Identifying Algorithmic Pricing Technology Adoption in Retail Gasoline Markets

Although firms have made use of pricing software for decades, recent technological advancements created a shift from mechanically set prices to artificial intelligence (AI)/machine learning (ML)-powered algorithms that can incorporate large quantities of data, learn, and autonomously make decisions. This new algorithmic pricing (AP) software has raised concerns about potential impacts on firm behavior and competition. A recent theoretical literature has shown that sophisticated pricing algorithms can increase retail prices either by facilitating/learning collusive behavior or by changing the nature of the game firms play (Calvano et al. 2020; Asker, Fershtman, and Pakes 2021, Miklos-Thal and Tucker 2019; Brown and MacKay, forthcoming). However, there has been no evidence of markets transitioning from one pricing technology to another to test theoretical predictions. This paper investigates pricing technology in the German retail gasoline market. In this market, according to trade publications and other sources, AP software became widely available beginning in 2017.1 However, despite the availability of comprehensive station-level pricing data, there is no direct information on which stations adopted the new technology. In this paper, we describe how we overcome this challenge and identify changes in pricing technology. new technology. In this paper, we describe how we overcome this challenge and identify changes in pricing technology.