Statistical Analysis of Price Fluctuations in the Online Market

In this paper we analyze price-setting behavior by stores and purchasing behavior by consumers by using prices indicated by virtual stores on the kakaku.com price-comparison site, and new data sets of corresponding consumer-clicking behavior recorded in one-second units. The following are the paper's most important findings. First, even if a store's price ranking (in terms of how low the store's price is) is not number one, the probability of clicks occurring is not zero. Nevertheless, if price ranking falls then the click probability falls, and an almost linear relationship then exists between the price ranking and the (logarithmic value of the) click probability. This linear relationship suggests that consumers have preferences for certain stores and that they select stores that display the lowest prices from among the group of their favorite stores. Second, the mean values of the prices displayed by individual stores follow a random walk with a drift pattern, which shows that most price fluctuations are caused by random increases and decreases in inventories held by stores. However, deviations from the random walk pattern appear when situations such as sharp price falls occur, suggesting that the strategic complementarity of the pricing of individual stores may lead to a price collapse.