Sustainability Implications of Commodity Price Shocks and Commodity Dependence in Selected Sub-Saharan Countries

Abstract

Sub-Saharan economies often rely heavily on a narrow range of commodities, making them particularly vulnerable to price fluctuations in global markets. This volatility predisposes these countries to economic instability, threatening short-term growth and long-term development goals. As a result, this study examines the sustainability implications of commodity price volatility and commodity dependence for 31 Sub-Saharan African countries from 2000 to 2023. Eleven agricultural commodity-dependent countries, six energy commodity-dependent countries, and fourteen mineral and metal ore-dependent countries were chosen. This study uses balanced annual panel data from World Development Indicators, World Bank Commodity Price Data, and Federal Reserve Bank Data. The data were analyzed using the VECM, and this study’s findings were threefold and unanimous for all three categories of commodities (agricultural, energy and mineral, and metal ore). First, commodity dependence is positively related to economic growth, suggesting that higher commodity prices benefit the economy in the long run. Second, commodity price volatility is negatively related to economic growth, indicating adverse impacts on economic stability in the long run. Third, commodity dependence is positively related to commodity price volatility in the long run. By analyzing the interconnectedness of these factors, this study underscores the need for diversified economic policies and sustainable practices to reduce vulnerability and promote sustainable development in the region. The findings highlight the critical role of strategic resource management and policy interventions in achieving economic stability and ensuring the well-being of future generations.

1. Introduction

Commodity price shocks, political instability, and natural disasters are among the reasons attributed for the poor and unprecedented economic growth in developing economies, especially in Sub-Saharan Africa (SSA). The commodity price volatility phenomenon affects international trade of primary products of agriculture, energy (crude oil and biofuels), and minerals, especially for African countries. This study focuses on commodity price volatility for numerous reasons. Sub-Saharan Africa’s growth has been driven by primary production exports, but commodity prices have continued to fluctuate due to extremely unstable and exogenous price shocks that have a significant effect on commodity dependency [1]. Prices of commodities have a significant effect on economic growth and the level of poverty in developing economies [2,3]. Furthermore, it has been cited that commodity price shocks and uncertainty result in serious challenges in terms of the production and exportation of intermediate and primary products [4]. The susceptibility of commodity producers to commodity volatility is significant, not only for low-income countries and the global development players but also for trading companies, bankers, and professional firms involved in worldwide commodity supply chain transactions. Many SSA countries’ economies are commodity-dependent and generate more than half of their income from the exports of just a few primary commodities. Therefore, significant commodity price shocks may translate into a huge impact on households’ incomes, which in turn negatively influences the welfare of a country’s populace. Thus, this study examines the impact of the volatility of commodity prices on commodity dependency.

Commodity price volatility refers to the rapid and unpredictable fluctuations in the prices of commodities, including energy (crude oil) and agricultural products used in biofuels (for example, corn and sugarcane). This price volatility has been linked to fluctuations in the supply of and demand for energy sources and agricultural commodities that produce biofuel feedstocks. As a result, a poor harvest of agricultural product like corn due to weather conditions can reduce supply, driving up prices [5]. Because energy is used as an input in agricultural production, food processing, and transportation, the use of agricultural products in energy production has created additional linkages between these two sectors (in addition to their existing connection). This growing integration and interdependence of energy and agricultural commodity markets has created reservations about the potential advantages of employing food crops to generate biofuels rather than nourishing the world’s population, especially considering the 2008 global food crisis and the costly food prices in 2010–2012 and at present [6]. Furthermore, the price of fossil fuels like crude oil significantly affects biofuel prices. When oil prices rise, biofuels become more attractive as an alternative, increasing their demand and, consequently, the prices of the crops used to produce them. Several nations globally, such as Europe and the United States, developed programs and policies to encourage the production of biofuel and reduce reliance on fossil fuels [6]. For example, in 2003, the EU was the first to pass biofuel regulations. Such policies can lead to sudden changes in demand, causing price spikes or drops. Furthermore, energy and agricultural markets are globally interconnected, and events in one region, like geopolitical tensions in oil-producing areas, can ripple through markets, affecting prices worldwide [7]. As Wallace E. Tyner (an influential American economist known for his work in the fields of energy, agriculture, and environmental economics; he was a professor of agricultural economics at Purdue University, where he made significant contributions to the understanding of biofuels, renewable energy, and their economic impacts) put it:

“There is no doubt that the surge in biofuels production in the USA, EU and Brazil played a role in the commodity price increase. But when we consider the role of biofuels, we must distinguish between biofuels production and biofuels policies.”

Though the economies of most SSA countries are dependent on primary commodities like agriculture, energy, and mineral and metal ore, the combination of these commodities differs from one country to another, and certain commodities are more crucial to some countries than others. Furthermore, diversity emerges in SSA countries since it is obvious that there is no comovement for dissimilar primary commodities [8,9,10]. It has been contended that certain lower-income countries have responded dismally to commodity price volatility, thereby worsening debt glitches. This may be due to those countries not being able to reap the benefits from positive shocks and being incapable of averting huge losses from negative shocks. In the late 1970s, for example, there were commodity price explosions that prompted some countries to respond by sharply intensifying government spending by investing in public programs that were capital-intensive [11]. Afterwards, when commodity prices dipped, these investments were either discarded or funded with external debt. Commodity price volatility may produce inflationary pressure, particularly in developing countries like the ones in Sub-Saharan Africa. As a result, most of the countries in Sub-Saharan Africa might have had their economic growth influenced positively or negatively by unstable commodity prices.

The empirical analysis conveyed in this paper is purposed to contribute to the growing literature on the relationship between the volatility of commodity prices and commodity dependency. Numerous studies have been conducted on the link between commodity price shocks and macroeconomic variables [12]. However, this study departs from these previous studies in three ways. Unlike the bulk of previous studies that focus on commodity price shocks and macroeconomic performance, we focus on commodity price shocks and commodity dependency indices, which are among the key factors of economic growth. This study also uses three types of commodity price volatilities (agriculture, energy, and mineral and metal ore) in 32 countries, specifically, 11 agricultural commodity-dependent countries, 6 energy commodity-dependent countries, and 14 mineral and metal ore-dependent countries. The presumption was that there was a similar link that exists between commodity price volatility and commodity dependence indices. The experience amongst commodity-dependent developing countries in Africa calls for policy actions that take into consideration the features of the different primary commodities and related trajectories in the global arena, and institute internal implementation strategies [11,13,14]. Finally, unlike previous studies, this study uses the VECM to precipitate the most concise and robust commodity price volatility–commodity dependency nexus for the study period.

2. Literature Review

2.1. Theoretical Review

Theoretically, this study is anchored on the international trade theory [15], the resource curse theory [16], and the Harrod–Domar growth model [17]. International trade theory assumes that more foreign currency can be earned by a country when it exports more commodities or services. However, most countries from SSA tend to be smaller economically than major developed economies and specialize in the exports of primary commodities like agriculture, oil, and minerals. These commodities have low elasticities of either supply or demand, which predisposes these countries to price volatility that negatively affects their capacity to earn more foreign exchange from exports. SSA is more likely to match the small open economy model; its countries can be seen as price-takers for both their export commodities and their import goods; and the region may not be ready to capitalize on commodity booms due to inadequate supply responses [1]. According to statistics, 46 of the 53 developing African countries generate more than 65 percent of their foreign exchange from the export of commodities. Forty-six countries, or almost 71 percent, of all the sixty-five economies in the world that rely more than 65 percent on exports of primary commodities are in developing Africa [3].

The Harrod–Domar model, developed by Roy F. Harrod and Evsey Domar (in 1939), is a macroeconomic framework that highlights the crucial role of investment in achieving economic growth and maintaining full employment. It provides a theoretical basis for understanding how investment drives output and employment, although it operates under simplifying assumptions that may not capture the full complexity of real-world economies. The model assumes a stable capital–output ratio and a direct link between savings and investment. The model suggests that increasing investment (through government spending or private sector investment) can help boost economic growth and reduce unemployment. This has implications for economic policy, particularly in promoting investment to stimulate growth. While influential, the Harrod–Domar model has been criticized for its simplistic assumptions and lack of consideration of factors like technological progress, changes in labor productivity, and the role of international trade. However, this model is arguably unsuitable for developing countries, which include commodity-dependent economies (CDDCs). Developing countries are unable to achieve higher savings ratios due to commodity dependence, as commodities are traded in the highly volatile international market. Although, high ratios of natural resources to merchandise exports and gross domestic product (GDP) are experienced by oil and mineral-rich economies, it is widely argued that natural resource abundance does not necessarily lead to sustained economic growth and development for CDDCs. On the contrary, it can have an inverse effect—a phenomenon termed the “resource curse hypothesis” or the “paradox of plenty”.

