Political scientists have long wondered why parties converge on the middle in theory but drift apart in practice. A study published June 23 in PLoS One by C. Campbell and G. J. Ackland of the University of Edinburgh’s physics side of the question, offers a model that can produce both outcomes.
The classic Hotelling-Downs model has parties adjusting their policies to win as many votes as they can. On a single left-right axis, that pushes them toward centrist platforms. It once fit reality well, at a time when the two major US parties advocated nearly identical policies. Over the following decades polarisation grew in the US and elsewhere, and the model stopped matching what voters saw.
Votes as energy
The authors borrowed from statistical physics. They treat votes the way a physicist treats potential energy, with parties moving through opinion space the way particles settle toward lower energy. Their simulation uses a Metropolis Monte Carlo method, written as open-source Python code.
They also went past one dimension. Two issues can be modelled at once, such as economic views from left to right and cultural views from liberal and multicultural to traditional and nationalist. The grid is square, so a voter can be extreme on both issues at the same time. Voters can be bunched in one peak or split between two.
What the simulations showed
Two-party systems came out less polarised, even when the voters themselves were split into two camps. Multiparty systems did the opposite. With more parties, they moved toward the peaks of the voter distribution and polarised, even when most voters held moderate positions.
The team then added two pressures: turnout and activists. Extreme actors are more likely to stand as candidates and to vote. Adding these effects produced more polarisation, even with only two parties. The authors say this suggests party polarisation can be driven by abstention, intra-party politics and turnout at the extremes.
In the two-party case, the winning party’s positions are more moderate than those of its supporters, and better represent the electorate as a whole. Polarisation cuts the other way. Individual voters find it easier to locate a party that matches their views, but the government, whether a winning party or a coalition, represents the population less well. That holds even when the population agrees on every issue.
The authors point out that earlier models with more than two parties or more than one policy dimension often failed to reach stable equilibria. Theirs supports both polarised and consensual outcomes, and they describe it as a possible basis for more complete models later.
The paper reports no specific funding and no competing interests. Code is on GitHub, and an online demo and a game version of the one-dimensional model are hosted by the University of Edinburgh physics department.
Sources
- A computational model of spatial politics: Hotelling-downs model as statistical physics
- A computational model of spatial politics: Hotelling-Downs model as statistical physics
