Can Machine Learning Aide in Cartel Detection?

Abstract

Recent research has focused on complex antitrust issues of stemming from corporate uses of “Big Data” and “Machine Learning” pricing algorithms. For instance, Whether could the pricing algorithms of two different companies ever could be said to be colluding with each other is just one of the many important questions being raised in this context.? In this short note, we want to explore the other side of the coin, and ask whether Big Data and Machine Learning could be used in the detection (and therefore in the defense) of cartels or other collusive, anti-competitive practices, and what, if anything, would be the role of the economist in such applications.

However, one wants to label the application of sophisticated pattern-matching algorithms to large data sets – “data mining,” Big Data, “artificial intelligence,” “machine learning” – it is often considered to be a field of expertise separate and distinct from traditional economics or even econometrics. We will not spend time developing a taxonomy over these different concepts (that itself being an interesting and nuanced exercise) but will simply refer to these collective practices as “machine learning,” a field (or maybe set of fields) often considered the domain of data scientists or computer scientists much more so than economists. Does this field have a home in cartel detection, and can (or should) the economist be excluded?

Keywords: Artificial Intelligence, Screens, Collusion, Detection

JEL Classification: C51, C52, D40, K21, L19