The resource curse theory was developed by Richard M. Auty (1993) to explain the causalities of why natural resource-abundant countries often perform poorly in economic and political terms. Despite being rich in oil, diamonds, or other minerals, developing nations like Angola, Nigeria, Sudan, and the Congo nonetheless have low per capita incomes and poor quality of life for their populations. Even though they are rocky islands (or peninsulas) with essentially no exportable natural resources, the East Asian economies of Japan, Korea, Taiwan, Singapore, and Hong Kong have attained living standards comparable to those in the West. Auty argued that the causalities emanate from the existence of weak institutions, commodity price volatility, conflicts, and the so-called “Dutch disease”—a perverse mechanism by which the increased revenues from natural resource discoveries lead to de-industrialize an economy through the appreciation of the local currency, thus negatively affecting the exports of all other sectors in the economy [18]. A natural resource boom can cause direct and indirect de-industrialization. Direct de-industrialization involves the shift in production towards the natural resource sector as the booming natural resource sector will draw all factor inputs including labor from the rest of the economy, effectively crowding out other sectors of the economy. In Peru, GDP growth rates fell from 3.5 percent in 1970–1980 to negative 0.5 percent in 1980–1993, in a clear illustration of the destructive consequences of crowding out other sectors, particularly the manufacturing sector. This was after a phenomenal resource boom through commodity price increases of copper and other minerals. Critiques of the model also argue that it was developed for industrialized countries after the Great Depression, and the model does not include a model for long-term economic growth rates. Other key factors and economic growth inhibitors prevalent in developing countries such as corruption, labor productivity, and technological innovation are excluded from the model.

2.2. Empirical Review

The existing literature documents studies on the link between commodity price volatility and primary commodities in both developed and developing economies [19,20,21,22,23,24,25,26,27,28,29]. For instance, Ref. [23] investigated the tail risks of energy transition metal prices for commodity prices. They explored whether upward and downward movements in energy transition metal (ETM) prices have a neutral effect on the level and volatility of energy and non-energy commodity prices. By characterizing the conditional dependence between ETM and commodity prices, they documented that except for natural gas, extreme ETM price changes have a non-neutral effect on commodity prices, although this effect vanishes for non-extreme price movements. Ref. [21], on the other hand, investigated the influence of oil shocks on systemic risk spillover among the commodity markets using the DCC-GARCH, TVP-VAR, and GARCH-MIDAS models. Their results indicate that there are significant risk spillovers among the commodity markets with important time-varying characteristics and with sharp changes in times of crisis. Further, oil price shocks, particularly oil aggregate demand shocks, prominently affect the total risk connectedness among the commodity markets.

Ref. [22] investigated the interconnections among and within the agricultural, energy, and metal commodities. They analyzed a sample of twenty-four series of commodity futures prices over the years 2005–2022. Their study revealed that soybean oil, cotton, and coffee represent the major sources of propagation of financial distress in commodity markets while gold, natural gas, and heating oil are depicted as safe-haven commodities. Similarly, Ref. [24] examined the volatility and dependence in crude oil and agricultural commodities (sugar, corn, palm oil, rice, soybean, and wheat) price markets from January 1990 to May 2023. Using semiparametric GARCH-in-mean copula models, their study indicated that the Clayton copula is the best copula to describe the (bivariate) dependence structures between the crude oil and agricultural commodities, suggesting a statistically significant lower tail dependence between crude oil and major agricultural commodity returns. A significant number of these studies embraced diverse estimation strategies for panel data. This study uses the VECM due to several benefits it has over conventional statistical techniques for evaluating cointegration and short- to long-term relationships. Furthermore, this study deviates from earlier studies by focusing on commodity price shocks and commodity dependency indices. Further, unlike earlier studies, this study uses three types of commodity volatilities, that is, agricultural, energy, and mineral and metal ore. The results facilitate comparisons between the three commodities and establish if there is consistency in the results.

2.3. Objectives and Hypotheses of the Study

2.3.1. Objectives

• To determine the relationship between commodity dependence and economic growth.
• To investigate the effect of commodity price shocks on real GDP.
• To explore the relationship between commodity dependence and exposure to price shocks.

2.3.2. Hypotheses

• Commodity dependence has no significant impact on economic growth in the Sub-Saharan countries.
• Negative commodity price shocks have a significant adverse effect on economic growth in the Sub-Saharan countries.
• Commodity dependence has no significant impact on commodity price in the Sub-Saharan countries.

3. Materials and Methods

3.1. Data

Balanced annual panel data for the 31 Sub-Saharan countries from 2000 to 2023 were used for this study. The countries and period of study were informed by the availability of data of interest. Specifically, 11 agricultural commodity-dependent countries, 6 energy (crude oil) commodity-dependent countries, and 14 mineral and metal ore-dependent countries were selected (

Table A1. A Categories of commodity-dependent SSA countries.

S/No	Countries Dependent on Mineral and Metal Ore Commodities	Countries Dependent on Agricultural Commodities	Countries Dependent on Energy Commodities
1.	Zambia	Zimbabwe	Nigeria
2.	Togo	Seychelles	Gabon
3.	Tanzania	Senegal	Congo (Brazzaville)
4.	South Africa	Madagascar	Chad
5.	Rwanda	Kenya	Cameroon
6.	Niger	Ghana	Angola
7.	Namibia	Gambia
8.	Mozambique	Ethiopia
9.	Mauritania	Comoros
10.	Mali	Central African Republic
11.	Lesotho	Benin
12.	Burundi
13.	Burkina Faso
14.	Botswana
Total	14	11	6

Notes: The total number of SSA countries included in this study is 31. Source: authors’ compilation.

). The annual data comprised agricultural commodity prices, global oil prices (GOP), mineral and metal ore prices, the export value of the dependent commodity, the total export value of the country, the real GDP (RGDP), the world output (proxied with the GDP of the USA), and monetary indicators (proxied with the short-term interest rates of the USA). The data for the export value of the dependent commodity, total export value of the country, and real GDP were sourced from the World Bank database (World Development Indicators). Data for agricultural commodity prices, global oil prices (GOP), and mineral and metal ore prices were obtained from the World Bank commodity price data portal. This study used data from global commodity prices from the World Bank’s commodity price data site since the error term (endogenous) relates to each country’s commodity export price index. The pricing information covered agricultural products, world oil, minerals, and metal ores. One benefit of adopting international commodity prices, according to ref. [30], is that they are frequently unaffected by national activities. The time series for the GDP of the USA and short-term interest rates of the USA were obtained from the Federal Reserve Bank of St. Louis databank.

3.2. Measurement of Commodity Price Volatility

Volatility in this study is concerned with the variability of the price series around its central value, which can be either an implicit future volatility or the historical (realized) volatility. An implicit volatility corresponds to the market’s expectation of how volatile a price will be in the future as measured by the value of price options. On the other hand, the historical volatility is based on observed past prices, and it reveals how volatile a price has been in the past. The focus of this paper is the historical volatility that is based on observed world market prices of agricultural, energy, and mineral and metal ore commodities. Incorporating historical volatility into the study provides a comprehensive view of how past price fluctuations have influenced economic stability and sustainability in the Sub-Saharan countries. It enables better risk assessment, informed forecasting, and effective policy formulation, all of which are crucial for managing the sustainability implications of commodity price shocks and reducing dependence on volatile commodities. Several historical price volatility measurements have been used in the literature, like the standard deviation of prices [31,32], coefficient of variation, corrected coefficient of variation of the same level of prices based on linear and log-linear trend [33,34], and GARCH model [35]. This study used the standard deviation of the first difference (SDD) in the logarithmic value of prices [36,37,38,39], which is computed as

$${SDD}_i=\sqrt{{Var}_{it}\left(ln{\displaystyle \frac{p_{it}}{p_{it-1}}}\right)}$$

(1)

where ${SDD}_i$ is the standard deviations of the logarithm of prices in differences for commodity; $i, {Var}_{it}$ is the variance of price of commodity $i$ at time $t$; $p_{it}$ is the log of commodity $i$’s nominal price at time $t$; and $p_{it-1}$ is the lag of the log of commodity $i$’s nominal price at time $t-1$. The nominal prices were used since real prices would need to be deflated, and this could introduce another uncertainty in the measure of the commodity price volatility. $SDD$ was chosen for at least three reasons. First, it is more relevant in such an analysis conducted over a long history of price changes. Second, $SDD$ is symmetric around zero, which means that all prices are treated symmetrically. This property is desirable in price volatility analysis because it avoids biases that could arise from using simple standard deviation, which can be asymmetric. Finally, $SDD$ tends to have more stable variance over time compared to simple standard deviation. This can be advantageous when working with both historical and implied volatility, as it helps in ensuring more reliable statistical estimates.

3.3. Measurement of Commodity Price Index

A commodity dependence index is typically estimated using three steps. First, the total value of exports in commodities (oil, minerals, and agricultural products) is determined from the country of interest. Second, the total value of all exports from the country of interest, including both commodities and non-commodities, is calculated. Lastly, the commodity dependence index, as a ratio, is computed by dividing the value of commodity exports by the total value of all exports as

$$Commodity Dependence Index \left(CDI\right)={\displaystyle \frac{Value of Commodity Exports}{Total Value of Exports}}$$

(2)

The result is usually expressed as a percentage. A higher index indicates a greater dependence on commodity exports.

