![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
In response to alleged citizen distrust of politics and parties, recent reforms have made electoral systems more candidate-centred. Rational choice theory claims that candidate-centred voting should make voting more rewarding while incentivising candidates to increase their campaigning. Behaviouralists point to the higher number of options, which would lead to abstention. This study investigates the reforms in Bremen and Hamburg, where the simple, single-vote, closed-list PR system was replaced with a multiple-vote preferential-list PR system in 2011. Employing a difference-in-differences design, the study reveals a significant decrease in voter turnout of at least ten percentage points in Bremen, Bremerhaven, and Hamburg, compared to the trend in comparable German districts. A cross-level analysis of 79 election surveys fielded since 2000 in all of Germany, comprising 97,849 respondents, highlights that the reforms have widened the participation gap between the lower- and higher-educated Bremen and Hamburg citizens compared to elsewhere in Germany. These findings contribute to research on electoral systems, choice, and voter turnout, emphasising the potential adverse consequences that can emerge from complex electoral system redesigns.
Acknowledgement
The author is most grateful for the guidance and feedback of Simon Hix and Miriam Golden. The paper significantly benefited from feedback and comments from Ruth Dassonneville, Filip Kostelka, Pedro Riera, the attendees at the Electoral Laws and the (In)equality of Democratic Representation panel of the APSA Annual Meeting 2023, the editor, and two anonymous reviewers.
Data Availability Statement
Data and replication material can be found on Cadmus, the European University Institute Research Repository, see https://cadmus.eui.eu/handle/1814/76670.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Supplemental Data and Research Materials
Supplemental data for this article can be accessed on the Taylor & Francis website, https://doi.org/10.1080/09644008.2024.2326457.
Notes
1 Renwick and Pilet count twenty-one such reforms in Europe between 1990 and 2009 (Renwick and Pilet Citation2016, 267). The appendix includes a list of recent candidate-centered electoral system reforms worldwide.
2 Still, voters’ effective influence on candidate selection is limited in what Däubler and Hix (Citation2018, 1803) call systems with ‘weakly flexible lists’, i.e. when other candidates than the a priori top-ranked candidate(s) fail to reach the threshold for getting elected.
3 However, within the rational-choice framework, one could also argue that the cost of informing oneself about all alternatives logically increases the more options there are, leading individuals to abstain (see Matsusaka Citation1995; Lassen Citation2005; Coupé and Noury Citation2004; Dalton and Wattenberg Citation2000). Feddersen and Pesendorfer (Citation1996) even suggest that abstention may be an optimal strategy for uninformed voters because it maximizes the probability that informed voters decide the election.
4 Another response is to rely on problem-solving strategies, or heuristics, to make reasonable decisions with minimal cognitive effort (Tversky and Kahneman Citation1974). For example, Lau and Redlawsk (Citation2010) show that voters rely on a candidate’s party affiliation, ideology, and endorsements from interest groups or other politicians. Augenblick and Nicholson (Citation2016) also find evidence for these heuristics, showing that choicefatigued voters will support the first candidate in a race and are likelier to vote ‘no’ on a proposition, thus endorsing the status quo. Voters have also shown an increased propensity to guide their decisions by information with seemingly less political relevance (e.g. Cunow et al. Citation2021). In Japan, for example, voters must write down the candidate’s name on a write-in ballot. Candidates with names spelled with simpler Kanji characters tend to get written down more than candidates with more complex names the greater the number of candidates one must choose between, all else equal (Muraoka Citation2021).
5 Bremen and Hamburg did not use such a system for their local elections before the 2011 reforms. Elections for the city councils of Bremerhaven and Bremen have been held concurrently since 1991, with few exceptions, and were also reformed to a multiple-vote preferential list in 2011. There is only no five percent threshold employed for city council elections, see https://www.wahlen.bremen.de/bremer-wahlen-6925. In Hamburg the elections for district assemblies - also held concurrently with State Parliament elections – were reformed in 2011 to multiple-vote preferential-list, similar to the system for state elections.
6 256,507 voters voted in favor, constituting 21.1 per cent of all eligible voters, thus passing the minimum threshold of 20 per cent before the referendum is considered valid.
