Compulsory civic duty and turnout: evidence from a natural experiment

Abstract

We study the effect of compulsory civic duty on turnout. We take advantage of a natural experiment in Spain, where electoral officers are selected through a lottery and being an electoral officer is compulsory for those drafted. We analyze whether acting as an officer changes turnout and attitudes in subsequent elections. Our findings show that compulsory civic duty has a strong short-term and positive effect on electoral participation, but has no impact on political attitudes. The positive effect of the civic duty treatment, however, fades after a few months. Our results have implications for theories about civic duty and the formation of political habits.

1. Introduction

Civic duty can be defined as a set of norms about how a “good citizen” should behave. As such, civic duty has different dimensions (Dalton 2008). Participating in the political process is one of them. Voting is generally considered a social good in the belief that a healthy functioning of democracy relies on citizens’ participation. However, political involvement among voters has been decreasing over the last few decades, and, at the same time, disenchantment with the way democracy works across the globe is widespread.¹ Although civic duty is considered a crucial determinant of voting behavior, not much attention has been paid to institutional factors -as opposed to individual factors- that can affect such sense of civic duty.

In this paper, we ask whether compulsory civic duty interventions can increase political participation -such as voting- potentially creating a virtuous democratic cycle. There are several studies that argue that engaging in certain civic activities increases political participation (e.g.: Finkel 1985; Putnam 2000; Gerber et al. 2003). As a result, there is a general belief that engaging in civic activities is a social good that generates positive externalities in the form of increased pro-social behavior and, in particular, voting. These claims have been sustained by evidence that shows strong correlations between different forms of civic and political engagement (Verba et al. 1996). These results, however, should not be interpreted causally, as, without an exogenous shock, the reasons that may lead an individual to engage in one civic activity may also explain involvement in others.

One of the few papers that provides causal evidence consistent with the participation hypothesis applied to turnout is Gastil et al. (2008), which shows that under certain conditions, individuals elected for jury duty in the United States were more likely to vote in the future. Leal and Teigen (2018) and Teigen (2006) show that people with previous military service are also more likely to vote. These two events, jury duty and military service, are large shocks to civic duty that involve months and possibly years of intense civic duty engagement that are likely to have profound impacts on individuals.

We know little, however, about how less time-consuming acts of civic duty related to the democratic process affect turnout. One way in which governments could engage citizens is through requiring some sort of compulsory participation or involvement in the system. In this regard, there is a literature analyzing compulsory voting itself (see, among others, Feitosa et al. 2020). Our paper is aimed at extending our knowledge on how compulsory civic duty affects voluntary political civic behavior by exploiting the randomization in civic duty involvement provided by a particular feature of Spanish elections: in Spain, elections are organized with the compulsory participation of hundreds of thousands of electoral administrators. These officers supervise both the voting process and the vote count. It is an intense one-day-long civic duty service. As being an electoral officer is mandatory and selection is through a lottery, Spanish elections provides an ideal setting to study the participation hypothesis.²

Our analysis is conducted during the 2019 election year, a time period that has ideal characteristics for testing the participation hypothesis. During 2019 Spanish voters were called to vote three times within a seven month period. First, they could vote in the general election in April. Less than a month later, in May, they were called to vote in local, European and, in some regions, regional elections. Finally, there was no agreement among political parties to form a government at the national level after the April election, so in November 2019 a new general election took place.

We find that randomly chosen electoral officers have a 7.5% higher probability of voting in the next election when compared to control individuals that had similar voting probabilities and attitudes towards voting and democracy prior to the compulsory civic duty treatment. Consistent with the nature of the treatment -it only lasts one day- the effect is not permanent: by the November 2019 election the effect had disappeared.

The results have important theoretical and practical implications. If the decline of societal involvement in the functioning of democratic systems is a concern, our results provide evidence that civic democratic behavior can be increased by administering small booster doses of compulsory civic duty among voters. Additionally, while voting and other forms of civic duty may not have as transformative effects as previously thought, they are still important to activate beliefs and to keep people's behavior consistent with beliefs.

