EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1433. Less than meets the eye: simultaneous experiments as a source of algorithmic seeming collusion
Invited abstract in session MC-14: Machine Learning for Electricity Market Applications, stream Energy Markets.
Monday, 12:30-14:00Room: 16 (building: 116)
Authors (first author is the speaker)
| 1. | Xavier Lambin
|
Abstract
This article reassesses the assertions regarding algorithmic collusion presented in Calvano et al. (AER 2020) and subsequent literature. We first reveal an important mistake in the original papers, which refutes the idea that the phenomenon of supracompetitive prices arises from collusive mechanisms where reward and punishment schemes support high prices. Then, our theoretical and simulation-based analysis suggests that both the high prices and the apparent punishment schemes result directly from simultaneous experimentation and the learning inertia inherent in most reinforcement learning techniques. The co-occurrence of high prices and apparent price wars is not causal and both stem from imperfect learning. We demonstrate that such seeming collusion can arise in memoryless environments, even with myopic agents, and much faster than previously thought. Our findings alert on the risk of qualifying as collusion a phenomena that is not, and invite new, simple approaches to address the issue of algorithmic supra-competitive pricing.
Keywords
- Machine Learning
- Economic Modeling
- Game Theory
Status: accepted
Back to the list of papers