Despite being a critical agricultural export for Indonesia, cocoa (Theobroma cacao) experiences severe price volatility, creating significant financial exposure for supply chain participants. Traditional statistical models relying on Gaussian or light-tailed distributions often fail to capture the true likelihood and magnitude of extreme price fluctuations, resulting in flawed risk mitigation. To address this, this study presents a robust tail risk assessment framework for Indonesian cocoa prices leveraging Extreme Value Theory (EVT). Specifically, the approach utilizes the Generalized Pareto Distribution (GPD) applied through the Peaks-Over-Threshold (POT) methodology. Threshold optimization is achieved via Mean Excess Plots and the Hill estimator, with GPD parameters derived using Maximum Likelihood Estimation (MLE).The proposed model measures tail risk by calculating Value-at-Risk (VaR) and Conditional Value-at-Risk (CVaR) across various confidence intervals (90%, 95%, 99%, 99.5%, and 99.9%) and generates return-level plots for long-term risk projection. When benchmarked against Normal and Student-t distributions, goodness-of-fit metrics—including Kolmogorov-Smirnov, Anderson-Darling tests, and Q-Q plots—demonstrate that the GPD-POT model more accurately captures the empirical tail behavior of cocoa log-returns. The calculated shape parameter (ξ = 0.2758) reveals a heavy-tailed distribution within the Fréchet domain, highlighting that extreme price shocks occur more frequently than thin-tailed models suggest. The resulting 99% VaR (6.301%) and CVaR (9.142%) serve as practical metrics for developing hedging strategies, setting trading risk limits, and pricing insurance premiums. Ultimately, this research delivers a statistically robust EVT framework for managing agricultural commodity risks in emerging economies.
Copyrights © 2026