An agent-based interaction model for Chinese personal income distribution

•The personal income distribution in China was studied by employing the empirical data.•Results from an agent-based model with mean-field approach and the model on networks, the simulation and empirical data have been compared.•Consider the high-order neighbors in networks as the global factor when agents exchange their wealth in this model.•To characterize the empirical findings, simulation on different networks has been studied.

The personal income distribution in China was studied by employing the data from China Household Income Projects (CHIP) between 19901990 and 20022002. It was observed that the low and middle income regions could be described by the log-normal law, while the large income region could be well fitted by the power law. To characterize these empirical findings, a stochastic interactive model with mean-field approach was discussed, and the analytic result shows that the wealth distribution is of the Pareto type. Then we explored the agent-based model on networks, in which the exchange of wealth among agents depends on their connectivity. Numerical results suggest that the wealth of agents would largely rely on their connectivity, and the Pareto index of the simulated wealth distributions is comparable to those of the empirical data. The Pareto behavior of the tails of the empirical wealth distributions is consistent with that of the ‘mean-field’ model, as well as numerical simulations.