Sales and Price Spikes in Retail Scanner Data

Abstract

Microeconomic pricing data display a prevalent pattern: Short term prices vary frequently and in large amounts around repeatedly observed, more stable levels. In response to this observation, authors have proposed various procedures for extracting hypothetical or non-sale price series. In this paper, I treat these filters as procedures for decomposing prices into a slow-moving attractor price'' component and a series of quickly-reversed temporary price shocks. I argue that sale and reference price filters each place implausible restrictions on one of these components. As a result, sale filters fail to exclude many small, temporary price while the formulation of reference prices creates spurious price spikes and sales. I propose a new price filter and show that, while many commonly cited pricing facts are robust to filter specification, the implications for price duration depend heavily on the choice of filter.