초록
영어
This paper focuses on forecasting national tax revenue processes using several time-series forecasting models and their combination. Possibility of structural change in tax revenue is tested, the out-of- sample forecast performance of several forecasts is compared, and model confidence sets are constructed. We find several structural changes or nonlinear trends in tax revenue series and related macro-variables. However it is still uncertain what kind of nonlinear trends is appropriate for the purpose of forecasting. Forecasts with some nonlinear trend corrections clearly dominate forecasts with a linear trend but there is some uncertainty in the optimal way of non-linearity correction.
한국어
본 논문은 시계열 예측모형들과 이들의 예측모형결합을 통해 국세수입을 추계하였다. 국세수입의 구조변화를 검정하였으며, 다양한 예측모형의 표본 외 예측성과를 비교하였고, 열위 예측모형을 제외한 예측모형신뢰집합을 구성하였다. 구조변화 검정결과 국세수입 및 관련 거시시계열의 구조변화 혹은 비선형추세가 확인되었다. 그러나 어떠한 비선형추세가 국세수입추계에 적절한가에 관한 불확실성이 존재한다. 비선형추세를 가정한 예측모형의 예측성과는 선형추세를 가정한 예측모형에 비해 우월한 성과를 보였으나, 국세수입추계를 위한 최적의 비선형추세설정의 불확실성은 남아있는 것으로 보인다.
목차
1. Introduction
2. Current Practices and Literature
2.1. Current Practices in Tax Revenue Forecasting
2.2. Previous Literature on Tax Revenue Forecasting
3. Data
4. Structural Change and Nonlinear Trends
4.1. Change Detection Methods and Overall Results
4.2. Structural Change in National Tax Revenue
5. Multi-variate Forecasting Models
6. Forecasting Performance Evaluation and Short-term forecasts
6.1. Forecasting Models and Methods for Structural Changes
6.2. Forecast Performance Evaluation Scheme
6.3. National Tax Revenue Forecasts Evaluations and 5-year Forecasts
7. Conclusion
Appendix A. Tables and Figures
Appendix B. Specification of Forecasting Models
References
초록