A. Muren J. of Economic Behavior Org. 41 2000 147–157 149
than two sellers tend to be more competitive than is predicted by the Cournot model. The results reached by Fouraker and Siegel provide a standard of comparison for both the price
setting and the quantity setting cases. Fouraker and Siegel report results from two triopoly quantity competition experiments,
which differ with respect to the amount of information provided to subjects. In the incom- plete information formulation, sellers know only their own profits for each quantity choice,
as dependent on the combined quantities of the other two sellers. In the complete infor- mation formulation, sellers know also the effect of their and the sum of their competitors’
quantity choices on the combined profits of the other sellers. In our capacity precommit- ment experiment the information to sellers is cast in terms of costs and demand instead of
profit tables, and sellers have identical costs, which means that they can infer the effects on other sellers’ profits. This suggests that of Fouraker and Siegel’s two Cournot triopoly ex-
periments, the complete information formulation would provide our most relevant Cournot outcome comparison.
Fouraker and Siegel summarize the results of their quantity competition experiments in terms of the support for each of three types of equilibria: cooperative, Cournot and
competitiverivalistic see Table 9.10, p. 150.
2
Their definition of ‘support’ is that the outcome supported an equilibrium solution if it was closer to the corresponding numerical
quantity prediction than to any other. There were 11 triopoly markets in their complete information experiment. The cooperative solution was supported by 0 markets, the Cournot
solution was supported by 5 markets, and the competitive solution was supported by 6 markets. Thus, the competitive solution was just dominant over the Cournot solution.
In the Bertrand competition case in Fouraker and Siegel there are also two experimental formulations of triopoly markets, differing in informational structure. In both formulations
the market outcome was strongly competitive, with the ruling price being equal to the Bertrand price in every market.
3
3. Experimental formulation
The subjects were all recruited from the group of students of economics at the introductory level at Stockholm University. The experiments were conducted in two rounds, during April
1998 and during May 1999. A total of 30 students participated. Each of the sessions had subjects divided into several markets. Subjects were informed
about the number of firms in their market three, but they were not told who else in the room was in the same market. Demand and cost information was common knowledge. The
demand side was simulated in a demand curve, since it was not part of the purpose of the experiment to investigate buyer behavior. In all markets the market demand curve was
P Q =
38 − Q, where P is market price and Q is total market output, and the cost of installing a unit of capacity was 10.
2
The results reported are from the 21st round of 22, where the fact that round 22 was the last one was not announced until the beginning of that period.
3
Here, results reported are from round 14 which again was the last period before the end-of-game period was announced.
150 A. Muren J. of Economic Behavior Org. 41 2000 147–157
The markets were all organized to simulate a Kreps and Scheinkman market in that sellers first chose capacities, next they were informed about the other sellers’ capacity choices and
then they set their prices. After each period all prices were listed. If needed, the rationing mechanism was that the seller with the lowest price in each market was the first-priority
seller, then the seller who set the second lowest price could sell and so on until all units demanded at the going price were sold.
In the first 1998 round a total of nine triopoly markets were run, six with subjects who did not have any experience of this kind of experiment and three with subjects who had
already participated in one of the previous unexperienced sessions. In the second 1999 round a total of seven markets were run, four inexperienced and three experienced. All
sessions had ten market periods. The time allowed between the respective inexperienced and experienced rounds was approximately two weeks.
Sessions were conducted with pen and paper. Subjects calculated their own earnings after each market period using pocket calculators. As stated in the instructions, subjects were
paid the accumulated earnings in all odd or even periods, where odd or even was determined by throwing dice at the end of the experiment. A participation payment of SEK 100 was
added 150 for the 1998 experienced markets. For recruitment reasons losses were not subtracted. All information was given at the beginning of the experiment. Sessions lasted
between 1 and 32 h including the reading of instructions the experienced sessions were quicker.
4
4. Equilibrium predictions