Digitalising Tax Compliance and Elevating Revenue Forecasting in Rwanda: Evidence from Statistical Modelling and Machine Learning

This paper explores the dual role of digitalisation and advanced forecasting models in enhancing tax compliance and revenue prediction in Rwanda. It investigates the impact of electronic billing machine (EBM) adoption on tax performance, the comparative accuracy of traditional econometric models (Bayesian VAR) versus machine learning models (NNAR and Ensemble), and the policy implications of these findings. Using data from 2010 to 2023, the analysis reveals that while EBM usage has expanded significantly, its impact on tax revenue is limited due to enforcement and implementation challenges. Regression results indicate that trade openness, financial development and effective governance positively influence tax revenue, whereas corruption and remittance inf lows pose challenges. Forecasting models indicate a moderately optimistic outlook for tax-to-GDP and trade integration, with NNAR outperforming other models in predictive accuracy. The study concludes with key policy recommendations focused on strengthening digital compliance infrastructure, addressing corruption, supporting financial sector development and leveraging machine learning for more accurate fiscal forecasting. These insights are vital for designing evidence-based tax reforms and achieving sustainable domestic resource mobilisation in Rwanda.