Simulation of Japanese National Diet Members Election System Using Agent-Based Modeling

Abstract

This article reports the investigation of the influence to the result of election with the change of electoral system. The latest election in Japan was held in 2017. In this election, the leading party, Liberal Democratic Party (LDP), got 48% of total votes and got 76% of seats. Some people considered this result as inadequate. In other words, current electoral system does not or cannot reflect the voters’ willing appropriately. What is the appropriate election system? Current Japanese election system uses the combination of single-seat system and proportional representation system. Is this the optimal system? This is our research question.In this article, we assume that there are distinct n political issues. Each people (voter), political party, and candidate has its own opinion. This opinion is represented by the value from \(-1\) to +1. We also represent each people, political party, and candidate as agents. Each agent has its own political position represented as a n-dimensional vector. All voters’ willing is represented as a composition vector of all voters’ political vectors. We compare the composition vector of all voters’ political vectors and that of all successful candidates. Difference of voters’ composite vector and successful candidates’ composite vector is observed when we change the electoral system. Our simulation results show that multiple-seat selection system is less reflects the people’s opinion and there is no significant difference in other electoral systems.KeywordsAgent-based modelingElection systemVector space modeling