2. Electoral officers in Spain: a random treatment in civic duty

On election day, Spanish voters are allocated to vote in a specific ballot box and polling station. Each ballot box is supervised by three officers. One of the officers acts as “president”, and the other two as assistants (“vocales”). These three poll workers (as well as two potential substitutes for each of them) are chosen by a lottery that takes place around a month before the election. All citizens between 18 and 70 are part of the lottery. Serving as an electoral officer is compulsory. Only those that are over 65 can be exempted of the duty if they apply for it within a week of being drafted. Beyond age motives, one can only be exempt from acting as an officer if very limited personal, family or professional exemption causes apply (e.g. being under medical leave, pregnant, handicapped, or imprisoned) and evidence must be submitted within a week from notice.³ There are very strict requirements for those claiming to be exempt and they are enforced severely. When a drafted individual does not show up, they are reported to the police, investigated, and, eventually, subject to criminal punishment that includes fines or even potential imprisonment of up to a year.⁴ Exemptions are, therefore, rare.⁵ All these requirements imply that being chosen as electoral officer can be considered a random assignment to treatment with high levels of compliance.⁶

The duties of an electoral officer imply receiving a one-day civic duty treatment that consists of being part of the democratic process. Officers report at the polling station one hour before the polling station opens to fill in official paperwork, prepare the ballot papers and boxes, and review that everything is ready. Officers also verify each voter's ID at the ballot box and keep a record to make sure that each voter only votes once and that the election takes place according to the law. Officers can only vote at the end of the election day, right after the polling station is closed to voters and right before starting the vote count. The officers then start the vote count and enter the results into a computer program where results from all polling stations are aggregated. They finish their duties by certifying the results and depositing the ballots at the provincial electoral authority.

In summary, the electoral officer treatment involves supervising the whole electoral process, making sure that elections are organized properly, seeing people vote throughout the day, talking to voters and other administrators about election-related matters, monitoring the vote and, at the end of the day, calling the official results.

3. Sample and design

The ideal setting to test if civic duty has an effect on turnout would be to randomize people receiving an exogenous shock in civic duty and then compare their posterior behavior with those that do not receive the treatment. Our empirical strategy fits this ideal design. Participating in the organization of elections is a clear shock in civic duty, and due to the institutional characteristics of Spanish democratic system, it is exogenous and random.

We focus on people randomly drafted to act as electoral officers during the April 2019 General Election and study their subsequent electoral behavior in the local, regional, and European elections in May 2019 and in the November 2019 General Election. In the April 2019 general election there were 60038 ballot boxes. Each ballot box is supervised by three officers, so 180114 citizens were recruited through a lottery to be part of the electoral administration- equivalent to 0.7% of the voting population below 65 years old.⁷

Our strategy was to identify a random sample of potential officers through a survey that was administered two weeks before the April 2019 election, right after the lottery that drafted officers and replacements. We restricted the target population to people under 65 to avoid the potential selection bias caused by people over 65 being able to opt out from the treatment. We performed the survey through the firm Netquest. Out of the total number of people included in the Netquest panel, we randomly contacted a total of 31,504 citizens. We were able to identify a sample of people that declared to have been contacted to be officers or to be replacements. At the same time, with the respondents that had not been drafted, we surveyed a representative sample of 1513 Spaniards using gender, age and regional quotas.

In the initial survey in April 2019, we asked both our potential treatments and controls about their pre-treatment attitudes and behavior, such as vote recall in the previous election (June 2016), their attitudes towards democracy, political parties, civic duty, voting and the electoral process. We also gathered information on several socio-demographic characteristics such as age, gender, education, occupation or income. Note that the initial pre-treatment sampling frame is designed in a way that allows us to identify both individuals that will not be treated (control group), and individuals that might be potentially treated (those initially chosen by the Lottery of the Electoral Administration as either officers or replacements), out of which we determine in the next wave the actual treatment group that served as officers.

A month later, the day after the May 2019 election, we contacted both samples again. We were able to survey 279 treated individuals⁸ and 1319 control individuals. The attrition was low: 11.51% and as 12,7% in each sample⁹ In this May survey, we asked both treated and control respondents about their turnout in the April 2019 and the May 2019 elections and about their attitudes towards voting and democracy to compare against the pre-treatment answers.

