Barryugo/FIEWS

Food Inflation Early Warning System — Nigeria # FIEWS — Food Inflation Early Warning System > **Can we detect a food inflation surge 3–6 months before it appears in Nigeria's CPI data?** FIEWS is a forecasting and risk-monitoring system for Nigerian food inflation. It integrates six macroeconomic and climate data sources into a monthly panel, applies four models to forecast inflation 3 and 6 months ahead, and produces a risk score, driver attribution, and scenario analysis designed for policymakers and researchers. --- ## Key Findings - Nigeria's food inflation has a **detectable 3–6 month early warning window** driven by exchange rate depreciation, fuel price shocks, and global commodity prices - Once an inflation episode begins, the **ACF half-life is ~14 months** — making early detection far more valuable than crisis response - The **2023–2024 crisis (peak: 40.9%)** was preceded by a clear fuel shock signal 3 months before the acute phase and FX pressure signals from Q4 2022 - **Current reading (April 2026): 16.1%** — Elevated risk zone, with 6-month forecast of 20.8% --- ## Charts Fig 1 — Food inflation timeline with regime annotations Fig 2 — Cross-correlation: leading indicators vs food CPI Fig 3 — Feature correlation heatmap Fig 4 — ACF: inflation persistence (half-life ≈ 14 months) Fig 5 — Walk-forward forecast vs actual (Gradient Boost) Fig 6 — Feature importance (permutation, 3-month horizon) --- ## Model Performance | Model | 3m MAE | 3m RMSE | 6m MAE | 6m RMSE | |---|---|---|---|---| | **ARIMA (Benchmark)** | **1.62pp** | 2.44pp | **1.62pp** | 2.47pp | | Ridge Regression | 4.95pp | 7.36pp | 8.60pp | 12.65pp | | Random Forest | 5.97pp | 7.72pp | 6.48pp | 8.05pp | | **Gradient Boost** | 5.43pp | 7.38pp | **5.78pp** | 7.66pp | > ARIMA dominates on MAE due to inflation's high autocorrelation (r=0.98 at lag 1). Gradient Boost provides the more useful policy output: probability scores, driver attribution, and scenario analysis. --- ## Leading Indicators | Indicator | Lead Time …