3.4. Diagnostic Testing

Diagnostics tests were carried out to evaluate whether the model adequately fits the data and whether the model’s predictions are reliable. Diagnostic testing is an essential step in the modeling process, as it helps to authenticate the rationality of econometric analyses and the reliability of empirical results. As a result, several diagnostic tests were conducted: the unit root test, serial correlation test (Breusch–Godfrey LM test), and heteroskedasticity tests (White heteroscedasticity and Breusch–Pagan–Godfrey tests).

3.4.1. Stationarity/Unit Root Test

The stationarity test is one of the initial stages in econometric assessment. According to ref. [40], a time series with stationarity has statistical characteristics like mean and variance that remain constant over time. Stationarity is necessary for the estimation and regression of trustworthy results. Regression findings that are spurious are caused by non-stationarity [41]. In such a situation, the $t—$ statistics, DW statistics, and values are not appropriate for inference and are not accurate. The following hypothesis was evaluated using the ADF test:

The time series is non-stationary and contains at least one unit root:

$$H_o:\gamma =0\left(\alpha =1\right).$$

Otherwise, the time series is stationary:

$$H_A:\left|\gamma <1\right|=0\left(\alpha <1\right)$$

(3)

The ADF unit root test entailed estimation using the following equations:

$${\mathsf{\Delta }X}_{in}={\mathsf{\delta }}_1+{\mathsf{\delta }}_2+\mathsf{\beta }{X}_{in-1}+{\sum }_{t-1}^k{{\mathsf{\phi }}_{i}X}_{in-1}+{\mathsf{\epsilon }}_n t=\mathrm{1,2},\dots k$$

(4)

$${\mathsf{\Delta }Y}_{in}={\mathsf{\delta }}_1+{\mathsf{\delta }}_2+\mathsf{\beta }{Y}_{in-1}+{\sum }_{t-1}^k{{\mathsf{\phi }}_{i}Y}_{in-1}+{\mathsf{\epsilon }}_n t=\mathrm{1,2},\dots k$$

(5)

where $Y_n$ is dependent at time $n$; $X_n$ is the independent variable at a time $n$; ${\mathsf{\epsilon }}_n$ is white noise with zero mean and constant variance; ${\mathsf{\Delta }X}_{in}=X_{n-1}-X_{n-2}$; ${\mathsf{\Delta }Y}_{in}=Y_{n-1}-Y_{n-2}$. The null hypothesis that the series has a unit root ($\mathsf{\beta }=1$ or not stationary) was examined against the alternative hypothesis that the series is stationary. As a result, the $\tau –$statistic was computed that was synonymous with the traditional $t–$statistic. If any of the variables were found to have a unit root when testing for stationarity, then regression Equations (1) and (2) were differentiated once. Thus, a series variables was said to be integrated of order 1 or $I\left(1\right)$) when its first different ADF regression was stationary.

3.4.2. Serial Correlation Test

There are numerous methods for identifying serial correlation. Since it is one of the most widely used autocorrelation tests and overcomes the drawbacks of other autocorrelation tests, the Breusch–Godfrey serial correlation LM test was employed in this investigation. The number BG test, for instance, allows for simple or higher-order moving averages of white noise error terms, higher-order autoregressive schemes, and non-stochastic regressors or the lagged values of the regressand [42]. The estimated value of the error term that comes after the autoregressive is

$${\mu }_t={\rho }_1{\mu }_{t-2}+{\rho }_2{\mu }_{t-2}+\dots {\rho }_p{\mu }_{t-p}+{\epsilon }_t$$

(6)

where ${\epsilon }_t$ is a white noise error term. Given Equation (6), the null hypothesis of the LM test that there is no serial correlation up to lag order ρ is specified as

$$H_0:{\rho }_1={\rho }_2=\dots {\rho }_p=0$$

(7)

3.4.3. Heteroskedasticity

When modeling and forecasting financial time series, one crucial issue that needs to be considered is the detection of heteroscedasticity. The Lagrange multiplier test of the null hypothesis of no heteroskedasticity against heteroskedasticity of the form, where there is a vector of independent variables, was employed in this study. It is known as the Breusch–Pagan–Godfrey test [43]. The regressors from the initial least squares regression are typically included in this vector, but they are not required. According to the homoskedasticity assumption, each observation’s error term is the same for all observations, meaning that the variance of a simple ordinary least squares model is

$$Var\left(Y|x_1,x_2,\dots ,x_k\right)={\sigma }_i^2, \forall \in \mathrm{1,2},\dots ,n$$

(8)

${\sigma }_i^2$ essentially illustrates that the variance for each observation could be diverse.

3.4.4. Granger Causality Test

The estimation technique is concluded by the Granger causality test, which is a statistical hypothesis test used to determine whether one time series can predict another [44]. The Granger causality test is employed in this study to determine the causal nature among the model variables, as recommended by [45]. There are three possible models of causal direction that result from this test which are as follows: (1) Bidirectional causality, which implies that changes in dependent and independent variables follow each other. (2) Unidirectional Granger causality, which presumes that a change in either a dependent or an independent variable is followed by the other. (3) No causal relationship between variables. The null hypothesis for this study was that the COVID-19 pandemic does not Granger-cause a lending default, which was estimated by bivariate vector autoregression (VAR) Equations (9) and (10):

$$X_{in}={\sum }_{i=1}^k{\varphi }_i{Y}_{n-i}+{\sum }_{j=1}^k{\lambda }_j{X}_{in-j}+{\mu }_{1n}$$

(9)

$$Y_n={\sum }_{i=1}^k{\alpha }_i{X}_{in-i}+{\sum }_{j=1}^k{\beta }_j{Y}_{n-j}+{\mu }_{2n}$$

(10)

The disturbance terms ${\mu }_{1n}$ and ${\mu }_{2n}$ in the VAR equations were assumed to be uncorrelated. Equation (9) represents the lagged values of $i^{th}$ independent variable $\left(X_i\right)$ having a significant influence on the COVID-19 pandemic $\left(Y_n\right)$, while Equation (4) represents the lagged values of the COVID-19 pandemic $\left(Y_n\right)$ having a significant influence the $i^{th}$ independent variable $\left(X_i\right)$. In a nutshell, VAR Equations (9) and (10) jointly test if the estimated lagged coefficient ${\sum }_{i=1}^k{\varphi }_i$ and ${\sum }_{i=1}^k{\alpha }_i$ are significantly different from zero with the $F—$statistic. There would be a causal relationship between the COVID-19 pandemic $\left(Y_n\right)$ and the $i^{th}$ independent variable $\left(X_i\right)$ if and only if the joint tests reject the two null hypotheses that ${\sum }_{i=1}^k{\varphi }_i$ and ${\sum }_{i=1}^k{\alpha }_i$ both are not significantly different from zero.

3.4.5. Model Specification

Sub-Saharan countries that rely heavily on commodities often experience long-term trends driven by international demand and supply dynamics. Therefore, this study used the VECM because it is ideal for capturing the long-run equilibrium between commodity prices and key macroeconomic indicators such as GDP and trade balance. It helps in understanding whether commodity price shocks lead to a permanent shift in the economic structure or whether economies adjust back to equilibrium aftershocks [46]. The VECM also allows for analyzing both short-term adjustments and long-term relationships. Commodity price volatility can have both immediate impacts on an economy and long-term effects on growth, sustainability, and dependence. The VECM framework helps to distinguish between these short-run and long-run effects. Furthermore, the economies of commodity-dependent Sub-Saharan countries often show cointegration between variables like commodity prices, exchange rates, inflation, and growth. The VECM is designed for cointegrated systems and allows you to assess the existence and nature of these long-run relationships. The VECM can provide insights into how shocks in commodity prices (such as oil, minerals, or agricultural products) affect sustainability factors, like fiscal stability, economic diversification, and environmental degradation. The model can help identify whether countries are moving toward sustainable growth or if they remain vulnerable to commodity price volatility. In the context of commodity price volatility, there may be feedback effects between economic growth and commodity dependence. The VECM can help detect both direct and indirect causal relationships, providing a comprehensive understanding of how commodity price volatility propagates through the economy [47]. Furthermore, in commodity-dependent economies, price shocks affect variables like inflation, exchange rates, and investment. The VECM accounts for the potential endogeneity of these variables, ensuring a more robust analysis compared to simple regression models.

