- Published by:
- Department of Treasury and Finance
- Date:
- 1 May 2024
Ching Hin (Jeffrey) Wong and Nathan La1 2
1. Department of Treasury and Finance
2. The authors appreciate the participants at the 17th Western Economics Association International (WEAI) Conference for their valuable comments and suggestions.
Author contact details: veb@dtf.vic.gov.au.
Disclaimer: The views expressed are those of the author and do not necessarily reflect the views of the Victorian Department of Treasury and Finance.
Suggested Citation: Ching Hin (Jeffrey) Wong and Nathan La (2024), Applying machine learning in tax revenue forecasting. Victoria’s Economic Bulletin, May 2024, vol 8, no 2. DTF.
Abstract
Accurate revenue forecasting is essential for effective government budget planning. This study investigates whether the use of machine learning methods can enhance the accuracy of payroll tax and land transfer duty revenue forecasts in Victoria. We compare the performance of nine different forecasting methods, including traditional econometrics models and machine learning algorithms, based on various forecast horizons, loss functions and sample periods. We find that while machine learning methods do not improve payroll tax revenue prediction, they do marginally outperform simpler methods in forecasting land transfer duty. This study shows that machine learning methods are more effective for tax lines that have higher volatility and are more sensitive to economic fluctuations.
1. Introduction
Reliably forecasting revenue is crucial for government budget planning. However, revenue forecasting is a complex task, complicated by the interplay of economic fluctuations and government policies.
2. Background
Background information for this research article.
3. An overview of forecasting algorithms
An overview of the two benchmarks employed in this study and machine learning methods.
4. The forecasting exercise
Information about the design of the study and forecast evaluation.
5. Main analyses
Information on the baseline results, and where the predictability of land transfer duty revenue comes from.
6. Additional analyses
Additional analyses, and going deeper into assessing their performance enhancement in specific scenarios.
7. Conclusion
Conclusion for this Victoria’s Economic Bulletin research article.
8. References
References for this Victoria’s Economic Bulletin research article.
Updated