Finally, we contacted all survey participants again after the November election. Our control sample in this round includes 1072 individuals (81.2% of the May sample), while our treatment sample consists of 230 individuals (82.3% of the May sample).¹⁰ Table A.1 in the Appendix presents the summary statistics. The balance test in Table A.3 shows that previous political behavior, ideology, pre-treatment attitudes and education levels are not significantly different across control and treated groups. This implies that our randomization worked well and we were not recruiting treated individuals that were more likely to vote already.

4. Results

We run four types of logit regression models, where the treated individuals are those that were electoral officers in the April 2019 general election. The first is a placebo model that measures the effect of being part of the electoral administration on turnout in the previous June 2016 general election.¹¹ There should be no treatment effect on voting in past elections as being an officer is a random event that should not condition past behaviors.

The second model measures the impact of being an electoral officer on the likelihood of voting in the April election (formulated as a turnout recall question in the May survey). We expect a large effect as drafted officers spend the day at the polling station. This higher chance of voting should be a result of two things. First, the civic duty treatment. Second, as they have to be present at the voting station the whole day, their cost of voting has exogenously been reduced to (almost) zero.

The participation hypothesis is tested in the third and fourth models. The third model captures the impact of the treatment on the likelihood of voting a month later in the European, local, and regional elections (in May 2019). We label those as the short-term effects of the treatment. The fourth model captures the impact of the treatment on the vote seven months later, on the November 2019 general election. We label those as the long-term effects on voting.

The results are displayed in Fig. 1. We show the average treatment effects with and without covariates.¹² Both types of models produce virtually identical results. The first marker shows no effect of performing as an electoral officer in April 2019 on having voted in the previous June 2016 election. That is, the placebo model -as expected- yields non-significant results. The placebo test deals also with potential concerns of over-reporting in the treatment sample. As our measure of turnout is self-reported, and it is not externally validated, the intervention may artificially boost the turnout responses (but not actual participation) among those that were chosen as electoral administrators. Seeing no differences in reported turnout for the 2016 elections is reassuring.

Figure 1. ATEs of Compulsory Civic Duty.

The second marker shows a large effect of the treatment on the April election. This is a relevant -though expected- result. For drafted officers it is compulsory to show up and spend the day in the polling station. As explained in the previous section, they can only vote at the end of the day, when the polling station closes and they still have to stay for the vote count. Voting therefore incurs no opportunity cost and, in addition, they might receive social pressure to vote. Our results show that officers increase their likelihood of voting by around 12%. In this model we still cannot disentangle which mechanism (civic duty, cost reduction or social pressure) explains the result.

The further two models test the participation hypothesis and capture the short-term and long-term effects of being an electoral officer. Out of three mechanisms that could explain higher voting in the previous model, two of them (social pressure and reduction of voting costs) are not present in the third and fourth models, which leaves the civic duty treatment as the only causal explanation for the result. We find first that April electoral officers were 8% more likely to vote in May. The effect is quite large when we consider that it is not far from the increase in the likelihood of voting in the April 2019 election, when they had the cost of voting suppressed and added social pressure. The effect is not explained by potential differential levels of turnout in General Elections vs. Local Elections, as turnout was virtually the same in both electoral contests (65.20% and 66.23%) and both treatment and controls are exposed to the same election effects. This result entails that the activation of civic duty through a compulsory service has positive effects on the political behavior a month later.

However, the treatment effect is virtually zero and non-significant in the November election. This suggests that the effects are not long-lasting. This is unsurprising given that the treatment we study only lasts one day. Finding that even a relatively low volt compulsory treatment does indeed last for weeks provides support to the participation hypothesis. The results are confirmed in Fig. 2 when we run a difference-in-difference estimation in which we control for individual level characteristics to exploit within-individual variation.¹³ - are also included in the online Appendix.

Figure 2. Diff-in-Diff Estimation.

These are relevant findings, as previous research only provided either correlational evidence, or evidence from rather life-changing interventions such as compulsory military service or compulsory voting.

