ABSTRACT The paper uses scanner data to measure the welfare effects of differing product availability across six Japanese regions. To eliminate the chain drift problem associated with the use of scanner data, various multilateral indexes were computed: GEKS, Geary–Khamis, and Weighted Time Product Dummy Hedonic price indexes. Chain drift can also be eliminated by estimating purchaser preferences using consumer demand theory. Thus, the paper also estimated linear preferences, CES preferences, and Konüs Byushgens Fisher (KBF) preferences using inverse demand functions, which dispenses with reservation price estimation. The various methods gave very different results, so the choice of method matters. A major problem with the Feenstra (1994) CES methodology for measuring the gains (or losses) of utility from new and disappearing products was illustrated. The results offer practical guidance for National Statistical Offices and researchers using scanner data to construct consumer price indexes in environments with high product turnover and regional heterogeneity.
Diewert et al. (2025) studied this question.