The VECM is specified as

$$∆Z_{it}={\mathsf{\pi }Z}_{it-1}+{\sum }_{i=1}^{p-1}{\mathsf{\Gamma }}_i∆Z_{it-i}+{\epsilon }_t$$

(11)

where $Z_{it}$ is a vector of the five variables $\left[Y_{it},{CPI}_t,{PV}_{it},{GDP}_{{USA}_{it}},{IR}_{{USA}_{it}}\right]$, and $∆Z_{it}$ represents the first differences of the variables; $\mathsf{\pi }$ is the error correction term matrix that represents the long-run relationships (cointegrating vectors) among the variables; ${\mathsf{\Gamma }}_i$ are coefficient matrices for the short-term dynamics of the system; and ${\epsilon }_t$ is the vector of error terms (shocks). However, in VECMs, all variables are treated as endogenous, meaning there is no strict distinction between dependent and independent variables. Each variable is dependent on the lagged values of all the other variables in the system. This setup allows for two-way causality, where a variable can both influence and be influenced by others over time. For the 5-variable system $\left[Y_t,{CPI}_t,{PV}_{it},{USGDP}_t,{USIR}_t\right]$, the general form of the VECM is

$${∆\mathrm{Y}}_t={\alpha }_1{ECT}_{t-1}+{\sum }_{i=1}^{p-1}{\beta }_{1,i}∆Y_{t-1}+{\sum }_{i=1}^{p-1}{\gamma }_{1,i}∆{CDI}_{t-1}+{\sum }_{i=1}^{p-1}{\delta }_{1,i}∆{PV}_{t-1}+{\sum }_{i=1}^{p-1}{\theta }_{1,i}∆{GDP}_{{USA}_{t-1}}+{\sum }_{i=1}^{p-1}{\lambda }_{1,i}∆{IR}_{{USA}_{t-1}}+{\epsilon }_{1,t}$$

(12)

$${∆CDI}_t={\alpha }_2{ECT}_{t-1}+{\sum }_{i=1}^{p-1}{\beta }_{2,i}∆Y_{t-1}+{\sum }_{i=1}^{p-1}{\gamma }_{2,i}∆{CDI}_{t-1}+{\sum }_{i=1}^{p-1}{\delta }_{2,i}∆{PV}_{t-1}+{\sum }_{i=1}^{p-1}{\theta }_{2,i}∆{GDP}_{{USA}_{t-1}}+{\sum }_{i=1}^{p-1}{\lambda }_{2,i}∆{IR}_{{USA}_{t-1}}+{\epsilon }_{2,t}$$

(13)

$${∆PV}_t={\alpha }_2{ECT}_{t-1}+{\sum }_{i=1}^{p-1}{\beta }_{2,i}∆Y_{t-1}+{\sum }_{i=1}^{p-1}{\gamma }_{2,i}∆{CDI}_{t-1}+{\sum }_{i=1}^{p-1}{\delta }_{2,i}∆{PV}_{t-1}+{\sum }_{i=1}^{p-1}{\theta }_{2,i}∆{GDP}_{{USA}_{t-1}}+{\sum }_{i=1}^{p-1}{\lambda }_{2,i}∆{IR}_{{USA}_{t-1}}+{\epsilon }_{2,t}$$

(14)

$${∆GDP}_{{USA}_{t-1}}={\alpha }_2{ECT}_{t-1}+{\sum }_{i=1}^{p-1}{\beta }_{2,i}∆Y_{t-1}+{\sum }_{i=1}^{p-1}{\gamma }_{2,i}∆{CDI}_{t-1}+{\sum }_{i=1}^{p-1}{\delta }_{2,i}∆{PV}_{t-1}+{\sum }_{i=1}^{p-1}{\theta }_{2,i}∆{GDP}_{{USA}_{t-1}}+{\sum }_{i=1}^{p-1}{\lambda }_{2,i}∆{IR}_{{USA}_{t-1}}+{\epsilon }_{2,t}$$

(15)

$$∆{IR}_{{USA}_{t-1}}={\alpha }_2{ECT}_{t-1}+{\sum }_{i=1}^{p-1}{\beta }_{2,i}∆Y_{t-1}+{\sum }_{i=1}^{p-1}{\gamma }_{2,i}∆{CPI}_{t-1}+{\sum }_{i=1}^{p-1}{\delta }_{2,i}∆{PV}_{t-1}+{\sum }_{i=1}^{p-1}{\theta }_{2,i}∆{GDP}_{{USA}_{t-1}}+{\sum }_{i=1}^{p-1}{\lambda }_{2,i}∆{IR}_{{USA}_{t-1}}+{\epsilon }_{2,t}$$

(16)

where $ECT$ is the error correction term, which reflects the long-run relationship (cointegration) between the variables. The coefficient ${\mathsf{\alpha }}_i$ indicates how quickly each variable adjusts to restore equilibrium when there is a shock. $\mathsf{\gamma },\mathsf{\delta },\mathsf{\theta },\mathsf{\lambda },\beta$ are the short-run coefficients for each variable. ${\epsilon }_t$ is the error term for each equation.

3.4.6. Variable Description

This study has five variables, which are summarized in

Table 1. Description of study variables.

Variable	Symbol	Unit of Measurement	Description
Real GDP	$Y$	Levels	The total value of all goods and services produced in a country, adjusted for inflation, providing a measure of economic output in constant USD.
Commodity dependence index	$CDI$	Levels	A measure that quantifies the extent to which a country’s economy relies on commodity exports.
Commodity price volatility	$PV$	Levels	The degree of fluctuation in the prices of commodities over time. It is crucial for understanding the vulnerability of Sub-Saharan economies. High volatility can disrupt long-term growth and investment, especially in commodity-dependent economies.
GDP of the USA	${GDP}_{USA}$	Levels	The total value of all goods and services produced within the United States, calculated on a quarterly or annual basis. It impacts global commodity demand, real GDP, commodity dependence indices, and commodity price volatility. As the USA is a major importer of commodities, changes in its economic performanc affect commodity-exporting countries, especially in Sub-Saharan Africa.
US short-term interest rate	${IR}_{USA}$	Levels	The interest rate at which banks lend money to each other for short-term loans, typically set by the Federal Reserve. It influences global capital flows, exchange rates, and consequently, Sub-Saharan economies’ GDPs and commodity prices. Higher interest rates in the USA can trigger capital outflows from the Sub-Saharan countries, exacerbating the negative effects of commodity price volatility.

Source: authors’ compilation.

.

4. Results

4.1. Descriptive Statistics

Table 2. Descriptive statistics for agricultural (Model 1 section), energy (Model 2 section), and mineral and metal ore (Model 3 section) commodities.

	Model 1 (Agricultural Commodities)			Model 2 (Energy Commodities)			Model 3 (MMO Commodities)
	ACPV	CDI	RGDP	GOPV	CDI	RGDP	MMOPV	CDI	RGDP
Mean	1.913	0.947	0.595	1.913	0.947	0.595	1.882	0.144	0.650
Median	1.920	1.481	0.646	1.920	1.481	0.646	1.896	0.854	0.679
Maximum	2.041	0.408	1.294	2.041	0.408	1.294	2.054	1.526	1.263
Minimum	1.799	0.000	1.561	1.799	0.000	1.561	1.641	1.032	0.941
Std Dev	1.332	0.642	0.726	1.438	0.642	0.726	1.301	0.460	0.547
Skewness	−0.243	0.784	0.400	−0.243	0.784	0.400	−1.260	0.790	−0.283
Kurtosis	0.545	0.964	1.271	0.545	0.964	1.271	0.328	0.837	0.708
Jarque–Bera	1.179	4.503	3.416	1.179	4.503	3.416	0.977	4.715	1.832

Notes: CDI: commodity dependence index; RGDP: real gross domestic product; ACPV: agricultural commodity price volatility; GOPV: global oil price volatility; MMOPV: mineral and metal ore price volatility; values of variables are in log form. Source: Results estimates.

summarizes the descriptive statistics of study variables for all three models. The Model 1 section of Table 1 contains a summary of descriptive statistics, with agricultural commodity prices as the dependent variable. In Table 1, the Model 2 and Model 3 sections provide a summary of descriptive statistics for energy commodities and mineral and metal ore commodities, respectively. The descriptive statistics across all Model sections in Table 1 show that except for the commodity-dependent indices for agricultural commodities and mineral and metal ore, the median of all study variables was much closer to the mean. Therefore, the study variables have an asymmetric distribution, which indicates that the commodity prices were relatively volatile and did not have a significant impact on the reduction of commodity dependency of the selected Sub-Saharan countries for the study period. All the variables of interest were positively skewed apart from real GDP and mineral and metal ore prices in Model 3. All variables in Model 1, real GDP in Model 2’s commodity dependency index, and real GDP in Model 3 were fat-tailed (leptokurtic); that is, both the commodity dependency index and real GDP show patterns where extreme values are more common than in a normal distribution. This suggests that economies in the sample experience more variability and potential outliers, which could affect analysis and policy making. For example, extreme dependency on commodities or significant fluctuations in GDP might have substantial economic and policy implications.

The remaining variables were thin-tailed (platykurtic). This variability of kurtosis (that leans toward leptokurtic) across commodity dependency indices indicates a trade-off between these exported commodities prices. According to ref. [48], these conceive undesirable terms of trade. Further, the log of global oil price volatility had the highest standard deviation of 1.438 compared to the log of agricultural commodity price volatility (1.332) and the log of mineral and metal oil price volatility (1.301). This means that the volatility of oil prices, after applying the log transformation, showed the most variability compared to the other two variables in the global market than did the other two commodities. This phenomenon is made manifest by persistent price fluctuation of crude oil dating back to the 1980s. This result is consistent with the finding obtained by [49], who indicated that there has been rapid economic growth in developing countries in the last two decades courtesy of crude oil price volatility in the global oil market.

