Improving the Forecasting of Budget Revenues by Tax Type

This article examines how budget revenue forecasting can be improved by modeling individual tax categories rather than relying primarily on aggregate revenue projections. Using official budget execution data for Uzbekistan for 2020–2025, the study analyzes corporate income tax, personal income tax, turnover tax, value added tax, excise tax, customs duty, property tax, land tax, subsoil use tax, and water use tax. Three transparent benchmark methods—linear trend extrapolation, compound annual growth, and median year-on-year growth—are assessed through rolling-origin validation. The exercise shows that forecast performance differs materially across taxes because tax bases, payment schedules, administrative changes, commodity exposure, and policy adjustments are heterogeneous. The paper therefore proposes a tax-specific forecasting architecture in which statistical models are selected by out-of-sample error, then supplemented by explicit tax-base assumptions, policy-costing adjustments, scenario ranges, and continuous forecast-error monitoring. The proposed framework is designed to improve both technical accuracy and the transparency of fiscal planning.