From the Sherman Act of 1890 through today, preventing and detecting collusion among competing firms has been a central issue for anti-trust and merger policy. For the design of such policy it is essential to understand when collusion can be expected to be likely and how it can be effectively discouraged. In the past theoretical research has focused primarily on understanding the mechanisms by which a collusive agreement can be sustained over time. In contrast, anecdotal evidence suggests that the primary impact of antitrust policy is to disrupt the process by which firms reach collusive agreements. Very little is known about how firms come to such agreements and it has proved extremely difficult to identify this from field data. This is therefore an environment in which an experimental approach can provide insights that would be otherwise unavailable.
The project consists of a series of experiments designed to study how firms reach collusive agreements. It focuses on the role of explicit communication between experimental subjects for this purpose and seeks to identify the features of communication that help foster collusion. All treatments start with an initial phase where no communication is available. A second phase then follows where communication is allowed. The treatments vary in the type of communication that is allowed for this second phase. The study's fundamental question is whether explicit communication is essential for achieving collusion. Expanding on this theme, the experimental design explores numerous questions about what features of communication matter most for reaching collusive outcomes:
* Theory suggests that messages need to communicate contingent strategies, but will simpler messages that are only about current prices or quantities suffice? * If only a small number of subjects can initially follow effective strategies for collusion, do their interactions with other subjects "infect" the rest of the population and lead to widespread collusion? * Do some individuals systematically use communication for deceptive purposes? Can this undermine the ability of 'honest' players to achieve collusive agreements? * Does allowing the possibility of communication to renegotiate limit the effectiveness of contingent messages? * How does complexity affect the relationship between communication and collusion? * Can a policy that that imperfectly targets and punishes communications about prices as in real anti-trust enforcement effectively reduce the incidence of collusion?
Broader Impact: The results of this project have important implications for anti-trust policy. Understanding how collusion arises makes it easier to determine where scarce enforcement resources should be focused and which market characteristics should be considered relevant for the assessment of coordinated effects analysis in mergers. Some of our treatments involve analyzing the content of subject's messages to determine what specific phrases are likely to yield collusive agreements. In particular, our experiments allow us to determine whether targeting enforcement on finding these types of messages can effectively combat collusion. The results of the study can also inform policy towards information exchange agreements between firms by helping to delineate the boundary between legitimate business activity and anticompetitive behavior.
