Inequality measures perform differently in global and local assessments: An exploratory computational experiment

• We simulate incomes distributed in networks using the Exponential Random Graph Model.
• We compare population and network inequality using a selection of inequality measures.
• Network inequality tends to be lower than population inequality for the Gini index.
• Network inequality tends to be lower when inequality of nodal degree increases.

Inequality measures are widely used in both the academia and public media to help us understand how incomes and wealth are distributed. They can be used to assess the distribution of a whole society–global inequality–as well as inequality of actors’ referent networks—local inequality. How different is local inequality from global inequality? Formalizing the structure of reference groups as a network, the paper conducted a computational experiment to see how the structure of complex networks influences the difference between global and local inequality assessed by a selection of inequality measures. It was found that local inequality tends to be higher than global inequality when population size is large; network is dense and heterophilously assorted, and income distribution is less dispersed. The implications of the simulation findings are discussed.