Inflation Measurement with Scanner Data and an Ever-Changing Fixed Basket

Abstract

[eng] Statistics Sweden introduced scanner data into parts of the consumer price index several years ago, with the concern to ensure comparability over time and between countries. In this article, we discuss the issue of preserving the fixed basket approach and whether the traditional manual item replacement strategy, with quality and quantity adjustments, is still a relevant method to ensure comparability despite the change in data collection mode and extensiveness of data. Biases from improper quantity adjustments are discussed and illustrated through numeric examples based on real changes in the Swedish market of daily necessity products. Manual adjustments of quality and quantity are implemented by following a small random sample of representative items, i.e. a fixed basket, which therefore leads to imprecision or variance in the consumer price index. This may be a questionable approach given the availability of census-like scanner data, thus the bias-variance trade-off is addressed. The sample size related variance is estimated through a jackknife method and contrasted with quality/quantity adjustments.