国库现金流预测存在的问题及方法研究——基于混频数据模型的实证分析

Research on the Problems and Methods of Treasury Cash Flow forecasting--Empirical Analysis Based on mixed data model

近年来,随着地方财政收入的不断增长,国库库存余额不断攀升,地方财政部门多采用开展国库现金管理的方式来实现国库资金的收益最大化,因此国库现金流预测对于提高国库现金管理水平具有重要意义。而混频数据模型(MIDAS)在高频数据向低频数据转化时,可以有效解决样本丢失的问题,因此相较于常用的指数平滑法和ARMA模型具有预测准确度高的优势。本文以吉林省2015-2019年国库现金流数据为例,运用MIDAS模型,对2020年国库现金流进行预测,同时就如何提高预测准确度提出政策建议。

In recent years,with the continuous growth of local fiscal revenue,the Treasury inventory balance is rising.Local financial departments often use treasury cash management to maximize the revenue of treasury funds.Therefore,treasury cash flow prediction is of great significance to improve the level of treasury cash management.The mixing data model(MIDAS)can effectively solve the problem of sample loss when converting high-frequency data to low-frequency data.Therefore,compared with the commonly used exponential smoothing method and ARMA model,Midas has the advantage of high prediction accuracy.Taking the treasury cash flow data of Jilin Province from 2015 to 2019 as an example,this paper uses Midas model to forecast the treasury cash flow in 2020,and puts forward policy suggestions on how to improve the accuracy of the forecast.