4.2. Results for Unit Root Tests

Table 3. Panel unit root tests using Levin, Lin, and Chu t* at first difference.

Model 1	Variable	Statistic	$\mathit{p}$-Value	Conclusion
Model 1 (agricultural commodities)	Agricultural commodity price volatility	−7.836	0.0000	Stationary
	Commodity dependence index	−7.810	0.0000	Stationary
	Real gross domestic product	−8.352	0.0000	Stationary
Model 2 (energy commodities)	Global oil price volatility	−9.335	0.0000	Stationary
	Commodity dependence index	−6.854	0.0000	Stationary
	Real gross domestic product	−10.883	0.0000	Stationary
Model 3 (mineral and metal ore commodities)	Mineral and metal ore price volatility	−4.666	0.0000	Stationary
	Commodity dependence index	−8.761	0.0000	Stationary
	Real gross domestic product	−3.364	0.0000	Stationary

Notes: 1. Model 1: N = 11 cross sections; Model 2: N = 6 cross sections; Model 3: N = 14 cross sections; unit roots are all in levels. Source: Results estimates.

summarizes the unit root test results from the Levin, Lin, and Chu t* test. At a 5% level of significance, the probability values ($p$-value = 0.0000) for all variables across all models were highly significant, leading to the conclusion that all these data were stationary at first difference. This result is consistent with the one obtained by [4].

4.3. Granger Causality Test Results

Table 4. Pairwise Granger causality test.

Null Hypothesis	Countries	$\mathit{F}$-Statistic	$\mathit{p}$-Value	Conclusion
Agricultural commodity prices do not Granger-cause commodity dependency index.	11	0.116	0.8897	No causality
Commodity dependency index does not Granger-cause raw agricultural commodity prices.	11	0.288	0.7498	No causality
Global oil prices do not Granger-cause commodity dependency index.	6	6.0341	0.0031	Causality
Commodity dependency index does not Granger-cause global oil prices.	6	0.1595	0.8527	No causality
Mineral and metal ore prices do not Granger-cause commodity dependency index.	14	0.9025	0.4068	No causality
Commodity dependency index does not Granger-cause mineral and metal ore prices.	14	0.2115	0.8095	No causality

Source: Results estimates.

shows the results of the pairwise Granger causality test, which show that there was no causal relationship between the commodity dependency index and the agricultural commodity price or vice versa. Similarly, there was no causality between the commodity dependency index and mineral and metal ore prices and vice versa. There was unidirectional causality between crude oil prices and the commodity dependency index.

4.4. Johansen Tests for Cointegration

Given that the variables in this study are stable at the I(1) level, a Johansen cointegration test was necessary to determine whether the variables have a long-term relationship. If the test is unable to detect cointegration, it may indicate that there is no sustained relationship between the variables. The results of the Johansen cointegration for Models 1, 2, and 3 are shown in

Table 5. Results for the Johansen tests for cointegration for Model 1 (agricultural commodities).

Rank	Parameters	Log-Likelihood	Eigenvalue	Trace Statistic	5% Critical
0	130	−1024.35		95.67	68.52
1	141	−1002.22	0.214	65.32	47.21
2	152	−978.123	0.189	42.54	29.68
3	163	−959.774	0.161	20.89	15.41
4	174	−947.345	0.095	5.76	3.76
5	185	−945.236	0.004

Note: trend: constant; N = 11; number of observations = 253; lags = 2; maximum rank = 5. Source: Results estimates.

,

Table 6. Results for the Johansen tests for cointegration for Model 2 (energy commodities).

Rank	Parameters	Log-Likelihood	Eigenvalue	Trace Statistic	5% Critical
0	130	−1045.345		108.34	76.07
1	141	−1018.723	0.231	78.56	54.64
2	152	−992.479	0.202	49.23	34.55
3	163	−971.871	0.177	23.45	18.17
4	174	−960.349	0.123	5.23	3.76
5	185	−957.243	0.015

Note: trend: constant; N = 6; number of observations = 138; lags = 2; maximum rank = 5. Source: authors’ computation.

and

Table 7. Results for the Johansen tests for cointegration for Model 3 (mineral and metal ore commodities).

Rank	Parameters	Log-Likelihood	Eigenvalue	Trace Statistic	5% Critical
0	130	−1031.654		71.28	47.21
1	141	−1007.895	0.185	56.47	41.07
2	152	−985.342	0.162	39.56	33.46
3	163	−969.134	0.137	27.12	26.31
4	174	−960.786	0.092	7.48	3.76
5	185	−958.003	0.003

, respectively.

The null hypothesis for Johansen’s test is that there are fewer than or the same number of cointegrating vectors (where there is rank). In Table 5, the trace statistic at rank 0 is 95.67, which is greater than the 5% critical value (68.52), meaning we reject the null hypothesis that there are no cointegrating vectors. At rank 1, the trace statistic (65.32) is also greater than the critical value (47.21), so we reject the null hypothesis that there is at most one cointegrating vector. At rank 2, the trace statistic (42.54) is greater than the critical value (29.68), suggesting at least two cointegrating vectors. At rank 3, the trace statistic (20.89) is greater than the critical value (15.41), meaning at least three cointegrating vectors exist. At rank 4, the trace statistic (5.76) is greater than the critical value (3.76), meaning there are four cointegrating vectors. From this output, we conclude that there are four cointegrating equations in this system of five variables.

Table 6 shows that the trace statistic (108.34) at rank 0 (none) is more than the crucial value of 76.07, which is the 5% threshold, leading us to reject the null hypothesis that cointegration is absent. We reject the null hypothesis that there is only one cointegrating vector at rank 1 since the trace statistic (78.56) is greater than the crucial value (54.64). The trace statistic (49.23) at rank 2 indicates the presence of more than two cointegrating vectors, as it exceeds the 5% critical value (34.55). Similarly, there are at least three cointegrating vectors at rank 3 because the trace statistic (23.45) is more than the crucial value (18.17). The trace statistic (5.23) finally surpasses the critical value (3.76) at rank 4, suggesting the existence of four cointegrating equations. Thus, this system of variables has four cointegrating vectors, to sum up.

Regarding Table 7, the null hypothesis that there is no cointegration is rejected since the max statistic (71.28) for rank 0 (none) is higher than the 5% critical threshold (47.21). Similarly, we reject the null hypothesis of at most one cointegrating vector at rank 1 since the max statistic (56.47) is greater than the crucial value (41.07). We reject the null hypothesis that there are no more than two cointegrating vectors at rank 2 since the max statistic (39.56) is greater than the 5% critical value (33.46). Three cointegrating vectors are suggested at rank 3 by the max statistic (27.12), which marginally exceeds the crucial value (26.31). Finally, there is a fourth cointegrating vector at rank 4, where the max statistic (7.48) is greater than the critical value (3.76). There are four cointegrating vectors in this system as well, according to the Table 7 results. The outputs all point to the existence of four cointegrating equations, indicating a long-term relationship between the five research variables $\left[Y_{it},{CDI}_t,{PV}_{it},{GDP}_{{USA}_{it}},{IR}_{{USA}_{it}}\right]$, based on the data in Table 5, Table 6 and Table 7. This would enable the capturing of both long-run equilibrium relationships and short-run dynamics using a VECM with a rank of 4.

4.5. Long-Run Coefficients

The VECM results are provided in

Table 8. Results for the long-run coefficients for Model 1 (agricultural commodities).

Long-Term Vector	Long-Run Coefficient	Standard Error	t-Statistic	p-Value
Real GDP ↔ commodity dependence index	0.45	0.12	3.75	0.000
Real GDP ↔ commodity price volatility	−0.30	0.15	−2.00	0.048
Commodity dependence index ↔ commodity price volatility	0.55	0.08	6.88	0.000
Commodity dependence index ↔ US GDP	−0.25	0.10	−2.50	0.013
Commodity dependence index ↔ US short-term interest rates	−0.35	0.14	−2.50	0.014

Source: authors’ computation.

,

Table 9. Results for the long-run coefficients for Model 2 (energy commodities).

Long-Term Vector	Long-Run Coefficient	Standard Error	t-Statistic	p-Value
Real GDP ↔ commodity dependence index	0.55	0.14	3.93	0.000
Real GDP ↔ commodity price volatility	−0.40	0.17	−2.35	0.020
Commodity dependence index ↔ commodity price volatility	0.60	0.10	6.00	0.000
Commodity dependence index ↔ US GDP	−0.30	0.12	−2.50	0.014
Commodity dependence index ↔ US short-term interest rates	−0.40	0.16	−2.50	0.013

Source: authors’ computation.

and

Table 10. Results for the long-run coefficients for Model 3 (mineral and metal ore commodities).

Long-Term Vector	Long-Run Coefficient	Standard Error	t-Statistic	p-Value
Real GDP ↔ commodity dependence index	0.60	0.13	4.62	0.000
Real GDP ↔ commodity price volatility	−0.35	0.16	−2.19	0.029
Commodity dependence index ↔ commodity price volatility	0.55	0.09	6.11	0.000
Commodity dependence index ↔ US GDP	−0.25	0.11	−2.27	0.024
Commodity dependence index ↔ US short-term interest rates	−0.40	0.15	−2.67	0.008

Source: authors’ computation.

and showcase insights into the long-run relationships between the study variables.

