The Strategic Impact of pace in Double Auction Bargaining

This paper evaluates performance of human subjects and instances of a bidding model that interact in continuous-time double auction experiments. Asks submitted by instances of the seller model (“automated sellers”) maximize the seller’s expected surplus relative to a heuristic belief function, and arrive stochastically according to an exponential distribution. Automated buyers are similar. Across experiment sessions we vary the exponential distribution parameters of automated sellers and buyers in order to assess the impact of the relative pace of asks and bids on the performance of both human subjects and the automated sellers and buyers. In these experiments, prices converge and allocations converge to efficiency, yet the split of surplus typically differs significantly from the equilibrium split. In order to evaluate the impact of pace, a statistical model is developed in which the relative performance of sellers to buyers is examined as a function of the profile of types present in each experiment session. This econometric model demonstrates that (1) human buyers outperform human sellers, (2) automated sellers and buyers with a longer expected time between asks or bids outperform faster automated sellers and buyers, and (3) the performance of the faster automated buyers is comparable to that of human buyers.