Directed Gossip Algorithms, Consensus Problems, and Stability Effects of Noise Trading
This paper provides a new algorithm that can be used to address cooperative control problems, and namely consensus problems, in networked multi-agent systems and elaborates on connections to the behavior and stability properties of social and economic systems. The paper thereby contributes both to systems theory and to computational financial economics. Motivated by the human communicational and computational limitations, as well as prior work in economics (Kirman, 1993), and engineering (Boyd et al., 2006), we introduce "Directed Gossip Algorithms" and interpret them in the context of financial markets. We propose that it is not the presence of a large number of agents who "follow the bandwagon" that causes excess volatility and bubbles, but rather the absence of fundamentalists is. Our theory may help explain why "thin" markets with a small number of participants are more volatile than developed markets.
Martin SCHMALZ, Masayuki FUJITA, Oliver SAWODNY
Agent-based Finance, Investor Sentiment, Market Dynamics, Consensus, Emergence