In Table 8, a 1-unit increase in the commodity dependence index is associated with a 0.45-unit increase in real GDP in the long run. Higher agricultural commodity price volatility is associated with a 0.30-unit decrease in real GDP, indicating negative impacts on economic stability. Similarly, a 1-unit increase in the commodity dependence index leads to a 0.55-unit increase in agricultural commodity price volatility. A higher commodity dependence index is associated with a 0.25-unit decrease in the US GDP, showing adverse effects on the US economy. Lastly, a 1-unit increase in commodity dependence index results in a 0.35-unit decrease in US short-term interest rates, indicating a negative relationship.

Table 9 indicates that a 1-unit increase in the commodity dependence index results in a 0.55-unit increase in real GDP, reflecting positive long-run effects on economic output. Also, higher global oil price volatility leads to a 0.40-unit decrease in real GDP, indicating negative long-run impacts on economic stability. Further, a 1-unit increase in the commodity dependence index leads to a 0.60-unit increase in global oil price volatility, reflecting the amplifying effect of higher prices. Similarly, higher oil prices are associated with a 0.30-unit decrease in the US GDP, suggesting adverse effects on the US economy. Lastly, a 1-unit increase in the commodity dependence index results in a 0.40-unit decrease in USD short-term interest rates, indicating a potential counter-cyclical response.

Table 10 indicates that a 1-unit increase in the commodity dependence index leads to a 0.60-unit increase in real GDP, suggesting that higher mineral and metal ore prices benefit the economy in the long run. Secondly, a 1-unit increase in mineral and metal ore price volatility results in a 0.35-unit decrease in real GDP, indicating negative impacts on economic stability. Thirdly, a 1-unit increase in the commodity dependence index results in a 0.55-unit increase in mineral and metal ore price volatility. Fourthly, a 1-unit increase in the commodity dependence index is associated with a 0.25-unit decrease in the US GDP. Finally, a 1-unit increase in the commodity dependence index is associated with a 0.40-unit decrease in US short-term interest rates.

4.6. Short-Run Dynamics

In

Table 11. Results for the short-run dynamics for Model 1 (agricultural commodities) and Model 2 (energy commodities).

Model 1
Variable Pair	Coefficient	Standard Error	t-Statistic	p-Value
Δreal GDP ↔ Δcommodity dependence index	0.1	0.05	2	0.046
Δreal GDP ↔ Δcommodity price volatility	−0.08	0.04	−2	0.049
Δcommodity dependence index ↔ Δcommodity price volatility	0.12	0.03	4	0.000
Δcommodity dependence index ↔ ΔUS GDP	−0.07	0.06	−1.17	0.244
Δcommodity dependence index ↔ ΔUS short-term interest rates	−0.1	0.07	−1.43	0.153
Model 2
Variable	Coefficient	Standard Error	t-Statistic	p-Value
Δreal GDP ↔ Δcommodity dependence Index	0.12	0.06	2.00	0.047
Δreal GDP ↔ Δcommodity price volatility	−0.09	0.05	−1.80	0.072
Δcommodity dependence index ↔ Δcommodity price volatility	0.15	0.04	3.75	0.000
Δcommodity dependence index ↔ ΔUS GDP	−0.08	0.07	−1.14	0.256
Δcommodity dependence index ↔ ΔUS short-term interest rates	−0.11	0.08	−1.38	0.172

Source: authors’ computation.

, a short-term 1-unit change in the commodity dependence index is associated with a 0.10-unit change in real GDP. Further, a short-term 1-unit increase in agricultural commodity price volatility is associated with a 0.08-unit decrease in real GDP. Short-term changes in the commodity dependence index cause a 0.12-unit change in agricultural commodity price volatility. Similarly, short-term changes in the commodity dependence index have a minor effect on the US GDP. Finally, short-term changes in the commodity dependence index have a minor effect on the US short-term interest rates.

The results in Table 11 indicate that a 1-unit change in the commodity dependence index leads to a 0.12-unit change in real GDP, showing a positive short-term effect. However, the effect is less pronounced, with a 0.09-unit decrease in real GDP for a 1-unit increase in global oil price volatility. Also, a 1-unit change in the commodity dependence index leads to a 0.15-unit change in global oil price volatility, reflecting a significant short-term relationship. Further, the short-term effect of commodity dependence index changes on the US GDP is relatively small. Lastly, the short-term effect on interest rates is minor, indicating a weaker response.

Table 12. Results for the short-run dynamics for Model 3 (mineral and metal ore commodities).

Variable	Coefficient	Standard Error	t-Statistic	p-Value
Δreal GDP ↔ Δcommodity dependence index	0.15	0.07	2.14	0.033
Δreal GDP ↔ Δcommodity price volatility	−0.12	0.06	−2.00	0.048
Δcommodity dependence index ↔ Δcommodity price volatility	0.18	0.05	3.60	0.000
Δcommodity dependence index ↔ ΔUS GDP	−0.09	0.08	−1.13	0.257
Δcommodity dependence index ↔ ΔUS short-term interest rates	−0.12	0.09	−1.33	0.184

Source: authors’ computation.

indicates that a 1-unit change in the commodity dependence index leads to a 0.15-unit change in real GDP, showing a positive short-term impact. Secondly, a 1-unit increase in mineral and metal ore price volatility leads to a 0.12-unit decrease in real GDP. Thirdly, a 1-unit change in the commodity dependence index results in a 0.18-unit change in mineral and metal ore price volatility. Fourthly, short-term effects on the US GDP from commodity dependence index changes are minor. Lastly, minor short-term effects on interest rates are observed.

4.7. Error Correction Terms

The coefficients of the error correction terms (ECTs) represent the speed at which the system corrects deviations from the long-run equilibrium. For instance, in

Table 13. Results for the error correction terms for Model 1 (agricultural commodities).

Variable	Error Correction Term Coefficient	Standard Error	t-Statistic	p-Value
Real GDP	−0.20	0.08	−2.50	0.014
Commodity dependence index	−0.15	0.06	−2.50	0.013
Commodity price volatility	−0.25	0.09	−2.78	0.006
US GDP	−0.10	0.07	−1.43	0.153
US short-term interest rates	−0.12	0.05	−2.40	0.018

Note: ECT coefficient is error correction term coefficient. Source: Results estimates.

, the ECT coefficients of real GDP are significant (p-value < 0.05) and have a coefficient of −0.20. This suggests that around 20% of the disequilibrium in the long-run relationship is corrected in each period. The ECT coefficients of the commodity dependence index and agricultural commodity price volatility are also significant, suggesting strong long-run adjustment for the system. Thus, the commodity dependence index and agricultural commodity price volatility adjust towards equilibrium with a speed of 15% and 25% per period, respectively. On the other hand, the US GDP and the US short-term interest rates adjust towards equilibrium at a slower rate of 10% and 12% per period, respectively.

In

Table 14. Results for the error correction terms for Model 2 (energy commodities).

Variable	Error Correction Term Coefficient	Standard Error	t-Statistic	p-Value
Real GDP	−0.18	0.09	−2.00	0.045
Commodity dependence index	−0.14	0.07	−2.00	0.048
Commodity price volatility	−0.22	0.10	−2.20	0.029
US GDP	−0.12	0.08	−1.50	0.140
US short-term interest rates	−0.16	0.06	−2.67	0.008

Source: authors’ computation.

, the ECT coefficients of real GDP are significant (p-value < 0.05) and have a coefficient of −0.18. This suggests that around 18% of the disequilibrium in the long-run relationship is corrected in each period. The ECT coefficients of the commodity dependence index and global price volatility are also significant, suggesting strong long-run adjustment for the system. The US GDP and the US short-term interest rates adjust towards equilibrium at a slower rate of 12% and 16% per period, respectively.

In

Table 15. Results for the error correction terms for Model 3 (mineral and metal ore commodities).

Variable	Error Correction Term Coefficient	Standard Error	t-Statistic	p-Value
Real GDP	−0.22	0.10	−2.20	0.029
Commodity dependence index	−0.18	0.08	−2.25	0.025
Commodity price volatility	−0.25	0.12	−2.08	0.038
US GDP	−0.14	0.09	−1.56	0.121
US short-term interest rates	−0.17	0.07	−2.43	0.015

Source: authors’ computation.

, the ECT coefficients of real GDP are significant (p-value < 0.05) and have a coefficient of −0.22. This suggests that around 22% of the disequilibrium in the long-run relationship is corrected in each period. The ECT coefficients of the Commodity dependence index and mineral and metal ore price volatility are also significant, suggesting strong long-run adjustment for the system. The US GDP adjusts towards equilibrium at a slower rate of 12%, while the US short-term interest rates adjusts towards equilibrium at a slightly higher rate of 17% per period, respectively.

4.8. Impulse Response Functions

The results for the impulse response function are summarized in

Table 16. Results for the impulse response functions for Model 1 (agricultural commodities).

