Modelling and Forecasting Nigeria's Tax Revenue: A Comparative Analysis of SARIMA and Holt-Winters Models

Abstract

Study’s Excerpt/Novelty

• This study offers an approach to forecasting Nigeria's tax revenue by applying SARIMA and Holt-Winters models to a comprehensive dataset spanning over three decades.
• The research rigorously evaluates these models using the Box and Jenkins methodology and identifies SARIMA (3,2,1)4 (0,1,1)4 as the most effective model based on the minimum AIC value, outperforming the Holt-Winters model in accuracy metrics (RMSE and MAE).
• The findings address the inadequacy of current forecasting methods and provide a robust model recommendation, crucial for enhancing fiscal management and economic planning in Nigeria.

Full Abstract

Tax revenue is a government's income from taxes imposed on individuals, businesses, and other entities.  It is a crucial funding source for public expenditures, such as infrastructure, education, healthcare, security, and social services.  Governments rely on accurate tax revenue projections for proper fiscal management and economic planning.  According to OECD and IMF reports 2023, tax revenue of Nigeria contributed 44.15% to total expenditure and only 10.86% to the Nation’s GDP.  This signifies inadequate utilization of appropriate models that may accurately and effectively forecast the tax revenue of Nigeria.  This study focused on modelling and forecasting the tax revenue of Nigeria, utilizing SARIMA and Holt-Winters models.  Quarterly tax revenue data from January 1990 to December 2022 was used.  The Box and Jenkins model identification, estimation, and forecasting procedures were followed accordingly.  And results revealed that the SARIMA (3,2,1)₄ (0,1,1)₄ model was selected as the best model among the various models identified based on minimum AIC value.  Similarly, the Multiplicative Holt-Winters Model was selected over the additive model on minimum AIC value.  The best-fitted models' performance was evaluated using an in-sample forecast with an 80% training set and a 20% validation set, enabling the assessment of forecast accuracy.  The results further revealed that the SARIMA model outperformed the Holt-Winters counterpart in forecasting the tax revenue of Nigeria because it minimized the evaluation criteria with an RMSE of 0.1654 and MAE of 0.0816.  The study recommended applying the SARIMA model in forecasting the tax revenue of Nigeria.