Academic journal article Atlantic Economic Journal

Some Experimental Evidence regarding Collusion

Academic journal article Atlantic Economic Journal

Some Experimental Evidence regarding Collusion

Article excerpt

Some Experimental Evidence Regarding Collusion

Experimental investigations of collusion have focused on two factors - the number of rivals in the market and the opportunity to collude. A number of studies beginning with Fouraker and Siegel [1963] and including Dolbear, Lave, et al. [1968] and Friedman and Hoggatt [1980] have discerned a tendency toward tacit collusion the smaller the number of rivals. Holt [1985] has observed tacit collusion in duopoly experiments. Isaac and Plott [1981], Issac, Ramey, and Williams [1984], and Isaac and Walker [1985] have investigated the relationship between the opportunity to collude and market outcome. This study presents some experimentally generated examples of both overt and tacit collusion.

I. The Experimental Environment

The experimental environment for the experiments is based on a model developed by Morrison [1979]. This model accommodates any number of rivals and has three solutions: (1) zero profit (long run competitive); (2) Cournot-Nash (noncooperative); and (3) collusion (cooperative). In terms of describing the properties of the model, if one starts from the competitive solution, any rival firm that raises price (other prices fixed) will increase its profit and also increase the demand of all of the other rivals. If the rivals' prices were fixed slightly above the competitive level this increase in demand would translate into increased profits for the rivals. Thus, starting at or near the competitive solution, there is a commonality of interest in raising prices. At the Cournot-Nash solution, by definition, any rival that changes its price in either direction will reduce its profit. However, at the Cournot-Nash solution there is still a commonality of interest in raising prices, but for price raising to be mutually beneficial, prices must be raised simultaneously or in collusion. Moving up from the Cournot-Nash solution, the collusion solution corresponds roughly to joint profit maximization for all of the rivals.

The Cournot-Nash solution is more profitable for all of the rivals than the competitive solution (infinitely so in ratio form), and the collusion solution is more profitable all around than the Cournot-Nash solution. But the collusion solution does not present the maximum possible profit for the individual rivals. For the rivals collectively the collusion solution has a prisoners' dilemma or "slippery slope" problem. At the collusion solution, if the other rivals hold their prices fixed, one rival can increase its profit by cutting price. If this rival were to cut to the point that maximized its profit, the rivals holding their prices fixed would receive less profit than in the Cournot-Nash solution. Their best, or least bad, noncooperative response is also that of cutting price.

Thus, if one rival breaks rank in a collusive situation, it might be expected that all of the rivals would slip down the profit hill if the collusion were tacit. In the case of overt collusion, if there were no enforcement mechanism and some rivals were obtuse, they would also slip down the profit hill.

In the context of this model, the experiments involved subjects serving as CEO's of the rival firms. For a succession of market periods each CEO turned in a price for his firm and then learned the price and quantity information for all of the firms before proceeding to the next market period. Subjects were rewarded on the basis of the profitability of their firms. For a complete description of the experimental procedure, see Kamarei [1985] or Morrison and Kamerei [forthcoming]. For the derivation of the profit equations, see the appendix to this paper.

II. Overt Collusion

The authors observed overt collusion in a series of experiments involving MBA students at Indiana University. In the set of experiments where overt collusion was observed, it had originally been intended to start the sequence of experiments with a six-firm experiment and work down to a two-firm experiment. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.