带税基的个人所得税分位回归预测模型

Individual Income Tax Quantile Regression Prediction Model with Tax Base

为了突破税收预测常用分析思路及方法的局限,更全面的反映数据全貌和回归信息,提出带税基的个人所得税分位回归预测模型.即以分位回归为基础模型,兼顾时间和税基因素,完成预测模型的构建,为预测领域提供一种新工具.利用1994-2014年的个人所得税和城镇居民人均可支配收入数据,以不带税基和带税基的模型为参照,比较所提出模型对2014年的预测效果,并给出2015年个人所得税的预测情况.研究发现,分位回归预测模型与线性回归模型揭示的变量间关系规律吻合,仍然保持较高的预测精度,对数据的展示更为详实,尤其在中段分位处,城镇居民人均可支配收入对个人所得税的影响表现的较为突出.

To break though the limits of general analysis ideas and methods on tax prediction and reflect the full view of data and regression information more comprehensively, the paper comes up with an individual income tax quantile regression prediction model with tax base. That is a new tool which is based on the quantile regression with consideration of time and tax base factors and accomplishes the establishment of prediction model. Take 1994-2014 individual income tax and per capita disposable income of urban households as example, the paper compares the quantile regression prediction result on 2014 with other methods with tax base or not and cubic curve and predicts the 2015 individual income tax. It shows that quantile regression model demonstrates the same relationship between covariates and response, reflects the full view of data comprehensively, keeps similar prediction accuracy and per capita disposable income of urban households effects individual income tax more at middle quantile.