Best-Seller pricing on amazon.com: A panel vector autoregressive approach

Amazon has created an ideal stop for one-stop shopping with its broad assortment of products sold by Amazon itself and other retailers. Its huge selection of products, big data-driven recommendation system, nice user interface, and many other factors entice consumers to shop there, and spend hours to discover items. A customer who visits Amazon.com is likely to buy unplanned items website recommends or that fulfill the condition for free delivery. High cross-selling potential of Amazon, and consumers’ high impulse buying potential facilitate using loss leader strategy. It is known that Amazon.com sells best-seller books at below cost, but there is limited understanding of the factors that influence pricing decisions of this company. In this study, we observe how key market characteristics impact discounting decisions of Amazon and how all these variables affect each other in this marketplace. We conduct Panel Vector Autoregressive modelling on a panel time series dataset with 15500 observations on 5 endogenous variables (discount, sales rank, list price, customer review and number of sellers) and 1 exogenous variable (physical format) of 500 books for 31 days. By using Panel Vector Autoregressive modelling, we also take the impact of previous days’ observations into consideration in explaining the relationship. Our results suggest that on Amazon.com discounts are deeper for books with better sales ranks, higher list prices, higher customer reviews, or lower number of sellers. We also demonstrate the effects of these variables to each other. Our study is among the few that observe dynamics of Amazon marketplac