Now we move to exploring whether the treatment entails any attitudinal change. Making individuals be part of the democratic process could affect their attitudes towards democracy, leading to an increase in participation (see Finkel 1985). Recent papers like Feitosa et al. (2020), using Chile as a case study, show that the sense of civic duty is affected by policies establishing or abolishing compulsory acts of civic duty such as compulsory voting. To test this mechanism we use the May 2019 survey to study responses to eleven questions about the degree of agreement with statements about voting, elections and democracy. The responses range from 0-I fully disagree- to 10-I fully agree.¹⁴. Our dependent variable is the change in attitudes between May and the same pre-treatment attitudes (surveyed before the election). We use the same covariates as in the previous models.

Our results are not consistent with this mechanism. We do not find any relevant or systematic change in attitudes or values. The systematic lack of statistical significance across all dependent variables in Fig. 3 suggests that citizens have not become much more familiarized with the political system, increased their political efficacy or even perceived that voting is more of a duty in May, after serving as electoral officer in the previous April election.¹⁵

Figure 3. May 2019 ATEs on attitudes and values.

Our findings are more likely to be explained through an activation mechanism suggested by the psychological literature. Within psychology there is research that explains the differences between values and beliefs on one side, and attitudes and behaviors consistent with those beliefs on the other side (Higgins 1996; Verplanken and Holland 2002). There is also work on the psychological mechanisms that trigger behavioral responses consistent with values and beliefs (Stapel and Tesser 2001; Verplanken and Holland 2002). For example, an individual may believe that exercising and eating healthy is important, but may only some times behave according to such belief. Whether the belief is “active” or not may depend on cues or events such as, for example, exposure to other people's behaviors consistent or inconsistent with the beliefs. It may also depend on previous behavior: engaging into a certain activity once may activate future behavioral responses on individuals. Within this literature, even small psychological interventions have been found to have sizable effects on activating behavior (Walton 2014). Importantly, these behaviors do not entail a change in the underlying values, only their activation.

Our results are consistent with this psychological activation mechanism: the civic duty treatment that we analyze leads officers to vote profusely on the day they serve as officers. We argue that it is this initial activation of voting caused by the treatment that also increases their probability to vote in the next election, leaving unchanged their underlying beliefs and attitudes. We confirm that this mechanism is at work in two further analyses. First, the activation works mainly for those that have pre-treatment attitudes that are already in favor of voting. In Fig. A.7 we show that it is those that already believe that voting is important who react to the treatment. In addition, a mediation analysis included in the online Appendix shows that most of the treatment's effect is an indirect effect through having voted in the election when they serve.

5. Concluding remarks

The literature has extensively assumed, based on correlations, that citizens that display more civic engagement tend also to participate more in politics. Our results show that there is indeed a causal relation between the two.

We have leveraged from a compulsory lottery system that selects electoral officers in Spanish elections. We have found positive and causal evidence of the participation hypothesis. Being drafted to serve as an officer in a polling station for one day, activates participation. Electoral officers are more likely to vote on the day they fulfill their compulsory civic duty and in upcoming elections in the short-term. The effects are not permanent, but they are important: even small interventions like the one analyzed here have effects on behavior that last for weeks. Importantly, serving as electoral officer does not entail any systematic attitudinal change neither in the short-run nor in the long-run.

These results have relevant implications. Our results support the participation hypothesis in a context of an intense one-day treatment. Citizens involved in civic duty activate their electoral participation in the short-run as a consequence of engaging in acts of civic duty, but for civic duty to exert effects over long periods of time it has to be activated frequently. It is left for further research to analyze to what extent recurrent small civic duty interventions may lead to a more stable long-term political engagement.

Finally, our research has limitations that are opportunities for future work. One limitation is that our study analyzes specific elections in which the context may play a role. For example, the November 2019 general elections took place only six months after the April 19th general election. Those who fulfilled their civic duty in April might be more frustrated than other citizens when the lack of agreement between political parties led to new elections in November. Future research should look at the role of the context on shaping the effect of civic duty interventions on political participation.