‘Friends and Neighbors’ Voting Under Mixed-Member Majoritarian Electoral Systems: Evidence from Lithuania

ABSTRACT

The concept of ‘friends and neighbors’ voting, in which voters are more likely to support locally-affiliated candidates, has a long tradition in the study of elections and representation. The concept was established in the study of American politics and has been tested in other political and social environments. The focus of ‘friends and neighbors’ voting has been on single and multi-seat constituency-based elections. This emphasis leaves open an important question: Does the concept carry over to party-list elections? This article takes advantage of the features of Lithuania's mixed-member electoral system to assess how ‘friends and neighbors’ voting may influence constituency and party-list races. We find that the magnitude of effect for ‘localness’ is greater when candidates dual-list in constituencies and on party lists. However, residency has a significant effect on voting outcomes even for candidates contesting only on the party list and for those in hopeless list positions, suggesting that ‘friends and neighbors’ voting may extend even to electoral systems that are not constituency-based.

Acknowledgements

The authors thank Monika Aleknaviciute, Alice Barker, and Maksym Palamarenko for research assistance. The research was supported, in part, by a grant from the National Science Foundation (SES – 0751662). The opinions expressed in this paper are the authors’ alone and do not reflect those of other individuals or organisations.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Erik S. Herron is the Eberly Family Professor of Political Science at West Virginia University. Emails: [email protected], [email protected]

Michael S. Lynch is Associate Professor of Political Science at the University of Georgia. Email: [email protected]

Orcid

Erik S. Herron http://orcid.org/0000-0002-6263-6464

Notes

1 Tingsten's study in Sweden predates Key.

2 The research design in Fernandez-Duran et al. (2004) focuses on constituency races.

3 See the thorough review of research in Evans et al. (2017).

4 In some cases, the SSD races are determined by majority-runoff rules, like in Lithuania.

5 See Herron et al. (2018a) for a review of the literature on mixed-member electoral systems. It notes that researchers are divided regarding the interaction between the components of the system, with some finding strong evidence of ‘contamination’ (e.g., Cox & Schoppa, 2002; Ferrara et al., 2005; Herron & Nishikawa, 2001; Krauss et al., 2012) and others finding limited or no evidence (e.g., Moser & Scheiner, 2005; Sieberer, 2010).

6 This paper adopts the naming conventions suggested by Herron et al. (2018b).

7 Optimally, we would control for the strength of local party branch organisations. Unfortunately, appropriate data are not available. While this raises endogeneity concerns, it is an issue that must be addressed in future research.

8 See Shugart and Wattenberg (2001) for an explanation of the terminology.

9 This finding is consistent with research about ballot position effects (Marcinkiewicz, 2014; Marcinkiewicz & Stegmaier, 2015).

10 Tavits (2010) shows that local connections benefit the candidate and the party. Our paper differs by showing relative effects of different features in a mixed-member electoral system. Another possible local effect that we cannot fully account for in our assessment is the extent to which candidates have been involved in local politics in elected or appointed positions, or as prominent civil society activists. These characteristics may influence performance, but we do not measure them in this analysis.

11 Electoral blocs must receive 7%.

12 Research shows that personal vote seeking is associated with improved list position (André et al., forthcoming; Crisp, Olivella, Malecki, & Sher, 2013) and that list position affects preference votes (Marcinkiewicz, 2014; Marcinkiewicz & Stegmaier, 2015).

13 The election code, in its current form, is available at: https://e-seimas.lrs.lt/portal/legalAct/lt/TAD/89e6d1a0f58111e5bf4ee4a6d3cdb874?jfwid=tu0odnkka.

14 Lithuania provides information on the initial and final list position for candidates, as well as the preference votes they garnered. See, for example, the Homeland Union results (http://www.vrk.lt/statiniai/puslapiai/2008_seimo_rinkimai/output_lt/rezultatai_daugiamand_apygardose/partijos_pirmumo_balsai3433.html).

15 Researchers have found that individual candidates in low list positions can receive ‘promotions’ for strong performance in flexible list systems (Crisp et al., 2013), a phenomenon that may travel.

16 We abbreviate the name Homeland Union – Lithuanian Christian Democrats throughout the article.

17 Candidates are likely to campaign close to their residence; thus effects may be localised (Gorecki & Marsh, 2012). Also, voters are likely to reward locals (André et al., 2012; Tavits, 2010).

18 Research on preference votes in Belgium has revealed that politically involved voters are more likely to participate and cast ballots beyond the top of the list. Candidates lower on the list receive support when they are known by voters (André et al., 2012).

19 The dependent variable is a count of preference votes, so we employ a count model. Given that the preference votes variable shows evidence of over-dispersion throughout the models presented, negative binomial models are reported. All models report negative binomial coefficients and robust standard errors clustered by candidates. Given the difficulty of interpreting the substantive impact of negative binomial coefficients, OLS models are presented in the online appendix in Tables A4, A5, and A6. While the dependent variable is not normally distributed, the results from these OLS models mirror the substantive results reported in the negative binomial models and the coefficients from these OLS models are directly interpretable as changes in the predicted number of preference votes.

20 Additional analysis reporting models for the Liberal and Central Union, Liberal Movement, Order and Justice Party, and the Rising Nation Party in 2008 and Labour Party, Liberal Movement, Order and Justice Party and Way of Courage Party in 2012 are located in the online appendix on Table A2.

21 Additional analysis reporting models for the Liberal and Central Union, Liberal Movement, and Order and Justice Party in 2008 and Labour Party, Liberal Movement, and Order and Justice Party in 2012 are located in the online appendix on Table A3.

22 This variable is omitted for the Homeland Union model in 2008 because no incumbents fit the category.

23 To generate the outcomes, we set the birth district and incumbency at zero, residency at one, age at 45, and gender at male. The results for dual-listed candidates uses the coefficients from Table 2. The results for non-dual listed candidates uses the coefficients from Table 3.

24 As noted above, the findings of this study provide additional support to the argument that ‘contamination effects’ cross from SSD to PR in mixed-member systems.

Additional information

Funding

The research was supported, in part, by a grant from the National Science Foundation (SES – 0751662).