Antitrust risk simulation in digital markets

Abstract

This study proposes a hybrid approach for assessing antitrust risk in digital markets, combining supervised machine learning (Random Forest) with Monte Carlo simulations. Using real data from the Administrative Council for Economic Defense (CADE), enriched with proxies for digital characteristics, it estimates the probability of high antitrust risk based on variables such as market share, network effects, entry barriers, and the use of pricing algorithms, employing Python for simulation. The model suggests that digital platforms with a market share above 40%, combined with increasing network effects and active use of pricing algorithms, have a probability greater than 85% of generating antitrust risks. The results demonstrate the predictive capability of the proposed system and its usefulness in providing insights for regulators and managers.