Abstract:
This paper studies the performance of four market protocols with regard to allocative efficiency and other performance criteria such as volume or volatility. We examine batch auctions, continuous double auctions, specialist dealerships, and a hybrid of these last two. All protocols are practically implementable because the space of messages for traders is simple. We test the protocols by running (computerized) experiments in an environment that controls for traders’ behavior and rules out any informational effect. We find that all protocols generically converge to the efficient allocation in finite time. An extended comparison over other performance criteria produces no clear winner, but the presence of a specialist is clearly associated with the best all-round performance.