Optimal Weights in a Two-Tier Voting System with Mean-Field Voters.

We analyse two-tier voting systems with voters described by a multi-group mean-field model that allows for correlated voters both within groups as well as across group boundaries. The objective is to determine the optimal weights each group receives in the council to minimise the expected quadratic deviation of the council vote from a hypothetical referendum of the overall population. The mean-field model exhibits different behaviour depending on the intensity of interactions between voters. When interaction is weak, we obtain optimal weights given by the sum of a constant term equal for all groups and a term proportional to the square root of the group's population. When interaction is strong, the optimal weights are in general not uniquely determined. Indeed, when all groups are positively coupled, any assignation of weights is optimal. For two competing clusters of groups, the difference in total weights must be a specific number, but the assignation of weights within each cluster is arbitrary. We also obtain conditions for both interaction regimes under which it is impossible to reach the minimal democracy deficit due to the negativity of weights.