POTENTIAL REASONS FOR CPI CHAIN DRIFT BIAS WHILE USING ELECTRONIC TRANSACTION DATA

Scanner data mean electronic transaction data that specify product prices and their expenditures obtained from supermarkets’ IT systems by scanning bar codes (i.e. GTIN or SKU). Scanner data are a relatively new and cheap data source for the calculation of the Consumer Price Index (CPI) and the biggest advantage of scanner data is the full product information they provide already at the lowest level of aggregation. Thus, the digitization of the public sector becomes not only something that is needed but an actual necessity resulting from organisational and economic premises (e.g.: reduction of costs or time related to obtaining data). One of main challenges while using scanner data is the choice of the right price index. The list of potential price indices, which could be used in the scanner data case, is quite wide, i.e. bilateral and multilateral indices are used in practice. One of the most important criterion in selecting index formula for scanner data case is the potential reduction of the chain drift bias. The chain drift occurs if the index differs from unity when prices and quantities revert back to their base level. In the paper we present situations on the market leading to the serious chain drift bias. Our main hypothesis is that lagging consumers’ reaction to price changes is the cause of the chain drift effect. Moreover, the article is an attempt to answer the question whether the correlation of prices and quantities may have an influence on the scale and sign of the bias of the measurement of price dynamics. The study focuses also on the scale of over- and underestimation the target full-window multilateral indices by their corresponding splicing extensions. Finally, the paper verifies a hypothesis that the identity test is a key property in reducing chain drift bias. In order to verify the above research problems, both empirical and simulation studies were carried out. Our main result is the confirmation of earlier suspicions that delayed consumer response and price-quantity correlation are determinants of chain drift bias.