Do Strategic Substitutes Make Better Markets? A Comparison of Bertrand and Cournot Markets

Recent experiments suggest that games where actions are strategic substitutes rather than strategic complements exhibit some desirable performance characteristics. This paper reports an experiment conducted to test whether these characteristics extend to differentiated product Cournot and Bertrand markets. We find that Cournot markets do not generally outperform Bertrand markets, and that the opposite is often true. Bertrand markets exhibit comparatively higher convergence levels and speeds, particularly when products are close substitutes. Bertrand sellers do engage in more signaling activity than Cournot sellers. Such efforts, however, affect market outcomes only occasionally, and only when products are differentiated. Analysis of individual decisions suggests that the observed differences in convergence levels and speeds are driven by a propensity for sellers to use forecast and inertia anchors as bases for action choices in addition to best replies. Given these propensities, Bertrand markets converge more rapidly and more completely to static Nash predictions, particularly when products are close substitutes, because the differences between the various anchors is smaller in Bertrand markets. .