markets.

The FCVAR model's ability to integrate long-term relationships and fractional dynamics offers a robust tool for
policymakers and investors alike. Policymakers can leverage these insights to design strategies that mitigate the impact of
energy price shocks on food markets, ensuring greater price stability and food security. For investors, the model
underscores the importance of monitoring energy-agriculture linkages and adopting tailored strategies based on
commodity-specific behaviors. Future research should extend this framework by incorporating external drivers such as
geopolitical risks, climate change, and technological advancements, which further shape the dynamics of global
commodity markets.

6. Conclusion and Policy Implications

This analysis underscores the economic linkage between the energy and agricultural sectors. The long-term
relationship implies that volatility in the oil market, often driven by geopolitical factors, supply chain disruptions,
or shifts in global demand, can significantly impact agricultural commodity prices. For example, a surge in oil
prices could elevate transportation and input costs (such as fertilizer), subsequently increasing the market
prices for agricultural products like cotton, sugar, and rice.

For policymakers, these insights reinforce the need for integrated risk management strategies that account for
the interconnectedness of energy and food prices. Food security initiatives and inflation control measures
should consider energy prices as a factor influencing food costs, particularly during periods of high oil price
volatility.

From an investment perspective, the FCVAR results offer guidance on portfolio diversification. The
interdependent nature of these markets suggests that traditional diversification may not provide complete
insulation from cross-sectoral price shifts. Therefore, investors seeking to hedge against commodity price
volatility should be aware of potential spillover effects from energy to agricultural markets.

In conclusion, the FCVAR model provides significant insights into the dynamics between Brent oil and
agricultural commodities, highlighting the interdependence of energy and food markets. This
interconnectedness calls for cross-sectoral awareness and strategic planning, particularly during times of
energy crises or agricultural supply disruptions. Monitoring energy market conditions could serve as an early
indicator of agricultural price trends, supporting more resilient and informed decision-making across these
sectors.

References

 Baillie, R. T., & Bollerslev, T. (1994). Cointegration, fractional cointegration, and exchange rate
dynamics. Journal of Finance, 49(2), 737–745. https://doi.org/10.1111/j.1540-6261.1994.tb05159.x
 Baffes, J. (2007). Oil spills on other commodities. Resources Policy, 32(3), 126–134.
https://doi.org/10.1016/j.resourpol.2007.08.004
 Johansen, S., & Nielsen, M. Ø. (2012). Likelihood inference for a fractionally cointegrated vector
autoregressive model. Econometrica, 80(6), 2667–2732. https://doi.org/10.3982/ecta9677
 Nazlioglu, S. (2011). World oil and agricultural commodity prices: Evidence from nonlinear causality.
Energy Policy, 39(5), 2935–2943. https://doi.org/10.1016/j.enpol.2011.03.001
 Nazlioglu, S., Hammoudeh, S., & Gupta, R. (2020). Oil and agricultural commodity markets:
Interdependence and implications. Energy Economics, 90, 104866.
https://doi.org/10.1016/j.eneco.2020.104866
 Serra, T., & Zilberman, D. (2013). Biofuel-related price transmission literature: A review. Energy
Economics, 37, 141–151. https://doi.org/10.1016/j.eneco.2013.02.014
 Tang, K., & Xiong, W. (2012). Index investment and financialization of commodities. Financial Analysts
Journal, 68(6), 54–74. https://doi.org/10.2469/faj.v68.n6.5
 Wright, B. D. (2011). The economics of grain price volatility. Applied Economic Perspectives and Policy,
33(1), 32–58. https://doi.org/10.1093/aepp/ppq033
 Zivot, E., & Wang, J. (2006). Modeling financial time series with S-PLUS. Springer.
https://doi.org/10.1007/0-387-32348-0
 Baumeister, C., & Kilian, L. (2016). Forty years of oil price fluctuations: Why the price of oil
may still surprise us. Journal of Economic Perspectives.
 Johansen, S. (1995). Likelihood-Based Inference in Cointegrated Vector Autoregressive
Models. Oxford University Press.
 Granger, C. W. J. (1986). Developments in the study of cointegrated economic variables.
Oxford Bulletin of Economics and Statistics.
 Ghoshray, A. (2011). A reexamination of trends in primary commodity prices. Journal of
Development Economics.