Email the details of this paper to a friend

We develop a variance screen for collusion that explicitly takes the geographic structure of a market into account. Using techniques from spatial statistics, we develop a method that allows us to determine whether outlets with relatively little price variation are randomly located, or whether such outlets cluster together, raising suspicions of a local cartel. We illustrate the use of our spatial screen with an application to the retail gasoline market, a market with a distinct spatial component.