Deterministic Models for Revenue Forecasting in Nigeria’s Federal Inland Revenue Service

Accurate revenue forecasting is fundamental to fiscal planning and budgetary stability in Nigeria, where tax revenue shortfalls against Federal Inland Revenue Service targets have persistently constrained government expenditure programmes. This study develops and evaluates deterministic revenue forecasting models using historical tax collection data from the FIRS spanning 2014 to 2023 across four major revenue heads: companies income tax, value added tax, petroleum profit tax, and withholding tax. Trend decomposition, regression-based forecasting, and autoregressive distributed lag models are specified and compared for predictive accuracy using expanding window out-of-sample evaluation over the 2021 to 2023 period. The ARDL model incorporating GDP growth, inflation rate, and oil price as macroeconomic conditioning variables achieves the lowest mean absolute percentage error of 9.7 percent in value added tax forecasting. The deterministic trend model performs best for petroleum profit tax, reflecting the strong explanatory power of oil price trajectories. Combined ensemble forecasts outperform individual model forecasts for three of four revenue categories. A five-year revenue projection under three macroeconomic scenarios is developed, providing a planning range for medium-term expenditure framework preparation. Keywords: revenue forecasting, ARDL model, FIRS, tax collection, fiscal planning.