| 408 | 7 | 124 |
| 下载次数 | 被引频次 | 阅读次数 |
大数据来源下CPI指数的创新编制,对及时了解新经济时代的物价走向和识别通胀危机、预测宏观经济拐点以实现我国通胀治理现代化、推动经济平稳和高质量发展具有重大意义。GEKS多边指数是近些年国际学术界重点研发的大数据热点价格指数,但其构造方法颇具争议。借助超市扫描大数据,就GEKS指数序列更新方法、窗口长度选择等学界难题开展理论与实证研究,获得了以下富有启发性的结论:①GEKS指数序列更新方法 2、3应用效果相对较差;②随着窗口长度的增加,GEKS环比价格指数会趋于单位值,不同更新方法下的GEKS链式指数也会呈现一定的趋同性;而GEKS指数的通胀趋势判断力却不受此影响,但更新方法的选择却会导致其不同的通胀趋势预测结果;③更新方法 4会随着窗口长度的增加而呈现更强的替代偏误,方法 1却没有出现明显的替代偏误。综合而言,更新方法1和13个月窗口长度应该是编制GEKS指数序列更为合理的组合方式。
Abstract：The innovative compilation of CPI index based on big data is of great significance for timely understanding the price trend in the new economic era,identifying the inflation crisis,predicting the economic turning point,realizing the modernization of China's inflation governance and promoting the stable high-quality development of the economy. The GEKS Multilateral Index is a big data-related hotspot price index researched and developed in the international academic community in recent years,but its construction method is quite controversial. Based on the big data of supermarket scanning,in connection with the unresolved problems such as its update method and window length selection,we draw some enlightening conclusions from the theoretical and empirical research: 1) The update methods 2 and 3 have relatively poor application effects. 2) With the increase of the window length,the GEKS sequential price index tends to the unit value,and the GEKS chain indexes under different update methods also show certain convergence. However,the GEKS index's inflation trend judgment is not affected by the window length,but the choice of update method can lead to different inflation trend prediction results. 3) Update method 4 will present stronger substitution bias with the increase of window length,while there is no obvious substitution bias in method 1. In general,it may be a more reasonable combination for update method 1 and 13 months' window length to compile GEKS index sequence.
[1]陈立双．CPI中基本价格指数的替代效应与选择问题[J]．漳州师范学院学报:哲学社会科学版，2013,27(4):41-48．
[2]李绍泰，刘建平．基于扫描数据的CPI基本分类指数编制方法研究[J]．统计研究，2015(3):3-11．
[3]李绍泰．应用扫描数据改进CPI编制的方法研究[D]．广州:暨南大学经济学院，2014．
[4]徐继华等．智慧政府---大数据治国时代的来临[M]．北京:中信出版社，2014:1-10．
[5]杨灿，孙秋碧．国民核算中的指数序列问题研究[J]．统计研究，2006(6):73-79．
[6]杨灿．现代指数形式理论评析[J]．厦门大学学报:哲学社会科学版，2002(3):32-40．
[7]Australian Bureau of Statistics.Making Greater Use of Transactions Data to Compile the Consumer Price Index[R].Information Paper6401.0.60.003,Canberra,ABS,November 29,2016．
[8]Balk M.On Curing the CPI’s Substitution and New Goods Bias[M].CBS,1999．
[9]Cavallo A,Rigobon R.The Billion Prices Project:Using Online Prices for Measurement and Research[J].Journal of Economic Perspectives,2016,30(2):151-78．
[10]Chessa A,Verburg J,Willenborg L.A Comparison of Price Index Methods for Scanner Data[C].the 15th Meeting of the Ottawa Group on Price Indices,2017:10-12．
[11]De Haan.A Closer Look at the Rolling Window GEKS Index with a Movement Splice[J/OL].https://www.business.unsw.edu.au/research-site/centreforappliedeconomicresearch-site/newsandevents-site/workshops-site/Documents/paper-de-haan-rwgeks-extension.pdf,2017-10-16．
[12]De Haan.A Framework for Large Scale Use of Scanner Data in the Dutch CPI[C].Paper Presented at the 14th Meeting of the Ottawa Group,Tokyo,2015:1-43．
[13]De Haan,Heymerik A,Der Grient.Eliminating Chain Drift in Price Indexes Based on Scanner Data[J].Journal of Econometrics,2011(1):36-46．
