Pricing Regimes in Disaggregated Data

This paper develops a test for changes in the distribution of good-level prices over time and applies it to grocery store data. The method is based on the Kolmogorov-Smirnov statistic, which measures the distance between two empirical distributions. This test is robust to different data generating processes and does not require specific a priori knowledge about patterns in the data. I find that the typical pricing regime lasts seven months yet consists of a small number of distinct prices: for the large majority of regimes, five or fewer unique prices account for more than 90% of the regime. The test provides a natural way to investigate the prevalence of sticky pricing plans, since the identified change points serve as estimates of transitions to new plans. I find strong evidence in favor of rigid pricing plans: in addition to rigidity of the modal price of each regime, 76% of product series exhibit some degree of within-regime rigidity among non-modal prices; conversely, only 18% of series consist entirely of one-to-flex regimes in which prices flexibly deviate from the rigid mode; the remaining 5% consist of single sticky price regimes.