Electoral Reform & Strategic Voting in Chilean Legislative Elections

ABSTRACT

This paper examines the effects of Chile's 2015 electoral reform, which increased the magnitude of districts in Chile's lower legislative chamber from two to between three and eight. This increase in district magnitude allows for a convenient ‘natural experiment’: post-reform voters should strategically vote at lower levels than pre-reform voters. I find strong evidence that voters have responded to Chile's electoral reform by engaging in less party-level strategic voting. However, I surprisingly find no evidence that the reform has decreased strategic voting for Chile's pre-electoral pacts. This study demonstrates that Chile's electoral reform is already affecting voter behaviour and is doing so in line with theoretical expectations.

KEYWORDS:

• Electoral systems
• Chile
• electoral reform
• strategic voting
• tactical voting

Acknowledgments

I thank Matthew Søberg Shugart, James Adams, Christopher Hare, Ethan Scheiner, and Oakley Gordon for their consultation on this project and their thoughtful feedback on this paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 District magnitude is understood to be the primary driver of electoral proportionality (e.g., Droop, 1881; Shugart & Taagepera, 2017), though the direction of this effect is conditioned by whether the electoral system employs a proportional or plurality formula.

2 The methodological advantages of the case study approach are well-explained by Eckstein (2000).

3 The applicability of Duverger's ‘law’ has been the subject of significant criticism and scholarship (e.g., Shugart & Taagepera, 2017).

4 ACE Project: Chile: A System Frozen By Elite Interests

5 Popularly referred to in Chile as the ‘sistema binominal’.

6 For additional yearly information on the Chilean party system in the Chamber, refer to the supplementary material.

7 The ‘effective number of parties’ is a measure developed by Laakso and Taagepera (1979) to weight the count of vote-earning or seat-earning parties by their relative strength.

8 Per Shugart and Taagepera (2017), the proportionality of an electoral system is linked to both district magnitude and assembly size.

9 Because a large number of voters decline to state their vote choice or party identification in the CSES, and the population weights correspond to the entire sample of voters, it is not feasible to determine the aggregate incidence of strategic voting in the 2005, 2009, and 2017 legislative elections.

10 1,200 voters in 2005, 1,200 in 2009, and 2,400 in 2017.

11 In the supplementary material, I conduct additional robustness checks where I further restrict the universe of respondents to those who had a candidate from their most-preferred party/coalition running in their district. The results from those analyses are substantively identical to those presented here.

12 While the Chilean Senate was also the subject of electoral reform, the CSES did not begin asking voters about their Senate vote choice until 2017. As such, I focus my analysis on Chamber elections.

13 Voters who named a coalition, not a party, when asked to name either the party they feel closest to or the party that most represents their views, and also do not have a most-liked party, are coded as missing data for the party-level strategic vote variable.

14 Voters who reported identifying with or voting for a party coded as ‘other’ in the CSES survey data are coded as missing data for the strategic vote variable.

15 Voters who reported voting for an coalition, not a party, are treated as missing data for the party-level strategic vote variable.

16 Though doing so excludes a majority of my sample, I also run my model of strategic party voting including only voters who named a party when asked if there is a party they feel closest to. The output for this model are in the supplementary material. The substantive interpretation of my models is unchanged by this sample restriction.

17 A voter's most-preferred party can remain the same across all three elections, but their most-preferred coalition may nevertheless shift under this coding scheme. This occurs because the partisan composition of Chile's electoral coalitions changes each election. To provide one example, the Chilean Communist Party was part of the ‘Juntos Podemos Más’ coalition in 2005, the ‘Concertación y Juntos Podemos’ coalition in 2009, and the ‘La Fuerza de la Mayoría’ coalition in 2017.

18 For example, consider a 2017 voter who gave 10 out of 10 ‘like’ scores to both the Socialist Party and the Communist Party (both members of the ‘La Fuerza de la Mayoría’ coalition). If they gave no other parties outside the ‘La Fuerza de la Mayoría’ coalition at 10 out of 10, the voter clearly prefers that coalition.

19 Voters for whom no district magnitude was listed are omitted from my sample.

20 While some previous studies have also accounted for political interest (e.g., Bolstad, Dinas, & Riera, 2013; Fisher, 2001; Meffert & Gschwend, 2011), such data are not available in the 2005 and 2009 CSES surveys.

21 The results of a hierarchical version of this model are reported in the supplementary material. The substantive results are identical to those presented here.

22 Because all Chamber elections are $\mathrm{M}=2$ prior to reform and all Chamber elections following reform are $\mathrm{M}>2$, it is not feasible to interact district magnitude with an electoral reform indicator variable. Instead, year fixed effects are included separately.

23 One might reasonably believe that effect of education and district magnitude on strategic voting is interactive, not additive. In the supplementary material, I re-run the party and coalition models interacting logged district magnitude and education. The effect of district magnitude on strategic voting is not significantly conditioned by education in the interactive model, even at lower district magnitudes.

24 For similar plots for the 2017-only models, see the supplementary material.

25 For more information on the intra-coalition politics of candidate nomination in these $\mathrm{M}=2$ districts, see J. M. Carey and Siavelis (2005).

26 Gschwend (2009) notes that small levels of strategic voting persist even at high district magnitudes. This aligns with my results.

27 As shown in the supplementary material, the findings presented here are not a product of increased candidate supply following the 2017 electoral reform. When the universe of respondents is restricted to those who had a candidate from their most-preferred party/coalition running in their district, the substantive results of logistic regression analysis are the same.

28 BBC News: ‘Jubilation as Chile votes to rewrite constitution’

29 Washington Post: ‘Chile to hold referendum on new constitution’ (11/15/2019)

30 Reuters: ‘Chile's govt in shock loss as voters pick independents to draft constitution’

Additional information

Notes on contributors

Isaac Hale

Dr. Isaac Hale is an Equity Research Postdoctoral Scholar at the UC Santa Barbara Blum Center on Poverty, Inequality, and Democracy. Prior to arriving at UCSB, he served as a Lecturer of Political Science at UC Davis. His research studies how racial attitudes and the design of electoral institutions can undermine representation in the US and other developed democracies. To learn more about Isaac, visit his website at www.isaacdhale.com.