Competing Against Simulated Equilibrium Price Dispersions:AnExperimentOn Internet-Assisted Search Markets

In a four-treatment experiment, we test some of the hypotheses inGarcia-Gallego et al.(2004) concerning competition among a number offirms of which some (or all) are indexed by a price-comparison enginefacilitating buyers' search process. In this paper, we isolateindividual behavior from noise due to other players' actions andlearning, facing each subject with simulated rivals whose prices areextracted from mixed strategy equilibrium distributions. We findsystematic deviations from both theoretical distributions and previousdata obtained in sessions where all players were human. Specifically,departures of experimental data from the corresponding theoreticalpredictions are enhanced in this setting as compared to our previousresearch in which all agents were represented by human players. Thissuggests that the divergence between theoretical and observed pricereported there should not be attributed to noisy learning and strategicuncertainty due to subjects' interaction with other players.Furthermore, economic tests on players' risk attitudes oganize pricingbehavior in meaningful, although not always compatible, ways.