Understanding the Challenges of Algorithmic Collusion
The conventional framework of antitrust laws typically presumes a “meeting of minds” as a prerequisite for collusion. Under Section 3(1) of India’s Competition Act, 2002, any agreement that leads to a significant negative impact on competition is prohibited. This provision, along with the broad interpretation of “arrangement, understanding, or action in concert” in Section 2(b), wasn’t designed to account for the complexities introduced by algorithmic pricing systems.
The Rise of Autonomous Algorithmic Collusion
Recent studies have demonstrated the potential for algorithms to autonomously engage in collusive behavior. For instance, a 2020 study by Calvano et al. showed that independent Q-learning agents could adopt strategies that mimic cartel-like behavior, leading to prices above competitive levels. Further, research by Assad and colleagues in 2024 highlighted how algorithmic pricing in German fuel retail increased margins significantly.
OECD’s Analysis and Proposed Solutions
The OECD’s 2017 report categorized the risks of algorithmic pricing into design, structure, and outcomes. While proposals like Harrington’s suggest focusing liability on algorithmic behavior rather than agreements, a clear operational framework remains absent. This article introduces a comprehensive approach involving a Structural Predisposition Test (SPT), a cumulative standard, and a proposed Section 3A to regulate algorithmic coordination without altering the existing agreement paradigm.
The Structural Predisposition Test
The SPT establishes a rebuttable presumption of anti-competitive coordination based on three cumulative conditions: algorithm design, market structure, and observable market effects. This test applies only when these conditions collectively indicate a likelihood of collusion, thus safeguarding lawful oligopolistic behavior from unfounded liability.
Algorithm Design
The first condition examines whether the algorithm can observe competitor pricing, optimizes long-term profits, adapts to rivals’ reactions, and lacks randomization. If these features are present, the algorithm may facilitate collusion, as evidenced by the Calvano experiment.
Market Structure
The second condition evaluates the market’s susceptibility to collusion, characterized by factors like high concentration and transparent pricing. A Herfindahl-Hirschman Index threshold of 2,000 is suggested to identify highly concentrated markets.
Observable Market Effects
The third condition requires consistent market effects indicative of collusion, such as elevated prices and reduced volatility, observed over a minimum six-month period.
Proposed Section 3A: A New Legal Pathway
Section 3A proposes a new statutory provision to address algorithmic pricing systems that adversely affect competition. It outlines conditions under which liability attaches, including the algorithm’s design, market conditions, and sustained collusive outcomes. Importantly, it allows for rebuttal by demonstrating independent algorithm operation or market conditions inconsistent with coordination.
Ensuring Constitutional Validity
The framework aligns with constitutional requirements, satisfying Article 14’s equality mandate and Article 19’s provisions on trade freedom with reasonable restrictions. The conditions are objective, promoting consumer protection without overreaching.
Ex-ante Governance and Collaboration
Recognizing the limitations of ex-post enforcement, the framework advocates for proactive measures. This includes forming a Digital Markets and Algorithmic Analysis Unit within the CCI, conducting mandatory audits in high-risk markets, and encouraging compliance-by-design obligations.
Conclusion
The current agreement-centric approach of Section 3 is inadequate for addressing autonomous algorithmic coordination. By introducing Section 3A, the framework fills a critical gap, targeting only those deployments where collusion-capable design, conducive conditions, and demonstrable effects converge. This ensures that legitimate pricing strategies remain unaffected, while providing a robust mechanism to curb algorithmic collusion.
Avichal Kumar is an LL.M. student at National Law University, Delhi.
