Who makes the pie bigger? An experimental study on co-opetition

Abstract

The tension between cooperation and competition that characterizes many business relationships is experimentally studied in a ‘pie’-creation game; value is created and increased through cooperation in a repeated prisoner's dilemma game. At the end, the player with the greater stake in the joint pie decides on the division of the pie. Three treatments of the pie-creation game are considered: in the first treatment, rivals create the pie; in the second, non-rivals create the pie; finally, in the third, the pie is created by subjects who do not know about the future pie-division. The data show that the competition for the right to split the pie biases behaviors towards defection when subjects play with their rival.

Keywords:

• competition
• cooperation
• co-opetition
• ambiguously repeated prisoner's dilemma
• experimental economics

Notes

1. Related literature considers that this term comes originally from Raymond Noorda in 1993 (e.g. Gee, 2000; Ketchen et al., 2004; Walley, 2007).

2. For instance, national railway companies of neighboring countries cooperate in the construction of international railways to compete for customers and freight transportation with other carriers; chip-manufacturers and software producers cooperate in advanced technology development.

3. Job-markets in universities frequently allow only external candidates to fill their advertised positions, and successful team-players may have comparative advantages over non-team-players.

4. Similar approaches have been followed in Palfrey and Rosenthal (1988) and Samuelson (1987) to provide a rationale for cooperation in the finitely repeated prisoner's dilemma game.

5. We chose the ambiguously repeated prisoner's dilemma game setting to have the same number of periods in each session and because it is a good presentation of the real world. Natural people and corporations usually do not know when a relationship ends and they do not assign probabilities to such events, if at all they rather assign maximum time length to such relationships.

6. Several articles have analyzed the effect of infinite and finite horizon on cooperation in prisoner's dilemma games (see among others Bruttel et al., 2007; Cooper et al., 1996; Dal Bó, 2005; Normann & Wallace, 2005; Selten & Stoecker, 1986).

7. The trials were aimed at reducing initial noise and confusion and have been reported in other laboratory studies (Andreoni, 1995).

8. However, a few observations regarding the pie-division game follow. Half of the dictators took 90% or more of the total, and 37% took it all. The average demand by the dictators were as follows; 79% in the control treatment (74% Germany, 83% Spain), 87% in the rival treatment (83% Germany, 92% Spain), and 83% in the non-rival treatment (80% Germany, 86% Spain) and overall. The differences in claims between the rival and the control treatment are significant (p = 0.031), as the two-tailed Wilcoxon-Mann-Whitney test suggests.

9. The random-effects logit regression of the relative choice of cooperate in a group per period {0, 0.5, 1} indicates that the decrease of cooperation levels is significant in the rival (non-rival) treatment; the p-value of the declining trend is 0.001 (0.004). The corresponding p-value for the control treatment is 0.269.

10. The random-effects logit regression involving country dummies on trend and intercept suggests no significant differences in the trend but all treatment dummies are significant on the intercept; the corresponding p-values are greater than 0.200 and smaller than 0.100, respectively. The one-tailed Mann-Whitney test results show that the differences are significant for each treatment; the p-values are 0.016, 0.015, and 0.021 for the control, the rival and the non-rival treatments.