Speaker
Description
This contribution presents a novel, fully automated system to forecast food price changes in fragile and conflict-affected economies, where official statistics are often delayed, disrupted, or unavailable. Using the case of Gaza during the 2023–2025 conflict, the system integrates cross-border trade data, exchange rates, and local retail food prices to quantify how international shocks translate into domestic consumer price volatility.
The methodology pools together six machine learning models and three time-series approaches to forecast monthly price changes of nine essential food commodities across twelve forecast horizons. Models are tuned via grid search and retrained in an expanding-window scheme, enabling adaptive, real-time forecasting as crises unfold.
Designed to be fully accessible, the system relies solely on publicly available data, offering a robust and interpretable tool where institutional infrastructure is weak or impaired.
Initial results reveal a shift from weakly correlated, seasonally stable price series to a new regime of synchronized inflation. Notably, imported items such as sugar, milk, and semolina show decoupling from regional market trends—highlighting the fragility of cross-border price transmission under conflict conditions.
In the short run, this methodology is aimed at supporting humanitarian agencies in anticipating food aid needs in Gaza and enabling local households to plan purchases strategically. In the longer run, it will offer a timely, accessible, and transparent short- and medium-term food price forecasts. As such, it could assist in real-time food security monitoring and resilience-building in regions facing climate disruption, economic fragility, or conflict—contributing to broader efforts toward environmental justice and equitable access to food.
