Remarks on Price Index Methods for the CPI Measurement Using Scanner Data

Anotace: Scanner data are a quite new data source for statistical agencies and the availability of electronic sales data for the calculation of the Consumer Price Index (CPI) has increased over the past 16 years. Scanner data can be obtained from a wide variety of retailers (supermarkets, home electronics, Internet shops, etc.) and provide information at the level of the barcode, i.e. the Global Trade Item Number (GTIN, formerly known as the EAN code). One of new challenges connected with scanner data is the choice of the index formula which should be able to reduce the chain drift bias and the substitution bias. In this paper, we compare several price index methods for CPI calculations based on scanner data. In particular, we consider bilateral index methods with chained versions of direct weighted and unweighted indices, and also selected multilateral index methods, i.e. the quality adjusted unit value method (QU method) and its special case (the Geary-Khamis method), the augmented Lehr method, the so called “real time index”, the GEKS method and the CCDI method. We consider different weighting schemes in quantity weights on the price index. We compare all these methods using a real scanner data set obtained from one supermarket chain.. The main aim of the paper is to show how big differences among bilateral and multilateral indices may rise while using real scanner data sets. In particular our results lead to the conclusion that the choice of the multilateral formula and the weighting scheme does matter in inflation measurement. It is shown that differences between values of all discussed formulas may exceed several percentage points even in the case of only one homogeneous group of products.