Do voters choose rationally, irrationally or at random? Data and theory for proportional elections

TL;DR

This paper investigates voting patterns in proportional elections, revealing a Zipf's law distribution, and develops a population genetics-inspired model to explain voter behavior and party dynamics.

Contribution

It introduces a novel mathematical model based on word-of-mouth voter grouping that explains election data and party behavior without relying on political content.

Findings

Election votes follow Zipf's law across multiple countries.

The model accurately reproduces observed vote distributions.

Party dynamics such as threshold effects and party number trends are explained.

Abstract

Data of proportional elections show a striking feature: If the parties are ranked according to the number of their voters, the number of votes grows exponentially with the rank of the party. This so-called Zipf's law has been reported before. We first show this correlation in results from recent elections in Germany, France and republican primaries in the USA. However the mechanism generating such feature remains so far unexplained. We develop a mathematical model of voter grouping that is only based on the word of mouth and not on political contents. The model is close to the infinite allele model and the Ewens sampling formula that are well known in population genetics. Strikingly, the model generates output agreeing very well with the observed election data. We further identify a cannibalism effect in Germany, whereby parties above the 5% threshold withdraw votes from parties just below the threshold. We demonstrate that the number of parties present in the parliament can be predicted given the number of parties standing for election. Finally, the steady loss of big-tent parties in Germany in the last 20 years relates to the increasing number of parties since 1975. We discuss the interpretation and consequences of these findings for modern democracies and the rationality or lack thereof of voters' choices.