Associate professor David Granlund, Umeå University, will present a method for estimating the probability of collusion from observed prices. The study is based on Swedish prices of off-patent pharmaceuticals and is joint work with professor Niklas Rudholm, HFI.

Knowledge about how common collusion is, how the risk of collusion is affected by the number of firms, and how much collusion increase prices is limited. Some information is gained from convicted cartels and experiments, but little is known about especially implicit collusion in real world settings.

The paper derives a method that uses probability theory to calculate the probability that collusion has existed, based on observed winning sequences. This involve defining the potential number of low-price bidders in each market and estimating autocorrelation in the probability of winning (i.e. placing the lowest bid) during competition. Then, we estimate the effect of the number of low-price bidders on the probability of collusion and how collusion affects prices.

Collusion is found to typically involve two or three firms, but the number of firms in a market can exceed the number of participants in the collusion. Preliminary results suggest that collusion increases prices more when it takes the form of bid-rotation - when colluding firms take turn to win - compared to when collusion is achieved through parallel bidding. Bid-rotation is also, by far, the more common form of collusion in the pharmaceutical markets studied.