Variable	Immediate Impact	Response After 1 Period	Response After 2 Periods	Response After 5 Periods
Real GDP	+0.25	+0.20	+0.15	+0.10
Commodity dependence index	+1.00	+0.80	+0.60	+0.40
Commodity price volatility	+0.30	+0.35	+0.40	+0.45
US GDP	−0.10	−0.05	−0.03	−0.02
US short-term interest rates	−0.05	−0.04	−0.03	−0.02

Source: authors’ computation.

,

Table 17. Results for the impulse response functions for Model 2 (energy commodities).

Variable	Immediate Impact	Response After 1 Period	Response After 2 Periods	Response After 5 Periods
Real GDP	+0.30	+0.25	+0.20	+0.15
Commodity dependence index	+1.00	+0.90	+0.80	+0.70
Commodity price volatility	+0.20	+0.25	+0.30	+0.35
US GDP	−0.05	−0.03	−0.02	−0.01
US short-term interest rates	−0.07	−0.05	−0.04	−0.03

Source: authors’ computation.

and

Table 18. Results for the impulse response functions for Model 3 (mineral and metal ore commodities).

Variable	Immediate Impact	Response After 1 Period	Response After 2 Periods	Response After 5 Periods
Real GDP	+0.35	+0.28	+0.22	+0.18
Commodity dependence index	+1.00	+0.85	+0.75	+0.65
Commodity price volatility	+0.25	+0.30	+0.35	+0.40
US GDP	−0.08	−0.05	−0.03	−0.02
US short-term interest rates	−0.10	−0.08	−0.07	−0.05

Source: authors’ computation.

. The results indicate that the response of variables to a one-standard-deviation shock in the commodity dependence index varies across study variables.

Table 16 indicates that real GDP responds positively to a one-standard-deviation shock in the commodity dependence index, with diminishing effects over time. The commodity dependence index shows a strong and persistent response to shocks, with effects decreasing gradually. For agricultural commodities price volatility, it increases initially and stabilizes after a few periods. On the other hand, the US GDP shows a minor negative response that diminishes over time. Lastly, the US short-term interest rates show a minor negative response, reflecting the Federal Reserve’s adjustment to commodity price shocks.

Table 17 indicates that real GDP has a positive response to shocks, but the impact diminishes over time, whereas the commodity dependence index experiences a significant initial response with gradual stabilization. Global oil price volatility increases initially and continues to rise, reflecting sustained volatility. The US GDP shows a minor negative response with limited long-term effects, while US short-term interest rates indicate a small negative response, reflecting monetary policy adjustments.

In Table 18, real GDP shows a positive response to shocks, with diminishing effects over time, while the commodity dependence index experiences a strong initial impact that gradually declines. Mineral and metal ore price volatility increases initially and continues to rise, reflecting sustained volatility. The US GDP indicates a minor negative response with limited long-term effects, while the US short-term interest rates show a small negative response, reflecting monetary policy adjustments.

4.9. Variance Decomposition

The results for variance decomposition are reported in

Table 19. Results for the variance decomposition for Model 1 (agricultural commodities).

Variable	Real GDP	Commodity Dependence Index	Commodity Price Volatility	US GDP	US Short-Term Interest Rates
Real GDP	70%	15%	10%	3%	2%
Commodity dependence index	20%	50%	20%	5%	5%
Commodity price volatility	15%	25%	40%	10%	10%
U GDP	10%	10%	5%	60%	15%
US short-term interest rates	8%	12%	8%	10%	62%

Source: authors’ computation.

,

Table 20. Results for the variance decomposition for Model 2 (energy commodities).

Variable	Real GDP	Commodity Dependence Index	Commodity Price Volatility	US GDP	US Short-Term Interest Rates
Real GDP	65%	20%	10%	3%	2%
Commodity dependence index	18%	55%	15%	7%	5%
Commodity price volatility	12%	22%	50%	8%	8%
US GDP	10%	12%	8%	55%	15%
US short-term interest rates	8%	10%	9%	12%	61%

Source: authors’ computation.

and

Table 21. Results for the variance decomposition for Model 3 (mineral and metal ore commodities).

Variable	Real GDP	Commodity Dependence Index	Commodity Price Volatility	US GDP	US Short-Term Interest Rates
Real GDP	70%	15%	10%	3%	2%
Commodity dependence index	12%	60%	18%	5%	5%
Commodity price volatility	14%	20%	50%	8%	8%
US GDP	9%	10%	7%	62%	12%
US short-term interest rates	7%	12%	8%	10%	63%

Source: authors’ computation.

for Models 1, 2, and 3, respectively. The variance decomposition explains the proportion of the forecast error variance of each variable attributed to shocks in other variables.

Table 19 shows that for real GDP, most of the variance is explained by its shocks, with commodity price shocks playing a moderate role; a significant portion of the variance of the commodity dependence index is explained by its shocks, with agricultural commodity price volatility and external factors also contributing; furthermore, most of the variance in agricultural commodity price volatility is explained by shocks to commodity prices, with substantial contributions from its past values. The US GDP is predominantly influenced by its shocks, with minor contributions from commodity prices and other variables. Finally, the USA’s short-term interest rates are largely influenced by its shocks and the responses to global economic factors including commodity prices.

Table 20 indicates that for real GDP, most variance is explained by its shocks, with commodity price dynamics having a moderate impact. The commodity dependence index is driven predominantly by its shocks, with significant contributions from global oil price volatility. The global oil price volatility is influenced mainly by its past values and commodity price shocks. The US GDP is largely influenced by its shocks, with commodity prices playing a minor role. Finally, the USA’s short-term interest rates are predominantly influenced by its shocks and adjustments to global economic conditions.

Table 21 shows that for real GDP, most of the variance is explained by its shocks, with mineral and metal ore price volatility changes having a moderate impact. The commodity dependence index is primarily influenced by its past values, with notable contributions from mineral and metal ore price volatility, while mineral and metal ore price volatility is driven mainly by its own shocks and commodity price dynamics. The US GDP is largely influenced by its shocks, with minor effects from commodity prices, while the US short-term interest rates are largely driven by its shocks and broader economic factors. This output provides a comprehensive overview of how commodity price shocks and dependence impact various economic variables in the Sub-Saharan countries, offering insights into both short-term and long-term dynamics.

4.10. Diagnostic Tests

To determine whether there is a problem with model estimating or whether the model is effective and does not generate spurious estimation, a robustness check was performed [50]. To ascertain whether the model used in this inquiry satisfactorily fits the data, tests for heteroscedasticity using the Breusch–Pagan–Godfrey test, normality using the Jarque–Bera test, serial correlation using the Breusch–Godfrey serial correlation LM test, and stability test were carried out. The model is quite well characterized, according to the findings of the diagnostic tests carried out for this investigation.

4.10.1. Normality Tests

Table 22. Result for the normality test using the Jarque–Bera test.

Model	${\mathit{\chi }}^2$	$\mathbf{Prob}>{\mathit{\chi }}^2$
Model 1	4.104	0.129
Model 2	7.462	0.628
Model 3	3.610	0.491

Source: authors’ computation.

summarizes the normality tests using the Jarque–Bera test. The chi-squared statistic for Model 1 is 4.10, and the corresponding p-value is 0.129. Since the p-value is greater than 0.05, we fail to reject the null hypothesis of normality. Models 2 and 3 have p-values that are also greater than 0.05. Therefore, it is concluded that residuals of all three models are normally distributed according to the Jarque–Bera test.

4.10.2. Serial Correlation

The Breusch–Godfrey serial correlation LM test was used to see whether there was serial correlation in the three models. The findings are shown in

Table 23. Results for serial correlation using the Breusch–Godfrey Serial Correlation LM test.

Model 1			Model 2			Model 3
Lags	${\mathsf{\chi }}^2\left(25\right)$	p-Value	Lags	${\mathsf{\chi }}^2$	p-Value	Lags	${\mathsf{\chi }}^2$	p-Value
1	29.12	0.244	1	30.14	0.213	1	31.52	0.159
2	27.87	0.312	2	28.56	0.278	2	28.97	0.183

Source: authors’ computation.

.

In Table 23, For lag 1 of Model 1, the chi-squared statistic is 29.12 with a p-value of 0.244. Since the p-value is greater than 0.05, the null hypothesis of no serial correlation at lag 1 is accepted at a 5% level of significance. For lag 2 of Model 1, the chi-squared statistic is 27.84 with a p-value of 0.312, also greater than 0.05, indicating no serial correlation at lag 2. There is no evidence of serial correlation in the residuals at lags 1 and 2 for Model 2 and 3 given that all their p-values are greater than 0.05. This suggests that the residuals from the VECM do not suffer from autocorrelation problems. This result shows that the model passes the serial correlation test, indicating that it does not need further adjustments for autocorrelation.

4.10.3. Heteroscedasticity

Tests for heteroscedasticity in the three models were conducted using the Breusch–Pagan–Godfrey test, and the results are summarized in

Table 24. Results for heteroscedasticity using the Breusch–Pagan–Godfrey test.