[14]De Haan,Krsinich F.Scanner Data and the Treatment of Quality Change in Nonrevisable Price Indexes[J].Journal of Business&Economic Statistics,2014(3):341-358．
[15]Diewert E.Chapter 7:Scanner Data and the Rolling Year GEKS Method for Constructing Indexes[C].the 32th General Conference of the International Association for Research in Income and Wealth held in Cambridge,MA,August 5-11,2012:1-12．
[16]Diewert E,Fox J.Substitution Bias in Multilateral Methods for CPI Construction Using Scanner Data[R].Vancouver School of Economics,2017．
[17]EC et al.System of National Accounts[M].United Nations,2008．
[18]Eltet9 O,K9ves P.On a Problem of Index Number Computation Relating to International Comparison[J].Statisztikai Szemle,1964,42(10):507-518．
[19]Gini C.On the Circular Test of Index Numbers[J].Metron,1931,9(9):3-24．
[20]ILO et al.Consumer Price Index Manual:Theory and Practice[M].Geneva:International Labour Office,2004．
[21]Ivancic L,Diewert E,Fox J.Scanner Data,Time Aggregation and the Construction of Price Indexes[J].Journal of Econometrics,2011(1):24-35．
[22]Ivancic L,Diewert E,Fox J.Using a Constant Elasticity of Substitution Index to Estimate a Cost of Living Index-From Theory to Practice[R].Discussion Paper,2010．
[23]Krsinich F.The FEWS Index:Fixed Effects with a Window Splice[J].Journal of Official Statistics,2016(2):375-404．
[24]Lloyd P.Substitution Effects and Biases in Nontrue Price Indices[J].The American Economic Review,1975,65(3):301-313．
[25]Melser D.Scanner Data Price Indexes:Addressing Some Unresolved Issues[J].Journal of Business&Economic Statistics,2018,36(3):516-522．
[26]Moulton B.Constant Elasticity Cost-of-living Index in Share Relative Form[M].Bureau of Labor Statistics,Washington DC,December,1996．
[27]Nygaard R.Chain Drift in a Monthly Chained Superlative Price Index[C].Article Presented at the Joint UNECE/ILOs Workshop on Scanner Data,Geneva,Switzerland,2010:1-17．
[28]Szulc B.Indices for Multiregional Comparisons[J].Przeglad Statystyczny,1964(3):239-254．
[29]Watanabe K,Watanabe T.Estimating Daily Inflation Using Scanner Data:A Progress Report[R].CARF Working Paper Series,CARF-F-342,February,2014.
(1)也即环比指数序列通过链锁所得到的指数，并据此实现商品价格水平的跨期比较。因此，链式指数通常被称为“间接比较指数”;与此相对应，定基指数则直接进行跨期比较，而被称为“直接比较指数”。
(2)主要指拉氏、帕氏以及对称性的Fisher、T9rnqvist和Walsh等加权指数。
(3)也即当所有商品价格都返回到基期水平时，总指数值不等于单位值的现象。
(1)需要说明的是，式(1)中l的取值范围是相对固定的。但随着时间推移，这一范围需要动态更新，不妨称其为动态基期。
(2)也即，时间互换测验、因素互换测验和循环性测验。
(3)该指数实际上包含P1TGEKS定基价格指数和PT(T+1)GEKS环比价格指数两部分。随着时间的推移，实践中需要针对后者构造环比价格指数序列，以进行链锁。
(4)如前所述，T为窗口长度(下同)，不过此处假设T为偶数，奇数情况的更新指数依此类推。
(1)方法2的应用效果不佳，此处暂不对其进行比较;方法3则需要相对较长的窗口长度，才能够对其开展更为有效的实证研究。而过长的窗口长度，容易导致商品项目匹配性和可比性等突出问题，从而在很大程度上影响其GEKS指数的稳定性和数据质量，该方法的应用效果可能并不尽人意，此处也未能对该方法展开相应的研究，图14亦同。
(2)由于窗口长度的不同，以及各组环比价格指数当期和比较基期的差异，P13,14GEKS,P14,15GEKS,P15,16GEKS，…，P35,36GEKS各组指数所包含的环比价格指数个数分别为1,2,3,4,5，…，23。
(1)之所以以该指数为基准测算GEKS指数替代偏误，是因为该指数可以相当可靠地判定它并无替代偏误(ILO等，2004)。
(2)详细推导过程，请参考相关文献，此处不再赘述。
(1)回归估计过程在此略去，具体的可参考陈立双(2013)有关文献。
(2)具体的推算过程相对复杂，论文中未能列出;详细的计算结果，亦可随时提供。
(1)与图2的分析类同，此处的分析中，我们同样排除对方法3的研究与比较。而对于该方法相关问题的深入探讨，将在后续进一步扩展和完善。
基本信息:
DOI：10.19343/j.cnki.11-1302/c.2020.04.002
中图分类号:F726;F224
引用信息:
[1]陈立双,祝丹.CPI之GEKS指数序列更新方法及窗口长度选择问题辨析[J].统计研究,2020,37(04):18-31.DOI:10.19343/j.cnki.11-1302/c.2020.04.002.
基金信息:
国家社会科学基金一般项目“大数据背景下线上CPI编制理论、方法与应用研究”(16BTJ028)
2020-04-25
2020-04-25