Determinants of volatility in the commodity markets

Amagbo Fru, Roland (2025) Determinants of volatility in the commodity markets. PhD thesis, Birkbeck, University of London.

Abstract

The high price of bread and other staple foods, and local food shortages were catalysts for the 1789-1799 French revolution and the recent 2010-2011 Arab Spring. Understanding determinants of prices in commodity markets, particularly in wheat markets, has significant implications for global food security, economic stability, and risk management. The increasing complexity of these markets, exacerbated by events such as the COVID-19 pandemic, geopolitical tensions, and supply chain disruptions, underscores the need for a deeper understanding of the factors driving volatility. Traditional approaches to analyzing market dynamics often overlook nonlinear and tail-specific behaviors, which are crucial for capturing the full spectrum of volatility. This gap in the literature calls for innovative methodologies capable of encapsulating the multifaceted nature of volatility and its interdependencies across commodity markets. This thesis addresses these challenges by using three key methodological innovations: the Quantile regression, the Quantile Autoregressive Distributed Lag (QARDL) and the Quantile VAR-DCC-GARCH framework. The Quantile regression is first used to capture the non-linear nature of the response of volatility to the levels of geopolitical risk. Complementing this, QARDL is next employed to analyze the short- and long-term effects of geopolitical risks (GPR) on wheat futures volatility, providing a comprehensive understanding of how these risks propagate through markets over time. Finally, we introduce the Quantile VAR-DCC-GARCH framework that integrates quantile regression, vector autoregression, dynamic conditional correlation, and GARCH models to capture tail-specific, time-varying, and nonlinear relationships between different commodity markets. Using these methodologies, the study investigates the impact of key determinants such as geopolitical risks, inventory levels, transportation costs, and speculative activity on wheat futures volatility from 2013 to 2023. Additionally, the analysis explores the interconnectedness of energy and agricultural markets, highlighting the role of ethanol as a significant risk transmitter under varying market conditions. The findings from the quantile regression and QARDL models reveal that geopolitical risks, inventory levels, supply chain activity, and speculative activity exhibit nonlinear, asymmetric effects that are significantly higher in the upper tail of wheat futures volatility. Geopolitical risks emerged as the dominant driver, with nonlinear short- and long-run effects that are particularly pronounced during heightened market uncertainty. Notably, a shock in geopolitical risk at the top quintile requires approximately 12 weeks to return to equilibrium, compared to 31 weeks for a shock at the 0.75 quantile, underscoring their persistent and long-lasting impact. Inventory levels display a counterintuitive positive relationship with volatility in extreme market conditions, suggesting that higher inventory levels are associated with increased volatility, contrary to conventional empirical findings. Transportation costs and speculative activity also show tail-dependent effects, with transportation costs unsurprisingly increasing volatility, while speculative activity generally plays a stabilizing role by reducing volatility. Regarding the analysis of interdependencies between commodity markets, Analyzing means and medians presents a different picture, since means, are more sensitive to outliers. For means our results indicate no return risk spillovers from energy to agricultural commodities over the entire study period, suggesting a decoupling of these markets. Specifically, we found no evidence of return spillovers from ethanol to corn during 2014–2019, consistent with the proposition by Hertel and Beckman (2011). For medians, our findings reveal that ethanol acts as a net transmitter of risk to both the WTI and corn markets, highlighting the nonlinear nature of return risk spillovers. This transmission intensifies in the upper tail, with WTI responding to both ethanol and corn, and corn responding to ethanol. In the lower tail, corn alone responds to ethanol, underscoring the asymmetric nature of return risk transmission. An analysis of volatility at the mean level reveals persistence within individual markets and the system as a whole across the periods studied. Sensitivity to short-term shocks was more pronounced before mid-2019 compared to afterward. At the median, the magnitude and significance of the coefficients were found to align closely with those observed at the mean. However, responses in the upper and lower tails are substantially stronger than those at the median, indicating a distinctly nonlinear reaction of volatilities and the system as a whole to both short- and long-term shocks. Furthermore, the connectedness between volatilities is asymmetric, particularly when comparing the upper and lower tails during extreme events, as seen after mid-2019. Dynamic correlations between commodity pairs exhibit significant volatility over time and across quantiles. For instance, the correlation between corn and ethanol has decreased in magnitude and volatility, further confirming the decoupling observed in return spillovers. The Quantile VAR-DCC-GARCH framework demonstrates its robustness in capturing these complex dynamics, providing insights that extend beyond traditional mean-based methods. These findings offer practical implications for policymakers, traders, and market participants, equipping them with advanced tools to manage risks and formulate strategies in increasingly volatile and interconnected markets. By addressing critical gaps in the literature, this research provides actionable insights into the evolving nature of commodity markets and the factors driving their volatility.

Metadata

Statistics

DownloadsShow export options

Export as XMLCSVJSON

Activity Overview

6 month trend

0Downloads

6 month trend

73Hits

Additional statistics are available via IRStats2.

Archive Staff Only (login required)

Edit/View Item