ABSTRACT
It has been suggested that one of the reasons why majoritarian electoral systems are associated with lower voter turnout in comparison to proportional electoral systems is that citizens in uncompetitive districts (“safe seats”) are not motivated to vote. This study brings this thesis into a new context and tests it with unique data. Mixed-member electoral systems have both majoritarian and proportional components. The relative importance of these components differs between mixed-member proportional (MMP) and mixed-member majoritarian (MMM) systems. I argue that, due to this difference, the impact of district-level competitiveness on turnout is stronger in MMM than in MMP. An analysis of district-level electoral data from four countries confirms this hypothesis. Findings from this study advance our theoretical understanding of voter participation and also of the functioning of mixed-member electoral systems.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes on contributor
Ko Maeda (Ph.D., Michigan State University, 2005) is an associate professor of political science at the University of North Texas, specializing in elections, party competition, and political institutions.
Notes
1. The sources are: http://www.electproject.org/Election-Project-Blog/2014generalelectionturnoutrates, accessed on June 19, 2015, and http://www.asahi.com/articles/ASGDH32GSGDHUTFK002.html, accessed on December 16, 2014.
2. The relationship does not necessarily hold in developing countries (Stockemer Citation2015).
3. More specifically, MMP systems were combined together with PR systems with upper-tier seats (such as Sweden’s) as “supradistrict corrections,” and MMM systems were not distinguished from other non-mixed systems.
4. For example, in Japan in 2005, the ruling Liberal Democratic Party (LDP) increased its average SMD vote share by 3.9 percentage points, which resulted in a dramatic jump in its SMD seat share, from 56% to 73% (Maeda Citation2006). The pattern continued in the subsequent elections (see, e.g. Maeda Citation2010).
5. In the most recent elections in Japan and South Korea, 62.1% and 82.0% of the seats were allocated to SMD, respectively.
6. The data for the Bundestag elections were obtained from Kollman et al. (Citation2014).
7. The data for the MMP elections from 1999 to 2011 were obtained from Kollman et al. (Citation2014), and the results of 1996 and 2014 elections were downloaded from the government’s website at http://www.electionresults.govt.nz/, accessed on March 13, 2015. New Zealand’s electoral system includes Maori districts that are reserved for its indigenous population. Since the turnout is significantly lower in Maori districts, I consider that they are not comparable with other districts, and thus they were not included in the analysis.
8. The data sources are: 1996 – obtained from the government through a direct correspondence; 2000 – http://www.shinmai.co.jp/senkyo/2000senkyo/, accessed on August 4, 2004; 2003 – http://www.sankei.co.jp, accessed on November 9, 2003; 2005 – http://www.sankei.co.jp, accessed on September 12, 2005; 2009 – http://www.nikkei.co.jp, accessed on August 30, 2009; 2012 – http://www.yomiuri.co.jp/election/shugiin/2012/kaihyou/, accessed on December 17, 2012; 2014 – http://www.yomiuri.co.jp/election/shugiin/2014/kaihyou/, accessed on December 15, 2014.
9. The data for the redistricting were obtained from the government’s website at http://www.soumu.go.jp/senkyo/senkyo_s/news/senkyo/shu_kuwari/index.html, accessed on March 7, 2015.
10. Since the correlation coefficients in this analysis aggregate all elections of different years, the distribution of the observations for each election need to be moved to the center. Specifically, the average values of the margin changes and turnout changes were calculated for each election, and the average values were subtracted from all values so that the center of the distribution for each election comes to zero.
11. Two hundred and sixty seven districts were excluded due to this reason. Two hundred and sixteen of them are from South Korea, and the rest (46) are from Japan. As a robustness check, I repeated the analysis while changing the threshold to 10% and 30% and obtained similar results, information is available upon request to the author.