This interdisciplinary honors thesis in Computational Economics investigates the different methods to compute auction equilibria and the impact of collusion on auction outcomes and the effectiveness of various machine learning algorithms in detecting collusive behavior using real-world datasets. We develop a program to analyze the Bayesian Nash equilibrium strategies of bidders in first-price and second-price auctions under scenarios with and without collusion. We further explore the performance of different machine learning algorithms, including Support Vector Machine (SVM), which demonstrates the highest F1 score in detecting collusion among the tested algorithms. The challenges associated with obtaining real-life auction data necessitate the use of synthetic data, providing a valuable resource for developing and validating anti-collusion algorithms in the future.This research contributes to a deeper understanding of auction dynamics and collusion, informing policymakers and regulators in designing robust auction mechanisms, implementing effective anti-collusion measures, and promoting fair and efficient market outcomes.
