A series of important recent experiments suggest that alterations in the strategic relationship of actions in a game can importantly affect behavioral outcomes. At issue is whether actions are strategic complements or strategic substitutes. Given an ordering of actions, actions are strategic complements when higher action choices by one player provide an incentive for others to make higher action choices as well. Actions are strategic substitutes when the reverse is true, that is, when higher action choices by one player provides an incentive for others to reduce their action choices.
Perhaps the readiest illustration of such an alteration in strategic relationships is the comparison of Bertrand and Cournot oligopolies. In a standard Bertrand environment, price choices are strategic complements. A seller's best response to, say, a positive price deviation by other sellers is to post a similar, albeit slightly lower own price. In a Cournot environment, quantity decisions are strategic substitutes. A seller's best response to a positive quantity deviation by other sellers is to reduce his or her own quantity.
Experiments conducted in carefully stylized environments show that when all else is held fixed, a context where actions are strategic substitutes tends to (a) converge more completely to static Nash equilibrium predictions, (b) adjust more quickly to anticipated nominal shocks and (c) be less prone to tacit collusion, than a comparable context where actions are strategic complements. These behavioral results have potentially important implications for industrial organization. Among other things, they suggest that an analyst's decision to characterize a specific context as involving Cournot rather than Bertrand interactions carries with it implications regarding stability, responsiveness and susceptibility to tacit collusion. These results also suggest that those institutions where actions are strategic substitutes may, as a matter of institutional design, be preferred.
Such implications come as something of a surprise to those familiar with the behavioral oligopoly literature. Several investigators have noted the slow and incomplete convergence in laboratory Cournot markets (e.g., Rassenti, Reynolds, Smith and Szidarosky, 2000, Offerman, Potters, Sonnemans, 2002), with the differences being particularly pronounced when compared to Bertrand markets (Davis, 1999, 2002).
Download
Do Strategic Substitutes Make Better Markets? A Comparison of Bertrand and Cournot Markets
