Study On Data Quality Evaluation Of Chinese Consumer Price Index

As an important index closely related to people’s livelihood,the change of the Consumer price Index is an important basis for formulating macroeconomic policies,and is of great significance to national economic management.The quality of CPI data is the focus of attention of the public and researchers.However,due to the imperfect statistical system,low transparency of data,differences in statistical methods between departments and other factors,the public often has a skeptical attitude after statistical institutions release data indicators.Therefore,this paper will actively explore the quality assessment method applicable to CPI data based on outlier diagnosis,which mainly includes the following two parts:The first part is to evaluate the data quality of the CPI total index through the econometric model analysis.The robust principal component regression model was used to identify the outliers in the CPI between 2001 and 2020 and the regional CPI in 31 provinces and cities in 2020.It was found that the CPI in 2002 and 2009 and the CPI in Shanxi and Yunnan in 2020 were outliers.According to the analysis,the CPI data of 2002 and 2009 are reliable,while the CPI data of Shanxi and Yunnan in 2020 have some problems.The second part is to evaluate the data quality of CPI subindex by statistical distribution test.Grubbs test and Dixon test were used to identify outliers in the CPI sub-index from 2001 to 2015 and the regional CPI sub-index from 2016 to 2020 in 31 provinces and cities.It was found that the residential index in 2009 was an outlier,while there were 8 outliers in the regional data.Through the analysis,the residential index of2009 is reliable,but there are some problems with the data in some areas.According to the empirical results,the overall CPI data in China is reliable,and only a few areas of data are diagnosed as outliers,which have reliability problems.On the other hand,it is also verified that the outlier diagnosis method based on robust principal component regression analysis and statistical distribution test is suitable for the quality assessment of CPI data.