The co-evolutionary dynamics of directed network of spin market agents

The spin market model [S. Bornholdt, Int.J.Mod.Phys. C 12 (2001) 667] is extended into co-evolutionary version, where strategies of interacting and competitive traders are represented by local and global couplings between the nodes of dynamic directed stochastic network. The co-evolutionary principles are applied in the frame of Bak - Sneppen self-organized dynamics [P. Bak, K. Sneppen, Phys. Rev. Letter 71 (1993) 4083] that includes the processes of selection and extinction actuated by the local (node) fitness. The local fitness is related to orientation of spin agent with respect to instant magnetization. The stationary regime characterized by a fat tailed distribution of the log-price returns with index $\alpha\simeq 3.6$ (out of the Levy range) is identified numerically. The non-trivial consequence of the extremal dynamics is the partially power-law decay (an effective exponent varies between -0.3 and -0.6) of the autocorrelation function of volatility. Broad-scale network topology with node degree distribution characterized by the exponent $\gamma=1.8$ from the range of social networks is obtained.