101 - 146
Jan 23, 2023
In 2020, Chile held a historic plebiscite that decided, by a broad margin, to hold a constitution-making process. The vote took place in a tumultuous context, marked by the 2019 social outburst and the COVID-19 pandemic. This article explains how these circumstances were associated with turnout, which increased slightly compared to previous elections. District-level regression analyses of Chile’s 345 municipalities suggest that different impacts of the pandemic were both positively and negatively associated with turnout. Additionally, exposure to both nonviolent and violent protests during the social outburst depressed turnout, particularly in more right-wing districts.
We thank the editors of Economía y Política and two anonymous reviewers for their careful reading of the manuscript and their helpful comments and suggestions.
1.- The plebiscite was originally scheduled to be held on 26 April 2020 but was postponed until 25 October due to the pandemic.
2.- Analyzing voting (a phenomenon based on individual behavior) using an aggregate measure may be an example of the ecological fallacy (Robinson 2009). However, in our case, it is conceptually more useful to consider that we analyze turnout, which is a collective phenomenon. As Franklin (2004: 16) explains, turnout “is a feature of an electorate not a voter. And, while it is true that electorates are made up of aggregates of voters, the process of aggregation is not simply one of adding up relevant features of the individuals who form part of it.”
3. The data and replication code is available on the Harvard Dataverse (Disi Pavlic, Sáez-Vergara and Godoy 2022).
4.- Epidemiological data comes from the Ministry of Science, Technology, and Innovation (MinCiencia 2020).
5.- Public Health experts sometimes raised concerns about the lack of correlation between changes in phases and the districts’ epidemiological situation, while public officials have also stated that they have relaxed measures when lockdowns are too long (Andrade 2021).
6.- Data is not available for 2020, but protests decreased after the first months of 2020 (during the Southern Hemisphere summer) and the beginning of the pandemic.
7.- We use the logarithm because municipal populations vary significantly in size, ranging from 213 inhabitants in Río Verde to 629,743 in Puente Alto, which could result in heteroscedasticity issues.
8.- Huber/White/sandwich estimators are used to address potential modelling misspecifications.
9.- The results of these models are also available in Table A1 of the Appendix.
10.- This is discussed further using predictive values in Figures 7 and 8 of the Robustness and Postestimation section.
11.- See Table A4 of the Appendix for the full results.
12.- This is done using the “lvr2plot” command on Stata, which cannot be used after robust standard errors. The plots are available on Figures A1, A2, A3 and A4 in the Appendix.
13.- Vitacura, Timaukel, Ollagüe, Torres del Paine, Providencia, Porvenir, María Elena, Arica, Santiago, Alto Biobío, Pica, San Juan de la Costa, Chonchi, Juan Fernández, and Independencia.
14.- The models can be found on Tables A2 and A3 of the Appendix.
15.- We exclude, therefore, the cumulative number of COVID-19 infections for turnout, and vote for Piñera, exposure to violent protests, and its interaction term for the difference with 2017.
16.- Perhaps places that were already more open to the constitution-making process become more politically disaffected when exposed to the disruption of demonstrations (Disi Pavlic 2021).
17.- To save space, the Figures showing adjusted values of the pandemic-related variables use models with the nonviolent protest variable.
18.- Young Chileans had shown signs of increased political engagement and organization since 2017, with a logic that did not follow the traditional cleavages of Chilean politics (Bellolio 2019)
19.- We would like to thank two anonymous reviewers for these suggestions.
20.- The full results can be found in Table A5 of the Appendix.