利用大数据编制CPI代表性项目抽样方法研究

Research on the Sampling Method of Items in the Preparation of CPI with the Support of Big Data

大数据为CPI编制中代表性项目的选择和权数的构建提供了新的路径和方向。针对“有价格大数据”和“无价格大数据”的基本分类,分别设计了代表性项目的抽样方案和指数的编制方案,并使用奶酪和啤酒基本分类数据编制了Jevons链指数、T?rnqvist链指数和RYGEKS指数,探讨排除抽样和与规模成比例概率序列抽样两种方法对指数编制结果的影响。结果显示:与规模成比例概率序列抽样易于控制各层的样本量,抽取的项目代表性更高。对于支出权数集中度低的基本分类,应使用与规模成比例概率序列抽样;对于集中度高的基本分类,两种抽样方式产生的偏离相差不大。采用与规模成比例概率序列抽样编制的T?rnqvist链指数与RYGEKS指数偏差都比排除抽样方法小,与规模成比例概率序列抽样是更优的选择。使用与规模成比例概率序列抽样编制的T?rnqvist逐月链指数与RYGEKS基准指数相比较,偏差较小且趋势走向基本一致;相较于Jevons链指数,T?rnqvist逐月链指数是更优的选择。

The"prices big data"provides a new path and direction to solve the existing problems of the CPI compilation,which can directly select representative items from the data and the construct weights.The sampling scheme of representative items and the compilation the indexes are designed in view of whether the basic classification has"prices big data",which selects the cheese and beer groups as the research objects and compiles the elementary aggregates indexes using the Jevons index,T?rnqvist index and RYGEKS index,in order to analyze the effects of different sampling methods on index.The results show that:The order pps sampling is easier to control the sample size.The selected items are more representative.For the basic classification of low concentrated expenditure weights,the order pps sampling should be the only choice.For the basic classification of high concentration,the deviation between the two sampling methods is small.Choose the T?rnqvist chain index and RYGEKS benchmark index,the deviation of the order pps are smaller than the cut-off sampling,the order pps is a better choice.Compared with the RYGEKS benchmark index,the T?rnqvist monthly chain index compiled by using the order pps sampling method has smaller deviation and basically the same trend.Therefore,compared with the Jevons chain index,the T?rnqvist monthly chain index is a better choice.