Detecting Collusion: The Gasoline Retail Margin in Québec and the Price-Fixing Cartel

Abstract

The purpose of this paper is to examine whether statistical tests can successfully detect collusion, especially in the recent Québec price-fixing cartel. The underlying idea is that retail gasoline margins might exhibit some abnormality during collusion time and certain statistical tests may be able to detect it. Several autoregression models are selected with the highest adjusted R2 for each city. However, the Chow breakpoint test, predictive failure test, and likelihood test all do not provide expected results. All three tests do not give direct evidence that retail gasoline margins behaved unusually during collusion in Sherbrooke relative to Chicoutimi. Therefore, I cannot conclude that statistical tests cannot be used to detect collusion.