Effects of price signal choices on market stability.pdfVIP

Eects of Price Signal Choices on Market Stability Hideyuki Mizuta Ken Steiglitz and Erez Lirov IBM Tokyo Research Laboratory Computer Science Department Princeton University April Abstract We study the eects of dierent price signals on the stability of an agentbased auctioncleared automated market using b oth simulation and analysis The results show that such markets can b e stabilized using a price signal that is the average of all bids and that this signal leads to more stable b ehavior if the bids are weighted inversely with the size of the oers This conrms the notion that information in the order b o ok is valuable and suggests new more stable algorithms for marketbased control Intro duction Agentbased market simulations are attracting increasing attention recently not only in economics to study markets themselves but also in computer science for distributed control of large systems Agentbased technology in computer systems oers a simple and ecient means for distributed optimized control of large and complex systems But markets inevitably entail the feedback of information in the form of price signals and like all feedback systems may exhibit unstable b ehavior It has b een shown in recent work see for example that valuebased traders can stabilize such markets and that trendbased traders can destabilize such markets causing b oth inationary and deationary price bubbles In this pap er we investigate as an alternative to value traders the eects of dierent price signals on the dynamic b ehavior of prices in an agentbased market simulation Simulation We implemented market simulations using b oth