지방세 세수예측방법의 비교･평가:서울시를 중심으로

초록

This study evaluates the accuracy of various forecasting methods using the total tax revenue and the automobile tax revenue of Seoul metropolitan city. The simple moving average, the exponential smoothing, the ARIMA, and the regression are used to forecast the tax revenues. The forecast models are constructed using 1975-2010 time series and the accuracy of the models is evaluated using 2011-2015 time series. MAPE and RMSE are used as the accuracy criteria. The evaluation results are as follows. First, while the exponential smoothing and the regression PW are best in total tax revenue forecasting, MA3 and MA5 make the most accurate prediction for the automobile tax revenue. Second, while time series models are better in the early period and the regression model is better in the later period in total tax revenue forecasting, MA3 and MA5 make the most accurate prediction for the automobile tax revenue during the entire period. Third, the change pattern of the forecast error differs by the forecast method and/or the tax revenue. Forth, the under-and overestimation tendency differs by the forecast method. Fifth, correlation analysis shows that while there is a statistically significant positive relationship between the RMSE of forecast models and that of forecast results in the total tax revenue, the relationship is insignificant in the automobile tax revenue.

키워드