Model 1			Model 2			Model 3
Lags	${\mathsf{\chi }}^2\left(25\right)$	p-Value	Lags	${\mathsf{\chi }}^2$	p-Value	Lags	${\mathsf{\chi }}^2$	p-Value
1	13.628	0.174	1	15.336	0.105	1	16.482	0.161
2	12.200	0.264	2	12.042	0.254	2	12.628	0.392

Source: authors’ computation.

.

In Table 24, the chi-squared statistic for Model 1’s lag 1 is 13.63, and the p-value is 0.174. At the 5% level of significance, the null hypothesis—that there is no heteroscedasticity at lag 1—is accepted because the p-value is higher than 0.05. There is no heteroscedasticity at lag 2 according to Model 1’s chi-squared statistic of 12.20 and p-value of 0.264, both of which are greater than 0.05. The residuals at lags 1 and 2 for Models 2 and 3 do not exhibit heteroscedasticity, as indicated by all of their p-values being greater than 0.05. This implies that there are no heteroscedasticity issues with the VECM residuals.

4.10.4. Stability Test

Table 25. Results for the stability test.

Model 1		Model 2		Model 3
Eigenvalue	Modulus	Eigenvalue	Modulus	Eigenvalue	Modulus
0.923	0.923	0.909	0.909	0.932	0.932
0.852	0.852	0.834	0.834	0.861	0.861
0.731	0.731	0.765	0.765	0.705	0.705
0.565	0.565	0.676	0.676	0.579	0.579
0.123	0.123	0.085	0.085	0.097	0.097

Source: authors’ computation.

indicates that all eigenvalues for Models 1, 2, and 3 lie inside the unit circle (i.e., their modulus must be less than 1). Given that all eigenvalues have moduli less than 1, this indicates that the VECM for all three models is stable and the system does not explode over time.

5. Discussion

In Table 8, a 1-unit increase in the commodity dependence index is associated with a 0.45-unit increase in real GDP in the long run. Higher agricultural commodity price volatility is associated with a 0.30-unit decrease in real GDP, indicating negative impacts on economic stability. Similarly, a 1-unit increase in the commodity dependence index leads to a 0.55-unit increase in agricultural commodity price volatility. The result of this study is consistent with [22], who investigated the interconnections among and within agricultural commodities and found that price volatilities of soybean oil, cotton, and coffee represent the major sources of propagation of financial distress in commodity markets. Ref. [51] examined the trend and pattern of agricultural price volatility and its seasonality using monthly data from January 2010 to December 2022. Using a fixed effects model, the results revealed that prices of vegetables were most volatile, followed by oilseeds and pulses, with volatility characterized with peaks during the pre-harvest and harvest periods and troughs during the post-harvest period. This implies that the volatility of agricultural commodity prices would have a different impact on commodity dependence depending on the type of crop and period of study. The findings may differ because this study included 31 countries, whereas [52] examined the 9 Sub-Saharan African countries that rely primarily on a single agricultural commodity. Ref. [53], on the other hand, discovered a positive correlation between agricultural commodity price volatility and growth.

Table 9 indicates that a 1-unit increase in the commodity dependence index results in a 0.55-unit increase in real GDP, reflecting positive long-run effects on economic output. Also, higher global oil price volatility leads to a 0.40-unit decrease in real GDP, indicating negative long-run impacts on economic stability. Further, a 1-unit increase in the commodity dependence index leads to a 0.60-unit increase in global oil price volatility, reflecting the amplifying effect of higher prices. This result is consistent with [21], who found out that oil price shocks, particularly oil aggregate demand shocks, prominently affect the total risk connectedness among the commodity markets. Ref. [22] indicated that energy commodity price volatility has a neutral effect on the propagation of financial distress in commodity markets.

Finally, Table 10 indicates that a 1-unit increase in the commodity dependence index leads to a 0.60-unit increase in real GDP, suggesting that higher mineral and metal ore prices benefit the economy in the long run. Secondly, a 1-unit increase in mineral and metal ore price volatility results in a 0.35-unit decrease in real GDP, indicating negative impacts on economic stability. This result is consistent with [22,23,54]. For instance, Ref. [23] found out that extreme energy transition metals (ETMs) price changes have a non-neutral effect on commodity prices, although this effect vanishes for non-extreme price movements. Ref. [54] looked at how commodity price shocks affected different sectors in Australia. The study used structural vector autoregressive estimation approaches. The study’s findings indicate that commodity price shocks have a significant impact on Australia’s mining, construction, and manufacturing industries. Ref. [22] indicated that mineral and metal ore commodity price volatility has a neutral effect on the propagation of financial distress in commodity markets. Overall, the VECM result (Table 6, Table 7 and Table 8) demonstrates that price volatility in agricultural commodities, global oil prices, and mineral and metal ore have significant and negative effects on commodity dependence.

6. Policy Implication of the Study

This study has several important implications for the economies of the Sub-Saharan countries. For instance, high volatility in commodity prices can lead to economic instability for countries heavily dependent on commodities due to fluctuations in their GDP growth rates, making long-term economic planning challenging [55]. Moreover, price volatility can cause unpredictable swings in government revenue, impacting public spending and investment in infrastructure, education, and healthcare. Similarly, commodity price volatility can lead to fluctuations in the exchange rates of commodity-dependent countries, thus affecting the balance of payments and foreign exchange reserves and thereby influencing the country’s ability to import goods and services. Furthermore, countries might face challenges in managing external debt, especially if they borrow in foreign currencies. Volatile commodity prices can make debt servicing more unpredictable and difficult. Similarly, volatile commodity prices can lead to an unpredictable trade balance. Countries might experience trade deficits during periods of low prices and surpluses during high prices, complicating trade policy and negotiations.

Commodity dependence can lead to unequal wealth distribution, where a small elite benefits from commodity exports while the majority remain impoverished. Price volatility can exacerbate this issue, as periods of low prices might lead to widespread economic hardship. Furthermore, fluctuating commodity prices can directly impact poverty levels. During price drops, incomes of those dependent on the commodity sector (like farmers or miners) decrease, leading to higher poverty rates. Therefore, this study recommends three policy implications. First, this study underscores the need for economic diversification. Relying on a narrow range of commodities makes economies vulnerable to price shocks [26].

Developing other sectors like manufacturing and services can mitigate these risks. Secondly, there is a need to implement policies that support sustainable agricultural practices to increase productivity and resilience to price shocks. This includes investments in research, development, and extension services. Further, the Sub-Saharan African countries need to create safety nets such as agricultural insurance schemes and price stabilization funds to protect farmers from extreme price volatility and ensure food security. These funds can be used to support the economy during downturns [56,57,58]. Thirdly, engaging in hedging strategies in international markets can help stabilize revenue from commodities by locking in prices for future sales [59]. Overall, understanding the connectedness of commodity price volatility and commodity dependence is crucial for the Sub-Saharan African countries to develop robust economic policies that enhance stability, promote sustainable growth, and improve the welfare of their populations.

7. Conclusions

This study explored the sustainability implications of commodity price shocks and commodity dependence in selected Sub-Saharan countries between 2000 and 2023, providing valuable insights into the economic and environmental impacts faced by these nations. In this regard, price volatilities of agricultural commodities, global oil prices, and mineral and metal ore prices were tested with their respective commodity price indices. The US GDP and US short-term interest rates were included in the model as control variables. This study had three null hypotheses; that is, commodity dependence has no significant impact on economic growth in the Sub-Saharan countries; negative commodity price shocks have a significant adverse effect on economic growth in the Sub-Saharan countries; and commodity dependence has no significant impact on commodity price in the Sub-Saharan countries. The vector error correction model (VECM) was employed for data analysis, and the results of this study were threefold for all three categories of commodities (agricultural, energy, and mineral and metal ore). First, the commodity dependence index is positively related to the real GDP, suggesting that higher commodity prices benefit the economy in the long run. Second, commodity price volatility is negatively related to economic growth, indicating adverse impacts on economic stability in the long run. Third, commodity dependence is positively related to commodity price volatility.

Countries heavily reliant on commodity exports are particularly vulnerable to global price fluctuations, which can lead to severe economic disruptions, reduced fiscal capacity, and increased vulnerability to external shocks. The analysis underscores the critical need for diversification strategies to mitigate the adverse effects of commodity dependence for the Sub-Saharan countries so they can enhance their resilience to price shocks and promote more sustainable economic growth. Policymakers should focus on implementing robust economic policies that support diversification, invest in human capital, and improve infrastructure to create a more resilient economic structure. In conclusion, addressing the sustainability challenges posed by commodity price shocks and dependence requires a multifaceted approach that combines economic diversification, environmental stewardship, and effective governance. By adopting these strategies, the Sub-Saharan countries can better navigate the complexities of the global commodity market and build a more sustainable and resilient future. Due to the paucity of data, this study was limited to only 31 countries, and future studies can consider studying all the Sub-Saharan countries. This study suggests two areas for further research. First, investigate how fluctuations in commodity prices impact progress toward specific SDGs, such as poverty reduction, food security, and climate action. Second, explore how changes in commodity prices influence environmental outcomes, including resource depletion, pollution, and ecosystem health.