7 Source: German Wikipedia, Public domain, via Wikimedia Commons. https://commons.wikimedia.org/wiki/File:HB-Wahlzettel%202011.jpg. For the Hamburg ballot used in 2011, see https://www.hamburg.de/contentblob/2753534/f8ec257836b8a08f1fe1520da5afb3b9/data/musterstimmzettelheft-wk-16.pdf, consulted on 15 January 2024. On the Hamburg ballot, voters first have five votes to distribute/cumulate over the parties and candidates at the state-wide level (Landeslisten) and then have five votes to distribute over parties and candidates in their local district (Wahlkreis).
8 Out of 408 thousand eligible voters in Bremen and 86 thousand in Bremerhaven.
9 A party’s total vote share consists of the votes for the list and votes for individual candidates. The system initially considered list votes first and then the candidate votes. E.g. when X per cent of a party’s total votes are list votes, the system allocates X per cent of the party’s seats in order of the list. The system allocates the remaining seats to not-yet-elected candidates with the highest candidate votes. Because the candidates ranked highest on the list tend to receive the most votes, they will usually have been allocated a seat already. Therefore, the initial procedure benefited lower-placed candidates, like the FDP candidate mentioned here. By reversing the allocation procedure, higher-placed candidates are once more favored, strengthening the party’s influence over which of their candidates wins.
10 In Bremen, the rate rose from 1.3 per cent in 2007 to 3.1 per cent in 2011, and in Hamburg, it increased from 1 to3 per cent.
11 Or Kreise, as they are known in North Rhine-Westphalia and Schleswig-Holstein.
12 The untreated member-states are Baden-Württemberg, Bayern (Bavaria), Berlin, Brandenburg, Hessen (Hesse), Mecklenburg-Vorpommern, Niedersachsen (Lower Saxony), Nordrhein-Westfalen (North RhineWestphalia), Rheinland-Pfalz (Rhineland-Palatinate), Saarland, Sachsen (Saxony), Sachsen-Anhalt (SaxonyAnhalt), Schleswig-Holstein, and Thüringen (Thuringia).
13 By generating an optimally weighted control unit, the SDID approach extends the principles of the synthetic control method (SC) (see Abadie Citation2021). However, SC requires that the treated unit(s) reside within a ‘convex hull’ formed by the control units. This requirement becomes a limitation in scenarios where, for instance, the treated unit(s) exhibit higher scores on the relevant variables than (most) control units. SC will fail to synthesise a control with identical pre-treatment trends in such instances. Indeed, estimates produced by the ‘synth’ package in Stata show that it is not possible to approximate the characteristics of Bremen and Hamburg using a weighted average of untreated districts (see Section 2.1 in the Supplementary Materials), in which case the use of synthetic controls is advised against (406). Where SC does not allow for level differences between treated and control units, SDID generates a synthetic control with parallel - rather than identical - trends, offering a more flexible analytical approach (see Arkhangelsky et al. Citation2021, 4092).
14 EKF002 of 73111 Wage and income tax statistics from Regionaldatenbank Deutschland.
15 See regionalstatistik.de/genesis/online. See Section 1.1 in the Supplementary Materials for all descriptive statistics.
16 In 1994, Bremen had a population of 549,357, Bremerhaven 130,400, and Hamburg 1,707,901. Subsetting is desirable because the covariates included in SDID only adjust for differences between treated and control units in before-to-after changes, not whether treated and control units have similar levels in covariates.
17 Under TWFE all control units are attributed the same weight, i.e. . All unit and time weights are listed under Sections 2.3 and 2.4, respectively, in the Supplementary Materials.
18 Refer to Section 3 of the Supplementary Materials for detailed survey references.
19 See Table 4 in the Supplementary Materials for all descriptive statistics.
20 Triple-difference estimates are hard to understand without predictive probabilities given by treatment group, period, and, in this case, education status, which will be presented in the following section. Therefore, I do not include fixed effects as in Equation (3) because a model with Land- and election-cycle fixed effects will not be capable of estimating these parameters. However, a fixed-effects model will arguably be more precise in estimating the treatment effect. Therefore, estimates from the model in Equation 4 but with Land- and time-fixed effects included can be found in the Supplementary Materials, Table 17. The results do not lead to a different interpretation of the below results.
21 Following notation from Olden and Møen (Citation2022), the estimate is derived after solving:
(5)
(5) The triple-differences estimate derived in a regression with unit- and time-fixed effects is 2.46, significant at the 0.001 level, See Table 17 in the Supplementary Materials.
Additional information
Funding
Notes on contributors
Victor Ellenbroek
Victor Ellenbroek is a PhD student at the European University Institute. His website can be found at www.victorellenbroek.com