Abstract
We study how information is learned and diffused through repeated bilateral trading in decentralized markets with asymmetric information. We combine laboratory experiments with a calibrated behavioural agent-based model and compare three benchmark network topologies—a ring, a small-world network, and a connected Erdös–Rényi-type random network—while holding the number of potential trading relationships fixed. The experimental evidence shows that network organization and the location of perfectly informed traders generate systematic differences in quotes, order flow, profits, market efficiency, and inequality. The joint pattern supports a gradual learning process based on repeated and interdependent local interactions rather than independent exposure to informed traders. The computational model uses order-flow imbalance as the learning channel, reproduces key experimental patterns, and makes it possible to trace the speed and location of belief updating. Simulations show that information remains more localized around insiders in regular networks, whereas random connections broaden its reach and make proximity to insiders less predictive of learning. Together, the experimental and computational results highlight how the architecture of decentralized trading relationships shapes both information diffusion and the distribution of informational advantages.
View Publication