An empirical game-theoretic analysis of credit network formation

The framework of credit networks provides a flexible and robust model of distributed trust, based on pairwise credit allocations representing commitments to allow transactions. Since issuing credit entails risks as well as benefits, it is unclear whether self-interested and autonomous agents will form viable credit networks. We tackle this question through an extensive simulation-based game-theoretic analysis of a 61-node credit network formation scenario, covering eight environments varying on information and cost-benefit parameters. We find that viable credit networks form in equilibrium given sufficiently high transaction value, or sufficiently low default risk. Although the amount of credit issued in equilibrium is significantly lower than in the social optimum, as is the social welfare achieved, this difference diminishes proportionally as the environment becomes more favorable overall.