مجلة التنويع الاقتصادي
Volume 7, Numéro 1, Pages 5-18
2026-06-30
Forecasting Global Food Price Pressure: The Predictive Limits Of Upstream Indicators Against An Autoregressive Benchmark
Authors : Zhang Juan . Yihuan Lin .
Abstract
This paper addresses a central tension in global food price forecasting: upstream cost and risk indicators are structurally important for explaining food-price pressure, yet their real-time forecasting value is far less secure. The evidence favours a parsimonious benchmark. Across the out-of-sample evaluation, the univariate AR(1) model delivers the lowest overall forecast risk, while Ridge and XGBoost remain statistically indistinguishable from it under modified Diebold-Mariano tests. More saturated specifications, particularly LASSO and ARDL-Full, lose forecasting power, indicating that additional lagged predictors can introduce estimation noise faster than usable signal in monthly food-price forecasting. Regime-specific results qualify but do not overturn this conclusion. XGBoost records a localized advantage during the Russia-Ukraine commodity shock, consistent with the ability of tree-based models to adapt to nonlinear cross-market stress, but this gain does not persist across the full sample or at longer forecast horizons. The findings separate diagnostic relevance from point-forecasting capacity. Energy, fertiliser, exchange-rate, supply-chain and geopolitical indicators remain valuable for interpreting global food-price pressure, especially during crisis episodes; however, their incremental one-month-ahead forecasting contribution is uneven. For real-time monitoring systems, richer information sets and machine-learning flexibility should be disciplined by transparent autoregressive benchmarks rather than accepted on the basis of economic plausibility alone.
Keywords
Food price forecasting ; AR(1) benchmark ; Upstream indicators ; XGBoost ; Forecasting parsimony
Les articles similaires
Loughreit Wided
.
Sahali Noureddine
.
Batache Abderrahmane
.
pages 80-96.
Hammoum Hocine
.
Yahiaoui Akli
.
Achour Bachir
.
pages 477-489.
Cherai Elhocine
.
Boudella Youcef
.
Barkat Karim
.
pages 101-126.
Tabarourt Allel
.
Leboukh Fatma
.
pages 312-327.
